What are the risks when investing in thin emerging equity markets: Evidence from the Arab world

Int. Fin. Markets, Inst. and Money 17 (2007) 102–123

What are the risks when investing in thin emergingequity markets: Evidence from the Arab world

Eric Girard a,∗, Mohamed Omran b

a School of Business, Finance Department, Siena College, 515 Loudon Road, Loudonville, NY 12211, USAb College of Management, Arab Academy for Science & Technology, Alexandria, Box 1029, Egypt

Received 21 April 2005; accepted 20 September 2005Available online 24 October 2005

Abstract

This study attempts to identify the risks involved when investing in five emerging Arab capital markets. Wefirst find that a constant beta is not a good proxy for risk in these thinly traded emerging markets. However,firms’ fundamentals and country risk rating factors prove significant in explaining the cross-sections of stockreturns. The paper provides three important contributions to the literature on asset pricing in emerging capitalmarkets: (i) we show how country risk ratings can be aggregated into a country risk factor; (ii) we add toa growing literature suggesting that, in markets other than the US, it is possible to find large and growthstocks to be riskier than small and value stocks; (iii) we determine that despite economic, financial andpolitical reforms, issues related to financial transparency and political instability are still powerful obstaclesto investments in these nascent emerging markets.© 2005 Elsevier B.V. All rights reserved.

JEL classification: F3; G1; N2

Keywords: CAPM; Multifactor model; Arab emerging markets

1. Introduction

Equity risk premiums are central components of every risk and return model in finance andare fundamental and critical components in portfolio management. While the return generatingprocess of individual stock is more established for developed markets, risk components thatdetermine risk premiums are difficult to evaluate in thin emerging markets, where the historical

∗ Corresponding author. Tel.: +1 518 783 4133.E-mail addresses: [email protected] (E. Girard), [email protected] (M. Omran).

1042-4431/$ – see front matter © 2005 Elsevier B.V. All rights reserved.doi:10.1016/j.intfin.2005.09.003

E. Girard, M. Omran / Int. Fin. Markets, Inst. and Money 17 (2007) 102–123 103

data tend to be limited. Our paper attempts to identify risks affecting long-run stock returns in anArab block comprised of five thin emerging Arab markets.

Many Arab countries have embarked on a process of privatization and stock market liberaliza-tion with the goal of deepening their markets and improving corporate governance for a nascentprivate sector. As a result, it is worthwhile to study the region on its own, and to see what impactthese reforms have had on the relationship between risk and return. In other words, do stockreturns reflect the ongoing infrastructure improvement in the Arab stock markets? What risks areinvolved when investing in these markets?

Our paper addresses a research area previously ignored in finance literature. We examinethe relationship between risk and stock returns in thinly traded emerging markets using notonly fundamental risk measures but also country risk scores. A previous study has investigatedthe relationship between volatility and returns in Arab equity indices (Girard et al., 2003a).Several papers have investigated the relationship between stock returns and fundamental riskattributes in emerging markets (Claessens et al., 1998; Lyn and Zychowicz, 2004; Ramchar-ran, 2004). Numerous papers have investigated the relationship between country index returnsand composite risks—i.e., a weighted average of the 22 risk scores used in our study (Erbet al., 1995, 1996a,b, 1998; Beakaert and Harvey, 2002, 2003; Harvey et al., 2002). Finally,several papers have specifically investigated the relationship between country index returnsand demographics or only one type of composite risk—i.e., political risk alone which is aweighted average of twelve risk scores used in this study (Diamonte et al., 1996; Erb et al.,1997).

We first consider the conventional approach to estimating risk premiums, which uses the CapitalAsset Pricing Model (CAPM) developed by Sharpe and Lintner, and we evaluate its weaknesses. Inessence, we follow Omran (2005) by testing for the null hypothesis of no cross-sectional abnormalreturns for a sample of more than 100 firms over 5 years. Second, we investigate a multifactorextension to the CAPM by showing how the cross-sections of 4 fundamental firm-specific and22 country-specific risk scores can be linearly related to risk premiums in five emerging Arabcapital markets. The firm-specific risk scores include beta, market-to-book value, size and industrytype while the country-specific risk scores comprise 12 political, 5 economic and 5 financial riskscores. In conducting the investigation, we employ a principal component analysis methodology(Chen et al., 1986; Groenewold and Fraser, 1997) to reduce the factor loading, and identify thesignificance of each risk factor’s effects on long-term stock risk premiums. Finally, as in Chen(1983) we test the information content of our multifactor expression as compared to three nestedmodels.

In our study: (i) we support, with strong evidence, an argument that stock returns in all Arabcountries do not follow the unconditional CAPM, but we reject a constant beta as a good proxyfor risk in determining stock returns and (ii) we find that fundamental attributes and coun-try risk scores are both significant in explaining the return generating process of individualfirms within our Arab block universe. In fact, we show that a model with both fundamentalsand risk scores is a significantly better explanatory tool than either the CAPM, or a modelwhich only includes a firm’s fundamentals, or a model based on country composite risk ratingsalone.

The remainder of the paper is organized as follows. Section 2 briefly discusses the relevantliterature. Data selection, research methodology and empirical models are described in Section 3.Section 4 provides analysis and interpretations of the empirical findings and Section 5 concludesthe paper.

104 E. Girard, M. Omran / Int. Fin. Markets, Inst. and Money 17 (2007) 102–123

2. Literature review

Asset valuation is based on deciding what risks explain required rates of return. Portfoliotheory tells us that only systematic risks (i.e., risks to which many securities are exposed) canbe associated with a premium in financial markets. This relationship has not been empiricallyaccepted because it is generally not clear how to measure risk. Researchers have long addressedthis issue using different versions of Sharpe and Lintner’s seminal CAPM. A plethora of articleshave investigated the cross-section risk premiums and beta in capital markets. CAPM tests haveled to the puzzling conclusion that the relationship between market premium and beta can bepositive, negative or insignificant. For instance, Girard et al. (2003b) suggest that many empiricalstudies provide evidence of a flat cross-sectional relationship, which has been attributed to constantpositive market price of risk limitation imposed by the CAPM. It is only true if we assume thatrisk aversion is intertemporally constant, that is, independent of the state of the economy andtime. We now know this is not true as many authors demonstrate that reward to risk changeswith domestic and world information variables (Beakaert and Harvey, 1995; He et al., 1996; DeSantis and Gerard, 1997). As a result, it is widely believed that multifactor models perform wellcompared to the CAPM and provide an attractive alternative to explain stock risk premiums.

When investing abroad, many different approaches have been proposed for pricing local finan-cial or real assets. For instance, a world CAPM states that a required rate of return depends on howthe investment contributes to the volatility of a well-diversified portfolio. Harvey (1991) showsthat it works for developed markets if (beta) risk is allowed to change through time. However, asthe model entails strong assumptions of perfect market integration, it fails in emerging marketsand is unreliable in smaller, less liquid developed markets. For instance, Erb et al. (1995) reportcountry beta less than one in most emerging markets, which is counter-intuitive as emergingcapital markets are typically more volatile than developed markets. Furthermore, the authors addthat it is not uncommon to observe another counter-intuitive relationship where country betas andreturns are inversely related.

A number of studies have shown that a time varying world beta reflects how investors expect tobe rewarded in connection with a change in risk in the world market (Beakaert and Harvey, 1995).However, less integrated markets are more likely to experience local market inefficiencies dueto barriers to portfolio investments across borders, currency risk, transaction cost differentials,insider trading, law enforcement differentials or infrequent trading—i.e., correlation with theworld portfolio is typically weaker. Beakaert and Harvey (1995) suggest that the CAPM needs tobe modified to account for partial or nascent financial integration. For instance, if a world CAPMholds in integrated markets and a local CAPM holds in segmented markets, this information canbe nested in a conditional beta CAPM. The degree of integration with the world financial marketwill determine what risks explain risk premiums in capital markets and a country asset pricingmodel should use a multifactor framework with local and common risk attributes. An alternativeapproach to price risk around the world has been suggested by Erb et al. (1995) who show thata country risk rating model has the advantage of including sovereign credit risk. The authorsprovide further explanations for the return generating process by exploring risk surrogates suchas political risk, economic risk, financial risk and country credit ratings from the InternationalCountry Risk Guide, the Institutional Investor’s Country Credit Rating, Euromoney’s CountryCredit Rating, Moody’s and Standard and Poor. They find that the International Country RiskGuide (ICRG) composite is highly correlated with Standard and Poor’s sovereign rating (morethan any other rating measures). More generally, they conclude that ratings predict inflation and


are correlated with wealth. They also observe that a lower rating (higher risk) is associated withhigher expected returns.

