· RThe International Journal Business of and Finance ESEARCH CONTENTS Risk Analysis Using Regression Quantiles: Evidence from International Equity Markets 1 Hongtao Guo, Miranda

RThe International Journal of

Business and FinanceESEARCH

CONTENTS

Risk Analysis Using Regression Quantiles: Evidence from International Equity Markets 1Hongtao Guo, Miranda S. Lam, Guojun Wu & Zhijie Xiao

Pre-Close Transparency and Price Efficiency at Market Closing: Evidence from the Taiwan Stock Exchange 17Cheng-Yi Chien

Chief Executive Compensation: An Empirical Study of Fat Cat CEOs 27Dan Lin, Hsien-Chang Kuo & Lie-Huey Wang

Bank Credit Lines and Overinvestment: Evidence from China 43Qianwei Ying, Danglun Luo & Lifan Wu

Long-Term Prior Return Patterns in Stock Returns: Evidence from Emerging Markets 53Sanjay Sehgal, Sakshi Jain & Pr Laurence the Porteu de la Morandiere

The Fama French Model or the Capital Asset Pricing Model: International Evidence 79Paulo Alves

Quality of Governance and the Market Value of Cash: Evidence from Spain 91Eloisa Perez-de Toledo & Evandro Bocatto

Is Inflation Always and Everywhere a Monetary Phenomenon? The Case of Nigeria 105Salami Doyin & Kelikume Ikechukwu

The Effects of Ownership Structure and Competition on Risk-Taking Behavior: Evidence 115from UAE Conventional and Islamic BanksHussein A. Hassan Al-Tamimi & Neila Jellali

VOLUME 7 2013NUMBER 2

The International Journal of Business and Finance Research ♦ VOLUME 7 ♦ NUMBER 2 ♦ 2013

1

RISK ANALYSIS USING REGRESSION QUANTILES: EVIDENCE FROM INTERNATIONAL EQUITY

MARKETS Hongtao Guo, Salem State University

Miranda S. Lam, Salem State University Guojun Wu, University of Houston

Zhijie Xiao, Boston College

ABSTRACT

In this paper we study risk management based on the quantile regression. Unlike the traditional VaR estimation methods, the quantile regression approach allows for a general treatment on the error distribution and is robust to distributions with heavy tails. We estimate the VaRs of five international equity indexes based on AR-ARCH model via quantile regressions. The empirical application show that the quantile regression based method is well suited to handle negative skewness and heavy tails in stock return time series. JEL: G110; G150; C18 KEYWORDS: Value at risk, international equities, quantile regression, risk analysis INTRODUCTION

he Value-at-Risk (VaR) is the loss in market value over a given time horizon that is exceeded with probability p, where p is often set at 0.01 or 0.05. In recent years, VaR has become a popular tool in the measurement and management of financial risk. This popularity is spurred both by the need

of various institutions for managing risk and by government regulations [see Dowd (1998), Saunders (1999), Blankley, Lamb and Schroeder (2000) for more detailed description of the SEC disclosure requirements]. Traditional methods of VaR estimation are either based on distributional assumptions such as normality, or nonparametric smoothing that suffers from curse of dimensionality. In this paper, we estimate VaR via quantile regression ARCH models. This model has the advantage of computational convenience, as well as the robustness properties of the quantile regression method. The estimation procedure can be easily implemented on a regular personal computer. The estimation programs are available in standard statistical packages such as S-Plus, and can also be easily written in other programming languages. In addition, since GARCH models can be asymptotically represented by ARCH processes, an ARCH model with an appropriate chosen number of lags can practically provide a good approximation .We estimate VaR in international equity markets using weekly return series for four major world equity market indexes: the U.S. S&P 500 Composite Index, the Japanese Nikkei 225 Index, the U.K. FTSE 100 Index, and the Hong Kong Hang Seng Index. We consider a combination of AR (in mean) and ARCH (in volatility) for the return series. The empirical results indicate that the quantile regression based method provides good coverage rates, and is better than the traditional estimation based on normality.The remainder of the article is organized as follows. Section 2 reviews relevant literature. Section 3 describes the quantile regression approach to VaR estimation, and provides data descriptions. Empirical results regarding estimated VaR and model performance are reported and discussed in section 4. Finally, Section 5 contains the concluding remarks.

T

H. Guo et al | IJBFR ♦ Vol. 7 ♦ No. 2 ♦ 2013

2

LITERATURE REVIEW Although VaR is a relatively simple concept, its measurement is in fact a challenging task. Currently there are two broad classes of methods in estimating VaR [see Beder (1995) and Duffie and Pan (1997) for surveys on this topic]. The first approach is based on the assumption that financial returns have normal (or conditional normal) distributions. Under this assumption, the estimation of VaR is equivalent to estimating conditional volatility of returns. Since there is a large and growing literature on volatility modeling itself, this class is indeed a large and expanding world (see, e.g., Jorian (1997)). However, there has been accumulated evidence that portfolio returns are usually not normally distributed. In particular, it is frequently found that market returns display negative skewness and excess kurtosis in the distribution of the time series. These findings suggest that VaR estimation by a more robust method is needed. The second class of VaR estimators is based on computing the empirical quantile nonparametrically (see, e.g. Jeong 2009); for example, using rolling historical quantiles. Although local, nearest neighbor and kernel methods are somewhat limited in their ability to cope with more than one or two covariates. Other approaches in estimating VaR include the hybrid method by Boudoukh, Richardson and Vhitelaw (1998), and the method based on the extreme value theory [see, for example, Boos (1984), McNeil (1998), and Neftci (2000)].We believe that the quantile regression method is well suited for estimating VaRs. Quantile regression was introduced by Koenker and Bassett (1978) and has now become a popular robust approach for statistical analysis. The quantile regression method is an extension of the empirical quantile methods. While classical linear regression methods based on minimizing sums of squared residuals enable one to estimate models for conditional mean functions, quantile regression methods offer a mechanism for estimating models for the conditional quantile functions, thus quantile regression is capable of providing a complete statistical analysis of the stochastic relationships among random variables (see, e.g. Powell (1986), Gutenbrunner and Jureckova (1992), Buchinsky (1994), and Koenker and Portnoy (1996) among others for subsequent development in quantile regression theory. In recent years, quantile regression estimation for time-series models has gradually attracted more attention. In particular, Koul and Saleh (1992) studied quantile regression methods for the traditional autoregressive processes and Koul and Mukherjee (1994) studied quantile regression in long-memory models. Portnoy (1991) studied asymptotic properties for regression quantiles with m-dependent errors, his analysis also allows for nonstationarity with a nonvanishing bias term. Koenker and Zhao (1996) extended quantile regression to ARCH models. Engle and Manganelli (1999) propose a CaVaR model based on the regression quantiles. Recently, Koenker and Xiao (2006) studied the quantile autoregression (QAR) models that can capture systematic influences of conditioning variables on the location, scale and shape of the conditional distribution of the response. Bouyé and Salmon (2008); Chen, Koenker and Xiao (2009) employ parametric copula models to generate nonlinear-in-parameters quantile autoregression models.The ARCH/GARCH models have been proved to be extremely successful in modeling financial returns. For this reason, much of the literature in VaR estimation considers ARCH type models. However, estimation of these models in the literature is usually based on the assumption that financial returns have normal (or conditional normal) distributions. There is accumulating evidence that financial time series display negative skewness and excess kurtosis. Extreme realizations of returns can adversely affect the performance of estimation and inference designed for Gaussian conditions; this is particularly true of ARCH and GARCH models whose estimation of variances are very sensitive to large innovations. For this reason, we propose using quantile regression methods to estimate VaR in ARCH models.


3

DATA AND METHODOLOGY Data The data used in the following empirical analysis are the weekly return series, from September 1976 to August 1999, of five major world equity market indexes: the U.S. S&P 500 Composite Index, the Japanese Nikkei 225 Index, the U.K. FTSE 100 Index, the Hong Kong Hang Seng Index, and the Singapore Strait Times Index. The FTSE 100 Index data are from January 1983 to September 1999. Table 1 reports some summary statistics of the data. Table 1: Summary Statistics of the Data

S&P 500 Nikkei 225 FTSE 100 Hang Seng SingaporeST Mean 0.0016 0.0011 0.0023 0.0029 0.0015 Std. Dev 0.0211 0.0242 0.0218 0.0393 0.0326

Max 0.0818 0.1214 0.0982 0.1542 0.1096

Min -0.1666 -0.1089 -0.1782 -0.3497 -0.4747

Skewness -0.5343 -0.2873 -1.0405 -1.1830 -2.6771

Excess Kurtosis 3.3987 3.1968 8.4345 7.0583 37.9887

AC(1) 0.0045 -0.0366 0.0534 0.1162 0.0685 AC(2) 0.0005 0.0971 0.0523 0.0922 0.0072 AC(3) 0.0084 0.0247 -0.0181 -0.0016 0.0324 AC(4) -0.0082 -0.0417 -0.0190 -0.0677 0.0101 AC(5) -0.0215 -0.0057 -0.0084 -0.0470 0.0448 AC (10) -0.0256 -0.0167 0.0125 -0.0241 -0.0213

This table shows the summary statistics for the weekly returns of five major equity indexes of the world. AC(k) denotes autocorrelation of order k. The sample period is from September 1976 to August 1999, except for FTSE 100 which starts in January 1983. The source of the data is the online data service Datastream. The mean weekly returns of the five indexes are all over 0.1% per week, with the Hang Seng Index producing an average return of 0.29% per week, an astonishing 32-fold increase in the index level over the 24-year sample period. In comparison, the Nikkei 225 index only increased by 3-fold. The Hang Seng's phenomenal rise does not come without risk. The weekly sample standard deviation of the index is 3.93%, the highest of the five indexes. In addition, over the sample period the Hang Seng suffered four larger than 15% drop in weekly index level, with maximum loss reaching 35%, and there were 23 weekly returns below -10%! As has been documented extensively in the literature, all five indexes display negative skewness and excess kurtosis. The excess kurtosis of Singapore Strait Times Index reached 37.99, to a large extent driven by the huge one week loss of 47.47% during the 1987 market crash. The autocorrelation coefficients for all five indexes are fairly small. The Hang Seng Index seems to display the strongest autocorrelation with the AR(1) coefficient equal to 0.116 and AR (2) coefficient equal to 0.092. Estimating VaR by Regression Quantiles For ease of exposition, we define Value-at-Risk as the percentage loss in market value over a given time horizon that is exceeded with probability p. That is, for a time series of returns on an asset {𝑟}𝑡=1𝑛 , find VaRt such that 𝑃𝑟(𝑟𝑡 < −𝑉𝑎𝑅𝑡|𝐼𝑡−1 ) = 𝑝, (1) Where 𝐼𝑡−1 denotes the information set at time t - 1. From this definition, it is clear that finding a VaR essentially is the same as finding a 100p% conditional quantile. Koenker and Bassett (1978) show how a


4

simple minimization problem yielding the ordinary sample quantiles in the location model can be generalized naturally to the linear model, generating a new class of statistics called regression quantiles. To motivate regression quantile, let's first consider estimating a simple pth sample quail- tile. It is clear that the estimator is the solution to the following minimization problem

𝑚𝑖𝑛𝑏 ∊ ℜ

[ ∑𝑡 ∊ {𝑡: 𝑟𝑡 ≥ 𝑏} 𝑝|𝑟𝑡− 𝑏| + ∑

𝑡 ∊ {𝑡: 𝑟𝑡 < 𝑏} (1 – 𝑝) |𝑟𝑡 − 𝑏| ]. (2)

When the quantile is the median, p= 0.5, we have an important special case: the estimator that minimizes the sum of absolute residuals - the median estimator. Such a device can be generalized to regressions. If we define a k by 1 vector of regressors,𝑥𝑡, and consider the regression model 𝑟𝑡 = 𝑏′𝑥𝑡 + 𝑢𝑡 (3) with i.i.d. residual {𝑢𝑡}, then, conditional on the regressor xt, the p-th quantile of rt is given by 𝐹𝑟𝑡−1(𝑝|𝑥𝑡 ) = 𝑖𝑛𝑓 {𝑦|𝐹𝑟𝑡(𝑦|𝑥𝑡) ≥ 𝑝 } = 𝑏′𝑥𝑡 + 𝐹𝑢−1(𝑝).

where 𝐹𝑢 (.) is the cumulative distributional function of the residual. Conventionally the first component of the regressors 𝑥𝑡 is an intercept term and we have 𝐹𝑟𝑡−1(𝑝|𝑥𝑡) = (𝑝) + = 𝑏 ,

Where 𝑏(𝑝) = (𝑏1 + 𝐹𝑢−1(𝑝), 𝑏2, … , 𝑏𝑘). The regression quantile process corresponding to model (3) is determined by the following optimization problem

𝑏� = 𝑎𝑟𝑔 𝑚𝑖𝑛𝑏 ∊ ℜ𝑘

� ∑𝑡 ∊ {𝑡: 𝑟𝑡 ≥ 𝑥𝑡𝑏} 𝑝|𝑟𝑡− 𝑥𝑡

′ 𝑏| + ∑𝑡 ∊ {𝑡: 𝑟𝑡 < 𝑥𝑡𝑏} �1 – 𝑝�|𝑟𝑡 − 𝑥𝑡′ 𝑏| �. (4)

The estimator 𝑏�(𝑝) generalizes the concept of pth sample quantile to the pth regression quantile. In this ease, the least absolute error estimator is the regression median. i.e., the regression quantile for p = 0.5. Koenker arid Bassett (1978) show that 𝑏�(𝑝) is a root-n consistent estimator of 𝑏�(𝑝). and √𝑛 �𝑏�(𝑝) − 𝑏(𝑝)�converges weakly to a normal distribution. The quantile regression theory can be extended to time series models with conditional heteroskedasticity. Consider a return process {rt} generated by the following regression model with conditional heteroscedasticity 𝑟𝑡 = 𝛼 ′𝑥𝑡 + 𝑢𝑡 (5)


5

where the error term satisfies 𝑢𝑡 = ( 𝛾0 + 𝛾1|𝑢𝑡−1| + … + 𝛾𝑞 �𝑢𝑡−𝑞�ℇ𝑡, (6) with 𝛾0 > 0, (𝛾1, . . . , 𝛾𝑞)

′ ∊ ℜ+𝑞 , then this is a time series with ARCH effect. Here we assume that the

innovations {ℇ𝑡 } have a general distribution F(.), including the normal distribution and other commonly used distributions in financial applications with heavy tails. In model (5), 𝑥𝑡 is the vector of regressors which may include lag values of the dependent variable. When 𝑥𝑡 = (1, 𝑟𝑡−1, . . . , 𝑟𝑡−𝑝)′, , model (5) reduces to the case of Koenker and Zhao (1996). By definition, VaRt at p-percent level is just the conditional quantile of 𝑟𝑡 in the model of (5) and (6) given information to time t - 1, i.e.𝐼𝑡−1. Thus, the conditional value at risk (VaRt) at p-percent level is −𝑉𝑎𝑅𝑡(𝑝) = 𝛼 ′𝑥𝑡 + � 𝛾0 + 𝛾1|𝑢𝑡−1|+ … + 𝛾𝑞 �𝑢𝑡−𝑞��𝐹−1(𝑝) = ⍺α′𝑥𝑡 + 𝛾(𝑝)′𝑍𝑡 Where 𝑍𝑡 = (1, |𝑢𝑡−1| , . . . , �𝑢𝑡−𝑞� )′ and 𝛾(𝑝)′ = (𝛾0,𝛾1, . . . , 𝛾𝑞)𝐹−1( 𝑝). Quantile regression provides a direct approach of estimating the γ(p) and other parameters, thus delivering an estimator of VaRt(p). In particular, the ARCH parameters 7, γ(p) can be estimated by the following problem

𝛾� (𝑝) = arg 𝑚𝑖𝑛𝛾 ∊ ℜ𝑘

[∑

𝑡 ∊ {𝑡:𝑢𝑡 ≥ 𝑧𝑡′𝛾} 𝑝�𝑢𝑡− 𝑧𝑡 ′ 𝛾� +

∑𝑡 ∊ {𝑡:𝑢𝑡 < 𝑧𝑡′𝛾}(1 – p)�𝑢𝑡 − 𝑧𝑡′ 𝛾�] (7)

Koenker and Zhao (1996) show that 𝛾�(p) is a root-n consistent estimator of y (p). In practice, we can replace 𝑢𝑡 and 𝑍𝑡 by their (say, OLS) estimators and, under mild regularity conditions, the resulting 𝑦�(p) is still a root-n consistent estimator of 𝛾 (p). Quantile regression method has the important property that it is robust to distributional assumptions. This property is inherited from the robustness property of the ordinary sample quantiles. Quantile estimation is only influenced by the local behavior of the conditional distribution of the response near the specified quantile.Computation of the regression quantiles by standard linear programming techniques is very efficient. It is also straightforward to impose the nonnegativity constraints on all elements of Υ. Barrodale and Roberts (1974) proposed the first efficient algorithm for 𝐿1- estimation problems based on modified simplex method. For very large quantile regression problems there are some important new ideas that speed up the performance of computation relative to the simplex approach underlying the original code. Portnoy and Koenker (1997) describe an approach that combines some statistical preprocessing with interior point methods and achieves faster speed over the simplex method for very large problems.


6

ARCH VaR Model Selection Given the model (5) and (6), if the lags are correctly selected we should have Pr { 𝑟𝑡 < 𝑉𝑎𝑅𝑡(𝑝)} = 𝑝 at the true parameter. As a result, {𝑒𝑡: 𝑒𝑡 = 𝐼[𝑟𝑡 < −𝑉𝑎𝑅𝑡(𝑝)] − 𝑝} should be i.i.d. In contrast, when the lags are incorrectly chosen, {𝑒𝑡} will be serially dependent. Therefore, to test the adequacy of lag choice, it suffices to check whether {𝑒𝑡} is i.i.d. There have been several statistical procedures for testing the i.i.d. assumption. In the case of Gaussian time series, the standardized spectral density captures all serial dependencies. Consequently, any deviation of the spectral density from uniformity is an evidence of serial dependence and thus we can test serial dependence in {𝑒𝑡} using the standardized spectral density approach (Hong 1996). More generally, for non-Gaussian time series, the higher order spectral method or the generalized spectral method may be used in testing serial dependence (Hong 1999). [see also Cowles and Jones (1937), L.jung and Box (1978), etc., for related topics.] Another popular method used in selecting lag length is to conduct sequential tests for the significance of the coefficients on lags. Such an approach provides a model selection strategy which chooses between a model with, say, k lags and a model with q=k + l lags. Koenker and Zhao (1996) show that a 𝑥2 test can be constructed for hypothesis of the type 𝐻0: 𝑅𝛾= 0. Under Ho, the following Wald statistic converges to a centered chi-square distribution with s degrees of freedom (where s is the number of restrictions) 𝑇𝑛 = 𝑛𝜔�−2(𝑅𝛾�(𝑝))′[𝑅𝐷�1−1𝐷�0𝐷�1−1𝑅′]−1 𝑅𝛾�(𝑝), (8) where 𝜔2 = 𝑝(1 − 𝑝)/𝑓(𝐹−1(𝑝))2, D0 = EZZ' and D1 = EZZ'/σ. This procedure can be applied to testing the significance of lag coefficients. If we are choosing between k lags and q = k +1 lags, let R be a diagonal matrix with the k+1 to q-th diagonal elements equal to ones and others equal to zeros, 𝑅 = 𝑑𝑖𝑎𝑔[0, … , 0, 1, … , 1], (9) then the corresponding statistic 𝑇𝑛 in (8) is used in testing the significance of coefficients 𝛾𝑘+1 ,. . . ,𝛾𝑞. In practice, we select a priori a big enough number𝑞𝑚𝑎𝑥 , then we choose the lag length from possible values {1, . . . , 𝑞𝑚𝑎𝑥}. The procedure starts with the most general model which has 𝑞𝑚𝑎𝑥 lags and tests whether the last lag coefficient is significant. If it is, then 𝑞𝑚𝑎𝑥 is chosen. Otherwise, we estimate the model with 𝑞𝑚𝑎𝑥- 1 lags. This is a sequential procedure which is repeated until a rejection occurs. RESULTS For each time series of the five international equity index, we first conduct model specification analysis and choose the appropriate lags for the mean equation and the ARCH component. Based on the selected model, we use Equation (1) to obtain a time series of residuals. The residuals are then used in the ARCH VaR estimation described in (7). Model Specification Analysis We conduct sequential tests for the significance of the coefficients on lags. The inference procedures we use here are asymptotic inferences. For estimation of the covariance matrix, we use the robust HAC (Heteroskedastic and Autocorrelation Consistent) covariance matrix estimator of Andrews (1991) with the data-dependent automatic bandwidth parameter estimator recommended in that paper. First of all, we choose the lag length in the autoregression


7

𝑟𝑡 = 𝛼0 + 𝛼1𝑟𝑡−1 + … + 𝛼𝑠𝑟𝑡−𝑠 + 𝑢𝑡, (10) using a sequential test of significance on lag coefficients. The maximum lag length that we start with is s = 9, and the procedure is repeated until a rejection occurs. Table 2 reports the sequential testing results for the S&P 500 index. Table 2: VaR Model Mean Specification Test for the S&P 500 Index

Round 1st 2nd 3rd 4th 5th 6th 7th 8th 9th α0 3.6193*** 3.5529*** 3.7008*** 3.5246** 3.6824** 3.7304*** 3.6843*** 3.6453*** 3.8125*** α1 -1.8916 -1.9735* -1.9903* -2.0020** -1.9996* -1.9735* -2.0147** -2.004** -2.104** α2 0.2432 0.2143 0.1474 0.0786 0.0872 0.0833 0.0935 0.0942 α3 -0.8676 -0.8220 -0.7795 -0.7899 -0.8123 -0.8162 -0.8162 α4 -0.1470 -0.1780 -0.1717 -0.1412 -0.1610 -0.1612 α5 -0.5730 -0.5771 -0.5940 -0.5670 -0.5677

α6 -0.6055 -0.5934 -0.6112 -0.6102

α7 0.1783 0.1895 0.1895 α8 1.3186 1.3191

α9 0.2034

This table reports the test results for the VaR model mean equation specification for the S& P500 Index. The number of lags in the AR component of the ARCH model is selected according to the sequential test. The table reports the t-statistic for the coefficient with the maximum lag in the mean equation. .*,**,***, indicate significance at the 10,5 and 1 percent levels respectively. The t-statistics of all coefficients are listed for nine rounds of the test. The significance level of the t-ratios are indicated in Table 2. *, **, *** indicates significance at 10, 5 and 1 percent level respectively. The t-statistic of the coefficient with the maximum number of lags does not become significant until S = 1, the 9th round. The preferred model is an AR(1) model. We then report the selected mean equations for all five indexes in Table 3. Table 3: ARCH VaR Models Selected by the Sequential Test

Index Mean Lag 5% ARCH Lag 1% ARCH Lag S&P 500 1 7 10 Nikkei 225 2 8 8 FTSE 100 1 6 6 Hang Seng 4 7 9 Singapore ST 1 7 9

This table summarizes the preferred ARCH VaR models for the five global market indexes. The number of lags in the mean equation and the volatility component of the ARCH model are selected according to the test. Our next task is to select the lag length in the ARCH effect 𝑢𝑡 = �𝛶0 + 𝛶1|𝑢𝑡−1| + ⋯+ 𝛶𝑞�𝑢𝑡−𝑞��. (11) Again, a sequential test is conducted using the results of (8). To calculate the t-statistic, we need to estimate 𝜔2 = 𝑝 (1 − 𝑝)/𝑓(𝐹−1(𝑝))2. There are many studies on estimating 𝑓(𝐹−1(𝑝)), including Siddiqui (1960) and Sheather and Maritz (1983). Notice that 𝑑𝐹

−1(𝑡)𝑑𝑡

= 1𝑓(𝐹−1(𝑡))

, (12)


8

following Siddiqui (1960), we may estimate (12) by a simple difference quotient of the empirical quantile function. As a result, 𝑓 (𝐹−1� (𝑡)) = 2ℎ𝑛

𝐹�−1 (𝑡+ ℎ𝑛)− 𝐹�−1 (𝑡− ℎ𝑛) , (13) where 𝐹�−1(t) is an estimate of F-1(t) and ℎ𝑛 is a bandwidth which goes to zero as n→∞. A bandwidth choice has been suggested by Hall and Sheather (1988) based on Edgeworth expansion for studentized quantiles. This bandwidth is of order 𝑛−1 3⁄ and has the following representation ℎ𝐻𝑆 = 𝑧𝛼

2 3⁄ [1.5𝑠(𝑡)/𝑠"(𝑡)]1 3⁄ 𝑛−1 3⁄ , (14) where 𝑧𝛼⍺ 𝑠atisfies Φ�𝑧𝛼⍺� = 1 − 𝛼/2 for the construction of 1 - α confidence intervals. In the absence of additional information, s(t) is just the normal density. Starting with 𝑞𝑚𝑎𝑥 =10, a sequential test was conducted and results for the 5% VaR model of the S&P 500 index are reported in Table 4. We see that in the fourth round, the t-statistic on lag 7 becomes significant. The sequential test stops here, and it suggests that ARCH(7) is appropriate. Table 4: 5% VaR Model ARCH Specification Test for the S&P 500 Index

Round 1st 2nd 3rd 4th 𝑦0 -20.621* -25.110* -27.081* -19.789* 𝑦1 2.8911* 3.3601* 3.2420* 3.1658* 𝑦2 1.9007 2.9561* 2.8366* 2.5561* 𝑦3 0.9982 1.0886 0.9567 1.2560 𝑦4 0.7737 1.0099 2.2672 1.5672 𝑦5 0.6919 0.8564 1.1111 0.8689 𝑦6 0.2336 0.3366 0.5244 0.2688 𝑦7 2.3406* 2.5219* 0.2318 2.8891* 𝑦8 0.4866 0.4688 1.3248 𝑦9 1.1644 0.9921

𝑦10 1.4665 This table reports the test results for the 5 % VaR model specification for the S& P500 Index. The number of lags in the volatility component of the ARCH model is selected according to the test. The table reports the t-statistic for the coefficient with the maximum lag in the ARCH equation. .*,**,***, indicate significance at the 10, 5 and 1 percent levels respectively. Based on the model selection tests, we decide to use the AR(1)-ARCH(7) regression quantile model to estimate 5% VaR for the S&P 500 index. We also conduct similar tests on the 5% VaR models for other four indexes and on the 1% VaR models for all five indexes. To conserve space we do not report the entire testing process in the paper. The results are available from the author. The mean equations generally have one or two lags, except the Hang Seng Index, which has a lag of 4 and displays more persistent autoreggressive effect. For the ARCH equations, at least 6 lags are needed for the indexes. The longest lag, at 10, is for the 10% ARCH VaR model for the S&P 500 index. The 1% VaR models require at least as many lags in the ARCH equation as the 5% VaR model. For the Nikkei 225 and FTSE 100 indexes, the lengths of the ARCH lags are the same for the 1% and 5% VaR models. Since the estimation program for the regression quantile VaR model is very efficient, lags up to 10 in the ARCH equation are very easy to handle. Estimated VaRs The estimated parameters for the mean equations for all five indexes are reported in Table 5. The constant term for the five indexes is between 0.1% for the Nikkei and 0.26% for the Hang Seng. As suggested by


9

Table 1, the Hang Seng seems to display the strongest autocorrelation and this is reflected in the 4 lags chosen by the sequential test. Table 5: Estimated Mean Equation Parameters

S&P 500 Nikkei 225 FTSE 100 Hang Seng Singapore ST 𝛼0 0.0021*** 0.0010*** 0.0021*** 0.0026*** 0.0015***

(0.0006) (0.0007) (0.0007) (0.0011) (0.0009)

𝛼1 -0.0558** -0.0327** 0.0553** 0.1090** 0.0655**

(0.0265) (0.0288) (0.0270) (0.0289) (0.0289)

𝛼2 0.0953** 0.0880**

(0.0288) (0.0291) 𝛼3 -0.0136**

(0.0291) 𝛼4 -0.0740**

(0.0289)

This table reports the estimated parameters of the mean equation for the five global equity indexes. The standard errors are in parentheses under the estimated parameters.*,**,***, indicate significance at the 10, 5 and 1 percent levels respectively. Tables 6 and 7 report the estimated ARCH parameters for the 5% VaR and 1% VaR models, respectively. The coefficients on the lagged absolute residuals are mostly positive. The negative coefficients are all statistically insignificant, with the exception of one. The selected ARCH models are relatively long, ranging from 6 lags to 10 lags. This is largely due to the fact that, when the conditional variances have relatively complicated structures, we usually need ARCH models with many lags to deliver good approximations of such general volatility models. Table 6: Estimated ARCH Equation Parameters for the 5% VaR Model

Parameter S&P 500 Nikkei 225 FTSE 100 Hang Seng Singapore ST

𝑦0 -0.0342*** -0.0395*** -0.0335*** -0.0637*** -0.0456*** (0.0017) (0.0022) (0.0014) (0.0032) (0.0025)

𝑦1 0.2129* 0.0645* 0.0506* 0.1691* 0.1089** (0.0672) (0.0554) (0.0697) (0.0795) (0.0489)

𝑦2

0.1103**

0.2005**

0.0595*

0.1092**

0.1559*

(0.0432) (0.0429) (0.0686) (0.0339) (0.0653) 𝑦3 -0.0196** 0.1043* 0.0298** 0.2282** 0.0223**

(0.0156) (0.0633) (0.0261) (0.0376) (0.0428)

𝑦4 0.1319*

0.0453**

0.0601*

0.0733**

0.1061*

(0.0842) (0.0390) (0.0883) (0.0296) (0.0813) 𝑦5 0.0167** 0.0996** -0.0174** 0.0235** 0.1479**

(0.0192) (0.0446) (0.0143) (0.0371) (0.0491)

𝑦6 0.0253* 0.0173** 0.0948** 0.0193* 0.0299**

(0.0941) (0.0326) (0.0478) (0.0530) (0.0206)

𝑦7 0.0002*** 0.2553** 0.0948** 0.0917** 0.1036** (6.92E-5) (0.0360) (0.0478) (0.0423) (0.0437)

𝑦8 0.1374** (0.0447)

This table reports the estimated parameters of the ARCH equation for the 5% VaR model for the five global indexes. The standard errors are in parentheses under the estimated parameters. .*,**,***, indicate significance at the 10, 5 and 1 percent levels respectively.


10

Based on our model estimation of the U.S. S&P 500 index, for the 5% VaR, the estimated VaRs generally range between 2.5% and 5%, but during very volatile periods they could jump over 10%, as happened in October 1987. The 1% VaRs lie above the 5% VaRs. The two series overlap each other most of the time, but they are very much separate from each other from 1992 to 1994 when overall market volatility is relatively low. During high volatility periods, there is high variation in estimated VaRs and 5% and 1% VaRs overlap each other more often. Certainly on a particular date, the 1% VaR lies above the 5% VaR. Table 7: Estimated ARCH Equation Parameters for the 1% VaR Model

Parameter S&P 500 Nikkei 225 FTSE 100 Hang Seng Singapore ST 𝑦0 -0.0523*** -0.0650*** -0.0551*** -0.1126*** -0.0797***

(0.0039) (0.0037) (0.0048) (0.0080) (0.0044)

𝑦1 0.2678** -0.0536 0.2842 0.3033 0.0893* (0.0679) (0.2367) (0.1013) (0.1449) (0.0564)

𝑦2 0.1783

0.2336**

0.0435

0.3158

0.3012*

(0.1291) (0.0259) (0.1492) (0.1858) (0.0980) 𝑦3

-0.0878 0.0929* 0.1582 0.3675 0.1141*

(0.2933) (0.0946) (0.1397) (0.1346) (0.0584)

𝑦4 0.1644* 0.0507 -0.0828 0.0463* 0.0562 (0.0540) (0.1561) (0.1424) (0.0642) (0.1029)

0.0971 0.0335* 0.1777 0.1013* -0.0710 (0.3457) (0.0527) (0.1312) (0.0658) (0.1522)

𝑦6 0.1557

0.0903**

0.1405*

0.0140

0.0183**

(0.1057) (0.0418) (0.0502) (0.2306) (0.0305)

𝑦7 0.1992

0.4277*

0.1121* 0.2293

(0.1533) (0.0733) (0.0659) (0.1208)

𝑦8 -0.0938*

0.1707*

0.0891 -0.0222

(0.0722) (0.0657) (0.3770) (0.1314)

𝑦9 -0.0394

0.4309 0.1320*

(0.1267) (0.1408) (0.0616)

𝑦10 -0.1892*

(0.0696)

This table reports the estimated parameters of the ARCH equation for the 1% VaR model for the five global indexes. The standard errors are in parentheses under the estimated parameters. .*,**,***, indicate significance at the 10, 5 and 1 percent levels respectively. The Japan-NIKKEI 225 results show that the estimated weekly 5% and 1% VaRs for the Nikkei 225 Index are quite stable and remain at the 4% and the 7% level from 1976 till 1982. Then the Nikkei 225 Index took off and appreciated about 450% over the next eight years, reaching its highest level at the end of 1989. This quick rise in stock value is accompanied by high risk, manifested here by the more volatile VaR series. In particular, the VaRs fluctuated dramatically, ranging from a low of 3% to a high of 15%. This volatility in VaR may reflect optimistic market outlook at times as well as worry about high valuation and the possibility of a market crash. That crash did come in 1990, and ten years later, the Nikkei 225 Index still hovers around at a level which is about half off the record high in 1989. The 1990s is far from a rewarding decade for investors in the Japanese equity market. The mean annual return from the Nikkei 225 Index is negative and risk is at a high level. Average weekly 5% VaR is about 5%, and about 7% for the 1% VaR. The variation in both series is also very high, bouncing between 13.5% and -1%. The estimated 5% and 1% VaRs for the U.K. Financial Times 100 Index appreciated 7-fold over the 16-year sample period. The 5% VaR is very stable and averages about 3%. They stay very much within


11

the 2% - 4% band, except on a few occasions, such as the 1987 global market crash. The 1% VaR is also more stable than that of the Nikkei 225 Index, mostly ranging between 4% and 8%. Compared with the SP 500 Index and the Nikkei 225 Index, the overlap between the 5% VaR and the 1% VaR is minimal. The Hong Kong Hang Seng Index produces an average return of 0.29% per week, an astonishing 32-fold increase in the index level over the 24-year sample period. The Hang Seng's phenomenal rise does not come without risk. We mentioned above that the weekly sample standard deviation of the index is 3.93%, the highest of the five indexes. In addition, the Hong Kong stock market has had more than its fair share of the market crashes. If we define a market crash as having the main index drop at least 15% in a week, then Hong Kong experienced four market crashes in 24-year sample period. The average 5% VaR over the sample is about 7%, and the average 1% VaR is about 12%, both the highest among the five indexes. The variation in the estimated VaR is huge, in particular the 1% VaRs. It ranges from 0% to 34%, the largest of the five indexes.Interestingly, for Singapore Strait Times index, the estimated VaRs display a pattern very similar to that of the U.K. FTSE 100 Index, although the former is generally larger than the latter. The higher risk in the Singapore market did not result in higher return over the sample period. Among the five indexes, the Singapore market suffered the largest loss during the 1987 crash, a 47.5% drop in a week. The market has since recovered much of the loss. Among the five indexes, the Singapore market only outperformed the Nikkei 225 Index over the entire 24-year sample period. Performance of the ARCH Quantile Regression Model In this section we conduct empirical analysis to help us understand the difference in dynamics between VaRs estimated by regression quantiles and those by volatility models with the conditional normality assumption. There are extensive empirical evidences supporting the use of the GARCH models in conditional volatility estimation. Bollerslev, Chou, and Kroner (1992) provide a nice overview of the issue. Furthermore, Engle and Ng (1993), Glosten, Jagannathan, and D. E. Runkle (1993), Bekaert and Wu (2000), and others have demonstrated that asymmetric GARCH models outperform those that do not allow the asymmetry, i.e., negative return shocks increase conditional volatility more than the positive return shocks. Therefore we estimate asymmetric GARCH(1,1) models and then produce VaR estimates by assuming conditional normality of the return. We also estimated several other ARCH models, with and without the asymmetric volatility specification. The regular GARCH(1,1) and asymmetric GARCH(1,1) produce similar performances in terms of the VaR test described below. We estimated the 5% VaRs of the S&P 500 Index estimated by the ARCH regression quantiles. and the asymmetric GARCH(1,1) model with the conditional normality assumptions. We see that these two series actually track each other pretty well, although the VaR series estimated by regression quantile seem to be higher than the GARCH VaR during low volatility periods. However, during very volatile markets, as during the 1987 market crash, the GARCH plus normality approach produces much higher VaR estimates. This could be due to the fact that large return shocks produce large volatility estimates in the GARCH setting. Value at risk after a market crash could be too high based on this approach. The quantile regression approach seems to generate an increased VaR at a more reasonable level. To measure the relative performance more accurately, we compute the percentage of realized returns that are below the negative estimated VaRs. The results are reported in Table 8. The top panel of the table presents the percentages for the VaRs estimated by the ARCH regression quantile model, and the bottom panel for the VaRs estimated by the asymmetric GARCH model with the conditional normal return distribution assumption. For the 1% VaR, we see that the regression quantile method produces the percentage that is closer to the 1% mark for all five series. The GARCH approach seems to underestimate the VaRs consistently. For the 5% VaR, the regression quantile method produces the percentage that is closer to the 5% mark for all series, except for the FTSE 100. But now the GARCH approach seems to overestimate the VaRs consistently. To look at this more closely we extend the analysis for the S&P 500


12

Index. We estimate VaRs using the two methods at 2%, 4%, 6%, 10%, 15%. Now we have a total of 7 percentage levels. The regression quantile method produces the closest percentage at all percentage levels, and the percentages scatter around the true value. However, the GARCH method seems to underestimate VaRs for the smaller percentages (1% and 2%), and overestimate VaRs for the larger percentages (larger than or equal to 4%). Table 8: VaR Model Performance Comparison

% VaR 1% 2% 4% 5% 6% 10% 15% VaR by Regression Quantile

S&P 500 1.339 1.925 4.168 5.2874 6.276 9.791 14.819

Nikkei 225 1.340 2.011 4.312 5.7084 6.581 10.210 14.564

FTSE 100 0.694 1.867 3.658 5.5868 5.232 8.951 12.372

Hang Seng 0.755 2.113 4.222 4.8902 5.512 9.348 13.558 VaR, by GARCH Normality Assumption S&P 500 1.1976 1.7964 3.1936 4.0918 4.9900 7.6846 12.4750

Nikkei 225 1.2974 1.9960 3.5928 4.6906 5.2894 8.5828 12.3752

FTSE 100 0.9980 1.6966 2.9940 3.4930 3.8922 6.5868 9.7804

Hang Seng 1.8962 2.8942 3.3932 3.6926 4.1916 7.1856 10.9780

VaR by RiskMetrics

S&P 500 0.4990 0.4990 0.6986 0.7984 0.7984 2.0958 3.5928

Nikkei 225 0.5988 0.7984 0.9980 0.9980 1.2974 2.2954 1.4910

FTSE 100 0.1996 0.1996 0.2994 0.7984 0.8982 1.7964 3.6926 Hang Seng 0.7984 0.8982 1.3972 1.3972 1.5968 2.6946 3.7924

This table reports the coverage ratios, i.e., the percentage of realized returns that are below the estimated VaRs. The top panel reports the performance of the VaRs estimated by the ARCH regression quantile model. The middle panel reports the results for VaRs estimated by the asymmetric GARCH model with the conditionally normal return distribution assumption. The bottom panel reports the results for VaRs estimated by the RiskMetrics method. The five indexes we analyzed are quite different in their risk characteristics as discussed above. The ARCH quantile regression approach seems to be robust and can consistently produce very good estimates of the VaRs at different percentage (probability) levels. The asymmetric GARCH model, being a very good volatility model, is not able to produce good VaR estimates with the normality assumption. The ARCH quantile regression model does not assume normality and is well suited to hand negative skewness and heavy tails. CONCLUDING COMMENTS In this paper we estimate value at risk using the quantile regression approach pioneered by Koenker and Bassett (1978). Comparing to the widely use RiskMetric method and other methods based on distributional assumptions, this method does not assume a particular conditional distribution for the returns. This is particularly important in VaR estimation because return data are well-known to be non-Gaussian. We apply the model to weekly return series of five major world equity market indexes: the U.S. S&P 500 Index, the Japanese Nikkei 225 Index, the U.K. FTSE 100 Index, the Hong Kong Hang Seng Index, and the Singapore Strait Times Index. The empirical results found that the quantile regression based method is more robust than RiskMetrics. These results regarding VaR estimation may have important implications for risk management practices.There are several directions for future research. First, our analysis in this paper is based on univariate analysis. Informative covariates may be introduced to improve the accuracy of estimation. Second, nonlinear models such as Copula models and GARCH


13

models may be considered to take into account nonlinearity in financial time series. We hope to explore these in future research. REFERENCES Andrews, D.W.K., (1991) "Heteroskedasticity and autocorrelation consistent covariance matrix estimation," Econometrica, 59, p. 817-858. Barrodale, I. and F.D.K. Roberts, (1974) "Solution of an overdetermined system of equations in the 11 norm," Communications of the ACM, 17, p. 319-320. Beder, T. S., (1995) "VAR: seductive but dangerous," Financial Analysts Journal, September-October 1995, p. 12-24. Bekaert, G., and G. Wu, (2000) "Asymmetric volatility and risk in equity markets," Review of Financial Studies 13, p. 1-42. Blankley, A., R. Lamb, and R. Schroeder, (2000) "Compliance with SEC disclosure requirements about market risk," Journal of Derivatives 7, Spring 2000, p. 39-50. Bollerslev, T., R. Y. Chou, and K. F. Kroner, (1992) "ARCH modeling in finance.” Journal of Econometrics 52, p. 5-59. Boos, D., (1984) "Using Extreme Value Theory to Estimate Large Percentiles," Technometrics 26, p. 33-39. Boudoukh, J., M. Richardson, and R. F. Whitelaw, (1998) "The best of both worlds." Risk 11, p. 64-67. Bouyé, E., Salmon, M., (2008) “Dynamic copula quantile regressions and tail area dynamic dependence in forex markets”, Manuscript, Financial Econometrics Research Centre, Warwick Business School, UK. Buchinsky, M., (1994) "Changes in the U.S. wage structure 1963-1987: application of quantile regression", Econometrica 62, p. 405-458. Chen, X., R. Koenker and Z. Xiao (2009) “Copula-based nonlinear quantile autoregression”, The Econometrics Journal, Volume 12, p. 50—67. Cowles, A., and H. Jones, (1937) "Some a posteriori probabilities in stock market action", Econometrica, 5, p. 280-294. Duffie, D., and J. Pan, (1997) "An overview of value at risk”, Journal of Derivatives. 4, 7-49. Dowd, K., (1998), Beyond Value at Risk: The New Science of Risk Management. John Wiley Sons, England. Engle, R. F., and V. K. Ng, (1993) "Measuring and testing the impact of news on volatility,” Journal of Finance 48, p. 1749-1778. Engle, R. F., and S. Manganelli, (1999) "CAViaR: Conditional autoregressive value at risk by regression quantiles," working paper, University of California, San Diego.


14

Glosten, L. R., R. Jagannathan, and D. E. Runkle, (1993) "On the relation between the expected value and the volatility of the nominal excess return on Stocks," Journal of Finance 48, p. 1779-1801. Gutenbrunner, C. and J. Jureckova, (1992), "Regression quantile arid regression rank score process in the linear model and derived statistics." Annal of Statistics. 20, p. 305-330. Hall, P., and S. J. Sheather, (1988) "On the distribution of a studentized quantile." Journal of Royal Statistical Society B. 50, 381-391. Hendricks, D., (1996) "Evaluation of value at risk models using historical data." FederalReserve Bank of New York Economic Policy Review 2, p. 39-69. Hong, Y., (1996) "Consistent testing for serial correlation of unknown form," Econometrica 64, p. 837-864. Hong, Y., (1999) "Hypothesis testing in time series via the empirical characteristic function: a generalized spectral density approach," Journal of American Statistical Association 94, p. 1201-1220. Jeong, S.O., Nonparametric estimation of VaR, Journal of Applied Statistics, 2009. Jorion, Philippe, (1997) “Value at risk: The new benchmark for controlling market risk”, Irwin Professional Pub., Chicago. Koenker, R.., and G. Bassett, (1978) "Regression quantiles," Econometrica 84, p. 33-50. Koenker, R. and Z. Xiao, (2006) “Quantile Autoregression”, Journal of the American Statistical Association, Vol. 101, No. 475, p. 980-1006. Koenker, R., and Q. Zhao, (1996) "Conditional quantile estimation and inference for ARCH models," Econometric Theory 12, p. 793-813. Koul, H., and Mukherjee, (1994), "Regression quantiles and related processes under long range dependence," Journal of Multivariate Statistics, 51, p. 318-337. Koul, H. and E. Saleh (1992), Autoregression quantiles and related rank-score process," Technical Report RM 527, Michigan State University. Ljung, G, and G. Box, (1978) "On a measure of lack of fit in time series models." Biometrica, 66, p. 265-270. McNeil, A., (1998) "Calculating (quantile risk measures for financial time series using extreme value theory." working paper, University of Zurich. Neftci, S., (2000) "Value at risk calculations, extreme events, and tail estimation,” Journal of Derivatives 7, Spring 2000, p. 23-37. Portnoy, S, (1991) “Asymptotic behavior of regression quantiles in non-stationary, dependent cases”, Journal of Multivariate Analysis 38, Issue 1, p. 100-113. Portnoy, S. L., and R., Koenker (1997) "The gaussian hare and the Laplacian tortoise: computability of squared-error vs. absolute-error estimator," Statistical Science 12, p. 279-300.


15

Powell (1986), "Censored regression quantiles"„Journal of Econometrics 32, p. 143-155. Saunders, A., (1999) Financial Institutions Management: A Modern Perspective, Irwin Series in Finance. Sheather, S., J., and J. S. Maritz, (1983) "An estimate of the asymptotic standard error of the sample median," Australian Journal of Statistics 25, p. 109-122. Siddiqui, M., (1960) "Distribution of quantiles in samples from a bivariate population," J. Res. Nat. Bur. Standards 64B, p. 145-150. BIOGRAPHY Hongtao Guo is Assistant Professor of Accounting, Bertolon School of Business, Salem State University, 352 Lafayette Street, Salem, Massachusetts, 01790,U.S.A.; [email protected], Miranda Lam is Associate Professor of Finance, Bertolon School of Business, Salem State University, 352 Lafayette Street, Salem, Massachusetts, 01790,U.S.A.; [email protected], Guojun Wu is Professor of Finance, Bauer College of Business, University of Houston, 334 Melcher Hall, Houston, Texas 77204, U.S.A.; [email protected], Zhijie Xiao is Professor of Economics, Boston College, 140 Commonwealth Avenue, Boston, Massachusetts, 02467, U.S.A.; [email protected].


17

PRE-CLOSE TRANSPARENCY AND PRICE EFFICIENCY AT MARKET CLOSING: EVIDENCE

FROM THE TAIWAN STOCK EXCHANGE Cheng-Yi Chien, Feng Chia University

ABSTRACT

This paper examines the impact of increased pre-close transparency on the effectiveness of stock closing call. On January 1, 2003, the Taiwan Stock Exchange increases pre-close transparency by disclosing the best five bid and ask prices and related unexecuted orders before market closing. At the same time, the Taiwan Stock Exchange does not disseminate any information about the limit-order book during the five-minute closing call period. This institutional change presents an opportunity to analyze how an increase in pre-close transparency affects informed trading and price efficiency near market closing. Empirical results show that an increase in pre-close transparency enhances the price efficiency of stock closing call, implying that informed traders will increase their trading during stock closing call following pre-close transparency increases. JEL: G14, G15, G18 KEYWORDS: Transparency, Closing call, Price efficiency, Taiwan Stock Exchange INTRODUCITON

his paper investigates whether an increase in pre-close transparency alters the effectiveness of stock closing call. Previous studies have examined the impact of stock closing call on price efficiency at market closing, and conclude that the introduction of stock closing call could enhance

market quality (e.g., Pagano and Schwartz, 2003; Huang and Tsai, 2008). On the other hand, prior studies have examined the effects of transparency on trading behavior of informed traders and the results to date are far from conclusive. One argues that informed traders trade more accurately in a transparent environment since they could tap the liquidity offered by the limit-order book more efficiently (e.g., Madhavan et al., 2005). The other suggests that informed traders prefer markets with less transparency to avoid revealing their private information (e.g., Chowdhry and Nanda, 1991; Comerton-Forde and Rydge, 2006). While there is no consensus in the literature on the influence of transparency on trading behavior of informed traders, this paper provides additional evidence regarding this issue by examining the impact of pre-close transparency on the efficacy of stock closing call. Starting from January 1, 2003, the Taiwan Stock Exchange (TWSE) enhances transparency by disclosing the best five bid and ask prices and related unexecuted orders in the trading period between 9:00 and 13:25. At the same time, the TWSE disseminates no information regarding the limit-order book during the five-minute closing call period surrounding this institutional change. Thus, this unique design provides an opportunity to examine how an increase in pre-close transparency affects informed trading near market closing and then price efficiency of stock closing call. The empirical results indicate that the quotation information, including bid-ask spreads and market depth, do not change significantly near market closing following pre-close transparency increases. However, this change in pre-close transparency results in larger trading volume, and then induces higher adjusted R-square of market model and less absolute return autocorrelation at market closing. The findings provide evidence that the increase in pre-close transparency could enhance the price efficiency of stock closing call. Thus, this paper conclude that, after pre-close transparency increases, informed traders will increase

T

CY Chien | IJBFR ♦ Vol. 7 ♦ No. 2 ♦ 2013

18

their trading during stock closing call when disclosing no limit-order book information at that time. The findings support Chowdhry and Nanda (1991) that informed traders prefer to trade in an opaque environment. The remainder of this study is organized as follows. Section 2 provides a brief literature review. Section 3 develops the research hypotheses. Section 4 describes how the sample is chosen and the methodology used in this paper. Empirical results are presented in section 5. Section 6 concludes the paper. LITERATURE REVIEW Previous research examines whether an introduction of stock closing call leads to an improvement in market quality. Most of previous literature concludes that stock closing call could enhance market quality. Hillion and Suominen (1998) examine the closing price behavior of the CAC 40 stocks on the Paris Bourse, and find that there exists high price volatility and bid-ask spreads near market closing due to price manipulation. Hillion and Suominen (1998) thus prompt the Paris Bourse to implement the closing call auction. Hillion and Suominen (2004) then develop a theoretical model of closing price manipulation and suggest that the call auction is the optimal closing mechanism because it could reduce price manipulation. Similarly, Pagano and Schwartz (2003) examine the impact of the introduction of stock closing call auction on market quality of Paris Bourse. Pagano and Schwartz (2003) find that the introduction of stock closing call reduces transaction costs and sharpens price discovery at the end of the day. In addition, Pagano and Schwartz (2005) examine the impact of stock closing call on price determination on the NASDAQ. Pagano and Schwartz (2005) find that stock closing call also improve the market quality for the stocks listed in the Russell 2000. Moreover, the introduction of stock closing call also improves the market quality in the Asia-Pacific stock markets. Comerton-Forde et al. (2007) examine whether the Singapore Exchange introduces opening and closing call auctions affect stock market quality, and find that the introduction of call auctions enhances the market quality at both market opening and closing. Huang and Tsai (2008) also examine the effects of the introduction of stock closing call on the TWSE. Huang and Tsai (2008) find that stock closing call could reduce the price volatility at market closing and enhances the market efficiency by reducing noise in stock closing prices. On the other hand, the effects of transparency have attracted considerable attention from practitioners, researchers, and policymakers. Transparency is defined as the availability of information regarding participants’ buy and sell orders before executed on the limit-order book (O’Hara, 1995). Prior studies assess the change in market quality associated with an increase in transparency; however, there is no consensus in the literature on whether increased transparency results in an improvement in market quality. Baruch (2005) constructs a theoretical model infer how an increase in transparency affects market quality. Baruch (2005) concludes that increased transparency reduces the bid-ask spread and increases the informational efficiency of stock price. Also, Boehmer et al. (2005) investigate the impact of the introduction of the OpenBook that provides limit-order book information to traders off the exchange floor on the NYSE. Boehmer et al. (2005) indicate that greater transparency leads to higher market liquidity and greater price efficiency. Examining a reduction in the transparency of the limit-order book on the Island ECN, Hendershott and Jones (2005) find that the dominant market for the three most active ETF’s decreases market quality following transparency decreases, and the market quality then improves when Island later redisplay its orders. Besides, Chung and Chuwonganant (2009) examine the influence of transparency on market quality using data surrounding the introduction of the SuperMontage which is a fully integrated order display and execution system for NASDAQ-listed stocks. Chung and Chuwonganant (2009) show that both bid-ask spreads and return volatility decline significantly after the implementation of the


19

SuperMontage, and conclude that the SuperMontage does improve the market quality on the NASDAQ. In contrast, Madhavan (1996) study the same question by theoretical model and their viewpoint is different from Baruch (2005). Furthermore, Madhavan et al. (2005) examine such an increase by empirical study to support the theoretical inference of Madhavan (1996). Madhavan et al. (2005) test whether an increase in transparency results in worse market quality after the Toronto Stock Exchange disseminates information about bid and ask depth at the top four price levels in the limit-order book. Madhavan et al. (2005) find that increased transparency results in higher trade execution costs and price volatility. In summary, previous studies have provided considerable insight into the efficacy of stock closing call. Previous research has also generated interesting findings about the effects of transparency. However, no work has been conducted into the influence of pre-close transparency on the effectiveness of stock closing call when disclosing no limit-order book information at that time. Thus, this paper attempts to bridge this gap. HYPOTHESES As suggested by Zhao and Chung (2006), informed traders buy when prices are below the estimates of fundamental value and sell when prices are above the estimates; hence, their trading could move prices to the estimates of fundamental value. Since informed traders could estimate values accurately, their trading then makes prices more informative and efficient. Based on Zhao and Chung (2006), two competing hypotheses have been offered to explain the impact of pre-close transparency on informed trading and then price efficiency near market closing. On the one hand, an increase in pre-close transparency may result in lower price efficiency at market closing. Madhavan et al. (2005) states that, if transparency increases informed traders’ expected profits by allowing them to tap the liquidity offered by the limit-order book more efficiently than in an opaque environment, then informed traders trade more accurately in a transparent environment, speeding up the process of price discovery. Since the TWSE increases pre-close transparency but disseminates no limit-order book information during stock closing call period, an increase of pre-close transparency should decrease price efficiency at market closing. On the other hand, Chowdhry and Nanda (1991) suggest that informed traders prefer markets with less transparency to avoid revealing their private information, implying informed traders prefer to trade in an opaque environment. Also, Comerton-Forde and Rydge (2006) and Grossman and Miller (1988) indicate that informed traders prefer to trade in an opaque market where they could retain their informational advantage. Since Zhao and Chung (2006) argue that the trading of informed traders facilitates price efficiency, this paper predicts that the opacity of stock closing call should attract more informed trading after pre-close transparency increases, and then results in higher price efficiency of stock closing call. DATA AND MATHODOLOGY Before the end of 2002, only the unexecuted orders of the limit-order book at the best bid and ask prices are disclosed. Due to the belief that increased transparency leads to a fairer and more efficient market, the TWSE enhances pre-close transparency by disclosing the best five bid and ask prices and related unexecuted orders on January 1, 2003. Nevertheless, the TWSE disseminates no information about the limit-order book during the five-minute closing call period at the same time. This makes the limit-order book at market closing a black box for traders. The price efficiency of stock closing call after pre-close transparency increases is therefore valuable to investigate.


