Who Herds? Who Doesn'T?

We build a simple model of analysts' propensity to herd. Using ideas from GMM and simulated method of moments, we estimate an analyst's herding propensity with I/B/E/S forecast data from 1989-2004. We find that, of the analysts whose herding propensity is defined by our model, 85% of them tend to herd while 5% of them tend to stand out from the crowd (i.e., 'anti-herd'). Out-of-sample tests validate our underlying model for analysts' behavior. Further cross-sectional analyses suggest that an analyst tends to herd if she issues less accurate forecasts in the past, has more analysts that issue forecasts before her, a longer forecast horizon, issues forecast less frequently, has less firm-specific experience, more general experience, follows more industries, works for a smaller brokerage house, and follows a firm with less volatile earnings and smaller size