An Examination of the Statistical Significance and Economic Implications of Model-Based and Analyst Earnings Forecasts
54 Pages Posted: 7 Jan 2013
Date Written: December 31, 2012
We address the demand for model-based earnings forecasts by proposing a cross-sectional model which incorporates three salient ideas. First, firm performance converges to expected levels over time; second, amounts from current financial statements are robust predictors of future performance; and third, ordinary least squares (OLS) estimation is unreliable in samples including extreme values. Accordingly, we estimate a cross-sectional earnings forecasting model based on least absolute deviations analysis (LAD), and include profitability drivers derived from financial statements as predictors. In terms of statistical significance, we find that these forecasts are more accurate than forecasts from three extant prediction models and consensus analysts’ forecasts. In terms of economic implications, we find that forecasts from our model have greater predictive ability for future abnormal returns than consensus analysts’ forecasts. Overall, our results are important because they document the usefulness of a cross-sectional earnings forecasting model for a broad range of diverse firms, including those with little or no analyst coverage.
Keywords: Earnings Forecasts, Financial Statement Analysis, Security Analysts
JEL Classification: M40, G11, G17
Suggested Citation: Suggested Citation