20

Thus, this paper utilizes intraday data to analyze the effects of an increase in pre-close transparency on the effectiveness of stock closing call in the Taiwan stock market. The data is taken from the Taiwan Economic Journal (TEJ) database, which provides the price and volume of executed orders and quotation information of limit-order book for each stock The sample period is divided into two sub-periods, namely, before and after the increase of pre-close transparency. Before, from October 8, 2002 to December 31, 2002 (a total of 60 trading days), represents only the unexecuted orders at the best bid and ask prices are disclosed. After, from April 1, 2003 to June 25, 2003 (a total of 60 trading days), represents the unexecuted orders at the best five bid and ask prices are disclosed. Due to the data availability on the TEJ database, the After period in this study is starting from April 1, 2003. Moreover, this paper selects sample stocks according to the following criteria: (1) the stocks are common stocks on the TWSE; (2) the stocks are traded both before and after pre-close transparency increases; and (3) the stocks are the top 50 largest firms in the testing sample. Then, this paper gathers data with each five-minute interval during the last half-hour of the selected days. To evaluate the impact of increased pre-close transparency on the effectiveness of stock closing call, this paper begins by calculating relative quoted spread and relative effective spread in each five-minute interval during 13:00-13:30. Adopting relative proxies could control for differences across stocks and time. Relative quoted spread and relative effective spread are the two standard measures of transaction costs. Relative quoted spread is estimated as the difference between the best ask price and bid price and then divided by the midpoint of the bid and ask price. In the interest of completeness, this paper also reports relative effective spread for each stock, measured as twice the difference between the transaction price and the midpoint of the bid and ask price and then divided by the midpoint of the bid and ask price. Similar to calculating spreads, market depth is measured by relative bid depth and relative ask depth. Relative bid depth for each stock is the quoted volume at the highest bid price in each five-minute interval divided by the total daily trading volume. Relative ask depth for each stock is the quoted volume at the lowest ask price in each five-minute interval divided by the total daily trading volume. This paper then utilizes relative trading volume to measure the impact of increased pre-close transparency on trading activities near market closing. Both trading volume measured by lots and trading volume measured by dollar are used in this study. The change of relative trading activities near market closing surrounding pre-close transparency increases could provide some evidence regarding the trading behavior of informed traders. Relative trading volume is the trading volume in each five-minute interval relative to the total trading volume for the entire day. Pagano and Schwartz (2003) propose a test of price efficiency via examining the price synchronicity across a set of stocks. An increase in the adjusted R-squares of market model would signal greater synchronicity in stock prices, thereby improving price efficiency. Thus, a comparison of the adjusted R-squares of market model at market closing surrounding pre-close transparency increases provides evidence regarding the impact of pre-close transparency on price efficiency of stock closing call. The market model is estimated for stock return in each five-minute interval during 13:00-13:30. The Taiwan Weighted Stock Index (TWSI) is used as a proxy for the market portfolio. Therefore, the market model could be estimated as follows:

itdmtdit1it0itd eRbbR ++= (1) where Ritd is the return of stock i in time interval t for day d, Rmtd is the corresponding market return, eitd is a random error term, and bit0 and bit1 are coefficients to be estimated.


21

Return autocorrelation is another measure of price efficiency. Boehmer et al. (2005) indicate that a more efficient price process would be closer to a random walk and therefore exhibit less return autocorrelation (both positive and negative). Thus, this paper estimates absolute value of return autocorrelation rolled on a five-minute interval from 13:00 to 13:30 as follow:

|)R,Corr(R| |AutoCorr| itdd1,-ti,it = (2) where AutoCorrit is the return autocorrelation of stock i in time interval t. RESULTS Table 1 and Table 2 present the preliminary analysis regarding the effects of the increase in pre-close transparency on quotation information near market closing. Table 1 shows the differences of bid-ask spreads during 13:00-13:30 surrounding pre-close transparency increases. Panel A of Table 1 indicate that the relative quoted spreads near market closing change insignificantly at the 5% level following pre-close transparency increases. Panel B of Table 1 also indicates that the change of the relative effective spreads nearing market closing are not significant after pre-close transparency increases. Table 1: Bid-ask Spread Near Market Closing before and bfter Pre-Close Transparency Increases

Time Before After △Diff. = Diff.After - Diff.Before

Panel A: Relative Quoted Spreads

(1) 13:00-13:05 0.0043 0.0043 (2) 13:05-13:10 0.0043 0.0042 (3) 13:10-13:15 0.0044 0.0043 (4) 13:15-13:20 0.0044 0.0043 (5) 13:20-13:25 0.0044 0.0043 (6) 13:25-13:30 0.0045 0.0045

Diff. = (6) - (5)

0.0001* (2.39)

0.0001** (3.87)

0.0000 (0.53)

Panel B: Relative Effective Spreads

(1) 13:00-13:05 0.0043 0.0042 (2) 13:05-13:10 0.0043 0.0042 (3) 13:10-13:15 0.0043 0.0042 (4) 13:15-13:20 0.0043 0.0042 (5) 13:20-13:25 0.0043 0.0043 (6) 13:25-13:30 0.0045 0.0045

Diff. = (6) - (5)

0.0002** (3.01)

0.0002** (4.11)

0.0000 (0.07)

This table presents the bid-ask spread near market closing before and after pre-close transparency increases. Before, from October 8, 2002 to December 31, 2002 (a total of 60 trading days), represents only the unexecuted orders at the best bid and ask prices are disclosed. After, from April 1, 2003 to June 25, 2003 (a total of 60 trading days), represents the unexecuted orders at the best five bid and ask prices are disclosed. Numbers in parentheses denote t-statistics. *Significant at the 5% level. **Significant at the 1% level. Table 2 shows similar results in the differences of relative bid depth and relative ask depth. Panel A or Panel B of Table 2 indicate that the relative bid depth and relative ask depth near market closing do not change significantly surrounding pre-close transparency increases. Combing the results of bid-ask spreads and market depth, we find that increased pre-close transparency has no significant impact on the quotation information of limit-order book during the closing call period.


22

Trading Activities This paper now examines the relative trading volume near market closing before and after pre-close transparency increases. Panel A of Table 3 shows that, from 13:20-13:25 to 13:25-13:30, the relative trading volume measured by lots significantly increases from 0.0332 to 0.0426 with the t-value of 5.90 in the before period and increases from 0.0311 to 0.0507 with the t-value of 11.11 in the after period. Furthermore, after pre-close transparency increases, the difference of relative trading volume measured by lots at market closing significantly increases from 0.0094 to 0.0196 with the t-value of 5.87. Similarly, Panel B of Table 3 shows that, after pre-close transparency increases, the difference of relative trading volume measured by dollar near market closing significantly increases from 0.0093 to 0.0196 with the t-value of 5.89. The results suggest that, following pre-close transparency increases, there exists a significant increase in trading activities during the closing call period. Table 2: Market Depth Near Market Closing Before and After Pre-Close Transparency Increases


Panel A: Relative Bid Depth

(1) 13:00-13:05 0.0379 0.0580 (2) 13:05-13:10 0.0346 0.0582 (3) 13:10-13:15 0.0369 0.0598 (4) 13:15-13:20 0.0387 0.0523 (5) 13:20-13:25 0.0381 0.0555 (6) 13:25-13:30 0.0364 0.0616

Diff. = (6) - (5)

-0.0017 (-0.51)

0.0061* (2.36)

0.0078 (1.83)

Panel B: Relative Ask Depth

(1) 13:00-13:05 0.0211 0.0564 (2) 13:05-13:10 0.0204 0.0546 (3) 13:10-13:15 0.0201 0.0544 (4) 13:15-13:20 0.0194 0.0540 (5) 13:20-13:25 0.0194 0.0537 (6) 13:25-13:30 0.0224 0.0532

Diff. = (6) - (5)

0.0030** (4.20)

-0.0005 (-0.23)

-0.0035 (-1.43)

This table presents the market depth near market closing before and after pre-close transparency increases. Before, from October 8, 2002 to December 31, 2002 (a total of 60 trading days), represents only the unexecuted orders at the best bid and ask prices are disclosed. After, from April 1, 2003 to June 25, 2003 (a total of 60 trading days), represents the unexecuted orders at the best five bid and ask prices are disclosed. Numbers in parentheses denote t-statistics. *Significant at the 5% level. **Significant at the 1% level. Price efficiency In order to test the price efficiency of stock closing call, this paper then examines the adjusted R-squares of market model and absolute return autocorrelations near market closing before and after pre-close transparency increases. Panel A of Table 4 presents the adjusted R-squares of market model during 13:00-13:30 surrounding the increase of pre-close transparency. The results indicate a significant improvement in price discovery at market closing after pre-close transparency increases. In the before period, the adjusted R-square reduces significantly from 0.1182 in 13:20-13:25 to 0.0620 in 13:25-13:30 with the t-value of -4.02. In the after period, the adjusted R-square increase significantly from 0.0477 in 13:20-13:25 to 0.2064 in 13:25-13:30 with the t-value of 6.56. As comparing the difference of the adjusted R-square between 13:20-13:25 and 13:25-13:30, this paper finds that the difference of the adjusted R-square increases significantly from -0.0562 to 0.1587 with the t-value of 9.53 following transparency increases. In particular, the smallest adjusted R-square in the before period occurs in the time interval 13:25-13:30, but the largest adjusted R-square in the after period occurs in the time interval


23

13:25-13:30. Thus, the evidence indicates that price efficiency enhances significantly during the closing call period after pre-close transparency increases, because a tighter fit between the individual stock returns and market returns at market closing in the after period. Table 3: Trading Volume Near Market Closing Before and after Pre-Close Transparency Increases


Panel A: Relative Trading Volume Measured by Lots

(1) 13:00-13:05 0.0186 0.0169 (2) 13:05-13:10 0.0189 0.0189 (3) 13:10-13:15 0.0203 0.0208 (4) 13:15-13:20 0.0223 0.0226 (5) 13:20-13:25 0.0332 0.0311 (6) 13:25-13:30 0.0426 0.0507

Diff. = (6) - (5)

0.0094** (5.90)

0.0196** (11.11)

0.0102** (5.87)

Panel B: Relative Trading Value Measured by Dollar

(1) 13:00-13:05 0.0186 0.0169 (2) 13:05-13:10 0.0188 0.0189 (3) 13:10-13:15 0.0203 0.0208 (4) 13:15-13:20 0.0222 0.0226 (5) 13:20-13:25 0.0332 0.0311 (6) 13:25-13:30 0.0425 0.0507

Diff. = (6) - (5)

0.0093** (5.87)

0.0196** (11.11)

0.0103** (5.89)

This table presents the trading volume near market closing before and after pre-close transparency increases. Before, from October 8, 2002 to December 31, 2002 (a total of 60 trading days), represents only the unexecuted orders at the best bid and ask prices are disclosed. After, from April 1, 2003 to June 25, 2003 (a total of 60 trading days), represents the unexecuted orders at the best five bid and ask prices are disclosed. Numbers in parentheses denote t-statistics. *Significant at the 5% level. **Significant at the 1% level. Panel B of Table 4 presents the absolute return autocorrelations near market closing before and after pre-close transparency increases. In a more efficient market mechanism, there exists less return autocorrelation, either positive or negative. Thus, absolute return autocorrelations provides another method to analyze price efficiency. Panel B of Table 4 shows that, in the after period, the absolute return autocorrelation becomes smaller significantly from 0.2735 in 13:20-13:25 to the lowest value, 0.1888, in 13:25-13:30 with the t-value of -3.70. After pre-close transparency increases, the difference of absolute return autocorrelation between 13:20-13:25 and 13:25-13:30 significantly decreases from 0.0086 to -0.0847 with the t-value of -3.52. The findings of absolute return autocorrelations indicate that the price efficiency during the closing call period indeed improve after increased pre-close transparency. Regression Analysis Table 5 presents the regression analysis of trading activities and price efficiency at market closing surrounding pre-close transparency increases. The following regression model is used in this study to access the effects of increased pre-close transparency on the effectiveness of stock closing call:

iiiii Ln(Size)(1/Price)DummyDV eββββ ++++= 3210 (3) where DVi is each dependent variable for firm i, including the differences of relative trading volume, adjusted R-square of market model, and absolute return autocorrelation between 13:20-13:25 and 13:25-13:30. The Dummyi is equal to 1 if the data is in the After period and is equal to 0 if the data is in the Before period. The inverse of price (1/Price)i and the natural logarithm of market capitalization


24

Ln(Size)i are used to control for heteroskedasticity caused by variation between different price levels as well as firm size. εi is the error term. Table 4: Price Efficiency Near Market Closing Before and After Pre-Close Transparency Increases


Panel A: Adjusted R-square of Market Model

(1) 13:00-13:05 0.3250 0.1126 (2) 13:05-13:10 0.2042 0.1526 (3) 13:10-13:15 0.2024 0.1117 (4) 13:15-13:20 0.1653 0.1185 (5) 13:20-13:25 0.1182 0.0477 (6) 13:25-13:30 0.0620 0.2064

Diff. = (6) - (5)

-0.0562** (-4.02)

0.1587** (6.56)

0.2149** (9.53)

Panel B: Absolute Return Autocorrelation

(1) 13:00-13:05 0.1878 0.2758 (2) 13:05-13:10 0.2139 0.2163 (3) 13:10-13:15 0.2129 0.2238 (4) 13:15-13:20 0.2083 0.2765 (5) 13:20-13:25 0.1930 0.2735 (6) 13:25-13:30 0.2016 0.1888

Diff. = (6) - (5)

0.0086 (0.39)

-0.0847** (-3.70)

-0.0933** (-3.52)

This table presents the adjusted R-squares of market model and absolute return autocorrelations near market closing before and after pre-close transparency increases. Before, from October 8, 2002 to December 31, 2002 (a total of 60 trading days), represents only the unexecuted orders at the best bid and ask prices are disclosed. From April 1, 2003 to June 25, 2003 (a total of 60 trading days), represents the unexecuted orders at the best five bid and ask prices are disclosed. Numbers in parentheses denote t-statistics. *Significant at the 5% level. **Significant at the 1% level. Table 5 shows that, even controlling firm characteristics, the relative trading volume measured by lots, relative trading volume measured by dollar, adjusted R-square of market model, and absolute return autocorrelation all change significantly at market closing after pre-close transparency increases. From 13:20-13:25 to 13:25-13:30, the relative trading volume measured by lots, relative trading volume measured by dollar, and adjusted R-square of market model significantly increase, but the absolute return autocorrelation decreases significantly. The findings of regression analysis also reveal that, after pre-close transparency increases, informed traders will increase their trading during the closing call period and then result in higher efficiency of closing price. CONCLUSION This paper analyzes the impact of increased pre-close transparency on the effectiveness of stock closing call. On January 1, 2003, the TWSE increase pre-close transparency by disclosing the best five bid and ask prices and related unexecuted orders. However, the TWSE disseminates no information about the limit-order book during the five-minute closing call period. Hence, this institutional change provides an opportunity to examine how an increase in pre-close transparency affects the price efficiency of stock closing call. According to Madhavan et al. (2005) and Chowdhry and Nanda (1991), this paper proposes two competing hypotheses to explain how an increase in pre-close transparency affect trading behavior of informed traders and then price efficiency during stock closing call. The empirical results show that, following pre-close transparency increases, the trading activities and price efficiency at market closing enhances on the TWSE. The relative trading volume and adjusted R-square of market model increase


25

significantly at market closing after pre-close transparency increases; furthermore, the absolute return autocorrelation becomes smaller at the end of the day. The findings suggest that increased pre-close transparency could enhance the price efficiency of stock closing call when disseminating no information about the limit-order book at that time. Thus, this paper concludes that informed traders will increase their trading during stock closing call following pre-close transparency increases, and this is consistent with the prediction of Chowdhry and Nanda (1991) that informed traders prefer to trade in an opaque environment. This paper examines the relationship between pre-close transparency and price efficiency at market closing in a comprehensive way. However, due to the limited availability of intraday data, this paper does not have access to information on each trader’s type (e.g., individual investors versus institutional investors) which could provide more detailed regarding the relationship of pre-close transparency and informational efficiency at market closing. How individual and institutional investors trade at marketing closing surrounding pre-close transparency increases is an interesting topic for future research. Table 5: Regression Analysis of Trading Activities and Price Efficiency at Market Closing Surrounding

Pre-Close Transparency Increases

Dependent Variable Constant Dummy 1/Price Ln(Size) R2

Panel A: Trading activities

Relative Trading Volume Measured by Lots -0.0246 (-1.30)

0.0101** (4.30)

0.0169 (0.35)

0.0030 (1.85) 0.19

Relative Trading Volume Measured by Dollar -0.0247 (-1.30)

0.0102** (4.32)

0.0169 (0.35)

0.0030 (1.85) 0.19

Panel B: Price efficiency

Adjusted R-square of Market Model -0.5433* (-2.48)

0.2118** (7.77)

1.3020* (2.32)

0.0385* (2.08) 0.42

Absolute Return Autocorrelation 0.0303 (0.12)

-0.0935** (-2.92)

0.1604 (0.24)

-0.0025 (-0.12) 0.08

This table shows the regression analysis at market closing surrounding pre-close transparency increases. The dependent variables are the differences of relative trading volume, adjusted R-square of market model, and absolute return autocorrelation between 13:20-13:25 and 13:25-13:30. The Dummy is equal to 1 if the data is in the After period and is equal to 0 if the data is in the Before period. Before, from October 8, 2002 to December 31, 2002, represents only the unexecuted orders at the best bid and ask prices are disclosed. After, from April 1, 2003 to June 25, 2003, represents the unexecuted orders at the best five bid and ask prices are disclosed. Furthermore, the inverse of price 1/Price and the natural logarithm of market capitalization Ln(Size) are used to control for heteroskedasticity caused by variation between price levels as well as firm size. Numbers in parentheses denote t-statistics. *Significant at the 5% level. **Significant at the 1% level. REFERENCES Baruch, S. (2005) “Who Benefits from an Open Limit-Order Book?,” Journal of Business, vol. 78(4), p. 1267–1306. Boehmer, E., Saar, G. and L. Yu (2005) “Lifting the Veil: An Analysis of Pre-Trade Transparency at the NYSE,” Journal of Finance, vol. 60(2), p. 783–815. Chowdhry, B. and V. Nanda (1991) “Multi-Market Trading and Market Liquidity,” Review of Financial Studies, vol. 4(3), p. 483–511. Chung, K. and C. Chuwonganant (2009) “Transparency and Market Quality: Evidence from SuperMontage,” Journal of Financial Intermediation, vol. 18(1), p. 93–111. Comerton-Forde, C., Lau, S. and T. McInish (2007) “Opening and Closing Behavior Following the Introduction of Call Auctions in Singapore,” Pacific-Basin Finance Journal, vol. 15(1), p. 18–35.


26

Comerton-Forde, C. and J. Rydge (2006) “The Influence of Call Auction Algorithm Rules on Market Efficiency,” Journal of Financial Markets, vol. 9(2), p. 199–222. Grossman, S. and M. Miller (1988) “Liquidity and Market Structure,” Journal of Finance, vol. 43(3), p. 617–633. Hendershott, T. and C. Jones (2005) “Island Goes Dark: Transparency, Fragmentation, and Regulation,” Review of Financial Studies, vol. 18(3), p. 743–793. Hillion, P. and M. Suominen (1998) Deadline Effect of an Order Driven Market: An Analysis of the Last Trading Minute on the Paris Bourse. Global Equity Markets Conference Proceedings, Paris–Bourse and NYSE edited, Paris. Hillion, P. and M. Suominen (2004) “The Manipulation of Closing Prices,” Journal of Financial Markets, vol. 7(4), p. 351–375. Huang, Y. and P. Tsai (2008) “Effectiveness of Closing Call Auctions: Evidence from the Taiwan Stock Exchange,” Emerging Markets Finance and Trade, vol. 44(3), p. 5–20. Madhavan, A. (1996) “Security Prices and Market Transparency,” Journal of Financial Intermediation, vol. 5(3), p. 255–283. Madhavan, A., Porter, D. and D., Weaver (2005) “Should Securities Markets Be Transparent?,” Journal of Financial Markets, vol. 8(3), p. 265–287. O’Hara, M. (1995) Market Microstructure Theory. Blackwell Publishers, Cambridge. Pagano, M. and R. Schwartz (2003) “A Closing Call’s Impact on Market Quality at Euronext Paris,” Journal of Financial Economics, vol. 68(3), p. 439–484. Pagano, M. and R. Schwartz (2005) “NASDAQ’s Closing Cross,” Journal of Portfolio Management, vol. 31(4), p. 100–111. Zhao, X. and K. Chung (2006) “Decimal Pricing and Information-Based Trading: Tick Size and Informational Efficiency of Asset Price,” Journal of Business Finance and Accounting, vol. 33 (5)&(6), p. 753–766. ACKNOWLEDGEMENT I thank the editor, Professor Terrence Jalbert, and two anonymous referees for helpful comments and suggestions. I also thank Hsiu-Chuan Lee and Shih-Wen Tai for their excellent research assistance. All remaining errors are my own. I am also grateful for the financial support from the National Science Council of Taiwan (NSC97-2410-H-035-033). BIOGRAPHY Cheng-Yi Chien is an Assistant Professor at the Feng Chia University. His research appears in journals such as Journal of Futures Markets, Accounting and Finance, and Asia-Pacific Journal of Financial Studies. Contact information: Department of Finance, Feng Chia University, 100, Wenhwa Road, Taichung, Taiwan. E-mail: [email protected].


27

CHIEF EXECUTIVE COMPENSATION: AN EMPIRICAL STUDY OF FAT CAT CEOS

Dan Lin, Takming University of Science and Technology Hsien-Chang Kuo, Takming University of Science and Technology

Lie-Huey Wang, Ming Chuan University

ABSTRACT

This paper empirically tests the determinants of executive pay. In order to gain more understanding of the fat cat problem that have been subject to hot debate, we also examine a sample firms that suffer from the “fat cat problem”, defined as firms with poor performance while their Chief Executive Offers (CEOs) receive high compensation. Based on a sample of 903 US firms between 2007 and 2010, we find that there is a substitution effect between CEO compensation and the level of CEO ownership and that larger firms give higher pay to their CEOs. When the sample is limited to fat cat companies only, we find that tenure and firm size are significantly positively associated with CEO compensation. The firm size, leverage ratio and investment opportunities are found to be significantly associated with the CEO total compensation when the sample is limited to fat cat companies in the financial services industries. Overall, firm size appears to be the most important determinant of CEO compensation and that there is a general lack of linkage between pay and performance. The evidence thus calls for public attention for reexamining the effectiveness of current pay system. JEL: G34, M52 KEYWORDS: Executive Compensation, Fat Cat, Pay-Performance Relationship INTRODUCTION

he issue of “fat cats” became a hot button issue during the recent financial crisis in 2007 and 2008. Blinder (2009) suggests that the “perverse” incentive built into the compensation plans of many financial firms is one of the most fundamental causes of the financial crisis and surprisingly

receives little public attention. The incentives given to Chief Executive Officers (CEOs) and other top executives of large banks or investment banks have encouraged the excessive risk-taking by top managers, leading to the financial crisis. Most financial institutions link incentives of executives to short-term securities trading performance. Executives are encouraged to engage in short-term gambles and to focus their attention on short-term objectives instead of achieving sustainable growth objectives (Abou-El-Fotouh, 2010). Specifically, these institutions have failed to recognize that high incentives could lead to uncontrollable risks and this problem has been blamed for causing the financial crisis. Additionally, there have been increasing concerns about the escalation in executive compensation (Dong & Ozkan, 2008). In particular, the substantial rises in executive pay have far exceeded the increases in underlying firm performance (Gregg, Jewell, & Tonks, 2005). The review on CEO compensation by Frydman and Jenter (2010) shows that there was a dramatic increase in compensation levels from the mid‐1970s to the early 2000s in the US. Especially in the 1990s, the annual growth rates were more than 10% by the end of the decade. The increase in executive compensation was evident in firms of all sizes with larger firms experiencing even greater growth. The high level of CEO pay in the US has therefore brought about considerable debate and a lot of attention from academia and policy makers regarding executive compensation, in particular, the pay-setting process and the effectiveness of the compensation contracts.

T

D. Lin et al | IJBFR ♦ Vol. 7 ♦ No. 2 ♦ 2013

28

In the US market, the regulations place a strong emphasis on shareholder protection and information disclosure. As a result, most US firms are characterized by dispersed share ownership and low managerial ownership (Core, Guay, & Larcker, 2003). Compensation contracts therefore become particularly important in aligning the interests of managers and shareholders. The level of executive compensation and the linkage between compensation and firm performance have been extensively researched while no consistent results have been reported. For example, studies by Murphy (1985), Jensen and Murphy (1990), Hubbard and Palia (1995), and Ozkan (2011) all find a positive relation between pay and performance, supporting the agency theory. The agency theory argues that managers are self-serving and therefore, formal mechanisms such as compensation contracts are required in an attempt to align the interests of managers with that of shareholders (Jensen & Meckling, 1976; Fama & Jensen, 1983). In contrast, Ozkan (2007) does not find a significant relationship between CEO compensation and firm performance based on a sample of large UK companies for the fiscal year 2003-2004. In a more recent study, based on a sample of 390 UK non-financial firms for the period 1999-2005, Ozkan (2011) reports a significantly positive relationship between firm performance and CEO’s cash compensation but an insignificantly positive relationship between firm performance and total compensation. Therefore, the objectives of this study are twofold: exploring the determinants the executive pay and examining a sample of “fat cat companies”, defined as having poor performance while giving high compensation to their CEOs. Specifically, to achieve the first objective, we examine whether CEO and board characteristics, including CEO experience, measured by CEO tenure and CEO age, CEO shareholdings, and board size, are related to CEO compensation for a sample US companies between 2007 and 2010. The second objective is achieved by analyzing the characteristics of fat cat CEOs and fat cat companies, and investigating the determinants of fat cat CEO compensation. Thus, although there has already been extensive research on executive compensation, this paper contributes to the literature by focusing on fat cats that have been subject to hot debate. The remainder of this paper is organized into six sections. In Section 2, we review the prior empirical literature on executive compensation. The hypotheses tested in this study are discussed in Section 3. In Section 4, we describe the data and sample and specify the model used in the tests. Empirical results for the full sample and fat cat companies are presented in Section 5. A conclusion is provided in Section 6. LITERATURE REVIEW ON EXECUTIVE COMPENSATION Most companies are characterized by the separation of ownership and control where the ownership is held by diverse shareholders and the control is in the hands of top executives. As a result, shareholders are not able to monitor managers’ actions directly. According to the agency theory, these companies are likely to suffer from agency problems. That is, managers as the agents may not always act in the interest of the shareholders (i.e., the principals), thereby giving rise to conflicts of interests. One important control mechanism to align the interests of shareholders and managers and to mitigate the agency problems is to structure CEO compensation so that changes in executive wealth are linked to changes in stock price. By creating a pay-performance linkage in compensation contracts, executives have more incentives to maximize shareholder wealth (Core et al., 2003). Moreover, the risks between the principals and agents can be shared more equitably (Jensen & Meckling, 1976; Holmstrom, 1979; Cordeiro & Veliyath, 2003). An early paper by Finkelstein and Hambrick (1988) provides a synthesis on CEO compensation and suggests that there are two main set of factors that affect CEO compensation: first, the market factors, including managerial labor market, marginal products of CEOs, CEO discretion, firm size, firm performance, and human capital; secondly, the power and preferences of the board and CEO.


29

Consistent with this view, Ozkan (2007) finds that corporate governance mechanisms have a significant effect on the level of CEO compensation. Specifically, measures of board and ownership structures are found to explain a significant amount of cross-sectional variation in the CEO total compensation. Prior empirical research on the pay-performance link is extensive but conflicting results have been reported. For example, Jensen and Murphy (1990) find that increases in shareholder wealth are positively related with CEO pay. Main et al. (1996) find that the relation between pay and performance becomes more significant when executive options are included in total compensation. In contrast, Brick et al. (2006) document that firm underperformance is related to excessive pay to managers and directors, providing evidence of cronyism between CEOs and directors. Gregg et al. (2005) based on a sample of large UK firms find a weak relationship between pay, measured by total board and highest director pay, and performance while board structure, firm size, industry and firm risk are all significant determinants of executive compensation. Ozkan (2007) finds a positive but insignificant relationship between performance and executive compensation. To extend on earlier research that often reports weak or statistically insignificant relationship between pay and performance, Barkema and Gomez-Mejia (1998) propose a general research framework and argue that the inclusion of other criteria (such as the market, peer compensation, and individual characteristics), a firm’s governance structure, and contingencies (such as a firm’s strategy, R&D level, market growth, industry concentration, regulation, and national culture), can enhance our understanding of the determinants of executive pay. Therefore, in addition to performance, there are other factors that can affect executive pay. As argued by agency theorists, the governance structure of firms can mitigate the potential agency problem between managers and shareholders arising from the separation of ownership and control, and therefore, influence the way firms set their compensation packages (Ozkan, 2011). In fact, the board of directors plays an essential role in setting CEO compensation (Finkelstein & Hambrick, 1988; Boyd, 1994; Barkema & Gomez-Mejia, 1998; Chhaochharia & Grinstein, 2009). One theoretical explanation for the rapid acceleration in CEO compensation in recent years is the rent extraction behavior of managers; that is, the managerial power hypothesis. The theory argues that the excessive CEO pay is due to the greater power of executives over directors that allows the former to set their own pay and extract rents (Bebchuk et al, 2002; Bebchuk & Fried, 2004). Thus, an implication of the theory is that enhancing the independence of the board will improve corporate governance and prevent managers from extracting rents in the form of higher pay (Guthrie, Sokolowsky, & Wan). Chhaochharia and Grinstein (2009) conduct tests on whether independent directors are better monitors of CEOs and find that non-independent directors are associated with excessive CEO pay, consistent with the agency theory. However, Guthrie et al. argue that the results of Chhaochharia and Grinstein’s (2009) study are driven by two extreme outliers. Guthrie et al. re-test Chhaochharia and Grinstein’s (2009) data after removing outliers and find contrasting results; that is, independent directors do not constrain CEOs from obtaining excessive pay. Both Cosh and Hughes (1997) and Core et al. (1999) also do not find support for the agency theory. In particular, they find that firms with a higher proportion of non-executive directors, which are expected to be associated with greater monitoring by the board of directors, tend to pay more to their CEOs. Moreover, Ozkan (2007) based on a sample of UK companies in the year 2003 finds that board and ownership structures are significantly associated with CEO's total compensation. Core et al. (1999) report that larger boards pay more to their CEOs in terms of both cash compensation and total compensation. Guest (2010) who examines a comprehensive and long period dataset of 1,880 UK firms over the period 1983-2002 also reports a positive relationship between board size and the rate of increase in executive compensation, providing support for the argument that large boards suffer from the problems of less efficient decision-making and poor communication. In addition, Guest (2010) reports that the


30

proportion of non-executive directors is negatively associated with the rate of increase in executive compensation and is positively related to the pay-performance link, highlighting the monitoring role of non-executive directors in setting executive pay. Further, Alonso and Aperte (2011) examine whether board independence and equity-linked compensation are alternative instruments of corporate governance based on a sample of European firms and find that these two mechanisms are complementary. Specifically, CEOs receive less cash compensation but higher equity-linked compensation when the proportion of non-executive directors is higher. Thus, this paper adds to the literature on executive compensation by examining the determinants of executive compensation and focusing on a sample of fat cat companies. HYPOTHESES This study tests if CEO and board characteristics, including CEO experience, CEO shareholdings, and board size, are related to CEO compensation. The hypotheses tested are outlined below. CEO Experience As CEOs build a power base and gain voting control over time, they may exert influence over board composition and consequently, demand compensation packages that serve their own interests rather than the shareholders’ (Hill & Phan, 1991; Cordeiro & Veliyath, 2003; Ozkan, 2011). The experience of a CEO may be measured by his/her tenure and age. As suggested by Finkelstein and Hambrick (1990), the tenure of an executive can also affect and proxy for his/her attitudes to risk. This is because long-tenured executives have established high firm-specific human capital and become less mobile (Finkelstein & Hambrick, 1990). Consequently, they will be unwilling to take on any unnecessary risks that are likely to bring more harms than benefits. Hill and Phan (1991) further argue that the positive relationship between pay and firm risk will be stronger the longer the tenure of the CEO. Hence, CEO experience, measured by CEO tenure and CEO age, is expected to be positively associated with CEO compensation. H1a: CEO tenure will be positively related to CEO compensation. H1b: CEO age will be positively related to CEO compensation. CEO Shareholdings The level of CEO shareholdings shows the extent to which the wealth of the CEO is linked with firm value and is related to the extent of agency problems faced by companies (Ozkan, 2007). CEOs with greater shareholdings in the firm have stronger incentives to boost the firm’s stock value. Consequently, less incentive compensation is needed for aligning the interests of CEO and shareholders. That is, CEO shareholdings can act as a substitute for CEO compensation (Cordeiro & Veliyath, 2003) and a negative relationship is expected between CEO compensation and CEO shareholdings. The study by Allen (1981) provides evidence supporting this view in that the level of CEO compensation is negatively related with the equity held by the CEO. Therefore, a negative relationship between CEO shareholdings and CEO compensation is proposed.

H2: CEO shareholdings will be negatively related to CEO compensation. Board Size The size of the board affects the effectiveness of the board in monitoring management. For example, when the board size grows large, more resource networks and independent and professional views can be brought to board. However, these advantages may be overwhelmed by the efficiency losses in


31

communication, decision-making and coordination between board members as the number of board members increases. In other words, a large board may in effect reduce the effectiveness of board monitoring and therefore be associated with higher CEO compensation. Based on a sample of 414 UK companies between 2003 and 2004, Ozkan (2007) finds supports for this view and reports that firms with larger board size are associated with higher CEO compensation, measured by total compensation and cash compensation. Therefore, this study expects a positive relationship between board size and CEO compensation. H3: Board size will be positively related to CEO compensation. DATA AND METHOD Data and Sample Selection The data used in this study are obtained from the Standard and Poor’s ExecuComp database. ExecuComp provides key financial information and compensation data of the top five executives and directors for each firm in the S&P 500, S&P Midcap 400, and S&P SmallCap 600. To be included in the sample, the sample firms must have all the required financial information. As the information on board of directors in the ExecuComp database is more complete from the year 2006 and onwards, the sample period for this study is between 2007 and 2010. The final sample includes 903 firms (or 3,612 firm-years). Model Specification and Variable Definitions The hypotheses are tested based on the following model using panel data estimation method as shown in Equation (1). The advantage of using panel data estimation is that it allows us to exploit time series variation in executive compensation, firm performance and other variables while controlling for unobserved time-invariant firm-specific effects. Therefore, the potential bias due to omitted variables can be eliminated (Ozkan, 2011).

tttititi

tititititi

IndustryYeareperformanc Firm)size Log(Firm)sizeLog(Board ngsshareholdi CEOage CEO tenureCEO)oncompensati Log(CEO

1,6,5,4

,3,2,1,,

+++++

+++=

−ββββββa

(1)

where CEO compensation is measured in two ways, CEO total compensation and CEO cash compensation. Ozkan (2011) suggests that firm performance may affect cash and equity-based components of compensation differently. Thus, it is important to incorporate multiple measures for compensation. The first measure, CEO total compensation, comprises salary, bonus, other annual payment, restricted stock grants, long-term incentive payouts, value of options granted and all other payments provided by ExecuComp database. The second measure, CEO cash compensation, consists of salary and bonus. Note that all compensation variables are log transformed so that extreme values would not drive the results. CEO tenure is measured by the number of years the CEO has held the position in a given company. CEO age is the age of the CEO. CEO shareholdings is calculated as shares owned by the CEO, excluding options that are exercisable or will become exercisable within 60 days, divided by the number of common shares outstanding. Board size is measured by the number of directors on the board. Firm size is measured by total assets. This controls for the fact that larger firms, which are typically more complex, will require directors to spend more time and put more effort in monitoring managers. In other words, larger firms are associated with greater complexity and information processing demands and therefore, CEOs of larger firms are expected to receive higher compensation (Smith & Watts, 1992; Core


32

et al. 2003). The study by Conyon (1997) reports a significantly positive relationship between firm size and CEO compensation levels. Hence, a positive relationship is expected between CEO compensation and firm size. Firm performance is measured by the return on average equity (ROE), which are lagged one year to reduce potential endogeneity; that is, to avoid measuring the effect of compensation on firm performance. The lagged performance measure can also account for the fact that CEO compensation paid in one year is usually determined by the firm performance in the previous year. Agency theory suggests that one way to align the interests of managers with that of shareholders is to tie compensation contracts to firm performance (Firth et al., 2006; Chhaochharia & Grinstein, 2009); that is, to create a pay-for-performance linkage. Thus, making the CEOs hold accountable for firm performance is essential for motivating the CEOs to initiate strategies that boost firm value. Hence, a positive relationship between CEO compensation and firm performance is expected. Year dummies are included to control for unobserved differences between years. The inclusion of these dummies can capture common factors that are driven by industry- and economy-wide effects. Industry dummies are based on the SIC division structure. As the pay of CEOs is likely to be set with reference to the pay of other CEOs in an industry, this variable controls for inter-industry differences in compensation levels. For example, Hilburn (2010) reports that directors of technology companies have higher pay than their counterparts at general industry companies. Moreover, previous literature has suggested that banks are likely to face greater potential conflicts of interests than industrial firms due to its distinct characteristics such as the existence of deposit insurance, high debt-to-equity ratios and asset-liability issues (Becher, Campbell II, & Frye, 2005). Therefore, the differences in industry structure, complexity and industry custom are likely to affect the level of compensation (Hempel & Fay, 1994). Sample Characteristics Table 1 presents the descriptive statistics for the full sample of 903 firms between 2007 and 2010. The average and median age of CEOs is both 55, ranging from 34 to 80. The mean CEO shareholdings is 1.51% and ranges from 0 to 75.8% of outstanding shares. CEO tenure, which measures the number of years the CEO has held the position in a given company, has an average of 7.3 years and ranges from 0 to 47 years. The mean (median) value of CEO cash compensation is $1.1 million ($876,000). The CEO total compensation has an average of $5.7 million, ranging from $30,000 to $128 million. The average board size is 9 and ranges from 3 to 32 directors. The average firm size is $17.9 billion if measured by total assets and $7.3 billion if measured by sales. The average ROA is 3.48% and ranges from -163% to 53%. The average ROE is 9.26% and ranges from -906% to 524%. The mean and median “average director compensation per board” is $175,000 and $161,000. The mean “total board compensation” is $1,574,000. Table 2 reports the correlations between variables. As expected, there is a high correlation between board size and total board compensation (0.503), and between CEO cash compensation and CEO total compensation (0.611).


33

Table 1: Descriptive Statistics of the Sample between 2007 and 2010 Mean Median Max Min SD CEO characteristics

CEO age 55 55 84 34 7 CEO shareholdings (%) 1.51 0.31 75.80 0.00 4.63 CEO tenure (years) 7.33 6.00 47.00 0.00 6.68

CEO compensation CEO cash comp ($000) 1,128 876 77,926 7 2,259 CEO total comp ($000) 5,731 3,915 128,706 30 6,636

Firm characteristics Board size 9 9 32 3 3 Total assets ($m) 17,936 3,072 2,175,052 10 88,762 Sales ($m) 7,334 1,872 425,071 0 23,660 ROA (%) 3.48 4.04 52.85 -163.38 10.86 ROE (%) 9.26 11.66 524.38 -906.03 33.54

Director compensation DIRCOMP_AVE ($000) 175 161 1,796 4 110 DIRCOMP_MAX ($000) 303 216 7,779 14 435 DIRCOMP_TOT ($000) 1,574 1,361 14,686 33 1,086

This table provides the descriptive statistics of the sample between 2007 and 2010. The sample includes 903 firms (or 3,612 firm-years). CEO shareholdings are calculated as shares owned by the CEO divided by the total number of common shares outstanding. CEO tenure is the number of years the CEO has held the position in a given company. CEO cash compensation includes salary and bonus. CEO total compensation includes salary, bonus, other annual, total value of restricted stock granted, total value of stock options granted (using Black-Scholes), long-term incentive payouts, and all other compensation. DIRCOMP_AVE is the average director compensation for each firm (or each board), that is, the per capita compensation of directors. DIRCOMP_MAX is the compensation of the highest paid director on each board. DIRCOMP_TOT is the total board compensation. Director compensation includes cash fees, stock awards, option awards, non-equity incentive plan, change in pension value and non-qualified deferred compensation earnings, and other compensation. Table 2: Correlation Matrix 1 2 3 4 5 6 7 8 9 10 CEO

age CEO holdings

CEO tenure Board

size Assets Sales ROE ROA CEOCOMP _CASH

CEOCOMP _TOT

1. CEO age 1 2. CEO

holdings 0.118 *** 1

3. CEO tenure 0.417 *** 0.377 *** 1 4. Board size 0.048 *** -0.221 *** -0.195 *** 1 5. Assets 0.013 -0.048 *** -0.056 *** 0.238 *** 1 6. Sales 0.038 ** -0.069 *** -0.075 *** 0.254 *** 0.417 *** 1 7. ROE -0.004 0.002 0.003 0.040 ** 0.018 0.081 *** 1 8. ROA -0.010 0.021 0.014 0.008 -0.011 0.070 *** 0.666 *** 1 9. CEOCOMP

_CASH 0.087 *** -0.019 0.051 *** 0.094 *** 0.130 *** 0.126 *** 0.020 0.008 1

10. CEOCOMP_TOT 0.089 *** -0.104 *** -0.015 0.284 *** 0.223 *** 0.330 *** 0.114 *** 0.091 *** 0.611 *** 1

This table reports the correlations of variables for a sample of 903 firms (or 3,612 firm-years) between 2007 and 2010. ***, ** and * indicate the significant level at the 1%, 5% and 10%, respectively. EMPIRICAL RESULTS Table 3 presents the regression results for the full sample on CEO compensation estimated using random effects model of the panel data estimation. The results demonstrate that CEO shareholdings are significantly negatively associated with CEO total compensation at the 1% level and the evidence is slightly weaker for CEO cash compensation, which is significant at the 10% level. The results provide support for the hypothesis that CEO shareholdings and CEO compensation contracts are substitute mechanisms for aligning the interests of CEO and shareholders (Cordeiro & Veliyath, 2003). In addition,


34

CEO age is significantly positively related to CEO cash compensation at the 1% level. Therefore, the results suggest that more experienced CEOs tend to receive higher cash-based compensation. Table 3: Regression Analysis of CEO Compensation Dependent Variable Log(CEO Total Compensation ) Log(CEO Cash Compensation) Intercept 4.160 *** 3.666 *** (6.535) (8.183) CEO TENURE 0.001 0.002 (0.384) (1.001) LOG(CEO AGE) 0.103 0.484 *** (0.823) (5.654) CEO HOLDING -0.009 *** -0.003 * (-3.281) (-1.679) LOG(BSIZE) 0.025 0.050 (0.455) (1.353) LOG(ASSETS) 0.409 *** 0.197 *** (30.603) (20.635) ROE t-1 0.000 0.000 (0.178) (-0.193) Industry and year dummies Yes Yes Adjusted R2 0.505 0.168 This table presents the regression analysis of CEO compensation for 903 firms (or 3,612 firm-years) between 2007 and 2010. The model is estimated using random effects model of the panel data estimation. The dependent variables are CEO total compensation and CEO cash compensation. CEO TENURE is measured by the number of years the CEO has held the position in a given company. CEO AGE is the age of the CEO. CEO HOLDING is calculated as shares owned by the CEO divided by the number of common shares outstanding. BSIZE is measured by the number of directors on the board. ROE is lagged one year. The t-statistics are presented in parentheses. ***, ** and * indicate coefficient is significant at the 1%, 5% and 10% level, respectively. Consistent with the expectation, firm size, measured by total assets, are significantly positively related to CEO compensation at the 1% level. Since larger firms are typically more complex, CEOs of these firms are more highly compensated. Interestingly, we find no evidence that firm performance, measured by ROE and ROA (not reported), is related to both measures of CEO compensation. This suggests that the pay-performance link does not exist and highlights the current concerns that pay to CEOs does not depend on firm performance. Since there have been harsh criticisms about fat cat CEOs, we also analyze “fat cat companies”, defined as firms that have low (or below median) performance but give high (or above median) compensation to CEOs. Accordingly, firms are categorized according to two factors, firm performance (measured by ROA) and CEO total compensation. Panel A of Table 4 shows that based on the two-way tabulation, there are 803 fat cat companies and 803 underpaid companies, defined as firms that have high (or above median) performance but give low (or below median) compensation to CEOs, during the sample period 2007-2010. Panel B of Table 4 shows that over the sample period, the number of fat cat companies increases at an increasing rate while the number of underpaid companies decreases at a decreasing rate. The results suggest the problem of “fat cat” highlighted in recent years has become worse over time and should receive more public attention.


35

Table 4: Fat Cat and Underpaid Companies Panel A: Identification of Fat Cat and Underpaid Companies Count ROA (No. of firm-years) Below median Above median Total Below median 1003 803 1806 CEO total compensation Above median 803 1003 1806 Total 1806 1806 3612 Panel B: Number of Fat Cat and Underpaid Companies Each Year Count (No. of firm-years) 2007 2008 2009 2010 Total No. fat cat companies 158 166 205 274 803

Annual % change 5.1% 23.5% 33.7% No. of underpaid companies 248 241 191 123 803

Annual % change -2.8% -20.7% -35.6% Panel A identifies fat cat and underpaid companies based on two dimensions, CEO total compensation and ROA. CEO total compensation includes salary, bonus, other annual, total value of restricted stock granted, total value of stock options granted (using Black-Scholes), long-term incentive payouts, and all other compensation. Panel B shows the number of fat cat and underpaid companies during the sample period. Fat cat companies are defined as having low performance (or below median ROA) but giving high (or above median) compensation to CEOs. Underpaid companies are defined as firms that have high performance (or above median ROA) but give low (or below median) compensation to their CEOs. Table 5 summarizes the number and percentage of fat cat companies and underpaid companies in each industry. If we do not consider industries that have small sample size, the Table shows that fat cat companies are concentrated in industry 5 (transportation, communications, electric, gas, and sanitary services) and industry 10 (finance, insurance, and real estate). On the contrary, the underpaid companies are concentrated in industry 7 (retail trade) and industry 8 (services). The fact that fat cat companies are concentrated in financial services industries coincides with the news critics that the fat cat problem is pervasive in financial companies. Table 5: Number and Percentage of Fat Cat and Underpaid Companies in Each Industry Industry High Comp % Low Comp % High Comp % Low Comp % Total

Low ROA High ROA High ROA Low ROA (firm- years) (fat cat) (underpaid)

1 4 50.0 0 0.0 4 50.0 0 0.0 8 2 49 23.1 47 22.2 88 41.5 28 13.2 212 3 6 13.6 15 34.1 10 22.7 13 29.5 44 4 204 15.2 332 24.8 511 38.1 293 21.9 1340 5 145 33.3 76 17.4 71 16.3 144 33.0 436 6 18 23.7 33 43.4 14 18.4 11 14.5 76 7 41 13.1 104 33.3 114 36.5 53 17.0 312 8 60 14.4 134 32.2 122 29.3 100 24.0 416 9 4 50.0 0 0.0 4 50.0 0 0.0 8

10 272 35.8 62 8.2 65 8.6 361 47.5 760 Total 803 22.2 803 22.2 1003 27.8 1003 27.8 3612

This table shows the number and percentage of fat cat and underpaid companies. Based on the SIC system, the industry classification in this study is defined as follows: (1) agriculture, forestry, and fishing; (2) mining; (3) construction; (4) manufacturing; (5) transportation, communication, electric, gas and sanitary services; (6) wholesale trade; (7) retail trade; (8) services; (9) public administration; and (10) finance, insurance and real estate.


36

Based on the two-sample t-test, Table 6 shows that compared with the rest of sample firms, CEOs of fat cat companies are less experienced with shorter tenure and have significantly lower shareholdings in the firms. These characteristics fit the descriptions of a fat cat CEO. In addition, fat cat companies have more directors on the boards and are larger in terms of firm size. Director compensation is also significantly higher for fat cat companies than the rest of firms. Table 6: Descriptive Statistics of Fat Cat Companies and Two-Sample T-Test Mean Comparison with the Rest of Sample Firms

Fat Cat Companies

(803 Firm-years) Rest of Sample

(2,809 Firm-Years) Mean Median Max Min SD Mean t-test CEO characteristics

CEO age 56 56 84 37 6 55 1.51 CEO holdings (%) 0.78 0.20 28.00 0.00 2.10 1.72 -5.11 *** CEO tenure (years) 6.88 5.00 46.00 0.00 6.14 7.46 -2.20 **

CEO compensation CEO cash comp ($000) 1,643 1,050 77,926 277 3,345 981 7.37 *** CEO total comp ($000) 9,085 7,012 112,465 3,917 7,573 4,772 16.87 ***

Firm characteristics Board size 10 10 32 4 3 9 14.04 *** Total assets ($m) 50,391 9,936 2,175,052 202 165,915 8,658 11.98 *** Sales ($m) 10,368 4,439 180,929 97 19,620 6,467 4.13 *** ROA (%) -1.14 1.47 4.03 -110.44 9.22 4.80 -14.03 *** ROE (%) -0.80 6.18 320.14 -906.03 43.48 12.13 -9.76 ***

Director compensation DIRCOMP_AVE ($000) 207 187 1,028 25 105 166 9.50 *** DIRCOMP_MAX ($000) 390 250 6,431 52 608 278 6.53 *** DIRCOMP_TOT ($000) 2,062 1,823 9,255 252 1,152 1,434 14.89 ***

This table presents the descriptive statistics of fat cat companies (or 803 firm-years), defined as firms that have poor performance but give high compensation to their CEOs, and the two-sample t-test of mean comparison between fat cat companies and the rest of sample firms. ***, ** and * indicate the significant level at the 1%, 5% and 10%, respectively. Table 7 then compares the characteristics of underpaid companies with the rest of sample firms using the two-sample t-test. The Table shows that underpaid CEOs are younger, have longer tenure and hold more shares in the companies than other CEOs. In addition, underpaid companies have smaller board size and firm size, and give lower pay to board of directors. In other words, the results suggest that underpaid companies are typically smaller firms and are more likely to expropriate younger CEOs who have worked in the companies for a longer period of time and have greater interest in the companies. Furthermore, to examine the determinants of fat cat CEOs’ compensation, panel estimation with the random effects model is used. Table 8 shows that CEO tenure, a measure of CEO experience, and firm size are significantly positively related with both measures of CEO compensation, consistent with the expectation. Thus, the results suggest that more experienced CEOs with longer tenure are more likely to receive higher pay and that larger firms give higher pay to their CEOs. While the corporate governance of manufacturing firms has received heightened attention, the corporate governance of financial companies has been less researched (John & Qian, 2003). According to John and Qian (2003), financial companies differ from manufacturing firms in two aspects. Firstly, financial companies are more highly regulated than manufacturing firms. Secondly, banks have much higher leverage than manufacturing firms. In addition, the remuneration policies of financial institutions have been blamed to contribute to the global financial crisis between 2007 and 2008 (Gregg, Jewell, & Tonks, 2011). Therefore, we conduct further analysis of fat cat companies that belong to financial services industry (i.e., industry 10: finance, insurance, and real estate).


37

Table 7: Descriptive Statistics of Underpaid Companies and Two-Sample T-Test Mean Comparison with the Rest of Sample Firms

Underpaid Companies

(803 Firm-Years) Rest of Sample

(2,809 Firm-Years) Mean Median Max Min SD Mean t-test CEO characteristics

CEO age 54 54 79 34 7 56 -6.22 *** CEO holdings (%) 2.67 0.51 61.39 0.00 6.85 1.18 8.08 *** CEO tenure (years) 7.85 6.00 45.00 0.00 7.41 7.19 2.48 ***

CEO compensation CEO cash comp ($000) 717 664 2,583 26 307 1,246 -5.88 *** CEO total comp ($000) 2,071 1,997 3,913 30 932 6,777 -18.55 ***

Firm characteristics Board size 7 7 28 3 2 9 -18.45 *** Total assets ($m) 1,659 836 27,397 57 2,390 22,589 -5.92 *** Sales ($m) 1,748 918 50,703 22 3,199 8,931 -7.65 *** ROA (%) 9.43 8.02 52.85 4.05 5.24 1.78 18.41 *** ROE (%) 19.69 15.71 220.85 4.99 16.13 6.27 10.13 ***

Director compensation DIRCOMP_AVE ($000) 141 130 795 11 79 184 -10.08 *** DIRCOMP_MAX ($000) 231 171 4,615 14 279 323 -5.30 *** DIRCOMP_TOT ($000) 1,048 989 5,563 33 638 1,724 -16.12 ***

This table presents the descriptive statistics for underpaid companies (or 803 firm-years), defined as firms that have high performance but give low compensation to their CEOs, and the two-sample t-test of mean comparison between underpaid companies and the rest of sample firms. ***, ** and * indicate the significant level at the 1%, 5% and 10%, respectively. Table 8: Regression Analysis of CEO Compensation for Fat Cat Companies Dependent Variable Log(CEO Total Compensation ) Log(CEO Cash Compensation) Intercept 6.793 *** 6.081 *** (8.328) (6.024) CEO TENURE 0.009 *** 0.007 ** (2.679) (1.796) LOG(CEO AGE) -0.130 0.047 (-0.708) (0.211) CEO HOLDING -0.002 0.013 (-0.198) (1.206) LOG(BSIZE) 0.044 0.080 (0.533) (0.801) LOG(ASSETS) 0.198 *** 0.151 *** (10.646) (6.519) ROE t-1 0.000 0.000 (-0.556) (-0.694) Industry and year dummies Yes Yes Adjusted R2 0.194 0.093 This table presents the regression analysis of CEO compensation for 373 fat cat companies (or 803 firm-years) between 2007 and 2010, using panel estimation with the random effects model. The dependent variables are CEO total compensation and CEO cash compensation. CEO TENURE is measured by the number of years the CEO has held the position in a given company. CEO AGE is the age of the CEO. CEO HOLDING is calculated as shares owned by the CEO divided by the number of common shares outstanding. BSIZE is measured by the number of directors on the board. ROE is lagged one year. The t-statistics are presented in parentheses. ***, ** and * indicate coefficient is significant at the 1%, 5% and 10% level, respectively.


