A Behavioral Model of Earnings Forecasts: Top Down versus Bottom Up

Abstract

Analysts' forecasts of earnings play a key role in determining stock market valuations. In recent research, forecast errors have played an important part in explaining momentum and mean-regression effects in asset returns. Given that substantial amounts of wealth ride on the accuracy of these forecasts, it is to be expected that they will conform closely to normative canons of optimality. The evidence suggests that this is not the case. Earnings forecasts have been found to be both biased and inefficient. In this paper, we focus on the earnings forecasts of major market indices such as the S&P500 made by bottom-up analysts (insiders) and top-down market strategists (outsiders). The bottom-up forecasts are aggregates of the forecasts of individual company earnings made by those analysts who follow each company, whereas top-down forecasts are the forecasts made by market strategists who estimate the earnings of the major market indices directly. We present a behavioral model of forecasting that characterizes monthly forecasts by three parameters. Although the general pattern of errors is similar, we find that incentives seem to matter. Insiders appear to be more optimistic than outsiders. Since neither top-down nor bottom-up forecasts is optimal, there is hope that a combination of these forecasts will reduce forecast errors. We confirm that this is indeed the case by generating hybrid forecasts that combine the two forecasts. We conjecture that such a method can be used to develop superior forecasts not just of the earnings of market indices, but of individual companies as well.