Gold and US money demand

Abstract

This letter is about the long run cointegration relation of the US money demand function that incorporates a gold price variable. A three-equation model is jointly constructed and estimated. The first equation has real gold prices, as a dependent variable, and real money, the real dollar index, a scale variable, and the lagged cointegration residual as independent variables. All the variables are in first-differences of the logs except the cointegration residual. The second equation is the cointegration regression with the same variables in log levels. And the third equation is a GARCH model of the conditional variance of residuals. Two different scale variables are chosen: the industrial production index and the real personal disposable income. Both variables produce close estimates. All coefficients are of the correct expected sign and are statistically different from zero. The evidence presented is highly supportive of the model. In particular we find long run money neutrality, and long run constant economies of scale for both scale variables. Moreover, both the short run and long run elasticities of the real dollar index are also unitary. Surprisingly real money and each one of the two scale variables, have no short run effects on the log of real gold prices, but have only long run effects. One can no more exclude gold from the US money demand without incurring a mis-specification. In this regard gold may be the missing variable that produces the structural breaks found in the literature.