Causality, Exogeneity, and Natural Rate Models: Reply to C. R. Nelson and B. T. McCallum

Abstract

First, in response to Nelson, I hope that the reader of my articles (Sargent 1976a, pp. 222-23; 1976b) will recognize that I did not assert that under general conditions the natural rate hypothesis is equivalent with the failure of the unemployment rate to be Granger caused by other variables. But it is untrue that there is no connection between the natural rate hypothesis and the Granger causality structure of blocks of and variables. Econometric evidence consistent with my main causality hypotheses (1976a) could be parsimoniously explained within the context of Lucas's formulation of the natural rate hypothesis, but would be very difficult to explain according to Keynesian macroeconomic models. This point of view is ably represented in McCallum's comment, so there is no need for me to expand upon it. The only analytical point on which Nelson takes issue with my papers is his assertion that . . exogeneity cannot be verified from nonexperimental data... Here his quarrel is with Sims (1972). Nelson's position on this matter is widely held but wrong. For Sims established that a necessary condition that x be exogenous with respect toy is thaty fails to Granger cause. In the first part of this reply, I identify what I believe is the element of truth in Nelson's positionnamely, that Sims's condition is not sufficient for x to be exogenous with respect to y-but indicate why Sims's test remains a valid test of the hypothesis of econometric exogeneity. In the second part, I take up some of the theoretical issues raised by McCallum. Here I undertake to characterize more extensively than I have done before the restrictions on equilibrium macroeconomic models necessary to deliver (block) exogeneity of real economic variables with respect to nominal variables.1