38

As the nature of financial services industry is different from that of other industries, such as the existence of deposit insurance, high debt-to-equity ratios and asset-liability issues (Becher et al., 2005), additional variables, including financial leverage and investment opportunities, are included as control variables in the following analysis of fat cat financial companies as shown in Equation (2). According to Fahlenbrach and Stulz (2011), one explanation of the financial crisis is that financial companies have excessive leverage. As for levered firms, shares are effectively options on the value of the assets, CEOs of financial companies attempt to increase the volatility of the assets and thereby, the value of their shares. Hence, the leverage ratio, defined as one minus the ratio of equity over assets, is added to the model as a control variable. Moreover, Finkelstein and Boyd (1998) suggest that the executive compensation is influenced by contingencies such as industry concentration and investment opportunities. Accordingly, we also include investment opportunities, measured by the firm’s market value divided by the book value of shareholders’ equity, as a control variable.

tti

titititi

tititititi

YearMB Leverageeperformanc Firm)size Log(Firm)size Log(Boardngsshareholdi CEOage CEO tenureCEO)oncompensati Log(CEO

,8

,71,6,5,4

,3,2,1,,

+

+++++

+++=

−

β

ββββ

βββa

(2)

where CEO compensation is measured by CEO total compensation and CEO cash compensation. CEO tenure is measured by the number of years the CEO has held the position in a given company. CEO age is the age of the CEO. CEO shareholdings is calculated as shares owned by the CEO, excluding options that are exercisable or will become exercisable within 60 days, divided by the number of common shares outstanding. Board size is measured by the number of directors on the board. Firm size is measured by total assets. Firm performance is measured by lagged return on average equity (ROE). Leverage is measured by one minus the ratio of equity over assets. MB is the investment opportunities, measured by the firm’s market value divided by the book value of shareholders’ equity. Year dummies are included to control for unobserved differences between years. Table 9 presents that panel estimation of CEO compensation for fat cat financial companies. The results show that fat cat financial companies with lower leverage ratio and higher investment opportunities give higher pay to CEOs, in terms of total compensation. The significant negative relationship between financial leverage and CEO total compensation is contrary to what is suggested by public critics that CEOs of financial companies have the incentive to take on excessive leverage to increase their compensation. In addition, firm performance, measured by lagged ROE, is found to be negatively associated with CEO cash compensation, significant at the 5% level. The results thus suggest that there is a strong positive relationship between the non-cash (or equity-based) compensation paid to fat cat CEOs in financial companies and firm performance. CONCLUSION The global financial crisis in 2008 sheds light on the significance of reviewing the compensation packages of top executive. Based on a sample of 903 US firms between 2007 and 2010, this study examines the determinants of CEO compensation and conducts further tests on a subsample of “fat cat companies”, which are defined as having poor performance while giving high compensation to their CEOs. The results show that CEOs with older age are associated with higher cash compensation. This finding provides support for the argument that CEO age is related to CEO’s ability to influence the board’s pay determination process. The results also suggest that there is a substitution effect between CEO compensation and the level of CEO ownership and that larger firms give higher pay to their CEOs.


39

Table 9: Regression Analysis of CEO Compensation for Fat Cat companies in Finance, Insurance and Real Estate Industries Dependent Variable Log(CEO Total Compensation) Log(CEO Cash Compensation) Intercept 8.421 *** 8.120 *** 7.827 *** 7.885 *** (5.754) (5.687) (4.178) (4.126) CEO TENURE 0.008 0.009 0.011 0.012 (1.230) (1.457) (1.432) (1.492) LOG(CEO AGE) -0.294 -0.255 -0.448 -0.481 (-0.793) (-0.695) (-0.944) (-0.980) CEO HOLDING 0.004 0.003 0.004 0.004 (0.250) (0.208) (0.190) (0.210) LOG(BSIZE) 0.009 -0.108 -0.010 -0.013 (0.069) (-0.808) (-0.060) (-0.074) LOG(ASSETS) 0.167 *** 0.263 *** 0.111 *** 0.140 *** (5.415) (6.546) (2.754) (2.578) ROE t-1 0.002 0.001 -0.004 ** -0.004 ** (0.999) (0.722) (-2.056) (-2.099) Leverage -0.919 *** -0.346 (-2.813) (-0.809) MB 0.076 *** 0.027 (2.552) (0.692) Year dummies Yes Yes Yes Yes Adjusted R2 0.172 0.214 0.059 0.056 This table presents the regression analysis of CEO compensation for 101 fat cat companies (or 272 firm-years) between 2007 and 2010 in finance, insurance and real estate industries, using panel estimation with the random effects model. The dependent variables are CEO total compensation and CEO cash compensation. CEO TENURE is measured by the number of years the CEO has held the position in a given company. CEO AGE is the age of the CEO. CEO HOLDING is calculated as shares owned by the CEO divided by the number of common shares outstanding. BSIZE is measured by the number of directors on the board. ROE is lagged one year. LEVERAGE is defined as one minus the ratio of equity over assets. MB is measured by the firm’s market value divided by the book value of shareholders’ equity. The t-statistics are presented in parentheses. ***, ** and * indicate coefficient is significant at the 1%, 5% and 10% level, respectively. Note that Hausman test has been conducted and the null hypothesis cannot be rejected. Moreover, this study finds that over the sample period, the number of fat cat companies, defined as having low firm performance while giving high compensation to CEOs, increases at an increasing rate and is concentrated in industry 5 (transportation, communications, electric, gas, and sanitary services) and industry 10 (finance, insurance, and real estate). In addition, fat cat CEOs are characterized by shorter tenure and lower shareholdings. The panel analysis shows that tenure and firm size are significantly positively associated with the compensation of fat cat CEOs. Furthermore, firm size, leverage ratio and market-to-book ratio are significantly associated with the total compensation of fat CEOs in the finance, insurance and real estate industries. Overall, we find that firm size is the most important determinant of CEO compensation, which is consistently significant throughout the analyses, and that there is a general lack of linkage between pay and performance even though the number of fat cat CEOs is increasing over the years. These results thus call for public attention that there is a strong need for reexamining the pay setting process and the effectiveness of current pay system. One limitation of this study is that due to the constraint on the availability of board of directors’ data, the sample period of this study is limited to four years only. Future research could extend the sample period by dropping the board size variable to see if similar results can be reached. REFERENCES Abou-El-Fotouh, H. (2010) “Cosmetic Corporate Governance: Will Companies Learn from the Global Financial Crisis?” Capital Business Magazine, Retrieved June 9, 2010 from: www.capital-me.com


40

Allen, M. (1981) "Power and Privilege in the Large Corporation: Corporate Control and Managerial Compensation," American Journal of Sociology, vol. 86, p. 1112-1123 Alonso, P. d. A. & Aperte, L. A. (2011) “CEO Compensation and Board Composition as Alternative Instruments of Corporate Governance,” Working Paper, University of Valladolid Barkema, H. G. & Gomez-Mejia, L. R. (1998) “Managerial Compensation and Firm Performance: A General Research Framework,” Academy of Management Journal, vol. 41(2), p. 135-145 Bebchuk, L. A., Fried, J. & Walker, D. (2002) “Managerial Power and Rent Extraction in the Design of Executive Compensation,” University of Chicago Law Review, vol. 69, p. 751-846 Bebchuk, L. A. & Fried, J. M. (2004) Pay without Performance: The Unfulfilled Promise of Executive Compensation. Cambridge, MA: Harvard Univ. Press Becher, D. A., Campbell II, T. L. & Frye, M. B. (2005) “Incentive Compensation for Bank Directors: The Impact of Deregulation,” Journal of Business, vol. 78, p. 1753-1777 Blinder, A. S. (2009, 28 May 2009) “Crazy Compensation and the Crisis,” Wall Street Journal, A.15 Boyd, B. K. (1994) “Board Control and CEO Compensation,” Strategic Management Journal, vol. 15(5), p. 335-344. Brick, I. E., Palmon, O. & Wald, J. K. (2006) “CEO Compensation, Director Compensation, and Firm Performance: Evidence of Cronyism?” Journal of Corporate Finance, vol. 12, p. 403-423 Chhaochharia, V. & Grinstein, Y. (2009) “CEO Compensation and Board Structure,” Journal of Finance, vol. 64(1), p. 231-261 Conyon, M. J. (1997) “Corporate Governance and Executive Compensation,” International Journal of Industrial Organization, vol. 15, p. 493-509 Cordeiro, J. J. & Veliyath, R. (2003) “Beyond Pay for Performance: A Panel Study of the Determinants of CEO Compensation,” American Business Review, vol. 21(1), p. 56-66 Core, J. E., Guay, W. R. & Larcker, D. F. (2003) “Executive Equity Compensation and Incentives: A Survey,” Federal Reserve Bank of New York Economic Policy Review, vol. 9, p. 27-50 Core, J. E., Holthausen, R. W. & Larcker, D. F. (1999) “Corporate Governance, Chief Executive Officer Compensation, and Firm Performance,” Journal of Financial Economics, vol. 51, p. 371-406 Cosh, A. & Hughes, A. (1997) “Executive Remuneration, Executive Dismissal and Institutional Shareholdings,” International Journal of Industrial Organization, vol. 15, p. 469-492 Dong, M. & Ozkan, A. (2008) “Institutional Investors and Director Pay: An Empirical Study of UK Companies,” Journal of Multinational Financial Management, vol. 18(1), p. 16-29 Fahlenbrach, R. & Stulz, R. M. (2011) “Bank CEO Incentives and the Credit Crisis,” Journal of Financial Economics, vol. 99, p. 11-26


41

Fama, E. F. & Jensen, M. C. (1983) “Separation of Ownership and Control,” Journal of Law and Economics, vol. 26, p. 301-325 Finkelstein, S. & Boyd, B. (1998) “How much does the CEO Matter? The Role of Managerial Discretion in the Setting of CEO Compensation,” Academy of Management Journal, vol. 41, p. 179-199 Finkelstein, S. & Hambrick, D. C. (1988) “Chief Executive Compensation: A Synthesis and Reconciliation,” Strategic Management Journal, vol. 9(6), p. 543-558 Finkelstein, S. & Hambrick, D. C. (1990) “Top-Management-Team Tenure and Organizational Outcomes: The Moderating Role of Managerial Discretion,” Administrative Science Quarterly, vol. 35(3), p. 484-503 Firth, M., Fung, P. M. Y. & Rui, O. M. (2006) “Corporate Performance and CEO Compensation in China,” Journal of Corporate Finance, vol. 12(4), p. 693-714 Frydman, C. & Jenter, D. (2010) “CEO Compensation,” Working Paper, Stanford University, Retrieved from: www.ssrn.com/abstract=1582232 Gregg, P., Jewell, S. & Tonks, I. (2005) “Executive Pay and Performance in the UK 1994-2002,” Working Paper, University of Bristol and University of Exeter Gregg, P., Jewell, S. & Tonks, I. (2011) “Executive Pay and Performance: Did Bankers' Bonuses Cause the Crisis?” Working Paper, Retrieved from: www.ssrn.com/abstract=1815210 Guest, P. (2010) “Board Structure and Executive Pay: Evidence from the UK,” Cambridge Journal of Economics, vol. 34(6), p. 1075-1097 Guthrie, K., Sokolowsky, J. & Wan, K.-M. “CEO Compensation and Board Structure Revisited” Journal of Finance, In Press Hempel, P. & Fay, C. (1994) “Outside Director Compensation and Firm Performance,” Human Resource Management, vol. 33(1), p. 111-133 Hilburn, W. (2010, 19 October 2010) “Trends in Director Compensation,” Bloomberg BusinessWeek Hill, C. W. L. & Phan, P. (1991) “CEO Tenure as a Determinant of CEO Pay,” Academy of Management Journal, vol. 34(3), p. 707-717 Holmstrom, B. (1979) “Moral Hazard and Observability,” Bell Journal of Economics, vol. 10, p. 74-91 Hubbard, R. G. & Palia, D. (1995) “Executive Pay and Performance: Evidence from the U.S. Banking Industry,” Journal of Financial Economics, vol. 39, p. 105-130 Jensen, M. C. & Meckling, W. H. (1976) “Theory of the Firm: Managerial Behavior, Agency Costs, and Ownership Structure,” Journal of Financial Economics, vol. 3(4), p. 305-360 Jensen, M. C. & Murphy, K. J. (1990) “Performance Pay and Top-Management Incentives,” Journal of Political Economy,” vol. 98(2), p. 225-264 John, K. & Qian, Y. (2003) “Incentive Features in CEO Compensation in the Banking Industry,” Federal Reserve Bank of New York Economic Policy Review, vol. 9(1), p. 109-121


42

Main, B. G. M., Bruce, A. & Buck, T. (1996) “Total Board Remuneration and Company Performance,” Economic Journal, vol. 106(439), p. 1627-1644 Murphy, K. J. (1985) “Corporate Performance and Managerial Remuneration: An Empirical Analysis,” Journal of Accounting and Economics, vol. 7(1-3), p. 11-42 Ozkan, N. (2007) “Do Corporate Governance Mechanisms Influence CEO Compensation? An Empirical Investigation of UK Companies,” Journal of Multinational Financial Management, vol. 17(5), p. 349-364 Ozkan, N. (2011) “CEO Compensation and Firm Performance: An Empirical Investigation of UK Panel Data,” European Financial Management, vol. 17(2), p. 260-285 Smith, C. & Watts, R. (1992) “The Investment Opportunity Set and Corporate Financing, Dividend, and Compensation Policies,” Journal of Financial Economics, vol. 32, p. 263-292 ACKNOWLEDGEMENT This work was fully supported by the National Science Council in Taiwan, under contract no.: NSC 100-2410-H-147-007. BIOGRAPHY Dan Lin is an assistant professor at the Takming University of Science and Technology, and can be reached at [email protected]. Hsien-Chang Kuo is a professor at the Takming University of Science and Technology, and can be reached at [email protected]. Lie-Huey Wang, the corresponding author of this paper, is an associate professor at Ming Chuan University, and can be reached at [email protected].


43

BANK CREDIT LINES AND OVERINVESTMENT: EVIDENCE FROM CHINA

Qianwei Ying, Si Chuan University, China Danglun Luo, Sun Yat-Sen University, China

Lifan Wu, California State University, Los Angeles USA

ABSTRACT

The paper investigates the relationship between bank credit lines and firms’ overinvestment for Chinese listed companies from 2001 to 2008. We find significant impacts of bank credit lines on firm overinvestment activities. Further, we find that overinvestment is mainly made by State-owned firms, and not privately-owned firms. State-owned firms have easier access to bank credit lines with cheaper cost than private-owned firms, and therefore are more likely to overinvest. The results suggest that concentration of credit lines among State-owned firms likely leads to low resource allocation efficiency. JEL: G21, G31, G38 KEYWORDS: Credit lines; Overinvestment; State-owned firms INTRODUCTION

ank credit lines have become a major source of funding for firms. Kashyap et al. (1993) reported that credit lines account for about 70% of U.S. small firms' financing. Sufi (2009) found that about 80% of bank lending to U.S. public firms is through credit lines. Jimenez et al. (2009) showed that

bank credit lines account for 42% of Spanish firms' bank financing. Similarly, credit lines have recently become increasingly popular in China corporate finance. About 5% of listed firms obtained credit lines in 2001. This number increased to 24% by 2009. Bank credit lines exceeded 1.5 trillion RMB for listed firms in 2009. Our sample shows firm credit lines account for 26% of total liabilities in 2009. Given the importance of bank credit lines for corporate financing, this paper empirically investigates the role of credit lines for Chinese firms. Credit lines themselves are not bank loans. But due to their funding speed and flexibility, users do not need to apply for the loan each time they need funding. Instead, they can easily and conveniently draw down unused lines of credit for their investment needs. However, this flexibility also provides opportunities to make undesirable investments. Jensen and Meckling (1976) showed that firms under financial distress may take advantage of bank lines of credit to pursue risky investments. Sufi (2009) pointed out that credit lines can make agency problems particularly severe, since banks can’t perfectly observe how lines of credit are used and can’t closely monitor borrower activities. Following these arguments, we examine how credit lines can lead to overinvestment for listed Chinese firms. In addition, we examine Chinese firms because credit lines are mostly issued to State-owned firms in China. In contrast to the U.S. private shareholding structure and the pyramidal family ownership structure in East Asia, State ownership is the dominant ownership structure in China. The government is the single largest shareholder in State-owned firms. Having both controlling rights and cash flow rights, the State not only plays a key role in corporate governance, but also appoints key executive positions and the board of directors for State-owned firms. In this environment, with less investor protection and external corporate control, conflict of interests between controlling shareholders (the State) and minority shareholders (outside investors) inevitably exists. Therefore, our paper intends to explore the functions of State ownership under these circumstances and to provide new evidence to the finance literature. The remainder of the paper is organized as follows. Section 2 provides a brief literature review. Section 3 describes data and develops testable hypotheses about the relation between lines of credit and

B

Q. Ying et al | IJBFR ♦ Vol. 7 ♦ No. 2 ♦ 2013

44

overinvestment. Section 4 presents empirical results, and concluding comments follow in Section 5. LITERATURE REVIEW Firms face trade-offs in choosing between spot loans and credit lines. When a firm takes out a credit line, it pays the setup fee and gets the loan commitment. Researchers have long studied the role of credit lines in the bank loan market, and have developed several competing arguments about why firms choose credit lines instead of spot loans.

Campbell (1978), Hawkins (1982), Melnik and Plaut (1986), and Sofianos et al. (1990) argued that credit lines serve as options which firms can employ to hedge against future uncertainty in the loan market. The credit lines give borrowers the right to borrow up to a specified amount of money, in exchange for the upfront fee, during a fixed period at a fixed rate. When a firm suffers a deterioration of creditworthiness, it may have difficulty obtaining spot loans. Having an unused line of credit would help lock in new funding. However, the main implication of option models is that optional use of credit lines is either exercised all or nothing, never left partially exercised. This prediction is contradicted by empirical data. Many firms use credit lines, yet rarely reach the limit. Thakor and Udell (1987), Maksinovic (1990), Boot et al. (1987, 1991), Berkovitch and Greenbaum (1991), Duan and Yoon (1993) and Morgan (1994) documented credit lines as optimal solutions to asymmetric information between banks and corporate clients. According to this view, some firms may have difficulty getting spot loans or borrowing because their assets are illiquid, their firm is too small or they have little track record. This adverse selection problem is mitigated by credit lines which allow borrowers to signal their quality to banks. Credit lines provide more protection to the banks than spot loans because banks may have the option to cut the unused credit line portion in the event of a change in the firms’ creditworthiness. The third view focuses on firms' investment opportunities. Avery and Berger (1991) conducted a survey and showed that flexibility and speed of action in seizing investment opportunities are main reasons for the use of credit lines. Martin and Santomero (1997) further modeled these features in the demand for credit lines and investment opportunities. Recently Sufi (2009) empirically examined firms’ profitability and flexibility of credit lines, and finds that banks only extend lines of credit to firms with high profitability and manage unused portions of lines of credit with covenants on profitability. He further suggests that lines of credit may play an instrumental role in firm investment policy and researchers may be able to examine how credit lines affect investment policy. These arguments are not mutually exclusive and they all likely to contribute to development of the bank loan market. The empirical evidence on these explanations is mixed, and most focuses on the relationship between credit lines and firms’ financing. For example, Ham and Melnik (1987), Berger and Udell (1995), Shockley and Thakor (1997), Dennis et al. (2000), Agarwal et al. (2004), Almeida et al. (2004), Faulkender and Wang (2006), and Jimenez et al. (2009) examined lines of credit as corporate liquidity and financing management. The purpose of this paper is to extend the current study of credit lines to examine the relationship between credit lines and overinvestment behavior. DATA AND METHODOLOGY Data We collect our sample and credit lines information from the bank loan data set of RSSET Database. Our sample selection is based on the following criteria: (1) Listed Chinese firms from year 2000 to 2008; (2) Excluding banks and financial institutions; (3) Excluding firms with negative net worth; and (4) Excluding firms with a bankruptcy filing. Our final data consists of 11,811 firm years. All other variables are collected from the Center of China Economic Research (CCER) Database.


45

Table 1 presents descriptive statistics of all sample firms with lines of credit from year 2001 to year 2008. As more firms obtain lines of credit each year, the proportion of lines of credit firms also increases. For example, about 5% of listed firms received lines of credit in 2001, and the number increased to more than 24%, 372 firms, 2008. During the same period, the total amount of credit lines issued to listed firms went up more than 10 times, from 31.2 billion RMB in 2001 to 333 billion RMB in 2008.

Table 1: Description of Credit Lines for Chinese Listed Firms

Year Total Firms Firms with Credit Lines

Ratio of Firms with Credit Lines to Total Firms

2001 1157 58 0.0501

2002 1215 133 0.1095

2003 1269 146 0.1151

2004 1349 146 0.1082

2005 1330 163 0.1226

2006 1383 179 0.1294

2007 1484 285 0.1920

2008 1537 372 0.2420

Table 1 shows descriptive statistics. Our sample selection is based on the following criteria: (1) Listed Chinese firms from year 2000 to year 2008; (2) Excluding banks and financial institutions; (3) Excluding firms with negative net worth; and (4) Excluding firms with bankruptcy filing. A total of 11,811 firm years are collected. In addition, lines of credit have become increasingly important in corporate financing management. Figure 1 describes the proportion of credit lines of firms relative to their total assets. It typically accounts more than 20% of total assets for most of our sample years, which exceeds the firms' cash position. Clearly lines of credit play an important role in firms’ liquidity management. Figure 1: Ratio of Credit Lines to Total Assets

The figure shows the ratio of credit lines to total assets for Chinese listed firms for the period of 2001 to 2008. The ratio of cash to total assets is also reported as the comparison.

To measure the credit line variable, we follow the Agarwal et al (2004) and Sufi (2009) methods. Specifically, we use two measurements: a dummy variable to indicate whether a firm obtains lines of credit (1= yes, and 0=no), and a ratio of lines of credit to the firm’s total assets to measure the level of credit lines.

0.100

0.120

0.140

0.160

0.180

0.200

0.220

0.240

2001 2002 2003 2004 2005 2006 2007 2008

Rat

io to

the

tota

l ass

ets

Year

CashBank Credit Lines


46

We measure overinvestment similarly to Richardson (2006). Overinvestment is defined as excess investment above a normal level. Underinvestment occurs if actual investment less than the normal level. We measure underinvestment as negative overinvestment in our analysis. We follow Richardson method (2006) to construct the overinvestment variable:

itit

itititititit

dicatorIndustryIntorYearIndicaInvtSizeAgeCashLeverageOPPInvt

eβ

βββββa

++++

+++++=

∑∑−

−−−−−

16

1514131211 (1)

where Invtit represents the investment expenditures of firm i at current year t, measured by the ratio of capital expenditures to total assets. 1itOPP − are the previous year investment opportunities. Investment opportunities are measured by Tobin’s Q (1969), Tobinit-1. Alternatively we measure investment opportunities by the growth rate of operating income, Growtht-1. 1itLeverage − is the previous year total debt ratio. 1itCash − is the previous year ratio of cash and cash equivalent to total assets. 1itAge − is age of the firm. Sizeit-1 is the firm size, measured by the logarithm of its total assets. 1itInvt − is the lagged investment. Year Indicator and Industry Indicator are dummy variables. The error terms of Equation（1）is our estimate of the overinvestment. Table 2 provides statistical description for these variables. Hypotheses and Methods Firms with credit lines have more ability to overinvest than firms without them. Since credit lines usually are cheaper than spot loans, they can act as alternative low cost loans. According to Sufi (2009), banks are likely to provide credit lines to financially healthy firms. Therefore, firms with credit lines usually have more free cash flow and liquidity than firms without. With more flexibility and lower cost of loans, firms with credit lines have more freedom to invest. Richardson (2006) pointed out that free cash flow can lead to overinvestment. We argue that lines of credit can also lead to overinvestment. Since credit lines give firms an option to freely borrow from a bank in any amount up to a specified limit at a specified price, firms are more likely use these options to explore favorable investment projects. Avery and Berger (1991) found that speed of action and pursuit of investment opportunities are primary reasons for credit line use. Duan and Yoon (1993) showed that low credit line interest rates cause firm overinvestment. Martin and Santomero (1997) further explored the relationship between investment opportunities and demand for lines of credit. Based upon these arguments, our first hypothesis is, H1: Credit lines provide incentive to overinvestment. State-owned firms are government controlled and their investment decisions are frequently influenced by government policies. Local government officials are interested in regional economic growth and GDP targets which are linked to promotion and benefits. Such performance pressure and incentives push government officials to take advantage of recourses in State-owned firms to achieve their political goals. Sometimes negative NPV investments are undertaken because of government official interference, as long as the project temporarily brings employment and growth to their area. Local branches of State-owned banks usually maintain close relationships with local government officials to gain the government support. Thus, State-owned firms can access lines of credits easily, and make more investments. Managers of State-owned firms have a stronger incentive to overinvest than private-owned firms. Their salaries are capped because of the State-owned nature. Therefore, pursuing expansion and investments is rational to obtain more implicit benefits. Based on these reasons, our second hypothesis is: H2: The State-owned firms are more likely to overinvestment than private-owned firms.


47

Table 2: Statistical Description of Variables

Variable Mean Minimum 25% Percentile

50% Percentile

75% Percentile

Maximum Standard Deviation

LOC 0.025 0.000 0.000 0.000 0.000 1.881 0.101

Excess 0.000 -0.343 -0.029 -0.011 0.016 0.662 0.054

Invt 0.064 0.000 0.015 0.043 0.089 0.749 0.067

FCF 0.021 -0.934 -0.030 0.052 0.128 0.372 0.198

HHI5 0.230 0.020 0.104 0.184 0.317 1.151 0.177

Magstk 0.015 0.000 0.000 0.000 0.000 0.502 0.073

Tobin 1.265 0.487 0.867 1.069 1.438 4.312 0.653

Growth 0.135 -1.220 -0.008 0.138 0.290 1.492 0.364

Leverage 0.481 0.079 0.349 0.486 0.616 0.922 0.186

Cash 0.159 0.004 0.074 0.129 0.213 0.562 0.118

Age 9.235 0.000 6.000 9.000 12.000 27.000 4.149

Size 21.236 19.124 20.556 21.130 21.829 24.171 0.989

The table reports summary statistics of variables. LOC is the ratio of credit lines to the total assets; Excess is the overinvestment variable using Richardson (2006)’s measurement; Invt is the investment expenditures measured by the ratio of capital expenditures to total assets; FCF is the free cash flow; HHI5 is the level of ownership concentration measured by the squared ratio of the first five largest shareholdings to the total outstanding shares; Magstk is the ratio of managers’ stock holding to total outstanding shares; Tobin is the ratio of firm's market value to its replacement costs. Growth is the growth rate of operating income; Leverage is the total debt ratio; Cash is the ratio of cash to the total assets; Age is the firm age. Size is the firm size measured by the logarithm of its total assets. To test our hypothesis, we first set up the following regression model to examine effects of credit lines on the overinvestment:

ititititititit SOEMagstkHHIFCFCreditlineExcess eβββββa ++++++= 54321 5 (2)

where, Excessit is the overinvestment variable; α is a constant term; Creditline is the lines of credit; FCFit is the free cash flow; HHI5 is the level of ownership concentration measured by the squared ratio of the first five largest shareholdings to the total outstanding shares; Magstk is the ratio of the managers’ stock holding; SOE is the ownership dummy variable (=1 if State-owned firm; and 0 otherwise); and ite is an error term. To control for factors, other than the level of credit lines, which can jointly affect firm’s overinvestment, we include in Equation (2) a set of control variables that are likely correlated with a firm’s credit lines. Richardson (2006) found that overinvestment activities are more concentrated in firms with large free cash flows. Therefore, we incorporate free cash flow as a control variable. In addition, we include corporate governance factors such as the ownership concentration, and managerial shareholding. Finally we include an ownership dummy variable to examine the overinvestment behavior of the State-owned and the private owned firms. In order to make more reliable statistical inference on Equation (2), we follow Peterson（2009’s suggestion to use robust standard errors clustered at the firm level. To investigate the impact of credit lines on overinvestment under different ownership structure, we add an ownership dummy variable jointly with bank credit lines, itit CreditlineSOE × , to Equation (2), and yield,

ititititititititit CreditlineSOESOEMagstkHHIFCFCreditlineExcess eββββββa +×++++++= 654321 5 (3)

In addition, we divide our whole sample into the State-owned firms and private firms to examine bank credit lines’ impacts on the overinvestment controlling free cash flow and other constraints. Specifically


48

we add a joint variable of the free cash flow with credit lines, FCF×Creditline, to Equation (2), which yields,

itititititititit CreditlineFCFMagstkHHIFCFCreditlineExcess eβββββa +×+++++= 54321 5 (4)

RESULTS We divide all firms into two groups, depending on whether they receive credit lines, to compare their variable characteristics. Table 3 provides the summary statistics for these two groups in first differences. All variables are statistically significant. We find that firms with bank lines of credit show significantly more overinvestment than firms without them (t-value =2.759). Similarly we find firms with lines of credit have more investment expenditures, larger size, and more free cash flow available. Table 3: T- Tests for Group Differences

With Credit Lines Without Credit Lines T-value

Excess 0.004 -0.001 2.759***

Invte 0.068 0.063 2.657***

Size 21.443 21.206 8.658***

FCF 0.053 0.016 6.590***

Leverage 0.511 0.477 6.710***

HHI5 0.245 0.227 3.674***

Magstk 0.029 0.013 7.284***

The table reports the T-test results on the differences between the group of firms with credit lines and those without credit lines. Excess is the overinvestment variable using Richardson (2006) measurement; Invt is the investment expenditures measured by the ratio of capital expenditures to total assets; FCF is the free cash flow; HHI5 is the level of ownership concentration measured by the squared ratio of the first five largest shareholdings to the total outstanding shares; Magstk is the ratio of managers’ stock holding. *** denotes significance at the 1％ level. Table 4 shows the impact of credit lines on overinvestment. Model (1)-(3) of Table 4 use the credit lines variable Creditdum to compare overinvestment behavior of firms with credit lines to firms without them. Creditdum is 1 if a firm gets lines of credit and 0 otherwise. Model (1) presents a significantly (at the 10% level) positive relationship between the access to credit lines and the overinvestment after controlling the free cash flow influence. Model (3) adds a joint variable, SOE×Creditdum, to analyze the difference in overinvestment behavior between State-owned and private-owned firms. We find that within the credit lines group, State-owned firms tend to overinvest more than the private-owned firms. Model (4)-(6) of Table 4 further investigate the impacts of the level of credit lines on overinvestment among firms with credit lines. We use the quantitative variable LOC, ratio of credit lines to the total assets, to measure the amount of credit lines. We find no significant impact credit line level on overinvestment. Given that only a small number of firms (less than 25% of listed firms) received credit lines, it appears that obtaining credit lines itself has strong influence on the firms’ overinvestment behavior, while the level of credit line has relatively weak influence. Table 5 divides firms within credit lines group into state-owned and private-owned firms to examine overinvestment behavior. Results from Model (1) and (2) indicate that only State-owned firms overinvest (t value = 0.005. No indication of overinvestment is found for private-owned firms. The results are consistent with our hypotheses that State-owned firms are more likely to overinvest than private-owned firms. Model (3) and (4) show the combined effects of bank credit lines and free cash flow on overinvestment. Both State and private-owned firms overinvest when free cash flows are high. From the joint effect variable, FCF×Creditdum, we find that credit lines accelerate free cash flow for State-owned firms, indicating State-owned firms have stronger incentives to overinvest than private firms.


49

Table 4: Test Results of Credit Lines on Overinvestment

Whole Sample Sample with Credit Lines (1) (2) (3) (4) (5) (6)

Creditdum 0.003* 0.003* -0.002 (1.79) (1.79) (-0.69)

LOC 0.008 0.011 0.024 (0.88) (1.21) (1.45)

FCF 0.038*** 0.038*** 0.039*** 0.069*** 0.077*** 0.077*** (10.23) (9.47) (9.49) (6.33) (6.02) (6.01)

SOE -0.001 -0.002 0.005 0.008 (-0.78) (-1.48) (1.43) (1.58)

HHI5 -0.005 -0.005 -0.006 -0.006 (-1.48) (-1.44) (-0.64) (-0.63)

Magstk 0.020** 0.023** 0.014 0.012 (2.16) (2.37) (0.86) (0.70)

SOE× Creditdum 0.008** (2.13)

SOE×LOC -0.018 (-0.86)

Constant -0.001** 0.000 0.001 -0.001 -0.005 -0.007 (-2.15) (0.10) (0.61) (-0.63) (-1.20) (-1.45)

N 9498 8866 8866 1408 1271 1271 R2 0.020 0.020 0.021 0.039 0.047 0.047 F 55.762 22.637 19.573 21.509 8.840 7.386

The table reports results for the regression: Excessit=α+β1 Credtilineit+β2FCFit+β3HHI5it+β4Magstkit+β5SOEit+β6SOEit×Creditlineit+εit.. Excess is the overinvestment variable. Creditline is the credit lines variable, either a dummy variable indicating whether the firm has access to lines of credit (Creditdum) or the ratio of lines of credit to firm’s total assets (LOC). FCF is the free cash flow; HHI5 is the level of ownership concentration; Magstk is the ratio of managers’ stock holding; SOE is the ownership dummy variable (=1 if state-owned; and 0 otherwise). Columns (1)-(3) show results for the full sample of listed firms. Columns (4)-(6) show results for the sample of firms with credit lines. The t-values are in the parentheses. ***, ** and * indicate significance at the 1, 5 and 10 percent levels respectively. Our previous analysis of the impact of credit lines is based upon the assumption that obtaining credit lines is exogenously given. However, approval of credit lines is not a random event. As Sufi (2009) and Jimenez et al. (2009) pointed out, firms’ ability to obtain credit lines is related to their own characteristics, such as profitability, growth potential, cash flow, debt ratio and firm size. To control for endogeneity, we use Heckman’s (1979) two-step method, Treatment Effects Model, to re-estimate the results. We first use the following Probit model to predict the access ability to credit lines,

itititit

itititititit

TangLeverageAgeSizeCflowTobinROAInvtCreditdum

eββββββββa

+++++++++=

−−−

−−−−−

181716

1514131211 (5)

From Equation (5), we get the Inverse Mills Ratio, and then take it as an extra control variable to add to Equation (2) and (3). The results are consistent with our previous estimates. Our estimate of overinvestment also depends on the measurement of investment opportunity. The Tobin’s Q measurement may be subject to some estimation error in Equation (1). Therefore, we replace Tobin’s Q with the growth rate of operating income to re-estimate the overinvestment in Equation (1), while at the same time we take into account for the endogenous selection bias. We do not find any significant change in the results after the replacement.


50

Table 5: Results of Credit lines on Overinvestment for the State-owned and Private-owned firms

(1) (2) (3) (4)

State-owned Private-owned State-owned Private-owned

Creditdum 0.005** -0.002 0.003 -0.003

(2.58) (-0.54) (1.60) (-1.10)

FCF 0.049*** 0.023*** 0.046*** 0.021***

(8.96) (4.30) (8.04) (3.87)

HHI5 -0.008* -0.000 -0.008* -0.001

(-1.78) (-0.03) (-1.76) (-0.09)

Magstk -0.042 0.030*** -0.046 0.028***

(-0.91) (2.98) (-1.01) (2.77)

FCF× Creditdum 0.042** 0.034*

(2.47) (1.75)

Constant -0.001 -0.000 -0.001 -0.000

(-0.76) (-0.15) (-0.69) (-0.08)

N 6290 2576 6290 2576

R2 0.026 0.016 0.027 0.017

F 22.477 8.735 19.387 7.902 The table reports results for following regression: Excessit =α+β1 Creddumit+β2FCFit+β3HHI5it+β4Magstkit+β5FCFit×Creditdumit+εit. . Excess is the overinvestment; Creditdum is a dummy variable indicating whether the firm has access to lines of credit; FCF is the free cash flow; HHI5 is the level of ownership concentration; Magstk is the ratio of managers’ stock holding. Columns (1) and (3) show results for the State-owned firms. Columns (2) and (4) show the results for the private-owned firms. The t-values are in the parentheses. ***, ** and * indicate significance at the 1, 5 and 10 percent levels respectively. CONCLUDING COMMENTS Existing financial research suggests that credit lines provide speed and flexibility for firms in pursuit of investment opportunities, and such flexibility can also lead to overinvestment. We conduct an empirical analysis on the role of credit lines in the overinvestment behavior of Chinese firms during the period between 2001 and 2008. We utilize Richardson’s (2006) method to estimate overinvestment and set up two hypotheses: Credit lines provide incentives to overinvest and State-owned firms are more likely to overinvestment than private firms. We find evidence that bank credit lines, similarly to free cash flows, are venerable to abuse by managers of firms in China and therefore, they can cause agency problems. These findings are consistent with Sufi’s (2009) arguments. However, we find such overinvestment behavior is limited to State-owned firms with credit lines. We find no evidence of overinvestment activities for private-owned firms. We attribute this fact to differential abilities for obtaining bank credit lines. Private-owned firms have a disadvantage in competing for bank credit lines comparing to State-owned firms, and therefore, they use approved credit lines more carefully. On the other hand, all major Chinese banks are owned or controlled by the government. They have a preference to provide credit lines to State-owned firms. With cheaper costs and easier access to credit lines, State-owned firms have a stronger incentive to overinvest. The results found in this paper show that excess liquidity and bank credit resources concentrated in State-owned enterprises lead to distortions in the efficiency of capital allocation and investment. Future reforms in the banking system and easier access to bank credit resources for private-owned firms are necessary in China. Although the focus in this paper is the linkage between credit lines and overinvestment, we do not rule out the possibility that bank credit lines can also increase investment efficiency by reducing financing constraints or facilitating cash management. This can even occur for State-owned enterprises in some cases. Thus, identifying the trade-off between the positive and negative effects of bank credit lines and examining the relationship between bank credit lines and investment


51

efficiency in different situations is an interesting topic for further investigation. REFERENCE Agarwal S., S. Chomsisengphet and J. Driscoll (2004) “Loan Commitments and Private Firms,” FEDS Working Paper No. 2004-27. Almeida H., M. Campello and M. Weisbach (2004) “The Cash Flow Sensitivity of Cash,” The Journal of Finance, Vol. 59 p. 1777-1804. Avery R. B. and A. N. Berger (1991) “Loan Commitments and Bank Risk Exposure,” Journal of Banking and Finance, Vol. 15(1) p. 173-192. Berger A. and G. Udell (1995) “Relationship Lending and Lines of Credit in Small Firm Finance,” Journal of Business, Vol. 68 p. 351-381. Berkovitch E. and S. Greenbaum (1991) “The Loan Commitment as An Optimal Financing Contract,” Journal of Financial and Quantitative Analysis, Vol. 26 p. 83-95. Boot A., A. V. Thakor and G. F. Udell (1987) “Competition, Risk Neutrality, and Loan Commitments,” Journal of Banking and Finance, Vol. 11 (September) p. 449-471. Boot A., A. V. Thakor and G. F. Udell (1991) “Credible Commitments, Contract Enforcement Problems and Banks: Intermediation as Credibility Assurance,” Journal of Banking and Finance, Vol. 15 (June) p. 605-632. Campbell T. S. (1978) “A Model of the Market for Lines of Credit,” Journal of Finance, Vol. 33 (March) p. 231-244. Dennis S., D. Nandy and I. G. Sharpe (2000) “The Determinants of Contract Terms in Bank Revolving Credit Agreements,” Journal of Financial and Quantitative Analysis, Vol. 35 p. 87-110. Duan J. C. and S. H. Yoon (1993) “Loan Commitments, Investment Decisions and The Signaling Equilibrium,” Journal of Banking and Finance, Vol. 17 (4) p. 645-661. Faulkender M. and R. Wang (2006) “Corporate Financial Policy and The Value of Cash,” Journal of Finance, Vol. 61 p. 1957-1990. Ham J. C. and A. Melnik (1987) “Loan Demand: An Empirical Analysis Using Micro Data,” Review of Economics and Statistics, Vol. 69 (November) p. 704-709. Hawkins G. D. (1982) “An Analysis of Revolving Credit Agreements,” Journal of Financial Economics, Vol. 10 (March) p. 59-81. Jensen M. And W. Meckling (1976) “Theory of The Firm: Managerial Behavior, Agency Costs, and Capital Structure,” Journal of Financial Economics, Vol. 3 p. 305-360. Jimenez G., J. Lopez and J. Saurina (2009) “Empirical Analysis of Corporate Credit Lines,” Review of Financial Studies, Vol. 22 p. 5069-5098. Kashyap A. K, J. C. Stein and D. W. Wilcox (1993) “Monetary Policy and Credit Conditions: Evidence from the Composition of External Finance,” American Economic Review, Vol. 83 (March) p. 78-98. Maksinovic V. (1990) "Product Market Imperfections and Loan Commitments," Journal of Finance, Vol.


52

45 (December) p. 1641-1655. Martin J. S. and A. M. Santomero (1997) “Investment Opportunities and Corporate Demand for Lines of Credit,” Journal of Banking and Finance, Vol. 21 (October) p. 1331-1350. Melnik, A. and S. Plaut (1986) “Loan Commitment Contracts, Terms of Lending, and Credit Allocation,” Journal of Finance, Vol. 41 (June) p. 425-435. Morgan D. P. (1994) “Bank Credit Commitments, Credit Rationing, and Monetary Policy,” Journal of Money, Credit and Banking, Vol. 26 (February) p. 87-101. Petersen M. A. (2009) “Estimating Standard Errors in Finance Panel Data Sets: Comparing Approaches,” Review of Financial Studies, Vol. 22 p. 435-480. Richardson S. (2006) “Over-Investment of Free Cash Flow,” Review of Accounting Studies, Vol. 11 p. 159-189. Shockley R. L. and A. V. Thakor (1997) “Bank Loan Commitment Contracts: Data, Theory, and Tests,” Journal of Money, Credit and Banking, Vol. 29 (November) p. 517-534. Sofianos G., P. Wachtel and A. Melnik (1990) "Loan Commitments and Monetary Policy," Journal of Banking and Finance, Vol. 14 (October) p. 677-689. Sufi A. (2009) “Bank Lines of Credit in Corporate Finance: An Empirical Analysis,” Review of Financial Studies, Vol. 22 p. 1057-1088. Thakor A.V. and G. F. Udell (1987) “An Economic Rationale for The Pricing Structure of Bank Loan Commitment,” Journal of Banking and Finance, Vol. 11(2) p. 271-289. Tobin J. (1969) “A General Equilibrium Approach to Monetary Theory,” Journal of Money, Credit and Banking, Vol. 1 p. 15-29. ACKNOWLEDGEMENT The authors are grateful for the financial support from National Natural Science Foundation (No. 70902024 and No. 71003108), and Guangdong Soft Science Foundation (No.2011B070400008, No.2011B070300024). BIOGRAPHY Qianwei Ying is Associate professor, Business School, Si Chuan University, 29 Wang Jiang Road, Chengdu, China, 610064. He can be reached at his email [email protected]. Danglun Luo is Associate Professor, Lingnan College, Sun Yat-sen University，135 Xingang Xi Road, Guangzhou, China, 510275. He can be reached at: [email protected]. Lifan Wu is Professor of Finance at California State University, Los Angeles. His research appears in journals such as Journal of Business Finance and Accounting, Global Finance Journal, Applied Financial Economics, International Journal of Finance, Asia-Pacific Financial Market, and Journal of Financial and Quantitative Analysis. He can be reached at: [email protected].


53

LONG- TERM PRIOR RETURN PATTERNS IN STOCK RETURNS: EVIDENCE FROM EMERGING MARKETS

Sanjay Sehgal, University of Delhi Sakshi Jain, University of Delhi

Pr Laurence the Porteu de la Morandiere, Group ESC Pau

ABSTRACT

In this paper, we identify long-term prior return patterns in stock returns for Brazil, Russia, India, China, South Korea, and South Africa (BRICKS) markets from January 1993 to February 2008. While Brazil, Russia and South Africa report momentum behavior, India, China and South Korea exhibit contrarian patterns for long-term prior return (24-60 months) as well as company characteristic(s) and prior return based portfolios. The CAPM is a poor descriptor of asset pricing as it doesn’t explain abnormal returns on these trading strategies for India and South Korea. It works well for other markets only for 24 and 36 months portfolio formation windows. The Fama-French (FF) model is able to explain most of the abnormal returns except 24-12-12 strategy for China and South Africa and 36-12-12 strategy for India. We find long-term prior return patterns in sector returns and that our augmented FF model, which contains a prior return sector factor, does a better job than the FF model. The research contributes to asset pricing and behavioral finance literature for emerging markets. Our findings shall be useful for global portfolio managers who analyze emerging markets, to combine them with mature markets for achieving risk diversification benefits. JEL: C51, C52, G12, G14, G15 KEYWORDS: CAPM, Momentum, Contrarian, Fama French model, and behavioral finance. INTRODUCTION

oldman Sachs in 2001 gave the acronym BRICs for the countries Brazil, Russia, India and China. The growth projections for these countries suggest they will become a world economic force. Goldman Sachs forecast that BRICs economies combined GDP can become larger than G7

economies (US, Japan, UK, France, Germany, Italy and Canada) by 2039. Recently, the investment banking industry has expanded the emerging markets basket from BRICs to BRICKS, by including South Korea (K) and South Africa (S). BRICKS represent one of the most active segments of emerging markets. They are on the radar of portfolio managers because emerging markets exhibit a low co-relation with mature markets. These markets have a history of providing high returns while ensuring risk diversification. Investment managers are continuously on the lookout for trading strategies that provide abnormal returns and analyze stock characteristics and return patterns for this purpose. In this paper, we extend the work of Sehgal and Jain (2010) from India to BRICKS capital markets. The study period is from January 1993 to February 2008 for all countries except Russia where the sample period is from January 2000 to February 2008 due to non-availability of data. The paper is motivated by the work of DeBondt and Thaler (1985, 1987) where the portfolio formation windows are greater than 12 months. We work on 24-60 month portfolio formation windows and skip 12 months between portfolio formation and holding windows to control for any short-term momentum effects on the lines of Fama and French (1996). Predicting returns on assets based on past returns has gained importance in recent years. Trading strategies have been developed to predict returns on stock that could lead to abnormal profits. Broadly,

G

S. Sehgal et al | IJBFR ♦ Vol. 7 ♦ No. 2 ♦ 2013

54

there are two trading strategies based on prior returns, one in which returns exhibit continuation (momentum) and the other in which returns have a tendency towards fundamental reversion in the long run (contrarian). These strategies have been found to be time dependent. Contrarian strategies perform well for very short term (up to 3 months), see Lo and MacKinlay (1990) and long term (3 years - 5 years), see De Bondt and Thaler, (1985, 1987) while momentum strategies perform well for short term (between 3months - 12 months), see Jegadeesh and Titman, (1993). Some researchers show that risk factors such as size, book-to-market equity, past sales growth, cash flow/price are related to firm’s average stock returns. Asset pricing models such as CAPM and Fama French three-factor model account for some of these risk factors, however others suggest that these abnormal returns have a behavioral explanation, that is, investors underreact or overreact to firm specific information. In the world of diversification managers are interested in holding portfolios across mature as well as emerging markets because of risk differentials in these markets. The asset pricing anomalies help managers develop trading strategies that provide extra normal returns and thus fund managers chase them to gain from these trading strategies. In this study, we test for abnormal returns for portfolios formed on the basis of long-term past returns (Return Portfolios) and company characteristic (Size, P/B, P/E, Dividend Yield and Past sales growth) as well as long-term past returns (Double Sorted Portfolios). The study also considers Triple Sorted Portfolios based on size and Price to Book (P/B) ratio and size and Price to Earnings (P/E) ratio characteristics and past returns. The paper also examines if the single factor (CAPM) or multi-factor (Fama-French three-factor) models are able to explain prior return patterns. Further, we test if the observed stock prior return effect is an outcome of sector prior return effect and can we use the sector information as a risk factor to form a four-factor model that it is able to explain returns. The research contributes to asset pricing anomaly literature for emerging markets. We find Brazil, Russia and South Africa report momentum behavior while India, China and South Korea show contrarian behavior for characteristic and long-term prior return based portfolios. The CAPM is a poor descriptor of cross-section of average returns in case of India and South Korea, for other BRICKS economies, explains for 24 and 36 portfolio formation windows. The Fama French model explains almost all the long-term prior return patterns in stock returns with the exception of China and South Africa for 24-12-12 strategy and India for 36-12-12 strategy. Employing an additional sector factor along with Fama French factors, we observe that the average returns are explained for 36-12-12 strategy in case of India and 24-12-12 strategy in case of South Africa. However, the abnormal returns for China in case of 24-12-12 strategy persist and remain an asset pricing puzzle. The remainder of the paper is organized as follows: Section 2 gives review of literature. Section 3 contains description of data and the methodology employed along with the empirical tests carried out. Section 4 summarizes the results and the interpretations. Section 5 provides concluding comments. LITERATURE REVIEW Two prior return patterns in stock returns namely contrarian and momentum have been empirically observed by researchers. While contrarian, implies return reversals, generally observed in long-term portfolio formation windows (24 months or more), momentum, implies continuation of returns, which has been observed for short-term portfolio windows (3-12 months). DeBondt and Thaler (1985, 1987) were first to document contrarian strategies. They report that individuals over react to information and subsequent price corrections lead to return reversals. This may imply a contrarian investment strategy based on buying past losers and selling past winners. For U.S. market, DeBondt and Thaler find that


55

portfolios of losers provide large returns even for more than five year portfolio formation windows. Chan (1988), Ball and Kothari (1989), Chopra, Lakonishok and Ritter (1992) have argued that systematic risk of contrarian portfolios can be attributed to the size effect. Conrad and Kaul (1993) examine the contrarian strategy and find that contrarian profits exist in long periods and tend to increase over time. Kaul and Nimalendran (1990) and Jegadeesh and Titman (1995) document that bid-ask spreads can explain short-term reversals which arise due to stock price over reaction to firm specific information. Return reversals may not be the only cause of contrarian profits. Both over-reaction and under-reaction (delayed reaction) of price to information can lead to contrarian profits; see Lo and MacKinlay (1990). Ball, Kothari and Shanken (1995) indicate that long-term reversals are based on microstructure biases, especially for low-priced stocks. The phenomenon of microstructure biases is related to book-to-market (B/M effects) as discussed by Chan, Hamao, and Lakonishok (1991), Fama and French (1992). Lakonishok, Shliefer and Vishny (1994), Schiereck, DeBondt and Weber (1999) report contrarian profits exist for long-term windows i.e. investing for 3 years to 5 years; earn excess returns of approximately 8 percent per annum. Momentum strategies have also attracted attention of investors because they exploit continuation patterns in stock returns. Jegadeesh and Titman (1993) document that in this strategy investors buy the decile of stocks with highest past returns and sell the decile of stocks with the lowest past returns. Stock momentum remains one of the most puzzling asset pricing anomalies. Chan, Jegadeesh and Lakonishok (1996) attribute momentum to market underreaction to firm specific information. Several behavioral theories have been developed to explain the momentum phenomenon suggesting that investors initially underreact and eventually overreact to firm specific news (Barberis, Shliefer and Vishny, 1998; Daniel, Kent and Subrahmanayam, 1998; and Hong and Stein, 1999). Hong Lim and Stein (2000) report that profitability of momentum strategies declines sharply with firm size and profitability has a positive impact for stocks that have low analyst coverage. Jegadeesh and Titman (2001) document that positive momentum returns may or may not be associated with post holding period reversals, suggesting that behavioral models provide partial explanation to behavioral theories. Fama and French (1996) argue that some other missing risk factor(s) may be the cause of profitability of momentum strategies. Conrad and Kaul (1998) argue that profitability of momentum strategies is due to cross-sectional variation in mean returns of individual securities. (Berk, Green and Naik, 1999; Chordia and Shivkumar, 2000) show that momentum profits are generated by time-varying expected returns. Lee and Swaminathan (2000) find price momentum strategy and volume based momentum strategy for American securities to be profitable over various portfolio formations and holding periods. Chan, Hameed and Tong (2000) find strong continuation in stock returns following an increase in trading volume and suggest that momentum profits arise from time-series predictability in stock market and very little from predictability in currency markets. Jegadeesh and Titman (2002) empirically demonstrate that cross-sectional differences in expected returns explain very little of momentum profits. Lewellen (2002) finds that momentum is explained by excess co-variance and not underreaction. Ahn, Conrad and Dittmar (2003) show their non-parametric adjustment of risk accounts for half the momentum profits. Scott, Stump and Xu (2003) document that after controlling for earnings-related news and stock's growth rates, interaction between momentum and volume largely disappears. Kent, Hirshleifer and Subrahmanayam (2004) propose a theory based on overconfidence and biased self-attribution to explain several securities return patterns. Antoniou, Lam and Paudyal (2007) report that some missing risk factor related to business cycle can probably explain momentum in European markets and behavioral models do not explain much of momentum. Chen, Chen, Hsin, and Lee (2010) examine relationship between price (return) momentum, earnings momentum and revenue momentum using US market data, and find all the three strategies to be profitable. Profits from price momentum strategy are largest and persistent followed by earnings momentum and revenue momentum.


