RDP 9003: The Balance of Payments in the 1980s Appendix: Feldstein-Horioka Tests

Feldstein and Horioka(1980) argue that in a world of perfect capital mobility, domestic savings and investment should be uncorrelated. The crux of the argument is that in the absence of capital flows domestic saving and investment are by definition equal. A rise in domestic investment opportunities will boost real interest rates to induce the extra saving needed to finance them. With perfectly mobile capital this need not be the case because a country then faces an elastic supply curve of funds from overseas. To test their proposition Feldstein and Horioka estimated;


I = gross national investment;
S = gross national saving; and
Y = gross domestic product.

They argued that if capital is perfectly mobile then β should equal zero. Their results showed that this was not the case and they therefore concluded that capital was not mobile.

It is now well known that a high correlation between domestic saving and investment does not necessarily imply limited capital mobility. Dooley, Frankel and Mathieson(1987) show that for saving and investment to be uncorrelated then investment must only depend on the domestic real rate of return, that the country be small so that the world real rate of return is exogenous and that capital be mobile so that the domestic and foreign real rate of return are equal. Violation of any one of these conditions will result in a correlation between domestic saving and investment. However, if each of these conditions hold then they argue that the correlation should be zero. This conclusion is also probably incorrect. This is because, in the limit, it implies that there is no mechanism that brings the current account into equilibrium or stablilises the net external debt to GDP ratio. For example, a correlation of zero implies that a country can persistently run extremely large current account deficits and indefinitely build-up external debt without this having any effect on expenditure patterns. That is, it depends on there being no effect on expenditure or policy from the diminution of net wealth caused by rising external debt. This does not seem to be a reasonable medium-term assumption. Once this channel is acknowleged then it is theoretically possible to have perfect capital mobility and still have a non-zero correlation between savings and investment. Nevertheless, estimating the Feldstein-Horioka equation remains useful in that lower (though non-zero) correlation remains consistent with the hypothesis of higher capital mobility.

In addition to these theoretical problems there are some econometric difficulties. In particular, domestic saving is endogenous which means that shocks that impinge on investment may also alter saving behaviour. The fact that saving and investment are procyclical is used as evidence of common influences. Edey and Britten-Jones(1990) show that this is undoubtedly the case for Australia. Unless the endogeneity of saving is accounted for in estimation the results may be biased toward showing a high degree of correlation.

Most subsequent studies have confirmed Feldstein and Horioka's original results. See Penati and Dooley(1984) and Dooley, Frankel and Mathieson(1987). However, Feldstein and Bacchetta(1989) and Frankel(1989) have found that the correlation between saving and investment has reduced over time and conclude that this is consistent with the move to more liberal financial markets.

Table 2 reports the results of testing equation 1 for Australia.

Instrumental Variable Regressions of the Investment Ratio (I/Y) against the National Savings Ratio (S/Y)
Inline Equation
Inline Equation Inline Equation Inline Equation DW
1963–1983 0.13
1963–1989 0.16

Footnotes *(**) denotes significantly different from zero at the five (one) per cent level.
All equations have been corrected for serial correlation.
Standard errors are in parentheses.
All data are annual.

The results in Table 2 show that the coefficient on domestic saving is significant in each period but falls as additional observations are added.[59] Thus when the period of reduced capital controls and the development of offshore $A markets is added to the original sample (1963–1983) it appears that the correlation between saving and investment declines. This is not inconsistent with the hypothesis that capital flows are now more mobile.


Following others in the literature the instruments chosen for domestic saving are the ratio of defence expenditure to GDP and the ratio of dependents (those 15 years or younger or 65 years or older) to the working age population. [59]