Rules versus Discretion: A Comment

Abstract

TN A recent article, Professor Franco Modigliani reported on some tests of monetary policy, the results of which support the use of by the monetary authority over automatic rules.' It will be contended, however, that these results depend on a number of questionable assumptions on which the tests have been based. The design of Modigliani's tests is very simple. First, he defines an approximation to a target money supply, Al. Second, using data for the postwar period, he calculates series of deviations from target M, one resulting from actual monetary policy and others implied by a number of automatic rules. Finally, from a comparison of these error series, he draws conclusions about the hypothetical relative performance of actual monetary policy and automatic rules. The major difficulty is to obtain a good approximation to target M. Of course, Modigliani is well aware of this problem, as his frequent warnings to the unwary reader testify. As we shall see, however, he fails, at critical points, to recognize the degree to which his results depend on the particular approximation that he has chosen. Target M is defined as the stock of means of payment necessary to support full-employment real GNP, with price stability. To approximate target Al, it is assumed that actual velocity equals target velocity, that the percentage shortfall from potential output equals the percentage shortfall from full employment, and that price stability means no increase in the GNP deflator. It follows that the calculated error of actual policy equals the increase in the price level less the shortfall of actual employment from full employment, all measured in percentage terms. For example, if prices rose by 1 per cent and employment was 2 per cent below full employment, then actual M was approximately 1 per cent below target M. Similarly, the calculated error of any rule equals the percentage change indicated by the rule less the target change, where the target change is the actual change in M less the actual error in M. For example, if the rule indicates an increase of 3 per cent and the target change is 4 per cent, then the M resulting from the rule is approximately 1 per cent below target M. A number of for a number of definitions for money (and near money) are explored. For present purposes it is sufficient to deal with one rule and one definition for money. The definition for money is currency plus demand deposits adjusted. The rule is Modigliani's second rule, which requires that the increase in M have two components -one to adjust for any gap that might have existed in the preceding period between actual and potential output, the other to allow for growth in capacity assumed equal to the long-run average. The data are six-month observations beginning with the second half of 1947 and ending with the second half of 1962. This longer period is broken into four subperiods to account for different policy objectives at different times. Of these, only one period, 1952-I to 1960-I, can properly be considered relevant to the rules versus discretion test. The data yield the results shown in Table 1. Modigliani concludes that this evidence supports the use of over a rule. As was indicated above, a number of modifications in Modigliani's approximation to target M may lead to contrary empirical