Researchers have traditionally inferred analysts’ coverage initiations using the first recommendation issued by an analyst, by a broker, or by both in IBES. Using a large hand-collected sample of analysts’ reports announcing coverage initiation, we examine measurement errors in these traditional methods of inferring initiations. We find that these methods all generate significant Type I (misclassification) and Type II (omission) errors, and the nature and degree of the errors vary systematically across the methods. We also show that the measurement errors result in significant sample biases and correcting for the errors can have a significant impact on research findings. We assess the effectiveness of the approaches that prior studies have used to mitigate the omission error in the traditional methods, and propose new approaches that can more effectively reduce the measurement errors.