International Capital Mobility and the Feldstein-Horioka Paradox: Evidence from Settler Economies in the Gold Standard Era
University of Adelaide Seminar Paper #97-06
Posted: 16 Feb 1998
Date Written: April 1997
Abstract
How do we know whether capital is more internationally mobile in some periods than in others or for some countries rather than others? Economists normally look at prices when examining market integration, but for capital markets it is difficult to test the law of one price with interest rates for identical assets in a common currency. To overcome the data constraint, Feldstein and Horioka (in Economic Journal, 1980) used quantity data to estimate the equation: I/Y = a + b. (S/Y) + e. If capital is immobile, investment must equal domestic saving, i.e., b=1, and if investment is unrelated to domestic saving, then b=0. Using data from OECD countries 1960-74, Feldstein and Horioka (FH) obtained values for b not significantly different from unity. The finding was seen as a paradox because capital markets were considered to be integrated in the 1960s.
Subsequent studies have generally confirmed the FH paradox. Values of b tended to be higher in the 1930s than before 1914 or after 1960, and they tended to be lower in time series studies but always significantly different from zero. The value of b has been interpreted as a measure of the degree of international capital mobility, bounded by zero and one. By contrast, FH tests for internal capital mobility (within U.S. states or U.K. regions) have found b=0, which has been interpreted as showing that the FH test is an appropriate measure of capital mobility.
This paper revisits the FH paradox by running the equation for three countries known to have been integrated into international capital markets: Argentina, Australia, and Canada in the pre-1914 gold standard era. Only Argentina passes the classic FH test of capital mobility, yet the other two countries clearly had mobile capital because a>0. The point is that b can take any value and be consistent with capital mobility; the highest estimated coefficient (b=2) is for Canada 1896-1914, where international capital mobility was amongst the highest ever.
The only absolute test possible with the FH equation is to test for capital immobility, with the null hypothesis of b=1 and a=0. If b differs significantly from unity, then capital must be mobile. Most studies since FH's original paper have indeed found b to differ significantly from unity and thus are evidence of capital mobility; there is no paradox. On the other hand, the value of b can give no clear guide to the degree of capital mobility. A value of b=1 is a necessary but not a sufficient condition for capital immobility. Capital can be mobile and b=1, or indeed b could equal any value. A value of b=0 is a sufficient, but not a necessary, condition for mobility. Using the range 0
JEL Classification: F32, F21, G15
Suggested Citation: Suggested Citation