http://clevelandfed.org/research/workpaper
Best available copy
firm i at time t, w∣t, in the portfolio is 1/n for t = l and
(4) wlt = wit-ι(1 + rit-l - rpt-t), for t = 2,∙∙∙∙,∏.
This portfolio weighting scheme assumes that if a firm pays a dividend, the
full amount of the dividend is reinvested without cost into the stock of the
firm. However, this portfolio weight adjustment is more realistic than one
that rebalances the portfolio daily to an equally weighted portfolio. In addi-
tion, this approach avoids factor-related biases that can arise from rebalanc-
ing portfolios to equally weighted portfolios on a daily basis (see Roll [24]
and Blume and Stambaugh [3D.
V. An Investigation of the Low-Priced Security Hypothesis
To test whether the TOY effect is a size-related effect or a low-priced secur-
ity effect, the sample is grouped into 10 MV portfolios on the basis of the
market value of the firm and into 10 PR portfolios on the basis of share
price. The portfolios are numbered on the basis of market value (price); MVl
(PRD is made up of the firms in the lowest market-value (price) decile and
MVlO (PRlO) is constructed from the firms in the highest market-value (price)
decile. In addition, 15 portfolios are constructed on the basis of size and
price. The data is sorted twice, first into size quintiles and then into
price quintiles. Five SIZE (PRICE) portfolios are formed from firms that are
in each size (price) quintile but not in the corresponding price (size) quin-
tile. For example, SIZEl (PRICED comprises firms in the lowest market-value
(price) quintile that are not in the lowest price (market-value) quintile.
Five MVPR portfolios are formed from the firms that are excluded from the SIZE
and PRICE portfolios. For example, SIZEl (PRICED and MVPRl contain the firms