Forecasting the one-year-ahead raw returns of a firm by estimating the equilibrium correction of the market to its time varying perception of the long run relationship between market and accounting values

Time series models based upon the reported accounting data of Abbott Laboratories Inc. are used to forecast its one-year-ahead raw returns over hold-out periods of 1, 4 and 10 years. The sample data is annual, from 1955 to 2003. A general-to-specific, testing down, modelling method is used. The models suggest market values are corrected for imbalances between expected and actual market values in the preceding year. All well specified models are in logs, implying a multiplicative statistical relationship between market and accounting values. The highest ranking model in terms of forecast root-mean-square-error criteria supports capitalised, reported earnings as the best estimate of year-end market perceptions of long run market value. Using book value of net assets, instead of earnings, produces forecasts of raw returns no better than a random walk model, i.e. book value is not value relevant in this case.