56

A number of authors have examined the behavior of foreign investors in emerging markets and conclude that investors in general adopt momentum based strategies, see Frankel and Schmukler (1996, 1998), Froot, Conell and Seaholes (2001), Richards (2002) and Kaminsky, Lyons and Schmukler (2002). Chui, Titman, and Kim (2000) document that momentum strategies are highly profitable when applied to eight Asian markets outside Japan. Kim and Wei (2002) find that foreign investors living outside Korea are more likely to employ momentum and herding trading than foreign individuals living in Korea as a result of information asymmetry. Hameed and Kusnadi (2002) report little evidence that momentum strategy when applied to individual stocks in six Asian markets yield significant profits. Lin and Swanson (2004) find evidence that foreign inflows have short-term positive impacts on local market returns, but find only minimal evidence that foreign investors employ momentum trading. Swanson and Lin (2005) investigate eighteen emerging markets (which include all the BRICKS countries) and eighteen developed markets over the period 1992 -2003. They conclude that U.S. investors tend to employ winners-momentum trading strategy (buy past winners) in emerging markets, developed markets and global market and employ losers-contrarian trading (buy past losers) in all the three markets segments. If abnormal returns are caused by time-series or cross-sectional patterns in expected returns then some asset pricing theory should be able to absorb the extra normal profits. The CAPM of Sharpe (1964) and Lintner (1965), and Fama French three-factor model (1993) are two theories, which try to explain cross section of average stock returns. The CAPM provides the linear relationship between expected return and risk of a financial asset. CAPM was unable to explain the market anomalies and hence Fama French (1993) developed three-factor model which states that the expected returns on a portfolio in excess of the risk free rate are explained by the sensitivity to three factors (market, size and value). Carhart (1997) employ a four-factor model to explain returns with an additional factor of one-year stock momentum along with Fama French factors, to capture cross-sectional return patterns as the Fama-French three-factor model could not explain the momentum anomaly. The average returns which are missed by standard asset pricing models could be explained by additional risk factor(s) or they may have a behavioral explanation. Many academicians have observed sector/industry and country patterns in stock returns and the fact that they may explain prior return patterns in individual security returns. Recently, research analysts have focused on specializing in economic sectors than individual companies (within a size/style bucket) as companies within a sector have much more in common in terms of their business model. Many studies have been conducted to understand whether the prior return effects can be attributed to companies belonging to winner sectors/industry or loser sectors/industry. Moskowitz and Grinblatt (1999) document strong momentum effect in industry components of stock returns. The industry momentum investment strategies appear to be profitable even after controlling for size, book-to-market equity, individual stock momentum, cross-sectional dispersion in mean returns and potential microstructure influences. Nijman, Swinkels and Verbeek (2004) investigate whether individual stock momentum in Europe is subsumed by country or industry momentum and suggest that positive expected excess returns are primarily driven by individual stock effects, while industry momentum plays a less important role, country momentum is even weaker. Scowcroft and Sefton (2005) analyze the value-weighted large-capitalization universe and find that price momentum is driven by industry momentum and not individual stock momentum whereas in the small-cap universe, stock-specific effects assume greater importance. Boni and Wamock (2006) report industry-based recommendation strategies and the short-term industry price momentum are explained by firms with more analyst coverage than firms with low analyst coverage. Menzly and Ozbas (2006) find strong cross-industry momentum for industries related to each other through supply chain. Chen, Benett and Zhang (2006) suggest investors should emphasize a sector based approach in developed countries but continue country-based allocation strategies for emerging


57

markets. Safieddine and Sonti (2007) report firms with the highest industry growth quintile have significantly higher momentum compared to industries in the lowest growth quintile. Liu and Zhang (2008) document the growth rate of industrial production is a risk factor in asset pricing tests that explains more than half of momentum profits. DATA AND METHODOLOGY Data comprises of monthly share prices adjusted for stock splits, stock dividends and rights issues for BRICKS markets and has been obtained from Thomson Reuters Datastream software. The sample period of the study is January 1993 to February 2008 except for Russia where the sample period is January 2000 to February 2008 due to paucity of data. Exhibit A shows the number of securities used for analysis along with market indices and their description for the sample countries. The companies account for a reasonable part of market capitalization and trading activity in their respective markets. Hence, the data set fairly represents market performance. Exhibit A: Data Description for Sample Countries

Country No. of Securities Market Index Index Description

Brazil 195 BRAZIL BOVESPA

BM&FBOVESPA S.A. is a security market index with base year 1968 and base value of 100. It is a total return index and handles about 85% of the total volume traded on country's nine stock exchanges.

Russia 75 RUSSIA RTS INDEX

The Russian Trading System Index is a capitalization-weighted index. The index was developed with a base value of 100 in 1995. It uses free float adjusted weights.

India 450 INDIA BSE-200 (SENSEX)

BSE-200 index is a free-float value weighted index that represents nearly 93% of the total market capitalization on the Bombay Stock Exchange. The financial year 1989-90 has been chosen as the base year.

China 600 SHANGHAI SE A SHARE

The Shanghai A-Share Stock Price Index is a market capitalization-weighted index. The index was developed with a base value of 100 on December 19, 1990. It comprises of all the A-shares which are restricted to trading by local investors and qualified institutional foreign investors.

South Korea 500 KOREA SE COMPOSITE (KOSPI)

The KOSPI 200 index consists of 200 Korean stocks which constitute 93% of the total market value on the Korea Stock Exchange. The index was developed with base value of 100 in the year 1990.

South Africa 250 FTSE/JSE Africa ALL SHARE

The FTSE/JSE All Africa Index Series is designed to represent the performance of the top African companies listed on Johannesburg Stock Exchange. Companies included consist of top 99% of the total pre-free float market capitalization. The FTSE/JSE Africa Index Series replaced the JSE Actuaries indices on the 24th of June 2002.

This table provides a data description for the sample countries. Monthly share prices for estimation purposes and further analysis have been converted to percentage monthly return series. Stylized portfolios are formed on the basis of past percentage returns (Percentage Returns estimation is based on capital gains component. There is no dividend component as in India, dividend yields of companies are very low, Gupta (2000). Also, all the Bombay Stock Exchange (BSE)-500 index series do not include dividends while computing index values. Hence, dividend inclusion in individual stock returns may bias the estimators of our proposed time series regressions) and characteristics such as the company Size, Price to Book (P/B) ratio, Price to Earnings (P/E) ratio, Dividend Yield and Past Sales Growth. Past Sales Growth is estimated as three year compounded growth rate in sales using the formula St+3= St (1+r)3, where St+3 and St are sales revenue in year t+3 and t respectively whereas r is compounded growth rate in sales. Annualized implicit yields on 91-day t-bills for each country available for all weekly auctions over the study period have been used. We select the implicit yield for the last week of each month to match month end closing prices of sample stocks. The end of month annualized implicit yield is divided by 12 to generate approximate monthly risk free yields.


58

Value-weighted market index is used as surrogate for aggregate economic wealth. Data for above said firm characteristics and market index has also been obtained from Thomson Reuters DataStream. Global Industry Classification System (GICS) was used for sector classification. GICS comprises of 10 sectors, 24 Industry groups, 68 industries and 154 Sub-industries. The data for sector and industry classification were taken from World Scope, Reuters Financials & Compustat Global. It was developed in response to the financial community’s need for one complete, consistent set of global sector and industry definitions. The GICS standard can be applied to companies globally, in both developed and developing markets. In our work, only information for sectors has been used. The 10 prominent sectors are Energy, Materials, Industrials, Consumer Discretionary, Consumer Staples, Health Care, Financials, Information Technology, Telecommunication Services and Utilities. Methodology We explore long term prior return patterns in asset returns, by forming three types of portfolios: 1) On the basis of average past returns of 24, 36, 48 and 60 months (single sorted portfolios), 2) On the basis of firm characteristics such as market capitalization, price to book (P/B) ratio, price to earnings (P/E) ratio, dividend yield and past sales growth (PSG) and average past returns of 24-60 months (double sorted portfolios), and 3) On the basis of size and price to book (P/B) ratio / size and price to earnings (P/E) ratio and prior returns of 24-60 months (triple sorted portfolios). Past sales growth (PSG) is computed as compounded growth rate in net sales three years prior to portfolio formation. For ranking we use size characteristic measured by market capitalization (Banz, 1981), and value characteristic measured by P/B (Chan, Hamao and Lakonishok, 1991), P/E (Basu, 1983) and PSG (Fama French, 1996). While the first two measures are scaled price variables, the third measure is a fundamental based proxy. We also sort securities on dividend yield as it may affect stock returns owing to differential treatment of dividend and capital gain income (Litzenberger and Ramaswamy, 1979). Jung and Shiller (2005) find that dividend yield ratios of individual companies have considerable ability to predict the future growth rate of real dividends (higher yields go together with lower future growth rates) The portfolios were formed on basis of (i months-j months-k months) strategy where i months involve portfolio formation period, ranging from 24-60 months, j months represent the 12 months that we skip between portfolio formation and portfolio holding period, while k is fixed at 12 months as portfolio holding period. We skip 12 months between portfolio formation and holding windows, as suggested by Fama and French (1996), to control for short-term momentum patterns that may hamper any clear discerning of long-term prior return patterns. We follow calendar year i.e. from January to December. The portfolio formation process of single sorted portfolios for 24 months-12 months-12 months (24-12-12) strategy is done as follows: In December of year t-2, the individual securities are ranked on the basis of past twenty four month’s average monthly past excess returns. The ranked securities are then classified into quintiles, P1 to P5. P1 and P5 comprises of bottom and top 20% stocks respectively, on basis of average past period returns. Equally weighted returns are then estimated for sample portfolios leaving a 12 month gap between portfolio formations and holding windows (i.e. January to December of year t-1). The portfolios are rebalanced in December of year t-1 and portfolio returns are estimated for year t. The process is repeated through the end of our sample period. Our single sorted portfolios are non-overlapping by construction. However, based on previous literature, we believe results for non-overlapping portfolios will not significantly differ from overlapping portfolios (Jegadeesh and Titman, 1993). Next, we sort sample companies on different company characteristics and observe prior return patterns within each characteristic group. Our analysis is inspired by past research which indicates the relationship between company characteristics and returns. We construct double sorted portfolios based on firm characteristics and long-term past excess returns for 24-12-12 investment strategy. In December of year t-


59

2, the sample securities are sorted into two groups, Small or S (bottom 50%) and Big or B (upper 50%) in case of company Size (measured by market capitalization) and Low or L (bottom 50%) and High or H (top 50%) in case of other company characteristics i.e. P/B ratio, P/E ratio, dividend yield and past sales growth. Within each characteristic group we construct three momentum portfolios i.e. (bottom (33⅓ %), middle (between 33⅓ % and 66 2/3%) and top (greater than 66 2/3%)) based on twenty-four months average past returns (t-2 and t-1 years). Equally weighted excess returns are estimated for sample portfolios skipping 12 months between portfolio formation and holding windows (i.e. January to December of year t-1) and the portfolios are rebalanced every 12 months based on double sorting criteria for the year t. The sub-portfolios are labeled as S1, S2, S3 and B1, B2, B3 for company size criteria and L1, L2, L3 and H1, H2, H3 for other company characteristics. Finally, we form Triple sorted portfolios based on size and price to book (P/B) ratio / Size and price to earnings (P/E) ratio and prior returns based on 24 months prior returns. The triple sorting procedure is done as follows: in December of year t-2, the sample securities are sorted on basis of company size into two groups, Small (S) and Big (B). Next, we regroup our sample stocks on basis of value factor ((P/B)/ (P/E)) and form two groups, Low (L) and High (H). We use intersection between the two criteria to form four portfolios, SL, SH, BL and BH. Within each four groups, we again construct three momentum portfolios as described for double sorted portfolios. The portfolios are labeled as SL1, SL2, SL3, SH1, SH2, SH3, BL1, BL2, BL3 and BH1, BH2, BH3. Estimation of 36 months-12months-12months (36-12-12), 48 months-12 months-12months (48-12-12) and 60months-12months-12months (60-12-12) investment strategies are done in similar manner. Average returns on sample portfolios are estimated to evaluate any discernible long-term prior return patterns in stock returns. The prior return patterns may be due to risk differences between the winners and the losers’ portfolios and hence any such abnormal profits should be explained by asset pricing model (s). If these abnormal returns persist even after risk compensation, they may warrant a behavioral explanation. Standard Risk models such as the one-factor Capital Asset Pricing model (CAPM) and three-factor Fama French model are used to predict relationship between returns on a portfolio and returns on the risk factor (s). CAPM provides linear relationship between returns on a financial asset and its sensitivity to returns on a broad based market portfolio. The excess return version of the market model is as follows: 𝑅𝑃𝑡 − 𝑅𝐹𝑡 = 𝛼 + 𝛽(𝑅𝑀𝑡 − 𝑅𝐹𝑡) + 𝑒𝑖 (1) Where, 𝑅𝑃𝑡 − 𝑅𝐹𝑡 = 𝐸𝑥𝑐𝑒𝑠𝑠 𝑅𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝑎 𝑝𝑜𝑟𝑡𝑓𝑜𝑙𝑖𝑜, 𝑅𝑀𝑡 − 𝑅𝐹𝑡 = 𝐸𝑥𝑐𝑒𝑠𝑠 𝑟𝑒𝑡𝑢𝑟𝑛 𝑜𝑛 𝑡ℎ𝑒 𝑚𝑎𝑟𝑘𝑒𝑡 𝑓𝑎𝑐𝑡𝑜𝑟, 𝛼 = 𝑀𝑒𝑎𝑠𝑢𝑟𝑒 𝑜𝑓 𝑎𝑏𝑛𝑜𝑟𝑚𝑎𝑙 𝑝𝑟𝑜𝑓𝑖𝑡𝑠, 𝛽 = 𝑆𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑠 𝑜𝑓 𝑠𝑡𝑜𝑐𝑘 𝑟𝑒𝑡𝑢𝑟𝑛𝑠 𝑡𝑜 𝑡ℎ𝑒 𝑚𝑎𝑟𝑘𝑒𝑡 𝑟𝑒𝑡𝑢𝑟𝑛𝑠, 𝑒𝑖 = 𝐸𝑟𝑟𝑜𝑟 𝑡𝑒𝑟𝑚. Alternatively, Fama French in 1993 developed a multi-factor model to explain CAPM anomalies. The model states that expected returns on a portfolio in excess of the risk free rate are explained by three factors: market, size and value. The size and value factors11 are constructed using the methodology given in Fama French (1993) paper. The Fama French Model is given as: 𝑅𝑃𝑡 − 𝑅𝐹𝑡 = 𝛼 + 𝛽(𝑅𝑀𝑡 − 𝑅𝐹𝑡) + 𝑠𝑆𝑀𝐵𝑡 + 𝑙𝐻𝑀𝐿𝑡 + 𝑒𝑖 (2)


60

Where: 𝑆𝑀𝐵 = Difference between returns on portfolio of small stocks firms and returns on portfolio of big stocks firm, 𝐻𝑀𝐿 = Difference between returns on a portfolio of high-book-to- market stocks and returns on a portfolio of low-book-to-market stocks, 𝑠 𝑎𝑛𝑑 ℎ = Sensitivity coefficients of SMB and HML respectively. All other terms are same as equation (1). We use LMH factor instead of Fama-French HML factor as we are using P/B as a value factor and not book to market, hence our interpretation of value factor shall be inverse to that of Fama French. SMBt is constructed such that it is independent of value factor: 𝑆𝐿+

𝑆𝑀+

𝑆𝐻

3−

𝐵𝐿+

𝐵𝑀+

𝐵𝐻

3 (3)

𝐿𝑀𝐻𝑡 is constructed such that it is independent of size factor: 𝑆𝐿+

𝐵𝐿

2−

𝑆𝐻+

𝐵𝐻

2 (4)

The double sorted size-value portfolios for calculating SMB and LMH are formed from intersection of the two size groups, small or S, (bottom 50%) and big or B, (top 50%) and three value groups, low or L bottom (33⅓ %), medium or M (between 33⅓ % and 66 2/3%) and high or H (greater than 66 2/3%). We regress, the excess returns on test portfolios on the excess return on market factor (CAPM specification) and returns on market, size and value factors (Fama French three model specification). Next, we evaluate any prior return patterns in sector data and whether the sector prior return factor is able to absorb past return patterns for China and South Africa 24-12-12 strategy and India, 36-12-12 strategy which are missed by Fama French three-factor model. For 24-12-12 strategies, in December of year t-2, we categorize the sample securities into 10 sectors according to Global Industry Classification System

(GICS). The excess monthly return for each sector is then calculated from January to December by taking the simple average of returns on securities that form part of each of these sectors. The individual sectors are then ranked on the basis of past twenty four month’s average monthly past excess returns. The ranked sectors are then classified into quintiles, K1 to K5. K1 and K5 comprise sectors with the lowest and highest average past returns respectively. Equally weighted excess returns are then estimated for sample portfolios by skipping 12 months between portfolio formations and holding windows (i.e. January to December of year t). The portfolios are then rebalanced until we reach the end of sample period. A similar construction procedure is followed for other long-term portfolio formation strategies. We form a sector prior return factor to test whether patterns observed in sectors can explain prior return patterns in stocks. The sector factor is formed on the arguments of Liu and Zhang (2008) who report that recent winners have temporarily higher loadings for growth rate of industrial production than recent losers, and the combined effect of growth rate of industrial production loadings and risk premiums account for more than half of momentum profits. They also suggest that expected-growth risk is priced and that the expected-growth risk increases with expected growth. Carhart (1997) augmented the F-F model by adding a stock momentum factor; following Fama French (1996) in which the three factors (market, size and value) could not explain momentum profits. However, the Carhart stock momentum


61

factor has a weak economic foundation and hence we augment the FF model with a sector prior return factor which is constructed by taking the difference in returns between the winner and loser sector portfolios on period to period basis. The four factor model is as follows: 𝑅𝑃𝑡 − 𝑅𝐹𝑡 = 𝛼 + 𝛽(𝑅𝑀𝑡 − 𝑅𝐹𝑡) + 𝑠𝑆𝑀𝐵𝑡 + 𝑙𝐿𝑀𝐻𝑡 + 𝑤𝑊𝑀𝐿𝑡 + 𝑒𝑖 (5) 𝑊𝑀𝐿 = 𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑓𝑖𝑟𝑚𝑠 𝑜𝑓 𝑤𝑖𝑛𝑛𝑒𝑟 𝑠𝑒𝑐𝑡𝑜𝑟 𝑎𝑛𝑑 𝑓𝑖𝑟𝑚𝑠 𝑜𝑓 𝑙𝑜𝑠𝑒𝑟 𝑠𝑒𝑐𝑡𝑜𝑟, 𝑤 = 𝑓𝑎𝑐𝑡𝑜𝑟 𝑠𝑒𝑛𝑠𝑖𝑡𝑖𝑣𝑖𝑡𝑦 𝑜𝑓 𝑊𝑀𝐿 𝑓𝑎𝑐𝑡𝑜𝑟. EMPIRICAL RESULTS Table 1 reports average unadjusted returns estimated as excess returns (portfolio returns minus risk free rate) for winner and loser portfolios (henceforth, to be referred as corner portfolios) for different portfolio formation periods. A positive value for a return differential implies momentum whereas a negative differential means contrarian. For prior return portfolios, we observe momentum behavior for Brazil, Russia and South Africa and this pattern persists even as we elongate the portfolio formation windows and it only dissipates for 60-12-12 strategies. On the other hand, India, China and South Korea report contrarian behavior and the long term contrarian patterns become stronger for China as we elongate the formation windows. The highest returns of 3.95% are reported by Russia for 48-12-12 strategies. For characteristic sorted portfolios, half of the BRICKS baskets (Brazil, Russia and South Africa) reports momentum behavior while the remaining markets (India, China and South Korea) exhibit contrarian behavior. Contrarian patterns however become dominant for all the sample countries as we elongate the formation windows beyond 24 months. Next, we evaluate if these long-term prior return patterns can be explained by standard risk models. Tables 2 and 3 provide results of CAPM and Fama French model respectively for test portfolios. The CAPM is able to explain most of the prior return patterns in case of Brazil, Russia, China and South Africa for 24 and 36 months portfolio formation strategies but it doesn’t do a good job for longer term portfolio formation strategies i.e. 48 and 60 months. In case of India and South Korea, the CAPM seems to be a poor descriptor of prior return patterns across all long-term portfolio formation strategies. For India and South Korea, we further observe that momentum patterns exist for many trading strategies involving long-term portfolio formation windows which are contradictory to DeBondt and Thaler (1985) evidence which shows that U.S. market generally follows long-term reversals. The Fama French model is expected to capture most of the returns that are missed by CAPM. In the FF model, besides market, size (SMB) and value (LMH) are the two underlying risk factors of special hedging concern to investors. We find the FF model is able to explain long-term prior return patterns in stock returns for BRICKS countries with the exception of China 24-and South Africa for 24-12-12 strategy and India for 36-12-12 strategy. While in Brazil, China and South Africa, portfolio returns mainly load on the size factor, in case of India and South Korea the value factor plays a dominant role and the size factor doesn’t explain cross-section of average returns. Both size and value factors absorb prior return patterns in the case of Russia. Thus, the FF model does a better job than CAPM in explaining long-term prior return patterns for BRICKS markets. However, given the few anomalies in case of India, China and South Africa, we construct a sector prior return factor in the next section and evaluate if it can explain returns that are missed by the FF model.


62

Table 1: Average Unadjusted Returns Estimated as Excess Returns

Panel A: Mean Excess Return on 24-12-12 Stylized Portfolios Panel B: Mean Excess Return on 36-12-12 Stylized Portfolios BRAZI

RUSS

IND

CHI

S.KOR

S.AFRI

BRAZI

RUSS

IND

CHI

S.KOR

S.AFRI

Return Portfolios Return Portfolios P1 0.020 0.016 0.04

0.026 0.033 0.023 P1 0.023 0.013 0.04

0.032 0.028 0.025

P5 0.034 0.018 0.12

0.023 0.022 0.014 P5 0.035 0.019 0.07

0.026 0.019 0.016 P5-

0.014 0.082 -.004 -

-0.003 0.011 P5-

0.012 0.038 -.007 -

-0.002 0.013

Characteristic Sorted Portfolios Characteristic Sorted Portfolios SIZE SIZE

S1 0.019 0.011 0.03

0.034 0.029 0.008 S1 0.025 0.036 0.03

0.030 0.028 0.010 S3 0.031 0.090 0.03

0.027 0.022 0.018 S3 0.029 0.058 0.03

0.024 0.022 0.016

B1 0.006 0.032 0.01

0.028 0.016 0.002 B1 0.005 0.030 0.02

0.021 0.017 0.005 B3 0.003 0.034 0.01

0.019 0.009 0.000 B3 0.004 0.031 0.01

0.016 0.012 -0.001

S3-

0.012 0.080 -.002 -

-0.006 0.010 S3-

0.003 0.022 -.003 -

-0.006 0.006 H3-

-0.003 0.002 -.003 -

-0.007 -0.002 H3-

-0.001 0.001 -.004 -

-0.005 -0.006

PB PB L1 0.016 0.035 0.02

0.024 0.026 0.009 L1 0.023 0.032 0.03

0.023 0.030 0.015

L3 0.032 0.036 0.02

0.024 0.023 0.015 L3 0.031 0.032 0.02

0.016 0.028 0.015 H1 0.008 0.032 0.01

0.030 0.011 -0.001 H1 0.011 0.037 0.01

0.021 0.013 -0.001

H3 -0.002 0.037 0.01

0.015 0.004 0.002 H3 0.000 0.031 0.01

0.013 0.008 0.004 L3-

0.016 0.001 -.002 0.000 -0.003 0.005 L3-

0.008 0.000 -.004 -

-0.003 0.000

H3-

-0.010 0.005 -.001 -

-0.007 0.003 H3-

-0.011 -0.006 0.00

-

-0.005 0.005 PE PE

L1 0.020 0.038 0.03

0.032 0.025 0.011 L1 0.021 0.041 0.03

0.027 0.028 0.014 L3 0.027 0.057 0.02

0.022 0.016 0.013 L3 0.030 0.051 0.03

0.016 0.020 0.018

H1 0.006 0.026 0.01

0.026 0.015 -0.002 H1 0.011 0.027 0.02

0.021 0.018 0.001 H3 0.005 0.019 0.01

0.015 0.009 0.000 H3 0.003 0.018 0.01

0.012 0.012 0.002

L3-

0.007 0.019 -.007 -

-0.009 0.002 L3-

0.009 0.010 -.002 -

-0.008 0.004 H3-

-0.001 -0.007 0.00

-

-0.006 0.002 H3-

-0.008 -0.010 -.003 -

-0.006 0.001

DYIELD DYIELD L1 0.008 0.029 0.01

0.023 0.013 -0.004 L1 0.008 0.043 0.01

0.018 0.013 0.000

L3 0.000 0.033 0.01

0.014 0.009 -0.001 L3 0.003 0.027 0.01

0.013 0.013 0.000 H1 0.010 0.031 0.02

0.029 0.020 0.008 H1 0.014 0.031 0.02

0.023 0.024 0.012

H3 0.008 0.044 0.02

0.030 0.015 0.009 H3 0.008 0.054 0.02

0.021 0.017 0.009 L3-

-0.008 0.004 0.00

-

-0.003 0.003 L3-

-0.005 -0.017 -.001 -

0.000 -0.001

H3-

-0.002 0.013 -.006 0.001 -0.005 0.001 H3-

-0.006 0.023 0.00

-

-0.006 -0.003 SALES SALES

L1 0.020 0.041 0.02

0.029 0.024 0.015 L1 0.020 0.029 0.02

0.031 0.022 0.016 L3 0.007 0.037 0.02

0.018 0.017 0.020 L3 0.009 0.023 0.02

0.015 0.018 0.015

H1 0.010 0.028 0.02

0.031 0.021 0.012 H1 0.011 0.031 0.02

0.029 0.021 0.015 H3 0.006 0.036 0.02

0.018 0.018 0.010 H3 0.004 0.023 0.02

0.019 0.019 0.010

L3-

-0.014 -0.005 0.00

-

-0.006 0.005 L3-

-0.010 -0.005 -.003 -

-0.004 -0.001 H3-

-0.003 0.008 -.005 -

-0.003 -0.001 H3-

-0.007 -0.009 -.007 -

-0.002 -0.005

SIZE PB SIZE PB SL1 0.027 – 0.03

0.024 0.033 0.023 SL1 0.040 – 0.04

0.031 0.036 0.029

SL3 0.069 – 0.03

0.020 0.029 0.038 SL3 0.059 – 0.04

0.018 0.034 0.036 SH1 0.026 – 0.03

0.017 0.018 0.004 SH1 0.019 – 0.04

0.019 0.023 0.001

SH3 -.001 – 0.02

0.021 0.009 0.007 SH3 0.000 – 0.03

0.017 0.010 0.005 BL1 0.005 – 0.01

0.017 0.022 0.004 BL1 0.005 – 0.02

0.021 0.026 0.007

BL3 0.003 – 0.01

0.016 0.015 0.004 BL3 0.010 – 0.02

0.016 0.024 0.002 BH1 -.003 – 0.00

0.015 0.012 -0.006 BH1 0.003 – 0.01

0.018 0.014 -0.003

BH3 0.000 – 0.01

0.006 0.005 -0.008 BH3 0.000 – 0.01

0.011 0.008 -0.006 SL3-

0.042 – 0.00

-

-0.004 0.015 SL3-

0.019 – 0.00

-

-0.003 0.007

SH3-

-.027 – -.006 0.005 -0.009 0.003 SH3-

-.020 – -.007 -

-0.013 0.004 BL3-

-.003 – -.004 -

-0.007 0.000 BL3-

0.005 – -.005 -

-0.002 -0.005

BH3-

-.003 – 0.00

-

-0.008 -0.002 BH3-

-.010 – -.002 0.003 -0.005 -0.003 SIZE PE SIZE PE

SL1 0.018 – 0.04

0.026 0.029 0.028 SL1 0.025 – 0.05

0.029 0.027 0.029 SL3 0.057 – 0.03

0.016 0.025 0.032 SL3 0.059 – 0.04

0.017 0.027 0.033

SH1 0.020 – 0.02

0.022 0.027 0.007 SH1 0.013 – 0.02

0.021 0.030 0.009 SH3 0.014 – 0.02

0.017 0.016 0.003 SH3 0.009 – 0.02

0.018 0.015 0.002

BL1 0.010 – 0.02

0.018 0.022 0.006 BL1 0.009 – 0.02

0.022 0.024 0.012 BL3 0.005 – 0.01

0.017 0.009 0.003 BL3 0.007 – 0.02

0.016 0.014 0.001

BH1 -.002 – 0.01

0.022 0.011 -0.008 BH1 -.002 – 0.01

0.020 0.014 -0.004 BH3 0.001 – 0.01

0.006 0.010 0.000 BH3 0.004 – 0.01

0.008 0.013 -0.004

SL3-

0.039 – -.001 -

-0.005 0.003 SL3-

0.034 – -.005 -

0.000 0.004 SH3-

-.006 – -.001 -

-0.011 -0.005 SH3-

-.004 – -.003 -

-0.015 -0.007

BL3-

-.005 – -.009 -

-0.012 -0.004 BL3-

-.002 – -.007 -

-0.010 -0.011 BH3-

-.010 – 0.00

0.016 -0.001 0.008 BH3-

-.030 – -.001 0.015 -0.001 -0.001


63

Table 1: Continued

Table 1: Panel C: Mean Excess Return on 48-12-12 Stylized

Table 1: Panel D: Mean Excess Return on 60-12-12 Stylized BRAZIL RUSSIA INDIA CHINA S.KOREA S.AFRICA BRAZIL RUSSIA INDIA CHINA S.KOREA S.AFRICA

Return Portfolios Return Portfolios P1 0.030 0.018 0.038 0.035 0.028 0.035 P1 0.028 0.021 0.054 0.038 0.030 0.022 P5 0.043 0.012 0.077 0.030 0.015 0.026 P5 0.049 0.011 0.083 0.032 0.015 0.017

P5-P1 0.012 0.039 -.005 -0.013 -0.003 0.003 P5-P1 0.022 0.029 -.006 -0.016 -0.001 0.000 Characteristic Sorted Portfolios Characteristic Sorted Portfolios

SIZE SIZE S1 0.031 0.042 0.038 0.029 0.036 0.014 S1 0.036 0.042 0.043 0.031 0.024 0.019 S3 0.035 0.052 0.038 0.022 0.034 0.013 S3 0.052 0.054 0.040 0.020 0.023 0.014 B1 0.005 0.032 0.024 0.021 0.032 0.006 B1 0.010 0.036 0.028 0.023 0.018 0.012 B3 0.004 0.034 0.022 0.013 0.020 -0.002 B3 0.008 0.034 0.022 0.014 0.012 -0.001

S3-S1 0.005 0.010 0.000 -0.007 -0.002 -0.001 S3-S1 0.017 0.012 -.003 -0.011 -0.001 -0.005 H3-H1 -0.001 0.002 -.002 -0.007 -0.011 -0.008 H3-H1 -0.002 -0.002 -.006 -0.010 -0.006 -0.013

PB PB L1 0.026 0.028 0.036 0.029 0.043 0.016 L1 0.032 0.029 0.039 0.035 0.025 0.021 L3 0.038 0.028 0.033 0.018 0.039 0.017 L3 0.049 0.028 0.036 0.021 0.028 0.020 H1 0.009 0.037 0.019 0.021 0.026 0.002 H1 0.017 0.037 0.025 0.022 0.016 0.008 H3 0.005 0.029 0.022 0.011 0.017 0.001 H3 0.009 0.029 0.022 0.013 0.010 -0.004

L3-L1 0.012 0.000 -.003 -0.010 -0.004 0.001 L3-L1 0.017 -0.001 -.003 -0.013 0.003 -0.002 H3-H1 -0.004 -0.008 0.002 -0.010 -0.009 -0.001 H3-H1 -0.007 -0.008 -.002 -0.009 -0.006 -0.012

PE PE L1 0.020 0.041 0.039 0.030 0.041 0.015 L1 0.025 0.039 0.043 0.034 0.023 0.021 L3 0.037 0.044 0.040 0.015 0.033 0.017 L3 0.047 0.041 0.040 0.020 0.021 0.018 H1 0.012 0.027 0.021 0.022 0.030 0.002 H1 0.016 0.026 0.025 0.025 0.019 0.008 H3 0.004 0.026 0.019 0.012 0.020 -0.001 H3 0.009 0.026 0.019 0.012 0.012 -0.005

L3-L1 0.017 0.003 0.001 -0.015 -0.008 0.002 L3-L1 0.022 0.002 -.003 -0.014 -0.002 -0.003 H3-H1 -0.009 -0.001 -.002 -0.010 -0.011 -0.003 H3-H1 -0.008 0.000 -.005 -0.013 -0.007 -0.013

DYIELD DYIELD L1 0.011 0.044 0.021 0.019 0.026 0.003 L1 0.016 0.041 0.027 0.023 0.016 0.007 L3 0.005 0.023 0.020 0.010 0.021 0.000 L3 0.005 0.026 0.019 0.012 0.014 0.003 H1 0.012 0.042 0.033 0.026 0.037 0.015 H1 0.018 0.032 0.033 0.032 0.026 0.022 H3 0.014 0.037 0.033 0.019 0.028 0.012 H3 0.021 0.038 0.034 0.022 0.017 0.015

L3-L1 -0.005 -0.021 -0.001 -0.009 -0.004 -0.003 L3-L1 -0.011 -0.015 -0.008 -0.011 -0.002 -0.005 H3-H1 0.002 -0.005 0.000 -0.007 -0.009 -0.003 H3-H1 0.003 0.005 0.001 -0.010 -0.009 -0.007

SALES SALES L1 0.020 0.024 0.028 0.036 0.035 0.021 L1 0.024 0.036 0.033 0.041 0.016 0.026 L3 0.014 0.025 0.022 0.016 0.028 0.015 L3 0.021 0.023 0.023 0.018 0.016 0.015 H1 0.015 0.025 0.031 0.033 0.032 0.014 H1 0.023 0.026 0.035 0.037 0.019 0.023 H3 0.017 0.027 0.026 0.017 0.027 0.007 H3 0.023 0.027 0.030 0.018 0.018 0.007

L3-L1 -0.006 0.001 -.006 -0.019 -0.008 -0.005 L3-L1 -0.002 -0.014 -.010 -0.023 0.001 -0.011 H3-H1 0.002 0.002 -.005 -0.016 -0.005 -0.007 H3-H1 0.000 0.000 -.005 -0.019 0.000 -0.016

SIZE PB SIZE PB SL1 0.041 – 0.046 0.033 0.047 0.028 SL1 0.046 – 0.047 0.039 0.026 0.031 SL3 0.069 – 0.048 0.021 0.044 0.033 SL3 0.092 – 0.050 0.024 0.036 0.030 SH1 0.027 – 0.041 0.022 0.029 0.003 SH1 0.031 – 0.049 0.024 0.021 0.008 SH3 0.006 – 0.032 0.018 0.022 0.002 SH3 0.009 – 0.040 0.021 0.014 0.000 BL1 0.006 – 0.029 0.023 0.039 0.005 BL1 0.011 – 0.033 0.025 0.020 0.012 BL3 0.008 – 0.030 0.013 0.031 0.004 BL3 0.012 – 0.028 0.016 0.018 0.009 BH1 0.004 – 0.016 0.017 0.024 0.004 BH1 0.013 – 0.019 0.018 0.014 0.007 BH3 0.009 – 0.019 0.008 0.015 -0.008 BH3 0.009 – 0.020 0.010 0.008 -0.007

SL3-SL1 0.028 – 0.002 -0.012 -0.003 0.004 SL3-SL1 0.046 – 0.003 -0.015 0.009 -0.001 SH3-SH1 -0.021 – -.009 -0.004 -0.007 -0.001 SH3-SH1 -.022 – -.009 -0.003 -0.007 -0.007 BL3-BL1 0.002 – 0.001 -0.009 -0.008 -0.001 BL3-BL1 0.000 – -.005 -0.009 -0.001 -0.003 BH3-BH1 -0.002 – 0.003 0.002 -0.009 -0.011 BH3-BH1 -.006 – 0.001 0.002 -0.006 -0.013

SIZE PE SIZE PE SL1 0.030 – 0.053 0.032 0.037 0.028 SL1 0.033 – 0.060 0.038 0.021 0.028 SL3 0.071 – 0.052 0.019 0.042 0.025 SL3 0.092 – 0.059 0.023 0.029 0.027 SH1 0.034 – 0.036 0.025 0.040 0.008 SH1 0.039 – 0.039 0.026 0.026 0.013 SH3 0.010 – 0.026 0.018 0.023 0.001 SH3 0.018 – 0.031 0.021 0.018 -0.001 BL1 0.008 – 0.032 0.022 0.038 0.010 BL1 0.019 – 0.035 0.023 0.022 0.016 BL3 0.009 – 0.028 0.013 0.024 0.002 BL3 0.013 – 0.030 0.017 0.012 0.008 BH1 -0.003 – 0.015 0.019 0.024 0.001 BH1 0.004 – 0.019 0.022 0.013 0.003 BH3 0.004 – 0.018 0.008 0.020 -0.002 BH3 0.006 – 0.018 0.007 0.011 -0.004

SL3-SL1 0.041 – -.001 -0.013 0.004 -0.003 SL3-SL1 0.059 – -.001 -0.015 0.008 0.000 SH3-SH1 -0.024 – -.010 -0.007 -0.016 -0.008 SH3-SH1 -.021 – -.008 -0.005 -0.008 -0.014 BL3-BL1 0.001 – -.004 -0.009 -0.014 -0.008 BL3-BL1 -.006 – -.005 -0.006 -0.010 -0.009 BH3-BH1 -0.023 – 0.002 0.017 -0.004 -0.004 BH3-BH1 -.022 – -.002 0.022 -0.002 -0.007

The table shows mean excess returns on the stylized portfolios. We construct four sets of quintile portfolios based on portfolio formation windows of 24, 36, 48, and 60 months and estimate for 12 months holding period after skipping one year between portfolio formations and holding windows to control for any short-term prior-return effects.


64

Table 2: CAPM Model for Test Portfolios.

CAPM Results : RPt - RFt = α + β (RMt - RFt)+ et Panel A: Mean Excess Return on 24-12-12 Stylized Portfolios Regressed on the Excess Return on the Market Factor

BRAZIL RUSSIA INDIA CHINA S.KOREA S.AFRICA α t(α) 𝑅�2 α t(α) 𝑅�2 α t(α) 𝑅�2 α t(α) 𝑅�2 α t(α) 𝑅�2 α t(α) 𝑅�2 RETURN PORTFOLIOS

P1 0.01 1.70* 0.36 0.03 2.38** 0.50 0.03 2.59** 0.00 0.02 2.90** 0.54 0.02 1.97** 0.56 0.01 1.16 0.18 P5 0.03 2.05** 0.21 0.10 1.39 0.02 0.02 2.41** 0.00 0.01 1.58 0.74 0.01 1.22 0.65 0.01 0.97 0.29 SIZE

S1 0.01 1.73* 0.34 -.02 -1.11 0.23 0.02 2.25** 0.23 0.02 2.78** 0.48 0.02 2.26** 0.44 0.00 0.00 0.17 S3 0.03 1.84* 0.13 0.07 1.52 0.01 0.02 2.25** 0.27 0.01 2.00** 0.60 0.02 1.99** 0.49 0.01 0.89 0.23 B1 0.00 0.08 0.51 0.01 1.88* 0.62 0.01 1.53 0.35 0.01 2.32** 0.64 0.01 1.35 0.69 -.01 -1.58 0.34 B3 0.00 -0.49 0.57 0.01 2.21** 0.72 0.01 1.03 0.48 0.00 1.16 0.77 0.00 0.51 0.76 -.01 -2.4** 0.46 PB

L1 0.01 1.32 0.36 0.01 0.90 0.46 0.02 2.03** 0.27 0.01 1.36 0.31 0.02 2.02** 0.53 0.00 -0.06 0.25 L3 0.03 1.64* 0.16 0.01 0.67 0.48 0.02 1.77* 0.28 0.01 1.36 0.43 0.02 2.19** 0.58 0.00 0.43 0.35 H1 0.00 0.36 0.42 0.01 1.72* 0.45 0.01 1.77* 0.39 0.01 2.24** 0.58 0.00 0.72 0.66 -.01 -1.88* 0.30 H3 -.01 -1.15 0.58 0.01 1.66* 0.69 0.01 1.41 0.48 0.00 0.39 0.68 0.00 -0.28 0.72 -.01 -1.64* 0.35 PE

L1 0.01 1.66* 0.37 0.01 1.63 0.58 0.02 2.49** 0.26 0.02 2.39** 0.41 0.02 1.86* 0.55 0.00 0.37 0.23 L3 0.02 1.28 0.17 0.02 0.94 0.36 0.02 1.77* 0.32 0.01 1.55 0.53 0.01 1.30 0.59 0.00 0.24 0.35 H1 0.00 0.17 0.47 0.01 0.58 0.32 0.01 0.90 0.36 0.01 1.70* 0.54 0.01 1.32 0.65 -.01 -2.0** 0.28 H3 0.00 -0.14 0.51 0.01 0.52 0.19 0.01 1.29 0.46 0.00 0.09 0.64 0.00 0.62 0.74 -.01 -1.94* 0.33 DYIELD

L1 0.00 0.42 0.44 0.01 1.01 0.48 0.01 1.09 0.40 0.01 1.59 0.62 0.01 1.15 0.69 -.01 -2.4** 0.31 L3 -.01 -0.86 0.54 0.01 0.85 0.64 0.01 1.24 0.49 0.00 0.07 0.73 0.00 0.63 0.73 -.01 -2.1** 0.35 H1 0.01 0.74 0.48 0.01 1.07 0.41 0.02 2.75** 0.28 0.02 2.15** 0.42 0.01 1.90* 0.61 0.00 -0.08 0.24 H3 0.00 0.40 0.51 0.02 2.63** 0.49 0.01 1.76* 0.39 0.02 2.51** 0.48 0.01 1.53 0.64 0.00 -0.21 0.31 SALES

L1 0.02 1.88* 0.39 0.02 1.66* 0.27 0.01 1.10 0.28 0.02 2.17 0.41 0.01 1.28 0.58 0.01 0.78 0.16 L3 0.00 0.29 0.44 0.01 1.31 0.46 0.01 1.37 0.32 0.01 0.79 0.48 0.01 1.07 0.64 0.01 1.24 0.26 H1 0.01 0.68 0.47 0.00 0.32 0.57 0.02 2.72** 0.35 0.02 2.11** 0.41 0.01 1.12 0.56 0.00 0.17 0.31 H3 0.00 0.17 0.46 0.02 1.98** 0.39 0.01 1.65* 0.35 0.01 1.08 0.60 0.01 1.20 0.77 0.00 -0.32 0.36 SIZE PB

SL1 0.02 1.96** 0.26 – – – 0.03 2.21** 0.19 0.02 1.49 0.27 0.03 2.02** 0.35 0.02 1.64* 0.05 SL3 0.06 1.83* 0.05 – – – 0.03 1.83* 0.10 0.01 1.01 0.40 0.02 2.20** 0.42 0.03 2.82** 0.07 SH1 0.02 2.15** 0.28 – – – 0.02 2.02** 0.23 0.01 1.02 0.51 0.01 1.24 0.50 0.00 -0.47 0.05 SH3 -.01 -0.59 0.26 – – – 0.02 1.71* 0.32 0.01 0.88 0.49 0.00 0.49 0.51 0.00 -0.52 0.14 BL1 0.00 -0.01 0.44 – – – 0.01 1.37 0.30 0.01 1.78* 0.59 0.01 2.02** 0.68 -.01 -1.07 0.30 BL3 0.00 -0.44 0.53 – – – 0.01 0.85 0.41 0.00 1.01 0.69 0.01 1.28 0.71 -.01 -1.45 0.42 BH1 -.01 -1.13 0.49 – – – 0.00 0.34 0.36 0.00 0.80 0.63 0.01 0.95 0.73 -.02 -2.6** 0.27 BH3 -.01 -0.80 0.55 – – – 0.01 1.34 0.50 0.00 -0.44 0.72 0.00 -0.24 0.77 -.02 -2.9** 0.30

SIZE PE SL1 0.01 1.16 0.21 – – – 0.03 2.46** 0.18 0.01 1.51 0.30 0.02 1.72* 0.39 0.02 1.91* 0.06 SL3 0.05 1.57 0.06 – – – 0.03 1.79* 0.10 0.00 0.48 0.47 0.02 1.69* 0.41 0.02 2.19** 0.07 SH1 0.02 1.77* 0.35 – – – 0.02 1.70* 0.24 0.01 1.11 0.50 0.02 2.12** 0.44 0.00 -0.32 0.12 SH3 0.01 0.93 0.14 – – – 0.01 1.54 0.31 0.00 0.65 0.46 0.01 1.62 0.56 -.01 -0.83 0.11 BL1 0.00 0.59 0.50 – – – 0.02 2.22** 0.29 0.01 1.38 0.61 0.01 1.85* 0.66 0.00 -0.84 0.34 BL3 0.00 -0.17 0.53 – – – 0.01 1.03 0.44 0.01 1.77* 0.78 0.00 0.47 0.71 -.01 -1.59 0.37 BH1 -.01 -1.12 0.54 – – – 0.00 0.73 0.35 0.01 1.62 0.59 0.00 0.75 0.70 -.02 -2.4** 0.18 BH3 -.01 -0.74 0.57 – – – 0.01 1.18 0.50 0.00 -1.09 0.65 0.00 0.63 0.78 -.01 -1.66* 0.27


65

Table 2: Continued

CAPM Results : RPt - RFt = α + β (RMt - RFt)+ et Panel B: Mean Excess Return on 36-12-12 Stylized Portfolios Regressed on the Excess Return on the Market Factor


P1 0.02 1.99** 0.39 0.03 2.46** 0.45 0.03 3.11** 0.00 0.02 2.45** 0.50 0.02 1.61 0.55 0.01 0.81 0.17 P5 0.03 1.97** 0.18 0.05 1.76* 0.19 0.03 2.59** 0.00 0.01 1.89* 0.72 0.01 0.99 0.65 0.01 1.00 0.28 SIZE

S1 0.02 2.54** 0.36 0.01 0.75 0.45 0.02 2.29** 0.25 0.02 2.53** 0.45 0.02 1.75* 0.43 0.00 0.35 0.15 S3 0.03 1.55 0.11 0.04 1.67* 0.03 0.02 2.25** 0.26 0.01 2.21** 0.58 0.01 1.55 0.48 0.00 0.65 0.23 B1 0.00 0.04 0.51 0.01 1.65* 0.64 0.01 1.77* 0.37 0.01 1.79* 0.65 0.01 1.01 0.66 -.01 -1.02 0.35 B3 0.00 -0.12 0.56 0.01 1.89* 0.72 0.01 1.06 0.47 0.00 1.07 0.77 0.00 0.41 0.77 -.01 -2.4** 0.44 PB

L1 0.02 2.27** 0.41 0.01 0.93 0.45 0.02 2.09** 0.28 0.01 1.30 0.36 0.02 1.96** 0.52 0.01 0.98 0.19 L3 0.03 1.46 0.13 0.01 0.59 0.42 0.02 1.71* 0.27 0.00 0.64 0.55 0.02 2.23** 0.59 0.00 0.45 0.37 H1 0.01 0.77 0.42 0.01 2.08** 0.52 0.01 1.69* 0.43 0.01 1.50 0.55 0.00 0.54 0.64 -.01 -1.9** 0.31 H3 0.00 -0.61 0.55 0.00 0.73 0.63 0.01 1.28 0.48 0.00 0.52 0.70 0.00 -0.16 0.74 -.01 -1.30 0.34 PE

L1 0.02 1.84* 0.39 0.01 2.23** 0.63 0.02 2.36** 0.28 0.02 1.81* 0.40 0.02 1.58 0.52 0.00 0.69 0.22 L3 0.03 1.33 0.14 0.01 0.69 0.35 0.02 2.23** 0.29 0.00 0.90 0.60 0.01 1.36 0.58 0.00 0.77 0.37 H1 0.01 0.93 0.49 0.01 0.54 0.31 0.01 1.54 0.39 0.01 1.38 0.54 0.01 1.28 0.63 -.01 -1.47 0.28 H3 0.00 -0.18 0.52 0.00 -0.68 0.68 0.01 0.97 0.48 0.00 0.14 0.62 0.00 0.58 0.74 -.01 -1.56 0.34 DYIELD

L1 0.00 0.56 0.46 0.02 1.99** 0.45 0.01 1.28 0.42 0.01 1.30 0.60 0.00 0.67 0.67 -.01 -1.75* 0.33 L3 0.00 -0.17 0.51 0.00 0.32 0.59 0.01 0.85 0.50 0.00 0.42 0.72 0.00 0.63 0.75 -.01 -1.78* 0.31 H1 0.01 1.35 0.46 0.01 1.03 0.46 0.02 2.58** 0.30 0.01 1.46 0.41 0.01 1.92* 0.62 0.00 0.61 0.21 H3 0.00 0.62 0.50 0.03 3.24** 0.46 0.02 2.30** 0.37 0.01 1.84* 0.62 0.01 1.22 0.63 0.00 -0.32 0.34 SALES

L1 0.02 1.74* 0.40 0.01 1.32 0.24 0.01 1.49 0.28 0.02 1.97** 0.34 0.01 1.03 0.55 0.01 0.86 0.15 L3 0.01 0.67 0.41 0.01 0.76 0.31 0.01 1.30 0.33 0.00 0.71 0.57 0.01 1.15 0.67 0.00 0.45 0.25 H1 0.01 0.69 0.41 0.00 0.58 0.70 0.03 3.16** -.01 0.02 1.88* 0.39 0.01 1.17 0.55 0.00 0.68 0.30 H3 0.00 -0.04 0.41 0.00 0.40 0.38 0.01 1.11 0.32 0.01 1.31 0.64 0.01 1.54 0.80 0.00 -0.46 0.36 SIZE PB

SL1 0.04 3.08** 0.26 – – – 0.04 2.80** -.01 0.02 1.75* 0.27 0.03 1.84* 0.36 0.02 2.28** 0.05 SL3 0.05 1.46 0.05 – – – 0.05 2.86** 0.03 0.01 0.67 0.47 0.03 2.33** 0.39 0.02 2.59** 0.11 SH1 0.02 1.67* 0.25 – – – 0.04 2.77** -.01 0.01 0.79 0.53 0.01 1.14 0.47 -.01 -0.72 0.05 SH3 0.00 -0.41 0.28 – – – 0.03 2.58** -.01 0.01 0.75 0.48 0.00 0.16 0.52 -.01 -0.67 0.12 BL1 0.00 0.13 0.44 – – – 0.03 2.50** -.01 0.01 1.78* 0.59 0.01 1.88* 0.67 0.00 -0.61 0.29 BL3 0.01 0.74 0.50 – – – 0.02 2.11** -.01 0.00 1.08 0.69 0.01 1.77* 0.74 -.01 -1.63 0.38 BH1 0.00 -0.09 0.43 – – – 0.02 2.51** -.01 0.01 1.19 0.59 0.00 0.52 0.71 -.01 -1.71* 0.21 BH3 -.01 -0.59 0.53 – – – 0.01 1.66* -.01 0.00 -0.07 0.69 0.00 -0.31 0.79 -.02 -2.5** 0.31

SIZE PE SL1 0.02 1.82* 0.22 – – – 0.05 3.10** -.01 0.02 1.73* 0.28 0.02 1.12 0.37 0.02 2.46** 0.05 SL3 0.06 1.53 0.04 – – – 0.05 2.58** 0.03 0.01 0.91 0.48 0.02 1.62 0.42 0.02 2.06** 0.14 SH1 0.01 1.24 0.31 – – – 0.03 2.53** -.01 0.01 0.89 0.49 0.02 1.95* 0.45 0.00 0.01 0.10 SH3 0.01 0.55 0.20 – – – 0.02 2.07** -.01 0.01 0.83 0.49 0.01 1.04 0.52 -.01 -0.84 0.10 BL1 0.00 0.53 0.50 – – – 0.03 2.82** -.01 0.01 2.01** 0.61 0.01 1.42 0.62 0.00 0.04 0.35 BL3 0.00 0.25 0.44 – – – 0.02 2.15** -.01 0.01 1.50 0.75 0.00 0.58 0.73 -.01 -1.75* 0.37 BH1 -.01 -0.78 0.47 – – – 0.02 1.89** -.01 0.01 1.33 0.59 0.00 0.67 0.67 -.01 -1.86* 0.19 BH3 0.00 -0.13 0.60 – – – 0.01 1.56 -.01 0.00 -0.71 0.65 0.00 0.48 0.78 -.02 -2.3** 0.31


66

Table 2: Continued

CAPM Results : RPt - RFt = α + β (RMt - RFt)+ et Panel C: Mean Excess Return on 48-12-12 Stylized Portfolios Regressed on the Excess Return on the Market Factor


P1 0.03 2.87** 0.39 0.03 2.28** 0.33 0.03 3.12** 3.11** 0.02 2.51** 0.48 0.02 1.99** 0.52 0.01 1.24 0.15 P5 0.04 2.12** 0.18 0.05 2.02** 0.31 0.03 2.81** 2,59** 0.00 0.97 0.72 0.01 1.68* 0.66 0.00 -0.31 0.36 SIZE

S1 0.03 3.21** 0.36 0.01 0.89 0.26 0.03 2.33** 0.24 0.02 2.43** 0.43 0.02 1.97** 0.41 0.01 0.72 0.15 S3 0.03 1.77* 0.10 0.04 1.45 0.03 0.03 2.57** 0.26 0.01 2.02** 0.54 0.02 2.34** 0.46 0.00 0.07 0.26 B1 0.00 0.20 0.49 0.01 2.02** 0.69 0.01 2.05** 0.36 0.01 1.77* 0.64 0.02 2.19** 0.69 -.01 -0.93 0.33 B3 0.00 0.04 0.57 0.01 2.16** 0.64 0.01 1.71* 0.48 0.00 0.61 0.76 0.01 1.00 0.78 -.02 -2.6** 0.42 PB

L1 0.02 2.69** 0.39 0.00 0.04 0.56 0.02 2.37** 0.28 0.02 1.92* 0.40 0.03 2.40** 0.48 0.01 0.93 0.17 L3 0.03 1.69* 0.13 0.00 0.48 0.45 0.02 2.13** 0.27 0.01 1.11 0.57 0.02 3.06** 0.59 0.00 0.47 0.38 H1 0.01 0.64 0.42 0.01 2.18** 0.53 0.01 1.65* 0.40 0.01 1.57 0.56 0.01 1.42 0.66 -.01 -1.54 0.31 H3 0.00 0.19 0.56 0.00 0.75 0.62 0.01 1.58 0.49 0.00 0.27 0.65 0.00 0.46 0.73 -.01 -1.69* 0.31 PE

L1 0.02 1.77* 0.36 0.02 2.04** 0.55 0.03 2.67** 0.28 0.02 2.24** 0.43 0.02 2.02** 0.50 0.00 0.63 0.21 L3 0.03 1.77* 0.16 0.02 1.83* 0.51 0.03 2.89** 0.30 0.00 0.72 0.62 0.02 2.35** 0.60 0.00 0.51 0.37 H1 0.01 1.25 0.49 0.01 0.70 0.30 0.01 1.58 0.38 0.01 1.53 0.56 0.02 2.23** 0.65 -.01 -1.41 0.27 H3 0.00 0.04 0.55 0.00 0.07 0.59 0.01 1.20 0.49 0.00 0.30 0.62 0.01 1.05 0.74 -.02 -2.2** 0.35 DYIELD

L1 0.01 1.02 0.44 0.02 2.35** 0.47 0.01 1.67* 0.42 0.01 1.40 0.56 0.01 1.86* 0.69 -.01 -1.45 0.31 L3 0.00 0.23 0.50 0.00 -0.51 0.61 0.01 1.25 0.51 0.00 0.12 0.67 0.01 1.28 0.76 -.01 -2.0** 0.35 H1 0.01 1.14 0.47 0.02 2.28** 0.46 0.02 3.13** 0.28 0.02 1.87* 0.42 0.02 3.03** 0.63 0.01 0.85 0.20 H3 0.01 1.46 0.52 0.02 1.77* 0.31 0.02 2.95** 0.38 0.01 1.42 0.62 0.01 2.28** 0.67 0.00 -0.07 0.33 SALES

