Financial integration and consumption risk sharing and smoothing

International Review of Economics and Finance 29 (2014) 585–598

Contents lists available at ScienceDirect

International Review of Economics and Finance

j ourna l homepage: www.e lsev ie r .com/ locate / i re f

Financial integration and consumption risk sharingand smoothing

Yui Suzuki⁎School of Diplomacy and International Relations, Seton Hall University, USA

a r t i c l e i n f o

⁎ School of Diplomacy and International Relations,fax: +1 973 275 2519.

E-mail address: [email protected] Estimates of the gains to international risk sharin

based models (Lewis, 2000).

1059-0560/$ – see front matter © 2013 Elsevier Inc. Ahttp://dx.doi.org/10.1016/j.iref.2013.08.005

a b s t r a c t

Article history:Received 25 April 2012Received in revised form 13 August 2013Accepted 14 August 2013Available online 27 August 2013

While paying careful attention to the stochastic properties of income process, this paper teststhe joint rational expectation and permanent income hypothesis (RE/PIH) to clarify how and towhat degree financial integration delinks national income and consumption. It is shown thatboth the OECD and the non-OECD countries benefit from financial integration in terms ofconsumption risk sharing and smoothing. The RE/PIH for the transitory income is not rejectedfor the OECD countries suggesting full consumption smoothing. Regression results also supportthe RE/PIH prediction that financial integration delivers even larger increases in consumptionresponding to positive shocks to income growth.

© 2013 Elsevier Inc. All rights reserved.

JEL classification:F41F36E21

Keywords:Consumption risk sharingConsumption smoothingFinancial integrationPermanent incomeTransitory income

1. Introduction

Theoretical models in open economy macroeconomics and international finance render clear predictions about the benefit offinancial integration in terms of consumption risk sharing and smoothing. In the typical dynamic stochastic general equilibrium(DSGE) models, trade in financial assets delinks national consumption levels from the country-specific component of outputfluctuations, which reduces the volatility of consumption relative to output and income (e.g. Backus, Kehoe, & Kydland, 1992;Mendoza, 1991). Although somewhat lower degrees of risk sharing and smoothing can be generated by introducing transactioncosts in the international financial markets and/or non-tradable and durable goods and services, they do not reverse thetheoretical prediction of the lower relative volatility of consumption. Although there are debates about larger or smaller potentialwelfare gain to international risk sharing,1 the theoretical prediction of the reduced relative consumption volatility throughinternational financial integration has been quite robust in numerical simulations.

Nevertheless, empirical facts reported in Kose, Prasad, and Terrones (2003) cast some doubt on this benefit of financialintegration in terms of consumption risk sharing and smoothing. They point out that the median of the ratio of the volatility of

Seton Hall University, 400 South Orange Avenue, South Orange, NJ 07079, USA. Tel.: +1 973 275 2215;

g based upon stock returns tend to find dramatically higher gains than do estimates from consumption-

ll rights reserved.

http://dx.doi.org/10.1016/j.iref.2013.08.005

mailto:[email protected]

http://dx.doi.org/10.1016/j.iref.2013.08.005

http://www.sciencedirect.com/science/journal/10590560

http://crossmark.crossref.org/dialog/?doi=10.1016/j.iref.2013.08.005&domain=pdf

586 Y. Suzuki / International Review of Economics and Finance 29 (2014) 585–598

total consumption to that of income has risen from the 1980s to the 1990s in more financially integrated economies (MFIEs),precisely in the period when financial integration should have paid off in offering better opportunities to alleviate country-specificshocks. Furthermore, they document that, although gross capital flows surged after the mid-1980s, industrial countriesexperienced virtually no major change in their average relative volatility over time since the 1970s. These findings run counter tothe notion that financial integration yields opportunities of consumption risk sharing and smoothing and beg the question againstthe capacity of international financial markets in providing better opportunities to attenuate country-specific shocks. Note that,since the relative volatility is used as a measure of the efficacy of consumption smoothing, these puzzling patterns are hardlyexplained away as just a consequence of boom and bust cycles possibly fomented by financial integration.2 In fact, Kose et al.(2003) conclude that the benefit of international financial integration in terms of improved consumption risk sharing andsmoothing accrues only beyond a certain threshold.

I hypothesize that the observed lack of reduction in the relative consumption volatility under international financial integrationis reconciled after careful deliberation on the types of income shock and process. The documented patterns of the relative volatilityof consumption to income are puzzling at a glance because they contradict the typical prediction of consumption risk sharing andsmoothing for transitory and random walk shocks. Consumption smoothing pertains to the rational expectation and permanentincome hypothesis (RE/PIH) in which consumption is smoothed intertemporally ex post by international lending and borrowing,bringing down the response of consumption to income shocks too. Moreover, volatility of consumption will be reduced whencountry-specific shocks are hedged ex ante by exchanging contingent assets, tapering off the response of consumption to incomeshocks. Nevertheless, while the RE/PIH points to consumption smoothing for a standard mean-reverting income process, it doespredict the opposite—larger adjustments in consumption than contemporaneous changes in income—and thus points to highervolatility of consumption, when income growths are positively autocorrelated. The RE/PIH prediction on the volatility ofconsumption indeed hinges on the underlying types of income shock and process.

With this conjecture in mind, this article empirically explores how international financial integration provides consumptionrisk sharing and smoothing opportunities and reconciles observed lack of decreased volatility of consumption under financialintegration. The goals are to investigate whether and to what extent countries are subject to credit constraints and to clarifyhow and to what degree international financial integration delinks national income and consumption after accounting forthe stochastic properties of income process. Specifically, I conduct a joint test of the RE/PIH, which is developed by Flavin (1981)and then extended by Elwood (1998) who argues that the specification of the income process is crucial. By explicitly consideringmultiple components of income which follow different processes, I will disentangle respective adjustment patterns ofconsumption responding to various types of income shocks and changes. To be more specific, the empirical results show whetherand to what extent consumption responds to three types of income shocks: the slope shock (the shock to income growth), thelevel shock (the random walk shock to the income level), and the transitory shock, as well as the predictable changes in thepermanent and transitory components of income. Furthermore, I probe how such responses are altered when countries becomemore integrated to the international financial markets.

Contrary to what we observe in the full sample countries case, the RE/PIH for the transitory income is not rejected for the OECDand the EU countries, suggesting full consumption smoothing and absence of excess sensitivity. Furthermore, I show the evidencethat not only the OECD but also the non-OECD countries benefit from international financial integration in terms of coping withthe income shocks. In particular, regression results support the RE/PIH prediction that financial integration delivers even largeradjustments in consumption in response to positive shocks to income growth, which can reconcile the puzzle of higher relativevolatility of consumption pointed out by Kose et al. (2003). Moreover, smaller adjustments in consumption responding to permanentshocks to the income level in more integrated countries suggest that countries are insured for idiosyncratic country-specific levelshocks through cross-holding of equities and securities.

