Review of Quantitative Finance and Accounting, 13 (1999): 137±151
# 1999 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.
Is the Term Premium a Risk Premium?
LOUIS H. EDERINGTON
University of Oklahoma, Michael F. Price College of Business Administration, Norman, OK 73019-0450
JEREMY C. GOH
University of South Florida, College of Business Administration, Tampa, FL 33620-5500
Drexel University, Lebow College of Business, Philadelphia, PA 19104
Abstract. This paper explores whether excess holding period returns on long vis-a-vis short-term securities
behave in a manner that is consistent with (1) market ef®ciency, (2) the time-varying-term-premium variant of the
expectations hypothesis, and (3) theories of the term premium that view it as a reward for risk bearing. Both
traditional and modern theories of the term premium imply that it should evolve fairly slowly over time as
attitudes toward risk and/or perceived covariances with wealth or consumption change. This implies that this
period's term premium should have some predictive ability for next period's. However, we ®nd that this quarter's
ex-post term premium has zero predictive ability. For monthly rates and returns, the evidence is less clear cut, but
again the implied term premia do not behave in a manner consistent with existing theories.
In seeming contradiction of the joint hypothesis of (1) market ef®ciency/rational
expectations and (2) the expectations model of the term structure, many studies (e.g.,
Shiller, Campbell, and Schoenholtz (1983), Fama (1984a) and Mankiw (1986)) have found
that excess holding period returns on long-term (vis-a-vis short-term) securities are
correlated with yield spreads observed at the beginning of the period. A common
explanationÐone which leaves the basic thrust of the joint hypothesis intactÐis that this
correlation is due to time varying term premia, e.g. Fama (1984a). In this paper we explore
whether the data are consistent with this explanation.
Both traditional and modern theories of the term premium (de®ned as the expected
excess holding period return) imply that, for short holding periods, this period's term
premium should have some predictive ability regarding next period's premium.
According to modern general equilibrium theories, for instance, the term premium will
only change as (1) market participants' levels of risk aversion change, or (2) the perceived
covariance of short and/or long rates with wealth or consumption changes. If these
parameters evolve reasonably slowly, as is usually assumed, next month's or quarter's
term premium should be partially predictable given knowledge of this period's. However,
we ®nd that this quarter's term premium has zero predictive ability. This implies that
either (1) term premia do not behave as implied by extant theories or (2) the joint
hypothesis of market ef®ciency and the expectations hypothesis doesn't hold. In contrast