L1 0.02 2.04** 0.41 0.00 0.59 0.52 0.02 1.81* 0.29 0.02 2.53** 0.41 0.02 1.56 0.53 0.01 1.34 0.11 L3 0.01 1.19 0.49 0.01 0.83 0.23 0.01 1.44 0.35 0.01 0.99 0.57 0.01 1.85* 0.69 0.00 0.25 0.37 H1 0.01 1.43 0.48 0.00 0.29 0.53 0.02 2.67** 0.36 0.02 2.44** 0.41 0.02 1.69* 0.53 0.00 0.26 0.27 H3 0.01 1.26 0.37 0.01 0.89 0.37 0.02 1.82* 0.33 0.01 1.24 0.62 0.01 2.10** 0.79 -.01 -0.99 0.34 SIZE PB

SL1 0.04 3.38** 0.26 – – – 0.03 2.22** 0.18 0.02 2.04** 0.32 0.03 1.96** 0.34 0.02 2.29** 0.03 SL3 0.07 1.63 0.06 – – – 0.04 2.26** 0.08 0.01 1.19 0.47 0.03 3.10** 0.40 0.02 2.65** 0.12 SH1 0.02 2.36** 0.27 – – – 0.03 2.02** 0.20 0.01 1.09 0.50 0.01 1.26 0.48 -.01 -0.66 0.08 SH3 0.00 0.46 0.25 – – – 0.02 2.12** 0.32 0.01 0.90 0.48 0.01 1.11 0.50 -.01 -1.04 0.13 BL1 0.00 0.25 0.43 – – – 0.02 2.19** 0.34 0.01 1.96** 0.58 0.02 2.34** 0.60 -.01 -0.88 0.27 BL3 0.00 0.56 0.50 – – – 0.02 2.47** 0.40 0.00 0.56 0.67 0.01 1.96** 0.73 -.01 -1.54 0.40 BH1 0.00 0.06 0.44 – – – 0.01 1.37 0.41 0.01 1.16 0.63 0.01 1.25 0.71 -.01 -1.20 0.29 BH3 0.00 0.52 0.51 – – – 0.01 1.24 0.51 0.00 -0.41 0.67 0.00 -0.05 0.79 -.02 -2.9** 0.31

SIZE PE SL1 0.03 2.25** 0.17 – – – 0.04 2.58** 0.22 0.02 2.06** 0.32 0.02 1.24 0.31 0.02 2.12** 0.04 SL3 0.07 1.69* 0.05 – – – 0.04 2.22** 0.07 0.01 1.37 0.51 0.03 2.38** 0.44 0.01 1.64* 0.13 SH1 0.03 3.22** 0.33 – – – 0.03 2.25** 0.21 0.01 1.37 0.51 0.03 2.46** 0.47 0.00 -0.17 0.11 SH3 0.01 0.74 0.20 – – – 0.01 1.47 0.33 0.01 1.01 0.47 0.01 1.58 0.50 -.01 -1.21 0.12 BL1 0.00 0.49 0.47 – – – 0.02 2.58** 0.33 0.01 2.10** 0.58 0.02 2.18** 0.63 .00 -0.38 0.32 BL3 0.01 0.59 0.49 – – – 0.02 2.23** 0.41 0.00 0.69 0.73 0.01 1.27 0.72 -.01 -1.91* 0.39 BH1 -.01 -0.82 0.48 – – – 0.01 1.00 0.36 0.01 1.27 0.61 0.01 1.47 0.67 -.01 -1.24 0.15 BH3 0.00 0.03 0.59 – – – 0.01 1.08 0.52 0.00 -0.57 0.65 0.00 0.74 0.78 -.02 -2.0** 0.28


67

Table 2: Continued

CAPM Results : RPt - RFt = α + β (RMt - RFt)+ et Panel D: Mean Excess Return on 60-12-12 Stylized Portfolios Regressed on the Excess Return on the Market Factor


P1 0.02 2.19** 0.45 0.03 2.01** 0.44 0.04 2.98** 0.01 0.02 2.36** 0.47 0.01 2.04** 0.48 0.01 1.93* 0.06 P5 0.04 2.05** 0.18 0.04 1.23 0.20 0.03 2.76** 0.00 0.00 0.54 0.68 0.01 1.26 0.58 0.00 -0.53 0.27 SIZE

S1 0.03 3.01** 0.37 0.01 0.92 0.26 0.03 2.27** 0.22 0.02 2.22** 0.42 0.02 2.13** 0.37 0.01 1.37 0.07 S3 0.04 2.09** 0.12 0.04 1.50 0.03 0.03 2.41** 0.25 0.01 1.28 0.57 0.02 2.05** 0.32 0.00 0.30 0.12 B1 0.00 0.20 0.51 0.02 2.50** 0.67 0.02 2.06** 0.33 0.01 1.76* 0.63 0.01 1.68* 0.63 0.00 0.17 0.23 B3 0.00 -0.25 0.52 0.01 2.06** 0.62 0.01 1.29 0.47 0.00 0.27 0.74 0.00 0.45 0.74 -.01 -2.1** 0.25 PB

L1 0.02 2.87** 0.45 0.00 0.40 0.53 0.03 2.29** 0.24 0.02 2.40** 0.47 0.02 2.31** 0.42 0.01 1.86* 0.08 L3 0.04 1.84* 0.13 0.00 0.52 0.44 0.02 2.08** 0.24 0.01 1.12 0.63 0.02 2.80** 0.44 0.01 1.16 0.21 H1 0.01 0.91 0.41 0.01 1.89* 0.48 0.01 1.97** 0.38 0.01 1.30 0.54 0.01 1.18 0.58 0.00 -0.40 0.19 H3 0.00 -0.01 0.55 0.00 0.67 0.59 0.01 1.27 0.48 0.00 0.17 0.65 0.00 0.13 0.67 -.02 -2.1** 0.18 PE

L1 0.02 1.65* 0.39 0.01 1.66* 0.62 0.03 2.64** 0.26 0.02 2.38** 0.48 0.02 1.98** 0.46 0.01 1.51 0.11 L3 0.04 1.81* 0.17 0.01 1.61 0.50 0.03 2.59** 0.26 0.01 1.29 0.67 0.01 1.84* 0.49 0.00 0.59 0.25 H1 0.01 1.02 0.50 0.01 0.83 0.24 0.01 1.67* 0.35 0.01 1.59 0.54 0.01 2.01** 0.60 0.00 -0.31 0.15 H3 0.00 -0.01 0.47 0.00 0.18 0.50 0.01 0.90 0.48 0.00 -0.04 0.62 0.00 0.76 0.67 -.02 -2.3** 0.17 DYIELD

L1 0.01 1.11 0.48 0.02 1.81* 0.49 0.01 2.00** 0.39 0.01 1.62 0.59 0.01 1.45 0.59 0.00 -0.46 0.18 L3 0.00 -0.41 0.50 0.00 0.02 0.54 0.01 0.75 0.50 0.00 0.16 0.66 0.00 1.01 0.73 -.01 -1.21 0.18 H1 0.01 1.16 0.45 0.01 1.24 0.53 0.02 2.80** 0.23 0.02 2.58** 0.51 0.02 3.03** 0.54 0.01 1.90* 0.13 H3 0.01 1.55 0.42 0.02 1.65* 0.31 0.02 2.69** 0.38 0.01 1.87* 0.68 0.01 1.60 0.54 0.01 0.76 0.15 SALES

L1 0.01 1.76* 0.49 0.02 1.98** 0.41 0.02 1.96** 0.24 0.03 2.55** 0.38 0.01 1.02 0.49 0.02 2.10** 0.03 L3 0.01 1.38 0.44 0.01 0.99 0.21 0.01 1.32 0.29 0.01 0.94 0.62 0.01 1.41 0.62 0.00 0.56 0.22 H1 0.01 1.59 0.46 0.00 0.45 0.62 0.02 2.52** 0.30 0.02 2.51** 0.41 0.01 1.60 0.53 0.01 1.99** 0.15 H3 0.01 1.18 0.36 0.01 0.54 0.42 0.02 1.89* 0.34 0.01 1.01 0.64 0.01 1.54 0.76 0.00 -0.61 0.17 SIZE PB

SL1 0.04 3.07** 0.24 – – – 0.03 2.02** 0.17 0.03 2.41** 0.38 0.02 1.94* 0.23 0.02 2.59** 0.03 SL3 0.08 1.86* 0.05 – – – 0.04 2.11** 0.07 0.01 1.18 0.51 0.03 3.01** 0.25 0.02 2.19** 0.16 SH1 0.02 2.28** 0.32 – – – 0.03 2.42** 0.20 0.01 1.17 0.50 0.01 1.70* 0.42 0.00 -0.27 0.09 SH3 0.00 0.27 0.33 – – – 0.03 2.37** 0.28 0.01 1.17 0.53 0.01 0.88 0.31 -.01 -1.13 0.10 BL1 0.00 0.29 0.49 – – – 0.02 2.28** 0.30 0.01 1.87* 0.60 0.01 1.85* 0.60 0.00 0.25 0.19 BL3 0.00 0.29 0.48 – – – 0.02 1.92* 0.36 0.00 0.80 0.70 0.01 1.44 0.68 0.00 -0.61 0.25 BH1 0.00 0.47 0.44 – – – 0.01 1.42 0.33 0.01 1.01 0.63 0.01 0.89 0.61 -.01 -0.77 0.22 BH3 0.00 -0.11 0.51 – – – 0.01 0.96 0.50 0.00 -0.37 0.68 0.00 -0.37 0.78 -.02 -2.4** 0.19

SIZE PE SL1 0.02 1.92* 0.28 – – – 0.04 2.49** 0.21 0.03 2.44** 0.37 0.01 1.30 0.24 0.02 2.26** 0.03 SL3 0.08 1.87* 0.05 – – – 0.05 2.34** 0.05 0.01 1.28 0.54 0.02 2.23** 0.24 0.02 1.80* 0.13 SH1 0.03 3.31** 0.42 – – – 0.03 2.22** 0.18 0.01 1.17 0.52 0.02 2.60** 0.42 0.00 0.38 0.11 SH3 0.01 1.06 0.22 – – – 0.02 1.57 0.37 0.01 1.18 0.52 0.01 1.46 0.32 -.01 -1.35 0.11 BL1 0.01 1.03 0.01 – – – 0.02 2.48** 6.68 0.01 1.87* 0.58 0.01 2.09** 0.63 0.00 0.65 0.24 BL3 0.00 0.33 0.49 – – – 0.02 2.09** 8.34 0.00 1.30 0.76 0.00 0.47 0.66 -.01 -0.90 0.27 BH1 0.00 -0.49 0.47 – – – 0.01 1.22 1.66 0.01 1.28 0.62 0.01 0.88 0.60 -.01 -1.11 0.18 BH3 0.00 -0.43 0.56 – – – 0.00 0.69 1.07 0.00 -0.82 0.62 0.00 0.21 0.76 -.02 -2.2** 0.20

This table reports Capital Asset Pricing model (CAPM) results. Excess returns on sample portfolios are regressed on excess returns of the market factor in the CAPM framework. Alpha (α) is a measure of extra normal performance whereas adjusted R-square is the goodness of fit measure. ** t -statistics are tested for significance at 5% level on 2-tail basis, *t-statistics are tested for significance at 10% level on 2-tail basis.


68

Table 3: Fama French model for Test Portfolios.

FF Three Factor Model Results : RPt - RFt = α + β (RMt - RFt)+ s SMBt + l LMHt + et Panel A: Excess Return on 24-12-12 Stylized Portfolios Regressed on the Excess Return on the Market (RM-RF) Factor

and Two Proxy Portfolios that Relate to Size (SMB) and (LMH) Factors BRAZIL RUSSIA INDIA CHINA S.KOREA S.AFRICA α t(α) 𝑅�2 α t(α) 𝑅�2 α t(α) 𝑅�2 α t(α) 𝑅�2 α t(α) 𝑅�2 α t(α) 𝑅�2 RETURN PORTFOLIOS

P1 0.01 0.64 0.42 0.01 0.85 0.40 0.03 2.60** -.01 0.02 2.79** 0.56 0.00 0.08 0.79 0.00 -0.49 0.24 P5 0.00 -.14 0.55 0.12 0.95 -.03 0.02 2.42** -.01 0.01 1.53 0.77 0.00 -.13 0.78 0.00 0.61 0.35 SIZE

S1 0.00 0.26 0.49 0.00 -0.12 0.26 0.01 1.22 0.72 0.02 2.63** 0.55 0.00 0.50 0.83 -.01 -1.95* 0.29 S3 -.01 -.87 0.61 0.08 1.16 -.01 0.01 1.24 0.73 0.01 1.89* 0.65 0.00 0.51 0.79 0.00 -0.09 0.30 B1 0.00 -.37 0.51 0.01 1.61 0.47 0.00 0.38 0.59 0.01 2.33** 0.64 0.00 -.22 0.76 -.01 -1.60 0.34 B3 -.01 -.91 0.58 0.01 2.00** 0.68 0.00 0.20 0.57 0.00 1.14 0.80 0.00 -.16 0.77 -.01 -1.41 0.50 PB

L1 0.00 -.14 0.49 0.00 0.34 0.37 0.00 0.79 0.72 0.00 0.83 0.76 0.00 -.63 0.84 -.01 -2.2** 0.35 L3 -.01 -1.3 0.66 0.01 1.54 0.48 0.00 0.35 0.72 0.00 0.92 0.80 0.00 0.28 0.79 0.00 -0.48 0.36 H1 0.00 -.16 0.46 0.02 2.15** 0.42 0.00 0.94 0.51 0.01 2.43** 0.64 0.00 -.23 0.72 -.01 -1.37 0.33 H3 -.01 -1.5 0.61 0.01 1.66* 0.65 0.00 0.84 0.53 0.00 0.35 0.72 0.00 -.86 0.77 -.01 -0.75 0.46 PE

L1 0.00 0.48 0.46 0.02 1.90* 0.53 0.01 1.54 0.69 0.01 2.29** 0.71 0.00 -.47 0.82 -.01 -1.57 0.34 L3 -.02 -1.7* 0.65 0.03 1.26 0.36 0.00 0.49 0.67 0.00 1.30 0.81 0.00 -.65 0.79 0.00 -0.36 0.37 H1 -.01 -.78 0.52 0.01 1.13 0.39 0.00 -0.19 0.54 0.01 1.66* 0.56 0.00 -.07 0.74 -.01 -1.79* 0.27 H3 -.01 -0.7 0.53 0.02 2.05** 0.15 0.00 0.65 0.52 0.00 -0.06 0.69 0.00 -.05 0.77 -.01 -1.53 0.43 DYIELD

L1 0.00 -.53 0.49 0.01 0.74 0.47 0.00 0.17 0.53 0.01 1.51 0.66 0.00 0.46 0.72 -.01 -2.1** 0.30 L3 -.01 -1.1 0.54 0.01 0.76 0.60 0.00 0.65 0.54 0.00 -0.06 0.76 0.00 0.18 0.75 -.01 -1.55 0.38 H1 0.00 0.13 0.50 0.01 1.28 0.35 0.01 1.92* 0.55 0.01 1.86* 0.61 0.00 0.37 0.71 0.00 -0.74 0.25 H3 0.00 -.21 0.53 0.01 1.59 0.51 0.00 0.80 0.58 0.01 2.62** 0.74 0.00 0.50 0.69 0.00 -0.27 0.30 SALES

L1 0.01 0.82 0.48 0.02 1.53 0.25 0.00 -0.76 0.66 0.01 1.90* 0.62 0.00 -.59 0.76 0.00 -0.42 0.20 L3 -.01 -.93 0.50 0.02 1.45 0.46 0.00 0.10 0.54 0.00 0.18 0.69 0.00 -.57 0.78 0.01 0.65 0.27 H1 0.00 -.27 0.51 0.00 0.12 0.56 0.01 1.63 0.62 0.01 1.95* 0.70 -.01 -1.1 0.77 0.00 0.00 0.30 H3 0.00 -.31 0.48 0.01 1.55 0.47 0.00 0.29 0.60 0.00 0.70 0.73 0.00 0.42 0.78 0.00 -0.02 0.39 SIZE PB

SL1 0.00 0.43 0.42 – – – 0.01 1.07 0.70 0.01 1.12 0.77 -.01 -1.0 0.86 -.01 -1.37 0.27 SL3 -.03 -1.3 0.67 – – – 0.00 0.61 0.77 0.00 0.27 0.82 0.00 0.14 0.77 0.01 0.61 0.16 SH1 0.01 1.43 0.44 – – – 0.01 0.77 0.66 0.00 0.84 0.73 0.00 0.03 0.75 -.01 -0.72 0.18 SH3 -.01 -1.4 0.39 – – – 0.00 0.46 0.63 0.00 0.53 0.69 0.00 -.57 0.72 0.00 -0.12 0.33 BL1 0.00 -0.5* 0.45 – – – 0.00 0.38 0.58 0.01 1.64* 0.62 0.00 0.08 0.80 -.01 -1.01 0.31 BL3 -.01 -1.1 0.54 – – – 0.00 -0.13 0.59 0.00 0.76 0.72 0.00 -.27 0.77 -.01 -1.02 0.41 BH1 -.01 -1.7* 0.53 – – – 0.00 -0.34 0.44 0.00 0.67 0.68 0.00 0.07 0.76 -.01 -2.0** 0.26 BH3 -.01 -.93 0.57 – – – 0.01 0.97 0.51 0.00 -0.58 0.74 0.00 -.51 0.78 -.01 -2.1** 0.35

SIZE PE SL1 0.00 -.22 0.34 – – – 0.01 1.39 0.70 0.01 1.15 0.77 0.00 -.68 0.82 -.01 -1.03 0.29 SL3 -.03 -1.5 0.65 – – – 0.01 0.55 0.69 0.00 -0.27 0.77 0.00 -.14 0.78 0.01 0.54 0.15 SH1 0.00 0.52 0.47 – – – 0.00 0.32 0.65 0.00 0.84 0.76 0.00 0.28 0.75 -.01 -1.24 0.14 SH3 0.00 -.09 0.21 – – – 0.00 0.33 0.58 0.00 0.23 0.69 0.00 0.73 0.70 -.01 -0.81 0.23 BL1 0.00 0.00 0.51 – – – 0.01 1.49 0.56 0.01 1.24 0.64 0.00 0.07 0.77 -.01 -0.95 0.35 BL3 -.01 -.86 0.55 – – – 0.00 0.27 0.54 0.01 1.63 0.78 0.00 -.75 0.75 0.00 -0.59 0.38 BH1 -.01 -1.3 0.54 – – – 0.00 -0.02 0.47 0.01 1.51 0.63 0.00 -.21 0.73 -.02 -2.4** 0.17 BH3 -.01 -.93 0.58 – – – 0.00 0.68 0.52 0.00 -1.45 0.71 0.00 0.33 0.79 -.01 -1.02 0.31


69

Table 3: Continued

FF Three Factor Model Results : RPt - RFt = α + β (RMt - RFt)+ s SMBt + l LMHt + et Panel B: Excess Return on 36-12-12 Stylized Portfolios Regressed on the Excess Return on the Market (RM-RF) Factor

and Two Proxy Portfolios that Relate to Size (SMB) and (LMH) Factors BRAZIL RUSSIA INDIA CHINA S.KOREA S.AFRICA α t(α) 𝑅�2 α t(α)

α t(α)

α t(α)

α t(α)

α t(α)

RETURN PORTFOLIOS P1 0.01 1.03 0.44 0.02 1.69* 0.57 0.03 3.21 0.00 0.01 2.25** 0.62 0.00 -.37 0.78 -.01 -0.73 0.23 P5 0.00 -0.30 0.56 0.05 1.02 0.08 0.03 2.66 0.00 0.01 1.75* 0.76 0.00 -.23 0.78 0.01 0.96 0.35 SIZE

S1 0.01 1.10 0.49 0.01 0.97 0.55 0.00 0.77 0.74 0.01 2.39** 0.63 0.00 -.10 0.80 -.01 -1.6* 0.30 S3 -.02 -1.58 0.65 0.04 1.19 0.04 0.00 0.77 0.72 0.01 2.09** 0.67 0.00 0.21 0.78 0.00 0.46 0.33 B1 0.00 -0.36 0.51 0.01 1.78* 0.54 0.00 0.61 0.60 0.01 1.71* 0.68 0.00 -.61 0.75 -.01 -1.02 0.35 B3 0.00 -0.53 0.58 0.01 1.24 0.72 0.00 0.18 0.56 0.00 0.97 0.81 0.00 -.17 0.79 -.01 -1.31 0.48 PB

L1 0.01 0.75 0.54 0.01 1.03 0.44 0.00 0.58 0.73 0.00 0.79 0.79 0.00 -.38 0.81 -.01 -1.05 0.31 L3 -.02 -1.6* 0.66 0.01 1.11 0.40 0.00 0.01 0.73 0.00 -0.15 0.80 0.00 0.76 0.77 0.00 -0.51 0.39 H1 0.00 0.33 0.46 0.02 2.50** 0.51 0.00 0.77 0.53 0.01 1.47 0.67 0.00 -.45 0.71 -.01 -1.54 0.33 H3 -.01 -1.18 0.59 0.01 1.15 0.56 0.00 0.61 0.52 0.00 0.36 0.75 0.00 -.80 0.79 0.00 -0.59 0.45 PE

L1 0.01 0.72 0.45 0.02 2.97** 0.60 0.01 0.99 0.72 0.01 1.61 0.77 0.00 -.62 0.77 -.01 -1.27 0.33 L3 -.02 -1.8* 0.66 0.02 0.82 0.34 0.01 0.98 0.66 0.00 0.35 0.76 0.00 -.19 0.78 0.00 0.22 0.40 H1 0.00 -0.17 0.55 0.00 0.38 0.41 0.00 0.48 0.55 0.01 1.20 0.63 0.00 -.21 0.73 -.01 -1.41 0.27 H3 -.01 -0.70 0.54 0.00 0.01 0.55 0.00 0.20 0.54 0.00 -0.19 0.71 0.00 0.10 0.77 -.01 -1.08 0.42 DYIELD

L1 0.00 -0.51 0.51 0.02 1.64* 0.44 0.00 0.39 0.53 0.01 1.16 0.65 0.00 -.07 0.70 -.01 -1.42 0.32 L3 0.00 -0.30 0.50 0.00 0.54 0.51 0.00 0.15 0.54 0.00 0.30 0.75 0.00 0.19 0.76 -.01 -1.45 0.35 H1 0.01 0.71 0.50 0.01 1.06 0.40 0.01 1.45 0.59 0.01 1.04 0.76 0.00 0.44 0.72 0.00 -0.18 0.22 H3 0.00 0.18 0.51 0.02 2.57** 0.44 0.01 1.17 0.59 0.01 1.65* 0.77 0.00 0.50 0.67 0.00 -0.47 0.34 SALES

L1 0.01 0.90 0.46 0.01 1.03 0.21 0.00 -0.05 0.63 0.01 1.68* 0.62 -.01 -1.0 0.74 0.00 -0.47 0.19 L3 0.00 -0.54 0.55 0.01 0.97 0.31 0.00 0.11 0.52 0.00 0.09 0.74 0.00 -.38 0.80 0.00 -0.03 0.25 H1 0.00 0.04 0.51 0.01 0.73 0.69 0.03 3.37** 0.03 0.01 1.65* 0.73 -.01 -1.0 0.75 0.00 0.58 0.29 H3 -.01 -0.52 0.44 0.00 0.03 0.42 0.00 -0.24 0.59 0.00 1.00 0.74 0.01 0.90 0.81 0.00 -0.39 0.40 SIZE PB

SL1 0.02 1.85* 0.44 – – – 0.04 2.47** -.01 0.01 1.65* 0.80 -.01 -.96 0.84 -.01 -0.97 0.30 SL3 -.04 -2.** 0.69 – – – 0.04 2.54** 0.02 0.00 -0.29 0.81 0.00 0.73 0.78 0.00 0.39 0.20 SH1 0.01 0.65 0.40 – – – 0.04 2.41** 0.00 0.00 0.49 0.74 0.00 -.28 0.75 -.01 -1.19 0.25 SH3 -.01 -1.28 0.38 – – – 0.03 2.39** -.02 0.00 0.37 0.69 0.00 -.75 0.73 0.00 -0.07 0.30 BL1 0.00 -0.37 0.45 – – – 0.02 2.23** -.01 0.01 1.64* 0.62 0.00 -.01 0.80 0.00 -0.68 0.30 BL3 0.00 0.20 0.51 – – – 0.02 2.03** -.03 0.00 0.84 0.72 0.00 0.61 0.78 -.01 -1.24 0.37 BH1 0.00 -0.55 0.45 – – – 0.02 2.31** -.02 0.01 1.17 0.64 0.00 -.31 0.74 -.01 -1.31 0.20 BH3 -.01 -0.74 0.55 – – – 0.02 1.74* -.02 0.00 -0.13 0.72 0.00 -.68 0.81 -.01 -1.8* 0.36

SIZE PE SL1 0.01 0.48 0.35 – – – 0.05 2.71** -.01 0.01 1.58 0.79 -.01 -1.5 0.79 0.00 -0.01 0.29 SL3 -.04 -1.9* 0.70 – – – 0.05 2.23** 0.02 0.00 0.22 0.75 0.00 0.02 0.81 0.00 0.39 0.21 SH1 0.00 0.45 0.38 – – – 0.03 2.18** -.01 0.00 0.46 0.78 0.00 0.11 0.77 -.01 -0.96 0.15 SH3 -.01 -0.79 0.32 – – – 0.02 1.91* -.02 0.00 0.43 0.70 0.00 0.47 0.65 -.01 -0.68 0.19 BL1 0.00 -0.05 0.50 – – – 0.02 2.56** 0.00 0.01 1.90* 0.63 0.00 -.29 0.73 0.00 -0.41 0.37 BL3 0.00 -0.51 0.47 – – – 0.02 2.05** -.03 0.00 1.32 0.76 0.00 -.44 0.76 -.01 -0.85 0.38 BH1 -.01 -0.99 0.47 – – – 0.02 1.73* -.02 0.01 1.21 0.64 0.00 -.46 0.72 -.01 -1.8* 0.19 BH3 0.00 -0.22 0.61 – – – 0.01 1.66* -.02 0.00 -1.03 0.70 0.00 0.15 0.79 -.01 -1.91* 0.32


70

Table 3: Continued

FF Three Factor Model Results : RPt - RFt = α + β (RMt - RFt)+ s SMBt + l LMHt + et Panel C: Excess Return on 48-12-12 Stylized Portfolios Regressed on the Excess Return on the Market (RM-RF) Factor

and Two Proxy Portfolios that Relate to Size (SMB) and (LMH) Factors BRAZIL RUSSIA INDIA CHINA S.KOREA S.AFRICA α t(α) 𝑅�2 α t(α)

α t(α)

α t(α)

α t(α) 𝑅�2 α t(α)

RETURN PORTFOLIOS P1 0.02 1.88* 0.44 0.02 1.45 0.37 0.04 3.07** -.02 0.01 2.01** 0.63 0.00 -0.03 0.79 0.00 -0.12 0.20 P5 0.00 -0.22 0.56 0.04 1.22 0.14 0.03 2.82** -.02 0.00 0.63 0.74 0.00 0.72 0.80 0.00 -0.09 0.44 SIZE

S1 0.01 1.79* 0.48 0.01 0.86 0.27 0.00 0.46 0.75 0.01 1.89* 0.63 0.00 -0.08 0.83 -.01 -1.15 0.28 S3 -.02 -1.42 0.66 0.04 1.13 0.04 0.01 0.89 0.73 0.01 1.57 0.62 0.01 1.37 0.80 0.00 -0.16 0.33 B1 0.00 -0.26 0.50 0.01 2.05** 0.60 0.00 0.54 0.62 0.01 1.59 0.68 0.00 0.60 0.77 -.01 -0.91 0.32 B3 0.00 -0.40 0.60 0.01 1.62 0.59 0.00 0.70 0.57 0.00 0.43 0.79 0.00 0.38 0.79 -.01 -1.04 0.46 PB

L1 0.01 1.44 0.49 0.00 0.21 0.53 0.00 0.59 0.71 0.01 0.94 0.78 0.00 0.07 0.83 -.01 -1.10 0.27 L3 -.02 -1.66* 0.69 0.01 0.63 0.41 0.00 0.27 0.69 0.00 -0.05 0.77 0.01 1.77* 0.78 0.00 -0.42 0.39 H1 0.00 0.11 0.46 0.02 3.06** 0.55 0.00 0.42 0.54 0.01 1.46 0.68 0.00 0.39 0.73 -.01 -0.94 0.34 H3 0.00 -0.33 0.61 0.01 1.16 0.54 0.00 0.75 0.53 0.00 0.03 0.72 0.00 0.16 0.78 0.00 -0.37 0.43 PE

L1 0.01 0.71 0.42 0.02 2.91** 0.53 0.01 1.02 0.73 0.01 1.48 0.76 0.00 -0.16 0.80 -.01 -1.31 0.32 L3 -.02 -1.42 0.68 0.01 1.84* 0.54 0.01 1.46 0.67 0.00 -0.12 0.74 0.01 1.16 0.79 0.00 -0.13 0.40 H1 0.00 0.28 0.53 0.00 0.00 0.46 0.00 0.05 0.60 0.01 1.21 0.69 0.00 0.69 0.76 -.01 -1.43 0.26 H3 -.01 -0.73 0.59 0.01 0.66 0.51 0.00 0.36 0.54 0.00 -0.17 0.70 0.00 0.62 0.77 -.01 -1.43 0.40 DYIELD

L1 0.00 0.13 0.47 0.02 1.89* 0.46 0.00 0.43 0.57 0.01 1.29 0.62 0.01 1.07 0.72 -.01 -1.16 0.30 L3 0.00 -0.06 0.51 0.00 -0.48 0.54 0.00 0.48 0.55 0.00 -0.03 0.70 0.00 0.80 0.78 -.01 -1.07 0.37 H1 0.00 0.42 0.50 0.02 2.48** 0.42 0.01 1.90* 0.55 0.00 0.84 0.75 0.01 1.51 0.75 0.00 0.24 0.19 H3 0.01 0.95 0.53 0.02 1.83* 0.41 0.01 1.82* 0.56 0.00 0.55 0.78 0.01 1.20 0.73 0.00 -0.01 0.32 SALES

L1 0.01 1.05 0.46 0.00 0.54 0.50 0.00 -0.11 0.67 0.01 1.89* 0.76 0.00 -0.40 0.74 0.00 0.19 0.13 L3 0.00 -0.23 0.58 0.01 1.06 0.23 0.00 0.10 0.52 0.00 -0.04 0.73 0.00 0.64 0.79 0.00 -0.22 0.37 H1 0.00 0.58 0.52 0.00 0.08 0.52 0.01 1.28 0.62 0.01 1.75* 0.72 0.00 -0.41 0.74 0.00 -0.05 0.27 H3 0.01 0.78 0.39 0.01 0.56 0.41 0.00 0.31 0.58 0.00 0.50 0.73 0.01 1.50 0.80 0.00 -0.53 0.39 SIZE PB

SL1 0.02 1.95* 0.45 – – – 0.00 0.34 0.74 0.01 1.12 0.79 -.01 -0.82 0.85 -.01 -0.73 0.27 SL3 -.05 -2.** 0.72 – – – 0.00 0.52 0.74 0.00 -0.30 0.80 0.01 2.03** 0.74 0.00 0.45 0.21 SH1 0.02 1.60 0.39 – – – 0.01 0.99 0.58 0.00 0.60 0.75 0.00 -0.27 0.77 0.00 -0.46 0.28 SH3 0.00 -0.44 0.33 – – – 0.01 1.47 0.54 0.00 0.30 0.73 0.00 0.55 0.74 -.01 -0.87 0.29 BL1 0.00 -0.17 0.43 – – – 0.01 0.99 0.58 0.01 1.70* 0.61 0.00 0.40 0.76 -.01 -0.78 0.26 BL3 0.00 -0.03 0.51 – – – 0.01 1.47 0.54 0.00 0.11 0.70 0.00 0.70 0.77 -.01 -0.93 0.39 BH1 -.01 -0.65 0.48 – – – 0.00 0.42 0.49 0.01 1.12 0.68 0.00 0.55 0.72 0.00 -0.58 0.29 BH3 0.00 0.42 0.52 – – – 0.00 0.66 0.53 0.00 -0.48 0.70 0.00 -0.35 0.81 -0.0 1.64* 0.37

SIZE PE SL1 0.01 0.83 0.31 – – – 0.01 0.86 0.72 0.01 1.17 0.75 -.01 -1.56 0.80 0.00 -0.47 0.28 SL3 -.04 -1.9* 0.73 – – – 0.01 0.52 0.67 0.00 0.33 0.71 0.01 1.64* 0.81 0.00 0.01 0.20 SH1 0.02 1.95* 0.42 – – – 0.00 0.45 0.66 0.00 0.78 0.79 0.00 0.62 0.80 -.01 -0.94 0.14 SH3 -.01 -0.56 0.32 – – – 0.00 -0.22 0.62 0.00 0.39 0.66 0.01 0.97 0.65 -.01 -1.09 0.23 BL1 0.00 -0.11 0.49 – – – 0.01 1.44 0.57 0.01 1.89* 0.60 0.00 0.48 0.75 -.01 -0.73 0.33 BL3 0.00 -0.13 0.52 – – – 0.01 1.35 0.50 0.00 0.18 0.76 0.00 0.30 0.75 -.01 -0.70 0.41 BH1 -.01 -1.05 0.48 – – – 0.00 -0.22 0.52 0.01 1.09 0.67 0.00 0.07 0.73 -.01 -0.97 0.14 BH3 0.00 -0.30 0.62 – – – 0.00 0.43 0.55 0.00 -0.86 0.70 0.00 0.45 0.79 -.01 -1.01 0.31


71

Table 3 Continued

FF Three Factor Model Results : RPt - RFt = α + β (RMt - RFt)+ s SMBt + l LMHt + et Panel D: Excess Return on 60-12-12 Stylized Portfolios Regressed on the Excess Return on the Market (RM-RF) Factor

and Two Proxy Portfilios that Relate to Size (SMB) and (LMH) Factors BRAZIL RUSSIA INDIA CHINA S.KOREA S.AFRICA α t(α)

α t(α)

α t(α) 𝑅�2 α t(α)

α t(α)

α t(α) 𝑅�2

RETURN PORTFOLIOS P1 0.01 1.34 0.48 0.03 1.81* 0.26 0.04 3.10** 0.02 0.01 1.54 0.63 0.00 0.47 0.69 0.00 0.46 0.14 P5 0.00 -0.29 0.58 0.05 1.01 0.10 0.03 2.74** 0.01 0.00 0.00 0.71 0.00 0.32 0.72 0.00 -.45 0.29 SIZE

S1 0.01 1.64* 0.48 0.02 0.93 0.27 0.00 0.32 0.76 0.01 1.28 0.65 0.00 0.39 0.75 -.01 -.66 0.27 S3 -.01 -1.02 0.71 0.04 1.18 0.04 0.00 0.62 0.72 0.00 0.68 0.64 0.01 1.05 0.69 -.01 -.50 0.15 B1 0.00 -0.03 0.51 0.02 2.50** 0.58 0.00 0.43 0.63 0.01 1.53 0.66 0.00 0.78 0.64 0.00 -.06 0.21 B3 -.01 -0.66 0.55 0.01 1.48 0.56 0.00 0.21 0.56 0.00 -0.02 0.77 0.00 -0.01 0.75 0.00 -.56 0.28 PB

L1 0.01 1.69* 0.51 0.01 0.63 0.50 0.00 0.37 0.73 0.01 1.30 0.75 0.00 -0.01 0.73 0.00 -.51 0.20 L3 -.02 -1.37 0.71 0.01 0.62 0.40 0.00 0.15 0.69 0.00 -0.04 0.77 0.01 1.05 0.68 0.00 0.05 0.21 H1 0.01 0.61 0.43 0.02 2.56** 0.57 0.00 0.67 0.56 0.01 1.17 0.66 0.01 0.98 0.61 0.00 0.04 0.27 H3 0.00 -0.42 0.59 0.01 0.94 0.51 0.00 0.40 0.53 0.00 0.01 0.71 0.00 -0.11 0.75 -0.0 -.95 0.25 PE

L1 0.01 0.78 0.43 0.02 2.51** 0.62 0.01 0.93 0.74 0.01 1.34 0.74 0.00 0.12 0.71 -.01 -.67 0.27 L3 -.02 -1.34 0.70 0.01 1.52 0.50 0.01 0.98 0.67 0.00 0.46 0.75 0.00 0.48 0.70 0.00 -.19 0.25 H1 0.00 0.33 0.52 0.00 0.29 0.38 0.00 0.10 0.59 0.01 1.08 0.66 0.01 1.04 0.63 0.00 -.49 0.15 H3 -.01 -0.67 0.52 0.00 0.40 0.39 0.00 0.00 0.54 0.00 -0.58 0.70 0.00 0.24 0.71 -.01 -1.2 0.23 DYIELD

L1 0.00 0.39 0.50 0.01 1.33 0.49 0.00 0.67 0.57 0.01 1.31 0.65 0.00 0.62 0.60 0.00 -.45 0.17 L3 -.01 -0.60 0.51 0.00 -0.06 0.45 0.00 -0.03 0.54 0.00 0.06 0.69 0.00 0.57 0.75 0.00 -.12 0.22 H1 0.01 0.73 0.49 0.02 1.96** 0.48 0.01 1.37 0.59 0.01 1.53 0.71 0.01 1.76* 0.59 0.01 1.30 0.11 H3 0.01 1.06 0.44 0.02 1.61 0.40 0.01 1.50 0.58 0.00 0.95 0.77 0.00 0.57 0.59 0.00 0.52 0.14 SALES

L1 0.01 0.85 0.52 0.01 1.40 0.42 0.00 0.02 0.67 0.01 1.36 0.74 0.00 -0.71 0.67 0.01 0.77 0.05 L3 0.00 -0.19 0.55 0.01 1.28 0.20 0.00 -0.15 0.51 0.00 -0.13 0.74 0.00 0.44 0.69 0.00 0.47 0.21 H1 0.01 0.99 0.49 0.00 0.46 0.61 0.01 0.94 0.66 0.01 1.40 0.70 0.00 0.52 0.56 0.01 1.30 0.14 H3 0.01 0.93 0.38 0.00 0.21 0.42 0.00 0.46 0.57 0.00 0.00 0.73 0.00 0.87 0.77 0.00 0.13 0.19 SIZE PB

SL1 0.02 1.64* 0.38 – – – 0.00 0.11 0.70 0.01 1.24 0.76 -.01 -0.87 0.78 0.00 -.29 0.26 SL3 -.04 -1.68* 0.74 – – – 0.00 0.28 0.72 0.00 -0.26 0.75 0.01 1.15 0.72 0.00 0.21 0.22 SH1 0.02 1.69* 0.40 – – – 0.01 0.64 0.73 0.00 0.70 0.72 0.01 1.20 0.67 -.01 -.57 0.31 SH3 0.00 -0.43 0.38 – – – 0.01 0.87 0.63 0.00 0.71 0.71 0.00 0.68 0.58 -.01 -1.1 0.22 BL1 0.00 -0.02 0.49 – – – 0.01 1.07 0.59 0.01 1.60 0.62 0.00 0.58 0.64 0.00 0.19 0.18 BL3 0.00 -0.11 0.50 – – – 0.01 0.82 0.53 0.00 0.58 0.71 0.00 0.34 0.70 0.00 -.09 0.24 BH1 0.00 0.11 0.46 – – – 0.00 0.30 0.48 0.01 1.06 0.69 0.00 0.70 0.60 0.00 -.16 0.23 BH3 0.00 -0.16 0.52 – – – 0.00 0.37 0.52 0.00 -0.36 0.72 0.00 -0.51 0.79 -.01 -.93 0.24

SIZE PE SL1 0.01 0.74 0.36 – – – 0.01 0.74 0.71 0.01 1.30 0.70 -.01 -0.94 0.71 0.00 0.00 0.26 SL3 -.04 -1.56 0.73 – – – 0.01 0.72 0.66 0.00 0.08 0.71 0.01 1.03 0.74 0.00 0.15 0.20 SH1 0.02 2.12** 0.50 – – – 0.00 0.38 0.68 0.00 0.30 0.79 0.01 1.23 0.68 0.00 -.48 0.14 SH3 0.00 -0.17 0.33 – – – 0.00 -0.19 0.65 0.00 0.50 0.69 0.01 0.76 0.51 -.01 -.97 0.20 BL1 0.01 0.67 0.52 – – – 0.01 1.26 0.59 0.01 1.77* 0.59 0.01 0.99 0.65 0.00 0.33 0.24 BL3 0.00 -0.34 0.53 – – – 0.01 1.11 0.51 0.00 1.09 0.76 0.00 -0.25 0.67 0.00 -.19 0.26 BH1 0.00 -0.59 0.46 – – – 0.00 -0.11 0.54 0.01 1.13 0.68 0.00 0.29 0.60 -.01 -1.1 0.17 BH3 0.00 -0.55 0.57 – – – 0.00 0.01 0.55 0.00 -0.86 0.68 0.00 -0.08 0.77 -.01 -.75 0.25

The table reports Fama French three-factor model results. Excess returns on the sample portfolios are regressed on the returns on market, size and value factor in the Fama French frame work. Alpha (α), the intercept term, shows risk adjusted returns while adjusted R-square is the goodness of fit measure. ** t -statistics are tested for significance at 5% level on 2-tail basis and *t-statistics are tested for significance at 10% level on 2-tail basis. Next, the results for mean excess returns on sectoral portfolios have been reported in Table 4. Russia and India exhibit strong momentum patterns for all portfolio formation (24-60 months) strategies. While Brazil reports weak momentum patterns for 24-48 months portfolio formation windows and weak contrarian patterns for 60 months formation windows. China and South Africa show weak contrarian patterns for all portfolio formation windows. There are no clear prior return patterns for South Korea. The highest momentum returns are reported by Russia of 4.14% for 36-12-12 strategies.


72

Table 4: Mean Excess Returns on Sectoral Portfolios

Mean Excess Returns on Sectoral Momentum Portfolios Panel A: 24-12-12 Stylized Portfolios

COUNTRY BRAZIL RUSSIA INDIA CHINA S.KOREA S.AFRICA K1 0.0195 0.0509 -0.0052 0.0234 0.0174 0.0162 K2 0.0190 0.0390 0.0062 0.0193 0.0201 0.0200 K3 0.0147 0.2456 0.0093 0.0179 0.0283 0.0025 K4 0.0127 0.0546 0.0129 0.0192 0.0159 0.0112 K5 0.0256 0.0738 0.0247 0.0195 0.0117 0.0092

K5 - K1 0.0062 0.0229 0.0299 -0.0040 -0.0057 -0.0071 EWI 0.0183 0.0928 0.0096 0.0199 0.0187 0.0118

Panel B: 36-12-12 Stylized Portfolios K1 0.0173 0.0358 -0.0055 0.0221 0.0286 0.0156 K2 0.0216 0.3585 0.0056 0.0152 0.0323 0.0192 K3 0.0181 0.0456 0.0084 0.0205 0.0234 0.0159 K4 0.0177 0.0341 0.0112 0.0200 0.0267 0.0048 K5 0.0279 0.0772 0.0220 0.0184 0.0296 0.0158

K5 - K1 0.0106 0.0414 0.0276 -0.0037 0.0010 0.0002 EWI 0.0205 0.1102 0.0083 0.0192 0.0281 0.0143

Panel C: 48-12-12 Stylized Portfolios K1 0.0140 0.0454 -0.0029 0.0221 0.0283 0.0213 K2 0.0265 0.0430 0.0058 0.0194 0.0320 0.0167 K3 0.0249 0.2240 0.0085 0.0164 0.0391 0.0058 K4 0.0130 0.0384 0.0108 0.0198 0.0214 0.0111 K5 0.0253 0.0809 0.0217 0.0171 0.0199 0.0114

K5 - K1 0.0113 0.0356 0.0245 -0.0050 -0.0084 -0.0098 EWI 0.0207 0.0863 0.0088 0.0190 0.0281 0.0133

Panel D: 60-12-12 Stylized Portfolios K1 0.0245 0.0525 -0.0004 0.0211 0.0116 0.0186 K2 0.0245 0.5795 0.0061 0.0231 0.0206 0.0169 K3 0.0299 0.0470 0.0084 0.0166 0.0152 0.0180 K4 0.0330 0.0441 0.0108 0.0183 0.0159 0.0117 K5 0.0189 0.0563 0.0202 0.0201 0.0257 0.0148

K5 - K1 -0.0056 0.0038 0.0207 -0.0010 0.0142 -0.0038 EWI 0.0262 0.1559 0.0090 0.0199 0.0178 0.0160

The table shows mean excess returns on sectoral momentum portfolios. We form five sectoral portfolios for each of the long-term portfolio formation windows (24,36, 48, and 60 months) and then estimate 12 month holding period returns for the sample portfolios after skipping one year between portfolio formation and holding windows for any short-term prior return effects. The empirical results based on four-factor model are given in Table 5. We find that the sector factor is able to capture average returns for 36-12-12 in case of India and 24-12-12 in case of South Africa. However, the abnormal returns for China in case of 24-12-12 strategy persist and hence continue to be an asset pricing puzzle. The Chinese anomaly may be an aberration or perhaps require some behavioral explanation. To conclude, our four-factor model does a better job than CAPM and the FF model in explaining prior return patterns in stock returns and hence it should be used as a baseline for evaluating investment strategies.


73

Table 5: Four Factor Model Results

Four Factor Model Results: RPt - RFt = α + β (RMt - RFt)+ s SMBt + l LMHt + w WMLt + et Excess Return on Stylized Portfolios Regressed on the Excess Return on the Market (RM-RF) Factor

and Three Proxy Portfilios that Relate to Size (SMB), (LMH) and Sector (WML) Factors Strategy 24-12-12 24-12-12 36-12-12

INDIA S.AFRICA CHINA α t(α)

α t(α)

α t(α)

RETURN PORTFOLIOS P1 0.03 3.05** 0.01 0.00 0.02 0.26 0.01 1.51 0.77 P5 0.03 2.96** 0.02 0.01 0.67 0.35 0.00 1.02 0.73 SIZE

S1 0.01 1.26 0.72 -0.01 -1.64* 0.29 0.01 2.30** 0.63 S3 0.01 1.07 0.74 0.00 0.15 0.30 0.01 2.02** 0.67 B1 0.00 0.46 0.59 -0.01 -1.27 0.34 0.01 1.86* 0.68 B3 0.00 0.02 0.58 -0.01 -1.31 0.49 0.00 1.14 0.81 PB

L1 0.00 0.82 0.71 -0.01 -1.89* 0.35 0.00 0.69 0.79 L3 0.00 0.21 0.72 0.00 -0.26 0.36 0.00 -0.21 0.79 H1 0.01 0.99 0.51 -0.01 -0.89 0.35 0.01 1.60 0.67 H3 0.00 0.62 0.54 0.00 -0.64 0.46 0.00 0.73 0.76 PE

L1 0.01 1.64* 0.69 -0.01 -1.18 0.34 0.01 1.51 0.77 L3 0.00 0.31 0.68 0.00 -0.10 0.37 0.00 0.23 0.76 H1 0.00 -0.09 0.54 -0.01 -1.44 0.28 0.01 1.41 0.64 H3 0.00 0.43 0.54 -0.01 -1.41 0.42 0.00 0.15 0.72 DYIELD

L1 0.00 0.28 0.53 -0.01 -1.82* 0.30 0.01 1.38 0.66 L3 0.00 0.40 0.56 -0.01 -1.24 0.39 0.00 0.50 0.75 H1 0.01 1.91* 0.55 0.00 -0.43 0.25 0.00 0.88 0.76 H3 0.00 0.72 0.58 0.00 -0.15 0.30 0.01 1.69 0.77 SALES

L1 0.00 -0.73 0.66 0.00 0.28 0.24 0.01 1.75* 0.62 L3 0.00 0.06 0.54 0.01 0.73 0.26 0.00 0.26 0.74 H1 0.01 1.65* 0.62 0.00 0.27 0.30 0.01 1.50 0.73 H3 0.00 -0.09 0.66 0.00 0.31 0.39 0.01 1.18 0.74 SIZE PB

SL1 0.01 1.10 0.70 -0.01 -1.18 0.27 0.01 1.67* 0.80 SL3 0.00 0.31 0.79 0.01 0.79 0.16 0.00 -0.20 0.81 SH1 0.01 0.80 0.65 0.00 -0.36 0.19 0.00 0.63 0.74 SH3 0.00 0.44 0.63 0.00 0.25 0.35 0.00 0.63 0.70 BL1 0.00 0.45 0.58 0.00 -0.53 0.33 0.01 1.71* 0.62 BL3 0.00 -0.20 0.59 0.00 -0.75 0.42 0.00 1.01 0.72 BH1 0.00 -0.24 0.44 -0.01 -1.48 0.29 0.01 1.45 0.66 BH3 0.00 0.73 0.54 -0.01 -1.98** 0.35 0.00 0.22 0.73

SIZE PE SL1 0.01 1.40 0.70 -0.01 -0.65 0.31 0.01 1.73* 0.79 SL3 0.00 0.30 0.70 0.01 0.81 0.16 0.00 0.27 0.75 SH1 0.00 0.37 0.65 -0.01 -1.28 0.14 0.00 0.53 0.78 SH3 0.00 0.22 0.58 0.00 -0.46 0.24 0.00 0.69 0.71 BL1 0.01 1.61 0.56 0.00 -0.73 0.35 0.01 1.94* 0.63 BL3 0.00 0.15 0.55 0.00 -0.34 0.39 0.01 1.48 0.76 BH1 0.00 0.04 0.47 -0.01 -1.85* 0.21 0.01 1.43 0.64 BH3 0.00 0.42 0.55 -0.01 -1.04 0.31 0.00 -0.65 0.72

The table shows the excess returns on the sample portfolios (that are not explained by the Fama French model) are regressed on the four risk factors including an additional sector prior return factor. Alpha (α) is a measure of extra normal performance and adjusted R-square is the goodness of fit measure. ** t -statistics are tested for significance at 5% level on 2-tail basis. *t-statistics are tested for significance at 10% level on 2-tail basis. CONCLUDING COMMENTS There is a large body of empirical research that deals with prominent asset pricing anomalies for both mature and emerging markets. Amongst the pricing anomalies, long-term mean reversal (contrarian) and short-term continuation (momentum) have received much attention over the last three decades. Academicians are still inconclusive about the sources of momentum/reversal profits. While some attribute it to risk factors, others believe that these extra normal returns are driven by some kind of behavioral biases.


74

In this paper, we examine for long-term prior return patterns in stock returns for BRICKS markets and whether they can be explained by standard asset pricing models such as the CAPM or the Fama French three-factor model. We form portfolios based on 24-60 months past returns and skip one year between formation and holding windows to control for any short-term momentum effects in these markets as documented by Sehgal and Jain, 2011. Four key propositions have been examined: (1) Do long-term portfolio formation strategies provide abnormal profits?, (2) Can these profits be absorbed by standard risk models like the CAPM or the Fama-French three-factor model?, (3) Are there any long-term prior return patterns in sector returns and (4) Can the prior returns patterns in stock returns that are missed by CAPM and the Fama French model be absorbed by introducing an additional sector prior return factor. For long-term prior return based portfolios, we observe momentum behavior for Brazil, Russia and South Africa and this pattern persists even as we elongate the formation windows whereas India, China and South Korea report contrarian behavior. For double and triple sorted portfolios (based on company characteristics and long-term prior returns), similar patterns are reported by the above mentioned countries. The CAPM is able to explain most prior return patterns in Brazil, Russia, China and South Africa for 24 and 36 months portfolio formation strategies but it doesn’t do a good job for longer term portfolio formation strategies i.e. 48 and 60 months. In case of India and South Korea, the CAPM seems to be a poor descriptor of prior return patterns across all long-term portfolio formation strategies. The Fama French model is able to capture long-term prior return patterns in stock returns for BRICKS countries that are missed by the CAPM, with the exception of China and South Africa for 24-12-12 strategy and India for 36-12-12 strategy. We explore if there are any prior return patterns in sector returns as was observed in case of stock returns. We find that Russia and India exhibit strong momentum patterns in sector for all portfolio formation windows (24-60 months) strategies. While Brazil reports weak momentum patterns for 24-48 months portfolio formation windows and weak contrarian patterns for 60 months formation windows. China and South Africa show weak contrarian patterns for all portfolio formation windows. There are no clear prior return patterns for South Korea. Given the few anomalies in case of India, China and South Africa, we augment the F-F model by including a sector prior return factor which is formed on the economic argument of Liu and Zhang (2008). The sector factor is able to capture average returns for 36-12-12 strategy in case of India and 24-12-12 strategy in case of South Africa. However, the abnormal returns for China in case of 24-12-12 strategy persist and hence continue to be an asset pricing puzzle. The unexplained returns even after controlling for sector factor in case of China may warrant a behavioral explanation or there may be some other missing risk factor(s) which may explain returns. Our findings are pertinent for portfolio managers and investment analysts who are continuously in pursuit of trading strategies that provide extra normal returns. From an academic point of view, we suggest that a sector factor should be used in the multi factor framework for explaining asset returns. Our research contributes to the asset pricing and behavioral finance literature for emerging markets. REFERENCES Ahn, D.H.; Conrad, J.; Dittmar, R.F. (2003). Risk Adjustment and Trading Strategies. The Review of Financial Studies , 16 (2), 459-485. Antoniou,A.;Lam,H.Y.T.;Paudyal,K. (2007). Profitabilty of Momentum Strategies in International Markets: The roel of business cycle and behavioral biases. Journal of Banking and Finance , 955-972.


