Have you ever wondered how much investors actually use SEC filings? I just read a paper with awesome new data that speaks to this question. The paper is entitled, “The Demand for Mandatory Disclosure: Evidence from Investors’ Use of the EDGAR Database,” and is coauthored by Michael Drake (BYU), Darren Roulstone (OSU), and Jacob Thornock (Univ. of Washington). The authors obtained the EDGAR server logs for the first half of 2008, “which record each ‘click’ … by a user to acquire a specific filing from EDGAR.” The dataset allows them to observe who is making the request (based on a partial IP address), when they make the request, which firm the request relates to, and the specific filing and filing date being requested. I know—wouldn’t you love to have this dataset?!
This paper reports data on the forms/filings that are most often requested and their relationship to firm information environment, extreme firm performance, and the sophistication of the investor base. Among other things, we learn that more than a million requests for SEC filings occur on a daily basis. We learn that 10% of the most highly demanded form types represent 95% of the requests. With over 200 forms/filings required by the SEC, this makes you wonder about the other 180 forms/filings. In fact, when I saw that statistic, it made me think of the disclosure framework project, which I know is trying to grapple with information overload. Wouldn’t you love to know which footnotes in a 10k are most often requested? Unfortunately, the data in this paper is not that granular.
Regardless, I thought I would post this paper for your general enjoyment. I also posted this paper because it reminds me of the usefulness of thinking outside of the data we normally use, and instead thinking of where you might be able to get data that would help you answer the questions you really care about. I’m sure there’s a string of papers that will come from this data, and I’m looking forward to seeing those papers. In the meantime, feel free to post your thoughts and reactions to this paper right here, including any thoughts on the questions you might ask if you had such a dataset. I look forward to your comments!