A criticism of the SEC that you'll sometimes hear is that it's mostly a bunch of lawyers, and two things that are broadly true of lawyers as a class is that they are good at close readings of dense texts and terrified of math. This means, some might say, that the agency is ill-equipped to regulate the high-tech quantitative world of modern finance. So it's obscurely pleasing to read that the SEC's office of quantitative research is rolling out a new program that applies high-tech quantitative methods to, basically, close reading of dense texts:
An initial step in the SEC's new effort [to crack down on accounting fraud] is software that analyzes the "management's discussion and analysis" section of annual reports where executives detail a company's performance and prospects.
Officials say certain word choices appear to reveal warning signs of earnings manipulation, and tests to determine if the analysis would have detected previous accounting frauds "look very promising," said Harvey Westbrook, head of the SEC's office of quantitative research.
Companies that bend or break accounting rules tend to play a "word shell game," said Craig Lewis, the SEC's chief economist and head of the division developing the model. Such companies try to "deflect attention from a core problem by talking a lot more about a benign" issue than their competitors, while "underreporting important risks."
It's also pleasing to hear that a CFO's guilty conscience over his earnings manipulation seeps directly into his prose. Though the article is a little light on the details of the SEC's earnings-manipulation model, which I guess makes sense, since "companies and their lawyers are expected to respond to the crackdown by trying to outsmart the agency's computers," which I would really like to see.1 That could be a mixed bag; the Journal hints that it might result in easier-to-read but more grandiose filings:2
In annual reports, executives tend to refer to themselves more frequently when their companies are doing well, University of Michigan accounting professor Feng Li said. Companies with poorer earnings often "file annual reports that are more difficult to read."
I for one have no problem with a move toward easy-to-read narratives of CEO heroism.
Actually, all in all I've mostly enjoyed the SEC's increasing reliance on its office of quantitative research, which brings a cheery efficient-markets fundamentalism to the SEC's work. Your hedge fund is outperforming? Must be fraud! You bought options before a takeover? You got inside information somehow! And now, the positive tone of your earnings doesn't match up with the obfuscation and evasion in your MD&A? Probably accounting fraud! Well, it probably is.
On the other hand, if you're an efficient-markets fundamentalist, is running screens on public information looking for public-company accounting fraud the best use of SEC resources? After all, "we're gonna try to stop accounting fraud" also means in part "we're going to give up on stopping circa-2007 fraud," which I guess is forward-looking of the SEC, though also perhaps driven by the fact that the statute of limitations has run out on most of the things you could've done in 2007.3 And whereas private lawsuits, and the linear and irreversible nature of time, will probably prevent any further fraud from being committed in 2007, the SEC obviously has other stuff on its plate too. Steve Cohen! Ponzi schemes! Steve Cohen! Flash crashes! Steve Cohen! Hedge funds fudging returns! Steve Cohen! Etc.
What those things have in common is that no one can short them: it's not really profitable for anyone to root out fraud in private placements, whether by companies or hedge funds or private equity funds, because you can't generally make money shorting them. Whereas anyone who builds a particularly good computer program to spot fraud in public company financial statements should be using it to generate short ideas, not giving it to the SEC. Once you've shorted the stock of the fraudulent company, of course, then you tell the SEC.
With a whole army of profit-motivated short sellers to do the screening for them, why should the SEC do the screening itself? Oh lots of reasons, probably, most of them centered around the gap between "fraud that is of interest to the SEC" and "fraud that is of interest to the SEC and also a promising short candidate."4
But also ... I dunno, if you're a real efficient-markets fundamentalist, don't you fundamentally not believe in short sellers doing work? Recent regulatory efforts include crackdowns on expert networks, scrutiny of abnormal hedge-fund performance, and more or less explicit statements that equity long-short funds are the favored target for regulatory crackdowns, unsystemic enough to prosecute but high-profile enough to be career-enhancing to prosecute. (Also: occasional bans on short selling.) The model of "guys sit around trying to figure out things about public companies that no one else knows, and then trade on those things," does not exactly seem to be highly favored; nor does the model of rewarding those guys with investigations of their short targets. In the SEC's world of perfect efficiency, it wouldn't be worth it for a hedge fund to do the work to spot accounting fraud. Which means the SEC will have to do it.
1.I want someone to build a computer program that translates any passage of straightforward financial description into, like, definitely not fraudulent and extremely fraudulent, depending I guess on the effect you're going for. Also, randomly, the Journal says "suspicious language or numbers in securities filings aren't necessarily illegal," which is just, I briefly enjoyed contemplating a world were they were.
2.The article quotes Columbia law professor John Coffee saying "As soon as the SEC suggests it's going to look at this in terms of the numbers of words, lawyers will be more loquacious," which I don't understand at all but which I chalk up to lawyers being afraid of math. Like, I don't think the SEC's screen for fraud is just "any 10-K under 150 pages is fraudulent"? Though maybe it is?
3.Not, like, murder, though. Keep that to yourself. Actually, generally speaking, probably don't go around talking about any illegal stuff you did in 2007. "Hey I read on Dealbreaker that I'm in the clear for all that fraud I did in 2007," is something you shouldn't be putting in an IM, particularly not on your Bloomberg. Not legal advice!
4.For instance, there's a whole category of, like, entirely non-substantive accounting fraud, like options backdating or whatever, that shouldn't cause the shares to go down but might nonetheless interest regulators. Though: why? More generally, a lot of prominent accounting frauds, Enron among them, could have been caught by the simple method of "paying attention when short sellers said they were frauds."