http://clevelandfed.org/research/workpaper α
Best available copy ~-'~
in the lowest market-value (price) qui ntile.8 The SIZE (PRICE) portfolios
are constructed to disentangle price (size) effects from size (price) effects
in the MV (PR) portfolios.
To investigate the presence of factor-related TOY premiums in the returns
of the portfolios, mean returns are computed for the MV, PR, SIZE, PRICE, and
MVPR portfolios over five subsample periods:
D the sample period = all but the last five observations in the sample;
2) the pre-yearend period = last five trading days of each calendar year;
3) the post-yearend period = first five trading days of each calendar year;
4) the TOY period = pre-yearend period + post-yearend period;
5) adjusted-year period = sample period - TOY period.
The last five observations are dropped when computing mean returns for each
subsample because they correspond to the pre-yearend period for 1982 and there
is no corresponding post-yearend period for 1983 in the sample. This partic-
ular partitioning of the sample is done for three reasons. First, the empiri-
cal evidence of Reinganum [22] and Keim [15] indicates that the bulk of the
TOY premium lies in the first five trading days of January. Second, Roll [23]
uses the 10 trading days centered on the end of the calendar year as the TOY
period. Finally, prior to the Tax Reform Act cf 1986, an installment-sale
option for capital gains was available to investors during the last five frad-
ing days of December.9
Table 1 indicates that there is a significant size- or price-related
effect in the returns of the MV and PR portfolios during the TOY and
post-yearend periods. We are unable to reject the hypothesis that the mean
returns are equal across size (price) deciles for the sample period, the
adjusted-year period, and the pre-yearend period for both the MV and PR
portfolios. Table 2 shows that once price is accounted for, the significant
size effect found during the TOY and the post-yearend periods disappears.