RDP 2007-10: Trade Costs and Some Puzzles in International Macroeconomics 6. What Aspects of the Puzzles Do the Trade Costs Explain?

The above results suggest that it is possible to find some evidence that trade costs are associated with correlations that have been linked with three puzzles in international macroeconomics. Important issues are: (a) whether trade costs appear to explain much of these correlations; and (b) whether there are only certain aspects of the correlations that trade costs explain.

The first issue can be addressed by comparing the coefficient on the main independent variable in the models with and without trade-cost controls. The correlation (or more precisely the regression coefficient) between saving and investment in the data (controlling only for fixed effects) is 0.44, but the results suggest that after also controlling for trade costs, this correlation falls to 0.19. This suggests that trade costs can account for about half of the puzzle. For the real exchange rate, the estimates suggest that accounting for trade costs can reduce the correlation with the lagged real exchange rate from about 0.9 to 0.8. The coefficient estimates imply that if trade costs were to fall to zero, the half-life of shocks to the real exchange rate would fall from about five to three years. So, for the Feldstein-Horioka and PPP puzzles, trade costs appear to explain around half of the puzzles. Trade costs appear to explain little of the consumption correlation puzzle with the preferred specification suggesting that accounting for trade costs has no effect on the correlation between a country's relative consumption and its relative output, with the correlation remaining stable at about 0.7 for the exchange rate measure.

Table 12 provides information for addressing whether there are only certain aspects of the correlations that trade costs explain by summarising results for different controls. It can be seen that while trade costs play a role in some specifications, they play a minimal role in others. For example, in cross-sectional regressions of investment on saving (similar to the original specification of Feldstein-Horioka), including trade-cost controls increases the extent of the puzzle. This provides further evidence that trade costs are unlikely to be a complete explanation of the puzzles. However, to the extent that theory suggests that other factors are likely to contribute to the correlation, this may not be surprising. For example, cross-country differences in financial systems and developments may explain why countries with high saving rates may also have high rates of investment.