Dissecting Anomalies (Digest Summary)

A great deal of academic research has reported the empirical existence of stock return
patterns that are considered anomalies. The authors investigate the pervasiveness of
seven anomalies for company size (i.e., market capitalization) groupings. In addition to
the insights provided by including three size groups—microcaps, small stocks, and
big stocks—in their analysis of return anomalies, the authors find that microcaps
have the largest effect on the size effect anomaly.

The authors use monthly return data from June 1963 through December 2005 for companies
trading on the NYSE, Amex, and (after 1973) NASDAQ. Size groups are formed each year at
the end of June by using as breakpoints the 20th and 50th percentiles for the market
capitalization variable for stocks trading on the NYSE. The NYSE breakpoints are used to
form the three size groups. The average monthly sample size is 3,060 companies. Although
the companies in the microcap group are about 60 percent of all stocks in the sample,
their market capitalization is only 3 percent of the total market capitalization of the
sample. The companies in the big-stock group constitute more than 90 percent of the
total market capitalization of the sample. Therefore, the big-stock group dominates the
returns of value-weighted indices, and the microcap group dominates the returns of
equally weighted indices. The microcaps also are found to have the largest standard
deviation of returns (and for each of the anomaly variables), thereby indicating that
microcaps influence tests of market efficiency.

The authors use two approaches to investigate stock return anomalies. In the sorts
approach, the average returns are measured net of size and market-to-book effects and
are examined within quintiles of the anomaly variable. In accordance with the anomaly
literature, the return of hedge portfolios, which is the long versus the short position
of extreme quintiles, is also analyzed. In the second approach, the slopes of the
cross-sectional regressions provide information related to the marginal effect of each
anomaly variable.

Regression analysis and the univariate sorts methodology show the returns for the anomaly
variables for net stock issues, accruals, and momentum to be pervasive in the three size
groups. That is, returns of the hedge portfolios formed from the sorts are found to be
large and significant (for both the equal-weighted and the value-weighted portfolios)
for the variables that proxy net stock issues, accruals, and momentum.

Using regression analysis, the authors find a positive relationship between momentum and
returns for all size groups. The momentum anomaly, however, is found to be half as
strong for the microcap group. A negative relationship is found for both the net stock
issues and the accruals for all size groups, and the relationship is weakest for the
accrual variable in the big-stock group.

The negative relationship between asset growth and returns is stronger for the microcap
group than the small-stock group but is not significant for the big-stock group. The
positive relationship between profitability and returns is found only for the
small-stock group. The slope in the regression analysis for the book-to-market ratio is
significant for the small-stock and the microcap groups. Because regression slope
coefficients are significant for the model with all companies, the authors suggest that
asset growth, profitability, and the book-to-market ratio still provide some unique
information in all size groups.

The authors argue that all their anomaly variables proxy to some degree for expected cash
flows in the traditional valuation model. Consequently, they conjecture, evidence of
return anomalies does not distinguish between how much of the return variation is the
result of the pricing of risk and how much is the result of market inefficiency.