Winners in American Banker's annual Wall Street Sharpshooters survey of accuracy in earnings forecasts were chosen using the statistical concept of relative error.

Scores are expressed as a percentage of the average error by the peer group of analysts forecasting quarterly results for a company. Calculations were by First Call Corp.

First Call subtracted analysts' estimates 30-days prior to earnings announcements from the actual earnings disclosed by each banking company. The data firm calculated the total error - a sum of the individual errors. It then divided the total by the number of analysts to produce the average error rate for each bank.

That was done for each quarter and divided by the number of quarters involved.

To determine the best analyst in each category, data were sorted according to analyst error rates and then divided by the number of quarters to produce average relative error.

Varying rates of error for different companies means the category winner may not be apparent from an individual analyst's error rates.

For inclusion in the rankings of analysts covering each bank, analysts had to have made at least two quarterly estimates on the bank this year.

For inclusion in overall rankings, money-center analysts had to have produced at least 10 quarterly estimates of those banks, and regional bank analysts, a minimum of 25 estimates in that larger category. For thrift institutions, it was 14 quarters; for general finance companies, 11 quarters; for government enterprises, six quarters; and for credit card issuers, nine quarters.

While a useful tool, statistical measurement, of course, is not necessarily the complete truth.

It should be noted that a cluster of highly accurate or highly erroneous estimates can skew the average forecast error for a company's earnings. When this happens, more than half the analysts could end up doing either better or worse than the average.

And in some cases, as with a company that has had volatile earnings or a "surprise" result not previewed for Wall Street more than 30 days before its release, most analysts may end up performing inconsistently across their category.

Subscribe Now

Access to authoritative analysis and perspective and our data-driven report series.

14-Day Free Trial

No credit card required. Complete access to articles, breaking news and industry data.