75

Ball, R.; Kothari, S. (1989). Nonstationary Expected Returns: Implications for Tests of Market Efficiency and Serial Correlation in Returns. Journal of Financial Economics , 25, 51-74. Ball, R.; Kothari, S.; Shanken, J. (1995). Problems in Measuring Portfolio Performance : An Application to Contrarian Investment Strategies. Journal of Financial Economics , 38, 79-107. Banz, Rolf W. (1981). The Relationship between Return and Market Value of Common Stocks. Journal of Financial Economics, 9, 3-18. Barferis, N.; Shliefer, A.; Vishny, R. (1998). A Model of Investor Sentiment. Journal of Financial Economics , 49, 307-343. Basu, S. (1983). The Relationship between Earning Yield, Market Value and Return for NYSE Common Stocks: Further Evidence. Journal of Financial Economics , 12, 129-156. Berk, J.;Green, J.; Naik, V. (1999). Optimal Investment, Growth Options, and Security Returns. Journal of Finance , 54, 1553-1607. Boni,L. ; Kent, L.W. (2006). Analysts, Industries, and Price Momentum. Journal of Financial and Quantitative Analysis , 41 (1), 85-109. Carhart, Mark M. (1997). On Persistance in Mutual Fund Performance. Journal of Finance , 52, 57-82. Chan, K. (1988). On the Contrarian Investment Strategy. The Journal of Business , 61 (2), 147-163. Chan, K.; Hameed, A.; Tong, W. (2000). Profitability of Momentum Stragegies in the International Equity Markets. Journal of Financial and Quantitative Analysis , 35 (2), 153-172. Chan, L.; Jegadeesh, N.; Lakonishok, J. (1996). Momentum Strategies. Journal of Finance , 51, 1681-1713. Chan. Louis K.C.; Hamao, Y.; Lakonishok, Josef. (1991). Fundamentals and Stock Returns in Japan. Journal of Finance , 46, 1739-1764. Chen, J.; Benett, A.; Zheng, T. (2006). Sector Effects in Developed vs. Emerging Markets. Financial Analysts Journal , 62 (6), 40-51. Chen,H.;Chen,S.; Hsin,C.;Lee.C. (2010). Price, Earnings and Revenue Momentum Strategies. http://papers.ssrn.com/sol3/papers.cfm/Abstract id=1571883 . Chopra, N.; Lakonishok, J.; Ritter, J. (1992). Measuring Abnormal Performance: Do Stocks overreact? Journal of Financial Economics , 31, 235-268. Chordia, T.; Shivkumar, L. (2002). Momentum, Business Cycle and Time, Varying Expected Returns. Journal of Finance , 57 (2), 985-1019. Chui,A.C.W., Titman, S., Kim, J.K.C. (2000). Momentum, Legal Systems and Ownership Structure: An Analysis of Asian Stock Markets.


76

Conrad, J.; Kaul, G.; Nimalendran M. (1991). Components of Short-Horizon Individual Security Returns. Journal of Financial Economics , 29, 365-384. Conrad, Jennifer; Kaul, Gautum. (1993). Long-Term Market Overreaction or Biases in Computed Returns? Journal of Finance , 48 (1), 39-63. Daniel, K.; Hiirshleifer, D.; Subrahmanyam, A. (1998). Investor Psychology and Security Marekt Under-and-over Reactions. Journal of finance , 53, 1839-1886. De Bondt, W.F.; Thaler, R. (1985). Do the Stock Markets Overreact? Journal of Finance , 40, 793-805. De Bondt, W.F.; Thaler, R. (1987). Further Evidence of Investor Overreaction and Stock Market Seasonality. Journal of Finance, 42, 557-581. Fama, Eugene F,; French, Kenneth R. (1996). Multi-factor Expalnation of Asset Pricing Anomalies. Journal of Finance , 51, 55-84. Fama, Eugene F.; French, Kenneth R. (1993). Common Risk Fators in the Returns on Stocks and Bonds. Journal of Financial Economics , 33, 3-56. Fama, Eugene F.; French, Kenneth R. (1992). The Cross-Section of Expected Returns. Journal of Finance , 47, 427-466. Frankel, J.A.;Schmukler, S.L. (1996). Country Fund Discounts, Asymetric Information and the Mexican Crisis of 1994: Did Local Residents Turn Pessimistic Before International Investors? Open Economics Review, 7, 511-534. Froot, K.; O Connell; Seasholes, P. (2001). The Portfolio Flows of International Investors. Journal of Financial Economics , 59, 2515-2547. Hameed,A.;Kusnadi,Y. (2002). Momentum Strategies:Evidence from Pacific Basin Stock Markets. Journal of Financial Research , 25 (3), 383-397. Hong, H.; Lim, J.C.; Stien, J. (2000). Bad News Travels Slowly: Size, Analyst Coverage, and the Profitability of Momentum Strategies. Journal of Finance , 55, 265-295. Hong, H.;Stien,J. (1999). A Unified theory of Underreaction, Momentum Trading and Overreaction in Assets Markets. Journal of Finance, 55, 265-295. Jegadeesh, N. ; Titman, S. (2001). Profitability of Momentum Strategies: An Evaluation of Alternative Explanations. Journal of Finance, 56, 699-720. Jegadeesh, N.; Titman, S. (2002). Cross-Sectional and Time Series Determinants of Momentum Returns. Review of Financial Studies , 15, 143-158. Jegadeesh, N.; Titman, S. (1995). Overreaction, Delayed reaction, and Contrarian Profits. Review of Financial Studies , 48, 973-993. Jegadeesh, N.; Titman, S. (1993). Returns to Buying Winners and Selling Losers: Implications of Stock Market Effeciency. Journal of Finance , 48, 65-91.


77

Jung, J.; Shiller, R. (2005). Samuelson's Dictum And The Stock Market. Economic Inquiry , 43 (2), 221–228. Kaminsky, G.;Lyons. R.K.;Schmukler, S.L. (2004). Managers, Investors, Crises: Mutual Fund Strategies in Emerging Markets. Journal of International Economics , 64, 113-134. Kaul, Guatum; Nimalendran, M. (1990). Price Reversals: Bid-ask Errors on Market Overreaction? Journal of Financial Economics , 28, 67-83. Kent, D.; Hirshleifer, D.; Subrahmanyam, A. (2004). A Theory of Overconfidence, Self-Attribution, and Security Market Under-and-overReactions. Finance 0412006, EconWPA . Kim, W.; Wei, S.J. (2002). Foreign portfolio investors before and during a crisis. Journal of International Economics , 56 (1), 77-96. Lakonishok, Josef; Shliefer, Andrei; Vishny, Robert W. (1994). Contraian Investment, Extrapolation and Risk. Journal of Finance , 49, 1541-1578. Lee, C.; Swaminathan. (2000). Price Momentum and Trading Volume. Journal of Finance , 55, 2017-2069. Lewellen, J. (2002). Momentum and Autocorrelation in Stock Returns. The Review of Financial Studies , 15, 533-573. Lin, A.Y.; Swanson, P. (2004). International Equity Flows and Developing Markets: the Asian Financial Market Crisis Revisited. Journal of International Financial Markets, Institutions & Money , 14, 55-73. Lintner, John. (1965). The Valuation of Risky Assets and the Selection of Risky Investment in Stock Portfolios and Capital Budgets. Review of Economics and Statistics , 47, 13-37. Litzenberg,R.H.;Ramaswamy,K. (1979). The effect of Personal Taxes and Dividend on Capital Asset Prices: Theory and Empirical Evidence. Journal of Financial Economics , 7 (2). Liu,L.X.;Zhang,L. (2008). Momentum Profits, Factor Pricing and Macroeconomic Risk. Review of Financial Studies , 41-66. Lo, A.; MacKinlay, C. (1990). "When are Contrarian Profits due to Stock market Overreaction?.". Review of Financial Studies (3), 175-206. Menzly,L.;Ozbas,O. (2006). Cross Industry Momentum. Moskowitz, T.J.; Grinblatt, M. (1999). Do Industries Explain Momentum? Journal of Finance , 44, 1249-1290. Nijman, T.;Swinkels, L.; Verbeek, M. (2004). Do Countries or Industries explain momentum in Europe? Journal of Empirical Finance , 461-481. Richards, A.J. (2002). Big Fish in Small Ponds:The Momentum Investing and Price Impact of Foreign Investors in asian Equity Markets. IMF and The Reserve Bank of Australia . Safieddine,A.;Sonti,R. (2007). Momentum and Industry Growth. Review of Financial Economics , 16 (2), 203-215.


78

Schiereck, D.; DeBondt, W.; Weber, M. (1999). Contrarian and Momentum Strategies in Germany. Financial Analyst Journal , 104-116. Scott, J.; Stump, M. ; Xu, P. (2003). News, Not Trading Volume, Builds Momentum. Financial Analysts Journal , 59 (2), 45-54. Scowcroft, A.; Sefton, J. (2005). Understanding momentum. Financial Analyst Journal , 61 (2), 64-82. Sehgal,S.; Jain, S. (2011). Long-term prior return patterns in stock and sectoral returns in India. Sharpe, W. (1964). Capital Asset Prices: A Theory of market Equilibrium under conditions of Risk. Journal of Finance , 19, 425-442. Swanson,P.;Lin,A.Y. (2005). Trading Behavior and Investment Performace of U.S. Investors in Global Equity Markets. Journal of Multinational Financial Management , 99-115. BIOGRAPHY Sanjay Sehgal is Ph.D. finance from Delhi School of Economics and post doctoral commonwealth research fellow from London School of Economics, UK. He is professor of Finance at Department of Financial Studies, South Campus, and University of Delhi, India, [email protected]. Sakshi Jain is Master’s in sciences graduate from Loughborough University and is currently pursuing doctoral research at Department of Financial Studies, South Campus, and University of Delhi, India, [email protected]. Pr Laurence the PORTEU de LA MORANDIERE is a professor of Finance, Groupe ESC Pau, Campus Universitaire - 3, Rue Saint John Perse, BP 7512 - 64075 PAU Cedex, FRANCE, [email protected].


79

THE FAMA FRENCH MODEL OR THE CAPITAL ASSET PRICING MODEL: INTERNATIONAL

EVIDENCE Paulo Alves, Lisbon Accounting and Management Institute (ISCAL), Lusofona University and

CMVM

ABSTRACT

This research paper attempts to evaluate the benefits of using the Fama and French Model by comparing them with those resulting from the use of the Capital Asset Pricing Model. Local, International, and European Monetary Union functional forms were considered, in an attempt to raise the following questions: Does the calculation method to determine size and financial distress premium have any significance for the financial analyst? Do the foreign risk premiums of the Fama and French Model have any importance for the financial analyst? Firstly, models based on European Monetary Union factors produce the worst results, independently of any Capital Asset Pricing Model or Fama and French Model consideration. Secondly, independently of the model, the expected return of big and low book-to-market stocks is more reliable. This is particularly observable for big firms, as it does not occur for low book-to-market firms using Fama and French Models. Finally, the Fama and French Model is notoriously preferable in comparison with the Capital Asset Pricing Model for small and high-book to market firms: in this case, the introduction of international factors increases the reliability of expected returns. JEL: G12; G15 KEYWORDS: CAPM, FFM, Local Factors, International Factors INTRODUCTION

he analysis of the cost of capital has, for the past decades, been a very important topic for academia and practitioners. It is well known that the expected return should be put to several different uses: in event studies, to parameterize managerial incentive schemes, to evaluate financial

assets, to assess the quality of investments, and to infer market efficiency. The improper (use of an) estimate of the cost of capital may mean accepting investment projects without quality, carelessly acquiring or selling shares (without adequate fundamentals), drawing the wrong conclusions from events studies, among other financial economics problems of great relevance. This is a matter of great importance for academia and practitioners, and also for society as a whole. Noticeably, bad investment decisions made by Governments, resulting from inadequate assessments capital costs, have a profound future impact on taxes and sovereign debts paid by citizens. In consequence of these concerns this research compares the Capital Asset Pricing Model (CAPM) and the Fama and French Model (FFM). Apparently, CAPM continues to be widely used (Bruner et al (1998), Graham and Harvey (2001) and Estrada (2011)), but this does not mean it has more explanatory power than the FFM. In this research we consider a sample of firms from ten countries of the European Monetary Union and use Griffin’s (2002) approach to evaluate both models. Fundamentally, he considers the FFM, that is, a model that expands CAPM by taking into account size and value factors in addition to the market risk factor of CAPM. Also, he expands on FFM by considering local, international, and global factors. In this research paper CAPM is used considering also local, international, and global risk premium. We analyze which has a greater explanatory power, considering the functional forms described above: Are local

T

P. Alves | IJBFR ♦ Vol. 7 ♦ No. 2 ♦ 2013

80

CAPMs better than local FFMs?; Do international factors grant any reliability?; Do global factors add any value?. The comparison between the two models is the main contribution of this research. This paper proceeds as follows. The next section presents the relevant literature and develops the scope of this research-paper. Subsequently we describe the data and methodology and discuss the results of our empirical tests. The final section concludes. LITERATURE REVIEW Since the 1960s, Sharpe (1964) and Lintner’s (1965) CAPM has been the most favored model to evaluate financial assets. However, since the 1970s and particularly in the 1980s, authors have identified many misspecifications of the CAPM. Basu (1977) found a positive connection between expected stock returns and earnings to price ratio. Banz (1981) concluded that small firms have, on average, higher risk-adjusted returns than large firms. Bhandari (1988) showed a positive connection between debt to equity and expected stock returns, even when controlling such variables as systematic risk, firm size and the January effect. Chan et al (1991), analyzing the relationship between expected stock returns and different fundamental variables, found a significantly positive impact on expected returns by market-to-book and cash flow yield. In the face of those misspecifications, Fama and French (1993), extended their 1992 research by developing an asset pricing model (FFM), in which the stock excess return is not only explained by market excess return, but also by two other variables: size (measured by market capitalization) and book-to-market ratio. We have two portfolios: a Small minus Big (SMB) portfolio and a High minus Low (HML) portfolio, depending respectively on market capitalization and book-to-market. Whereas book-to-market is related to financial distress problems, size is associated with profitability. Smaller stocks lead to lower earnings than larger stocks, and consequently to a higher expected return, after book-to-market’s control. On the other hand, book-to-market is related to financial distress problems. Firms with high book-to-market systematically present lower earnings on book equity, indicating signals of financial distress problems. The two factors have been criticized since the mid 1990s. Berk (1995) claims that size does not result from misspecification of CAPM, but it is a consequence of economic risk. If two firms have the same size at time t and consequently the same expected cash-flows at time t+1, the firm with higher risk will have a lower market value in that period; Lakonishok et al (1994) explain that high book-to-market stocks (or value stocks) do not present higher average returns than growth stocks as a reward for bearing a higher risk, but as a result of systematic mispricing by naive investors, who tend to extrapolate past earnings growth into the future, leading to under-pricing of value stocks and over-pricing of growth stocks. Fama and French (1998) expand the debate between growth and value stocks to thirteen major capital markets around the world. They find that in twelve markets - Italy is the exception - there is a value premium; moreover, they confirm that value stocks present higher returns than growth stocks and conclude that the world CAPM does not capture the referred premium, reasserting the CAPM misspecification. Still in the international field, Griffin (2002), resorting to the three-factor model of Fama and French (1993), compares that FFM, using country factors and global factors, and concludes that the former explains excess stock returns with more accuracy. Moerman (2005), using a similar approach to the one adopted in this research, but using monthly returns, concludes that the Local FFM outperforms the EMU FFM. It must be highlighted, however, that there is an important difference between both research papers. Whilst we debate the use of CAPM and FFM, he focuses solely on FFM. In fact, the main objective of this research paper is to evaluate the benefits of using FFM in comparison with CAPM. This field of investigation has also been done by many other authors: Bartholdy and Peare (2005) concluded that the small gain of using FFM in terms of explanatory time does not justify the work involved in calculating two more factors; Gharghori et al (2009), comparing the results of both models,


81

also concluded that the performance of the FFM is less than satisfactory in Australia; Vassalou (2003), considering a sample of 10 countries, concluded that FFM explains asset returns better than the CAPM; Kothari and Warner (2001), evaluating mutual fund performances, found that procedures based on the FFM are somewhat better than CAPM-based measures; Estrada (2011) considers that value and size matters and practitioners should understand and know how to apply the FFM. DATA AND METHODOLOGY Data was downloaded from Datastream (DS) and includes a significant number of firms from the following EMU members: Austria, Belgium, Finland, France, Germany, Ireland, Italy, the Netherlands, Portugal, and Spain. Luxembourg, a founding member, is excluded as a result of its small capital market. Greece was also not included because it only adopted the Euro currency at the beginning of 2001. Additionally, we also exclude (1) firms from the financial sector, since they have some capital requirements which offer them special features, and (2) firms whose book-to-market is negative, indicating some financial distress problems. This analysis focuses on the period from 1990 to 2003 and represents the debates, the arrangements and the results from the euro implementation. The period reflects the preliminary discussion of the single currency, from 1990 to 1995; a second period, from 1996 to 1998, characterized by many economic policies introduced by local countries in order to enter (in) the single currency; and finally, a period of four years of results from the euro implementation. Panel A of Table 1 reveals a stable market share among those countries in the period between 1990 and 2003. France and Germany are the biggest markets with more than half the EMU market capitalization. Italy, the Netherlands, and Spain are median-sized markets. Their market shares vary from 8% to 16%. Austria, Belgium, Finland, Ireland, and Portugal are the smallest markets. All of them present less than 5% of the EMU portfolio. Austria, Ireland, and Portugal, with less than 2%, are particularly small. Table 1: Sample Description by Countries

Panel A: Datastream Country Weights (%) AU BG FL FR GR IR IT NL PT SP 1.2 4.3 2.9 25.2 27.5 1.4 12.1 15.7 1.2 8.3

Panel B: Number of Firms AU BG FL FR GR IR IT NL PT SP 39 52 68 350 391 25 114 88 34 60

Panel C: Median Market Capitalization by Firm (€ millions) AU BG FL FR GR IR IT NL PT SP 45 131 112 65 63 97 148 145 47 262

Panel D: Median Book-to-Market by Firm AU BG FL FR GR IR IT NL PT SP 0.82 0.61 0.63 0.58 0.51 0.59 0.75 0.54 0.94 0.66

AU, BG, FL, FR, GR, IR, IT, NL, PT, and SP are respectively Austria, Belgium, Finland, France, Germany, Ireland, Italy, the Netherlands, Portugal, and Spain. Panel A shows Datastream country weights in the EMU portfolio. Panel B indicates the annual average number of firms by period used to build the size and the distress risk premiums. Panel C indicates the median size of firms in Panel B. Panels D indicates the median book-to-market of firms in Panel B. Panel B shows the number of firms by country. Germany and France respectively with 391 and 350 firms (61% of the sample), on an annual average, are the most represented countries. Austria, Ireland and Portugal are the least represented. For example, the sample only considers, on average, 25 Irish firms per year. The expressive representation of the biggest markets can be explained as a reaction to the development of some new markets, particularly the Neuer Market and the Nouveau Marché, the German and French regulated platforms, created respectively in 1997 and in 1996, which target young, small, and high growth stocks (e.g., technology, biotechnology, media and financial services stocks). The remaining countries created secondary markets with the same objective, although without the same success.


82

Panel C shows the median market capitalization by firm, downloaded from DS, during each sub-period. The large number of new firms in French and German stock markets caused a decrease in the median firm size. Spanish firms experienced an increase in their market capitalization, as a result of a comparatively lower increase in the number of firms. Austrian and Portuguese stock markets are characterized by a relatively large number of small firms. Book-to-market by firm, also downloaded from DS, is exhibited in Panel D. Austrian and Portuguese stocks present the highest median value for the book-to-market ratio. This paper uses the three-factor model of Fama and French (1993) with the adjustments adopted by Griffin (2002). The main objective is to clarify whether local or global factors are forces that might best explain stock returns. Basically, FFM is built using the following procedure: (i) We use weekly returns from January of 1990 to December of 2003; (ii) The market excess return (MER) is obtained through the difference between the stock market return and the risk free asset. Datastream (DS) stock market indices, German Deutschmarks denominated, are used as a proxy of local market return. DS indices were chosen because they represent, in general, more than 99% of local market value. Germany Euro one-month interest rate is used as the risk free asset; (iii) Stocks were classified by market capitalization in June of year t, using the sample median value, dividing them into Big (B) and Small (S) portfolios - delisted firms were also ranked, avoiding survivorship bias; (iv) Independently of (iii), the sample is divided into three groups of stocks (using the 30% and 70% percentiles), according to their book-to market, using the preceding values of December (year t-1) for that ratio, creating the high (H), medium (M), and low (L) book-to-market portfolios; (v) Portfolios are reallocated on an annual basis; (vi) Portfolios are value-weighted and amount to six: SL, SM, SH, BL, BM, and BH; (vii) Size premium is obtained, controlling the firm’s book-to-market, from the difference between S ((SL+SM+SH)/3) and B ((BL+BM+BH)/3), resulting in portfolio SMB (small minus big); (viii) Book-to-market equity is obtained, controlling the firm’s size, through the difference between H ((SH+BH)/2) and L ((SL+BL)/2), resulting in portfolio HML (high minus low). Next, following Griffin (2002), different functional forms of FFM, using either global or local factors or both, are reported. First, a model based on EMU factors is presented in German Deutschmarks: 𝑟𝑖,𝑡 = 𝛼𝑖 + 𝑏𝑖𝐸𝑀𝐸𝑅𝑡 + 𝑠𝑖𝐸𝑆𝑀𝐵𝑡 + ℎ𝑖𝐸𝐻𝑀𝐿𝑡 + 𝜀𝑖,𝑡 (1) in which 𝑟𝑖,𝑡 is the weekly excess stock return, 𝑏𝑖, 𝑠𝑖, and ℎ𝑖 are the unconditional sensitivities of asset i to the factors 𝐸𝑀𝐸𝑅𝑡, and, 𝐸𝑆𝑀𝐵𝑡, and 𝐸𝐻𝑀𝐿𝑡 represent the EMU factors. They are calculated considering the countries’ weight in the EMU portfolio, in which 𝐸𝑀𝐸𝑅𝑡 = 𝑊𝐷𝑡−1 𝐷𝑀𝐸𝑅𝑡 + 𝑊𝐹𝑡−1 𝐹𝑀𝐸𝑅𝑡. 𝑊𝐷𝑡−1 and 𝑊𝐹𝑡−1 are respectively the weight of local and foreign portfolios in the EMU portfolio in the week t-1. The same procedures for the size and distress premium are used. This research also considers an international model, based on local and international sensitivities: 𝑟𝑖,𝑡 = 𝛼𝑖 + 𝑏𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝑀𝐸𝑅𝑡 + 𝑠𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝑆𝑀𝐵𝑡 + ℎ𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝐻𝑀𝐿𝑡 +𝑏𝐹𝑖 𝑊𝐹𝑡−1 𝐹𝑀𝐸𝑅𝑡 + 𝑠𝐹𝑖 𝑊𝐹𝑡−1 𝐹𝑆𝑀𝐵𝑡 + ℎ𝐹𝑖 𝑊𝐹𝑡−1 𝐹𝐻𝑀𝐿𝑡 + 𝜀𝑖,𝑡 (2) in which 𝐷𝑀𝐸𝑅, 𝐷𝑆𝑀𝐵, 𝐷𝐻𝑀𝐿 are local factors and 𝐹𝑀𝐸𝑅 𝐹𝑆𝑀𝐵, and 𝐹𝐻𝑀𝐿 are international factors. Finally, a local model is shown, in which the international factors do not play any role: 𝑟𝑖,𝑡 = 𝛼𝑖 + 𝑏𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝑀𝐸𝑅𝑡 + 𝑠𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝑆𝑀𝐵𝑡 + ℎ𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝐻𝑀𝐿𝑡 + 𝜀𝑖,𝑡 (3) Thus, if model (2) does not grant any explanatory power to model (3), there are signs suggesting that the excess return is fundamentally explained by local factors.


83

EMPIRICAL RESULTS The analysis debates the results of CAPM versus FFM considering different portfolios (High, Low, Small, and Big). The absolute value of the intercept or Jensen’s alpha, meaning the pricing error, and the adjusted R², indicating explanatory power are used to evaluate the robustness of each model. The discussion is carried out based on the following procedures: First, market excess return, size, and distress risk premiums, which are used in the local FFM application, are presented (see table 2); Secondly, the results obtained for local, international, and EMU CAPM models, using High, Low, Small, and Big portfolios, ranked by quintiles, are confronted (see tables 3 and 4); Thirdly, the previous exercise is repeated, now considering FFM, in order to assess how accurate the models based on EMU factors are and to evaluate how size and financial distress international premiums increase the accuracy of FFM (see tables 5 and 6); and finally, we compare CAPM with FFM specifications in order to evaluate how useful both models are. Table 2 shows the weekly local market risk premium or domestic market excess return (DMER), size (DSMB), and distress risk premium (DHML) by country from 1990 to 2003. Table 2: Descriptive Statistics of Variables

DMER SMB HML Mean Stdev Mean Stdev Mean Stdev Austria -0.025 1.707 0.067 2.311 0.234 2.572 Belgium 0.000 2.436 0.128 2.026 0.050 2.258 Finland 0.220 4.538 0.148 2.810 -0.102 3.207 France 0.045 2.770 0.281 3.163 0.144 3.479 Germany 0.010 2.724 0.140 2.044 0.259 1.880 Ireland 0.090 2.615 0.055 2.878 0.056 2.917 Italy 0.018 3.131 0.108 2.059 0.000 2.262 Netherlands 0.055 2.545 0.097 2.056 -0.004 2.178 Portugal -0.017 2.352 0.078 2.394 -0.092 2.677 Spain 0.080 2.699 0.014 2.236 0.249 2.365 E.M.U. 0.053 2.433 0.142 1.735 0.120 1.638

Domestic market excess return (DMER) is obtained, considering a DS country indices and Germany Euro-mark one month, as proxies for market return and risk-free-asset. Small minus big (DSMB) is the return difference between S (small firms) and B (big firms) domestic portfolios. High minus low (DHML) is the return difference between H (high book-to-market firms) and L (low book-to-market firms) domestic portfolios. EMU results are value-weighted. Variables are weekly means, calculated on a value-weighted basis. Results are a weekly percentage. Finland, with a 0.22% (1.59%) of weekly (annual) risk premium presents the highest value. The remaining countries present weaker results. Some of the smallest markets present the poorest performance. Austria, Belgium, and Portugal present a weekly DMER of -0.025%, 0.000%, and -0.017% respectively. The equity risk premium varies from 0.010% (Germany) to 0.080% (Spain). These figures are abnormally low, when compared with the traditional results for equity risk premium. Damodoran (1992), advises an annual equity risk premium of 4.5%-5.5%, for developed markets with limited listings, and 3.5%-4.0% for Germany. However, some facts, namely the tech bubble in the late 1990s and ensuing fall in early 2000, offer a valuable explanation for such trend. Size premium (DSMB) reveals, in line with DMER, a consistent behavior across European countries. It is possible to observe signs of the existence of that type of premium. Size premium is particularly high in Finland, France, and Germany. In the French case, the difference between Small and Big portfolios excess return is 0.281%, on a weekly basis. The emergence of platforms for small and medium enterprises in these countries, particularly France and Germany, may be an explanation for what happened. Relating to book-to-market premium (DHML), our results are less consistent than those obtained by Fama and French (1998). They find a book-to-market premium in 11 out of 12 stock markets of their own sample, whereas we only find the book-to-market premium in 6 - Austria, Belgium, France, Germany, Ireland, and Spain – out of the 10 stock markets analyzed. Previous findings are the result of institutional and economic changes that European capital markets witnessed after the single currency. In fact, the


84

single currency and subsequent lower interest rates, in addition to the high-tech euphoria during the second half of the 1990s, can explain the stock price behavior of growth firms. In all probability, asset managers did not use the more advisable fundamentals, namely the cost of equity of growth firms. They estimated a lower cost of equity for growth stocks, which substantially increased their market prices; that is, asset managers used a lower estimate for cost-of-equity, creating the ideal conditions for stock prices to overreact. That explains what happened after 1999, a sustainable correction of stock market throughout this period of time, which would end in 2002. Tables 3 and 4 show how international factors in comparison with local factors, regardless of the type of stock, have a small role in the reliability of some estimates of the CAPM. The introduction of international factors sometimes deteriorates the estimates. This can be observed in Table 3 for high book-to-market stocks (Jensen’s alpha increases, in average, from 0.324 to 0.325 with the inclusion of the international factor). However, this result is not extensible to all countries. For example, Austria benefits from the inclusion of international factors. Table 3: Excess Returns of High and Low Portfolios using CAPM

Local CAPM International CAPM EMU CAPM |α| Adj. R2 |α| Adj. R2 |α| Adj. R2

High Austria 0.382 0.229 0.371 0.235 0.333 0.084 Belgium 0.293 0.387 0.288 0.389 0.259 0.269 Finland 0.335 0.080 0.352 0.165 0.363 0.163 France 0.455 0.444 0.453 0.444 0.440 0.397 Germany 0.404 0.350 0.409 0.351 0.380 0.277 Ireland 0.504 0.046 0.504 0.045 0.534 0.016 Italy 0.013 0.589 0.008 0.591 0.038 0.370 Netherlands 0.052 0.236 0.068 0.268 0.076 0.265 Portugal 0.482 0.009 0.476 0.012 0.481 0.012 Spain 0.319 0.409 0.319 0.408 0.360 0.282

Mean 0.324 0.278 0.325 0.291 0.326 0.213 Low Austria 0.029 0.434 0.034 0.436 0.073 0.121 Belgium 0.162 0.532 0.163 0.532 0.130 0.323 Finland 0.207 0.743 0.215 0.750 0.420 0.336 France 0.185 0.808 0.185 0.808 0.167 0.706 Germany 0.095 0.520 0.073 0.548 0.057 0.531 Ireland 0.242 0.338 0.245 0.382 0.278 0.284 Italy 0.125 0.685 0.120 0.688 0.097 0.515 Netherlands 0.066 0.701 0.067 0.701 0.099 0.575 Portugal 0.186 0.554 0.182 0.564 0.210 0.241 Spain 0.044 0.525 0.040 0.530 0.000 0.413

Mean 0.134 0.584 0.132 0.594 0.153 0.404 t-stat 3.27*** 3.33*** 2.70**

High and Low portfolios excess returns are dependent variables representing the top and bottom quintile. Variables are value-weighted, calculated on a weekly basis. DS country indices are used as a local market proxy. Germany Euro-Mark one-month is the risk-free asset proxy. The method of estimation is ordinary least squares, using the Newey and West (1987) covariance estimator that is consistent in the presence of both heteroskedasticity and autocorrelation of unknown form. Domestic Model is a result of regression 𝑟𝑖,𝑡 = 𝛼𝑖 + 𝑏𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝑀𝐸𝑅𝑡+𝜀𝑖,𝑡, where ri,t is the portfolio (High or Low) excess return in period t, DMER is the domestic excess return, and αi is a constant. wDt-1 is the weight of a local portfolio in EMU. b Di is the unconditional sensitivity of asset i to the factor. EMU Model is a result of regression: 𝑟𝑖,𝑡 = 𝛼𝑖 +𝑏𝑖𝐸𝑀𝐸𝑅𝑡 + 𝜀𝑖,𝑡, where EMERt represents the EMU factor. It is also calculated using a value-weighted basis. 𝐸𝑀𝐸𝑅𝑡 = 𝑊𝐷𝑡−1 𝐷𝑀𝐸𝑅𝑡 +𝑊𝐹𝑡−1 𝐹𝑀𝐸𝑅𝑡, where wDt-1 and wFt-1 are respectively the weight of local and foreign portfolios in the EMU portfolio in the week t-1. International Model is the result of regression: 𝑟𝑖,𝑡 = 𝛼𝑖 + 𝑏𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝑀𝐸𝑅𝑡 + 𝑏𝐹𝑖 𝑊𝐹𝑡−1 𝐹𝑀𝐸𝑅𝑡 + 𝜀𝑖,𝑡. *, **, and ***, indicate significance at the 10, 5 and 1 percent level. On the other hand, when the model is based on EMU factors, the reliability of the CAPM is considerably worse, particularly for big and low book-to-market stocks. The Jensen’s alpha of EMU CAPM increases


85

in comparison with Local CAPM from 0.134 to 0.153 for low book-to-market stocks (see Table 3). For big stocks it changes from 0.111 to 0.130 (see Table 4). The CAPM model, regardless of its functional form, presents better adherence to big and low book-to-market stocks. The results presented in Tables 3 and 4 show a lower Jensen’s alpha for that type of stocks rather than small and high book-to-market stocks, with statistical significance (see table 3 and 4, particularly t-stat). For example, for local model low-book-to-market firms with 0.134 of Jensen’s alpha compares with 0.324 for high-book-to-market firms (t-stat = 3.27). Table 4: Excess Returns of Small and Big Portfolios Using CAPM

Local CAPM International CAPM EMU CAPM |α| Adj. R2 |α| Adj. R2 |α| Adj. R2

Small Austria 0.362 0.062 0.363 0.061 0.335 0.010 Belgium 0.255 0.084 0.252 0.085 0.243 0.064 Finland 0.570 0.037 0.581 0.062 0.596 0.060 France 0.530 0.254 0.522 0.271 0.520 0.269 Germany 0.654 0.123 0.653 0.122 0.642 0.106 Ireland 1.250 0.011 1.252 0.011 1.270 0.009 Italy 0.370 0.205 0.368 0.204 0.353 0.146 Netherlands 0.233 0.204 0.240 0.218 0.249 0.210 Portugal 0.927 0.010 0.919 0.013 0.930 0.011 Spain 0.599 0.118 0.603 0.119 0.631 0.095

Mean 0.575 0.111 0.575 0.116 0.577 0.098 Big Austria 0.092 0.768 0.085 0.774 0.037 0.224 Belgium 0.067 0.729 0.071 0.732 0.037 0.402 Finland 0.155 0.769 0.165 0.783 0.338 0.384 France 0.167 0.954 0.168 0.954 0.149 0.817 Germany 0.109 0.907 0.098 0.917 0.068 0.830 Ireland 0.187 0.621 0.188 0.624 0.248 0.295 Italy 0.061 0.835 0.065 0.836 0.036 0.538 Netherlands 0.065 0.905 0.064 0.905 0.104 0.701 Portugal 0.135 0.488 0.135 0.487 0.162 0.152 Spain 0.069 0.790 0.068 0.790 0.120 0.532

Mean 0.111 0.777 0.111 0.780 0.130 0.487 t-stat 4.60*** 4.62*** 4.17***

Small and Big portfolios excess returns are dependent variables. The functional forms are similar to those applied in table 3. *, **, and ***, indicate significance at the 10, 5 and 1 percent level The inclusion of international factors in FFM, contrarily to CAPM, improves the reliability of the model. For all kinds of stocks, the introduction of international factors reduces the Jensen’s alpha (see tables 5 and 6). For example, the Jensen’s alpha of small stocks decreases from 0.358 to 0.307 (see table 6). Also for FFM does a model based on EMU factors deteriorate the reliability of the estimates in comparison with their peers. Big stocks are the exception as the Jensen’s alpha is lower than that obtained for local and international FFM (see table 6) Unlike the CAPM, the reliability of estimates for high and low book-to-market stocks is constant using FFM (see table 5, namely the t-stat’s). The same does not occur when big and small stocks are taken into account. In this case, the Jensen’s alpha of big socks is considerably lower, with statistical significance, than the one observed for small stocks (see table 6). In Table 7 we compare CAPM and FFM for all stocks by functional form. In the case of the local form, the FFM only presents a lower Jensen’s alpha, with statistical significance, for small stocks (t-stat = 1.77).


86

For the international form, an improvement is observed in high and small stocks. Relatively to EMU factors, there is no evidence, with statistical significance, of more reliability using FFM, independently of the type of stock considered. Table 5: Excess Returns of High and Low Portfolios Using FFM

Local FFM International FFM EMU FFM |α| Adj. R2 |α| Adj. R2 |α| Adj. R2

High Austria 0.279 0.340 0.226 0.348 0.230 0.104 Belgium 0.289 0.526 0.224 0.538 0.190 0.307 Finland 0.266 0.369 0.215 0.421 0.269 0.197 France 0.452 0.497 0.361 0.524 0.377 0.440 Germany 0.157 0.633 0.167 0.635 0.236 0.388 Ireland 0.325 0.208 0.283 0.213 0.394 0.050 Italy 0.032 0.731 0.073 0.734 0.173 0.412 Netherlands 0.055 0.459 0.026 0.485 0.017 0.308 Portugal 0.055 0.593 0.138 0.605 0.247 0.051 Spain 0.182 0.549 0.147 0.552 0.254 0.323

Mean 0.212 0.484 0.190 0.500 0.239 0.258 Low Austria 0.027 0.497 0.004 0.499 0.164 0.143 Belgium 0.183 0.601 0.159 0.604 0.121 0.322 Finland 0.244 0.748 0.259 0.757 0.528 0.380 France 0.177 0.834 0.210 0.842 0.233 0.747 Germany 0.173 0.585 0.089 0.638 0.022 0.596 Ireland 0.271 0.366 0.233 0.418 0.210 0.309 Italy 0.154 0.768 0.121 0.786 0.070 0.554 Netherlands 0.066 0.723 0.065 0.725 0.113 0.575 Portugal 0.194 0.559 0.165 0.581 0.134 0.278 Spain 0.050 0.588 0.009 0.621 0.009 0.454

Mean 0.154 0.627 0.131 0.647 0.160 0.436 t-stat 1.10 1.30 1.36

High and Low portfolios excess returns are dependent variables. They represent the top and bottom quintile. Variables are value-weighted, calculated on a weekly basis. DS country indices are used as local market proxy. Germany Euro-Mark one-month is the risk-free asset proxy. The method of estimation is ordinary least squares, using the Newey and West (1987) covariance estimator that is consistent in the presence of both heteroskedasticity and autocorrelation of unknown form. Domestic Model is a result of regression 𝑟𝑖,𝑡 = 𝛼𝑖 + 𝑏𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝑀𝐸𝑅𝑡 + 𝑠𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝑆𝑀𝐵𝑡 +ℎ𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝐻𝑀𝐿𝑡 + 𝜀𝑖,𝑡, where ri,t is the portfolio (High or Low) excess return in period t, DMER is the domestic excess return, DSMB is the return difference between S (local small firms) and B (logal big firms), DHML is the return difference between H (high book-to-market firms) and L (low book-to-market firms), and αi is a constant. wDt-1 is the weight of a local portfolio in EMU. b Di, sDi,, and h Di are the unconditional sensitivities of asset i to the factors. EMU Model is a result of regression: 𝑟𝑖,𝑡 = 𝛼𝑖 + 𝑏𝑖𝐸𝑀𝐸𝑅𝑡 +𝑠𝑖𝐸𝑆𝑀𝐵𝑡 + ℎ𝑖𝐸𝐻𝑀𝐿𝑡 + 𝜀𝑖,𝑡 , where EMER, ESMB, and EHML represent EMU factors. They are calculated considering the countries weight in the EMU portfolio. Thus, we have, for example, 𝐸𝑀𝐸𝑅𝑡 = 𝑊𝐷𝑡−1 𝐷𝑀𝐸𝑅𝑡 +𝑊𝐹𝑡−1 𝐹𝑀𝐸𝑅𝑡, where wDt-1 and wFt-1 are respectively the weight of local and foreign portfolios in the EMU portfolio in the week t-1. International Model is the result of regression: 𝑟𝑖,𝑡 = 𝛼𝑖 + 𝑏𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝑀𝐸𝑅𝑡 +𝑠𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝑆𝑀𝐵𝑡 + ℎ𝐷𝑖 𝑊𝐷𝑡−1 𝐷𝐻𝑀𝐿𝑡 + 𝑏𝐹𝑖 𝑊𝐹𝑡−1 𝐹𝑀𝐸𝑅𝑡 + 𝑠𝐹𝑖 𝑊𝐹𝑡−1 𝐹𝑆𝑀𝐵𝑡 +ℎ𝐹𝑖 𝑊𝐹𝑡−1 𝐹𝐻𝑀𝐿𝑡 + 𝜀𝑖,𝑡.*, **, and ***, indicate significance at the 10, 5 and 1 percent level.


87

Table 6: Excess Returns of Small and Big Portfolios using FFM Local FFM International FFM EMU FFM

|α| Adj. R2 |α| Adj. R2 |α| Adj. R2 Small Austria 0.148 0.269 0.079 0.270 0.183 0.030 Belgium 0.223 0.218 0.186 0.226 0.172 0.096 Finland 0.317 0.496 0.289 0.501 0.413 0.141 France 0.423 0.400 0.339 0.457 0.371 0.398 Germany 0.473 0.334 0.424 0.351 0.476 0.240 Ireland 0.882 0.310 0.885 0.307 1.175 0.019 Italy 0.153 0.561 0.104 0.574 0.097 0.276 Netherlands 0.174 0.336 0.122 0.353 0.121 0.280 Portugal 0.299 0.311 0.120 0.326 0.567 0.052 Spain 0.487 0.340 0.523 0.341 0.498 0.113

Mean 0.358 0.357 0.307 0.371 0.407 0.164 Big Austria 0.125 0.803 0.113 0.809 0.031 0.256 Belgium 0.087 0.779 0.084 0.780 0.045 0.409 Finland 0.173 0.781 0.163 0.794 0.399 0.409 France 0.170 0.955 0.177 0.955 0.201 0.828 Germany 0.152 0.916 0.144 0.924 0.080 0.830 Ireland 0.286 0.694 0.246 0.701 0.171 0.311 Italy 0.101 0.850 0.078 0.854 0.006 0.542 Netherlands 0.082 0.918 0.075 0.918 0.135 0.706 Portugal 0.148 0.487 0.109 0.497 0.076 0.189 Spain 0.106 0.807 0.073 0.818 0.106 0.539

Mean 0.143 0.799 0.126 0.805 0.125 0.502 t-stat 2.92** 2.21** 2.63**

Small and Big portfolios excess returns are dependent variables. The functional forms are similar to those applied in table 5. *, **, and ***, indicate significance at the 10, 5 and 1 percent level

Table 7: Excess Returns of All Portfolios using FFM and CAPM

High Low Small Big Local CAPM 0.324 0.134 0.575 0.111 Local FFM 0.209 0.154 0.358 0.143 t-stat 1.67 -0.57 1.77* -1.33 International CAPM 0.325 0.132 0.575 0.111 International FFM 0.186 0.131 0.307 0.126 t-stat 2.27** 0.03 2.10* -0.65 EMU CAPM 0.326 0.153 0.577 0.130 EMU FFM 0.239 0.160 0.407 0.125 t-stat 1.43 -0.11 1.18 0.10

This table compares the Jensen’s alpha of CAPM and FFM, considering different perspectives: Local, International and EMU models .*, **, and ***, indicate significance at the 10, 5 and 1 percent level. CONCLUSION This research compares the reliability of CAPM and FFM, considering a sample of firms from ten countries of the European Monetary Union, between January 1990 and December 2003. We have resortedto Griffin’s (2002) approach to evaluate both models. Fundamentally, he expands on FFM considering local, international and global factors. In this research, CAPM is used also considering local, international, and global risk premium and its comparison with the different functional forms of FFM, previously described, is its main contribution. Our goal has been to shed some light on which functional forms of both models have more explanatory power and to answer the questions that we put forward at the outset of our research.


88

First, models based on EMU factors, a global functional form, independently of any CAPM or FFM consideration, produce the worst results. In fact, the reliability of such models is negligible when compared with the other functional forms. Secondly, independently of the functional form of both models, the expected return of big and low book-to-market stocks is more reliable. This is particularly observable for big firms. Finally, FFM is notoriously preferable in comparison with CAPM for small and high-book to market firms: in this case, the introduction of international factors increases the reliability of expected returns. Summing up (After careful consideration of all these different scenarios), we advise using FFM for small and high book to market firms. The nature of the sample, with very different firms from very different countries, limits the reliability of size and financial distress premiums, but it would be worse to solely consider firms from France and Germany. In that case, we would only draw conclusions regarding those two countries, far from being international evidence. In the future, we plan to introduce winner minus loser premium to FFM and, resorting to Griffin’s (2002) approach, develop a study very similar to this one. REFERENCES Banz, R. (1981) “The Relation between Return and Market Value of Common Stocks,” Journal of Financial Economics, vol. 9, p. 3-18 Basu, S. (1977) “Investment Performance of Common Stocks in Relation to their Price-Earnings: A Test of the Efficient Market Hypothesis,” Journal of Finance, vol. 32, p. 663-682 Bartholdy, Jan & Peare, P (2005) “Estimation of Expected Return: CAPM vs. Fama and French,” International Review of Financial Analysis, vol. 14 (4), p. 407-427 Berk, J. (1995) “A Critique of Size-Related Anomalies,” Review of Financial Studies, vol. 8, p. 275-286 Bhandari, L. (1988) “Debt/Equity Ratio and Expected Common Stock Returns: Empirical Evidence,” Journal of Finance, vol. 43, p. 507-528 Bruner, R. F., K. Eades, R. Harris & R. Higgins (1998) “Best Practices in Estimating the Cost of Capital: Survey and Synthesis”. Financial Practice and Education, vol. 8(1), p. 13-28 Chan, L., Y. Hamao, & Lakonishok, J. (1991) “Fundamentals and Stock Returns in Japan,” Journal of Finance, vol. 46, p. 1739-1764 Damodaran, A. (1992) Investment valuation (John Wiley). Estrada, J. (2011) “The Three-Factor Model: A Practitioner's Guide,” Journal of Applied Corporate Finance, vol. 23(2), p. 77-84 Fama, E. F. & French, K.R. (1992) “The Cross Section of Expected Stock Returns,” Journal of Finance, vol. 47, p. 427-465 Fama, E. & French, K. (1993) “Common Risk Factors in the Returns on Stocks and Bonds,” Journal of Financial Economics, vol. 33, p. 3-56


89

Fama, E., & French, K. (1998) “Value versus Growth: The International Evidence,” Journal of Finance, vol. 53, p. 1975-1999 Gharghori, P., Lee, R. & Veeraraghavan, M. (2009) “Anomalies and Stock Returns: Australian Evidence,” Accounting and Finance, vol. 49(3), p. 555-576 Graham, J. R. & Harvey, C. R. (2001) “The Theory and Practice of Corporate Finance: Evidence from the Field”. Journal of Financial Economics. vol. 60(1-2), p. 187-243 Griffin, J. (2002) “Are the Fama and French Factors Global or Country Specific?,” Review of Financial Studies, vol. 15, p. 187-243 Kothari, S. & Warner, J. (2001) “Evaluating Mutual Fund Performance,” Journal of Finance, vol. 56, p. 1985-2010 Lakonishok, J., Shleifer, A. & Vishny, R. (1994) “Contrarian Investment, Extrapolation, and Risk,” Journal of Finance, vol. 49, p. 1541-1578 Lintner, J. (1965), “The Valuation of Risk Assets and the Selection of Risky Investments in Stock Portfolios and Capital Budgets,” Review of Economics and Statistics, vol. 47, p. 13-37 Moerman, G. (2005) “How Domestic is the Fama and French Three-Factor Model? An Application to the Euro Area, Working Paper,” Erasmus Research Institute of Management Newey, W. & West, K. (1987) “A Simple, Positive Semi-Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix,” Econometrica, vol. 55, p. 703-708 Sharpe, W. (1964) “Capital Asset Prices: A Theory of Capital Market Equilibrium Under Conditions of Risk,” Journal of Finance, vol. 19, p. 425-442 Vassalou, M., (2003) "News Related to Future GDP Growth as a Risk Factor in Equity Returns,” Journal of Financial Economics, vol. 68, p. 47-73 ACKNOWLEDGMENTS We thank the Journal editors, Terrance Jalbert and Mercedes Jalbert, two anonymous referees, Rui Alpalhão, Paulo Francisco, and Rita Pires dos Reis. BIOGRAPHY Dr. Paulo Alves is a Senior Financial Economist at the Portuguese Securities and Exchange Commission (CMVM). He is also an Assistant Professor of Finance at Lisbon Accounting and Management Institute (ISCAL) and Lusofona University. Numerous articles stemming from his research have been published in journals such as the Journal of Multinational Financial Management, the Applied Financial Economics, the International Research Journal of Finance and Economics, the ICFAI Journal of Applied Finance and the Portuguese Journal of Management Studies. He can be contacted at: Lisbon Accounting and Management Institute (ISCAL), Av. Miguel Bombarda, no. 20 1069 - 035 Lisbon. Phone: 00351 91 7308049. Email: [email protected]


91

QUALITY OF GOVERNANCE AND THE MARKET VALUE OF CASH: EVIDENCE FROM SPAIN

Eloisa Perez-de Toledo, MacEwan University Evandro Bocatto, MacEwan University

ABSTRACT

We examine the value shareholders attribute to one euro of extra cash held by Spanish firms and how corporate governance impacts this value by comparing the value of cash for companies with good and poor governance. The results show that one euro of extra cash is valued at a considerable premium at companies with good governance. Moreover, the presence of future growth opportunities intensifies this effect. Our results also suggest that the conflict between shareholders and debt holders is more severe in Spain than in the U.S. as investors apply a stronger discount for leverage when valuing Spanish firms. JEL: G11, G34 KEYWORDS: Corporate governance, cash holdings, firm value, future growth opportunities, leverage. INTRODUCTION

he economic relevance of cash holdings has increased in recent years. Cash and cash equivalents represented 17% of total assets held by corporations worldwide in 2007 (Ammann, Oesch, and Schmid (2010)). Chang and Noorbakhsh (2009) show a consistent increase in cash holdings from

9% of total assets in 1985 to 17% in 2004. How firms use cash and cash equivalents impacts firms’ performance and affects firms’ market value (Dittmar and Mahrt-Smith (2007), Acharya, Almeida, and Campello, (2007), Almeida, Campello, and Weisbach (2004)). However, cash and cash equivalents can be diverted by bad-intentioned managers to appropriate themselves of high liquidity level benefits (Jensen (1986) and Myers and Rajan (1998)). Efficient corporate governance structures mitigate these problems, as they provide the necessary mechanisms for controlling and monitoring firms’ use of cash reserves. The present study aims to provide empirical evidence on the interaction between the quality of governance and the market value of cash of Spanish publicly traded firms. The relevance of Spain lies on a series of factors: the relatively early-stage of development of its financial markets (Demirguc-Kunt and Maksimovic (1998)), the participation of individual investors is amongst the lowest in Europe, as a consequence the banking sector is of great importance in financing firms, family businesses or family controlled businesses compound the majority of listed firms, hence highly concentrated ownership structures are the rule. As Ocaña, Peña and Robles (1997) show, the market for corporate control in Spain is still incipient, for that it is not an import governance mechanism as in Anglo-Saxon countries. In fact, the percentage of hostile takeovers registered in the Spanish market (e.g. 4%, Fernandez and Gomez-Anson (1999)) is not comparable to the ones in major markets such as the U.S. (e.g. 47%, Cotter, Shivdasani and Zenner (1997)) or the U.K. (e.g. 25%, Franks and Mayer (1996)). As external governance mechanisms are rare, the main governance mechanism is the concentration of ownership (Leech and Manjón (2002)). Some of these characteristics of Spain are common features of other Western European countries. Therefore the present study contributes to the literature on the role of governance structures in the valuation of cash holdings in Europe. Using a sample of 98 non-financial Spanish firms listed in the Madrid Stock exchange in the period between 2003 and 2007, we find that the same euro of cash has more value in companies with good governance and in companies with future growth opportunities. The results show that shareholders value

T

E. Perez de Toledo & E. Bocatto | IJBFR ♦ Vol. 7 ♦ No. 2 ♦ 2013

92

an extra euro of cash at as high as €1.02 for good governance firms and as low as €0.57 for companies with poor governance. The presence of future growth opportunities accentuates this difference. As firm value is determined in part by how investors expect cash to be used, the presence of future growth opportunities acts as a moderator in the relationship between firms’ quality of governance and the market value of cash. The results indicate that investors reward companies that accumulate cash with the objective of investing in positive net present value (NPV) projects in the near future, and penalize firms that hoard cash simply to increase management’s discretion without any investment perspective. Moreover, we find that the negative effect of leverage in the valuation of the companies is more pronounced in Spain than in the U.S. Our results indicate that the value of one extra euro of cash for an all-equity financed Spanish firm is 49 cents higher than for a firm with 10% leverage. In the U.S. the discount applied for leverage is of only 14 cents of dollar, as reported by Faukender and Wang (2006). We interpret this result as evidence that the conflict of interest between shareholders and debt holders is more severe in Spain than in the U.S., as previously proposed by de-Miguel and Pindado (2001). The paper is organized as follows. The next section offers a revision of the literature and presents the hypotheses. Part 3 describes the data and the main methodological approach used in the study. Part 4 analyses the results and, finally, Part 5 concludes the paper. LITERATURE REVIEW In classical valuation models, cash is defined as ‘negative debt’ since cash balances are used exclusively to pay back debt. Therefore only net leverage is relevant to firm value. This approach has important and restrictive assumptions, such as that raising new capital is costless and frictionless. However, recent studies have shown a different reality where cash assumes a central role in firms’ financial strategy. Acharya, Almeida and Campello (2007) propose a theory of cash–debt substitutability and identify a hedging motive behind financially constrained firms’ cash and debt management, indicating that cash and debt are used optimally by firms depending on their free cash flow generation and access to credit. Moreover, Almeida, Campello and Weisbach (2004) and Denis and Sibilkov (2010) show that financially constrained firms save a portion of their cash flow to hedge future investment against income shortfalls, an indication that cash plays a central role in firms’ financial strategy as it is used as a hedge against future shortages of capital. Besides the hedging motive, there are two classical motives for holding cash: the transaction motive and the precautionary motive. The transaction motive assumes that a certain level of cash holdings is required to support the day-to-day activities of the firm and that cash cannot be raised instantaneously, thus firms hold a certain level of cash to meet their cash flow needs (Frazer (1964) and Keynes (1936)). The precautionary motive to hold cash states that firms accumulate precautionary financial slack in anticipation to new investment opportunities when external finance is costly (Myers and Majluf (1984)). The problem with the precautionary motive is that shareholders may want the firm to distribute all surpluses of corporate liquidity avoiding thus the possibility of cash being invested in low (or even negative) yield investment opportunities (i.e. poorly performing mergers and acquisitions, as described by Kim, Mauer and Sherman (1998)). The transaction motive is related to the trade-off theory, as the transaction costs to raise external funds sometimes exceed the rate of return of the firm’s pool of projects. It takes firms to reject investment opportunities that they would otherwise accept. Therefore, the two main benefits of holding cash are to avoid the transaction costs associated with the issue of new finance, and to use cash to finance activities and investments when other sources of capital are not available. Transaction costs are alleged to be the major determinant of the level of cash holdings. If the marginal cost of raising one euro of cash is too high, firms would prefer to hold more cash than firms facing lower transaction costs when raising external capital. Similarly, large firms face lower transaction costs compared to small firms, for this reason it is expected that big companies hold less cash than small


93

companies. In Spain, the empirical evidence on the transaction costs of new issues is provided by De-Miguel and Pindado (2001) who show that Spanish firms bear considerable transaction costs when they decide to adjust their debt ratio in the previous period to their target level in the current period. They also find an inverse relationship between debt and cash flow. They argue that this inverse relationship arises in the presence of asymmetric information and could take firms to face the underinvestment problem. They interpret this result as evidence of the pecking order (Myers and Majluf, 1984). For this reason, firms with good investment opportunities would be better off by creating financial slack (accumulating cash) in the previous period to finance future investment opportunities in the next. Cash Holdings and Governance The extant literature on the valuation of cash holdings and its relationship with governance shows mixed results. Harford (1999) finds that firms with excess cash spend more on acquisitions and Harford, Mansi and Maxwell (2008) extend his results by adding corporate governance and show that poorly governed firms dissipate more cash in acquisitions. On the other hand, Opler, Pinkowitz, Stulz and Williamson (1999) and Mikkelson and Partch (2003) do not find evidence that poor governance firms hold more cash. The underlying assumption of these studies is that there is an ‘optimal’ level of cash and managers often deviate from this level for a reason, the precautionary motive or the transaction motive. More recently, some studies have taken a different perspective. Instead of assuming a pre-established ‘optimal’ level of cash, the authors examine how managers spend it, and empirically assess the value shareholders attribute to one euro of cash holdings and what factors determine its value (Acharya, Almeida and Campello, 2007, Dittmar and Mahrt-Smith, 2007, Faulkender and Wang , 2006, Opler, Pinkowitz, Stulz and Williamson, 1999, Pinkowitz, Stulz and Williamson, 2006 and Pinkowitz and Williamson, 2004). Almeida, Campello and Weisbach (2004) show firms with greater frictions in raising outside financing save a greater portion of their cash flow as cash than those with fewer frictions. Recent studies by Faulkender and Wang (2006) and Pinkowitz and Williamson (2004) report evidence consistent with the view that cash holdings are more valuable for constrained than unconstrained firms. Collectively, these studies support the view that higher cash holdings are more valuable for financially constrained firms. An alternative view is that high cash holdings increase agency problems in constrained firms. The evidence on this view is also mixed. Harford (1999) and Dittmar, Mahrt-Smith and Servaes (2003) provide support for the hypothesis that cash hoarding by firms is value reducing and can be a result of agency problems inside corporations. Mikkelson and Partch (2003) argue that a policy of high cash holdings is not necessarily value reducing and may be an operating necessity. Other studies show that cash is associated with other corporate variables like firm value, bankruptcy risk and firm’s quality of governance (Attig, El Ghoul, Guedhami and Rizeanu (2011), Dittmar and Mahrt-Smith (2007) and Harford, Mansi and Maxwell (2008)). In this new context, corporate liquidity is not inversely related to debt, as proposed by the traditional view, but rather used as substitutes in the design of firms’ optimal financial policy. Empirical Predictions and Hypotheses Traditionally cash holdings are assumed to be zero NPV investments. Therefore one euro of cash should add one euro to firm’s market value. Nonetheless, information asymmetries, transaction costs, and taxes create a deviation from this hypothetical parity. As Myers and Majluf (1984) propose, financial slack has value because it allows firms to undertake positive NPV projects they would otherwise give up. When companies have good investment opportunities but the cost of new issues is prohibitive due to the signaling effect contained in new issues, the market value of one euro of cash should be higher than one. The main hypothesis of our study is that the value shareholders attribute to one euro of extra cash is determined by specific firm-characteristics, in particular the quality of the firm’s governance system and the presence of future growth opportunities. We expect to find cross-sectional differences in the market


94

value of cash holdings for companies with good and poor governance. Hence, we analyze the interaction between cash and governance and its impact on the market value of Spanish companies. While governance at the firm level changes slowly, cash holdings experience a considerable variation over time, offering a statistically powerful test for measuring the impact of governance on the use and destination of cash flows (Chi (2005) and Dittmar and Mahrt-Smith (2007)). Next, we control for future growth opportunities and examine what is the impact of the presence of growth opportunities (as a proxy for future investment) on the market value of an extra euro of cash holdings for firms with good and poor governance. Therefore, we are interested in empirically testing the following hypotheses: H1: The market value of an extra euro of cash holdings is higher for companies with good quality governance. The same euro of cash is worth more in a firm with good quality of governance than in a firm with poor governance. Shareholders will value the same euro differently because agency costs are higher for the latter. Besides, in companies with bad quality of governance the problem of overinvestment can be more acute so it is expected that shareholders apply a discount for the level of cash companies already have on hand in the beginning of the year, that takes to the second hypothesis. H2: the value of an extra euro of cash holdings decreases with the level of the firm’s cash position in the beginning of the year. The third hypothesis is related to the conflict of interest that may arise between shareholders and debt holders. In firms with high leverage ratios, debt holders may capture most of the investment projects’ future benefits, so shareholders have an incentive to either underinvest or to take on overly risky projects causing the underinvestment and the asset substitution problems. For high leveraged firms the probability that debt holders will receive most of the cash flows generated by the new investments takes shareholders to apply a discount on financial leverage. Thus, while an increase in cash will produce an increase in the value of the firm, because debt holders will capture most of its value in firms with high leverage, we expect the coefficient for the interaction between cash and leverage to be negative, as stated in the following hypothesis: H3: An extra euro of cash holdings is valued at a lower value by shareholders in companies with high leverage ratios. Finally, because financial slack may be valuable for firms with future growth opportunities, it is expected that, ceteris paribus, shareholders place a greater value on the same euro of cash holdings for firms with future growth opportunities. It takes to the formulation of the following hypothesis: H4: The value of an extra euro of cash holdings is higher for companies with future growth opportunities. DATA AND METHODOLOGY Our sample is composed by all publicly traded companies listed at the mercado continuo of the Madrid Stock Exchange from 2003 to 2007 for which data is available (however, as the model employed uses some lagged variables, the data collection starts in 2002). To be consistent with the previous literature we exclude the financial services industry where liquidity is determined mostly by regulatory agencies. The final sample is composed by 98 companies with 490 firm-year observations. The main data source is the Spanish Securities Exchange Commission and the Madrid Stock Exchange databases, and the Corporate Governance Report released by the companies. All financial and accounting data was obtained from the database OSIRIS publicly listed companies worldwide (Bureau Van Dijk). The data was collected annually at the end of each fiscal year. For missing data, firms’ annual financial reports were used.