Erb et al. (1996b) investigate how ICRG composite risk scores (political, financial and eco-nomic risk) explain the cross-sections of expected returns on IFC country indices. They constructtwo portfolios of countries which are rebalanced every 6 months. One includes countries thathave experienced a decrease in risk rating, while the other contains those that have experiencedan increase in risk ratings. They find that economic and financial risks are the most significantfactors determining expected returns in developed markets, while political risk has some marginalexplanatory power in emerging equity markets. They also investigate the relationship betweenworld beta, index volatility, index aggregate book-to-price value (a fundamental attribute at thecountry level) and composite risk scores. Their findings suggest that composite risk scores arehighly correlated with country fundamentals. The authors conclude that their findings explainwhy value-oriented strategies generate high average returns.

At the stock level, empirical research has shown that some fundamental firm-specific factors– size, leverage, book value to market value of equity – are more suited to describe the cross-sections of US and Japanese stock returns (Chan et al., 1991; Aggarwal et al., 1992; Fama andFrench, 1992). For instance, Fama and French (1998), Patel (1998) and Rouwenhorst (1999)report a premium for small and value stocks in foreign developed markets. However, Claessenset al. (1998), Lyn and Zychowicz (2004) and Ramcharran (2004) describe mixed results for therelationship between fundamental attributes and returns in emerging markets. In some cases, theauthors find positive relationships between size and returns as well as a positive relationshipbetween price-to-book value and returns, which is contrary to the conventional belief that smalland value firms are riskier. Nevertheless, all of the studies mentioned above have shown that firms’fundamentals such as size and price-to-book value explain stock returns in capital markets.

In sum, a “multifactor” CAPM approach can consist of adding various risk premiums togetherto calculate the rewards that investors will want to receive in exchange for bearing the risksattached to a security. Prior research suggests that traditional dimensions of risk in the worldfinancial market (such as beta vis-a-vis the world market) can be sufficiently systematic to receivea reward. At the country index and stock levels, individual stock risk premiums can be explainedusing political risk, economic risk, financial risk and fundamentals (such as a stock beta vis-a-visthe local market, a firm’s size and a firm’s market-to-book ratio).

3. Data and methodology

3.1. Data

Data are obtained from the International Finance Corporation (IFC) over the period 1997–2001.The stock markets of Egypt, Jordan, Morocco, Saudi Arabia and Tunisia are investigated.1 Webelieve that those five countries dominate Arab stock markets as they have the biggest and mostactive markets in the region, and consequently provide a representation of Arab firms’ behaviorin relation to stock valuations (Bolbol and Omran, 2005).

1 We limit our sample to firms in these countries because some countries have not yet established stock markets (Iraq,Libya, Syria and Yemen), and other countries have established stock markets only recently (Algeria, Sudan, Qatar andUnited Arab Emirates). For the rest of the Arab countries, though stock markets do exist, data on listed firms could notbe easily obtained. For instance, Kuwait has a large and active stock market and would have been a relevant inclusion inthe study, but Kuwaiti firms are not listed in the IFC index.


Firms listed in IFC indices are targeted rather than those listed in local market indices, for anumber of reasons. First, IFC indices are widely accepted in the international investment industry,forming the basis for index funds and structured financial instruments. Second, IFC indices aremore reliable because they list firms based on market size, trading activity and sector represen-tation, whereas local indices include a large number of firms that are traded infrequently. Localindices may therefore be misleading. Finally, the IFC provides a price index for each firm that isadjusted for dividend payments, stock splits, capital increases and any other event, all making foran accurate calculation of stock returns.

As seen in Table 1 , Panel A, the total number of listed firms in the IFC is 1592. However,out of the original sample (159), complete accounting data was not available for a maximumof 33 firms in 1998 and a minimum of 29 firms in 2001; hence, we have to exclude thosefirms from the analysis. Thus, the maximum sample in a given year (2001) contains 130 firmsand the minimum sample contains 99 firms (1997). We also note that 60% of the firms inour sample are from Egypt and Jordan. Panels B–G show that firms’ sizes are heterogeneousacross countries. Saudi Arabian firms have the highest average market capitalization and bookvalue. Median stock returns tend to be similar to the means and the standard deviations ofreturns is low. Stock returns do not differ in accordance with calculation method, whether CRor BHR, where CR implicitly assumes a more active strategy with monthly portfolio rebal-ancing, while BHR implies a more passive strategy. Last, Jordanian firms exhibit the lowestmarket-to-book ratio (just close to one), which might imply that they are the least profitableand have little potential for growth relative to the entire sample. By contrast, the Saudi Arabianfirms have the highest market-to-book ratio, indicating higher profitability and more potential forgrowth.

Other risk measures are obtained from the ICRG managed by the PRS group. This is thesame country risk provider used by and recommended by Erb et al. (1995, 1996a,b, 1998).These authors examine many providers of country risk data (Bank of America World Informa-tion Services, Business Environment Risk Intelligence, Control Risks Information Services, theEconomist Intelligence Unit, Euromoney, Institutional Investor, S&P Rating Group, the ICRG,Coplin-O’Leary Rating System and Moody Investors Services) and conclude that only the ICRG,composite, political, financial and economic risk scores contain information that explain indexreturns.

ICRG risk scores are grouped into three categories: 12 political, 5 financial and 5 economic.The ICRG ranks risks from a high score, indicating a low risk, to a low score, indicating a highrisk. A summary of each country’s average risk ratings from 1997 to 2001, as well as definitionsfor each risk scores are provided in Table 2. Risk scores for the United States are also reportedfor comparison purposes.

The five markets have typically lower composite, political and economic risk ratings as com-pared to the US, indicating that they are riskier. However, financial risk rating is lower in theUS as compared to Saudi Arabia, Jordan and Egypt. In many instances, the US has a lower riskrating than other Middle Eastern and North African countries. To name a few, debt servicing andexternal conflict risk rating are lower in the US than in the five MENA countries, and governmentstability risk rating is higher in Tunisia than in the US.

Two conclusions can be drawn from these observations: (i) each risk score includes informationthat cannot be aggregated in a composite measure and (ii) particular risk factors have greater

2 No data were available for Saudi Arabia’s firms for the year 1997 as the IFC added the country by the end of 1997.


Table 1Data and descriptive statistics

Number of listed firms Number of useable firms

1997 1998 1999 2000 2001

Panel A: number of firmsEgypt 66 45 45 49 44 47Jordan 41 35 35 36 36 36Morocco 18 8 13 14 14 14Saudi Arabia 21 0 20 18 20 20Tunisia 13 11 13 12 13 13

Total 159 99 126 129 127 130

β CR CCAPMR BHR BHCAPMR MC BV MV/BV

Panel B: all countries (132 firms, 601 observations)Mean 0.54 −0.06 0.06 −0.06 0.07 430 215 1.95Median 0.49 −0.09 0.08 −0.12 0.08 66 49 1.41Standard deviation 0.44 0.32 0.17 0.33 0.17 1388 740 1.91Minimum −0.36 −0.79 −0.47 −0.64 −0.43 1 2 0.19Maximum 2.05 1.09 0.66 1.63 0.88 14864 8772 17.50Standard skewness 3.46 5.73 −0.14 15.80 6.02 64.27 88.71 38.00Standard kurtosis −0.57 4.74 6.80 20.70 11.40 238.60 454.64 110.74

Panel C: Egypt (49 Firms, 220 observations)Mean 0.26 −0.10 0.08 −0.10 0.09 127 60 2.51Median 0.24 −0.16 0.10 −0.19 0.10 57 37 2.01Standard deviation 0.32 0.38 0.12 0.40 0.12 153 71 2.48Minimum −0.34 −0.78 −0.44 −0.64 −0.41 2 3 0.21Maximum 1.31 1.06 0.53 1.63 0.64 750 503 17.50Standard skewness 6.01 5.65 −4.84 11.66 −0.47 11.04 19.36 19.10Standard kurtosis 5.92 2.89 9.44 13.08 11.54 8.86 42.94 42.77

