Working on a unbalanced sample of OECD countries spanning the period 1970-2003, this paper contributes to the empirical literature on the Feldstein-Horioka puzzle by making three main innovations: First, it goes beyond the traditional national-level investment-saving equations to estimate, for the first time, regressions at the institutional sector level (households, corporations, and government). Second, it explores the implications of giving separate consideration to current account deficits and surpluses. Lastly, it uses advanced panel data techniques to deal with endogeneity and to distinguish long- and short-run effects. After discarding the influence of common factors, the conclusions are that: (i) The national Feldstein-Horioka coefficient is in the vicinity of 0.5, but sectoral coefficients are much lower and even insignificantly different from zero; (ii) Such positive and significant national coefficient would not reflect frictions in international credit markets but just a fiscal current account targeting policy; (iii) Nevertheless, when the sample is split into deficit and surplus years, a higher and significant correlation emerges for the former at the national, household, and corporate level, implying that credit imperfections still play a role for the private but not for the public sector. Equally noteworthy, household correlation is still positive and significant, yet lower, for surplus years; and (iv) Against the background of a unitary long-run coefficient to satisfy the intertemporal budget constraint, the long-run relationship is 0.75 for national data, 0.6 for the corporate sector, and marginally or non-significant at the household and government level.
