http://clevelandfed.org/research/workpaper
Best available copy ∙∣
(between one-fifth and one-tenth) the magnitude of β2 for the SIZE port-
folios. A look at the other market-value quintiles shows that β3 is roughly
the same for MV and SIZE portfolios in each specific market-value quintile,
while β2 tends to be higher (lower) for the SIZE portfolios than the MV
portfolios in the low (high) market-value quintiles. Therefore, the failure
of the cross-equation equality restriction for β3 cannot account for the
disappearance of the size-related shift in β during the post-yearend period
once price is accounted for.12
VI. Conclusion
Recorded-price errors are potential sources of misspecification in joint tests
of the CAPM and market efficiency. We show that if the recorded-price errors
are sufficiently nonrandom, they can lead to biases in returns and in the
estimated coefficients of the market model. From this standpoint this paper
is an extension of the work of 81ume and Starnbaugh.
The second contribution of this paper is that it provides an explanation
of the TOY effect that is consistent with both the CAPM and market effi-
ciency. We find that the TOY effect is a price-related effect and that size
appears to be proxying for price during the TOY period. We propose and test
the LPSH, which argues that the TOY effect is due to nonrandomness in
recorded-price errors induced by tax-related flow-supply pressures at the end
of the calendar year. Tests of both raw returns and regression coefficients
from the market model fail to reject recorded-price errors as the source of
the TOY effect. This errors-in-variables explanation for the anomalous be-
havior of stock returns during the TOY period is consistent with both the CAPM
and market efficiency.