The policy content of the yield curve slope.

I. Introduction

Recent studies (Bernanke, 1990; Estralla and Hardouvelis, 1991; Stock and Watson, 1993; Bernanke and Blinder, 1992) have shown that changes in the slope of the yield curve contain information about future real economic activity. Other studies (Mishkin, 1990; Fama, 1990) have shown that changes in the slope of the yield curve contain information about future inflation. Stock and Watson (1989) suggest including the yield curve slope in the index of leading indicators. Although these studies all indicate that there is information contained in the yield curve slope which helps predict output growth and inflation, it does not necessarily follow that this information is consistently reliable (in a statistical sense) or that it can be exploited by policymakers for the purpose of managing aggregate economic activity. This paper addresses three important questions pertaining to the usefulness of the yield curve slope as a predictor of real output growth and inflation. First, has the predictive ability of the yield curve slope remained constant over time? Second, does the yield curve slope provide information over and above what is already contained in monetary policy indicators? Third, does the predictive ability of the yield curve slope vary systematically with changes in monetary policy regimes?

Others have looked at these questions. Bernanke (1990) found that the yield curve slope loses much of its predictive power when other interest rates and interest rate spreads are added to a forecasting regression for real output. He also found that in samples which included data beyond 1980, interest rates and interest rate spreads lost much of their predictive power. Estralla and Hardouvelis (1991), on the other hand show that the slope of the yield curve retains its predictive power when the federal funds rate (their proxy for the stance of monetary policy) is added to the vector autoregression thus concluding that the yield curve slope contains independent information on the future course of the economy.

Theoretically, it is equally plausible that movements in monetary policy are exogenous with respect to the yield curve slope or movements in the yield curve slope are exogenous with respect to monetary policy. What complicates this issue further is that monthly innovations to the yield curve slope and monetary policy indicators (such as the federal funds rate) are contemporaneously correlated. Thus, it is not possible to identify the independent information contained in the yield curve slope without imposing contemporaneous restrictions on the residual correlations between yield curve innovations and monetary policy innovations. To impose such restrictions would of course defeat the purpose of this exercise since it is equivalent to assuming that one of these variables is exogenous with respect to the other.(1) Thus, standard econometric identification techniques cannot be employed to determine which variable represents structural innovations: the federal funds rate or the slope of the yield curve.

To avoid this problem altogether I use a different method of identification. Namely, I make use of the fact that previous studies have identified 3 distinct monetary policy regimes during the post-Treasury-Fed accord period: 1955.01-1979.09, 1979.10-1982.09 and 1982.10-present. I estimate a Federal Reserve reaction function in the spirit of Bernanke and Blinder (1992) and investigate whether the ability of the yield curve to predict output growth and inflation is systematically related to the Fed's reaction function over these three distinct monetary policy regimes. By using information from these previously identified monetary policy regimes I am able to, in a sense, hold the effects of changing policy regimes constant and measure whether the information content of the yield curve varies across regimes.

The empirical results reported here suggest that when the Federal Reserve sets policy to counter movements in a particular variable (either inflation or output growth) then the slope of the yield curve loses its predictive power for that variable. For example, the Federal Reserve appeared to lean against monthly movements in inflation before October 1979 and ignore them afterwards. The yield curve slope contains no independent information about the future course of inflation before October 1979 but does afterwards. These results suggest that the slope of the yield curve contains independent information about a variable only when the Fed is not reacting to that variable. Thus, it appears that monetary policy is exogenous with respect to the yield curve.

This finding is also very similar to the results obtained by Mankiw and Miron (1986) in a study of the predictive quality of the yield curve for interest rates. They found that prior to the founding of the Fed the yield curve slope predicted movements in interest rates and after the founding of the Fed it did not. They attribute the difference in predictive power to the Fed's policy of smoothing interest rates.(2)

This paper is organized as follows. The empirical methodology and results are presented in Sections II and III. Section II presents a series of Granger Causality test results over various monetary regimes. Section III presents results from a Federal Reserve reaction function and relates those results to changes in the information content of the yield curve slope. Section IV summarizes and concludes the paper.

II. Empirical …