This paper relates to a vast number of preceding empirical literature discussing the lack of international consumption risk sharingand smoothing.3 It is typically concluded that the fruits of better consumption risk sharing are, at best, limited to the developedcountries and little evidence of the gains in the developing countries has been found. Lewis (1996) argues that not only capitalmarketrestrictions but also the nonseparability between tradables and nontradables are required to explain lack of consumption risk sharing.More recently, Artis and Hoffmann (2008) illustrate that the OECD countries have becomemore insured against permanent shocks;however, basic measures of risk sharing have not picked up this change because of the great moderation and synchronization of thebusiness cycles. Bekaert, Harvey, and Lundblad (2006) show that equity market liberalization and capital account openness areassociatedwith declines in consumption growth volatility suggesting improved risk sharing; however, such results areweaker for theliberalizing emergingmarkets. Kose, Prasad, and Terrones (2007) argue that the developed countries have come to deal with incomeshocks better during the recent period of globalization,while the developing countries have been shut out of this benefit. On the otherhand, Sørensen and Yosha (1998), applying risk sharing regression proposed by Asdrubali, Sørensen, and Yosha (1996), quantify thatonly 40% of shocks to GDP are insured in the European and the OECD countries. Bai and Zhang (2012) argue that there have been nosubstantial improvements in the degree of international risk sharing in the last two decades, even in the developed countries. Thesefindings are perplexing when we recall recent rapid enhancement in the cross border transactions of financial assets and thepotentially large welfare gains through financial integration. In regard to such discrepancy, this study contributes to the debates by

2 And in fact, as it is presented in Ćorić and Pugh (2013), the decades preceding the recent financial crisis and global downturn were a period of unusually mildoutput volatility for many developed and developing market economies.

3 Backus et al. (1992) point out that consumption correlations observed in the data are too low to be explained by a standard DSGE model with completemarkets (the quantity anomaly).

587Y. Suzuki / International Review of Economics and Finance 29 (2014) 585–598

demonstrating that not only the OECD but also the non-OECD countries gain from international financial integration. I owe thefindings, in particular, to two points in my approach, which are unique in the literature. Unlike in Kose et al. (2007), I explicitlyconsider simultaneous shocks to the trend and the cycle which potentially lead to offsetting multiple innovations in consumption.I also incorporate a de facto measure of financial integration in the regression as opposed to Bekaert et al. (2006) who study de juremeasures.

The rest of this paper is organized as follows: Section 2 discusses the RE/PIH predictions for consumption path under variousincome processes. Section 3 reviews ways to measure the degree of international financial integration. Section 4 presents theempirical framework and methodology. Section 5 shows and discusses main regression results. Section 6 exhibits regressionresults without interaction terms to support the discussion in the previous section. Section 7 discusses potential asymmetrybetween positive and negative income shocks/changes. Finally, Section 8 draws conclusions.

2. Predictions for consumption path under risk sharing

This section reviews the RE/PIH prediction for the consumer's responses to income changes and affirms that they indeed hingeon the underlying income processes considered. These predictions are relevant to the main subject of this paper, namely toexplore how international financial integration shifts the responses of consumption to income shocks and changes and enhancesconsumption risk sharing.

I consider a standard macroeconomics model where consumption is governed by the permanent income, i.e. the consumermaximizes a utility function by

subjec

Et

X∞s¼0

1þ ρð Þ−sU Ctþs� � ð1Þ

t to the intertemporal budget constraint

EtX∞

s¼01þ rð Þ−s Ytþs−Ctþs

� �þAt ≥ 0 ð2Þ

Y is the income, A is the non-human wealth, and r is the constant and known real interest rate. I assume that the utility
wherediscount rate equals the real interest rate (r = ρ) and the U(.) is quadratic. I consider the optimal response of consumption toa income shock when the income follows each of the three processes; (i) a standard AR(1); (ii) a random walk; and (iii) anARIMA(1,1,0).
ið ÞYt ¼ ϕYt−1 þ εt 0 bϕ b 1ð Þ ð3Þ

iið ÞYt ¼ Yt−1 þ ηt ð4Þ

iiið ÞΔYt ¼ ϕΔYt−1 þ ζ t 0 bϕ b 1ð Þ ð5Þ

When the income follows (i) the AR(1) process, the income shock (εt) is a transitory one and does not persist into theindefinite future. Therefore, if the consumer experiences such an unexpected shock, he or she will adjust consumption only by asmall fraction of the income change and spread this temporary increase/decrease across the rest of lifetime to smooth outconsumption. By solving the model, the optimal consumption response to transitory income shocks, when income follows theAR(1) process, is shown to be

ΔCt ¼r

1þ r−ϕ

� �εt b εtð Þ: ð6Þ

Notice that, since the ϕ is less than one, consumption responds less than the transitory income shock.In contrast, when the income follows (ii) the random walk, an income change is permanent providing the consumer with

more/less income in every future period from which to consume. Therefore, a consumer experiencing an unexpected randomwalk shock (ηt) should increase/decrease consumption by the same amount as the increase/decrease in the income. By pluggingϕ = 1 to the one for a transitory shock, I can derive the optimal consumption response, as

ΔCt ¼ ηt: ð7Þ

When it comes to (iii) the ARIMA(1,1,0) or positively autocorrelated shock to income growth, once-and-for-all positiveincome shock has persistent effect on the income growth and the final effect of a blip in income is larger than the initial blip.Therefore, when a consumer experiences the unexpected shock to income growth (ζt), consumption rises/falls by more than theinitial increase/decrease in income. By solving the model, it is shown that the optimal response of consumption is

ΔCt ¼1þ r

1þ r−ϕ

� �ζ t Nζ tð Þ: ð8Þ

y,c AR(1) (transitory shock)

c

y

t = 0 t

y,c Random Walk (level shock)

t = 0 t

y,c ARIMA(1,1,0) (positively autocorrelated permanent shock)

yc

t = 0 t

y, c

Fig. 1. RE/PIH prediction on consumption path upon being struck by income shocks under different income process.


Notice that, for any positive ϕ (0 b ϕ b 1), consumption responds more than the income shock and the higher the ϕ, thegreater the response. The impulse response functions for the three income processes, along with the RE/PIH prediction for theoptimal consumption paths are illustrated in Fig. 1.

We care only unexpected shocks; if the income change was expected, it already should have been incorporated into theconsumer's choice of the optimal consumption path over the lifetime. That is, the RE/PIH predicts that consumer does not react topredictable income changes under full consumption risk sharing. Nevertheless, it is widely known that consumption responds toomuch to expected changes in income and such issue has been referred to as excess sensitivity puzzle.

Taking into account the shocks to income growth, we understand that dealingwith shocks to income does not necessarilymeanmoderating consumption volatility. On one hand, the RE/PIH prediction for a transitory income shock and predictable incomechanges support the conventional notion that more financially integrated countries should benefit by stabilizing consumptionvolatility. On the other hand, the RE/PIH predicts that instantaneous increase in consumption should be greater than that of incomeand thus international financial integration points to higher consumption volatility, when the positively autocorrelated permanentshocks to income growth are present.4 It is not a desk theory; Aguiar and Gopinath (2007) show that the income process may becharacterized by the trend shocks rather than the transitory shocks, particularly in the emergingmarket countries. Then, the puzzlein Kose et al. (2003), lack of reduction in the consumption volatility under financial integration, can be reconciled—countries mayhave managed the permanent shocks to income growth to increase consumption volatility.

4 Related to this, Kaminsky, Reinhart, and Vegh (2004) report procyclical capital inflow in most of the developed and developing countries. They provide threepotential underlying reasons: (i) dominating investment effect in a standard international business cycle model with physical capital stock, (ii) inter-temporaldistortion in consumption imposed by temporary policies, and (iii) varying risk premium over the business cycle.


3. Measuring the degree of international financial integration

Economists have reached a broad consensus that international financial integration has dramatically advanced in the past fewdecades. After the developed countries implemented liberalization of capital transactions in the 1970s and 1980s, a number of thedeveloping countries adopted liberalization in the late 1980s and the early 1990s. It then follows that not only the developedcountries but also the developing countries, to a greater or lesser extent, are increasingly more integrated to the internationalcapital markets in the 1990s and thereafter. This means that the national economy is connected to the rest of the world intwo-way. While vigorous participation by foreign investors heightens foreigners' share of domestic assets, domestic entities holdmore assets abroad.

