Why You Should Read Those Boring 10-K Filings
Differences from one year to another contain useful information for predicting future earnings.
What happens when publicly traded companies issue their annual 10-K reports? Empirical evidence suggests that the stock market pays immediate attention to the tables, but less to the text. Yet the information in the text eventually is reflected in stock prices. In other words, for a significant period of time, price-relevant information about public companies is effectively hiding in plain sight.
Over the past two decades, the amount of text included in 10-Ks has grown substantially. For example, for the typical company, the number of words quadrupled, to more than 60,000 in 2017 from about 15,000 in 1995. Over roughly the same period, the amount of boilerplate and repetition increased; for the median firm, redundant sentences also quadrupled. Finally, the text has become more consistent from year to year, as the share that changed from one report to the next shrank.
A team of economists from Harvard Business School and the University of Illinois at Chicago has studied the value of the textual disclosures in these reports. The authors find that although there is little immediate impact, modifications to the text are highly predictive of changes in the company's stock price over the months after the report, as the implications slowly and gradually become apparent to the market.
The vast majority of the text changes are concentrated in the Management Discussion and Analysis (MD&A) of the 10-K. These disclosures also tend to be more negative than positive, perhaps because the reports are typically drafted by lawyers who tilt toward disclosing negative trends more than positive ones. When the authors applied natural language text processing to evaluate the changes, they found that 86 percent reflected negative sentiment shifts and only 14 percent positive shifts. Furthermore, the text differences contain useful information for predicting future earnings: Changes in the 10-K written text today predict earnings surprises in the future.
Given this negative bias to the textual changes and their ability to predict future earnings, the study shows that companies with 10-K text modifications experience noticeably lower future stock returns than other firms. For example, the authors construct a portfolio that goes long on companies with no material textual changes and shorts firms that contain such changes. That portfolio earns an abnormal positive return of up to 7 percent per year above the market.
The negative future returns are, perhaps not surprisingly, worse when the changes involve legal proceedings and risk factors rather than other textual changes. For companies with “positive” sentiment shifts in the text of their 10-Ks, by contrast, the future excess stock market returns are positive. Other research suggests that the market response to changes in the footnote text is a similar puzzle, in the sense that footnote text shifts are only slowly reflected in stock prices (though the eventual impact on stock returns from footnotes is not as large as with text changes in the main document).
So why don't more people make money from changes in the 10-K text? After all, the modifications in both the document itself and in the footnotes are freely and widely available upon release, but they take time to affect stock prices — and until that price change happens fully, a careful reader of text changes can in effect predict the market, at least to some degree. None of the studies has a great explanation, beyond limited attention spans and the complexity of the documents.
The Harvard-Illinois study has some suggestive evidence that lack of attention is indeed the culprit. In particular, when more investors download both a given year’s 10-K and the previous year’s (presumably making text comparison easier) simultaneously, the future negative stock returns in response to text changes were much smaller than when only the current 10-K was downloaded. This evidence is clearly only suggestive, for many reasons.
Whatever the explanation, the evidence suggests 10-Ks contain market-relevant information. And that information is freely available to any investor.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
To contact the editor responsible for this story:
Max Berley at firstname.lastname@example.org