Panel D: Jordan (36 firms, 178 observations)Mean 0.49 −0.10 0.03 −0.11 0.03 103 70 1.01Median 0.47 −0.10 0.06 −0.11 0.06 21 20 0.94Standard deviation 0.37 0.24 0.09 0.24 0.08 354 176 0.53Minimum −0.36 −0.79 −0.34 −0.59 −0.31 1 2 0.19Maximum 1.38 1.09 0.18 1.56 0.17 2656 1289 3.44Standard skewness 1.82 2.92 −9.01 12.53 −8.42 31.35 27.88 6.22Standard kurtosis 0.26 9.59 9.14 39.01 7.86 91.79 76.26 5.81

Panel E: Morocco (14 firms, 63 observations)Mean 1.07 0.03 0.08 0.06 0.10 455 154 2.94Median 0.97 −0.08 0.00 −0.10 −0.01 317 122 2.76Standard deviation 0.37 0.35 0.27 0.37 0.30 469 162 1.56Minimum 0.48 −0.79 −0.40 −0.59 −0.39 19 11 0.81Maximum 2.05 0.71 0.65 0.93 0.77 2007 729 8.56Standard skewness 2.92 −0.08 0.78 2.00 1.30 5.04 6.47 3.34Standard kurtosis 1.33 −0.54 −1.78 −0.46 −1.57 3.78 6.71 2.53

Panel F: Saudi Arabia (20 firms, 78 observations)Mean 0.92 0.03 0.05 0.04 0.07 2243 1118 1.73Median 0.97 0.06 0.08 0.04 0.07 884 482 1.69Standard deviation 0.24 0.28 0.23 0.28 0.23 3251 1787 1.05Minimum 0.33 −0.59 −0.43 −0.47 −0.38 30 51 0.39Maximum 1.30 0.65 0.48 0.80 0.55 14864 8772 4.64


Table 1 (Continued)

β CR CCAPMR BHR BHCAPMR MC BV MV/BV

Standard skewness −2.61 −0.50 −1.03 1.51 0.22 7.50 11.96 3.11Standard kurtosis 0.23 −0.59 −1.00 −0.25 −0.97 7.05 19.93 0.46

Panel G: Tunisia (13 firms, 62 observations)Mean 0.67 −0.02 0.03 −0.02 0.05 136 108 1.91Median 0.67 −0.04 0.02 −0.05 0.02 125 87 1.59Standard deviation 0.39 0.29 0.22 0.28 0.25 87 88 2.10Minimum −0.35 −0.75 −0.47 −0.62 −0.43 23 17 0.34Maximum 1.40 0.55 0.66 0.64 0.88 572 350 14.57Standard skewness −2.24 −0.56 1.42 1.35 3.07 7.49 4.11 15.34Standard kurtosis 2.28 0.42 0.17 0.44 1.85 16.05 1.40 41.09

The table shows descriptive statistics of the sample firms. In Panel A, we provide the number of firms listed in theInternational Finance Corporation (IFC) indices for each country as of the end of 2001, along with the final sample offirms we used in the analysis. In the following panels we provide measures of central tendency, variability and shape ofvariables of the study for the sample firms. We present the mean, median, standard deviation, minimum and maximumvalues of each variable. We also list the standardized skewness and the standardized kurtosis, which can be used todetermine whether these performance measures are normally distributed. β is the market risk, CR the cumulative return,CCAPMR the cumulative return calculated based on the CAPM, BHR the buy-and-hold return, BHCAPMR the buy-and-hold return calculated based on the CAPM, MC the market capitalization, BV the book value and MV/BV is the ratio ofmarket value to the book value.

bearing on business or investments, and a composite risk rating should give greater weight tothese variables.3

3.2. Methodology

The purpose of this paper is to identify risks in emerging markets of the Arab world. First, wetest for the validity of the CAPM by investigating unconditional beta as a good proxy for risk toexplain stock returns. We compare the yearly actual returns on stocks with their returns based onthe CAPM. Significant differences between both returns would imply that a constant beta is not agood proxy for risk. Second, we investigate a multifactor extension to the CAPM by showing howthe cross-sections of 4 fundamental firm-specific risk scores (beta, leverage, size and industrytype) and 22 country-specific risk scores (12 political, 5 economic and 5 financial) can be linearlyrelated to risk premiums in five emerging Arab capital markets. We follow a principal componentanalysis methodology to reduce the factor loading, and identify the significance of each riskfactor’s effects on long-term stock risk premiums (Chen et al., 1986; Groenewold and Fraser,1997). Finally, as in Chen (1983), we test the information content of our multifactor expression ascompared to three nested models—i.e., the CAPM, a three-factor model that includes beta, sizeand the market-to-book value ratio, and a three-factor composite risk model.

Instead of using monthly returns, we employ long-run returns as the CAPM in a long-run model.Due to the lack of consensus on the appropriate way of calculating long-run returns (Barber and

3 Composite political, economic and financial risk scores assume a fixed weighting scheme among their respectiveconstituents. Why should corruption risk weight more than bureaucratic risk in the making of political risk? Why shouldthe risk of government instability weight more than the risk of inflation in the making of a composite risk rating? Thereare no obvious theoretical answers to these questions and it is empirically misleading to allow for a fixed weighting ofrisk scores.


Table 2Average ICRG country risk scores

Risk rating Category Egypt Jordan Morocco Saudi Arabia Tunisia USA

GST Political risk 10.55 10.28 10.49 10.20 10.93 10.34SOE Political risk 5.80 5.00 4.28 6.00 5.97 9.79IPR Political risk 9.07 9.58 8.59 7.95 9.28 10.23ICO Political risk 8.07 9.32 9.89 9.99 11.17 11.01XCO Political risk 9.63 10.32 9.72 10.26 10.98 8.84COR Political risk 1.99 3.43 3.00 2.00 2.98 4.02MIL Political risk 3.00 5.00 4.00 5.00 4.00 5.93RT Political risk 2.01 3.00 4.00 3.00 5.01 5.91LO Political risk 4.00 4.00 6.00 5.00 5.00 6.00ET Political risk 6.00 5.07 5.00 5.00 5.00 4.95DA Political risk 2.12 4.00 2.75 0.00 1.55 5.78FD Financial risk 6.66 4.53 5.00 9.23 5.63 9.32DS Financial risk 8.68 8.00 7.13 9.73 7.93 7.00CAX Financial risk 11.10 12.03 11.77 12.06 11.64 10.03LIQ Financial risk 2.57 3.28 3.20 2.18 1.35 0.64XSTB Financial risk 9.16 9.73 9.14 9.87 8.55 9.12POP Economic risk 0.87 0.84 0.74 2.96 1.46 4.82GDPG Economic risk 8.73 8.43 8.07 6.96 8.87 8.07INF Economic risk 8.74 8.82 9.51 9.93 9.24 9.43BBA Economic risk 7.00 6.88 6.01 5.89 5.85 7.87CAG Economic risk 10.47 11.34 10.66 10.94 10.26 10.55

Political risk Composite risk 64.25 71.11 69.73 66.4 73.87 87.05Financial risk Composite risk 38.48 37.84 36.89 43.54 35.69 37.61Economic risk Composite Risk 36.81 36.97 36.16 37.76 36.67 40.36Composite risk Composite risk 69.77 72.98 71.41 73.88 73.13 82.54

