Quantifying Qualitative Survey Data: New Insights on the (Ir)Rationality of Firms' Forecasts (ESCoE DP 2021-14)


Quantifying Qualitative Survey Data: New Insights on the (Ir)Rationality of Firms’ Forecasts (ESCoE DP 2021-14)

By Alex Botsis, Christoph Görtz, Plutarchos Sakellaris

Go to next section

Using a novel dataset that combines firms’ qualitative survey-based sales forecasts with their quantitative balance-sheet data on realized sales, we document that only major forecast errors (those in the two tails of the distribution) are predictable and display autocorrelation. This result is a particular violation of the Full Information Rational Expectations hypothesis that requires explanation. In contrast, minor forecast errors are neither predictable nor autocorrelated. To arrive at this finding, we develop a novel methodology to quantify qualitative survey data on forecasts. It is generally applicable when quantitative information, e.g. from firm balance sheets, is available on the realization of the forecasted variable. To explain our empirical result, we provide a model of rational inattention. When operating in market environments where information processing is more costly, firms optimally limit their degree of attention to information. This results in larger absolute forecast errors that become predictable and autocorrelated.