Understanding the Backus-Smith puzzle: It's the (nominal) exchange rate, stupid

Backus, Kehoe and Kydland (BKK 1992) demonstrated that if international capital markets are complete, consumption growth correlations across countries should be higher than their corresponding output growth correlations. In stark contrast to the theory, however, in actual data the consumption growth correlation is lower than the output growth correlation. By assuming trade imperfections due to non-traded goods, Backus, D.K., Smith, G.W. [1993 Consumption and real exchange rates in dynamic economies with non-traded goods. Journal of International Economics 35(3-4), 297-316] showed that there is an additional impediment at work that can lower the consumption growth correlation. While their argument was successful in partially explaining the puzzlingly low cross country correlation of consumption growth rates, it contributed to generating another puzzle because the data forcefully show that consumption growth is negatively correlated with the real exchange rate, which is also a violation of the theory. Using data for OECD countries, we decompose real exchange rate growth into its nominal exchange rate growth and inflation differential components, and find that nominal exchange rate movements are the main source for the Backus-Smith puzzle. We demonstrate the robustness of this finding by examining sub-samples of the data, by allowing for imperfect risk sharing due to 'rule of thumb' consumers, and by examining intranational data across the U.S. states where the nominal exchange rate is fixed.