Review of Quantitative Finance and Accounting, 11 (1998): 37–51
© 1998 Kluwer Academic Publishers, Boston. Manufactured in The Netherlands.
The Economic Signiﬁcance of the Cross-Sectional
Autoregressive Model: Further Analysis
School of Business, Hebrew University of Jerusalem
KOK CHEW LIM
Department of Economics and Finance, City University of Hong Kong, Tat Chee Ave, Kowloon, Hong Kong
Abstract. We reexamine whether investors can gain abnormal returns using the cross-sectional autoregressive
model of stock returns. We ﬁnd that the pattern of abnormal returns obtained is inconsistent over the time period
1934–94. We adjust for the higher commission costs in the pre-May 1 1975 period, a point overlooked in
Jegadeesh (1990), by assuming a conservative one-way transaction cost of 0.75%. For the post-May 1 1975
period, we use a one-way transaction cost of 0.25%. The results show that investors who invest only on the long
side would earn insigniﬁcant ‘after-transaction cost’ abnormal returns in the post-World War II period, 1946–94.
The ‘after-transaction cost’ abnormal return from the short strategy is about 0.5% for the period 1946–94. This
article shows that an investor would not earn abnormal returns using this model considering that it is more costly
in practice to sell securities short and that most investors would not earn interest on short sale proceeds.
Key words: Economic signiﬁcance, autocorrelations in stock returns, transaction costs
Recent research has shown that stock prices do not follow random walks (see, for ex-
ample, Fama and French (1988), Conrad and Kaul (1989), Jegadeesh (1990), Lo and
MacKinlay (1988), McQueen and Thorley (1991), and Poterba and Summers (1988)).
Such evidence implies that stock returns are predictable.
Fama and French, and, Poterba
and Summers ﬁnd evidence of systematic reversals in stock returns over three- to eight-
year intervals. Conrad and Kaul propose a time-varying weekly expected return model,
and show that daily, weekly and monthly returns are positively autocorrelated. Using a
simple speciﬁcation test based on variance estimators, Lo and MacKinlay ﬁnd that weekly
holding-period returns exhibit positive serial correlation. McQueen and Thorley employ
Markov chains to analyze stock prices pattern. They ﬁnd that annual returns exhibit
signiﬁcant nonrandom walk behavior in the sense that low returns tend to follow runs of
high returns, and high returns tend to follow runs of low returns.
While the evidence of predictable behaviour in stock prices reported by Fama and
French, Conrad and Kaul, Lo and Mackinlay, Poterba and Summers, and McQueen and
Thorley does not clearly suggest that economically signiﬁcant abnormal returns can be
earned by investors, Jegadeesh, using one-way transaction cost of 0.25% of the principal
value, shows that the zero-investment strategy of the cross-sectional autoregressive model
@ats-ss9/data11/kluwer/journals/requ/v11n1art3 COMPOSED: 03/17/98 9:02 am. PG.POS. 1 SESSION: 15