Testing saving and investment rates to understand capital mobility and current account solvency

Abstract

Feldstein and Horioka (1980) motivated the international finance literature by claiming a least squares regression of domestic investment rates on domestic savings rates is an informative measure of capital mobility. Their method stirred up controversy when they interpreted a high correlation between savings and investment rates as evidence of capital immobility, creating the famous Feldstein-Horioka puzzle. Current research starts with the Feldstein-Horioka result and shifts focus toward measuring short and long-run adjustments to external imbalances. The literature has implemented dynamic time-series and panel estimators to test the relationship. Following recent literature, each chapter in this dissertation jointly focuses on the adjustment process of current account imbalances and the conditions required for capital mobility.

The intent of this study is to show through the use of new estimation techniques previous results have been largely misguided. The starting point for this analysis is a thorough review of three key equations used in saving-investment regressions. The three models in question are an ordinary least squares model, error correction model, and an autoregressive distributive lag estimator. Each model is tested for stability, and it is found that a number of countries have an unstable relationship. One argument for the instability results is the presence of structural breaks. Previous literature has found that both variables follow non-stationary processes, but when using more powerful unit root tests and controlling for level shifts, both variables appear stationary. If each variable is stationary then previous methods assuming non-stationarity will produce incorrect inferences. Each series is optimally estimated for structural breaks, and through a mean differencing process the savings-investment coefficient is significantly reduced. Additionally, removing the exogenous breaks and using the lower frequency components allows for modeling the short-run current account adjustment process. Finally, the results are extended to measure the relationship in a panel framework using dynamic panel estimators and threshold effects. After controlling for structural breaks the coefficient decreases and exhibits a downward trend. The remaining correlation can be explained through trade openness and country size measures.