Errors in recorded security prices and the turn-of-the year effect

http://clevelandfed.org/research/workpaper
Best available copy

related effects. To test for size- and price-related slope seasonality, we
test cross-equation equality restrictions on the βs in the regression of
equation <5) for the MV, PR, SIZE, PRICE, and MVPR portfolios. The test
results appear in table 6.

As seen in tables 3, 5, and 6, there are significant size-related effects
in the estimates of β₂ for the MV portfolios, although β₂ does not exhibit
a significant size-related effect for the SIZE portfolios. The rejection of
the cross-equation equality restriction on β₂ for the MV portfolios, com-
bined with the inability to reject the cross-equation equality restriction on
β₂ for the SIZE portfolios, is evidence that the size-related effect in 82
for the MV portfolios is actually a price-related effect. In contrast, look-
ing at tables 4 through 6, we see a significant price-related effect in the
estimates of β₂ for the PR and PRICE portfolios. The failure to reject the
cross-equation equality restriction on β₂ for the SIZE portfolios while
rejecting it for the MV, PR, PRICE, and MVPR portfolios is evidence that the
factor-related slope seasonality is an effect related to price but not
size. " This is consistent with the LPSH.

One could argue that we are overstating the significance of the tests of
the cross-equation equality restrictions for β₂ because β₂ is the shift in
β during the post-yearend period and we reject the cross-equation equality
restriction for β₃ (which is our estimate of the market model β exclusive of
the TOY shifts). This may indicate that the relative and not the absolute
shifts in β₂ are important in determining whether size or price i s driving
the seasonality in β. However, closer inspection of the results in tables 3

and 5 indicates that this is not a problem. The coefficient for β₃ for the

lowest (highest)  market-value quintile of the MV portfolios is close to that

of β₃ for SIZEl (SIZES). On the other hand, the coefficient for β₂ for

the lowest (highest) market-value quintile of the MV portfolios is twice

<<   12   13   14   15   16   17   18   >>

More intriguing information