Ebook The Lack of International Consumption Risk Sharing: can Inflation Differentials and Trading Costs help explain the Puzzle?
Risk sharing between regions and nations has been the focus of much empirical research over the last decade. The main conclusion that emerges from this literature is that regions within a country share a lot more consumption risk than do countries.
Most of the evidence on international and interregional risk sharing is based on panel regressions of real idiosyncratic consumption growth on other idiosyncratic variables, mostly national or regional output. The motivation behind such risk sharing regressions is that in a world with complete capital markets, countries and regions will insure completely against any idiosyncratic risk. If furthermore, trade in goods markets is frictionless so that prices equalize across countries and regions, then, ex post, there should not be any correlation between a country’s or region’s relative output and consumption. The size of the regression coefficient of idiosyncratic consumption on idiosyncratic output can therefore be interpreted as a measure of the deviation from the complete markets outcome.
This paper argues that the results obtained from such regressions in international data are not generally comparable to those based on regional data. The reason for this is that for most countries consumer price indexes are not available at the regional level. Therefore, in regional data, the commonly applied procedure is to transform nominal into real quantities by deflating with the country-wide CPI. This practice preserves fluctuations in the relative value of consumption across regions. In this paper, we advocate this practice also for international data sets. Earlier studies that have examined risk sharing in international data have typically deflated the data with national (i.e. country-specific) CPIs. In this way, only fluctuations in the relative quantities but not in the relative value of consumption are preserved.
What may at first sight appear as a measurement issue is, in fact, an important conceptual difference: in addition to quantity (i.e.: capital income and credit) flows, fluctuations in relative consumer prices may constitute a separate channel of risk sharing. I add such a price channel to the popular variance decomposition by Asdrubali, Sørensen and Yosha (1996). Whereas earlier versions of this decomposition have focused on the relative importance of capital income and credit flows for risk sharing, the version suggested here also allows to gauge the contribution of price fluctuations. In the framework of this decomposition, it is straightforward to give economic meaning to the different practices of deflating regional and international data: the procedure commonly used on international data simply eliminates fluctuations in relative purchasing power and therefore does not pick up their contribution to risk sharing. Since fluctuations in relative prices are particularly important at the international level where goods markets are relatively segmented, one may expect that the omission of this channel may have a particularly pronounced effect on estimates of risk sharing obtained from international data sets.
We find this conjecture confirmed in our empirical investigation, for which we use data from the Penn World tables for 22 industrialized countries from 1973-2000. Our results show that once the price channel is accounted for in a comparable way, the coefficients estimated from risk sharing regressions are similar in regional and international data sets. Hence, conceptual differences in the preparation of the data used in estimation seem to explain why most studies find very little risk sharing in international data and a lot in regional data.
Does this finding suggest that there is no lack of international risk sharing? To explore the anatomy of this result further, we compare our international results to evidence obtained from regional data sets from Australia, Canada, Germany and Italy — countries for which consumer price data can be obtained at a regional level. We find that regions within countries achieve most of their risk sharing through quantity (income and credit flows), very much as the earlier literature has documented. Also, quantity flows between countries are small, again in line with virtually all of the extant literature. In this sense there is a clear lack of international consumption risk sharing. The reason why we still find a small coefficient when our version of the risk sharing regression is performed on international data, is that international inflation differentials covary strongly with the relative value of a country’s output. This channel, on the other hand, is virtually absent in regional data, presumably because the cross-regional dispersion of consumer prices is low.
Posted in :
