Unanticipated inflation and unanticipated money growth and short-run real output response : dynamic rational expectations models and empirical tests

The study uses U.S. quarterly data from 1956 through 1979. Subperiod analyses, from 1956-1967 and from 1968-1979, are also undertaken. Two indices of economic activity are used, real GMP and the employment rate. Two inflation variables, the CPI and the GNP deflator, and two nominal monetary aggregates. Ml and the monetary base, are used.

Two transmission mechanisms are invoked to explain how the unanticipated variables "cause" output: 1) for the inflation models, the Lucas Hypothesis, based upon relative/absolute price confusion and information lags, is used, and, 2) for the money models, the traditional Real Balance Effect, where unanticipated increases in nominal money lead directly to more spending and output, is used. While the inflation models can appeal directly to the Lucas Hypothesis, a judgement as to which mechanism best describes the money-to-output linkage is made inferentially by comparing the inflation and money regression results.

Expectations and expectation formation play a major role in this study. The questions addressed in this regard are: 1) how are expectations about inflation and money growth rates formed, and, 2) how do these variables introduce a dependence of present output activity on past realizations of these expectations. A chief assumption of this study is that expectations are formed "rationally" according to Muth.

The statistical methodology used here is divided into two parts:

1) the generation of the unobservable unanticipated inflation and money growth rate variables by Box-Jenkins univariate time series methods, and,

2) the use of these constructed variables as independent regressors in distributed lag equations specifically designed to examine the lag structure connecting present output activity with past (random) forecasting mistakes. The Box-Jenkins technique is appropriate here because it con forms in important respects to the basic precepts of the Muth hypothesis and because the estimated ARIMA models can be assumed to approximate the forecasting "rule" which mimics the rational forecasting behavior of economic agents. The ARIMA models employ a varying parameter technique so as to capture any structural changes in the true process generating the actual inflation and money series over the period. To further justify the use of the univariate forecasting filters, the residuals of the ARIMA inflation and money models are subjected to extensive cross-correlation tests to determine if any lead/lag covariation can be found; in all cases the hypo thesis of cross-filter residual dependency is rejected. The major findings of this study are as follows.

1. For the 1956-1979 period, unanticipated inflation or money growth is not neutral in the short-run, but is directly related to short-run out put response. This finding is substantiated using either real GNP or the employment rate as the index of real economic activity. While the results of the inflation regressions support the Lucas Hypothesis, they do not support the current claim that excessive "noise" in the price signaling mechanism over the period has resulted in a positively-sloped Phillips curve.

2. A second major finding of this study is that output response due to exogenous inflation and money shocks is time distributed, with the effects on output first rising and then falling in magnitude. This finding leads this study to conclude that the conventional (i.e., comparative static) Phillips curve is more properly specified when the statistical form by which it is estimated allows for lagged or "persistence" effects in out put response patterns.

3. This study finds that unanticipated money produces a longer-lasting and more pronounced effect on real output than unanticipated inflation. [The Ml/GNP and monetary base/GNP lags take thirteen and sixteen quarters, respectively, to terminate; the Ml/employment rate relationship requires seventeen quarters. Conversely, unanticipated CPI or deflator inflation affects GNP for only five quarters; for the employment rate, the lag is six quarters.] These results support the conclusion that the Real Balance Effect associated with unanticipated additions to nominal money growth have a more powerful influence on output than inflation-induced changes in perceived relative prices. A related conclusion is that unanticipated money is transmitted to output primarily through unanticipated increases in liquidity in the system, with little reliance on the Lucas price confusion linkage. 4. Over the two subperiods, 1956-1967 and 1968-1979, this study finds that real output is more responsive to unanticipated inflation and un anticipated money growth in the second period than in the first period. These results, therefore, do not support the suspicion that the inflation-unemployment or the money-unemployment "tradeoff" has all but disappeared in the U.S. in the mid- to late 1970's.