Common correlated effects and international risk sharing

• DOI: http://doi.org/10.1111/infi.12119
• Author: Peter Fuleky,Qianxue Zhao,Luigi Ventura
• Published date: 01 March 2018
• Date: 01 March 2018

DOI: 10.1111/infi.12119

ORIGINAL ARTICLE

Common correlated effects and international

risk sharing

Peter Fuleky

1,2

|

Luigi Ventura

3

|

Qianxue Zhao

2

1

UHERO, University of Hawaii at

Manoa, Honolulu, Hawaii

2

Department of Economics, University of

Hawaii at Manoa, Honolulu, Hawaii

3

Department of Economics and Law,

Sapienza, University of Rome, Rome,

Italy

Correspondence

Peter Fuleky, UHERO, University of

Hawaii at Manoa, Honolulu, HI.

Email: fuleky@hawaii.edu

Abstract

Correct assessment of consumption risk and its international

diversification has important policy implications. However,

existing studies of international risk sharing rely on the

unrealistic assumptions that all economies are characterized

by symmetric preferences and uniform transmission of

global shocks. We relax these homogeneity constraints and

compare our proposed approach with the conventional ones

using a 44-year panel of 120 countries. Our results confirm

that consumption is only partially smoothed internationally

and risk sharing is directly related to the level of development,

but we do not detect a significant increase in risk sharing

during the surge in financial globalization over the last four

decades.

1

|

INTRODUCTION

Since the early contributions by Cochrane (1991), Mace (1991), and Obstfeld (1994), a number of

consumption risk sharing tests have been presented in the literature. Published research finds excess

sensitivity of consumption to income shocks and this has been interpreted as lack of international risk

sharing. To maintain analytical tractability, the derivation and implementation of risk sharing tests

usually relies on several homogeneity assumptions that are unlikely to hold in worldwide panels: all

economies are assumed to be characterized by symmetric preferences and uniform transmission of

global shocks. The extent of risk sharing is then estimated by a panel data regression of cross-

sectionally demeaned consumption on cross-sectionally demeaned income. However, if the

homogeneity assumptions underlying the analysis are violated, the results may be biased.

We extend the existing literature by taking into account various sources of heterogeneity.

Specifically, we allow for cross-country variation in preferences, endowments, structure of the

economy—all of which affect the transmission of global and regional

1

shocks to countries—and in the

International Finance. 2018;21:55–70. wileyonlinelibrary.com/journal/infi © 2017 John Wiley & Sons Ltd

|

55

extent to which consumption reacts to income fluctuations. Considering these sources of heterogeneity

is worthwhile for the following reasons:

1. The underlying theory suggests that if a country has full access to international risk sharing

opportunities, consumption will be independent of idiosyncratic income shocks. However, this does

not necessarily imply uniform consumption growth around the world. Country-level and global

consumption will move in lockstep if preferences are symmetric across countries, but they will

diverge if risk aversion, endowments, or discount factors are heterogeneous (Obstfeld, 1989).

Hence, even under perfect risk sharing, consumption paths can differ from each other due to

heterogeneous preferences.

2. If risk sharing is imperfect, consumption will also be affected by idiosyncratic income fluctuations.

This in turn raises the question of how to isolate idiosyncratic income shocks from global ones. Due

to differences in their productive and financial structure, regulations, and their participation in

international trade, countries may be affected by aggregate shocks to varying degrees. Accordingly,

idiosyncratic fluctuations are obtained by controlling for the extent of global shocks transmitted to

individual countries; an extent that is unlikely to be uniform throughout the world (Giannone &

Lenza, 2010).

3. Finally, because of differences in the quality of smoothing channels, the effects of idiosyncratic

income shocks on consumption may also vary across countries.

All three sources of cross-sectional heterogeneity have important consequences. We argue that the

appropriate method for filtering out the unobserved common factors from the observed variables

should allow for the heterogeneity of countries in terms of their preferences and exposure to aggregate

risk. Therefore—at odds with the existing literature—we let global factors have country-specific

loading coefficients. In addition, we relax the homogeneity assumption behind pooled or fixed effects

estimation and employ a mean-group type estimator that is robust to heterogeneous country

characteristics. Due to these refinements, the proposed approach can isolate idiosyncratic fluctuations

and is less susceptible to bias than the cross-sectional demeaning method.

We contribute to the existing literature in several important ways. First, we highlight the

shortcomings of the conventional approach to analyse international risk sharing. Second, we

re-evaluate earlier results on the lack of perfect risk sharing using a more flexible econometric model

that isolates idiosyncratic fluctuations in the data. Specifically, we are the first to apply the common

correlated effects (CCEs) estimator of Pesaran (2006) to the analysis of international risk sharing.

Third, while earlier studies have focused on smaller more homogeneous sets of economies, our large

panel of 120 countries allows us to analyse risk sharing along a variety of economic characteristics.

Fourth, we look at the change in the extent of risk sharing over the past 40 years, but find no evidence to

support the notion that financial globalization has led to an increase in international consumption

smoothing.

2

|

THE CONVENTIONAL APPROACH

Regression-based risk sharing (or consumption insurance) tests are based on the null hypothesis of

market completeness, or the possibility to redistribute wealth (hence, consumption) across all date–

event pairs. Under market completeness, the solution to the representative agent's maximization

problem ensures that marginal utility growth is equalized across agents and depends on aggregate

factors but not on individual shocks (Cochrane, 1991; Mace, 1991; Obstfeld, 1994). Assuming CRRA

56

|

FULEKY ET AL.