Since financial account liberalization frequently spurs the foreign investment, de jure indicators of capital account openness,which follow the time schedule of liberalization policy implementation, are often quoted in the literature. They are typicallybinary variables taking unity when certain restrictions are non-existent and zero otherwise, or indexes constructed on suchbinary variables or laws and regulations coded.5 Although the odds are that financial account liberalization will be followed byincreased cross border holdings of assets, liberalization itself is not sufficient to induce a surge of capital flow. The existence ofpersistent impediments, such as country risk perceived by foreign investors and home bias by domestic asset holders, requiresubstantial lead time before witnessing the spur. Moreover, as Prasad, Rogoff, Wei, and Kose (2003) point out by referring tocapital flights from Latin American countries in the 1970s and 1980s, cross border holdings of assets can be increased withoutrelevant liberalization policy when capital control is less effective. For these reasons, a de jure measure of financial accountliberalization may not be the ideal one in light of the objective of this paper.

I, therefore, consider de facto measures of financial integration, which are constructed based on actual capital flows andother realized macroeconomics variables, instead. The current account balance in the balance of payment and the saving-investment correlation are typical examples of the de facto measures. It stands to reason that the current account balance cantake greater surplus and/or deficit with freer capital flows. Moreover, Feldstein and Horioka (1980), referring to the nationalincome identity, interpret the saving-investment correlation as an indicator of capital mobility.6 However, since I explicitlyconsider multiple income processes which have potentially offsetting impacts on the optimal consumption response and so thecapital flow and savings paths, year to year flow data are not appropriate to the empirical analysis. Therefore, I employ astock-based measure proposed by Lane and Milesi-Ferretti (2001, 2007). They estimate gross external assets and liabilitiesincluding portfolio equity, FDI, debt, financial derivatives and official reserves, via the accumulation of the corresponding inflowsand outflows as well as relevant adjustments for valuation changes, and argue that the sum of them, normalized by GDP, isa volume-based measure of international financial integration. The measure, denoted by LMF hereafter, in a country i at a yeart is thus

5 The(AREAEconstrua quant

6 Altetemporsaving-

7 TheZealand

8 It isand listIndonesUrugua

LMFi;t ¼FAi;t þ FLi;t

GDPi;t; ð9Þ

FA (FL) denotes the stock of total foreign assets (liabilities) in current US$.
whereTable 1 reports simple averages of LMF from the 1970s to the 2000s for the full sample countries, the OECD countries,7 and the
non-OECD countries, as well as the EU countries and the emerging market (EM) countries.8 Overall, international financialintegration measured by LMF has advanced since the 1970s. The average LMF for the full sample countries has grown steadily andmore than tripled from 77.1% in the 1970s to 250.9% in the 2000s. Looking at LMF by groups, the progress in the OECD and the EUcountries far outpaced that in the non-OECD countries. While there were no major gaps between groups in the 1970s and the1980s, LMF started to accelerate in the OECD and the EU countries in the 1990s and further in the 2000s. Predictably the EUcountries persistently record higher degree of financial integration than the other OECD countries. Although it is not as intensiveas in the OECD counties, the non-OECD countries also have shown steady increase in general since the 1970s. Viewed bysub-groups of the non-OECD countries, it appears no evidence that the EM countries are more integrated than the othernon-OECD countries. The EM countries rather fell behind in the 1990s presumably due to sequence of turmoil in the internationalfinancial markets, however their average LMF went back up and boosted further in the 2000s.

former is, for example, the IMF measure of exchange restrictions reported in Annual Report on Exchange Arrangements and Exchange RestrictionsR) and the official equity market liberalization indicator by Bekaert, Harvey, and Lundblad (2005), and the latter includes Chinn and Ito (2002), whoct an index based on the AREAER binary series to measure the intensity of capital controls, and Quinn (1997), who codes the laws of 64 countries to createitative measure of the regulation of international financial transactions.rnative interpretations, such as due to permanent shifts in the rate of technological progress, population growth with imperfect labor mobility, andary demand and productivity shocks, are possible for the observed high saving-investment correlation, as surveyed in Tesar (1991), suggesting that theinvestment correlation does not necessarily offer strong evidence for the question of international capital mobility.OECD countries include Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Japan, Netherland, New, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom (UK), and United States of America (USA), which were the OECD members in the 1970s.difficult to make an exact list of emerging market countries; I follow an investment information source, namely JP Morgan's EMBI global sovereign spread,32 countries (Algeria, Argentina, Brazil, Bulgaria, Chile, Colombia, Cote D'Ivore, Croatia, Dominican Republic, Ecuador, Egypt, El Salvador, Hungary,ia, Korea, Lebanon, Malaysia, Mexico, Morocco, Panama, Pakistan, Peru, Philippines, Poland, Russia, South Africa, Thailand, Tunisia, Turkey, Ukraine,y, Venezuela) as the emerging market countries.

Table 1Indicator of international financial integration (LMF).

1970s 1980s 1990s 2000s

Full sample countries (120 countries) 77.1 121.9 153.7 250.9OECD countries (22 countries) 73.7 124.8 212.6 521.2

(of which) EU countries (14 countries) 75.7 132.1 232.5 598.9Non-OECD countries (98 countries) 78.3 121.0 139.0 181.9

(of which) Emerging market countries (32 countries) 83.1 124.4 118.1 147.2

Note: LMF is a ratio of the sum of total foreign assets and liabilities to nominal GDP. This table reports LMF in percentage.Data Source: External Wealth of Nations Mark II data (Lane and Milesi-Ferretti (2007)) and Balance of Payment Statistics database (IMF).


4. Estimation model

The empirical analysis consists of two stages. The first stage decomposes GDP into permanent and transitory componentsand specifies their processes using the state-space/unobserved component model. I assume particular models for these twocomponents respectively and obtain maximum likelihood estimates through applying the Kalman filter to generate the predictionerror decomposition of the likelihood function.9 Note that the autoregressive coefficients vary by country, since I consider andestimate country-specific income processes. The second stage inquires how consumption responds to various types of incomeshocks and changes applying the first stage results and, in particular, whether excess sensitivity is detected after accounting forthe stochastic nature of income path. Moreover, by including the interaction terms between income shocks/changes and LMF,I explicitly investigate the effect of financial integration on the validity of the RE/PIH and on more successful consumption risksharing and smoothing.

First, GDP (yi,t) is decomposed into two components of permanent income (yPi,t) and transitory income (yTi,t) as,

wheredistrib

ΔyPi;t

ζi,t − 1

incomthe slofully a

9 For

yi;t ¼ yPi;t þ yTi;t; ð10Þ

subscripts i and t denote country and year, respectively. It is assumed that the permanent component follows the local
wherelinear trend model, in which both the slope and the level are stochastic, à la Harvey (1989),
yPi;t ¼ yPi;t−1 þ μ i;t−1 þ ηi;t ηi;t eNID 0;σ2ηi

� �;

μ i;t ¼ μ i;t−1 þ ζ i;t ζ i;t eNID 0;σ2ζ i

� �;

ð11Þ

μi,t − 1, the slope of yPi,t, follows a random walk. ηi,t and ζi,t are the serially independent disturbances from normalutions with zero mean. Then, the change in yPi,t has three elements as

