Stock Selection in Unscreened and Screened Universes
In the previous section we examined the financial characteristics of un-
screened and socially screened stocks, finding, as did Kurtz and DiBar-
tolomeo (1996), that larger, more value-oriented stocks are excluded by social
screening. Can a composite stock selection model, using value and growth
factors, be effective in selecting securities that outperform the market in a so-
cially screened universe? Let us propose to use a quantitative model for all se-
curities publicly traded on any exchange during the 1987–1996 period. The
model has seven variables: six value factors and the composite, proprietary
growth variable developed in Chapter 8. The six value factors are earnings-
to-price, book value-to-price, cash flow-to-price, sales-to-price, dividend
yield, and net current asset value. The earnings, book value, cash flow, and
sales variables are traditional fundamental variables examined in the invest-
ment literature, as discussed in Chapter 8. The traditional theory of value in-
vesting holds that securities with higher earnings, book value, cash flow, and
sales are preferred to those securities with lower ratios, respectively. The net
current asset value is the current assets of a firm less its total liabilities. A firm
is hypothesized to be undervalued when its net current asset value is less than
its stock price (Graham, Dodd, and Cottle 1962; Vu 1990).
Stone, Guerard, Gultekin, and Adams (2002) applied a four-factor risk
model using a response surface methodology and found no differences be-
tween portfolios constructed using a composite stock selection model in
socially screened and unscreened universes.
Financial economists have studied the effectiveness of consensus (mean
values of forecasts) for more than 30 years in the United States, producing a
huge literature exceeding 400 articles, summarized in Keon (1996). A con-
sensus has yet to develop as to whether analysts’ forecasts add value, that is,
create excess returns. It has been shown that analysts’ forecasts are generally
more accurate than time series models, but it has not been consistently shown
that the more accurate forecasts produce statistically significant excess re-
turns; see Brown (1993) for an excellent survey of the literature on earnings
forecasting. In this study we analyze three possible sources of excess returns
from analysts’ forecasts: (1) the forecasts themselves, (2) the changes in the
mean values of earnings forecasts relative to the stock price, and (3) the
breadth of the forecasts, where breadth is defined to be the monthly net num-
ber of analysts raising the forecast divided by the total number of forecasts. It
is possible that the forecasts themselves may not produce excess returns; that
is, simply buying securities forecasted to have the highest growth in earnings
for the current fiscal year (FY1) or next fiscal year (FY2) may not add value.