95

A governance index (GOV-I) is constructed as a proxy for the quality of governance. The index was created based on a questionnaire with 25 binary objective questions. The questions were developed based on the recommendations of the Spanish Code of Best Practices proposed by the Olivencia and Aldama Committees. Spanish publicly traded companies are requested to release a Corporate Governance Annual Report since 2003. These reports are our main source of governance data, which is collected annually. The construction of the index is straightforward, we first code the 25 variables as 1 or 0 depending on whether the firm complies with a specific corporate governance standard or not. Each positive answer adds one point to the index, and the companies present a corporate governance level that ranges, in theory, from 0 to 25. The answers to all questions were obtained exclusively from secondary data, as the main objective was to measure companies’ degree of transparency and the easiness of access to any relevant governance information. The index is composed by four dimensions in order to assess good governance practices: (1) access and content of the information; (2) board structure; (3) ownership structure and control; and, (4) transparency. Appendix A provides the questions compounding the index. Table 1 reports summary statistics for the main variables used in the study. Panel A refers to all sample firms. On average, Spanish firms held 9.3% of total assets in cash and cash holdings between 2003 and 2007. However, as the median is 5.5% the distribution is right-skewed with the median firm holding about €45 million in cash. Regarding the dependent variable, the excess stock return, the median firm presents a –4.84% 1-year excess stock return while the mean is zero, which is consistent with excess return distributions that are right-skewed. The average sample firm has a debt-to-market value ratio of 20.2%. The median firm has sales growth (measured by the geometric average of the last three years) of about 16%, a considerably high ratio considering that Spain is a western European country and that our study is focused on large firms. An interesting phenomenon is related to sales growth. It has increased consistently over the sample period, from 6.8% in 2003 to as high as 23.3% in 2007, an indication of the booming period that preceded the 2008 financial crisis. Panel B reports the average cash holdings scaled by total assets by industry, as well as the logarithm of total assets, our proxy to firm size. Transport and Engineering services are the industries that held more cash during the sample period. Broadcast/Media, Construction/Concessions and Utilities are the sectors that held less cash. These sectors are traditionally composed by very large firms that normally do not face restrictions to raise external capital, which explains their lower levels of cash. Table 2 presents a correlation matrix for selected variables. Regression Specification To measure the value effects of corporate governance on cash resources, we use the model proposed by Faulkender and Wang (2006) and extended by Dittmar and Mahrt-Smith (2007) to include governance. The excess return for firm i during year t less the return of stock’s i benchmark portfolio, as defined by Fama and French (1993), is intended to be the measure of change in firm value. The dependent variable is the stock return and the independent variables are the change in cash, both by itself and its interaction with (1) the quality of governance (proxy by the GOV-I); (2) the lag value of cash (Ci,t-1); and, (3) the leverage ratio. The change in cash is normalized by beginning-of-period equity value in order to capture the euro (€) change in shareholder value resulting from one euro change in the amount of cash held by the firm. To determine the effect of governance, we allow for the interaction between the change in cash with the governance index (GOV-I) and other measures of governance used as a proxy of governance: board ownership and blockholdings (the sum of all shareholders with 5% or more ownership stake on the firm). Finally, the model includes variables that control for changes in profitability, investment and financing strategies (control variables). The method used for estimating the effect of cash holdings on firm value is the generalized least squares (GLS) regressions with random effects.


96

Table 1: Summary Statistics Panel A. All Sample Firms

Mean Median Standard

Minimum Maximum Cash/ Total Assets 2003 0.084 0.059 0.081 0.001 0.405 Cash/ Total Assets 2007 0.097 0.056 0.126 0.001 0.708 Cash/ Total Assets 2003-2007 0.093 0.055 0.111 0.000 0.712 GOV-I index 13.85 14.00 3.1 7 22 GOV-I (%) 0.554 0.560 0.124 0.28 0.88 3-year Sales Growth 2003 0.07 0.03 0.36 -0.46 3.03 3-year Sales Growth 2007 0.23 0.13 0.48 -0.35 3.90 3-year Sales Growth 2003-2007 0.18 0.10 0.48 -0.75 5.97 LN(Assets) 2003 14 13.6 2.1 9.7 19.7 LN (Assets) 2007 14.5 14.1 2.2 10.5 20.6 LN (Assets) 2003-2007 14.2 13.8 2.2 9.5 20.6 Market-to-Book Value 2003 4.23 1.69 15.9 0.24 154.37 Market-to-Book Value 2007 3.71 2.51 4.42 1.59 36.12 Market-to-Book Value 2003-2007 3.92 2.20 8.44 0.24 154.37 Change in M-B Value 0.374 0.230 0.722 -0.726 7.542 Leverage to market value ratio 2003 0.228 0.165 0.194 0.000 0.677 Leverage ratio (MV) 2007 0.206 0.160 0.191 0.000 0.825 Leverage ratio (MV) 2003-2007 0.202 0.162 0.178 0.000 0.825 Panel B. Industry Average LN(Assets) Cash/Total Assets Utilities/Oil, Gas, Water 16.6 3.6% Iron and Steel 14.3 5.9% Machinery-Industrial/Specialty 13.4 12.3% Construction 16.0 6.1% Chemicals 14.2 4.1% Engineering Services 13.7 24.8% Food and Beverage 13.0 8.1% Apparel and Textile 13.2 19.2% Paper and Paper Products 12.9 5.8% Chemicals-Diversified 12.2 10.5% Retail-Misc./Specialty 13.8 7.5% Restaurants and Hotels 14.9 5.1% Transport 14.1 37.7% Construction-Concessions 15.2 3.2% Other Services 12.4 6.8% Broadcasting Media 14.4 1.9% Electronics/Computer/Communication 14.0 6.4% Aerospace/Defense 18.1 12.3% Panel A shows descriptive statistics for selected variables employed in the study. The variables are the ratio of cash holdings to total assets (Cash/TA), the Governance Index (GOV-I) in absolute value and in percentage, sales growth (the geometric average of the last three years sales growth), the natural logarithm of total assets (LNAssets) as a proxy for firm size, the market to book value ratio (M-B) as well as the change in the M-B value ratio, and the leverage ratio (total debt/[total debt + market value of equity]).Panel B provides the industry average for the natural logarithm of total assets (LNAssets) and the ratio of cash holdings to total assets (Cash/TA) for the sample period (2003-2007). Table 2: Correlation Coefficients Lever Lag Cash Cash/NA MV/BV Size Sales

Growth GOV-I BOwn

LagCash 0.15*** 1 Cash/NA -0.22*** 0.15*** 1 MV/BV -0.29*** -0.18*** 0.24*** 1 Size 0.41*** 0.05 -0.03 0.04 1 Growth 0.12** -0.05 0.14*** 0.06 0.20*** 1 GOV-I 0.12*** 0.13*** 0.07 0.25*** 0.34*** 0.24*** 1 BOwn 0.09** 0.05 -0.05 -0.07 -0.05 -0.01 -0.02 1 Block -0.07 0.03 -0.01 -0.05 0.05 0.07 0.00 0.04 This table provides Pearson correlations for selected variables used in the study: Lever is the debt ratio to market value of assets (total debt/[total debt + market value of equity]), LagCash (Casht-1) is the level of cash at the beginning of year t, Cash/NA is cash and cash equivalents divided by Net Assets (Total Assets net of cash holdings), MV/BV is market value of assets divided by the book value of assets, Size is the log of Total Assets, Sales Growth is the geometric average of last three years sales growth, GOV-I is the governance index, BOwn is the percentage of board ownership, Block is the sum of all shareholders with 5% or more ownership stake on the firm. ***,**,* indicate significance at the 1, 5 and 10 percent level respectively.


97

The dependent variable is the stock’s excess return for year t which is defined as the return of the stock during the calendar year t less stock’s i benchmark portfolio return during the same period. The benchmark portfolios are formed on size and book-to-market value following Fama and French (1993). All studies that try to capture the dollar change in the firm value resulting from a change in one dollar of cash holdings use this methodology proposed by Daniel and Titman (1997) and used by Grinblatt and Moskowitz (2004), Faulkender and Wang (2006), Dittmar and Mahrt-Smith (2007), and Pinkowitz, Stulz and Williamson (2006). Table 3: Research Variables Variable Measure Code Governance Index (GOV-I) Index composed by 25 binary questions GOV-Ii,t Blockholdings Σ % of shares owned by the controlling shareholders

(shareholders with more than 5% stake on the firm) by the end of year t Blocki,t

Board ownership Σ % of shares owned by the members of the board by the end of year t Bowni,t Cash Cash and Cash equivalents in year t Ci,t Leverage Total debt / (Total debt + Market value of equity) in year t Li,t Dividends Dividends paid in year t Di,t Earnings Earnings before extraordinary items in year t Ei,t Interest Interest expenses in year t Ii,t Stock return Stock annual return ri,t Portfolio return Fama and French (1993) benchmark portfolio return R,t Market capitalization Market value of equity = stock price times the number of shares outstanding

by the end of year t Mi,t

Net Assets Total assets net of cash in year t NAi,t New finance Net new equity issues plus net new debt issues in year t NFi,t This table provides a summary of all variables used in equation (1) and a description of how each variable is calculated. A portfolio return is a value-weighted return based on the market capitalization of the firms. The excess return for firm i is the difference between the benchmark return for this company’s stock and the return of the stock. The dependent variable is calculated by simply subtracting the portfolio return to which stock i belongs from its realized return during year t. The main specification used in the study is the model proposed by Dittmar and Mahrt-Smith (2007) as follows, which is estimated using panel data random effects:

+∆

+∆

+∆

+∆

+∆

+=−−−−−− 1,

,5

1,

,4

1,

,3

1,

,2

1,

,10,,

ti

ti

ti

ti

ti

ti

ti

ti

ti

tiBtiti M

DM

IMNA

ME

MC

Rr γγγγγγ

+∆

×+∆

×++++−−−

−

−−

−

1,

,,10

1,

,

1,

1,9

1,

,8,7

1,

1,6

ti

titi

ti

ti

ti

ti

ti

titi

ti

ti

MC

LM

CMC

MNF

LMC

γγγγγ

titi

titi M

CGOVI ,

1,

,,11 εγ +

∆×+

− (1)

Where ∆X indicates a change in X from year t – 1 to t. The dependent variable is the excess stock return ri,t - Ri,t, where ri,t is the stock return during year t and Ri,t is the benchmark portfolio return calculated for the companies in the sample following Fama and French (1993) methodology. The independent variables are: Mi,t is the market value of equity. Ci,t is cash and cash equivalents. Ei,t is earnings before extraordinary items. Earnings is calculated following Fama and French (1998) as earnings before extraordinary items plus interest, deferred tax credits, and investment tax credits. NAi,t is net assets (total assets net of cash), Ii,t is interest expenses, Di,t is dividend payments, Li,t is Debti,t / (Debti,t + Mi,t), to measure leverage and is calculated as total debt (short term debt + long term debt) divided by the sum of total debt and market value of equity, NFi,t is new finance (net new issues of equity + net new issues of debt), GOVi,t is the Governance Index (GOV-I) and the other variables used as a proxy of governance (board ownership and


98

blockholdings). The variables are described in Table 3. The other variables control for profitability (Ei,t), investment (NAi,t) and financing (Ii,t, Di,t, Li,t and NFi,t) strategies. The initial prediction is that coefficient γ11 (GOV-I) is positive and statistically significant which means that we expect the interaction between changes in cash and corporate governance to be statistically significant (H1). The interaction between the change in cash with governance is calculated by assigning a dummy variable that takes the value of one if the company is in the top or bottom tercile of the GOV-I, zero otherwise. This dummy is multiplied by the change in cash (GOV-I*∆Ct). We also expect the interaction of changes in cash with the initial cash level and with leverage, coefficients γ9 (Ci,t-1) and γ10 (Li,t) respectively, to be negative and statistically significant (H2 and H3). Finally, we expect that the presence of future growth opportunities is a moderator factor in the relationship between the value of cash and the quality of governance (H4). All independent variables (except leverage) are deflated by the lagged market value of equity (Mi,t-1). EMPIRICAL RESULTS Table 4 reports the estimated coefficients for different specifications of the model. Column [1] shows the regression coefficients of the model without governance. The main objective with this specification is to measure the marginal value of cash for the average Spanish firm allowing only for the interaction of cash with the level of cash the firm has on hand in the beginning of the year (Ct-1*∆Ct) and with leverage (Lt*∆Ct). The results show that an extra euro of cash is valued by investors at €0.93 for the average firm. It is expected that one euro of cash held by the firm is valued at a discount due to taxes (at the shareholders level) and transaction costs incurred to transfer cash from the company to its shareholders (via dividends or share repurchases). This finding is consistent with the results obtained by Faulkender and Wang (2006) and Pinkowitz and Williamson (2004) for the U.S. The first find that the value of one extra dollar of cash in the mean U.S. firm is valued at $0.94 and the latter report a value of $1.20 for the same dollar of extra cash. However, the latter do not account for the interaction of the level of cash and leverage with the change in cash, which may explain the difference. Regarding the interaction of cash with leverage, the negative effect of leverage on the value of the firm is more pronounced for the Spanish firms than for its American counterparties. Considering an all-equity firm in Spain, the value of one extra euro of cash for this firm is 49 cents higher than for a firm with 10% leverage. For the U.S., Faulkender and Wang (2006) report a discount of only 14 cents of dollar for leverage. This result corroborates the hypothesis that the conflict of interests between shareholders and debt holders is more severe in Spain than in the U.S., as previously proposed by De-Miguel and Pindado (2001). In Column [2] we introduce the interaction between governance and the change in cash. The results indicate that the marginal value of cash is sensitive to the firm’s quality of governance. The estimated coefficients of the interactions of cash with the level of cash (Ct-1*∆Ct) and with leverage (Lt*∆Ct) are statistically significant. The economic interpretation of these coefficients is as follows, considering the sample mean for cash, leverage and governance (reported in Table 5), for an average firm with cash holdings of 14% of equity, financial leverage of 20.2% and average quality of governance (55.4% in the GOV-I), the marginal value of its cash is valued at €0.79. However, the same euro of cash is valued at as high as €1.02 for companies with good governance (companies in the top tercile of the distribution of the GOV-I) and as low as €0.57 for companies with bad governance (companies in the bottom tercile of the distribution of the GOV-I). We interpret this result as evidence of agency theory as investors apply a discount to bad governance firms. The discount on poor governance is the cost of agency imposed by investors on firms with information opacity. When information asymmetries are high, investors fear managers will employ the extra cash in value destroying activities. Similarly, minority shareholders may pay a premium for firms with good governance, because it is less likely that majority shareholders will expropriate their wealth through the appropriation of the benefits of control. Our results are consistent with the results of Dittmar and Mahrt-Smith (2007) for the U.S. They report a marginal value of cash for


99

the average governance firm of $1.09, a value as high as $1.62 for companies with good governance and as low as $0.42 for companies with bad governance. Table 4: Regression Results – The Impact of Governance on the Marginal Value of Cash

Dependent Variable = Excess Stock Return (ri,t - Ri,t)

Independent Variables [1] [2] [3] [4] [5]

∆Ct 2.054*** (0.000)

1.163** (0.021)

1.884*** (0.000)

1.820*** (0.000)

1.436*** (0.007)

GOV-I*∆Ct 1.472*** (0.001)

1.806*** (0.000)

Block*∆Ct 0.118 (0.790)

Block 0.483*** (0.006)

BOwn*∆Ct 0.596 (0.244)

BOwn 0.004 (0.977)

∆Et -0.002 (0.975)

0.013 (0.852)

-0.001 (0.984)

0.003 (0.963)

-0.816*** (0.001)

∆NAt 0.074 (0.129)

0.097** (0.047)

0.054 (0.266)

0.070 (0.155)

0.185*** (0.003)

∆It 0.314 (0.476)

0.188 (0.668)

0.254 (0.568)

0.395 (0.377)

1.457*** (0.005)

∆Dt 1.125 (0.229)

1.042 (0.261)

1.005 (0.280)

1.117 (0.235)

3.023*** (0.007)

NFt -0.006 (0.889)

-0.021 (0.637)

0.001 (0.987)

-0.009 (0.833)

-0.082 (0.141)

Ct-1 0.295** (0.014)

0.378*** (0.002)

0.296** (0.013)

0.288** (0.018)

0.596*** (0.000)

Ct-1*∆Ct -0.929*** (0.005)

-0.780** (0.018)

-0.889*** (0.007)

-1.082*** (0.002)

-1.892*** (0.000)

Lt 0.0189 (0.902)

0.003 (0.983)

0.059 (0.700)

0.026 (0.865)

-0.400** (0.033)

Lt*∆Ct -4.888*** (0.000)

-5.352*** (0.000)

-4.661*** (0.000)

-5.260*** (0.000)

-4.855*** (0.000)

Intercept -0.097** (0.026)

-0.107** (0.014)

-0.201*** (0.000)

-0.098** (0.036)

-0.093* (0.062)

Observations 490 490 490 490 380 R-squared 0.09 0.11 0.09 0.09 0.20

This table presents the return regressions of Faulkender and Wang (2006) and Dittmar and Mahrt-Smith (2007) models. The dependent variable is the excess stock return of firm i relative to the portfolio return calculated according to Fama and French (1993). All variables except for Lt are deflated by the lagged market value of equity (Mt-1). ∆ indicates the change from previous year and the regressions are calculated on a panel of 98 non-financial Spanish listed firms at the Mercado Continuo of the Madrid Stock Exchange between 2003 and 2007. The dependent variables are described in Table 3. The first figure in each cell is the regression coefficient. P-values based on robust standard errors are reported in the brackets. ***, **, and * indicate significance at 1, 5 and 10 percent levels respectively. Columns [3] and [4] report the coefficients for the regressions using other governance variables, specifically board ownership and block ownership (the sum of all shareholders with 5% or more ownership stake on the firm). The results show that both variables have a positive impact on firm value albeit not statistically significant. Finally, column [5] reports the results of the regression for a sub-sample of companies, only companies with positive sales growth. We use this measure as our proxy for future growth opportunities. It is expected that the same euro of extra cash have greater value for companies with future growth opportunities, because they will need capital to finance their growth process and because raising external capital is costly. An analysis of the estimated coefficients indicate that the effect of governance is accentuated in companies with positive sales growth, as they will need cash to invest in future projects and continue growing. In this case, shareholders value the fact that managers’ interests are aligned. However, more pronounced is the increase in the coefficient of the interaction of the level of cash with the change in cash (Ct-1*∆Ct), which shows again the importance of the conflict between


100

shareholders and debt holders. In the presence of growth opportunities, this conflict can take to two well-described problems: the underinvestment problem and the asset substitution problem. Table 5: The Marginal Value of Cash to Shareholders

Sample Means Marginal Value of €1 of Cash Level cash at the beginning of the year (Ct-1) 0.140 Companies with good governance €1.02 Leverage (Lt) 0.202 Average governance company €0.79 Quality of governance (GOV-Ii) 0.554 Companies with bad governance €0.57 Top tercile (GOV-Ii) 0.713 Bottom tercile (GOV-Ii) 0.409

This table reports sample means for the level of cash in the beginning of the year, leverage, and the quality of governance (average, top and bottom tercile). The means are used to calculate the marginal value of €1 of cash for the average, good and bad governance company. For example, for a company with average cash holdings of 14% of equity, average financial leverage of 20.2% and average quality of governance (55.4% in the GOV-I), the marginal value of its cash is valued at €0.79 (=1.163+(-0.78*14%)+(-5.352*20.2%)+1.472*55.4%). The first is motivated by the fact that debt holders will capture most of the project’s benefits due to the payment of interests. It takes to the second problem, that is likely to happen when shareholders of a highly leveraged firm may prefer riskier projects because they will profit from any upside (which is more likely to happen), but any downside is shared with the debt holders. It creates an incentive for shareholders of high leverage firms to take on overly risky projects (asset substitution) and/or pass up positive NPV projects (underinvestment). The effect of leverage (Lt*∆Ct) remains the same. Thus, if we consider only companies with positive growth prospects, the value of an extra euro of cash for the average firm is valued at €1.19, and as high as €1.48 for companies with future growth opportunities and good governance and as low as €0.93 for companies with future growth opportunities and bad governance. Our results corroborates the hypothesis proposed by Myers and Majluf (1984) that financial slack has value when future growth opportunities are present and this evidence supports hypothesis 4 (H4). Intuitively firms with positive NPV investment projects that need to be financed in the next year should save cash (through internally generated funds), instead of distributing cash to shareholders (i.e. via dividends) and raising new capital at a much higher cost (ignoring other factors like the availability of capital in times of crisis, for example). The results of the regressions reported in Table 4 show empirically that shareholders attribute special value to good governance under the presence of future growth opportunities, as they believe cash will be used to maximize firm value. Previous research has shown that governance improves firm value. We show that this positive effect is also observed through the interaction between financial slack and governance, and strengthened by the moderating effect of the presence of future growth opportunities. CONCLUSIONS Our objective with this paper was to analyze the interaction between cash and governance and its impact on the market value of Spanish firms using the model developed by Faulkender and Wang (2006) and adapted by Dittmar and Mahrt-Smith (2007) to include governance. We thus extend the analysis to include the presence of future growth opportunities as a moderator factor in this relationship, due to its influence on the level of cash held by firms. Our results show that Spanish investors attribute a different value to the same euro of extra cash at companies with good and poor quality of governance, and that, in the presence of future growth opportunities, this difference is accentuated. We show evidence that investors apply a considerable premium (discount) on good (bad) governance companies. When considering future growth opportunities, the results show that a higher premium is paid to good governance, which suggests that shareholders believe the benefits of holding cash to finance future investment offset the potential agency costs associated with it. Hence, our results support the agency theory and the pecking order hypothesis, as they


101

show that companies accumulate cash in the previous period to finance their growth in the next period. We also find evidence to support the hypothesis that financial slack has value in the presence of future growth opportunities. Nonetheless, the results show that the conflict between shareholders and debt holders may be more severe in Spain than in the U.S. This conflict can take to the underinvestment and asset substitution problems in the case of highly leveraged firms, and for this reason, investors apply a stronger discount for leverage when valuing Spanish firms. This study expands the literature on cash holdings and on corporate governance by providing empirical evidence on the value shareholders attribute to the marginal value of cash for companies with different levels of quality of governance in Spain. One possible limitation of our study is the fact that our analysis is focused on a single country. Nonetheless, the specifics of the Spanish case are common features of other European countries. Our findings may, for this reason, be applied to other realities with similar characteristics. This article opens important possibilities for future research. One suggestion would be to analyze the importance of financial slack in times of environmental jolts (i.e. financial/credit crisis) and how it affects firm value creation during and after the jolt. APPENDIX A: Questions Used in the Construction of the Governance Index (GOV-I)

Dimension of governance

# Questions

Access and Content of the Information

1 Does the company website provide information about its governance system? 2 Does the company have an English version of its website where results are promptly updated? 3 Does the company have an Investors Relation Department? 4 Does the company analysts’ presentations with which investors can prepare financial projections? 5 Does the company disclosure information about its next or tree-year ROA or ROE targets? 6 Does the company publish/announce quarterly reports within two months of the end of the quarter? 7 Does the company promote analysts’ and investors’ meetings on a regular basis (i.e. when they publish the

Annual Report)? 8 Is the public announcement of results promptly published in the web page of the company?

Board Structure 9 Is the Chairman an independent, non executive director? 10 Does the CEO serve on no more than one additional board of other public company? 11 Is the board composed by no less than 5 and more than 15 members? 12 Is shareholder approval required for changing the board size? 13 The company does NOT have any Golden Parachute Provision approved for the senior executives? 14 Does the board include no direct representative of banks and other large creditors of the company? (having any

representatives is negative) 15 Do independent, non-executive directors account for more than 50% of the board? 16 Are board members elected annually (they have a unified mandate of one year and the reelection is not

automatic?) 17 The Chairman of the board and the CEO are not represented by the same person. 18 Does the board have at least one female director?

Ownership Structure and Control

19 Do directors receive part of their remuneration in stocks/stock options? 20 Is directors’ stock ownership less than 35% or more than 70% of total outstanding shares? 21 Does the Chairman have Casting Vote? 22 Does the company offer tag along to the minority shareholders?

Transparency 23 Does the company define any rules to ensure that the auditor does not perform any other services for the company (e.g. consulting)?

24 Does the company publish in the Annual Report information about its risk management system? 25 Are the audit committee and the nominating committee exclusively composed by outside directors?

This table provides the questions compounding the governance index (GOV-I) constructed as a proxy for the quality of governance. The questions were developed based on the recommendations of the Spanish Code of Best Practices. For constructing the index we first code the 25 questions/variables as 1 or 0 depending on whether the firm complies with each corporate governance standard or not. Each positive answer adds one point to the index, and the companies present a corporate governance level that ranges from 0 to 25. The answers to all questions is obtained exclusively from secondary data (firms’ websites and the Corporate Governance Annual Report released by the companies) REFERENCES Acharya, V. V., H. Almeida, and M. Campello, 2007, "Is cash negative debt? A hedging perspective on corporate financial policies," Journal of Financial Intermediation 16, 515-554.


102

Almeida, H., M. Campello, and M. S. Weisbach, 2004, "The Cash Flow Sensitivity of Cash," The Journal of Finance 59, 1777-1804. Ammann, M., D. Oesch, and M. M. Schmid, 2010, "Corporate Governance and Firm Value: International Evidence," SSRN eLibrary. Available at http://ssrn.com/abstract=1692222 Attig, N., S. El Ghoul, O. Guedhami, and S. Rizeanu, 2011, "The governance role of multiple large shareholders: evidence from the valuation of cash holdings," Journal of Management and Governance 1-33. Chang, K., and A. Noorbakhsh, 2009, "Does national culture affect international corporate cash holdings?," Journal of Multinational Financial Management 19, 323-342. Chi, J. D., "Understanding the Endogeneity between Firm Value and Shareholder Rights," Financial Management, Vol. 34, No. 4, Winter 2005. Cotter, J. F., A. Shivdasani, and M. Zenner, 1997, "Do independent directors enhance target shareholder wealth during tender offers?," Journal of Financial Economics 43, 195-218. Daniel, K., and S. Titman, 1997, "Evidence on the Characteristics of Cross Sectional Variation in Stock Returns," Journal of Finance 52, 1-33. de Miguel, A., and J. Pindado, 2001, "Determinants of capital structure: new evidence from Spanish panel data," Journal of Corporate Finance 7, 77-99. Demirguc-Kunt, A., and V. Maksimovic, 1998, "Law, finance, and firm growth," Journal of Finance 53, 2107-2137. Denis, D. J., and V. Sibilkov, 2010, "Financial Constraints, Investment, and the Value of Cash Holdings," Review of Financial Studies 23, 247-269. Dittmar, A., and J. Mahrt-Smith, 2007, "Corporate governance and the value of cash holdings," Journal of Financial Economics 83, 599-634. Dittmar, A., J. Mahrt-Smith, and H. Servaes, 2003, "International Corporate Governance and Corporate Cash Holdings," Journal of Financial & Quantitative Analysis 38, 111-133. Fama, E. F., and K. R. French, 1993, "Common risk factors in the returns on stocks and bonds," Journal of Financial Economics 33, 3-56. Fama, E. F., and K. R. French, 1998, "Taxes, financing decisions, and firm value," Journal of Finance 53, 819-843. Faulkender, M., and R. Wang, 2006, "Corporate Financial Policy and the Value of Cash," Journal of Finance 61, 1957-1990. Fernández, A. I., and S. Gómez-Ansón, 1999, “Un estudio de las Ofertas Públicas de Adquisición en el mercado de valores español,” Investigaciones Económicas, 23, 471-495. Franks, J., and C. Mayer, 1996, "Hostile takeovers and the correction of managerial failure," Journal of Financial Economics 40, 163-181.


103

Frazer, W. J., Jr., 1964, "The Financial Structure of Manufacturing Corporations and the Demand for Money: Some Empirical Findings," Journal of Political Economy 72, 176-183. Grinblatt, M., and T. J. Moskowitz, 2004, "Predicting stock price movements from past returns: the role of consistency and tax-loss selling," Journal of Financial Economics 71, 541-579. Harford, J., 1999, "Corporate Cash Reserves and Acquisitions," Journal of Finance 54, 1969-1997. Harford, J., S. A. Mansi, and W. F. Maxwell, 2008, "Corporate governance and firm cash holdings in the US," Journal of Financial Economics 87, 535-555. Jensen, M. C., 1986, "Agency Costs of Free Cash Flow, Corporate Finance, and Takeovers," The American Economic Review 76, 323-329. Keynes, J. M., 1936, The general theory of employment, interest and money, Macmillan, London. Kim, C. S., D. C. Mauer, and A. E. Sherman, 1998, "The Determinants of Corporate Liquidity: Theory and Evidence," Journal of Financial & Quantitative Analysis 33, 335-359. Leech, D., and M. C. Manjón, 2002, "Corporate Governance in Spain (with an Application of the Power Indices Approach)," European Journal of Law and Economics 13, 157-173. Mikkelson, W. H., and M. M. Partch, 2003, "Do persistent large cash reserves hinder performance?," Journal of Financial and Quantitative Analysis 38, 275-294. Myers, S. C., and N. S. Majluf, 1984, "Corporate Financing and Investment Decisions When Firms Have Information That Investors Do Not Have," Journal of Financial Economics 13, 187-221. Myers, S. C., and R. G. Rajan, 1998, " The Paradox of Liquidity," Quarterly Journal of Economics 113, 733-771. Ocaña, C., J. I. Peña, and D. Robles, 1997, "Preliminary evidence on takeover target returns in spain: A note," Journal of Business Finance & Accounting 24, 145-153. Opler, T., L. Pinkowitz, R. Stulz, and R. Williamson, 1999, "The determinants and implications of corporate cash holdings," Journal of Financial Economics 52, 3-46. Pinkowitz, L. E. E., and Williamson, R. (2004). “What is a dollar worth? The market value of cash holdings,” Unpublished working paper, Georgetown University. Pinkowitz, L. E. E., R. E. N. Stulz, and R. Williamson, 2006, "Does the Contribution of Corporate Cash Holdings and Dividends to Firm Value Depend on Governance? A Cross-country Analysis," Journal of Finance 61, 2725-2751. ACKNOWLEDGEMENT The authors are grateful to two anonymous reviewers and the managing editor for helpful comments during the peer review process. The authors thank participants of the Collaborative Research Forum with the Athabasca University and the 2009 European Academy of Management Conference in Liverpool, for excellent comments on early versions of the paper. All errors remain ours. This research received financial support from the MacEwan Research, Scholarly Activity and Creative Achievements Fund.


104

BIOGRAPHY Dr. Eloisa Perez-de Toledo is Assistant Professor of Accounting and Finance at the School of Business, MacEwan University in Edmonton, Canada. She can be reached at MacEwan School of Business, Rm 5-225K, 10700 – 104 Avenue, Edmonton, AB, Canada T5J 4S2 or by e-mail: [email protected] Dr. Evandro Bocatto is Assistant Professor of Management at the School of Business, MacEwan University in Edmonton, Canada. He holds a PhD from ESADE Business School, Barcelona, Spain. He can be contacted at MacEwan School of Business, Rm 5-252E, 10700 – 104 Avenue, Edmonton, AB, Canada T5J 4S2 or by e-mail: [email protected]


105

IS INFLATION ALWAYS AND EVERYWHERE A MONETARY PHENOMENON? THE CASE OF NIGERIA

Salami Doyin, Lagos Business School, Nigeria Kelikume Ikechukwu, Lagos Business School, Nigeria

ABSTRACT

In response to shocks, emanating from the global financial crisis of 2007-2008 the Central Bank of Nigeria has continuously used tight monetary policy instrument to check volatility in the general price level. The success of using monetary policy tool to influence the movement of key macroeconomic aggregates in Nigeria rests solely on the question of whether inflation is driven purely by changes in monetary aggregates. Using quarterly time series data for Nigeria over the period 1970 to 2011, we test the quantity theory relationship between money and price movement to establish if inflation is always and everywhere a monetary phenomenon. Using the autoregressive distributed lag (ARDL) modeling approach we obtained a robust estimate for Nigeria. The result of the study shows that inflation is not always and everywhere a monetary phenomenon in the case of Nigeria raising serious doubt on the continuous use of monetary policy tool to achieve price stability in Nigeria. JEL: C22, E52, E63, G28 KEY WORDS: Money Supply, Monetary Policy, Policy Regulation, Time series Model INTRODUCTION

he Nobel Prize winning economist Milton Friedman once postulated, “Inflation is always and everywhere a monetary phenomenon” (Friedman, 1956). His argument was anchored on the classical quantity theory of money which establishes the existence of a direct functional

relationship between money supply and the general price level with aggregate income and the velocity of transaction remaining constant (Lothian, 2009; Selgin, 2008). Given these perceived relationship, the Central Bank uses monetary policy instrument to influence the availability and cost of credit with the ultimate objectives of achieving price stability, sustainable economic growth, balance of payment equilibrium and full employment level (Mishkin, 2000). In recent years, a growing consensus has emerged for price stability to be the overriding long-run goal of monetary policy (Mishkin, 1998). However, the effective prediction of the relationships between inflation and money supply depends largely on the existence of a stable and predictable relationship between monetary aggregates, inflation and the output in the economy. If the money market is largely underdeveloped and the relationship between the chosen monetary aggregates and the ultimate policy objective are weak, monetary targeting becomes a very weak instrument (Panzera, 2011; Nachega, 2001). In Nigeria, the focus for monetary policy since the inception of the Central Bank of Nigeria (CBN) in 1959 has been to stamp out incipient inflationary risk and maintain a sustainable growth in output level. To achieve these intermediate and broad policy frameworks, the country has experimented two different regimes of monetary policy-the exchange rate targeting regime (1959-1973) and the monetary targeting regime (1974-to date) (CBN, 2008). The shift from exchange rate targeting to monetary targeting regime in 1974 was done with the aim of mitigating inflationary pressures arising from increased public expenditures. This was anchored on the premise that inflation is a monetary phenomenon. Evidences from monetary targeting in Nigeria has however shown that monetary policy had always encountered problems and the ultimate target of low and stable price levels enacted by successive administrations may be driven by some forms of structural rigidities inherent in the Nigerian economy (CBN 2008). Consequently, this brings to focus the issue of

T

S. Doyin & K. Ikechukwu | IJBFR ♦ Vol. 7 ♦ No. 2 ♦ 2013

106

whether inflation is always and everywhere a monetary phenomenon in Nigeria and the rational of pursuing stability in the general price level with the use of purely monetary aggregates. The aim of this study is to test the quantity theory relationship between money and prices in Nigeria with a view of establishing whether the notion that inflation is always and everywhere a monetary phenomenon holds true for Nigeria. Following the introductory section, the other parts of the paper is structured into four sections; section two is the literature review while section three is expressed views on the data set and methodology of the study. In section four, we present the empirical analysis while section five is the conclusion and policy implications of the study. LITERATURE REVIEW The quantity theory of money holds that an expansion of the money stock does not increase output in the long run with a focus on average inflation and money growth over successive time intervals (Friedman, 1984). Consequently, it has been confirmed that price increase in the short run due to scarcity in the supply of some essential products may not have an influence on the rates of inflation in the long run. This is because such rates of inflation is controlled by contractionary monetary policies (Barro and Sala-i-Martin, 1995; Rolnick and Weber, 1998; Kermal, 2006). Some scholars are of the opinion that inflation is a monetary phenomenon, taking into consideration longer run studies and that it is greatly encouraged by structural policy issues (Khan, 1980; Grauwe and Polan, 2005; Khan and Schimmelpfennig, 2006). Kermal (2006) stated that long run money supply impact the inflation rates and that the quantity theory of money holds in the long run, emphasizing that inflation is a monetary phenomenon. In the short run, he noticed that the tendencies for money to influence inflation is not quick, he disclosed that it takes approximately three quarters. He noticed that it occurs with persistent rates of inflation and consistent shocks emanating from gross domestic product, money supply and price in the economy. Cecchetti (2000) revealed that high and persistent inflation in most economies, act as a repressive tax, with impending consequences for those who are asset-poor and hold their entire savings only in cash. In addition, inflation was to be harmful to economic and financial sector growth, impede resource allocation and societal welfare (Whitehead, 1976). While Aisen and Jose’Veiga (2006) evidences showed the effect of inflation on the economy may be politically costly for the government due to its socio-economic impact in the country. To effectively evaluate the impact of inflation and put in place policies to guide its effect on the economy, it is necessary to disclose, if inflation everywhere is a monetary phenomenon (Friedman, 1956). To justify this fact, it is essential to first analyze if constraints in supply side factors cause inflation to increase and persist without any form of monetary accumulation (Bernanke, 2005). To analyze the link between inflation, the growth rate of money and the inflationary experiences of a stated set of economies, they used the panel-data technique to test for the quantity theory relationship between money and inflation in accordance to the Friedman’s principle (Friedman, 1956). This principle stated changes in money supply growth lead to equal proportional changes in the inflation rate, through the forces of the Fisher effect, in the nominal interest rate (Grauwe and Polan, 2005). Subsequently, long run money affected the price level and not the level of output. Thus, inflation in these economies was a monetary phenomenon. The most appropriate solution to redeem this situation will be controlling the supply of money in circulation (King, 2001). Subsequently, they resolved that there is a strong positive relationship between long run inflation and the growth rate of money, such that when money growth increases by distinct percentage, the rates of inflation also raise by the same proportion. Hence, in the long run, there is neutrality between money growth and output growth from one perspective and the velocity of changes from another perspective (Grauwe and Polan, 2005). This strong link was due to the levels of hyperinflation in the data set of countries used in the model. In addition, they disclosed that inflation and money growth for low-inflation countries is weak.


107

Examining core inflation as the component of measured inflation that has no medium and long run impact on real output, in accordance with the Phillips curve analysis of co-movement in inflation and output. Quah and Vahey (1995) introduced the vector auto-regression model with restriction on the dynamic process. Goodfriend (2000), King (1999), Blejer and Leon (2000), and Blejer, Ize, Leone and Werlang (2000) supported the view that this study method is important because it will analyze the efficiency of monetary policies to stabilize prices. They disclosed that monetary factors are the main determinants of recent surge in inflation. In addition, the economic growth variable (GDP growth) and the prices of major consumables matters while exchange rate appreciation play very little roles. In the long run, movement in the rates of inflation and the growth in money supply, follow a one to one relationship, relative to real income. This relationship was also same for growth in real income and the velocity of money. From their analysis, it stands sure, that increases in money supply are the main cause of inflation. Since the proportional relationship between the excess money supply is over that of the output growth and the velocity growth. Therefore, these studies proposed that for policy makers to control for this situation, they should put in place tight monetary policy measures. Consequently, such monetary policy formulation must strictly take into consideration activities in the real and financial sectors and treat them as constraints on policy. METHODOLOGY Data Source The data used for this study cover 42 years period (166) quarterly observations. It begins in the first quarter of 1970 and ends in the second quarter of 2011. The source of the data is from the Central Bank of Nigeria Statistical Bulletin and the Published bulletin of the National Bureau of Statistics Nigeria. The two major variables used in the study are the money supply variable (Broad Money M2) and the Price variable represented by CPI (consumer price index). Model Specification We link the theoretical base of the views that inflation is always and everywhere a monetary phenomenon to one of the oldest theories in economics-the classical quantity theory of money and the work of Friedman (1963). In its simplest form, the quantity theory of money states that there is a direct proportional relationship between changes in money supply growth and inflation. This presupposes that growth in money supply follows an equal change in inflation and the force of the Fisher effect, in the nominal interest rate. Using the quantity theory of money, we will attempt to explain the extent to which monetary forces trigger changes in price movement in the economy. Equation 3.1 below expresses the famous equation of exchange

MV=PY (3.1)

Where M is a suitable measure of money supply (in the case of Nigeria, we use M2-broad money supply which is a better measure of the stock of money supply.)

V is the income velocity of money obtained by Y*P/M (See Muskin, 2008; P. 19)

P is the aggregate price level represented by the CPI, which is a measure of the general price level

Y is the real gross domestic product GDP.

Expressing equation 3.1 in growth form, we denote its logarithm for, in lower case as:


108

m + v = p + y (3.2)

From equation 3.2, we derive the inflation equation as:

p = m - y + v (3.3)

From equation 3.3, we can obtain three basic elements of the quantity theory of money. That;

(i) There exist a long run proportional relationship between growth in money supply and growth in the general price level.

(ii) A permanent increase in the growth rate of money supply leaves output and velocity unaffected in the long run.

(iii) From the quantity theory of money, we can ascertain the time it takes growth in the general price level to respond to changes in the growth in money supply and output.

To obtain a reliable estimate of the short run and long run relationship between growths in general price level and growth in money supply, we will make use of the autoregressive distributed lag (ARDL) modeling technique. The ARDL statistic approach is much more flexible than the other methods available for conducting co-integration test such as the residual based Engle-Granger (1987) test, the maximum likelihood based Johansen (1991; 1995) test and the Johansen-Juselius (1990) test. This is because it can be applied to variables with different order of integration (Pesaran and Pesaran 1997) and it takes sufficient number of lags to capture the data generating process in a general-to-specific modeling framework (Laurenceson and Chai 2003). Therefore, to test the proposed of the inflation-money supply relationship drawn from equation 3.1, under the assumption that velocity of money (V) and income (Y) is constant. We express the linear form of the model as:

ttttt UgVgYgMgP ++++= 3210 2 αααα (3.4)

Where;

gPt = Rate of change in the level of consumers price index CPI

gM2t = Rate of change in broad money supply

gYt = Rate of change in real GDP

gVt =Rate of change in velocity of money

U= Stochastic disturbance factor.

On apriori, in the long run 1α =1, 2α < 0, 3α > 0, while gM2 and gY are uncorrelated.

Theoretically, there is a direct functional relationship between money supply and the general price level. However, the money supply variable may influence the general price level with a time lag. This allows us to incorporate lags of money supply in the regression. Furthermore, the price variable may correlate with its lag, suggesting that the lags of the price variables should be included in the regression model. The inclusion of lags of the dependent variable and lags of the explanatory variables into the regression motivates the commonly used ARDL (p,q) model or the unrestricted ECM model defined as follows:


109

t

q

j

q

jjtjjtj

jt

q

jjjt

p

jjttttt

ugVgY

gMgPgVgYgMgPgP

∑ +∑ ∆+∆+

∆∑+∆∑++++=∆

−

=

−

=−−

−

−

=−

−

=−−−−

1

1

1

1

1

1

1

114131211 22

δψ

λγαααα

(3.5)

In equation 3.5, the term in summation signs represents the error correction dynamics while the term with the coefficientα corresponds to the long run relationship. The ARDL method estimates (P+1)k number of regressions in order to obtain the optimal lag for each variable where p is the maximum number of lags to be used and k is the number of variables in the equation. Given that we are using quarterly data, we select the fourth lag as our maximum lag (P) following Pesaran and Pesaran (1997) to test the robustness and reliability of the regression estimate. Empirical Analysis Before estimating the ARDL model, we tested for the presence of unit roots among the variables with the aid of the Augmented Dickey-Fuller test of unit roots. Table 1 reports the results of the unit root test. The result shows the growth in price level, growth in money supply; output growth and growth in velocity of transaction were all stationary at levels. Therefore, there is no need to difference the variables in the model. Although, cointegration test methods based on Johansen (1991; 1995) and Johansen and Juselius (1990), requires that all the variables be of equal degree of integration, this is however not a requirement for the ARDL approach which combines variables irrespective of their order of integration (Pesaran and Pesaran 1997). Table 1: Unit Root Test on Variables with Intercept and a Linear Trend

Augmented Dickey-Fuller (ADF) Test Variables Levels Status gP -4.958* I(1) gY -12.905* I(1) gM2 -14.262* I(1) gY -14.288* I(1)

Note* This table shows the results of the Augmented Dickey-Fuller (ADF) Unit Root test, which indicate that the level of each variables are integrated or stationary at their individual levels. Given these results, each variable satisfies the requirement to be included in the long-run co integration model. * indicates the significance at 1 percent level. Table 2 display the short run dynamic model. The coefficient of money supply is not statistically significant at the second and third quarter lags but is significant at the first lag at the 10 percent levels. However, the variable has the wrong sign indicating that a change in money supply in the previous quarter has a statistically significant negative impact on change in the current price level. In the third lag, the growth in money supply exhibit a positive impact on changes in the current price level but it was not statistically significantly. Subsequently, changes in output growth had a negative and insignificant impact on current price level in the first, second and third lag respectively. The variable had the right sign. Recall that the quantity theory of money predicts that over a significantly long period, changes in the growth of money do not affect output growth. The velocity variable had the right sign in the second and third lag and passed the test of statistical significance at the 5 percent levels only in the second lag. The lag price variable is the most statistically significant variable in the model affecting current change in the price level. The variable had a negative but statistically significant impact on current changes in price level, passed the test of statistics significance at the 1 percent level in the first, second, and third lags respectively.


110

Table 2: Estimated Short Run Coefficients

Dependent Variables gP n=166 after Adjustment (1970Q2-2011Q3) Regressors Coefficients Standard Error T-Statistic Prob. C -0.0054 1.1119 -0.0048 0.9961 ∆gP(-1) -0.5424 0.1172 -4.635*** 0.0000 ∆gP(-2) -0.6442 0.1217 -5.292*** 0.0000 ∆gP(-3) -0.5261 0.1136 -4.630*** 0.0000 ∆gM2(-1) -0.1598 0.0817 -1.956* 0.0524 ∆gM2(-2) -0.1116 0.0833 -1.339 0.1826 ∆gM2(-3) 0.0163 0.0172 0.9485 0.3444 ∆gY(-1) -0.0076 0.0262 -0.2930 0.7699 ∆gY(-2) -0.0553 0.0351 -1.576 0.1172 ∆gY(-3) -0.0437 0.0315 -1.387 0.1677 ∆gV(-1) -0.0113 0.0169 -0.669 0.5044 ∆gV(-2) 0.1395 0.0681 2.047*** 0.0425 ∆gV(-3) 0.1160 0.0744 1.560 0.1209 ∆gV(-3) 0.1160 0.0744 -3.689*** 0.0003 Ecm(-1) -0.5189 0.1406 -3.689*** 0.0003 R-Squared 0.57 R-Bar-Squared 0.54 F-Stat. 15.36*** DW-Statistic 1.97

Note: 113322113322113322113322110 222 −−−−−−−−−−−−− +∆+∆+∆+∆+∆+∆+∆+∆+∆+∆+∆+∆+=∆ tttttttttttttt ECMVgVgVgYgYgYgMgMgMgPgPgPgP δγγγφφφβββαααα This table shows

the short-run ARDL (p, q) regression estimates of the model over the adjusted sample period of 1970:Q2-2011:Q3. The independent variables are lag growth in price level (gP), lag change in money supply (gM2), lag change in real GDP (gY) and lag change in velocity (gV). The table displays the outcome of the estimated Short-run coefficients of variable in the model expressed in section three. The figures in each cell in column four are the t-statistics, ***, and * indicate significance at 1 and 10 percent levels respectively. The coefficient of ECMt-1 is relatively large in magnitude and is statistically significant at the 1 percent level. It demonstrates the existence of long run relationship between the variables, with the coefficient term -0.5189, suggesting a fast adjustment process. Approximately 51 percent of disequilibrium of the previous quarter’s shock adjusts back to equilibrium in the current quarter. Overall, the result shows that changes in lag past price levels and the velocity of transaction are the most significant variable influencing the current price movement in the short run negating the monetarist claim that inflation is always and everywhere a monetary phenomenon.The result of the long run relationship displayed in Table 3 shows that the growth in money supply had the wrong sign and is not statistically significant as shown by the probability value of 0.9013. The coefficient value of –0.0017 is close to 1 as indicated by the quantity theory of money specification but the negative sign is at variance with the long run proportionality relationship between growth in money supply and growth in price level. Table 3: Estimated Long Run Coefficients

Dependent Variable gP n=166 after Adjustment (1970Q2-2011Q3) Regressors (1) Coefficients (2) Standard Error (3) T-Statistic (4) Prob. (5) gM2 -0.0017 0.0138 -0.1242 0.9013 gY -0.1858 0.0194 -9.580*** 0.0000 gV 0.0141 0.0123 1.1469 0.2531 C 4.699 0.9053 5.191*** 0.0000 R-Squared 0.36 R-Bar-Squared 0.35 F-Stat. 30.70*** DW-Statistic 2.17

Note: ugVgYgMgPt ++++= 3210 2 αααα This table shows the Long-run regression estimates over the adjusted sample period of 1970:Q2-2011:Q3. The independent variables are change in money supply (gM2), change in real GDP (gY) and change in velocity (gV). The table displays the outcome of the estimated long-run coefficients of the equation expressed in section three. The figures in each cell in column 4 and column 5 are the t-statistics and their respective probabilities. *** indicate significance at 1 percent level. The coefficient of output growth had the expected sign; it is low in value and is statistically significant at the 1 percent level indicating that changes in output in the long run has a significant negative effect on growth in the general price level. The velocity variable had the right sign but failed the test of significance at the 1 percent and 5 percent levels of significance The R square value and the adjusted R square values of 0.36 and 0.35 show the model had a poor fit. This implies that over 65 percent systematic changes in price level unaccounted for by the model. The F-statistic value of 30.7 shows the overall model has a good fit and the Durbin-Watson value of 2.17 shows the likely absence of serial correlation in the model.