¼ μ i;t−2 þ ζ i;t−1 þ ηi;t: ð12Þ

and ηi,t, which hold new information making up yPi,t, are the slope and level shocks. Note that ζi,t − 1 represents a shock toe growth affecting the slope permanently, while ηi,t stands for a one-time level shift in income which has no implication forpe. On the other hand, μi,t − 2 is past information known at period t representing a change in permanent income that can benticipated. Therefore, the RE/PIH predicts that consumption should respond to an unanticipated slope shock (ζi,t − 1) and ahock (ηi,t), but not to a predictable change in the permanent income (μi,t − 2).
level s
We also assume that the transitory component (yTi,t) follows an AR(1) process,

yTi;t ¼ ϕTiy

Ti;t−1 þ εi;t εi;t eNID 0;σ2

εi

� �; ð13Þ

ϕTi is the autoregressive coefficient and εi,t is the serially independent disturbance from a normal distribution with zero
where
mean. I call this stochastic disturbance, εi,t, as the transitory shock which constitutes a part of the change in the transitoryincome (ΔyTi,t). The remaining elements of ΔyTi,t, calculated as ϕT

iΔyTi,t − 1 − εi,t − 1, are a function of the model parameter andlagged values of income change and disturbance standing for a predictable change in the transitory income. The RE/PIH thuspredicts that consumption should respond to a transitory shock (εi,t) but not to a predictable change in the transitory income(ϕT

iΔyTi,t − 1 − εi,t − 1).The state-space/unobserved component model jointly estimates the autoregressive coefficient and the error terms. Thereby I

identify the slope, level, and transitory shocks (ζi,t − 1, ηi,t, and εi,t) as well as the predictable changes in the permanent andtransitory components of income (μi,t − 2 and ϕT

iΔyTi,t − 1 − εi,t − 1).Second, I examine the joint test of the RE/PIH after accounting for the stochastic properties of income process to explicate how

international financial integration bears upon consumption risk sharing. In particular, I elucidate the excess sensitivity ofconsumption and clarify the presence of full consumption smoothing for the transitory income—whether the consumption

more on the theory of parameter estimation in the state-space/unobserved components model using the Kalman filter, see Harvey (1989).


responds to the predictable changes in the transitory income. Moreover, I expound on how international financial integrationattunes anomaly of unreduced consumption volatility under financial integration by examining the RE/PIH prediction—consumption should come to respond more to the unexpected slope shocks as countries get more integrated to the internationalcapital market. Specifically, I consider following regression model, using the information obtained in the first stage,

10 TheShentonmodel s11 Wein the tmortgaasset pr12 Thewith a s

shock c

Δci;t ¼ kc þþβSLP1 SLPSHKi;t þ βSLP

2 SLPSHKi;t � LMFi;t� �

βLVL1 LVLSHKi;t þ βLVL

2 LVLSHKi;t � LMFi;t� �

þγP1PRDPMNTi;t þ γP

2 PRDPMNTi;t � LMFi;t� �

þβT1TRSTSHKi;t þ βT

2 TRSTSHKi;t � LMFi;t� �

þγT1PRDTRSTi;t þ γT

2 PRDTRSTi;t � LMFi;t� �

þδ LMFi;t þ vi þ ytð Þ þ ui;t;

ð14Þ

kc is a constant term, which helps capturing the effects of subsequent generations' increased consumption due to greater
wherelifetime wealth from the increased productivity over time. Inclusion of country fixed effects (vi) controls for unobservedtime-invariant country-specific effects. Finally, yt is the annual dummy variables that control for potential world-widemacroeconomic shocks, while ui,t is an error term including measurement error and preference shocks. SLPSHK, LVLSHK, andTRSTSHK are the slope shock, the level shock, and the transitory shock respectively. PRDPMNT and PRDTRST are the predictableincome changes in the permanent component and the transitory component respectively.
Data sources are the World Bank's World Development Indicators (WDI) database, Lane and Milesi-Ferretti (2007), and IMF'sBalance of Payment Statistics (BOPS) database. I employ annual per capita real GDP and private consumption in constant US$ in2000 taken from theWDI. LMF is constructed based on foreign assets and liabilities available in External Wealth of Nations Mark IIdata by Lane and Milesi-Ferretti (2007) and the BOPS database. The sample consists of 120 countries, of which 22 are the OECDcountries, 32 are the EM countries, and 66 are the other non-OECD countries. The sample period is from 1980 to 2011, startingwhen LMF began to diversify across the OECD and the non-OECD countries. The data set is not a balanced panel, since data aremissing in some countries and years.

5. Financial integration and consumption risk sharing

Statistical properties of income shocks and changes estimated by the state-space/unobserved component model in the firststage are reported in Table 2.10 A pronounced finding is the highly volatile shocks and income changes in the non-OECD countries.Looking at the standard deviations of slope shock (SLPSHK) and level shock (LVLSHK), the OECD and the EU countries appear torecord higher volatilities. However, they are affected by the sizes of income level and growth which is evident in the predictablechange in the permanent component of income (PRDPMNT). Consistently with the conventional wisdom, once adjusted for thedifference in their sizes, the EM and the other non-OECD countries showmuch higher volatilities in SLPSHK and LVLSHK as well asPRDPMNT. The EM and the other non-OECD countries also exhibit much more volatile transitory shocks (TRSTSHK), whichamplifies the volatilities of the predictable changes in the transitory income (PRDPTRST) with high AR(1) correlations. Theseobserved higher volatilities of shocks and income changes in the non-OECD countries suggest that there is huge room forconsumption risk sharing through international financial integration in the non-OECD countries.

Table 3 presents the parameter estimates of the Eq. (14) by the country fixed effects model and, as is evident in the table,inclusion of year dummy variables does not change the results.11 Before diving into the main findings of the regression analyses,let me sort out what we learn from the RE/PIH in Section 2. The RE/PIH predictions for the optimal consumption responses are

ΔCt ¼ r1þr−ϕT

h iεt bεtð Þ for a transitory shock (εt); ΔCt = ηt for a level shock (ηt); and ΔCt ¼ 1þr

r

� ζ t Nζ tð Þ for a slope shock (ζt) in

approximation.12 On the other hand, the RE/PIH predicts that consumption should not react to predictable changes in permanentand transitory components of income, since they should have already been factored in the optimal consumption path. These RE/PIH predictions are not likely to hold due to, among other reasons, credit constraints, however, if international financialintegration relaxes liquidity constraints, it is expected that we observe β2

SLP N 0, γ2P b 0, β2

T b 0, and γ2T b 0.

results are checked and passed by basic goodness-of-fit and serial correlation criterion. On the other hand, the results occasionally fail to satisfy Bowman–normality test statistic. However, according to Koopman, Harvey, Doornik, and Shephard (2006), this is a common occurrence caused by outliers and thehould not necessarily be rejected. As suggested by Koopman et al. (2006), the results are checked and confirmed by looking at the auxiliary residuals.observe significant coefficients only for the year dummy variables of 1986, 2005, 2006, and 2007, which are all positive, for the full sample (not reportedable). Interestingly, these years are just before large scale stock markets crashes around the world in 1987 (Black Monday) and in 2008 (Subprimege crisis and Bankruptcy of Lehman Brothers). The wealth effect might increase consumption disproportionately in these years before the corrections inices.slope shock considered in the local linear trend model is, in a sense, the case of ϕ = 1 in the ARIMA(1,1,0) process. Since the income will not convergelope shock, rigorously speaking, the model cannot be solved in the way we explored in Section 2. However, the optimal consumption response to a slope

an be investigated by plugging ϕ = 1 (or near value) to ΔCt ¼ 1þr1þr−ϕ

h iζ t which must be greater than ζt.

Table 2Statistical properties of income shocks and changes.

