Using Shrinkage Estimators to Improve Upon Time-Series Model Proxies for the Security Market's Expectation of Earnings

Abstract

Time-series models (Box and Jenkins [1976] and Nelson [1973]) of earnings expectations have frequently been used in capital market research to produce proxies for the security market's assessment of earnings surprise surrounding corporate earnings announcements (see, e.g., Ball and Brown [1968], Beaver, Clarke, and Wright [1979], and Foster, Olsen, and Shevlin [1984]). These time-series models are used to proxy for the market's assessment of unexpected earnings in situations wherever current analysts' forecasts are not available (e.g., small firms and quarterly earnings). In a recent survey paper, Bathke and Lorek [1984] suggest that three parsimonious time-series models have been shown to be useful forecasting models of quarterly earnings per share, viz. those of (1) Foster [1977], (2) Griffin [1977] and Watts [1975], and (3) Brown and Rozeff [1979]. Using different samples of firms drawn from different time periods, three classes of seasonal autoregressive integrated moving average (SARIMA) models have been identified as being the best univariate time-series