On measuring the impact of controls: A comment

Abstract

After specifying and estimating an econometric wage-price model [Gordon, (1971)], I applied it in two papers [Gordon (1972), (1973)j to measure the impact of the Nixon administration control program in effect during 1971-73. The most striking conclusions were (1) that the price controls had held down the nonfarm price level by 3.2 percent by the end of the first two years of controls, and (2) that the wage controls had no direct effect on wages, since all the moderation of wages during the period was attributable to the indirect effect of lower prices. Far from claiming that this evidence demonstrated that the controls were a success (as Oi implies), I explicitly concluded that they had failed: The evidence that price controls had the effect of holding down inflation in the 1971-73 period should not be misinterpreted as an indication that the controls ‘succeeded.’ The effect of controls worked not by moderating the behavior of wages relative to prices, but rather by squeezing profit margins sufficiently to hold the price level below its freemarket value. This is not a situation which would be expected to last indefinitely, and hence the very fact of short-run ‘success’ for the control program guarantees its long-run failure....If...profit margins are eventually going to return to their no-controls level, there will be a catch-up period after the controls are lifted during which the rate of inflation will be substantially faster than would have occurred during that period without the controls. [Gordon (1973), pp. 777-81 Walter Oi deserves praise for the clarity of his careful review and critique of my model and its application to the controls issue. His strongly stated negative conclusion regarding the validity of my results is based on two major criticisms. First, the underlying theoretical structure of the model is unacceptable to him, because while “the forces that fundamentally drive the inflation process are our monetary and fiscal policies,” these policies, he claims, are not included among the set of explanatory variables in the model. Second, even if the theoretical rationale of the model were to be accepted, the instability of crucial parameter estimates during the sample period prohibits extrapolations of the estimated model into the post-sample period either to measure the effects of controls or for any other purpose.