A re-examination of the rational expectations hypothesis of the term structure: reconciling the evidence from long-run and short-run tests

Abstract

This paper explains the contradictory findings from long-run and short-run tests of the rational expectations hypothesis of the term structure. Recent research suggests that, while the long-run tests support the theory, the short-run tests do not. Our results which are based on US Treasury bills rates confirm this and suggest that the conflict in the evidence can be explained by the presence of a stationary, time-varying, term premium. Nevertheless, 'news' about future changes in short rates are far more important than the risk premium in explaining interest rate movements at the short end of the term structure, and comprise the main source of the single factor explanation that emerges from our results. © 1998 John Wiley & Sons, Ltd.KEY WORDS: term structure; rational expectations; cointegration; VAR; GMM SUMMARYThis paper attempts to determine whether the apparently contradictory findings from long-run and short-run tests of the rational expectations hypothesis of the term structure (REHTS) can be reconciled by allowing for the presence of a risk premium.The REHTS states that long-run interest rates reflect rational expectations about future short rates. Recent research on U.S. data has found that while there is general support for the REHTS in long-run based on the evidence from cointegration analysis there is much less support in the short run. Two main explanations of the failure of the short-run tests have been advanced. The first is that long rates not only contain information about future short rates, but also about risk (or term) premia. The second is the over-reaction hypothesis, that long rates over react to expected future short rates.The central assumption in our argument is that the risk premia are stationary. It then follows that if returns are non-stationary, cointegration analysis on the returns will yield super-consistent estimates of their long-run relations even if the risk premia are omitted. In contrast, models based on the predictive power of the term spread for short-run changes in returns will involve stationary variables and hence their estimates will be subject to omitted variables bias as a consequence of ignoring a risk premium. Thus, if the risk premia are stationary, the long-run evidence should provide a more reliable basis for testing the REHTS than short-run evidence.The long-run restrictions of the REHTS are tested using the data of McCulloch and Kwon (1993) which are annualized and continuously compounded monthly estimates of the yields to maturity for zero-coupon US government bonds for the period 1946.12 -1991.02. We carry out a cointegration analysis based on three returns (the yield to maturity and the holding period return on an n-period bond and the return on a one period bond). We find that the restrictions predicted by the REHTS are satisfied, implying that both the term spreads for different maturities and the term premia are stationary.If the REHTS holds then the short-run evidence can be used to provide more information about the term premia. We use a three variable V...

The Expectations Hypothesis of the Term Structure : A Look at the Polish Interbank Market

“……”

Central Bank of Ireland

“…The authors also find that surprises in one period excess returns are due to news about future short rates and not due to revisions about future term premia. These results are supported by recent evidence on US Treasury bills by Tzavalis and Wickens (1998), which shows evidence of a TVTP.…”

“…Tzavalis and Wickens (1998) also find similar results from their variance decomposition of the excess holding period return, using US treasury bills.…”

“…As a comparison to previous studies, namely Tzavalis and Wickens (1998) and Cuthbertson and Nitzsche (1998), we also compare the time series behaviour of the unexpected return eh(n,t+1) = h(n,t+1) -E t h(n,t+1)…”

How Did We Get to Where We Are Now? Reflections on 50 Years of Macroeconomic and Financial Econometrics

“…Whereas Campbell and Shiller () found that long rates under‐reacted to short rates, Campbell and Shiller () and Hardouvelis () found that long rates over‐reacted to short rates. However, Tzavalis and Wickens (, ) found that, once allowance is made for a term premium, long rates react to short rates as theory predicts. More generally, it is necessary to explain how short rates combine to give the three factors.…”