FORECASTING INTEREST-RATES AND YIELD SPREADS - THE INFORMATIONAL CONTENT OF IMPLIED FUTURES YIELDS AND BEST-FITTING FORWARD RATE MODELS

Forecasts of interest rates for different maturities are essential for forecasts of asset prices. The growth of derivatives markets coupled with the development of complex theories of the term structure of inte rest rates have provided forecasters with a rich array of variables fo r predicting interest rates and yield spreads. This paper extends prev ious work on forecasting future interest rates and yield spreads using market data for T-bills, T-Notes, and Treasury Bond spot and futures contracts. The information conveyed in technical models that use marke t data is also assessed, using a recent innovation in interest rate mo delling, the maximum smoothness approach. Forecasts from this model ar e compared with predicted yields and yield spreads derived from future s prices as well as with those of the random walk model. The results s how some evidence of market segmentation, with more arbitrage evident for nearby maturities. Market participants appear to show a greater de gree of consensus on short-term interest rates than on longer-term int erest rates. There is some indication that forecasts from the futures markets are marginally better than those provided by those of the maxi mum-smoothness approach, consistent with the informational advantages of futures markets. Finally, futures and maximum-smoothness market for ecasts are shown to outperform those of the random walk model. (C) 199 7 by John Wiley & Sons, Ltd.