Where Fraud Lives and Why
This paper has been getting some attention lately for its eye-catching estimates: 11% of publicly traded companies are committing securities fraud every year, with an annual cost of over $700bn. The paper is a classic use of two powerful tools for coming up with useful generalizations:
- There's a natural experiment: when Arthur Andersen collapsed, its former clients all had to find new auditors. The new auditors were more likely to hunt for fraud, both because they didn't have the institutional disincentive that spotting a fraud they'd previously allowed looked bad, and because Arthur Andersen's own accounting was tainted.
- If we treat this as a meaningfully random sample, we can extrapolate to get some kind of base rate for fraud. There are many reasons to doubt this: what if Enron was a representative case of Arthur Andersen's auditing? What if they focused on industries with lots of fraud? The paper actually does a great job of addressing most of these concerns (Andersen-audited firms had fewer accounting restatements than similar companies audited by other major firms, for example, and adjusting for industry exposure didn't have a big impact).
The paper notes that fraud is cyclical, and that the natural experiment occurred towards the middle of the cycle, so it's not too tough to get a full-cycle estimate of fraud from that snapshot. But what if fraud is cyclical, but not purely so, and Arthur Anderson’s collapse just part of a cycle within a secular trend towards less fraud? There are good reasons to think that accounting fraud at US public companies is not as common as it used to be. The equilibrium for fraud is set by two forces: how rewarding it is for companies to engage in it, and how rewarding it is to catch it:
- Sarbanes-Oxley created a bunch of annoying disclosure requirements for companies, but it also made it much easier to go to prison for committing fraud. And while some of the big accounting frauds were of the making-up-numbers variety, many of the problems caught in the early 2000s were more about smoothing out numbers than manufacturing new profits out of nothing.
- The growth of long/short funds as a sub-asset class meant that there was a larger dollar value of short positions, and thus more demand for reasons that would make a stock plummet. (Aggregate short interest has been rising for decades.)
- Cheaper and better data meant that some symptoms of books-cooking—companies booking more accounts receivable relative to actual cash revenue, for example—get incorporated into systematic models. And then the long/short process kicks in: if a stock looks cheap in part because systematic models are betting against its low-quality earnings, a fundamentals-oriented analyst might look into it as a long and determine that it's a short instead. (A below-market P/E is only as good as the E!)
- Investors increasingly expect mature companies to return all of their earnings in the form of buybacks, and will evaluate these companies on the basis of free cash flow rather than reported earnings. If a company is persistently unable to convert stated earnings into actual cash, or unwilling to release cash from the balance sheet to buy back stock, investors will wonder why.
- The market has also gotten more liquid, and the market in ideas is more liquid, too; early-2000s fraud was partly ubiquitous because the people best positioned to spot it either had fees to worry about (if they were sell-side analysts) or couldn't spread their ideas widely (buy-side analysts). Today, short theses and rebuttals (or lack thereof!) spread fast through FinTwit.
So generally speaking public markets in the US have gotten more hard-headed about frauds. Today, if you hear about a massive accounting misstatement plunging a company into bankruptcy overnight, it's more likely to be in an emerging market or in Europe. Look at Wikipedia's table of big accounting frauds you’ll find the same thing—it was utterly dominated by American companies in the early 2000s, but post-crisis the only frauds involving American companies are 1) Autonomy, a British company whose accounting issues were discovered after it was acquired by the American HP, and 2) Wells Fargo, where the fraud was perpetuated by lower-level employees against management (albeit because management set up terribly perverse incentives to do so).
But since fraud is a human problem, and not purely a matter of better accounting standards, it's not likely to have just gone away. But if the rate of accounting problems among big publicly-traded companies is lower than the 11% number cited in the paper, the question isn't "why did it disappear?" but rather "where did it go?" And we can take our list of trends against fraud and invert them:
- Sarbanes-Oxley does apply to private companies, but only on the penalty side, not the disclosure side. But accounting frauds in private companies are often less visible; many investments go to zero, anyway, and it's less embarrassing for everyone involved not to say why.
- There are no short-sellers in private markets. There have been efforts here, but they don't work out because the market doesn't clear ("everyone wanted to short Theranos, Dropbox and WeWork"). The closest you can get to shorting is to pass on a round and then brag about it later. Big deal: I didn't invest in FTX, either.
- There's less data available on private companies, though the rise of alternative data tools means it's easier to get decent proxies.
