How to Fake Sincerity in Reporting Quarterly Results

Michael Giberson

The Wall Street Journal reports on research analyzing quarterly reports (nearly 1/2 a million of them from over a 27-year period) which discovered strong circumstantial evidence that companies tweak their results to improve reported earnings.  The evidence? The number 4 occurs less frequently than chance would dictate in the tenths of a cent digit for quarterly earnings.  Pushing an initially calculated 0.4 cent result just a little will get it to 0.5 cent, and 0.5 cent can be rounded up to the nearest whole cent, which can make the difference between making or just missing a quarterly target.

The study, conducted by folks at Stanford, also found that that companies that later restate quarterly earnings or become charged with accounting violations report far fewer 0.4 cent results on average, compared to other firms.

The article quotes one of the study’s author, Stanford law professor Joseph Grundfest, as suggesting the pattern may be “a leading indicator of a company that’s going to have an accounting issue.” Maybe so. But now that this study has been publicized, it suggests a way to fake a signal of quality: just fudge them up to a 0.4 cent end. Be suspicious of a company that hasn’t had a 0.4 earnings result for the last ten years if they start reporting 0.4 cent results all of the time.

2 thoughts on “How to Fake Sincerity in Reporting Quarterly Results

  1. There was a study done years ago on this issue that any random number series based on sequential data (say, street addresses on a random page of the phone book, as opposed to throws of a die) can be audited in this fashion, since such sequences should show an exponential decay in the frequency of the digits in their analyses – the idea being that a street is more likely to have an house with an address starting with the number 2 than 3, and more likely to have a 1 than a 2 – but people who are making up a series of addresses are unlikely to get their mix of guesses right to show this impact. (The key is that you look not at the probability of seeing a specific number like 17, but rather at the digits – so a 17 counts as one “1” and one “7” – and you would expect to see a lot more 1s than 7s in any truly random sample of street addresses.

    Presumably, earnings forecasts would show the same propensity. In which case you need not look for 4s necessarily, but rather any pattern of digits that don’t show a statistically likely decay in the probabilities for higher digits.

Comments are closed.