The Stability of the Demand for Money: Evidence from the Post-1973 Period: A Reply
回应货币需求函数是否在1973年后发生偏移的争论，通过使用永久GNP替代实测GNP来估计货币需求方程，发现预测误差更小，并探讨了不同估计技术的影响。
Recent studies by Enzler, Johnson, and Paulus (1976), and Goldfeld (1976) have offered extensive evi. dence consistent with the claim that the demand for money function has shifted. In particular, both of these studies find that specifications of the demand for money equation, when fitted beyond 1973 consistently overpredict the actual demand for money and that the forecast errors are large. However, a recent paper by Hamburger (1977), using an alternative specification of the demand for money, provides evidence that shows improvement over the results given by Goldfeld. Clearly, the resolution of the question of stability has important implications for the appropriate conduct of monetary policy. The purpose of this paper is to show that the conventional demand for money equation which employs permanent GNP instead of measured GNP (as employed by Goldfeld) results in smaller forecast errors when the equation is extrapolated beyond 1973. It is also shown that while this result holds under a variety of parameter estimation techniques, demand equations estimated using more sophisticated techniques appear to perform better. Thus, the evidence in this paper has implications for the appropriate estimation as well as the proper specification of the demand for money. Since the main issue is the stability of the demand for the MI stock of money, the following analysis is restricted to estimating the money demand equation for MI only. There has been little controversy regarding the demand for the broader money stocks, M2 or M3. In section I, we specify our demand for money equation and estimate it using the familiar Cochrane-Orcutt (CORC) technique. We then extrapolate the estimated equation over the 1974:1 to 1979:4 post-sample period. In section II, we reestimate the demand for money using the estimation technique recently developed by Michio Hatanaka (1974) and again examine the postsample forecasts provided by this superior estimation procedure. The implications of these results and the major conclusions of the paper are discussed in section III. I.