111

Recall that, the quantity theory predicts that over a sufficiently long period, changes in growth rate of money supply do not affect output growth. This is the neutrality position. To test this position, we estimated output growth as a linear function of growth rate of money supply. Table 4 is the OLS result of money supply output growth relationship. Table 4: OLS estimation of Output Growth on Money Supply

Dependent Variable gP n=166 after Adjustment (1970Q2-2011Q3) Regressors Coefficients Standard Error T-Statistic Prob. gM2 -0.0171 0.0545 -0.3135 0.754 C 6.590 3.595 1.833*** 0.068 R-Squared 0.00059 R-Bar-Squared -0.0054 F-Stat. 0.098 DW-Statistic 2.009

Note: ugMgPt ++= 210 αα This table displays the OLS estimate of output growth and money supply in Nigeria over the adjusted sample period of 1970:Q2-2011:Q3. The figures in each cell in columns 4 and 5 are the t-statistics and their respective probabilities. *** indicate significance at 1 percent level. The result shows the effect of higher money growth on output growth is negative and not statistically significant. This confirms the quantity theory of money prediction that an expansion of money stock does not increase output in the long run and the findings is in line with the conclusion of Barrow and Sala-i-Martin (1995). CONCLUSION This study originally set out to address the monetarist claim that inflation is always and everywhere a monetary phenomenon. Using quarterly data obtained from the Central Bank of Nigeria Statistical Bulletin over the period (1970Q1-2011Q2), we tested the quantity theory of money proposition that there exists a long run proportional relationship between money growth and inflation and neutral relationship between money growth, output growth and velocity. From the regression results obtained, we find a weak negative and statistically insignificant relationship between long run money supply growth and inflation negating the quantity theory of money proposition that the relationship is one of proportionality. This is in line with earlier studies for low inflation countries and EMU countries (Grauwe and Polan 2005). Secondly, we find money growth and income growth to be weakly linked in the long run suggesting that monetary policy tools may not be effective in controlling and influencing macroeconomic aggregates in Nigeria. Finally, we found that there exist a long run relationship between money growth, inflation, output growth and velocity of transaction. The adjustment process between the short run and long run period is fast. Specifically, nearly 51 percent of disequilibrium of the previous quarter’s shock adjusts back to equilibrium in the current quarter. This study has serious policy implications for policy makers in Nigeria and other low-income countries that have continuously based their monetary policy strategy on the premise that “inflation is always and everywhere a monetary phenomenon”. Our result indicates that this is not true for Nigeria and that the continuous use of monetary policy tool to maintain price stability is not likely to yield the desired medium to long-term monetary policy goals. Limitations In this study, we carried out a dynamic modeling of price movement to determine whether inflation is purely a monetary phenomenon in Nigeria. This is to ascertain if the theoretical concept as proposed could be justified for the case of Nigeria. We noticed that an in-depth study, which will consider money market operations, issuance of government securities in primary market, repurchase agreement, interest payment of government domestic debt and sales of foreign exchange, is necessary in order to clarify pending issues on inflation and monetary policy in Nigeria. However, subsequent studies will address these shortcomings.


112

REFERENCE Aisen, A and F. Jose’Veiga (2006) “Political Instability and Inflation Volatility,” IMF Working Paper, WP/06/212, September. Barro, R. and X. Sala-i-Martin (1995), Economic Growth, MIT Press, Cambridge, MA. Bernanke, B. S. (2005) “Inflation in Latin America-A New Era?” Remarks at the Stanford Institute for Economic Policy Research Economic Summit, February 11. Blejer, M. I., A. Ize, A. M. Leone, and S. Werlang (2000) Inflation Targeting in Practice: Strategic and Operational Issues and Application to Emerging Economies, Washington, D. C.: International Monetary Fund. Blejer, M. I., and A. M. Leone (2000) “Introduction and Overview,” In Blejer, et al. Inflation Targeting in Practice: Stategic and Operational Issues and Application to Emerging Economic. Washington, D. C.: International Monetary Fund. Cecchetti, S. G. (2000) “Making Monetary Policy: Objectives and Rules,” Oxford Review of Economic Policy Vol. 16, No. 4, p. 43-59. Engle, R. F. and C. W. J. Ganger (1987) “Cointegration and error correction: Representation, estimation and testing,” Econometrica Vol. 55, p. 251-276. Fabio, S. P. (2011) “Price Stability and Financial Imbalances: Rethinking the Macro-financial Framework after the 2007-8 Financial Crisis,” Macroeconomic and Monetary Economics Working Paper SES No. 423, December. Friedman, M. (1984) “Lessons from the 1979-82 Monetary Policy Experiment,” American Economic Review, 74, p. 397-400. Friedman, M. (1956) “The Quantity Theory of Money-A Restatement,” In Friedman, Milton, ed., Studies in the Quantity Theory of Money, Chicago: University of Chicago Press, p. 1-21. Friedman, M. and A. Schwartz. (1963) A Monetary History of the United States, 1867-1960, Princeton University Press, Princeton, NJ. Goodfriend, M. (2000) Maintaining Low Inflation: Rationale and Reality. In Blejer, et al. Inflation Targeting in Practice: Strategic and Operational Issues and Application to Emerging Economies, International Monetary Fund. Gruwe, P. D., and M. Polan (2005) “Is Inflation Always and Everywhere a Monetary Phenomenon?” Scandinavian Journal of Economics Vol. 107, No. 2, p. 239-259. Johansen, S. (1995) Likelihood-Based Inference in Cointegrated Vector Autoregressive Models, New York: Oxford University Press. Johansen, S. (1991) “Estimation and Hypothesis Testing of Cointegration Vectors in Gaussian Vector Autoregressive Models,” Econometrica, Vol. 59, No. 6, p. 1551-1580.


113

Johansen, S. and K. Juselius (1990) “Maximum likelihood estimation and inference on cointegration with application to the demand for money,” Oxford Bulletin of Economics and Statistics Vol. 52, p. 169-210. Kemal, M. A. (2006) ‘Is Inflation in Pakistan a Monetary Phenomenon? The Pakistan Development Review 45:2, p. 213-220. King, M. (1999) Challenges for Monetary Policy: New and Old, in New Challenges for Monetary Policy, Proceedings of the Symposium sponsored by the Federal Reserve Bank of Kansas City. King, M. (2001) “No Money, No Inflation: The Role of Money in the Economy,” Economie internationale, Vol. 88, p. 111-132. Khan, A. H. and A. Schimmelpfennig (2006) “Inflation in Pakistan: Money or Wheat?”, International Monetary Fund, Working Paper 06/06. Khan, M. S. (1980) “The Dynamics of Money and Price and the Role of Monetary Policy in SEACAN Countries,” SEACEN Occasional Paper, December. Laurenceson, J. and J. C. H. Chai (2003) Financial Reform and Economic Development in China, Univeristy of Queensland, Australia, Edward Elgar Publishing, Inc., UK. Lothian, J. R. (2009), “Milton Friedman’s monetary economics and the quantity-theory tradition,” Journal of International Money and finance Vol. 23, p. 1086-1096. Mishkin, F. S. (2008), ‘The Economics of Money, Banking and Financial Markets. Pearson-Addison Weasley. Mishkin, F. S. (2000) “What Should Central Bank Do?,” Business and Economics Working Papers, Columbia University, March. Mishkin, F. S. (1999) “International Experiences with Different Monetary Policy Regimes,” NBER Working Paper Series, Working Paper 7044, March. Nachega, J. A. (2001) “A Co-integration Analysis of broad money demand in Cameroon,” IMF Working Paper, WP 61/26, March. Nadeem, U. H. and Q. Abdul (2006) “Inflation Everywhere is a Monetary Phenomenon: An Introductory Note,” The Pakistan Development Review, Vol. 45, No. 2, p. 170-183. Pesaran, M. H. and B. Pesaran (1997) Working with Microfit 4.0: Interactive Analysis; Oxford University Press. Quah, D. and S. P. Vahey (1995) “Measuring Core inflation,” The Economic Journal, Vol. 105, No. 432, September, p. 1130-1144. Rolnick, A. and W. Weber (1998) “Money, Inflation, and output under Fiat and Commodity Standards,” Federal Reserve Bank of Minneapolis Quarterly Review 22, 11-17. Selgin, G. (2008) “Milton Friedman and the Case Against Currency Monopoly,” Cato Journal, Vol. 28, No. 2, Spring/Summer.


114

Whitehead, L. (1976) “The Political Causes of Inflation,” Political Studies, 27(4), 564-577, December. ACKNOWLEDGEMENT The authors would like to thank the editor and the anonymous reviewers for their valuable comments that have led to tremendous improvement in the research manuscript. Many thanks go to the management of Lagos Business School for funded this research. BIOGRAPHY Doyin Salami holds a doctoral degree of Queen Mary College, University of London. He is a full-time faculty member at Lagos Business School (LBS), Pan-African University. He leads sessions in economic environment of business and had served as director of programs for five years until January 2005. He is a member of the Monetary Policy Committee of the Central Bank of Nigeria and had been a member of the Federal Government’s Economic Management Team. Dr Salami’s research interests include issues in corporate long-term financial management; macroeconomic policy; corporate competitiveness and risk management; and characteristics of small and medium enterprises (SMEs). +234 803 5767 562, email: [email protected] Ikechukwu Kelikume is currently a doctoral student of the Swiss University of Economics (SMC University, Switzerland) and leads sessions in microeconomic and macroeconomic environment of business at Lagos Business School. He research and consults in areas, which include macroeconomic modeling, financial and monetary economics as well as econometrics and quantitative method in economics. +234 813 7978 069, email: [email protected]


115

THE EFFECTS OF OWNERSHIP STRUCTURE AND COMPETITION ON RISK-TAKING BEHAVIOR: EVIDENCE FROM UAE CONVENTIONAL AND

ISLAMIC BANKS Hussein A. Hassan Al-Tamimi, University of Sharjah Neila Jellali, Institut Supérieur de Gestion de Tunis

ABSTRACT

The objective of this study is to examine the effect of ownership structure and competition on risk-taking behaviour of UAE banks during the period 1998–2010. The study covers 15 national banks, including eleven conventional banks and four are Islamic banks. The proportion of ownership by government, private sector and institutional ownership measures ownership structure. Concentration is used as a measure of competition. Three control variables are also included in the analysis economic condition, bank size and profitability. The main findings of this study are that UAE conventional banks are riskier than Islamic banks; concentration of UAE conventional national banks is negatively associated with bank risk-taking, but this inverse relationship is not confirmed in the case of Islamic banks; and the private ownership of UAE national banks is negatively associated with bank risk-taking. Finally, the results indicate that there is a significant difference between UAE conventional banks and Islamic banks regarding risk-taking behaviour. JEL: G20, G21 KEYWORDS: Ownership Structure, Competition, Risk-taking, UAE Conventional National Banks,

UAE Islamic banks INTRODUCTION

t the end of 2010, the UAE had 51commercial banks, of which 23 were national banks and the remaining 28 were foreign banks. Among the national banks, there were eight Islamic banks. The total assets of the national banks increased from AED 162.9 billion in 1998 (about US$ 44.4

billion) to AED 1,373.5 billion (about US$ 374 billion) in 2010. The total assets of Islamic banks increased from AED 9.2 billion in 1998 (about US$ 2.5 billion) to AED 262 billion (about US$ 71.4 billion) in 2010. The proportion of UAE Islamic banks’ assets increased from 4.1 percent of the UAE banking sector’s total assets and 5.65 percent of the UAE national banks’ assets in 1998 to 16.3 percent and 19.1 percent in 2010, respecively (Emirates Banks Association ). According to the UAE Central Bank, the number of branches of UAE Islamic banks in 2010 was 247 compared with 507 branches of the conventional banks, representing 28.6% of the total branches of the UAE commercial banks. The objective of this study is to examine the effect of ownership structure and competition on risk-taking behaviour of UAE banks. The current study represents an attempt to investigate the effect of concentration and the three types of ownership structure, government, instituttional and private sector, on risk-taking behaviour of UAE banks. Boubakri et al. (2005) indicate that the impact of concentration and the three types of ownership structure on risk taking is crucial in the context of banks. In the current study a comparison was made between the two sets of banks in UAE, the conventional and Islamic banks. The paper is organized as follows. In the following section we discuss the literature related to the ownership structure and concentration on risk-taking behaviour. This section is followed by an exposition

A

H. A. H. Al-Tamimi & N. Jellali | IJBFR ♦ Vol. 7 ♦ No. 2 ♦ 2013

116

of the empirical model and data. The fourth section is devoted to discussion of the empirical findings. In the final section a brief summary of the paper and conclusions concerning the main results are provided. LITERATURE REVIEW A large number of empirical studies address the effect of ownership structure and concentration on risk-taking behaviour. However, there is no such known study in the case of UAE banks, which represents the main motivation for the current study. The following is a summary of the main findings of some of the related studies. Hassan et al. (2005) investigated the impact of ownership structure and regulation on the risk-taking behaviour of commercial banks and savings and loans in the U.S. The authors found a positive relationship, but only for lower levels of ownership concentration. The results also indicate there are no significant risk-differentials between commercial banks and savings and loans. In addition, there are no significant risk-differentials between depository institutions that are state-chartered and those that are chartered nationally. Teresa and Dolores (2008) analyze the determinants of risk-taking in Spanish financial intermediaries, with special emphasis on the ownership structure and size of the different entities. It was found that the specific legal configuration of Spanish savings banks may lead them to differ from commercial banks in their risk behaviour. In particular, they may invest in riskier projects. Zou and Adams (2008) investigated the effect of corporate ownership on a firm’s equity risk and stock returns in China. They found that the various types of corporate ownership have important, but different impacts on equity risk and returns. Companies with more state ownership have higher risk and lower returns. In contrast, companies with more legal-person ownership tend to have lower risk and higher stock returns. Foreign and managerial ownership are found to have little effect on firms’ equity risk and returns. Kalluru (2009) examined the effect of ownership on performance and risk of commercial banks in India. The study, using t-test, fixed effects and random effects models, examines whether there exists any significant difference in performance and risk among state-owned banks, domestic private banks and foreign banks. The results indicate significant differences in the performance and risk, and foreign banks were more profitable and more risk-taking than the other two sets of banks. Bank capital and demand deposits were positively associated and loans were negatively associated with bank profitability, whereas size of banks and growth rate of economy were negatively associated with bank risk. Huang and Ming Huang (2009) examined the influences of ownership structure on the capital structure of Chinese-listed companies. The key factors used were state ownership, institutional ownership, and the risk of default. The results confirm that the expected default risk is important in explaining debt decisions. The results also indicate that ownership by the state and by institutions has a positive effect on corporate leverage in high-leveraged companies but not in low-leveraged firms. The authors defined state ownership as the proportion of shares owned by the government, while institutional ownership was measured as the ratio of shares owned to outstanding shares by domestic, foreign and founding institution investors. Chunet et al. (2011) investigated the effects of managerial ownership on the risk-taking behaviour of Korean and Japanese banks. The main finding is that managerial ownership alone does not affect either the risk or the profit levels of Korean banks. Whereas an increase in managerial ownership adds to the total risk of Japanese banks, increased risk-taking behaviour does not produce higher levels of profit for Japanese banks. Chou and Lin (2011) examined bank’s risk-taking and ownership structure of Taiwan banks. The main objective of their study was to investigate the effects of specific types of ownership on the risk-taking behaviours of banks under differential ownership structures. The results show that banks with higher inside management ownership and higher government ownership have higher overdue loans (higher risk) and lower capital adequacy ratios. Banks with higher foreign institution ownership and stronger relative governance strength are associated with lower overdue loans (lower risk) and higher regulatory capital. Fazlzade and Mahboubi (2011) investigated the role of ownership structure on firm performance of 137 listed firms of the Tehran stock exchange within the period 2001 to 2006. The ownership structure included ownership concentration, institutional ownership and institutional


117

ownership concentration. They concluded that ownership concentration doesn’t have any significant effect on firm performance but the two other variables are significant. Institutional ownership has positive significant effect on firm performance but the effect of concentrated institutional ownership is negative. The negative association between market concentration and bank risk taking has been established by many others including Dam, et al., (2011) and Repullo (2004). However, Forssbæcka and Shehzad (2011) tested the effect of banking-sector competition on bank risk-taking on a sample of around 400 European banks during the period 1995-2010. The results reveal a relatively clear indication of a positive competition-risk effect on risk-taking behaviour of European banks. From the above literature review, the following conclusions can be derived: 1.) Privatization improves the banking performance and decreases its exposure to risk, 2.) There is a negative relation between private ownership and bank risk, 3.) Managers holding a high level of shares, are exposed to a high level of risk, 4.) Companies with more state ownership tend to have higher risk and lower returns, 5.) Foreign banks are more profitable and more risk-taking than state and private banks 6.) Ownership by the state and by institutions has a positive effect on corporate leverage in high-leveraged companies, 7.) Managerial ownership alone does not affect either the risk or the profit levels, 8.) Higher government ownership involves higher overdue loans (higher risk), 9.) Higher foreign institution ownership involves lower overdue loans (lower risk), 10.) Ownership concentration doesn’t have any significant effect on firm performance, 11.) Institutional ownership has positive significant effect on firm performance and 12.) There is a negative relation between competition and bank risk. Based on the literature review, the following three hypotheses are formulated: H1: Competition among UAE national banks is negatively associated with bank risk-taking. It is assumed this will be a negative relationship because when market share is small (market concentration is low), banks are motivated to engage in more risky activities in order to increase their market share. Dam et al.(2011) indicate in this regard that banks become more risky as their markets share becomes more concentrated or when the market concentration is low, banks invest in more risky assets. The same finding has been reached by Jiménez et al., (2007), which is the negative relationship between competition and bank risk. H2: Private ownership of UAE national banks is negatively associated with bank risk-taking. If private ownership is dominant, banks are expected to be more conservative and take fewer risks than those with government-dominated ownership. La Porta et al. (2002) and Cornett et al. (2003) conclude that public-owned banks take more risks than other banks. They mention that the behaviour of public-owned banks is justified by political and social objectives, which reflects the crucial role of banks in the economy. On the other hand, banks highly dominated by government ownership are less risky because they are politically protected from a lack of financial resources (see Kwan, 2004). H3: Government ownership of UAE national banks is positively associated with bank risk-taking. It is assumed that the domination of government ownership encourages banks to take more risks in order to be able to implement their political and social role. H4: There is a significant difference in the level of risk-taking between the UAE conventional banks and Islamic banks. There are some similarities between the products and services of conventional and Islamic banks. Meanwhile there are some differences between these two types of banks, the key difference being that


118

Islamic banking is based on a Shariah foundation (Islamic principles). Therefore all dealings, transactions, business approaches, product features and investment focus are derived from the Shariah law, which leads to significant difference in many aspects of the operations from conventional banking. Based on these differences, it is assumed that there is a significant difference in the level of risk-taking between UAE conventional banks and Islamic banks. DATA AND METHODOLOGY The data used in this study were for the period 1998–2010 and the banks covered were 15 national banks, of which 11 were conventional banks and the remaining four were Islamic banks. The study did not cover all national banks because some of them are new or because they are small and the data are incomplete. The foreign banks were not included because it is difficult to get data about ownership structure as they represent branches of foreign banks. Furthermore, the national banks’ proportion of total assets was 78.1 % in 2010 which reflects the domination of the national banks in the UAE’s banking industry. To test the study’s hypotheses, the following two regression models were used: RISK= f (ECON, SIZE, ROA, CON) RISK = f (ECON, SIZE, ROA, GOG, INSIT, PRIV) Where:

- RISK – is a measure of risk = the risk-weighted assets / total assets;

- ECON - is a measure of economic conditions = GDP growth rate;

- SIZE - is a measure of banks’ size measured by total assets;

- ROA- is a measure of banks’ profitability

- CONT- is a measure of banks’ concentration;

- GOV- is a measure of government ownership

- INSIT- is a measure of institutional ownership;

- PRIV- is a measure of private sector ownership.

In addition, a dummy variable was used as an independent variable to reflect the bank type (TYPE) of which 0 was allocated to Islamic banks and 1 to conventional banks. The dependent variable in the two models was risk-taking behavior and it was measured by dividing the risk-weighted assets / total assets. This method was used by Shrieves and Dahl (1992), Jacques and Nigro (1997) and Murinde and Yaseen (2004). Concentration (CONT) was used as a measure of competition. The Herfindahl index was used in this regard (www.wikipedia.com) and calculated by the sum of the squares of the market share. The second measure was the proportion of shares owned by government, institutions and private sector (see Gursoy and Aydogan, 2002; Yi Huang et al., 2009; Hassan et al., 2005; Meca, 2009; Fazlzadeh et al., 2011). Regarding the control variables, three common variables were used; the first one was GDP growth rate which reflects economic conditions as there is a positive relationship between economic growth and financial development (see for example Wang, 2009 and Beck et al., 2008). The second control variable was total assets (the bank size) as larger banks would be able to diversify their assets risk (see Sanders et al., 1990 and Kalluru, 2009). The third control variable was profitability measured by ROA; it is assumed that a bank with high earnings might have higher risk (see Yi Huang et al., 2009


119

and Ta Ho and ShunWu, 2006) The data used in this study were mainly obtained from four sources: the UAE Central Bank annual reports and statistical bulletins, the UAE commercial banks annual reports published by the Emirates Banks Association, and the BankScope and ORISIS databases. Table 1 provides descriptive statistics for concentration, ownership and the risk factor. Table 1: Descriptive Statistics

N Minimum Maximum Mean Std. Deviation privcon 13 .37 .40 .3797 .01147 insitcon 13 .19 .22 .2074 .01058 govcon 13 .40 .42 .4092 .00643 privis 13 .53 .53 .53 .00000 insitis 13 .09 .09 .09 .00000

govis 13 .38 .38 .38 .00000 riskcon 13 .77 .92 .8198 .04847 riskis 13 .65 .78 .7073 .04699 contcon 13 .40 .49 .4476 .02405 contis 13 .00 .02 .0122 .00744

PRIVCON- is a measure of private sector ownership(Conventional banks); INSITCON- is a measure of institutional ownership (Conventional banks);; GOVCON- is a measure of government ownership(Conventional banks); PRIVIS- is a measure of private sector ownership(Islamic Banks); INSITIS- is a measure of institutional ownership(Islamic Banks); GOVIS- is a measure of government ownership (Islamic Banks); RISKCON – is a measure of Conventional banks’ risk = the risk-weighted assets / total assets; RISKIS – is a measure of Islamic banks’ risk = the risk-weighted assets / total assets; CONTCON- is a measure of Conventional banks’ concentration; CONTIS- is a measure of Islamic banks’ concentration. RESULTS The purpose of this study was to compare the ownership structure, competiton and risk-taking behaviour between the two types of commercial banks in the UAE, Islamic banks and conventional banks. It is worth mentioning here the main features of these two sets of banks regarding ownership structure and the level of risk. The ownership structure of both Islamic banks and conventional banks is almost the same. For the risk factor, the conventional banks are riskier than Islamic banks as expected, as the latter are more conservative. Table 2 reveals the ratio of the risk-weighted assets / total assets; it can be seen that the average of this ratio during the period 1998–2010 was 82 percent for the conventional banks compared with 70.7 percent for the Islamic banks. Table 2: Ratio of risk-weighted assets to total assets of UAE commercial banks

Year Islamic Banks Conventional Banks 1998 .68 .78 1999 .67 .81 2000 .68 .81 2001 .65 .78 2002 .68 .80 2003 .67 .77 2004 .72 .78 2005 .66 .78 2006 .78 .84 2007 .73 .84 2008 .75 .90 2009 .78 .92 2010 .77 .85 Average .71 .82

Table 2 shows the ratio of the risk-weighted assets / total assets, the average of this ratio for the period 1998–2010 was 82 percent for the conventional banks and 70.7 percent for Islamic banks.


120

The First Model The first model includes the bank risk-taking measured by the risk-weighted assets / total assets as a dependent variable and four independent variables; competition among UAE banks measured by concentration (CONC) and three control variables namely, economic conditions (ECON) measured by economic growth, the banks’ size (SIZE) measured by total assets, profitability measured by ROA. Table 3-a and Table 3-b provide a summary of the regression results of the first model for the two sets of banks. It can be seen from Table 2-a that the explanatory power of Table 3-a: Summary of Regression Results-UAE Conventional Banks

Beta t (Constant) 7.356 ECON .160 1.085 SIZE .667 4.396** ROA -.282 -1.312 CONC -.192 -1.172 R .942 R Square .887 Adjusted R Square .831 Std. Error of the Estimate 01994

Dependent Variable: RISK(the risk-weighted assets / total assets) **Statistically significant at the 5 percent level Note: This table shows the regression estimates of the equation: RISK= f (ECON, SIZE, ROA, CON)).The table reveals the coefficient values, the t-statistics and the significant level.

the adjusted 2R explained 83.1.8% of the variation of conventional national banks’ risk factor and 59% in the case of Islamic banks. The estimated coefficient of concentration (CON) was, as expected, negative but statistically insignificant. This result is expected because the conventional national banks were highly concentrated, and therefore there was no need to engage in more risky activities. However, the estimated coefficient of concentration (CON) of Islamic banks was unexpectedly positive, but statistically insignificant. The result was not expected in the case of Islamic banks because of the low concentration of these banks, which would lead them to engage in more risky activities. These results are consistent with the conclusions reached by Body Dam et al. (2011 and Jiménez et al.,(2007),). The results partially confirm the first hypothesis in which it is assumed that concentration among UAE banks is negatively associated with bank risk-taking. Table 3-b: Summary of Regression Results Islamic Banks

Beta t (Constant) 25.225 ECON .122 .556 SIZE .701 1.373 ROA -.027 -.141 CONC .152 .297 R .852 R Square .727 Adjusted R Square .590 Std. Error of the Estimate .03009

Dependent Variable: RISK(the risk-weighted assets / total assets) Note: This table shows the regression estimates of the equation: RISK= f (ECON, SIZE, ROA, CON)).The table reveals the coefficient values and the t-statistics. The Second Model This model includes the bank risk-taking as a dependent variable measured by the risk-weighted assets / total assets and six independent variables, including three control variables namely, economic conditions (ECON) measured by economic growth, the banks’ size (SIZE) measured by total assets and profitability measured by ROA; in addition to the ownership structure measured by three variables, namely the proportion of government ownership (GOVS), private sector ownership (PRIVS) and institutional


121

ownership (INSTS). Table 3-a and Table 4-b show the results of the regression model. It can be seen from the table that the adjusted R square is 89.9% in the case of conventional banks and 75.3% for Islamic banks. This indicates that the six independent variables explain 89.9% and 75.3% of the risk-taking behaviour by UAE conventional national banks and Islamic banks respectively. For conventional banks, the estimated coefficients of two independent variables were, as expected, negative and statistically significant at the 5 percent level in the case of private ownership and 10 percent in the case of institutional ownership. For Islamic banks, the estimated coefficients of private ownership and institutional ownership were as expected negative and statistically significant at the 10 percent level. In general, the results confirm the second hypothesis which states: Private ownership of UAE national banks is negatively associated with bank risk-taking. This finding is consistent with the conclusions of Table 4-a: Summary of Regression Results-UAE Conventional Banks

Beta t (Constant) 2.049 ECON .284 2.704** SIZE .977 2.767** ROA -.384 -2.448** GOVS .293 1.117 PRIVS -.978 -2.612** INSTS -.899 -2.006* R .974 R Square .949 Adjusted R Square .899 Std. Error of the Estimate .01544

Dependent Variable: RISK(the risk-weighted assets / total assets)n bNote: This table shows the regression estimates of the equation: RISK = f (ECON, SIZE, ROA, GOG, INSIT, PRIV) The table reveals the coefficient values, the t-statistics and the significant level. **Statistically significant at the 5 percent level *Statistically significant at the 10 percent level La Porta et al. (2002) and Cornett et al. (2003). The estimated coefficient of government ownership is as expected positive but statistically insignificant in the case conventional banks, but it is unexpectedly negative and statistically significant at 10 percent in the case of Islamic banks. The results did not confirm hypothesis three which states: Government ownership of UAE national banks is positively associated with bank risk-taking. This is consistent with conclusion of Kwan (2004). Table 4-b: Summary of Regression Results- UAE Islamic Banks

Beta t (Constant) 2.217 ECON -.064 -.334 SIZE 2.058 3.462* ROA -.132 -.812 GOVS -2.074 -2.110** PRIVS -2.404 -1.979** INSTS -.897 -2.226** R .936 R Square .876 Adjusted R Square .753 Std. Error of the Estimate .02337

Dependent Variable: RISK(the risk-weighted assets / total assets) Note: This table shows the regression estimates of the equation: RISK = f (ECON, SIZE, ROA, GOG, INSIT, PRIV) The table reveals the coefficient values, the t-statistics and the significant level. **Statistically significant at the 5 percent level *Statistically significant at the 10 percent level An attempt has been made to examine the above mentioned model by considering the data for all UAE banks( conventional and Islamic banks). Table 5 reveals the results of the test. It can be seen that all the coefficient values are statistically insignificant, which gives the support of dividing the sample into two groups, conventional and Islamic banks. Finally, the difference between the UAE conventional national banks and Islamic banks regarding risk-taking behaviour was examined. Table 4 shows the results of One-Way ANOVA analysis for the differences between UAE Islamic banks and the conventional banks regarding risk-taking behaviour to


122

test hypothesis four which states: There is a significant difference in the level of risk-taking between the UAE conventional banks and Islamic banks. It can be seen from the table that there is as expected a significant difference between the UAE conventional banks and Islamic banks regarding risk-taking behaviour, that is statistically significant at the 1 percent level. The results are expected because of the nature of operations, activities and the risk exposure of each type of bank. These results confirm hypothesis four. Table 5: Summary of Regression Results- UAE Banks

Beta t (Constant) .728 ECON 1.691 1.116 SIZE -.454 -.388 ROA -.150 -.667 GOVS .375 .974 PRIVS 1.691 1.116 INSTS -.724 -.981 R .954 R Square .911 Adjusted R Square .786 Std. Error of the Estimate .02083

Dependent Variable: RISK(the risk-weighted assets / total assets) Note: This table shows the regression estimates of the equation: RISK = f (ECON, SIZE, ROA, GOG, INSIT, PRIV) The table reveals the coefficient values, the t-statistics and the significant level. Table 5: The Results of Analysis of Variance for Islamic Banks and Conventional Banks

Sum of Squares df Mean Square F Sig. RISK Between Groups

.082

1

.082

36.080 .000

Within Groups .055 24 .002 Total .137 25

RISK is the risk-weighted assets / total assets) The two groups are conventional banks and Islamic banks CONCLUDING COMMENTS The objective of this study was to examine the effect of the ownership structure and competition on risk-taking behaviour of the UAE banks during the period 1998–2010. Concentration was used as measure of competition, whereas the proportion of ownership by government, private sector and institutional ownership was used in the case of ownership structure. In addition, we used three control variables: economic condition, bank size and profitability. To test the study’s hypotheses, two regression models were used, in which concentration and ownership were used alternatively as a dependent variable in the two models.. The main findings of this study are: 1) that the UAE conventional banks are riskier than the Islamic banks, which is to be expected because the latter are more conservative; 2) that the ownership structure of UAE conventional national banks is negatively associated with bank risk-taking when concentration is used as a measure of ownership structure, but that this inverse relationship is not confirmed in the case of Islamic banks; 3) that the private ownership of UAE national banks is negatively associated with bank risk-taking; but 4) that the results did not support the assumption that government ownership of UAE national banks is positively associated with bank risk-taking in the case of both sets of banks, the conventional and Islamic banks; and finally 5) that there is as expected a significant difference between UAE conventional banks and Islamic banks regarding risk-taking behaviour. Among the limitations of this research is the data availability, as the data was available somehow for a short period. In addition, the study did not cover all the national banks because of insufficient data. For further research, it is highly recommended to cover all UAE banks and it is interesting to examine the same topic to include commercial banks of the Gulf region..


123

REFERENCES Beck, T.Demirguc-Kunt, A., Laeven, L.,Levine, R.(2008), “Finance, Firm Size, and Growth”, Journal of Money, Credit & Banking (Blackwell),Vol.40(7), p.1379-1405. Bor-Y.,B.and Ming, H.C.(2009), “The Influences Of Ownership Structure: Evidence From China”, Journal of Development Areas, Vol.45(1), p.209-227. Boubakri, NR., Cosset J.C., Fischer, K. and Omrane, G. (2005), “Privatization and bank performance in developing countries”, Journal of Banking and Finance, Vol.29, p. 2015-2041. Chou S.and Lin F. (2011), “Bank's risk-taking and ownership structure – evidence for economics in transition stage”, Vol.43(12), p. 1551-1564 Chun, S.E, Nagano, M., Lee, M., ( 2011), “Ownership Structure and Risk-taking Behaviour: Evidence from Banks in Korea and Japan”, Asian Economic Journal, Vol. 25( 2) , p. 151-175. Cornett, M.., L.G. Khaksari, S. And Tehranian, H. (2003), “The Impact of Corporate Governance one Performance Differences in Privately-Owned versus State-Owned Banks: International year Comparison”, Working Paper, Boston College. Dam, K., Villar,M.E. Pages,S.S(2011), “On the Relationship between Market Concentration and Bank Risk Taking”, Working paper, No.36, Universitat de les Illes Balears, Departament d'Economía Aplicada. Emirates Bank Association(2009), “Financial Position of Commercial Banks in the UAE”, Abu Dhabi. Forssbæcka, J. and Shehzad, C.T.(2011), “Competition and risk-taking in European Banking”, http://snee.org/filer/papers/632.pdf. Jiménez, J. Lopez, J.A., and Saurine,J.(2007), “How Does Competition Impact Bank Risk-Taking?”, Federal Reserve Bank of san Francisco, Working Paper Series, No.23. Meca, E.G. and Ballesta,J.P.(2011), “Firm value and ownership structure in the Spanish capital market”, Corporate governance, Vol.11(1), p. 41-53 Fazlzadeh, A. Hendi, A. and Mahboubi,K.(2011), “Performance in Listed Firms of Tehran Stock Exchange Based on the Type of the Industry”, International Journal of Business and Management, Vol.6(3), p. 249-266. Gursoy, G. and Aydogan, K. (2002), “Equity Ownership Structure, Risk Taking, and Performance”, Emerging Markets, Finance and Trade, Vol.38, p. 6-25. Hassan,M.K.,Karels,G.V., Wilcox,S.E.(2005), “Impact of Ownership Structure and Regulation on the risk-Taking Behavior of Depository Institutions”, Journal of Financial Management and Analysis, Vol.18(2), p. 34-40. Jacques, K. and P. Nigro (1997), “Capital Risk-based, capital portfolio risk, and bank: With simultaneous equations approach”, Journal of Economics and Business, 49, 533-547.


124

Kalluru ,S.R.(2009), Ownership Structure, “Performance and Risk in Indian Commercial Banks”, The IUP Journal of Applied Finance, Vol.15(8,), p. 31-45. Kirchmaier T and Grant J (2004), “Who Governs? Corporate Ownership and Control Structures in Europe”, available at http://ssrn.com/abstract=555877 Kwan, S.H (2004), “Risk and return of publicly held versus privately owner banks”, FRBNY, Economic policy review, September,97-107. La Porta, R., Lopez-de-Silanes, F., Shleifer, A.(2002), “Government ownership of banks”, Journal of Finance,Vol.57, p. 265–301. Meca ,E. And Ballesta,J. P.S.(2011), “Firm value and ownership structure in the Spanish capital market”, Corporate Governance,Vol.11(1), p. 41-53 Murinde V. and Yaseen H. (2004), “The impact of Basle capital Regulations Agreement one bank and risk behaviour: 3D evidence from the Middle East and North Africa (CARRIED OUT) area”, Third International Conference of the Centre for Regulation and Competition, Working paper. Repullo, R.( 2004), “Capital Requirements, Market Power, and Risk-Taking in Banking”, Journal of Financial Intermediation, Vol.13, p. 156–182. Saunders, A., E. Strock, and NR. G. Travlos (1990), Ownership structure, “Deregulation, and bank risk taking”, Journal of Finance, Vol.45, p. 643-654. Shrieves, R. E. and Dahl D., (1992), “The relationship between risk and capital in commercial banks”. Journal of Banking and Finance, Vol.16, p. 439- 457. Teresa, G. and Dolores, R.M.(2008), “Risk-taking behaviour and ownership in thebanking industry: The Spanish evidence”, Journal of Economics & Business, Vol.60(4), p. 332.YiHuang, B., MinLin,C. Ming Huang,C.(2011), “The influences of ownership structure: Evidence from China”, Journal of Development Areas,Vol.45(1), p. 209-227. Wang ,Fuhmei (2009), “Financial Distorstions and Economic Growth: Empirical Evidence”,FulEmerging Markets Finance &Trade,Vol.45(3), p. 56-66. Zou , H and Adams,M.B.(2008), “Corporate ownership, equity risk and returns in the People’s Republic of China”, Journal of International Business Studies, Vol.39, p. 1149– 1168. BIOGRAPHY Hussein A. Hassan Al-Tamimi is a Professor of Finance at Department of Accounting, Finance and Economics. Professor Al-Tamimi can be contacted at College of Business Administration University of Sharjah, P.O. Box 27272, Sharjah, United Arab Emirates, E-mail: [email protected]

Neila Jellali can be contacted at Unité de recherche finance et stratégies des affaires, Institut Supérieur de Gestion de Tunis. 41, rue de la liberté cite bouchoucha, 2000 le Bardo. E-mail: [email protected]

REVIEWERS The IBFR would like to thank the following members of the academic community and industry for their much appreciated contribution as reviewers.

Hisham Abdelbaki, University of Mansoura - EGYPT

Isaac Oluwajoba Abereijo, Obafemi Awolowo University

Naser Abughazaleh, Gulf University for Science and Technology

Nsiah Acheampong, University of Phoenix

Vera Adamchik, University of Houston-Victoria

Iyabo Adeoye, National Horticultural Research Instittute, Ibadan, Nigeria.

Michael Adusei, Kwame Nkrumah University of Science and Technology

Moh'd Ajlouni, Yarmouk University

Sylvester Akinbuli, University of Lagos

Anthony Akinlo, Obafemi Awolowo University

Yousuf Al-Busaidi, Sultan Qaboos University

Khaled Aljaaidi, Universiti Utara Malaysia

Hussein Al-Tamimi, University of Sharjah

Paulo Alves, CMVM, ISCAL and Lusofona University

Ghazi Al-Weshah, Albalqa Applied University

Glyn Atwal, Groupe Ecole Supérieure de Commerce de Rennes

Fabiola Baltar, Universidad Nacional de Mar del Plata

Susan C. Baxter, Bethune-Cookman College

Nagib Bayoud, Tripoli University

Ahmet Bayraktar, Rutgers University

Myrna Berrios, Modern Hairstyling Institute

Kyle Brink, Western Michigan University

Karel Bruna, University of Economics-Prague

Priyashni Chand, University of the South Pacific

Yahn-Shir Chen, National Yunlin University of Science and Techology, Taiwan

Wan-Ju Chen, Diwan College of Management

Bea Chiang, The College of New Jersey

Te-Kuang Chou, Southern Taiwan University

Shih Yung Chou, University of the Incarnate Word

Monica Clavel San Emeterio, University of La Rioja

Caryn Coatney, University of Southern Queensland

Michael Conyette, Okanagan College

Rajni Devi, The University of the South Pacific

Leonel Di Camillo, Universidad Austral

Steven Dunn, University of Wisconsin Oshkosh

Mahmoud Elgamal, College of Business Administration - Kuwait University

Esther Enriquez, Instituto Tecnologico de Ciudad Juarez

Ernesto Escobedo, Business Offices of Dr. Escobedo

Zaifeng Fan, University of Wisconsin whitewater

Olga Ferraro, University of Calabria

William Francisco, Austin Peay State University

Carmen Galve-Górriz, Universidad de Zaragoza

Blanca Rosa Garcia Rivera, Universidad Autónoma De Baja California

Peter Geczy, AIST

Lucia Gibilaro, University of Bergamo

Carlos Alberto González Camargo, Universidad Jorge Tadeo Lozano

Hector Alfonso Gonzalez Guerra, Universidad Autonoma de Coahuila

Hongtao Guo, Salem State University

Danyelle Guyatt, University of Bath

Shahriar Hasan, Thompson Rivers University

Zulkifli Hasan, Islamic University College of Malaysia

PENG HE, Investment Technology Group

Niall Hegarty, St. John's University

Claudia Soledad Herrera Oliva, Universidad Autónoma de Baja California

Paulin HOUANYE, University of International Business and Education, School of Law

Daniel Hsiao, University of Minnesota Duluth

Xiaochu Hu, School of Public Policy, George Mason University

Biqing Huang, Angelo State University Member, Texas Tech University System ASU Station #10908

Jui-Ying Hung, Chatoyang University of Technology

Fazeena Hussain, University of the South Pacific

Shilpa Iyanna, Abu Dhabi University

Shilpa Iyanna, Abu Dhabi University

Sakshi Jain, University of Delhi

Raja Saquib Yusaf Janjua, CIIT

Tejendra N. Kalia, Worcester State College

Krishna Kasibhatla, North Carolina A&T State University

Gary Keller, Eastern Oregon University

Ann Kelley, Providence college

Ann Galligan Kelley, Providence College

Ifraz Khan, University of the South Pacific

Halil Kiymaz, Rollins College

Susan Kowalewski, D'Youville College

Bamini KPD Balakrishnan, Universiti Malaysia Sabah

Bohumil Král, University of Economics-Prague

Jan Kruger, Unisa School for Business Leadership

Christopher B. Kummer, Webster University-Vienna

Mei-Mei Kuo, JinWen University of Science & Technology

Mary Layfield Ledbetter, Nova Southeastern University

John Ledgerwood, Embry-Riddle Aeronautical University

Yen-Hsien Lee, Department of Finance, Chung Yuan Christian University

YingChou Lin, Missouri University of Science and Technology

Shulin Lin, Hsiuping University of Science and Technology

Melissa Lotter, Tshwane University of Technology

Xin (Robert) Luo, Virginia State University

Andy Lynch, Southern New Hampshire University

Eduardo Macias-Negrete, Instituto Tecnologico de Ciudad Juarez

Abeer Mahrous, Cairo university

Gladys Marquez-Navarro, Saint Louis University

Jesús Apolinar Martínez Puebla, Universidad Autónoma De Tamaulipas

Cheryl G. Max, IBM

Francisco Jose May Hernandez, Universidad del Caribe

Aurora Irma Maynez Guaderrama, Universidad Autonoma de Ciudad Juarez

Romilda Mazzotta, University of Calabria

Mary Beth McCabe, National University

Avi Messica, Holon Institute of Technology

Scott Miller, Pepperdine University

Cameron Montgomery, Delta State University

Sandip Mukherji, Howard University

Tony Mutsue, Iowa Wesleyan College

Cheedradevi Narayanasamy, Graduate School of Business, National University of Malaysia

Erwin Eduardo Navarrete Andrade, Universidad Central de Chile

Dennis Olson, Thompson Rivers University

Godwin Onyeaso, Shorter University

Bilge Kagan Ozdemir, Anadolu University

Dawn H. Pearcy, Eastern Michigan University

Eloisa Perez, MacEwan University

Iñaki Periáñez, Universidad del Pais Vasco (spain)

Pina Puntillo, University of Calabria (Italy)

Rahim Quazi, Prairie View A&M University

Anitha Ramachander, New Horizon College of Engineering

Charles Rambo, University of Nairobi, Kenya

Prena Rani, University of the South Pacific

Alma Ruth Rebolledo Mendoza, Universidad De Colima

Kathleen Reddick, College of St. Elizabeth

Maurizio Rija, University of Calabria.

Carmen Rios, Universidad del Este

Matthew T. Royle, Valdosta State University

Tatsiana N. Rybak, Belarusian State Economic University

Rafiu Oyesola Salawu, Obafemi Awolowo University

Paul Allen Salisbury, York College, City University of New York

Leire San Jose, University of Basque Country

Celsa G. Sánchez, CETYS Universidad

I Putu Sugiartha Sanjaya, Atma Jaya Yogyakarta University, Indonesia

Sunando Sengupta, Bowie State University

Brian W. Sloboda, University of Phoenix

Adriana Patricia Soto Aguilar, Benemerita Universidad Autonoma De Puebla

Smita Mayuresh Sovani, Pune University

Alexandru Stancu, University of Geneva and IATA (International Air Transport Association)

Jiří Strouhal, University of Economics-Prague

Qian Sun, Kutztown University

Diah Suryaningrum, Universitas Pembangunan Nasional Veteran Jatim

James Tanoos, Saint Mary-of-the-Woods College

Jeannemarie Thorpe, Southern NH University

Ramona Toma, Lucian Blaga University of Sibiu-Romania

Jorge Torres-Zorrilla, Pontificia Universidad Católica del Perú

William Trainor, East Tennessee State University

Md Hamid Uddin, University Of Sharjah

Ozge Uygur, Rowan University

K.W. VanVuren, The University of Tennessee – Martin

Vijay Vishwakarma, St. Francis Xavier University

Ya-Fang Wang, Providence University

Richard Zhe Wang, Eastern Illinois University

Jon Webber, University of Phoenix

Jason West, Griffith University

Wannapa Wichitchanya, Burapha University

Veronda Willis, The University of Texas at San Antonio

Bingqing Yin, University of Kansas

Junye Yu, Louisiana State University

REVIEWERS The IBFR would like to thank the following members of the academic community and industry for their much appreciated contribution as reviewers.

Haydeé Aguilar, Universidad Autónoma de Aguascalientes

María Antonieta Andrade Vallejo, Instituto Politécnico Nacional

Olga Lucía Anzola Morales, Universidad Externado de Colombia

Hector Luis Avila Baray, Instituto Tecnologico De Cd. Cuauhtemoc

Graciela Ayala Jiménez, Universidad Autónoma de Querétaro

Ana Cecilia Bustamante Valenzuela, Universidad Autonoma De Baja California

Carlos Alberto Cano Plata, Universidad De Bogotá Jorge Tadeo Lozano

Alberto Cardenas, Instituto Tecnologico De Cd. Juarez

Edyamira Cardozo, Universidad Nacional Experimental De Guayana

Sheila Nora Katia Carrillo Incháustegui, Universidad Peruana Cayetano Heredia

emma casas medina, Centro de Estudios Superiores del Estado de Sonora

Benjamín Castillo Osorio, Universidad Cooperativa De Colombia y Universidad De Córdoba

Benjamin Castillo Osorio, Universidad del Sinú-Sede Monteria

María Antonia Cervilla de Olivieri, Universidad Simón Bolívar

Cipriano Domigo Coronado García, Universidad Autónoma de Baja California

Semei Leopoldo Coronado Ramírez, Universidad de Guadalajara

Esther Eduviges Corral Quintero, Universidad Autónoma de Baja California

Dorie Cruz Ramirez, Universidad Autonoma Del Estado De Hidalgo /Esc. Superior De Cd. Sahagún

Edna Isabel De La Garza Martinez, Universidad Autónoma De Coahuila

Javier de León Ledesma, Universidad de Las Palmas de Gran Canaria - Campus Universitario de Tafira

Hilario Díaz Guzmán, Universidad Popular Autónoma del Estado de Puebla

Cesar Amador Díaz Pelayo, Universidad de Guadalajara, Centro Universitario Costa Sur

Avilés Elizabeth, CICESE

Avilés Elizabeth, CICESE

Ernesto Geovani Figueroa González, Universidad Juárez del Estado de Durango

Ana Karen Fraire, Universidad De Gualdalajara

Carmen Galve-Górriz, Universidad de Zaragoza

Teresa García López, Universidad Veracruzana

Helbert Eli Gazca Santos, Instituto Tecnológico De Mérida

Denisse Gómez Bañuelos, CESUES

Ana Ma. Guillén Jiménez, Universidad Autónoma de Baja California

Ana Ma. Guillén Jiménez, Universidad Autónoma de Baja California

Araceli Gutierrez, Universidad Autonoma De Aguascalientes

Andreina Hernandez, Universidad Central de Venezuela

Arturo Hernández, Universidad Tecnológica Centroamericana

Alejandro Hernández Trasobares, Universidad de Zaragoza

Alma Delia Inda, Universidad Autonoma Del Estado De Baja California

Terrance Jalbert, The IBFR

Gaspar Alonso Jiménez Rentería, Instituto Tecnológico de Chihuahua

Lourdes Jordán Sales, Universidad de Las Palmas de Gran Canaria

Santiago León Ch., Universidad Marítima del Caribe

Graciela López Méndez, Universidad de Guadalajara-Jalisco

Virginia Guadalupe López Torres, Universidad Autónoma de Baja California

Angel Machorro Rodríguez, Instituto Tecnológico de Orizaba

Cruz Elda Macias Teran, Universidad Autonoma de Baja California

Aracely Madrid, ITESM, Campus Chihuahua

Deneb Magaña Medina, Universidad Juárez Autónoma de Tabasco

Carlos Manosalvas, Universidad Estatal Amazónica

Gladys Yaneth Mariño Becerra, Universidad Pedagogica y Tecnológica de Colombia

Omaira Cecilia Martínez Moreno, Universidad Autónoma de Baja California-México

Jesus Carlos Martinez Ruiz, Universidad Autonoma De Chihuahua

Alaitz Mendizabal, Universidad Del País Vasco

Alaitz Mendizabal Zubeldia, Universidad del País Vasco/ Euskal Herriko Unibertsitatea

Fidel Antonio Mendoza Shaw, Universidad Estatal De Sonora

Juan Nicolás Montoya Monsalve, Universidad Nacional de Colombia-Manizales

Jennifer Mul Encalada, Universidad Autónoma De Yucatán

Alberto Elías Muñoz Santiago, Fundación Universidad del Norte

Bertha Guadalupe Ojeda García, Universidad Estatal de Sonora

Erika Olivas, Universidad Estatal de Sonora

Erick Orozco, Universidad Simon Bolivar

José Manuel Osorio Atondo, Centro de Estudios Superiores del Estado de Sonora

Luz Stella Pemberthy Gallo, Universidad del Cauca

Andres Pereyra Chan, Instituto Tecnologico De Merida

Iñaki Periáñez, Universidad del Pais Vasco (spain)

Adolfo León Plazas Tenorio, Universidad del Cauca

Hector Priego Huertas, Universidad De Colima

Juan Carlos Robledo Fernández, Universidad EAFIT-Medellin/Universidad Tecnologica de Bolivar-Cartagena

Humberto Rosso, Universidad Mayor de San Andres

José Gabriel Ruiz Andrade, Universidad Autónoma de Baja California-México

Antonio Salas, Universidad Autonoma De Chihuahua

Claudia Nora Salcido, Facultad de Economía Contaduría y Administración Universidad Juarez del Estado de Durango

Juan Manuel San Martín Reyna, Universidad Autónoma de Tamaulipas-México

Francisco Sanches Tomé, Instituto Politécnico da Guarda

Edelmira Sánchez, Universidad Autónoma de Ciudad Juárez

Deycy Janeth Sánchez Preciado, Universidad del Cauca

María Cristina Sánchez Romero, Instituto Tecnológico de Orizaba

María Dolores Sánchez-Fernández, Universidade da Coruña

Luis Eduardo Sandoval Garrido, Universidad Militar de Nueva Granada

Pol Santandreu i Gràcia, Universitat de Barcelona, Santandreu Consultors

Victor Gustavo Sarasqueta, Universidad Argentina de la Empresa UADE

Jaime Andrés Sarmiento Espinel, Universidad Militar de Nueva Granada

Jesus Otoniel Sosa Rodriguez, Universidad De Colima

Edith Georgina Surdez Pérez, Universidad Juárez Autónoma de Tabasco

Jesús María Martín Terán Gastélum, Centro de Estudios Superiores del Estado de Sonora

Jesús María Martín Terán Gastélum, Centro de Estudios Superiores del Estado de Sonora

Jesus María Martín Terán Terán Gastélum, Centro de Estudios Superiores del Estado de Sonora

Maria De La Paz Toldos Romero, Tecnologico De Monterrey, Campus Guadalajara

Abraham Vásquez Cruz, Universidad Veracruzana

Angel Wilhelm Vazquez, Universidad Autonoma Del Estado De Morelos

Lorena Vélez García, Universidad Autónoma de Baja California

Alejandro Villafañez Zamudio, Instituto Tecnologico de Matamoros

Hector Rosendo Villanueva Zamora, Universidad Mesoamericana

Oskar Villarreal Larrinaga, Universidad del País Vasco/Euskal Herriko Universitatea

Delimiro Alberto Visbal Cadavid, Universidad del Magdalena

HOW TO PUBLISH

Submission Instructions

The Journal welcomes submissions for publication consideration. Authors wishing to submit papers for publication consideration should visit our website at www.theibfr.com/journal.htm, under “How to Submit a Paper.” Complete directions for manuscript submission are available at the Journal website www.theIBFR.com/journal.htm. Papers may be submitted for initial review in any format. However, authors should take special care to address spelling and grammar issues prior to submission. Authors of accepted papers are required to precisely format their document according to the guidelines of the journal.

There is no charge for paper reviews. The normal review time for submissions is 90-120 days. However, authors desiring a quicker review may elect to pay an expedited review fee. Authors of accepted papers are required to pay a publication fee based on the length of the manuscript. Please see our website for current publication and expedited review rates.

Authors submitting a manuscript for publication consideration must guarantee that the document contains the original work of the authors, has not been published elsewhere, and is not under publication consideration elsewhere. In addition, submission of a manuscript implies that the author is prepared to pay the publication fee should the manuscript be accepted.

Subscriptions

Individual and library subscriptions to the Journal are available. Please contact us by mail or by email to: [email protected] for updated information.

Contact Information

Mercedes Jalbert, Managing Editor The IBFRP.O. Box 4908Hilo, HI [email protected]

Website

www.theIBFR.org or www.theIBFR.com

Review of Business & Finance Studies Review of Business & Finance Studies (ISSN: 2150-3338 print and 2156-8081 online) publishes high-quality studies in all areas of business, finance and related fields. Empirical, and theoretical papers as well as case studies are welcome. Cases can be based on real-world or hypothetical situations.

All papers submitted to the Journal are double-blind reviewed. The Journal is distributed in print and through SSRN and EBSCOhost Publishing, with nation-wide access in more than 70 countries. The Journal is listed in Cabell’s directory.

The journal accept rate is between 15 and 25 percent

Business Education & AccreditationBE A

AT Accounting

Taxation&

Accounting and Taxation (AT)

Accounting and Taxation (AT) publishes high-quality articles in all areas of accounting, auditing, taxation and related areas. Theoretical, empirical and applied manuscripts are welcome for publication consideration.

All papers submitted to the Journal are double-blind reviewed. AT is listed in Cabell’s and Ulrich’s Periodicals Directory. The Journal is distributed in print, through SSRN and EBSCOHost publishing, with presence in over 70 countries.

The journal acceptance rate is between 5 and 15 percent.

Business Education and Acreditation (BEA)Business Education & Accreditation publishes high-quality articles in all areas of business education, curriculum, educational methods, educational administration, advances in educational technology and accreditation. Theoretical, empirical and applied manuscripts are welcome for publication consideration.

All papers submitted to the Journal are double-blind reviewed. BEA is is listed in Cabell’s and Ulrich’s Periodicals Directory. The Journal is distributed in print, through SSRN and EBSCOHost publishing, with presence in over 70 countries.

The journal acceptance rate is between 15 and 25 percent.

PUBLICATION OPPORTUNITIES

Documents

· RThe International Journal Business of and Finance ESEARCH CONTENTS Risk Analysis Using Regression Quantiles: Evidence from International Equity Markets 1 Hongtao Guo, Miranda