Lundholm and Rogo (2016; hereafter, LR) present some intriguing new evidence that analyst forecasts may be too variable. I enjoyed reading the paper. The empirical work is well done, and concerns with forecast staleness and stationarity assumptions are all carefully handled. It seems clear from the evidence that the frequency of variance bound violations (VBVs) is well above zero for a significant number of firm-quarters. Evidently, at least some of the time, the population of one-year-ahead earnings forecasts varies too much, given sensible statistical bounds established by past realizations of earnings changes.
In this commentary, I summarize the main findings in the study and reflect on how these findings might be related to other big-picture issues of concern to accounting researchers. Specifically, I discuss what we might reasonably infer from these findings about the rationality of analyst forecasts, excessive volatility in stock returns, and the broader role of analysts...