The table shows average country risk ratings from 1997 to 2001 in each of the countries studied and the United States.Government Stability (GST): risk associated with a government’s ability to carry out its declared program(s), and itsability to stay in office. Socioeconomic Conditions (SOE): risk associated with general public satisfaction with thegovernment’s economic policies. Investment Profile (IPR): risk associated with expropriation, taxation, repatriation ofcapital and labor costs. Internal Conflict (ICO): risk associated with political violence and its impact on governance.External Conflict (XCO): risk to both the incumbent government and inward investment. Corruption Risk (COR): riskassociated with corruption within the political system. Military in Politics (MIL): risk associated with military involvementin politics. Religious Tensions (RT): risk associated with the domination of a single religious group or the suppressionof religious freedom. Law and Order (LO): risk associated with the weakness and partiality of a legal system, andthe lack of observance of the law. Ethnic Tensions (ET): risk associated with tensions within a country attributable toracial, nationality, or language divisions. Democratic Accountability (DA): risk associated with a government that is notresponsive to its people. Foreign Debt as a percent of GDP (FD): risk associated with gross foreign debt in a given year,converted into US dollars. Foreign Debt Service as a percentage of Exports of Goods and Services (DS): risk associatedwith foreign debt service per year, in US$. Current Account as a percentage of Exports of Goods and Services (CAX):risk associated with the annual current account deficit, in US$. Net International Liquidity as Months of Import (LIQ):risk associated with the total estimated official reserves for a given year, in US$. Exchange Rate Stability (XSTB): riskassociated with the appreciation/depreciation of a currency against the US$ (against the DM for the US). GDP Per Head(POP): risk associated with a low GDP per head for a given year, converted into US$. Real GDP Growth (GDPG): riskassociated with a percentage increase or decrease in the estimated GDP, at constant 1990 prices. Annual Inflation Rate(INF): risk associated with annual inflation rate (the unweighted average of the Consumer Price Index). Budget Balanceas a percentage of GDP (BBA): risk associated with a government budget deficit for a given year in the national currency.Current Account as a percentage of GDP (CAG): risk associated with the current account balance deficit for a given year,converted into US$.


Lyon, 1997; Kothari and Warner, 1997; Brav and Gompers, 1997; Lyon et al., 1999), we usetwo methods to calculate long-run returns—i.e., cumulative return method (CR) and buy-and-hold return method (BHR). CR implicitly assumes a more active strategy with monthly portfoliorebalancing, which, arguably, might not be a realistic ex-ante trading strategy. On the other hand,BHR means that no survivorship bias or look-ahead bias is involved and implies a more passivestrategy. Given the fact that each method answers different questions and yields different results,it is necessary to consider both methods in calculating long-run returns. This is reinforced by astudy by Barber and Lyon (1997) which found that both methods are biased predictors of long-runabnormal returns and that test statistics designed to detect abnormal returns are consequently oftenmispecified.

The first step in comparing the yearly actual return on a given firm with what it should yieldbased on the CAPM, is to calculate its monthly return and its beta. Monthly return on a given firm(Ri,t) is calculated by taking the logarithmic difference of the price index over 1 month. Then,the CAPM return for a given firm (CAPMRi,t) is calculated using the short-term 1-month rate forbank deposits as a proxy for the risk free rate and the slope obtained from regressing the firm riskpremium on the market risk premium4 as a proxy for the beta of the firm.

We compute actual and CAPM cumulative returns (CRs) and buy-and-hold returns (BHRs)over a 1-year interval (252 trading days) for every firm. We test the null hypothesis that the cross-sectional average abnormal returns over 12 months equal zero for a sample of n firms. If the nullhypothesis is rejected, this means that firms, on average, exhibit negative or positive abnormalreturns, implying that beta is not a good proxy to reflect stock returns.

Under the null hypothesis, the parametric test statistic follows a Student’s t-distribution if thesample is normally distributed. Given the fact that returns are not normally distributed in somecases, an alternative non-parametric test statistic is the Wilcoxon signed-rank test, which tests thenull hypothesis that the median abnormal return is equal to zero.5 Even though the non-parametrictest statistic is less sensitive to the presence of outliers, it is less powerful than the t-test if the datacome from a single normal distribution. We therefore use both test statistics for the robustness ofthe results. However, the findings from the parametric test should be treated with caution if theircorresponding variables are not normally distributed.

Finally, we investigate the linear relationship between k risk factors and long-term returnfollowing a two-step procedure suggested by Chen et al. (1986) and Groenewold and Fraser(1997). First, we consider expected long-term returns previously computed using the CR and BHRmethods. Second, we determine their cross-sectional relationships with k risk factors comprisedof some or all of 26 risk scores (beta6; the logarithm of a firm’s market capitalization; a leverageratio of market value to book value; a dummy variable that takes one if firm i belongs to thenon-financial sector and zero otherwise7; 12 ICRG political risk scores; 5 ICRG economic riskscores; 5 ICRG financial risk scores).

4 We mean by the market index the International Finance Corporation Global (IFCG) index for each country exceptTunisia, in which we use the IFC Frontier (IFCF) index because it is not included in the IFCG index.

5 The results given in Table 1 show that the values of the standardized skewness and standardized kurtosis for somevariables are outside the range of +2 or −2, which means these variables are not normally distributed.

6 Given the fact that beta tends to be stable over a 5-year period, we use the beta value that we calculate for each year.7 It is obvious that risk and return are not the same across industries. For example, we cannot group financial firms,

in particular banks, with manufacturing firms. Each type of industry has its own characteristics. Again, for example,banks do not have leverage while other firms do. Service firms have lower fixed assets and tangible assets compared withmanufacturing, while construction firms have their own characteristics. In sum, it is necessary to control for the industrydifferences.


In order to avoid arbitrary weighting of risk scores by using a composite measure, we utilize aprincipal component analysis (also known as factor analysis) to select the main risk drivers withina risk category. Our country risk factor model (CRFM) should have each asset return linearlyrelated to k factors plus its own idiosyncratic disturbance as follows:

Ri = α +k∑

i=1

λiZi + εCRFM,i (1)

where Ri,t is a vector of long-term stock risk premiums (using CR and BHR) of our sample firmsfor 5 years, Zi a vector of common risk scores factors for each return and λi is a vector of riskpremiums associated with the risk factors.

Most likely, some risk variables are highly correlated with each other, which makes themredundant. We use a Principal Component Analysis to create a grouping or factor that capturesthe essence of these variables.

Finally, we compare the CRFM in Eq. (1) to three nested models proposed in the literature. Thefirst is the CAPM, the second is a three-factors model with firm’s fundamentals (FM, thereafter)and the third is a country risk composite model (CRCM) which relates return to political risk(PR), economic risk (ER) and financial risk (FR) as in Erb et al. (1996a), i.e.,

Ri = λ0 + λ1β + εcapm,i (2)

Ri = λ0 + λ1β + λ2MVBV + λ3Size + εFM,i (3)

Ri = λ0 + λ1 ln(PR) + λ2 ln(ER) + λ3 ln(FR) + εCRCM,i (4)

Following the approach used by Chen (1983), we use three tests to compare the explanatorypower of the CAPM, FM and CRCM as compared to the CRFM—namely, the Davidson andMacKinnon (1981) equation, the posterior odds ratio and a residual analysis.

Davidson and MacKinnon’s (1981) equation estimates the proportion of information (α, theeffectiveness of a model as compared to a competing model) unexplained by a competing model(CAPM, FM and CRCM in our case) which is explained by the CRFM. The test consists inmeasuring α as follows:

Ri = αRCRFM + (1 − α)RCompeting + εi (5)

The next test computes a ratio for posterior odds of the CRFM over a competing model. Theratio is computed as follows:

ρ =[

ESSCompeting

ESSCRFM

]N/2

NKCompeting−KCRFM

2 (6)

where ESS is the error sum of squares, N the number of observations and K is the dimension ofrespective models.

In the last test, the residuals from the competing models are used for performance analysis.If a competing model is not mispecified, its residuals will behave like white noise with zeromean across time. Thus, if expectations in the market are rational and a competing model is notmispecified then

εCompeting,i = ⌊Ei − Ei(Competing)

⌋ + Vi (7)


where Ei is the market rational expected return, Vi the error term (white noise), Ei(Competing)the expected return from the competing model and εCompeting,i represents the residuals fromthe competing model. If the model is correct then Ei = Ei(Competing) and εCompeting,i shouldbehave like white noise and should not be priced by any other model. If εCompeting,i is priced byour CRFM, then the competing model is mispecified. As a result, the test consists of running aregression between εCompeting,i and the factor loadings of the CRFM. We also run a regression ofthe residuals of the CRFM (εCRFM,i) with the factor loadings of the three competing models tocheck for information missed by the CRFM.

4. Results

We start by comparing the yearly actual returns of stocks with their returns based on the CAPM.Should significant differences be found between both returns, this would imply that a constantbeta is not a good proxy for stock returns.