1980–2011

Mean Std. Dev. Min Max

Whole sample countries (120 countries)SLPSHK 1.18 100.30 −1463.56 1373.34LVLSHK −1.82 217.23 −2947.53 1515.17PRDPMNT 129.46 227.72 −1818.46 1782.25TRSTSHK −0.39 93.12 −2194.77 1779.70PRDTRST 0.19 39.81 −719.73 1010.36

OECD countries (22 countries)SLPSHK −7.44 187.94 −1463.56 1373.34LVLSHK −5.06 410.34 −2947.53 1515.17PRDPMNT 350.83 308.39 −1818.46 1782.25TRSTSHK −0.15 41.80 −399.31 282.20PRDTRST −0.16 3.21 −21.85 19.52

EM countries (32 countries)SLPSHK 3.49 53.92 −400.00 778.21LVLSHK −1.78 90.17 −522.02 536.53PRDPMNT 66.95 115.42 −497.16 587.89TRSTSHK −0.68 52.98 −1011.56 285.12PRDTRST 0.41 22.00 −213.84 349.63

Other non-OECD countries (66 countries)SLPSHK 3.87 48.75 −428.80 936.32LVLSHK −0.31 115.77 −1177.83 1155.43PRDPMNT 62.24 156.75 −613.11 1423.09TRSTSHK −0.32 124.51 −2194.77 1779.70PRDTRST 0.22 54.75 −719.73 1010.36

EU countries (14 countries)SLPSHK −6.78 118.81 −1350.30 920.58LVLSHK −7.56 394.20 −2947.53 1312.04PRDPMNT 325.13 256.92 −1547.20 1737.94TRSTSHK −0.27 23.47 −265.85 110.89PRDTRST −0.19 3.05 −21.85 19.52


We have discussed the RE/PIH prediction in a standard model with constant interest rate—that is, the consumer is involved inconsumption risk sharing through borrowing and lending. However, the international capital market provides equity instrumentsbesides debt instruments, and they lead to extra predictions for the optimal consumption response. When the consumer isinsured by equity instruments, the consumption responds less even to unexpected income shocks. In particular, the consumptionshould be adjusted less responding to a level shock and so we expect β2

LVL b 0.The point estimates for the response of consumption to income shocks/changes in a typical country, which I call marginal

propensity to consume (MPC) hereafter, are computed based on the coefficient estimates and LMF value as β1 + β2 ∗ average[LMF] for the unanticipated income shocks and γ1 + γ2 ∗ average[LMF] for the predicted income changes and are reported inTable 4. For the full sample regressions, as expected, the consumption is responding to five income shocks as well as twopredictable income changes. In particular, the consumption reacts to a predictable change in the transitory income (PRDTRST)indicating the excess sensitivity. Moreover, the MPCs for the slope shock (SLPSHK) are below one and are not necessarily greaterthan the MPCs for the predictable change in permanent income (PRDPMNT). These are at odds with the RE/PIH predictions andremind us that the countries are subject to credit constraints and do not fully succeed in consumption risk sharing. Therefore, inexploring how the countries benefit from international financial integration in terms of consumption risk sharing, we payattention to the coefficients for the interaction terms; β2

SLP for the slope shock (SLPSHK), β2LVL for the level shock (LVLSHK), γ2

P forthe predictable change in the permanent income (PRDPMNT), β2

T for the transitory shock (TRSTSHK), and γ2T for the predictable

change in the transitory income (PRDTRST).Most striking consequences in Table 3 emerge in transitory components. Consistently with the RE/PIH prediction,

international financial integration de-links income shocks/changes in transitory component and consumption; the consumptionresponds less to a transitory income shock as well as a predictable change in transitory income in more integrated countries. Itis inferred that the relaxed liquidity constraint alleviates excess sensitivity of consumption and contributes to consumptionrisk sharing and smoothing. Another striking finding is that international financial integration lessens the response of con-sumption to the level shocks, at odds with the RE/PIH. This implies that well-integrated countries are insured against theidiosyncratic country-specific level shocks through prearranged cross border holdings of equities and securities. On the otherhand, there is no clear evidence for the effect of financial integration on consumption risk sharing against the slope shocks. Since theslope shocks stand for the permanent shocks to income growth, financial integration should deliver even sharper adjustments ofconsumption responding to slope shocks. Although signs are consistent with this RE/PIH prediction, estimates are not statisticallysignificant.

Table 3Consumption growth, income shocks and changes, and international financial integration.

Whole period (1980–2011)

Full sample OECD & EM Non-OECD

SLPSHK (β1SLP) 0.396a 0.404a 0.375a 0.378a 0.752a 0.783a

(0.089) (0.090) (0.126) (0.129) (0.253) (0.247)SLPSHK ∗ LMF (β2

SLP) 0.020 0.019 0.014 0.014 −0.169b −0.181b

(0.018) (0.018) (0.018) (0.018) (0.082) (0.083)LVLSHK (β1

LVL) 0.425a 0.429a 0.429a 0.429a 0.524a 0.546a

(0.068) (0.066) (0.065) (0.062) (0.108) (0.108)LVLSHK ∗ LMF (β2

LVL) −0.026c −0.024c −0.026b −0.025b −0.060b −0.057c

(0.014) (0.013) (0.011) (0.011) (0.029) (0.031)PRDPMNT (γ1

P) 0.463a 0.480a 0.454a 0.470a 0.601a 0.623a

(0.080) (0.081) (0.080) (0.081) (0.151) (0.142)PRDPMNT ∗ LMF (γ2

P) 0.002 0.002 0.004 0.004 −0.022a −0.022a

(0.005) (0.005) (0.006) (0.005) (0.003) (0.003)TRSTSHK (β1

T) 0.544a 0.539a 0.765a 0.751a 0.466a 0.473a

(0.153) (0.163) (0.206) (0.216) (0.140) (0.148)TRSTSHK ∗ LMF (β2

T) −0.021b −0.020b 0.000 −0.001 −0.016b −0.015c

(0.008) (0.009) (0.099) (0.095) (0.007) (0.008)PRDTRST (γ1

T) 0.877a 0.909a −1.143 −0.994 0.996a 0.989a

(0.074) (0.075) (1.589) (1.546) (0.042) (0.058)PRDTRST ∗ LMF (γ2

T) −0.036a −0.040a 1.931 1.821 −0.044a −0.045a

(0.005) (0.005) (1.556) (1.505) (0.003) (0.004)LMF (δ) −1.350 −0.995 0.380 1.837 3.349a 3.142a

(1.710) (1.821) (3.173) (3.194) (0.766) (0.731)CONSTANT 0.617 9.553 6.271 9.529 −14.736 −3.263

(19.911) (10.542) (32.764) (15.436) (13.618) (9.429)Year dummy Yes No Yes No Yes NoMean LMF 1.928 1.928 2.123 2.123 1.563 1.563R-squared (within) 0.432 0.421 0.505 0.487 0.412 0.399Number of observations 3011 3011 1562 1562 2324 2324Number of countries 120 120 54 54 98 98

Notes:Upper rows are coefficients and lower rows in parentheses are robust standard errors clustered in countries.

a Estimates significantly different from zero at the 1% level.b Estimates significantly different from zero at the 5% level.c Estimates significantly different from zero at the 10% level.


Tables 3 and 4 also report the results for the sub-sample regressions to explore potential differences in the estimates by thechoice of sample countries. I consider two sub-samples; the OECD and the EM countries, which have been more popular in theinternational capital market, and the non-OECD countries (the EM and the other non-OECD countries).13

First of all, consumption risk sharing against the level shocks through international financial integration is shown to beuniversal regardless of the degrees of development of countries and financial integration. Moreover the MPCs for the three shocksand the predictable changes in permanent income for sub-sample regressions are also very similar to the ones for the full samplecountries. On the other hand, it turns out that the effect of international financial integration on consumption risk sharing againstthe transitory shocks and the predictable changes in transitory income vary in the degree of development. It appears thatinternational financial integration is more important in reducing excess sensitivity among the non-OECD countries. With regardto the two MPCs in transitory component, while estimates in the regressions among the non-OECD countries are quite similar tothe ones in the full sample regressions, these MPCs among the OECD and the EM countries are less reliable due to the insignificantestimates for β2

T and γ2T.

