A decade ago, the demand for money was one of the least controversial topics in macroeconomics, both in its underlying theory and in the stability and plausibility of empirical coefficient estimates. While the theory of the long-run demand for money remains essentially intact, a cloud of uncertainty now hangs over the the short-run demand for money. This originates partly in the 1976 study by Goldfeld. He found, for the United States, too little money and too much velocity in the mid-1970s. The more recent puzzle of too much money and too little velocity in 1981-83 has only added to the uncertainty.

In a recent CEPR Discussion Paper, Research Fellow Robert Gordon suggests a new interpretation of the short-run demand for money in the United States. He emphasizes the interrelationships among four variables that enter the standard money demand function - the nominal money supply, real output, the price level, and the interest rate. These relationships include the short-run Phillips curve, explaining price changes as depending on output and (implicitly) past changes in money; a money supply function which relates the money supply to the short-run monetary base, interest rates, reserve requirements, and the discount rate; and the central bank's reaction function which relates the monetary base to the determinants of money demand. Gordon also considers an equation relating the rate of change of money to past monetary changes and unemployment, which some authors use to proxy the concept of 'anticipated monetary change'.

Gordon argues that the existence of these 'other' relationships means that previous short-run money demand functions may be better viewed as 'interesting reduced forms' rather than as characterizing a structure derived from the theory of portfolio behaviour. Shifts in their coefficients may reflect not changes in portfolio behaviour but rather movements of variables in the 'other' equations that are incorrectly omitted from the equation explaining real balances (e.g., supply shocks and price controls in the Phillips curve equation); instability in the coefficients in the 'other' equations; or a shift in control regimes by the central bank.

Both Laidler and Gordon have suggested that the short-run money demand function may be partly a Phillips curve in disguise. Sluggish adjustment of real balances may reflect inertia in aggregate price adjustment, as well as inertia in portfolio adjustment. Some of the post-1973 instability in the short-run money demand function may be a side effect of shifts in the Phillips curve that occurred as a result of supply shocks in 1973-75.

The recognition of inertia in the inflation process leads Gordon to doubt that a short-run structural demand for money function can be identified. The usual function explains real balances as depending on output, interest rates and lagged real balances. But if prices are sticky, then the short-run adjustment to changing output and interest rates must be achieved by the nominal money supply. If the central bank, in an attempt to stabilize interest rates, allows the money supply to respond instantly and fully to changes in output and interest rates, then these passive shifts in the money supply function will indeed trace out the short-run money demand function. But if the central bank abandons interest rate stabilization and instead targets monetary growth, then roles are reversed. Output and interest rates become endogenous variables responding to money. There is widespread agreement that over time the Federal Reserve shifted its emphasis from interest rate stabilization to monetary targeting. If so, then coefficients in conventional estimates on the 'demand for money' may actually represent a shifting mixture of demand and supply responses!

Gordon's paper also provides new estimates of money demand equations for the United States. He analyses a variety of dynamic adjustment processes, including the 'error-correction' model advocated by Hendry, Davidson and their collaborators. Each model is subjected to simulations over the decade since 1973 and the period following the shift in monetary control regimes in late 1979. The specifications he considers exhibit greater structural stability after 1973 than the standard partial adjustment model. Shifts in coefficients as the sample period is extended beyond 1973 are consistent, Gordon argues, with the interpretation that the real balance equation no longer traces out structural demand parameters, but rather a mixture of demand and supply responses.