As shown in Table 3, it seems that the unconditional beta is not a good proxy for risk to explainstock returns in all Arab equity markets. The mean (median) abnormal CR and BHR are −0.12(−0.10) and −0.12 (−0.13), respectively. The parametric test statistics are statistically significantat the 1% level for both models (CR and BHR), and the non-parametric Wilcoxon signed-rank testconfirms the same findings with the same significance level. When we examine each stock marketseparately, we find that Egypt and Jordan, which dominate our sample firms, have identical resultsto those of all countries. Both countries show negative mean (median) abnormal returns at −0.18(−0.18) and −0.13 (−0.11), respectively, using the CR method and −0.26 (−0.29) and −0.14(−0.13), respectively, using the BHR method. Results in other countries mirror those obtained forEgypt and Jordan. Qualitatively, the results mostly reproduce the same outcomes, but quantitativedifferences are revealed. Specifically, the non-parametric Wilcoxon signed-rank test failed to passthe critical values of significance at any level, while the parametric t-test is negatively significantat the 10% level, using the CR method only. However, results for Saudi Arabia and Tunisia aremostly negatively significant when we employ the non-parametric test. In sum, strong evidencesupports the belief that beta is not constant and/or is not a good proxy in determining stock returns.

In conducting the principal component analysis to select the risk scores by factors, we followthe methodology proposed by Seiler (2004). The matrix X in our test is a (601, 26) matrix formedby 26 risk scores vectors (each vector has 601 components). The Kaiser–Meyer–Olkin test (KMO)value for the sample is very high (0.971) and the Barlett test of sphericity is significant at the 1%level, indicating that the factor analysis is an appropriate technique for our data.

Table 4 presents the results from the factor analysis. Panel A shows the number of commonfactors found using a VARIMAX rotation. In the first column (components), we list the newlyextracted factors. In the second column (eigenvalues), the eigenvalues represent the proportion oftotal variance in all the variables accounted for by that factor. To decide the number of factors toretain, we use the Kaiser criterion which involves dropping the eigenvalues less than one—i.e.,unless a factor extracts at least as much as the equivalent of one original variable, we drop it. Inthe third column (percentage of variance), these values are expressed as a percentage of the total.As we can see, factor 1 accounts for 24.13% of the variance, factor 2 comprises 21.95% and soon. The fourth column (cumulated percentage) contains the cumulative variance extracted andshows that the six dominant factors whose eigenvalues are more than one, add up to 85.25% ofthe total variance. These factors can be considered as the six major risk factors that characterizethe five Arab countries.

E.G

irard,M.O

mran

/Int.Fin.M

arkets,Inst.andM

oney17

(2007)102–123

113

Table 3Abnormal returns for the sample firms

Cumulative return Buy-and-hold return

Number ofobservationsabove median(below median)

Mean abnormalreturns (median)

T statistic fordifference inmeans

Wilcoxon testfor differencein medians

Number of obs.above median(below median)

Mean abnormalreturns (median)

T statistic fordifference inmeans

Wilcoxon testfor differencein medians

All countries 191 (400) −0.12 (−0.10) −9.66*** −10.17*** 164 (437) −0.12 (−0.13) −9.72*** −11.64***

Egypt 56 (164) −0.18 (−0.26) −7.01*** −6.78*** 50 (170) −0.18 (−0.29) −6.86*** −7.52***

Jordan 57 (121) −0.13 (−0.11) −7.34*** −7.09*** 42 (136) −0.14 (−0.13) −7.71*** −8.04***

Morocco 27 (36) −0.05 (−0.02) −1.71* −1.15 27 (36) −0.04 (−0.02) −1.49 −1.18Saudi Arabia 28 (50) −0.02 (−0.03) −1.13 −1.69* 26 (52) −0.03 (−0.03) −1.22 −2.00**

Tunisia 23 (39) −0.06 (−0.08) −1.59 −2.27** 19 (43) −0.07 (−0.09) −1.92* −2.95***

The table shows the results of the parametric t-test and the non-parametric Wilcoxon signed-rank test for abnormal returns of the sample firms. The abnormal returns are calculatedas the differences between the yearly CR (BHR) of each individual firm and its CCAPMR (BHCAPMR). We provide the number of observations where their abnormal returnsare above (below) the median, the mean and median values of abnormal returns, and the t and z statistics values with their significance level. For the parametric (non-parametric)test, we list the results under the null hypothesis that the mean (median) abnormal return = 0.0 vs. the alternative hypothesis that the mean (median) abnormal return �= 0.

* Refer to 10% significance level.** Refer to 5% significance level.

*** Refer to 1% significance level.


Table 4Factor analysis and component matrix

Component Eigenvalues % of Variance Cumulative %

Panel A: total variance explained with rotation sums of squared loadings1 6.03 24.13 24.132 5.49 21.95 46.083 3.96 15.84 61.914 2.18 8.72 70.635 1.92 7.68 78.316 1.73 6.93 85.25

Social tensionrating (factor 1)

Financialtransparencyrating (factor 2)

Investmentpotential rating(factor 3)

Autarkyrisk rating(factor 4)

Political stabilityrating (factor 5)

Fundamentalattributes risk(factor 6)

Panel B: rotated component matrixBETA 0.49 0.03 0.00 0.08 0.09 0.61BBA −0.53 0.05 0.73 −0.05 0.05 −0.03CAG 0.21 0.04 0.89 −0.07 0.34 0.03CAX 0.32 −0.05 0.29 0.27 0.44 0.08COR 0.58 0.76 −0.11 0.23 −0.04 −0.01DA −0.25 0.92 0.07 −0.13 0.07 −0.10DS −0.26 0.87 0.13 0.03 0.22 0.04ET 0.93 0.23 0.10 −0.18 −0.15 −0.06XCO 0.18 0.17 0.19 −0.81 0.18 0.01FD −0.02 0.96 −0.15 −0.03 0.12 0.05GDPG −0.26 −0.22 0.81 0.13 −0.12 −0.08GST −0.03 −0.17 0.35 0.10 0.78 −0.03IND −0.28 0.05 −0.03 −0.31 0.16 −0.44INF 0.94 0.10 −0.01 −0.20 0.03 0.06ICO 0.68 −0.12 −0.46 0.50 −0.07 0.05LIQ 0.09 −0.51 0.33 −0.62 0.31 0.02IPR −0.18 −0.10 0.87 0.11 −0.22 −0.04LO 0.79 −0.56 −0.12 −0.03 0.18 −0.01MIL 0.92 0.24 −0.07 0.07 0.25 0.10MVBV −0.34 −0.02 −0.11 −0.13 0.09 0.72POP 0.32 0.61 −0.35 0.12 0.32 0.19RT 0.74 −0.46 −0.09 0.43 −0.04 0.05Size 0.16 0.43 0.01 −0.02 0.05 0.75SOE −0.38 0.79 0.06 0.20 −0.30 0.06XSTB 0.28 0.22 0.70 −0.51 −0.21 0.04

The table shows the factor analysis and the component matrix. The extraction method is the Principal Component Analysis.The rotation method is Varimax with Kaiser normalization. Rotation is converged in seven iterations. We select individualrisk scores with a cut-off at 0.7. The selected bold scores are further averaged to determine each factor’s composite score.

Table 4, Panel B shows the loading of each risk score variable within each factor. Interpretationand naming of the factors is not straightforward as it depends on the particular combination ofobserved variables that correlate highly with each factor. In order to minimize the subjective natureof the principal component analysis, we carefully follow the procedure described in Hair et al.(1992), Tabachnick and Fidell (1996) and Seiler (2004). Furthermore, we only consider individualrisk score loadings with “excellent” correlation. Comrey and Lee (1992) define an “excellent”condition for a loading greater than 0.7 (or smaller than −0.7)—i.e., 50% overlapping variance.


Each factor’s composite score is determined by taking into account the risk scores that loadhighly in it. Accordingly, following Seiler (2004), each factor’s score is computed using a sum-mated scale methodology where selected loadings within each factor are averaged to determine afactor score.

Having arrived at a satisfactory number of statistically significant factors, we determine appro-priate names for each factor using substantive interpretation based on the significant higherloadings with similar signs (as suggested by Hair et al., 1992). To interpret a factor, we try tounderstand the underlying dimension that unifies the group of variables loading on it. As shownin Table 4, Panel B, the factors form coherent groups of selected associated variables, which canbe given appropriate names, and describe risk in the Arab block. Each of the six constructs isbriefly reviewed and named below.

