Nicholas Bloom

Kellogg Labor Seminar Series - (Click for series website)

Speaker: Nicholas Bloom from Stanford University

Title: Rationalizing Firms Forecasts

Abstract: We partner with a US payments company to run a 5-year, 10 wave panel survey of incentivized sales forecasts on 6,594 firms. These firms are size representative of US employee firms. We identify issues of idiosyncratic bias, predictable errors and over-precision, that lead to extremely poor forecast accuracy. These biases are substantially larger than those found in the prior literature studying large public firms. Second, we run four experiments to try and improve forecasts, focusing on (i) data, (ii) incentives, (iii) training, and (iv) contingent thinking. These interventions have little impact in reducing biases or improving overall accuracy. This suggests that most firms have substantial and persistent forecasting biases, which is important for modelling micro and macro policy interventions.