The idea that agents attempt to insure their consumption streams against individual income or wealth fluctuations is a pervasive feature of many modern dynamic macro models. It arises in closed economy models where there is heterogeneity in income or preferences across agents, in open economy models where countries with heterogeneous income streams borrow and lend internationally, and in models where consumption insurance may be a trigger (or a deterrent) for long-term growth.

In Discussion Paper No. 1074, Research Fellow Fabio Canova and Morten Ravn empirically examine whether consumption insurance holds at an international level. They begin by showing the implications of international risk sharing in a very simple theoretical model and extend the analysis to cover more complicated models. The basic implication of the analysis is that monotonic transformations of aggregate consumption must be highly correlated across countries, even when preferences are time non-separable, when there are non-separabilities across goods, when leisure choices are included and when there are non-traded goods. They then derive some testable implications of the theory and compare our testing approach to others which exist in the literature.

Four major results stand out from the empirical analysis. First, aggregate consumption appears to be fully insured against shocks to real, fiscal, monetary and demographic variables. Second, aggregate consumption co-varies with some lagged country-specific demographic and labour market variables over medium-long cycles. Third, two other implications of the risk sharing proposition (a set of cross-equation constraints across the moments they test and the restriction that consumption correlations across countries must be perfect) are, in many instances, rejected regardless of the type of fluctuations they considered. Fourth, there is very little evidence of risk sharing in the very long-run.

International Consumption Risk Sharing
Fabio Canova and Morten O Ravn

Discussion Paper No. 1074, September 1994 (IM)