GE and Benford
GE was recently placed in the cross-hairs of famed Madoff whistleblower Harry Markopolos. According to his claim, GE has a minor $38 Billion accounting problem. That’s a lot of digits for a former stalwart of American capitalism.
I like accounting, but not as much as I like math. Had I known about the field of forensic accounting years ago, maybe my interest in accounting would be greater. But I didn’t, so I ended using my math skills as a trader.
Benford’s Law is one of the primary tools that forensic accountants use to search for fraud. It has a fascinating history that I won’t delve into but I encourage you to google it. Benford categorized something that had been learned in the late 1800s: digits within numbers have a distribution. Armed with this distribution, you can determine if certain numbers, say within accounting statements, are appearing an expected number of times, or not.
Forensically, if certain digits have a distribution far from the expected, one could infer that the numbers had been doctored in some way. This was the case with Enron, they doctored their numbers pretty heavily back in the 2000s before they collapsed into bankruptcy. Enron’s infamy boosted the popularity and usage of Benford’s Law.
Forbes posted an article about Markopolos’s claim which got me thinking about Benford. It seems the major accounting problems Mr. Markopolos has pointed out are items that would be debatable between professional accountants and consultants, not necessarily outright Enron-type fraud. I wanted to see if Benford validated Markopolos's assertion or not.
In the Enron days it would have been hard to run Benford’s law against all the companies in the S&P 500, but not with today's computing power. I loaded up the last 5 years of annual and quarterly financial statements for all the S&P 500 companies then calculated the mean squared error for their digits versus the Benford expectation. It took my Linux box a few hours to crunch through the data.
Here’s what I found. GE’s financial statement numbers are pretty close to what you would expect per Benford’s law. In fact, better than most in the index. Scaled charts below.
This could mean GE is not committing fraud and Mr. Markopolos is wrong, or, GE is committing fraud, knows about Benford’s Law, and has turned the “Benford Flag” to “On” in their Quickbooks application. Kidding aside, we now know that the GE's numbers haven’t been grossly cooked and that Benford doesn't validate Mr. Markopolos's claim.
Time and tides expose those who are swimming naked. Guess we'll see soon if GE brought their suit.