6. Consumption risk sharing by country groups

Table 5 reports the results of the regressions without interaction terms with LMF for the OECD, EM, and the EU countries,which show variations in the degree of international financial integration. Although the sample countries are different from thoseof the regressions with interaction terms, the estimates for the coefficients in Table 5 are comparable with the five MPCs in Table 4and, in a sense, clarify the fruits of financial integration in terms of consumption risk sharing.

Most noticeably, while the consumption is responding to all the five shocks and income changes in the EM countries, no excesssensitivity is detected in the OECD and the EU countries. In the OECD and the EU countries, which are more integrated, theconsumption is not significantly adjusted to a predictable change in transitory income (PRDTRST). Related to this, we also observe

13 Since the regression model primarily intends to identify how international financial integration helps consumption risk sharing, the sample countries ideallyrecord larger variations in LMF. These two sub-sample regressions, rather than the regression for each of the country groups (e.g. the OECD countries), benefitvariations in LMF across the country groups, which we discuss in the Section 3, in addition to the ones within country groups.

Table 4MPCs for income shocks and changes.

RE/PIH prediction Country Whole period (1980–2011)

MPC for SLPSHK (β1SLP + β2

SLP ∗ average[LMF])Δc = (1 + r)/r (N1) Full sample 0.434 0.440

(0.000) (0.000)OECD & EM 0.403 0.408

(0.000) (0.000)Non-OECD 0.488 0.501

(0.007) (0.005)Year dummy Yes No

MPC for LVLSHK (β1LVL + β2

LVL ∗ average[LMF])Δc = 1 Full sample 0.374 0.384

(0.000) (0.000)OECD & EM 0.374 0.376

(0.000) (0.000)Non-OECD 0.430 0.457


MPC for PRDPMNT (γ1P + γ2

P ∗ average[LMF])Δc = 0 Full sample 0.467 0.484

(0.000) (0.000)OECD & EM 0.463 0.479

(0.000) (0.000)Non-OECD 0.566 0.589


MPC for TRSTSHK (β1T + β2

T ∗ average[LMF])Δc = r/(1 + r − ϕ) (b1) full sample 0.504 0.501

(0.000) (0.001)OECD & EM 0.765 0.749

(0.000) (0.000)Non-OECD 0.441 0.449


MPC for PRDTRST (γ1T + γ2

T ∗ average[LMF])Δc = 0 Full sample 0.807 0.832

(0.000) (0.000)OECD & EM 2.957 2.872

(0.095) (0.092)Non-OECD 0.928 0.919


Notes: Upper rows are point estimates and lower rows in parentheses are P-values of the F-tests for the hypotheses that MPCs are equal to zero.


greater response of consumption to a transitory shock (TRSTSHK) in the OECD than in the EM countries―it might be thatconsumers in the OECD countries tend to factor in subsequent predictable income changes upon facingwith a persistent transitoryshock. The MPC for the transitory income shocks in the OECD countries are estimated to be somewhat larger than the RE/PIHprediction, however, they are not necessarily too large given the very persistent transitory income processes.14 Furthermore theMPC for the level shocks (LVLSHK) is estimated to be below one, contrary to the RE/PIH prediction, and those in the OECD and theEU countries are smaller than in the EM countries. It is inferred that more integrated OECD and EU countries are insured bycontingent bonds and so the consumption responds less to level shocks.

Table 5 also presents the regressions results for the sub-period of post 1990, when European countries have facilitatedintegration through creating the EU and the euro (Gehringer, 2013). In this sub-period, the consumption does not significantlyrespond to the transitory income shocks (TRSTSHK) in the EU countries. Although it is not consistent with the RE/PIH prediction,it can be interpreted as the consumption risk sharing through contingent bonds. The EU countries have accelerated financialintegration and becomemore insured against the unexpected transitory income shocks in the 1990s and the 2000s. This takes thedebate one step further from the finding by Lewis (1996) where she shows that the risk sharing can not be resolved by either thecapital market restrictions or nonseparabilities between tradables and nontradables. In contrast, my approach, by explicitlyconsidering the permanent and transitory components of income, indicates full consumption risk sharing in the EU countrieswithout dealing with disaggregated consumption data.

14 AR(1) coefficients often exceeds 0.98. The model prediction of MPC is 0.73 with r = 0.04 and ϕ = 0.985.

Table 5Consumption growth and income shocks and changes (without interaction terms with LMF).

Whole period (1980–2011) Post 1990 period (1990–2011)

OECD EM EU OECD EM EU

SLPSHK (β3SLP) 0.454a 0.467a 0.834a 0.817a 0.438a 0.431a 0.449a 0.461a 0.845a 0.826a 0.420a 0.403a

(0.078) (0.085) (0.114) (0.105) (0.130) (0.117) (0.079) (0.086) (0.132) (0.126) (0.140) (0.121)LVLSHK (β3

LVL) 0.322a 0.320a 0.580a 0.584a 0.370a 0.331a 0.318a 0.314a 0.622a 0.617a 0.332a 0.301a

(0.042) (0.038) (0.064) (0.062) (0.048) (0.033) (0.041) (0.036) (0.066) (0.059) (0.035) (0.025)PRDPMNT (γ3

P) 0.402a 0.439a 0.773a 0.753a 0.518a 0.550a 0.401a 0.439a 0.793a 0.770a 0.510a 0.542a

(0.055) (0.054) (0.114) (0.106) (0.083) (0.093) (0.057) (0.056) (0.094) (0.091) (0.094) (0.099)TRSTSHK (β3

T) 1.196a 0.995a 0.742a 0.728a 0.974b 0.415 1.178a 1.026a 0.777a 0.754a 0.795 0.456(0.305) (0.291) (0.057) (0.058) (0.403) (0.247) (0.314) (0.284) (0.070) (0.073) (0.460) (0.336)

PRDTRST (γ3T) −2.151 −1.827 0.829a 0.828a −2.687 −2.653 −1.380 −1.550 0.792a 0.817a −1.704 −3.255

(3.540) (2.616) (0.032) (0.036) (4.434) (2.907) (3.499) (2.678) (0.051) (0.052) (5.727) (4.223)CONSTANT 31.173 26.809 −7.326 −6.092 −44.829 −16.941 32.326 30.640 −7.744 −8.435 −39.190 −19.353

(59.270) (18.892) (21.624) (7.248) (38.865) (29.437) (60.457) (19.273) (21.031) (7.598) (39.495) (31.601)Year dummy Yes No Yes No Yes No Yes No Yes No Yes NoMean LMF 3.165 3.165 1.304 1.304 3.583 3.583 3.994 3.994 1.338 1.338 4.536 4.536R-squared (within) 0.487 0.440 0.713 0.702 0.697 0.624 0.483 0.440 0.729 0.720 0.720 0.666Number of observations 687 687 903 903 440 440 668 668 664 664 308 308Number of countries 22 22 32 32 14 14 22 22 32 32 14 14

Notes:Upper rows are coefficients and lower rows in parentheses are robust standard errors clustered in countries.



