Identification of Analyst Coverage Initiations

Researchers have traditionally inferred analysts’ coverage initiations using an analyst’s and/or a broker’s first recommendation for a firm in IBES. Using a large hand-collected sample of analyst reports announcing coverage initiations, we document that 17% of the traditionally-inferred initiations are not actual initiations (Type I error) and that 50% of the actual initiations are not identified using the traditional approach (Type II error). We show that these measurement errors lead to systematic biases in a variety of firm and analyst characteristics and in the distribution of analysts’ recommendations, and that correcting for these measurement errors changes an important prior finding. We assess the effectiveness of the methods that prior studies have used to mitigate the measurement errors, and suggest a new method that improves upon these prior methods