We name the first factor social tension rating because the contributing variables emphasizeissues associated with political repression (military in politics, and law and order) as well as eco-nomic and social hardship (inflation, religious tensions and ethnic tensions and internal conflicts).This factor accounts for 24.13% of the variance. The factor loadings are positive, thus it followsthe ICRG scale—i.e., a high value indicates a low risk and a low value indicates a high risk.

The second factor consists of five significant variables: indebtment ratings, debt servicingratings, corruption, democratic accountability and socio-economic conditions. Ultimately, thisrisk factor is named financial transparency rating, as all five variables relate to financing andcorruption. This factor accounts for 21.95% of the variance. The factor loadings are also positive,so a high value indicates a low risk and a low value indicates a high risk on the ICRG rating scale.

The third factor grouping consists of five variables related to the investment environment inthe Arab block. Four of the variables focus strongly on economic issues related to economicgrowth, trade deficit and currency depreciation. The other variable, the investment profile rating,is obviously related to the other four variables. In sum, we name this factor investment potentialrating. This factor accounts for 15.84% of the variance. Factor loadings are positive and a high(low) value indicates a low (high) risk as on the ICRG scale.

The fourth factor relates to economic and financial segmentation and is summed up as anautarky risk rating. This risk factor takes into account issues of trade barriers, quotas and foreigninvestment restrictions. It accounts for 8.72% of the variance. The factor loading is negative anda high value indicates a high risk and a low value indicates a low risk.

The fifth factor is a political stability rating as it addresses the stability of the current regimeand the longevity of the laws passed or initiated. This factor accounts for 7.68% of the variance.It has a positive factor loading and a high (low) value indicates a low (high) risk.

The last factor is a firm’s fundamentals rating. It includes market-to-book value ratio and size,both of which have a positive factor loading. This factor accounts for 6.93% of the variance. Asin Fama and French (1992), a high (low) score should relate to a low (high) risk.

The next step is to identify which of these factors can explain long-term returns. We use astepwise regression to identify the significant factors that explain long-term returns. Due to theinversed scale of ICRG risk scores, we hypothesize a negative relationship between long-termreturns and factors 1–3 and 5. A positive relationship should be seen between factors 4 and long-term returns because of a negative factor loading. Finally, according to Fama and French (1992),factor 6 should have a negative relationship with stock returns, as small and value firms are riskier.

The stepwise regressions’ findings are shown in Table 5 . We report each regression by addinga risk factor based on its explanatory power (R-squared with the dependent variable). Coeffi-cients, t-statistics, standardized coefficients and the variance inflation factors are reported foreach regression.

116E

.Girard,M

.Om

ran/Int.F

in.Markets,Inst.and

Money

17(2007)

102–123

Table 5Stepwise cross-sectional regressions between long-term return and risk factors

Dependent variable

CR BHR

Constanta Factor 6a Factor 5a Factor 2a Factor 4a Factor 3a Factor 1a R2 Constanta Factor 6a Factor 5a Factor 2a Factor 4a Factor 3a Factor 1a R2

Model 1Coefficient −0.300*** 0.080*** 0.131 −0.329*** 0.086*** 0.156t-stat −10.84 9.558 −11.495 10.534SCoef 0.364 0.395VIF 1.000 1.000

Model 2Coefficient 1.110*** 0.071*** −0.133*** 0.171 1.360*** 0.080*** −0.159*** 0.212t-stat 4.266 9.044 −5.464 5.171 10.015 −6.456SCoef 0.339 −0.205 0.366 −0.236VIF 1.015 1.015 1.015 1.015

Model 3Coefficient 1.882*** 0.078*** −0.154*** −0.118*** 0.186 2.151*** 0.087*** −0.181*** −0.121*** 0.223t-stat 5.537 9.712 −6.200 −3.484 6.272 10.679 −7.189 −3.546SCoef 0.373 −0.237 −0.138 0.4 −0.269 −0.137VIF 1.084 1.080 1.150 1.084 1.080 1.150

Model 3Coefficient 1.575*** 0.0790*** −0.158*** −0.132*** 0.0418** 0.192 1.846*** 0.088*** −0.185*** −0.135*** 0.0427** 0.234t-stat 4.345 9.837 −6.373 −3.850 2.370 5.044 10.805 −7.359 −3.905 2.339SCoef 0.376 −0.244 −0.154 0.088 0.404 −0.275 −0.152 0.085VIF 1.086 1.085 1.185 1.031 1.086 1.085 1.185 1.031

Model 4Coefficient 1.587*** 0.079*** −0.150*** −0.129*** 0.043** −0.014 0.199 1.871*** 0.0870*** −0.168*** −0.129*** 0.043** −0.028* 0.239t-stat 4.375 9.793 −5.673 −3.754 2.409 −0.889 5.12 10.746 −6.296 −3.734 2.426 −1.807SCoef 0.375 −0.231 −0.151 0.090 −0.035 0.401 −0.25 −0.146 0.088 −0.069VIF 1.088 1.232 1.194 1.033 1.144 1.088 1.232 1.194 1.033 1.144

E.G

irard,M.O

mran

/Int.Fin.M

arkets,Inst.andM

oney17

(2007)102–123

117

Table 5 (Continued)

Dependent variable

CR BHR

Constanta Factor 6a Factor 5a Factor 2a Factor 4a Factor 3a Factor 1a R2 Constanta Factor 6a Factor 5a Factor 2a Factor 4a Factor 3a Factor 1a R2

Model 5Coefficient 1.605*** 0.079*** −0.150*** −0.131*** 0.043** −0.014 −0.002 0.199 1.922*** 0.087*** −0.169*** −0.134*** 0.045** −0.029* −0.005 0.239t-stat 3.414 9.782 −5.632 −3.088 2.120 −0.833 −0.059 4.059 10.737 −6.264 −3.136 2.190 −1.714 −0.170SCoef 0.375 −0.231 −0.152 0.091 −0.036 −0.003 0.401 −0.251 −0.151 0.092 −0.072 −0.009VIF 1.089 1.251 1.802 1.369 1.381 1.991 1.089 1.251 1.802 1.369 1.381 1.991

Ri = α +k∑

i=1

λiZi + εCRFM,i.

All five regressions (models 1–5) are estimated using white heteroskedasticity coefficient covariance. SCoef is the standardized coefficient and VIF is the Variance InflationFactor. Factor 6 is the “firm’s fundamentals rating.” Factor 5 is the “political stability rating.” Factor 2 is the “financial transparency rating.” Factor 4 is the “autarky risk rating.”Factor 3 is the “investment potential rating.” Factor 1 is the “social tension rating.”

a Independent variables.* Refer to 10% significance level.

** Refer to 5% significance level.*** Refer to 1% significance level.


Our cross-sectional results suggest that only four risk factors explain stock cumulative returns(CR), including firms’ fundamentals, political stability, financial transparency and autarky risk.However, five risk factors provide explanations for buy-and-hold returns (BHR)—namely, firms’fundamentals, political stability, financial transparency, autarky risk and investment potential. Asa result, active and passive portfolio strategies are mostly affected by the same risks; however,investment potential is only priced by the passive investor and not by the active investor.

The variance inflation factors for each independent variable are extremely low in the two models(less than 1.24, that is, more than 80% of the variance of each independent variable is not shared byother independent variables) indicating that the CRFM is not likely affected by multicollinearity.Standardized coefficients indicate that firms’ fundamentals are the most important drivers of riskpremiums. It is also worth noting that the combined effect of country risk scores has a greaterimpact on risk premiums than firms’ fundamentals alone (in each model the sum of the absolutevalues of the standardized coefficients associated with country risk factors is greater than thestandardized coefficient for fundamentals risk). Finally, the R-squared for each equation indicatesthat about 20% of the variations in the risk factors explain the variation in long-term risk premiums(R-squared is 19.9% for CR and 23.9% for BHR).