7. Asymmetry between positive and negative shocks

One relevant question is whether international financial integration contributes to consumption risk sharing and smoothingsymmetrically in conjunction with the positive and negative income shocks/changes. To address this question, the regressionwith interaction terms is repeated but allowing asymmetric consumption response to positive and negative income shocks/changes. In particular, five income shocks and changes are replaced by two separate variables respectively, for positive andnegative shocks/changes.15 Table 6 presents the estimates for the interaction terms to show how international financialintegration helps consumption risk sharing for the positive and negative income shocks/changes as well as the results of F-test forthe hypothesizes that financial integration contributes symmetric way to consumption risk sharing against the positive andnegative income shocks/changes.

The results support both symmetric and asymmetric effects of international financial integration on consumption risk sharingdepending on the types of income shocks and changes. The asymmetry is remarkable for the slope shocks (SLPSHK), particularlyamong the OECD and the EM countries. Consistently with the RE/PIH prediction, international financial integration enablescountries to borrow to expand consumption instantaneously more than the income growth upon experiencing a positive shock toincome growth. The puzzling absence of reduced consumption volatility under financial integration can be reconciled byconsidering this mechanism. Indeed, the regression results point to a possibility that the higher and unreduced volatilities ofconsumption growth are rather the results of risk sharing against the permanent trend shocks through international financialintegration and are not a testimony to the tenuous benefit of financial integration. We also observe asymmetry for the transitoryshocks (TRSTSHK). It follows that the consumption responds less to a positive transitory shock in more integrated countriesimplying that more integrated counties are involved in savings through accumulating foreign assets (or reducing foreign debts) inthe face of a positive transitory shock. On the other hand, same countries do not necessarily increase debts to supportconsumption upon experiencing a negative transitory shock. It might be that countries find it less burdensome to liquidizedomestic assets, because more foreign borrowings potentially uplift risk premium particularly for developing countries.

In contrast, we observe more symmetry for the level shocks (LVLSHK) and the predictable changes in transitory income(PRDTRST). By construction, the countries can deal with the level shocks only through equity instruments and so it is expectedthat β2

LVL positive and β2LVL negative match up. The hypothesis that these estimates are equal is hardly rejected. We also observe that

international financial integration reduces excess sensitivity of consumption in the face of both positive and negative predictablechanges in transitory income in general and, in particular, among the non-OECD countries.16 The greater absolute values of theestimates for γ2

T negative imply that international financial integration is more helpful in allowing countries to borrow to supportconsumption in the face of negative predictable income changes rather than to save responding to positive income changes.

15 Namely, income shock/changepositive (e.g. SLPSHKpositive) equals income shock/change (SLPSHK) if the income shock/change is positive and zero otherwise;income shock/changenegative (e.g. SLPSHKnegative) equals income shock/change if the income shock/change (SLPSHK) is negative and zero otherwise.16 The estimates for regressions among the OECD and the EM countries are less reliable because, as we see in the Section 6, the OECD countries tend not to adjustconsumption responding to a predictable change in transitory income.

Table 6Positive and negative income shocks/changes and international financial integration.

Whole period (1980–2011)

Full sample OECD & EM Non-OECD

SLPSHKpositive ∗ LMF (β2SLP positive) 0.046b 0.046b 0.055a 0.055a −0.184 −0.174

(0.019) (0.019) (0.020) (0.018) (0.136) (0.142)SLPSHKnegative ∗ LMF (β2

SLP negative) −0.003 −0.004 −0.023 −0.020 −0.265 −0.264(0.021) (0.021) (0.014) (0.015) (0.232) (0.235)

F-test [β2SLP positive = β2

SLP negative] prob N F 0.176 0.180 0.004 0.008 0.737 0.714

LVLSHKpositive ∗ LMF (β2LVL positive) −0.020 −0.015 −0.016 −0.013 −0.087c −0.084c

(0.015) (0.015) (0.014) (0.014) (0.046) (0.048)LVLSHKnegative ∗ LMF (β2

LVL negative) −0.032b −0.032b −0.030b −0.032b −0.036 −0.035(0.014) (0.013) (0.013) (0.013) (0.027) (0.028)

F-test [β2LVL positive = β2

LVL negative] prob N F 0.550 0.387 0.408 0.298 0.322 0.350

PRDPMNTpositive ∗ LMF (γ2P positive) 0.011c 0.011 0.018 0.015 −0.019c −0.018c

(0.006) (0.007) (0.011) (0.010) (0.010) (0.010)PRDPMNTnegative ∗ LMF (γ2

P negative) 0.014 0.014 0.018 0.021 −0.006 −0.017(0.016) (0.016) (0.018) (0.018) (0.103) (0.106)

F-test [γ2P positive = γ2

P negative] prob N F 0.879 0.871 0.984 0.793 0.903 0.988

TRSTSHKpositive ∗ LMF (β2T positive) −0.050a −0.048a −0.053 −0.058 −0.024a −0.023a

(0.017) (0.016) (0.108) (0.102) (0.007) (0.007)TRSTSHKnegative ∗ LMF (β2

T negative) 0.000 0.000 0.006 0.006 −0.011 −0.011(0.013) (0.013) (0.054) (0.044) (0.015) (0.015)

F-test [β2T positive = β2

T negative] prob N F 0.045 0.042 0.499 0.467 0.533 0.556

PRDTRSTpositive ∗ LMF (γ2T positive) −0.027b −0.033b −2.229b −2.146b −0.043a −0.043a

(0.013) (0.013) (0.951) (1.045) (0.009) (0.008)PRDTRSTnegative ∗ LMF (γ2

T negative) −0.052a −0.053a 4.076b 3.740b −0.054a −0.056a

(0.019) (0.019) (1.649) (1.628) (0.009) (0.011)F-test [γ2

T positive = γ2T negative] prob N F 0.383 0.478 0.007 0.018 0.457 0.362

Year dummy Yes No Yes No Yes NoMean LMF 1.928 1.928 2.123 2.123 1.563 1.563R-squared (within) 0.464 0.455 0.558 0.540 0.431 0.418Number of observations 3011 3011 1562 1562 2324 2324Number of countries 120 120 54 54 98 98

Notes:Upper rows are coefficients and lower rows in parentheses are robust standard errors clustered in countries.Regressions include all the positive and negative shocks and income changes as well as LMF. Only interaction terms are reported.



8. Conclusion

This paper empirically shows that international financial integration improves consumption risk sharing and smoothing in awide range of countries from the OECD to the non-OECD countries. International financial integration has advanced dramaticallyin the last several decades. Many countries, including those classified as the emerging market and the developing countries,liberalized their capital transactions, and as a consequence, surges in the cross-border holdings of financial assets have beenobserved since the 1980s and, in particular, in the post 1990s globalization period. However, the international capital market hasbeen severely constrained by various types of incompleteness, and, on top of them, the emerging market countries are subject tothe risk of sudden stop or large scale reversal of capital flows. Thus, policymakers often regard high uncertainty in the accessibilityof the international capital markets during economic downturns and so are skeptical about the gain from international financialintegration in terms of consumption risk sharing. Contrary to such concerns, this paper provides favorable evidence for thebenefits of international financial integration. The results of regressions elicit an important policy implication. Cross-borderholdings of financial assets have been beneficial to public welfare through consumption risk sharing and smoothing in both theOECD and the non-OECD countries, despite the incomplete and precarious natures of the international capital market.