- Startups are not expected to return capital. It's a bad sign if they do. They're often valued either based on strategic considerations or starting with a multiple of sales—a dollar of sales is much easier to fake than a dollar of earnings or cash flow, so the incentive to do so is strong.
- The idea market in startups is liquid when it comes to successes, but it would be pretty tacky for a VC to write a long blog post explaining why they passed on a live deal. (That memo may exist internally, but to the extent that it's shared it's in the form of a quick summary over Twitter DM or Signal.)
JPMorgan Chase's writedown of their fintech acquisition Frank is a great case study in all of these forces. The NYT has a good story digging into the details: Frank's founder is a serial exaggerator whose self-promotion veered into fraud (once again, if the rate of continuous improvement in public perception to be maintained exceeds what the fundamentals can deliver, compound interest works its ruthless magic). The company was valued at a high multiple of what turned out to be a flexible metric, total email addresses captured. And there were alternative datasets that could have pointed to problems: given the likely number of student aid applicants in the US, Frank's numbers implied that it had reached near-dominant market share in the category with little marketing. Meanwhile, its monthly site traffic was not enough to have acquired that sizable a customer list over Frank's entire existence. So it could have been caught, if the buyer had been looking for fraud. But one paradox of frauds and cheats in general is that lying is less than half the work—most of the effort is in appearing not to need to lie. The more impressive a company looks, the more embarrassing the basic due diligence questions are.
A down market and a series of high-profile failures might give private markets the same kind of natural experiment that Arthur Andersen's failure did for public markets. Due diligence checklists will get longer and more thorough, and new funding rounds will feel more like a cross-examination and less like a party. One reason for a high base rate of fraud is that at least some of it stems from inattention rather than malice—the Arthur Andersen study finds that most of the frauds were fairly minor, and could be more the result of poor internal metrics than of intent to mislead. But either way, standards will get higher, and private companies will need to step up their efforts accordingly.
Any institution that fundamentally sells some kind of imprimatur has the perverse economic incentive to spend a while making its reputation as good as possible and then to basically sell it off and try to maximize gains while people took a while to catch on. The more sterling the reputation, the stronger the temptation; any long-lived institution has to devote a lot of its energy to struggling against this. It's fortunate for the auditing industry that there isn't a member of the Big Four that's considered notably better than the rest—to the extent that there was one historically, it was Arthur Andersen. ↩︎
Of course, one of these can easily evolve into the other. If your fraud consists of smoothing the path to 15% annual growth in earnings per share, but what your business actually produces is more like 10% +/- 20% growth in earnings per share, a yawning gap will open between what the "smoothed" number is and what the business can produce. Sometimes that gap gets filled by getting more inventive and more fraudulent. ↩︎
Management's responsibility here is a bit like corruption in emerging markets: if bureaucrat's wages don't keep pace with market wages, the only people who want to work for the government are the ones who are deeply patriotic or the ones who expect to make up the difference with bribes. And once an organization has a critical mass of corrupt people, going along with them tends to be necessary for career advancement. This can be the most cost-effective way to run things, at least for a while—in some countries, bribes may be cheaper to collect than taxes—but it's important for anyone running this system to be aware of it. Wells Fargo executives had to wonder, from time to time, why employees were so willing to go through the grueling process of relentlessly cross-selling. ↩︎
In fact, accounting standards can make it worse! Companies used to ignore depreciation, and once they started recognizing it they could play games with improperly slow depreciation. ↩︎
A Word From Our Sponsors
Here's a dirty secret: part of equity research consists of being one of the world's best-paid data-entry professionals. It's a pain—and a rite of passage—to build a financial model by painstakingly transcribing information from 10-Qs, 10-Ks, presentations, and transcripts. Or, at least, it was: Daloopa uses machine learning and human validation to automatically parse financial statements and other disclosures, creating a continuously-updated, detailed, and accurate model.
If you've ever fired up Excel at 8pm and realized you'll be doing ctrl-c alt-tab alt-e-es-v until well past midnight, you owe it to yourself to check this out.
A useful tool in finance is the idea that market dislocations can persist when investor populations are different. Junk bonds, for example, produced great returns when almost no one in the bond business thought like equity investors and looked for potential upside rather than the avoidance of downside. It also applies to companies; in energy, the frackers were such voracious consumers of capital that they've had to become closely attuned to their investor base, and that investor base (still!) isn't excited for them to spend money. (To paraphrase T. Boone Pickens, the best place to convert $1 of assets into more than $1 of future oil revenue isn't the Permian, it's the NYSE.) Other parts of the industry are showing different behaviors: day rates for offshore drillships are back to 2014 levels after a long and painful slump ($, WSJ). The magnitude of an industry upcycle is partly a function of the length of the previous downcycle; the longer the industry has gone without additional capital expenditure, and the larger the number of retirements since the last boom, the more extreme the supply/demand mismatch can be when things turn around.
