The use of two-year-ahead forecasts, revisions, and breadth does not out-
perform the one-year-ahead forecasts, a result consistent with Guerard,
Gultekin, and Stone (1996).
Earnings forecasts themselves do not add value, a result consistent
with Elton, Gruber, and Gultekin (1981) and Blin, Bender, and Guerard
(1997). Earnings breadth produces higher turnover than analysts’ forecasts
or revisions, but generally enhances excess returns relative to the other
forms of earnings forecast variable—a conclusion supported in the regres-
sion results of Guerard, Gultekin, and Stone (1996) and in Chapter 8. The
proprietary, composite earnings forecast variable produces somewhat
higher turnover than the individual forecast variable, but much higher ex-
cess returns. In the simulations, we assume that a portfolio manager tightly
constrains portfolio security weighting and industry weights to be very
similar to the S&P 500 index. A composite earnings forecasting frame-
work similar to that in Wheeler (1995) substantially dominates the use of
individual forecast variables. The equally weighted proprietary growth
model produces 362 basis points of excess returns.
The results of this study are more consistent with those of Wheeler
(1994), Guerard and Stone (1992), and Chapter 8, in that analysts added
significant value, than with the results of the earlier studies of Cragg and
Malkiel (1968) and Elton, Gruber, and Gultekin (1981). Perhaps the value
of analysts was significant because we used a broader definition of earnings
forecasting and because earnings rose substantially during the 1982–1994
period. An annual regression of ranked achieved earnings growth on
ranked total returns produced positive and statistically significant coeffi-
cients on the achieved earnings variable in 12 of the 13 years. Earnings are
certainly a major determinant of stock prices—a result consistent with
Graham, Dodd, and Cottle (1962), Niederhoffer and Regan (1972), and
Elton, Gruber, and Gultekin (1981).
We have shown that earnings forecasts, breadth, and revisions en-
hance returns relative to using only historical, value-oriented data. Now
we address the question of using equally weighted or regression-
weighted composite models. The application of the Beaton-Tukey outlier-
adjustment procedure to equation (10.1) during the 1987–1996 period in
estimating equation (10.1), using the proprietary growth variable, pro-
duced the scaled regression coefficients, where the value variable weights
average approximately 65 percent during the period. The proprietary
growth variable weight approaches 0.50 during the 1990–1994 period
and averages 0.35, quite consistent with the Guerard (1990) and Guer-
ard, Takano, and Yamane (1993) estimations. The excess returns of the ro-
bust-weighted composite model are 635 basis points, exceeding the equally
backadmin
(backadmin)
#1