𝔖 Bobbio Scriptorium
✦   LIBER   ✦

Predicting long-term earnings growth: Comparisons of expected return models, submartingales and value line analysts

✍ Scribed by M. S. Rozeff


Publisher
John Wiley and Sons
Year
1983
Tongue
English
Weight
744 KB
Volume
2
Category
Article
ISSN
0277-6693

No coin nor oath required. For personal study only.

✦ Synopsis


This paper derives four-five year predictions of growth rates of accounting earnings per share implicit in four expected return models commonly used in financial research. A comparison of such growth rates with those produced and reported by Value Line analysts and those generated by a submartingale model revealed the following: two expected return models-the Sharpe-Lintner-Mossin model and the Black modelwere significantly more accurate than the submartingale model, though not significantly more accurate than the other return models. However, the growth rate forecasts provided by Value Line significantly outperformed all the other models tested-none of which relied on the direct input of a security analyst.

KEY WORDS Forecasting Earnings growth Comparisons Empirical study Analysts Value Line

An extensive body of literature evaluates the short-run (less than 15 months) earnings forecasts of security analysts and time-series models.' The importance of this subject to accounting and finance is that a variety of applications such as firm valuation, cost of capital, and event studies require the measurement of earnings expectations. However, except for a recent paper by Moyer et al. (1983), little work has been done to this point in studying long-run earnings forecasts.

Moreover, a potential source of earnings forecasts-expected return models-has been overlooked. This paper evaluates the accuracy of long-term forecasts of growth rates of annual earnings per share. Six sources of forecasts are used: a submartingale model, the Value Line Investment Survey, and four expected return models. Each expected return model is combined with the Gordon-Shapiro constant growth model. Further, certain expected return models use the beta coefficient and, as such, lend insight into the usefulness of beta in a forecasting context. The paper comprises three sections. Section 1 describes the six forecasting sources and states the