A formal test of the RE/PIH is conducted employing a de factomeasure of international financial integration to investigatewhetherand to what extent the countries are subject to credit constraints and to clarify how and to what degree international financialintegration de-links national income and consumption. I particularly argue that we must be attentive to the stochastic properties ofincome process, since the RE/PIH prediction for the innovation in consumption differs among underlying income processes. Inspiredby Aguiar and Gopinath (2007)who emphasize the importance of the shocks to trend growth in the emergingmarket countries, twounobserved components of income (i.e. the permanent component and the transitory component) and three types of income shocks


(i.e. the level shock, the slope shock and the transitory shock) are considered. The state-space/unobserved component model jointlyestimates the autoregressive parameter and the three types of shocks which constitute the two components of income as well asthe predictable income changes. Then, I shed light on the excess sensitivity of consumption by probing into the presence of fullconsumption smoothing as well as on the anomaly of unreduced consumption volatility by examining whether financial integrationhelps consumption to respond more to the unpredicted permanent shock to income growth.

I show that the presence of consumption risk sharing and smoothing varies by the degree of financial integration and thecomponents of income. Contrary to what we observe in the full sample countries case, the excess sensitivity of consumption wasnot detected in the OECD countries. Moreover, it is shown that the EU countries have been under full consumption risk sharing forthe unexpected transitory shocks in the post 1990 periods. In contrast, the consumption responds to the predictable changes intransitory income in significant manner in the non-OECD countries. On the other hand, even the OECD countries are shown to befar from successful in dealing with the permanent shocks to income growth.

The regression results also point to the fact that international financial integration helps consumption risk sharing andsmoothing. More integrated countries are better able to cope with various kinds of income shocks in general. Consistently withthe RE/PIH prediction, the consumption is much less responsive to unanticipated transitory income shocks in more integratedcountries. Moreover, both the OECD and the non-OECD countries come to adjust consumption less responding to the permanentrandom walk shocks, as they get more integrated, in particular, through exchanging equities and securities. Furthermore, theregression results support the RE/PIH prediction that financial integration delivers even larger increases in consumptionresponding to positive shocks to income growth. This effect is estimated to be particularly clear among the OECD and theemerging market countries and thus can reconcile the anomaly of unreduced consumption volatility under international financialintegration, which is documented in Kose et al. (2003).

The new approach to the joint RE/PIH test draws different conclusions in terms of the capacities to deal with the country-specific risks and the implication of international financial integration from the ones in previous literature. It has become aconventional belief that the potential gains from global risk sharing are quite sizable for the OECD countries and even much largeramong the developing countries (e.g., Athanasoulis & van Wincoop, 2000; van Wincoop, 1999) and, although still being far fromperfect, international financial integration has substantially advanced in the last couple of decades. On the other hand, manyempirical works have found less evidence of the benefit from international financial integration in terms of consumption risksharing and smoothing, and thus understanding risk sharing patterns has been of considerable interest. In this sense, the results ofthis article, pointing to improved consumption risk sharing and smoothing not only in the OECD but also in the non-OECDcountries, create a stir in the debates.

Acknowledgements

This article is based on my doctoral dissertation submitted to the University of Michigan in September 2007. I am grateful toLinda Tesar, Jan Svejnar, Kathryn Dominguez, Jing Zhang, and Christopher House for their guidance and encouragement. I thankthe seminar participants at the University of Michigan, the Bank of Japan, and the Seton Hall University for valuable suggestionstoo. I also appreciate the significant comments provided by the editor and the anonymous referees. Any remaining errors aresolely the author's responsibility.

References

Aguiar, M., & Gopinath, G. (2007). Emerging market business cycles: The cycle is the trend. Journal of Political Economy, 115(1), 69–102.Artis, M. J., & Hoffmann, M. (2008). Financial globalization, international business cycles and consumption risk sharing. The Scandinavian Journal of Economics,

110(3), 447–471.Asdrubali, P., Sørensen, B. E., & Yosha, O. (1996). Channels of interstate risk sharing: United States 1963–1990. Quarterly Journal of Economics, 111(4), 1081–1110.Athanasoulis, S. G., & van Wincoop, E. (2000). Growth uncertainty and risksharing. Journal of Monetary Economics, 45(3), 477–505.Backus, D. K., Kehoe, P. J., & Kydland, F. E. (1992). International real business cycles. Journal of Political Economy, 100(4), 745–775.Bai, Y., & Zhang, J. (2012). Financial integration and international risk sharing. Journal of International Economics, 86(1), 17–32.Bekaert, G., Harvey, C. R., & Lundblad, C. (2005). Does financial liberalization spur growth? Journal of Financial Economics, 77(1), 3–55.Bekaert, G., Harvey, C. R., & Lundblad, C. (2006). Growth volatility and financial liberalization. Journal of International Money and Finance, 25(3), 370–403.Chinn, M. D., & Ito, H. (2002). Capital account liberalization, institutions and financial development: Cross country evidence. NBER working paper (no. 8967).Ćorić, B., & Pugh, G. (2013). Foreign direct investment and output growth volatility: A worldwide analysis. International Review of Economics and Finance, 25, 260–271.Elwood, S. K. (1998). Testing for excess sensitivity in consumption: A state-space/unobserved components approach. Journal of Money, Credit, and Banking, 30(1),

64–82.Feldstein, M., & Horioka, C. (1980). Domestic saving and international capital flows. The Economic Journal, 90(358), 314–329.Flavin, M. A. (1981). The adjustment of consumption to changing expectations about future income. Journal of Political Economy, 89(5), 974–1009.Gehringer, G. (2013). Growth, productivity and capital accumulation: The effects of financial liberalization in the case of European integration. International

Review of Economics and Finance, 25, 291–309.Harvey, A. C. (1989). Forecasting, structural time series models and the Kalman filter. Cambridge, UK.: Cambridge University Press.Kaminsky, G. L., Reinhart, C. M., & Vegh, C. A. (2004).When it rains, it pours: Procyclical capital flows and macroeconomic policies. NBERWorking Paper (No. 10780).Koopman, S. J., Harvey, A. C., Doornik, J. A., & Shephard, N. (2006). STAMP: Structural time series analyser, modeller and predictor. London: Timberlake Consultants Press.Kose, M. A., Prasad, E., & Terrones, M. (2003). Financial integration and macroeconomic volatility. IMF working paper (WP/03/50).Kose, M. A., Prasad, E., & Terrones, M. (2007). How does financial globalization affect risk-sharing? Patterns and channels. IMF working paper (WP/07/238).Lane, P. R., & Milesi-Ferretti, G. M. (2001). The external wealth of nations: Measures of foreign assets and liabilities for industrial and developing countries. Journal

of International Economics, 55(2), 263–294.Lane, P. R., & Milesi-Ferretti, G. M. (2007). The external wealth of nations mark II: Revised and extended estimates of foreign assets and liabilities, 1970–2004.

Journal of International Economics, 73(2), 223–250.Lewis, K. K. (1996). What can explain the apparent lack of international consumption risk sharing? Journal of Political Economy, 104(2), 267–297.

http://refhub.elsevier.com/S1059-0560(13)00088-9/rf0020




























Lewis, K. K. (2000). Why do stocks and consumption imply such different gains from international risk sharing? Journal of International Economics, 52(1), 1–35.Mendoza, E. G. (1991). Real business cycles in a small open economy. American Economic Review, 81(4), 797–818.Prasad, E., Rogoff, K., Wei, S., & Kose, M. A. (2003). Effects of financial globalization on developing countries: Some empirical evidence. IMF working paper.Quinn, D. (1997). The correlates of changes in international financial regulation. The American Political Science Review, 91(3), 531–551.Sørensen, B. E., & Yosha, O. (1998). International risk sharing and European monetary unification. Journal of International Economics, 45(2), 211–238.Tesar, L. (1991). Savings, investment and international capital flows. Journal of International Economics, 31(1–2), 55–78.van Wincoop, E. (1999). How big are potential welfare gains from international risksharing? Journal of International Economics, 47(1), 109–135.








Documents

Financial integration and consumption risk sharing and smoothing