THE SUPPLY OF MONEY AND COMMON STOCK PRICES: FURTHER OBSERVATIONS ON THE ECONOMETRIC EVIDENCE

Abstract

RECENT, well-publicized attempts to employ econometric methods to isolate the determinants of common stock prices in the United States have aroused considerable interest in the investment community. Models developed by Keran [7], Hamburger and Kochin [4], and Homa and Jaffee [6], each of which assigns an important role to the linkage between the money supply and the level of common stock prices,' appear to have met with considerable success in explaining the behavior of Standard and Poor's Composite Index. The Hamburger-Kochin model, for example, explains a remarkable 99 per cent of the variance of this index, using quarterly data, during the 1956-1970 period. Homa and Jaffee, in addition, perform a series of simulation experiments which suggest that an investment strategy incorporating the forecasts generated by their model can outperform, in terms of both mean and variance of return, a simple buy-and-hold alternative. The purpose of this paper is to evaluate the potential contribution of the Keran, Hamburger-Kochin, and Homa-Jaffee models to the problem of forecasting the level of common stock prices. To this end, the models are reestimated using both Canadian and American data, and then subjected to a series of tests designed to measure their structural stability and sensitivity to possible specification error. This preliminary evaluation is followed by a more detailed analysis of their out-of-sample forecasting accuracy. In brief, both theoretical and empirical considerations suggest that the extraordinary success of these models in tracking the behavior of stock prices during the sample period may be illusory. The forecasting experiments documented in the text serve to confirm the suspicion. Ex post forecasts of Standard and Poor's Composite Index generated by the three models for the 1970:1 to 1972:2 period, for example, prove to be inferior to a naive no-change extrapolation. Comparable forecasts of the level of Canadian stock prices, measured by Statistics Canada's Investors Index, also fail to pass the most rudimentary tests of forecasting accuracy. Although the forecasting experiments cover a period of only ten quarters, one's confidence in their pessimistic conclusions is enhanced by the numerous limitations of these models analyzed earlier in the text. Finally, the fact that these models do not appear to have captured stable structural relationships suggests that one should not attach undue importance