In general, the signs associated with the country risk factors are consistent with the theory.Indeed, a negative relationship exists between returns and political stability, financial transparencyand investment profile—these three risk factors follow the ICRG scale where a high value indicatesa low risk. The positive relationship between autarky risk and stock returns is also logical—it hasa negative loading indicating that a high value is consistent with high risk. Finally, the positiverelationship between fundamental risk and return is opposite to our expectations. Fundamentalrisk has a positive loading in size and market-to-book value; Fama and French (1992) in the USand Chan et al. (1991) and Aggarwal et al. (1992) in Japan suggest that this relationship shouldbe negative. However, Harvey and Roper (1999) report small positive relationships between sizeand returns in Asian emerging markets. Claessens et al. (1998), Ramcharran (2004) and Lyn andZychowicz (2004) report a positive relationship between returns, and size and market-to-bookvalue in some emerging markets. Several arguments have been put forth to explain these findings.Harvey and Roper (1999) argue that market growth has led to the mobilization of new capital andan increase in the number of firms rather than an increase in value. Indeed, the total number of firmsin our study increases from 99 in 1997 to 130 in 2001. Furthermore, due to either the restrictionson debt financing in Muslim countries, or the immature debt markets, small firms have a capitalstructure made up principally of equity, while larger firms with their international exposure canmore easily gain access to leverage. For instance, Bolbol and Omran (2005) indicate that onlylarge firms have higher leverage ratios in Arab markets. Claessens et al. (1998) also suggest thatthe market microstructure causes these substantial differences and that regulatory and tax regimesforce investors to behave differently in nascent markets. The authors also hypothesize that thepositive relationships between returns and size and market-to-book value can be attributed to thesegmentation of financial markets.

The next and final step of our study is to examine how the CRFM compares to three nestedmodels. In Table 6 , we report the cross-sections of the CAPM, a three-factor CAPM, a compositerisk model and the CRFM. We also include the results from the Davidson and Mackinnon posteriorodds ratio and residual tests.

As seen from Table 6, the cross-sections of beta indicate that local beta is priced in these marketswith an explanatory power of about 2%. Also, when fundamentals are included in the cross-sectional regression, beta becomes insignificant and the explanatory power of the model increasesto 13% for CR and to 15% for BHR. Interestingly, the composite country risk cross-sections

E.G

irard,M.O

mran

/Int.Fin.M

arkets,Inst.andM

oney17

(2007)102–123

119Table 6Multivariate cross-sectional regression analysis of the determinants of stock returns

CR CAPM FM CRCM CRFM

Panel A: when CR is the dependent variableConstant −0.121*** (−5.866) −0.311*** (−9.525) −1.23** (−2.546) 1.575*** (4.345)Beta 0.107*** (3.623) 0.019 (0.61)Size 0.038*** (4.582)MV/BV 0.036*** (5.189)Political risk rating 0.0078* (1.725)Economic risk rating −0.0064 (−0.816)Financial risk rating 0.0023*** (3.475)Factor 6: firms’ fundamentals 0.079*** (9.837)Factor 5: political stability rating −0.158*** (−6.373)Factor 2: financial transparency rating −0.132*** (−3.85)Factor 4: autarky risk rating 0.042** (2.37)N 601 601 601 601R-squared 0.021 0.133 0.020 0.198F-stat 13.128 30.536 4.151 36.696

Davidson and Mackinnon test—α 1.033*** 1.013*** 0.975***

Posterior odds ratio 5.40E 5.15E+08 4.44E+24RT: competing models’ residuals explained by CRFM factors—R2 0.152 0.072*** 0.177***

CRFM factors explaining competing models’ residuals 6***, 5***, 2** 5***, 2***, 4** 6***, 5***, 2**

RT: CRFM residuals explained by competing models’ factors—R2 0.001 0.002 0.005Competing models’ factors explaining CRFM residuals None None None

Panel B: when BHR is the dependent variableConstant −0.119*** (−5.572) −0.33*** (−9.929) −1.303** (−2.611) 1.871*** (5.12)Beta −0.112*** (3.682) 0.015 (0.472)Size 0.042*** (4.883)MV/BV 0.043*** (5.978)Political risk rating 0.0071 (1.527)Economic risk rating −0.0043 (−0.524)Financial risk rating 0.024*** (3.535)Factor 6: firms’ fundamentals 0.087*** (10.746)Factor 5: political stability rating −0.168*** (−6.296)Factor 2: financial transparency rating −0.129*** (−3.734)

120E

.Girard,M

.Om

ran/Int.F

in.Markets,Inst.and

Money

17(2007)

102–123

Factor 4: autarky risk rating 0.043** (2.426)Factor 3: investment potential rating −0.028* (−1.807)N 601 601 601 601R-squared 0.022 0.157 0.022 0.239F-stat 13.555 36.953 4.510 37.32

Davidson and Mackinnon test—α 1.039*** 1.012*** 0.985***

Posterior odds ratio 1.32E+27 3.90E+10 7.86E+29RT: competing models’ residuals explained by CRFM factors—R2 0.189*** 0.095*** 0.215***

CRFM factors explaining competing models’ residuals 6***, 5***, 2**, 3* 5***, 2**, 4**, 3* 6***, 5**, 2**, 3*

RT: CRFM residuals explained by competing models’ factors—R2 0.003 0.004 0.001Competing models’ factors explaining CRFM residuals None None None

The table shows the results from multivariate cross-sectional regression analyses of the determinants of yearly stock returns. The following models are estimated: CAPM,Fundamentals Model, Country Risk Composite Model and CRFM. We also include the results from the Davidson and Mackinnon tests, Posterior Odds Ratios and R-squaredfrom Residual Tests (RT). Regressions are estimated using white heteroskedasticity coefficient covariance.CAPM: Ri = λ0 + λ1β + εcapm,i.FM: Ri = λ0 + λ1β + λ2MVBV + λ3Size + εFM,i.CRCM: Ri = λ0 + λ1 ln(PR) + λ2 ln(ER) + λ3 ln(FR) + εCRCM,i.

CRFM: Ri = α +k∑

i=1

λiZi + εCRFM,i

Figures between parentheses are t statistics.* Refer to 10% significance level.

** Refer to 5% significance level.*** Refer to 1% significance level.


indicate that only composite financial risk is priced by CR and BHR, while political composite isonly priced at the 90th percentile by CR. The explanatory power of the model is approximately2%. It is worth noting that the positive signs are not logical. This anomaly may possibly beexplained by the high variance inflation factor for each independent variable (approximately 2.5,not reported but available upon request). Also, political, economic and financial risk scores assumea fixed weighting scheme among their respective constituents and there is no obvious theoreticalrational for this. Furthermore, several risk scores constituting each composite rating are negativelycorrelated, indicating that, at the composite level, the effect of some risk scores will offset otherrisk scores.

All tests of comparison between the models unambiguously demonstrate that CRFM is the bestmodel. For instance, alphas in DM tests are close to the unity (the CRFM is 100% more effectivethan the three nested models) and are significant at the 99th percentile. Posterior Odd Ratiosare highly in favor of the CRFM. The residual tests indicate that the CRFM’s factors providesignificant additional information to the CAPM (factors 6, 5 and 2 for CR and factors 6, 5, 2 and3 for BHR), FM (factors 5, 2 and 4 for CR and factors 5, 2, 4 and 3 for BHR) and CRCM (factors6, 5 and 2 for CR and factors 6, 5, 2 and 3 for BHR). The CAPM, FM and CRCM do not provideadditional information to the residuals of the CRFM.

5. Conclusion

Our study attempts to identify the risks involved when investing in five emerging Arab capitalmarkets. We first find that a constant beta is not a good proxy for risk in these thinly tradedemerging markets, so we turn to a multifactor representation of the return generating processand find that firms’ fundamentals and country risk rating factors are significant in explaining thecross-sections of stock returns. Furthermore, we show that a pricing model including both firm’sfundamentals and country risk rating factors has a significantly better explanatory power than theCAPM, a model only including a firms’ fundamentals, or a model based on country compositerisk ratings.

Our paper provides three important contributions to the literature on asset pricing in emergingcapital markets. Firstly, we show that country risk rating cannot be arbitrarily aggregated into acomposite risk rating as individual risk rating can be negatively related and, therefore can offseteach other’s effects; however, country risk ratings can be combined into a country risk factor usinga factor analysis. Secondly, we add to a growing literature base suggesting that, in markets otherthan the US, it is possible to find large and growth stocks to be riskier than small and value stocks.Thirdly, although many Arab countries have embarked on a process of privatization and stockmarket liberalization to deepen their markets and improve corporate governance, issues related tofinancial transparency and political instability are still powerful obstacles to investments in thesenascent emerging markets.

