On Unmodeled Breaks in the Turn of the Year, Turn of the Month, and January Effects

Date01 November 2017
Published date01 November 2017
The Financial Review 52 (2017) 725–747
On Unmodeled Breaks in the Turn of the
Year, Turn of the Month, and January
Russell P. Robins
Tulane University
Geoffrey Peter Smith
Arizona State University
We produce convincing new evidence that the turn of the year (TOY), turn of the month
(TOM), and January effects are critically dependent on the sample period over which they are
estimated. The TOY effect is significant in the value-weight portfolio from 1962 to 1997. It
becomes insignificant in the medium-size portfolio after 1994 and in the equal-weight and
low-size portfolios after 1997. The TOM effect becomes insignificant in the value-weightand
high-size portfolios after 1978, in the equal-weight and medium-size portfolios after 1997, and
in the low-size portfolio after 1998. January effects are significant in some subperiods but not
Keywords: structural changes, turn of the year effect, turn of the month effect,January effect
JEL Classifications: G10, G14, G19
Corresponding author: Arizona State University,WP Carey School of Business, Department of Finance,
300 E Lemon St, Tempe, AZ 85287; Phone: (480) 965-8623;Fax: (480) 965-8539; E-mail: gps@asu.edu.
We gratefully acknowledgethe helpful comments and suggestions of the anonymous referees. Of course,
we are responsible for any remaining errors.
C2017 The Eastern Finance Association 725
726 R. P.Robins and G. P. Smith/The Financial Review 52 (2017) 725–747
1. Introduction
The well-known turn of the year (TOY) and turn of the month (TOM) effects
are longtime violations of the efficient-market hypothesis. First recorded by Wachtel
(1942) and reintroduced by Rozeff and Kinney (1976) and Keim (1983), the TOY
effect is an anomaly because up to half the annual return on stocks occurs in the
month of January alone. The TOM effect of Ariel (1987) is an anomaly because
mean stock returns are positive and significant at the beginning of the month and
indistinguishable from zero at the end of the month. Research on TOY/TOM effects
continues despite a longstanding expectation of their imminent demise.1
In light of efficient markets, we question whether TOY/TOM effects are persis-
tent economic phenomena or nothing more than Type I errors due to sample period
selection. If TOY/TOM effects are persistent economic phenomena, then their eco-
nomic and statistical significance are not contingent on the subperiod over which they
are estimated. If, on the other hand, TOY/TOMeffects are contingent on the subperiod
over which they are estimated, then the model used to estimate them is incomplete if
it does not account for the unmodeled breaks in the data generating process.
The purpose of this study is to test if TOY/TOM effects are contingent on
unmodeled breaks in the data generating process and to find the critical break dates
that delineate each subperiod. Todo so, we apply the multiple structural change model
of Bai and Perron (1998, 2003) (Bai/Perron) to TOY/TOM effect regressions over
approximately 90 years from 1926 to 2015–2016. Bai/Perron treats unknown break
dates in a time series regression as parameters to be estimated. Thus our approach
is not susceptible to data mining and Type I errors due to sample period selection.
Bai/Perron also allows for general forms of serial correlation and heteroskedasticity
in the errors, lagged dependent variables, trending regressors, as well as different
distributions for the errors and the regressors across segments. It is thus appropriate to
apply Bai/Perron to stock return data even if the data are nonnormal, heteroskedastic,
and autocorrelated.
The data we study are the daily returns on the CRSP value- and equal-weight
NYSE/Amex/NASDAQ market portfolios and on the high-, medium-, and low-size
stock portfolios from the Internet data library of Ken French. High-, medium-, and
low-size refer to portfolios built from the top 30%, middle 40%, and bottom 30%
of NYSE/Amex/NASDAQ stocks by market equity. We study the size-related stock
portfolios because Sikes (2014) and many others suggest that TOY/TOM effects
are chiefly size-related phenomena. We also add a self-financing low-minus-high
arbitrage portfolio by subtracting the high-size stock portfolio returns from the low-
size stock portfolio returns. We use all the available data so the sample periods are
from January 1, 1926 to December 31, 2015 for the market portfolios and from
July 1, 1926 to November 30, 2016 for the size-related stock portfolios and for the
1See, for example, Zhang and Jacobsen (2013), Lynch, Puckett and Yan (2014), Sharma and Narayan
(2014), and Urquhart and McGroarty (2014).

To continue reading

Request your trial

VLEX uses login cookies to provide you with a better browsing experience. If you click on 'Accept' or continue browsing this site we consider that you accept our cookie policy. ACCEPT