The savings-investment puzzle is one of the most important regularities in international finance, which originally identified by Feldstein and Horioka (1980). Feldstein and Horioka find that a cross-country regression of average investment rates on average savings rates results in a regression coefficient close to one. They argue that this regression coefficient should be close to zero in a frictionless open economy world and suggest that this large regression coefficient indicates a high degree of financial frictions in the world economy. Their finding and claims about substantial financial frictions in international financial markets raise a huge debate in the literature. The profession has responded to the Feldstein-Horioka finding from two different approaches.
First, empirical research attempts to refute the finding either by studying different data samples and periods, by adding other variables to the original OLS regression, or by using different estimation methods. However, it turns out that the high correlation between savings and investment is remarkably robust in both OECD and non-OECD countries, though the FH coefficient has tended to decline in recent years. In this paper, we confirm the Feldstein-Horioka finding with a more recent data set with a sample of 57 countries over the period from 1960 to 2000.
Second, theoretical research tries to refute the claims that the regression coefficient on savings is close to zero in a frictionless world, and thus the Feldstein-Horioka finding provides no information about financial frictions. However, most quantitative studies in the literature have focused on a variant of the original finding, that is, the high time-series correlation between savings and investment rates at business cycle frequencies within nations, instead of the cross-section Feldstein Horioka finding.
Baxter and Crucini (1993) and Finn (1990) show that the high time-series correlation between savings and investment rates at business cycle frequencies can arise naturally in a quantitative complete markets model (frictionless capital markets with a complete set of assets). Mendoza (1991) show that the most popular incomplete markets model, the bond model in which countries can only trade state-uncontingent bonds, can also quantitative produce the high correlation between savings and investment in time series. Thus, the high time-series correlation offers no insight about financial frictions.
Download
Can Financial Frictions Account for the Cross-Section Feldstein-Horioka Puzzle?
