An equation can do a good job of forecasting without affording us much insight into underlying causes. We can forecast tomorrow’s weather, for example, by assuming it will resemble today’s. But if we content ourselves with this model, we will miss many of the important causal elements behind weather, including the seasons.

A currency model may happen to predict future currency demand well, without explaining anything about the reason for this demand and, in fact, while ignoring many important causal elements. For example, demand deposits have an implicit yield measured by the value of services (such as check transferal and safekeeping) per unit time divided by the average amount on deposit. This yield has been rising over the past quarter century; one would thus expect that, other things equal, the demand for demand deposits would rise and the demand for currency would fall.

That Professor Garcia’s currency demand model has successfully predicted currency demand despite the fact that it leaves out this important causal element (among others) can be explained by two facts. One, both the amount of currency and the model’s two principal explanatory variables (GNP and interest rates) have been rising rapidly over the time period. Two, the model also includes as an explanatory variable the previous period’s currency amount; i.e., today will be sunny because yesterday was sunny. Thus its success gives us no assurance that it includes the important explanatory factors — including the subterranean economy.