Risk factors, in particular political risk, are likely to remain significant in Arab stock markets.Finance literature shows that changes in political risk in general tend to have a strong effect on localstock market development and excess returns in emerging economies, suggesting that political riskis a priced factor (Oijen and Perotti, 2001). In this context, the Arab economies are no exception.Here, and to nobody’s surprise, the Arab world does not fare well, having a relatively closed andhighly concentrated political system with a poor mode of national governance. Consequently, weexpect to find that any changes in political risk in these countries will be strongly associated withgrowth in stock market development indicators and that the economic impact appears to be verylarge. Our analysis of the influence of political risk on stock market development is also related


to recent research on the link between the legal institutional framework and corporate finance inwhich La Porta et al. (1997, 1998) find that countries with lower quality of legal rules and lawenforcement have smaller and narrower capital markets. Demirguc-Kunt and Maksimovic (1998)also show that firms in countries with highly rated effectiveness of their legal systems are able togrow faster by relying more on external finance.

It seems then that political risk has strong implications for stock market development. Given thegrowing literature suggesting that the development of financial markets supports economic growth(see Levine and Zevros, 1998), the problem of political risk has an important policy implication forgrowth in these thinly traded Arab markets. A great need exists to improve political risk in theseArab countries in order to attract more investment and better allocation of resources through stockmarkets. To achieve this, more institutional (stock market) reforms are needed. These could tacklethe issues of improving the institutional and legal frameworks—accountability, transparency anddisclosure, corruption, rule of law and contract enforceability, among others.

References

Aggarwal, R., Ramesh, P.R., Hiraki, T., 1992. Price/book value ratios and equity returns on the Tokyo stock exchange:empirical evidence of regularities. Financial Review 27 (4), 589–605.

Barber, B., Lyon, J., 1997. Detecting long-run abnormal stock returns: the empirical power and specification of teststatistics. Journal of Financial Economics 43 (3), 341–372.

Beakaert, G., Harvey, C., 1995. Time-varying world market integration. Journal of Finance 50 (2), 403–444.Beakaert, G., Harvey, C., 2002. Research in emerging markets finance: looking to the future. Emerging Markets Review

3 (4), 429–448.Beakaert, G., Harvey, C., 2003. Emerging markets finance. Journal of Empirical Finance 10 (1), 3–56.Bolbol, A., Omran, M., 2005. Investment and the stock market: evidence from Arab firm-level panel data. Emerging

Market Review 6 (1), 85–106.Brav, A., Gompers, P., 1997. Myth or reality? The long-run underperformance of initial public offerings: evidence from

venture capital and non-venture capital-backed companies. Journal of Finance 52 (5), 1791–1822.Chan, L.K.C., Hamao, Y., Lakonishok, J., 1991. Fundamentals and stock returns in Japan. Journal of Finance 46 (5),

1739–1764.Chen, N.F., 1983. Some empirical tests of arbitrage pricing. Journal of Finance 38 (5), 1393–1414.Chen, N.F., Roll, R., Ross, S.A., 1986. Economic forces and the stock market. Journal of Business 59 (3), 383–403.Claessens, S., Dasgupta, S., Glen, J., 1998. The cross-section of stock returns: evidence from emerging markets. Emerging

Markets Quarterly 2, 4–13.Comrey, A.L., Lee, H.B., 1992. A First Course in Factor Analysis, second ed. Hillsdale, Erlbaum, NJ.Davidson, R., MacKinnon, J., 1981. Several tests for model specification in the presence of alternative hypotheses.

Econometrica 49 (3), 781–793.Demirguc-Kunt, A., Maksimovic, V., 1998. Law, finance and firm growth. Journal of Finance 53 (6), 2107–2137.De Santis, G., Gerard, B., 1997. International asset pricing and portfolio diversification with time-varying risk. Journal of

Finance 52 (5), 1881–1912.Diamonte, R., Liew, J.M., Stevens, R.L., 1996. Political risk in emerging and developed markets. Financial Analysts

Journal 52 (3), 71–76.Erb, C., Harvey, C., Viskanta, T., 1995. Country credit risk and global portfolio selection. Journal of Portfolio Management

9 (Winter), 74–83.Erb, C., Harvey, C., Viskanta, T., 1996a. Expected returns and volatility in 135 countries. Journal of Portfolio Management

22 (3), 46–58.Erb, C., Harvey, C., Viskanta, T., 1996b. Political risk, financial risk and economic risk. Financial Analysts Journal 52

(6), 28–46.Erb, C., Harvey, C., Viskanta, T., 1997. Demographics and international investment. Financial Analysts Journal 53 (4),

14–28.Erb, C., Harvey, C., Viskanta, T., 1998. Risk in emerging markets. The Financial Survey (July–August), 42–46.Fama, E.F., French, K.R., 1992. The cross-section of expected stock returns. Journal of Finance 47 (2), 427–465.Fama, E.F., French, K.R., 1998. Value versus growth: the international evidence. Journal of Finance 53 (6), 1975–1999.


Girard, E., Omran, M., Zaher, T., 2003a. On risk and return in MENA capital markets. International Journal of Business8 (3), 285–314.

Girard, E., Rahman, H., Zaher, T., 2003b. On market price of risk in Asian capital markets around the Asian flu. InternationalReview of Financial Analysis 12 (3), 241–265.

Groenewold, N., Fraser, P., 1997. Share prices and macroeconomic factors. Journal of Business Finance and Accounting24 (9), 1367–1383.

Hair, J., Anderson, R., Tatham, R., Black, W., 1992. Multivariate Data Analysis, third ed. Macmillan, New York, NY.Harvey, C., 1991. The world price of covariance risk. Journal of Finance 46 (1), 111–157.Harvey, C., Roper, A., 1999. The Asian bet. In: Harwood, Alison, Litan, Robert E., Pomerleano, Michael (Eds.), The

Crisis in Emerging Financial Markets. Brookings Institution Press, pp. 29–115.Harvey, C., Solnik, B., Zhou, G., 2002. What determines expected international asset returns? Annals of Economics and

Finance 3 (2), 249–298.He, J., Kan, R., Ng, L., Zhang, C., 1996. Tests of relations among marketwide factors, firm-specific variables, and stock

returns using a conditional asset pricing model. Journal of Finance 51 (5), 1891–1908.Kothari, S., Warner, J., 1997. Measuring long-run horizon security price performance. Journal of Financial Economics 43

(3), 301–340.La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1997. Legal determinants of external finance. Journal of

Finance 52 (3), 1131–1150.La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1998. Law and Finance. Journal of Political Economy 106

(6), 1113–1155.Levine, R., Zevros, S., 1998. Stock markets and economic growth. American Economic Review 88 (4), 537–558.Lyn, E., Zychowicz, E., 2004. Predicting stock returns in the developing markets of eastern Europe. The Journal of

Investing 13 (2), 63–72.Lyon, J., Barber, B., Tsai, C., 1999. Improved methods for tests of long-run abnormal stock returns. Journal of Finance

54 (1), 165–201.Oijen, P., Perotti, E., 2001. Privatization, market development and political risk in emerging economies. Journal of

International Money and Finance 20 (1), 43–69.Omran, M., 2005. Underpricing and long-run performance of share issue privatizations in the Egyptian stock market.

Journal of Financial Research 28 (2), 215–234.Patel, S., 1997. Cross-sectional variation, in emerging markets equity returns, January 1988–March 1997. Emerging

Markets Quarterly 2, 57–70.Ramcharran, H., 2004. Returns and pricing in emerging markets. The Journal of Investing 13 (1), 45–55.Rouwenhorst, G., 1999. Local return factors and turnover in emerging stock markets. Journal of Finance 54 (4), 1439–1464.Seiler, M.J., 2004. Performing financial studies. In: A Methodological Cookbook. Pearson Prentice Hall, Upper Saddle

River, NJ.Tabachnick, B., Fidell, L., 1996. Using Multivariate Statistics, third ed. Harper Collins, New York, NY.

Documents

What are the risks when investing in thin emerging equity markets: Evidence from the Arab world