Fictitious dividend cuts in the CRSP data

Abstract

We compare CRSP dividend data to data from firms' financial statements in the SEC's EDGAR database. Data screens typically used in the extant literature produce a sample in which about half of the apparent reductions in dividends per share are not actually dividend cuts. Many of the fictitious cuts are due to CRSP misclassifying special dividends as regular dividends. We demonstrate that the fictitious cuts can affect point estimates and inferences in studies of the information content of dividends. We provide a simple filter that eliminates most of the fictitious cuts in our sample – remove dividend cuts from the sample that are preceded by a dividend increase in the prior quarter. This filter produces empirical results that are similar to our sample of true cuts.