The Feldstein–Horioka puzzle in South American countries:a time-varying approach

Abstract

Inasmuch as the investment-savings correlation is related to capital immobility and each country has specific and time-varying capital account liberalization and financial deregulation processes, we might not expect a constant and identical correlation for a group of countries. We performed both a constant (OLS) and a time-varying (Kalman filter) estimation of the investment-savings coefficient for Argentina, Brazil and Chile. As expected, the time-varying model presented a better fit, but our findings suggested that the investment-savings correlation is a misleading measure of capital mobility.

Notes

¹ See Frankel (1986), Obstfeld (1986), Dooley et al. (1987), Bayoumi (1990), Wong (1990), Tesar (1991), Vamvakidis and Wacziarg (1998) and Sinha and Sinha (2004).

² Results are available from the authors upon request.

³ We also modelled the state variable as an AR(2) process. However, in all three cases, the results did not change significantly, while the AR(1) model presented the smaller mean squared error. For those reasons, we do not report the AR(2) estimates.

⁴ Repayment of the external debt requires a decrease in domestic absorption, i.e. in private consumption (C) and/or in public consumption (G) and/or in domestic investment (I). Since domestic savings rate = (Y − C − G)/Y = (I/Y) + [(X − M)]/Y, where (X − M) = net exports, a decrease in domestic absorption tends to lead to a decrease in the investment-savings coefficient.

⁵ In an OECD study, Hussein (1998) also found some countries with coefficients greater than one.