(Andrew at Yet Another Value Blog has a good in-depth series on offshore for those interested in more.)
Apple the Outlier
Of big tech companies, Apple was unusually slow to expand headcount in the last few years, and is now the only one not to have announced layoffs ($, WSJ). This comparatively cautious strategy is one reason their shares have held up better than large-cap rivals (since January 1, 2022, Apple's total return is -24%; Google is down 32%, and other large tech companies look worse than that).
At this point, avoiding layoffs might be more of a long-term strategic choice than a short-term financial move; at a company that's still fairly good at avoiding leaks, a slight increase in attrition would not be hard to hide. Meanwhile, being able to tell prospective employees that the company avoided layoffs is a strong selling point—which is especially valuable right now, when so many talented people are looking for their next role. For investors, this would show up as worse short-term margins, but a comparative advantage in talent-hoarding is a valuable long-term asset.
Google is increasingly relying on outside companies to sell ads, rather than handling it internally. This is partly because it's much better, for PR purposes, to reduce payments to a supplier than to lay off more employees. It's hard for companies to have many categories of superstar employee without keeping headcount extremely low—at some point, it's either a company where the salespeople are in charge or one where the tech people are. (It's hard enough to deal with hierarchies within those groups.) So Google is making a prudent choice here.
Interestingly, one possibility on the ad sales front is that self-service advertising can be much more effective with LLMs. Ad systems for large, mature companies are necessarily complex, so it's easy to waste money. But much of the information Google needs to target ads is still controlled by the advertisers themselves. An LLM is a decent tool for collapsing that asymmetry through conversational back-and-forth, and it's also a way to get a substantial sample size without worrying that the cost per query will make LLM use uneconomical.
Working as Intended
Crypto-focused banks are borrowing billions from Federal Home Loan Banks to make up for deposit shortfalls as their crypto customers vaporize ($, WSJ). It's important to note that this is exactly what such systems are for. The point of bank regulation is 1) keep banks from taking too many risks, but 2) ensure that, if they do wind up taking risks that put them at risk of a liquidity problem, give them emergency funding so they aren't forced to liquidate. This does illustrate that crypto-related deposits are much less sticky than traditional deposits, and that it would be consistent to treat them differently for regulatory purposes. But a meta principle of banking regulation is that we shouldn't delay putting out fires because we're putting the finishing touches on our new fire-safety lecture.
Twitter's active employee count is down by 80% since Elon Musk took control, though Elon Musk cites a higher number. This is now past the 75% layoff number that Musk allegedly floated and then denied. It may be easy to underestimate attrition when layoffs come along with a culture shock: employees who didn't sign up for a 50% leaner and 100% meaner Twitter might leave and make it leaner still. At this point, it's become a very interesting experiment in how small a team it takes to run a service with hundreds of millions of users, at least for a while. But it's easy to overfit: if a large fraction of the headcount is redundant—having multiple people who can each perform the same critical role in case one of them leaves for example—then headcount reductions will always look good at first and much worse over time.
Companies in the Diff network are actively seeking talent! If you're interested in exploring growth opportunities at unique companies, please reach out. Some top current roles:
- A fintech company using AI to craft new investment strategies seeks a Brokerage Engineer / Trader with 2+ years of experience in trading or operations for equities or crypto. This is a technical role—FIX proficiency required, Python skills a plus. (NYC)
- A well funded early stage startup founded by two SpaceX engineers is building the software stack for hardware companies. They're looking for a backend engineer who can build services that quickly process large amounts of data. (Los Angeles)
- A company bringing machine learning tools to everyone is looking for experienced ML engineers with strong product sense. (Remote)
- A new service that's trolling the dating market with a better product and better monetization is looking for a full-stack founding engineer. (Los Angeles)
Even if you don't see an exact match for your skills and interests right now, we're happy to talk early so we can let you know if a good opportunity comes up.
If you’re at a company that's looking for talent, we should talk! Diff Jobs works with companies across fintech, hard tech, consumer software, enterprise software, and other areas—any company where finding unusually effective people is a top priority.