Man versus Machine Learning: The Term Structure of Earnings Expectations and Conditional Biases

Abstract We introduce a real-time measure of conditional biases to firms’ earnings forecasts. The is defined as the difference between analysts’ expectations and statistically optimal unbiased machine-learning benchmark. Analysts’ are, on average, biased upward, bias that increases in forecast horizon. These are associated with negative cross-sectional return predictability, short legs many anomalies contain firms excessively optimistic Further, managers companies greatest upward-biased forecasts more likely issue stocks. Commonly used linear models do not work out-of-sample inferior those analysts provide.