http://clevelandfed.org/research/workpaper
Best available copy
12
As seen in tables 3 through 5, there appears to be significant seasonality
in the estimate of I3 during the beginning of the calendar year. The estimate
of the slope-dummy variable for the post-yearend period,^, is positive and
significant for all portfolios except for MVlO and PRICE5 wherezfh is posi-
tive and insignificant, MVPR5 wherezβ42 is negative and insignificant, and
PRlO WherezB42 is negative and significant. However, we find very little
evidence of pre-yearend slope seasonality. zβ4. is significantly different
from zero only for MVlO, PR9, PRlO, PRICES, MVPR5, and PRICEl. is nega-
tive and significant for the first four and positive and significant for
PRICEl. F-Tests for the equality of βf and B2 fail to reject the restric-
tion only for PRlO, PRICEI, and MVPRS. Because mean daily returns are higher
during both the pre- and post-yearend periods, rejecting the hypothesis that
B2 = O while failing to reject the hypothesis B, = O is inconsistent with
the hypothesis that increased systematic risk during the TOY is the source of
anomalous TOY returns.
The insignificance of Bt in the majority of the regressions is not
inconsistent with the LPSHls error-in-variables explanation for observed
increases in B during the TOY period. The insignificance of B, may indicate
that the majority of recorded-price decreases, on stocks that are tax-loss
selling candidates, have already occurred by the pre-yearend period. This
would reduce the degree of recorded-price bias in the portfolio returns and
the market proxy return, and therefore the bias in β. In fact, Roll [231 pro-
vides evidence that the recorded prices of tax-loss selling candidates start
to readjust toward their true price on the last trading day of the year.
For the LPSH to be accepted, B2 should show a price-related bias. That
is, we should observe more slope seasonality for low-priced portfolios than
for high-priced portfolios. In addition, we should not observe size-related
slope seasonality in B2 in the SIZE portfolios where we control for price-