Given Awbury’s focus on credit and related risks, we have long been paranoid about missing something “obvious” when analyzing, say, a set of financial accounts.
Our team has been around long enough to remember the likes of Enron, Worldcom and Parmalat; and the sense that in hindsight, the accounting and business frauds should have been obvious, in spite of the fact of information asymmetry between “insiders” and “outsiders”.
Therefore, we are always interested in research that might help to identify and flag situations in which “something does not add up”, thereby leading to closer examination. Naturally, forensic accountants make a living from analyzing fraud and financial malfeasance, but they are almost invariably brought in after the fact, once a problem has been identified.
So, we read with interest of work being undertaken in various quarters intended to provide mechanisms to identify problems ex ante, rather than ex post facto. It seems that, in the era of Big Data and increased computing power, a number of these are based upon the so-called Benford’s Law (also called the First Digit Law, and pre-figured by work done by Newcomb), which posits:
“In lists of numbers from many real-life sources of data, the leading digit is distributed in a specific, non-uniform way.”
So, for example, in any random set of numbers one might expect each number from 1 to 9 to be distributed uniformly (i.e., roughly 11.1% each.) However, somewhat surprisingly, that is not the case. In fact, the number 1 tends to appear some 30.1% of the time and each number from 2 to 9 with gradually decreasing frequency down to number 9 at a mere 4.6%.
This means that the financial data of companies whose accounts do not conform with the Law may be concealing some form of accounting manipulation. Of course, this then begs the question of whether a truly sophisticated financial fraudster, being aware of this fact, could “game” the algorithms being used to detect potential anomalies!
In reality, while a quantitative or statistical approach is no doubt valuable, fraud or malfeasance is committed by human beings (at least in advance of General AI), who do so for many different reasons, not just personal gain. Alignment of incentives, and appropriate governance structures and checks and balances are essential, but not sufficient, in any business; while the human tendency to look for patterns, or to prefer smooth rather than volatile progressions (say, in revenue growth or earnings) can lead to ignoring the fact that the real world is a messy and disorderly place. As the infamous Madoff case demonstrated, smooth and stable returns over lengthening periods of time often mask wrongdoing. And anyone who has been in business long enough knows that a forecast becomes inaccurate the moment it is finalized and published.
Accounting can be mind-numbingly complex, while accounting conventions can also depart from reality. This means that, while one still needs to be able to understand the detail, it is also important to look at the context of a set of accounts and the real-world factors that will have influenced the underlying business, so that one stands a reasonable chance of recognizing when something simply does not make sense, or look right.
In sum, there is no one technique or approach which is likely to identify financial or accounting fraud. One needs to cast the net a widely as possible.
As we said at the start, at Awbury we are constitutionally paranoid; and aim to bring to bear the full range of our knowledge and experience to minimize the risk that we will miss something that should have been “obvious”.
The Awbury Team