Those of us who have been associated with audit and investigations in one form or another over the years will always be on the lookout for something that gives us the edge over the fraudster.

One such approach that has recently gained favour in financial markets is based on Benford’s Law which examines the frequency distribution of leading digits in numerical data sets.

It has been around for a long time, but only recently has it been applied in accounting and in the financial sector: Insurance companies have started using it to detect false claims, the IRS to detect tax fraud, and the Big 4 accounting firms to detect accounting irregularities.

It is named after an early-20th-century British scientist and states that in lists of numbers from any naturally occurring data source, be it credit card charges, procurement entries, cash receipts, whatever, the first digit for each number will be 1 (for example, 1, 157, 1,820) about 30% of the time. The first digit will be 2 about 18% of the time, and each successive number will represent a progressively smaller proportion, to the point where 9 will occur as the first digit less than 5% of the time.

This distribution has been found to hold for a practically limitless array of data sets, for example: The length of rivers (in feet and in meters), the population of cities and countries, trading volume on stock exchanges, the number of ranking points for tennis pros, the molecular weights of chemicals, the height of the world’s tallest buildings, etc. (in many ways similar to the Pareto principle, a.k.a 80/20 rule)

Accounting variables should also be distributed in accordance with Benford’s Law—and they are, as long as there has been no conscious manipulation of the data and this also holds true even if the figures are converted from one currency to another. It is fair to say that if a set of accounting data deviates from Benford’s Law, that can be taken as evidence of manipulation.

Hence when you are reviewing company’s financial statements or an inventory listing or even sales invoices. If an unusually high number of first digits in the accounting data are 7s, 8s, or 9s, it may indicate a conscious effort by someone to finesse the numbers to achieve the desired financial results or create false data.

This statistical tool, which would appear to have limitless applications if you are using CAATS might be supplemented by verbal clues when it comes to questioning suspects.

I hear of two accounting academics who have analysed the transcripts of nearly 30,000 conference calls by U.S. CEOs and CFOs from 2003 to 2007. Drawing on psychological studies that show how people’s speech patterns change when they lie, they have discovered several verbal cues that could have tipped off a listener that something was not quite right with the company’s accounts.

For example, in companies that were later forced by the SEC to make major restatements of key financials, managers displayed the following patterns:

  • They referred to shareholder value relatively seldom (perhaps to minimise the risk of a lawsuit).
  • They used extremely positive words (for example, instead of describing something as “good,” they called it “fantastic”).
  • They avoided use of the word “I” in favour of the third person.
  • They used fewer hesitation words, such as “um” and “er” (which might suggest that they were coached in their deceptions).
  • They used obscenities more frequently.

With a little coaching, you can of course be taught to avoid those.

I think that one of the most interesting points is the last one, on the subject of obscenities. Over many years of dealing with very senior managers I have noticed that one or two used obscenities to frighten people, to force them to make decisions in the heat of the moment. With hindsight, they were often hiding something, even if it was only poor decision making on their part, rather than fraud. An interesting indicator to watch out for.


Article by Chris Hollands, a director of TomJak Ltd, a company which specialises in audit training and consultancy