A Model for Ex Ante Real Interest Rates and Derived Inflation Forecasts

Abstract

A methodology for estimating expected real interest rates and making inflation forecasts, together with appropriate confidence intervals, is presented. The role of price expectations in determining real interest rates is also analyzed. Structural breaks in the process generating real interest rates are detected in 1980 and 1983 by tracking the accuracy of interest-rate forecasts. Analysis of these structural breaks reveals that the break at the beginning of 1983 returns the process to its pre-1980 formulation. During 1953-1985, price forecasts derived from the interest-rate model are found to be unbiased once the structural breaks in the real rate process are handled through heteroscedastic corrections. Invoking the rational-expectations hypothesis, we construct a general yet parsimonious dynamic model for estimating economic agents' anticipations of the real rate of interest. By using quarterly data for 1953-1985 the model is efficiently estimated by two-step two-stage least squares. Results provide support for a random-walk process and for the existence of a negative price expectations (Tobin-Mundell) effect on real interest rates. The U.S. economy was subjected to a number of shocks in the early 1980s. A new chairman took the helm of the Federal Reserve System (Fed) in 1979, and almost immediately the Fed announced that it would change its operating procedure from controlling interest rates to targeting the money supply. The Fed then returned to its pre-1979 operating procedure in October 1982. Added to these changes were the tightening of credit controls in the first half of 1980, the advent of the large fiscal deficits of the Reagan administration, and a fall in petroleum prices. To some degree, all of these shocks probably contributed to the greater volatility of the economy in the first few years of the 1980s. In fact, some researchers have recently argued that there have been shifts in the process generating real interest rates in October 1979 and 1982 coincident with the Fed's changes in operating procedure. To investigate the existence and timing of any breaks in the early 1980s we constructed exact (small-sample) confidence intervals for the interest-rate forecasts of investors and tracked them with actual interest rates. We identified a break in the first quarter of 1980 when the actual interest rate no longer fell within the confidence interval for the forecast, and another break in the first quarter of 1983. Adapting the model to account for these structural shifts, we show that, although the 1980-1982 period is significantly different from the pre-1980 period (in terms of parameter values), the post-1982 period is not significantly different from the pre-1980 period. We also construct investors' ex ante forecasts of inflation from the interest-rate model and show them to be unbiased, whereas the inflation forecasts from the Livingston survey are biased. The mean squared prediction error of inflation forecasts improves significantly when price expectations are included as a determinant of ex ante real interest rates. We also find that the forecasting performance of this model compares favorably with the inflation forecasts from the American Statistical Association/National Bureau of Economic Research Business Outlook Surveys.

Key Words:

• Forecasting accuracy
• Judgmental surveys
• Rational-expectations hypothesis
• Structural breaks
• Two-step two-stage least squares