An investigation of expectation errors appears to refute the unbiased expectations hypothesis. The authors consider the source of these errors and the circumstances in which they arise.
The unbiased expectations hypothesis, presumably a neutral forecasting tool of future spot rates, has long been considered to be flawed. The authors’ contribution to the literature on this topic takes into account irrationality and pseudo peso bias as the sources of such errors. The former contributes to deviations from rational behavior during market crises, whereas the latter has merit primarily in periods of calm markets.
How Is This Research Useful to Practitioners?
The unbiased expectations hypothesis states that forward rates are unbiased predictors of future spot rates. But events over the past 20 years have led to an increased focus on the observation that markets often behave irrationally. Recent history has provided the basis for extensive academic research, which the authors contribute to by focusing on the source of expectation errors that call the hypothesis into question. At a high level, they attribute expectation errors to rational investors anticipating events that fail to occur during a period under investigation (what they call the “peso problem”) and when investors’ sentiment governs decision making and trading (irrationality).
The analysis reviews weekly Eurodollar futures over a 12-year period that encompasses several market disturbances. These data form the basis for the authors’ research. They test the unbiased expectations hypothesis of the term structure of US interest rates to see how and at what point it fails to explain some of its anomalies. They then propose a methodology to break down deviations from the hypothesis and analyze the two sources of expectation errors—namely, pseudo peso bias and irrationality—and offer an explanation for them. Their work differentiates itself from the current body of literature through its focus on a more robust and dynamic set of data, longer periods and subperiods over which to observe them, the disaggregation of expectation errors into components of irrationality and pseudo peso bias, an emphasis on the difference between market expectations and rational expectations, and the use of robustness tests to confirm the underlying causes of variations in expectation errors.
The authors’ research confirms a widely held view in academia that markets and investors often behave irrationally. Their conclusions would be of interest to students of behavioral finance, traders, analysts, and econometricians. A deeper understanding of the drivers of investor behavior could better inform trading and forecasting decisions, and it could help regulators have a well-informed perspective in their responses to market disruptions. Finally, market historians may find the authors’ work a good addition to the ongoing history of financial crises.
How Did the Authors Conduct This Research?
The authors’ work is a complex form of attribution analysis. They consider a broad dataset of weekly Eurodollar futures rates that extend from 3 months to 10 years during 1998–2010, a period that covers the seminal events of the internet bubble, the Great Recession, the Federal Reserve’s quantitative easing, and the introduction of and issues surrounding the European Economic and Monetary Union. Eurodollar futures offer current information on expectations of future spot rates, in contrast to survey data used in many previous analyses, and have extremely high trading volume.
The authors establish a forecast horizon to test the unbiased expectations hypothesis using these futures contracts. They observe greater predictive power of the forward premium to explain future changes in longer-term rather than shorter-term rates. But for subperiods that involve financial crises, the predictive power of the forward premium is weaker. To proxy for investor sentiment, the authors create a sentiment index, a multifactor model consisting of direct sentiment measures (e.g., American Association of Individual Investors sentiment survey data) and indirect sentiment measures (principal component analysis for several financial market variables that reflect trading activity). They discuss model specifications for estimating expectation errors that account for unexpected macroeconomic surprises, deviations from rational investor behavior, and a robustness check in the form of a lagged expectation error to determine the degree to which deviations from rationality persist.
The findings affirm the explanatory power of investor sentiment in subperiods as well as in longer periods but particularly during periods of financial unrest. In contrast, the pseudo peso problem plays a role in calmer periods.
The literature evaluating the merits of the unbiased expectations hypothesis is extensive. The authors make a novel contribution by investigating the sources of errors in the hypothesis, but their conclusions are familiar. Sentiment carries extraordinary explanatory power of investor behavior and is robust to different sample periods researched in this study. The notion of market participants behaving as rational actors seems to have been disproved again.