Incentives, Information Production, and Bias : Evidence from Analyst Text

I study how analysts' performance depends on their incentives using a regulatory shock and the textual content of analyst reports. My results focus on two aspects of performance, information production and bias. After incentives are reduced, analysts produce less information that is also less biased. My identification of analyst incentives uses the Global Research Settlement as a shock that affected analysts at investment banks, but not other analysts, in a difference-in-difference design. I find that analyst reports become more similar to one another after the shock, an indication of less information production. Additionally, I find that text exhibits lower overall sentiment, which is driven by an increase in negativity, and that reports contain fewer markers of bias (less weaseling). The results highlight a mechanism through which there is a trade-off between information production and bias