Beating a Random Walk
45 Pages Posted: 21 Sep 2017 Last revised: 8 Aug 2018
Date Written: August 7, 2018
As a crucial input to many valuation models, earnings forecasts are important to many practitioners and academics. Unfortunately, there is a large sample of firms that analysts do not cover, and analysts’ earnings forecasts are less accurate than a random walk at long horizons. Recent work by Hou, van Dijk, and Zhang (2012) and Li and Mohanram (2014) suggested the use of cross-sectional models to produce earnings forecasts. Several studies immediately used these models because of the obvious advantage that forecasts can be formed for a sample that is much greater than the sample of firms covered by analysts. Unfortunately, these models also produce earnings forecasts significantly worse than random walk forecasts. We present a simple and intuitive modification to these models – the use of quantile rather than OLS regressions in the prediction model – that produces earnings forecasts significantly better than a random walk. Subsequent analysis suggests that this simple modification produces earnings forecasts that lead to more accurate return forecasts, and better represents market expectations.
Keywords: earnings, forecasts
JEL Classification: G14
Suggested Citation: Suggested Citation