Measuring Relative Bias and Accuracy with Crowdsourced Earnings Forecasts

Abstract

A growing number of studies use crowdsourced data to draw inferences regarding information relevance. To bolster research using crowdsourced data and to allow researchers to draw stronger inferences regarding information relevance, we examine the reliability of online biographies using earnings forecasts provided by Estimize contributors. We examine if: (1) biographical information provided by Estimize contributors are reliable; (2) forecast quality is conditional on whether contributors provide their biographical information and names; and (3) contributors who provide their biographical information but withhold their identities make forecasts with different characteristics than those who provide their biographical information and identities. We find Estimize buy siders behave similarly to buy siders documented in prior studies, and Estimize sell siders (especially brokers) are similar to sell siders documented in prior studies. We show that, relative to other Estimize contributors, brokers’ forecasts are more akin to IBES in that they are: made closer in time to IBES forecasts, more likely to be within one penny of IBES forecasts, and as biased as IBES forecasts. We find that contributors who reveal their biographical information are more active on the Estimize platform and issue higher quality forecasts. Finally, we document that known brokers are more pessimistic than anonymous brokers.