May 15, 2014

A Filter for Financial Fraud

A new technique uses the power of statistical analysis to uncover likely accounting and financial fraud. The tool could help the SEC — as well as auditors and investors — to quickly and inexpensively root out wrongdoing.

Print this story

The Idea

Use a technique from statistical analysis to quickly identify likely accounting misstatements.

The Research

During the last decade, the Securities and Exchange Commission (SEC) has dramatically scaled back the unit responsible for investigating accounting fraud and increased its focus on misdeeds in the financial sector that have emerged as a result of the financial crisis that started in 2008. Reported instances of accounting fraud and financial restatements have dropped, but many experts cite lax enforcement as a result of the cuts as the cause of this drop, not a decrease in wrongdoing.

Professor Dan Amiram, who specializes in accounting, taxation, and business law, along with Columbia Business School doctoral candidate Ethan Rouen and Zahn Bozanic of Ohio State University, adapted a statistical technique that has been used to detect other types of artificial data manipulation that could help the auditors, investors, and the SEC quickly — and inexpensively — detect accounting fraud and financial irregularities. The technique acts as a filter, sifting through financial statements for signs of likely manipulation to point investigators toward those that should be more thoroughly investigated.

Practical Applications

Auditors, Financial Regulators, Investors

You can use this research to detect likely instances of accounting fraud and financial irregularities in financial statements. Using methods from statistical analysis, the researchers developed a measure, the Financial Statement Divergence (FSD) score, that allows them to compare the distribution of certain numbers in a firm’s annual financial statement to that of Benford’s distribution — a theoretical expected distribution frequency. In this case, the distribution is applied to the frequency with which the first digits of numbers in a financial statement should appear. For example, the first digits of the numbers in the statement should appear with decreasing frequency, with the number 1 appearing about 30 percent of the time, the number 2 appearing about 17.6 percent of the time, and so on.

The researchers found that about 84 percent of existing recent financial statements conform to Benford’s Law, or the law of first digits. They measured how much the nonconforming statements diverged from the expected distribution — it’s this divergence that makes up the FSD score. The higher the FSD score, the greater likelihood of misstatements, a signal that should trigger a deeper investigation for wrongdoing. The study shows that firms that correct their misstated financial statements have lower FSD score after they corrected the wrongdoing. The study also shows that FSD score predicts actual SEC enforcement actions against misstating firms.

Dan Amiram is assistant professor of accounting at Columbia Business School.

Read the Research

Amiram, Dan, Zahn Bozanic, and Ethan Rouen. “Financial Statement Irregularities: Evidence from the Distributional Properties of Financial Statement Numbers.” Working paper, Columbia Business School, 2014.