How good are analysts at detecting accounting fraud? Not that great, according to new research.

A joint study by Julie Cotter from the University of Southern Queensland in Australia, and Susan M. Young from Baruch College at City University of New York found that they're not real flash.

Their paper, Do analysts accept accounting fraud? found that analysts respond to some frauds and not others. And they are still fooled by fictitious accounts.

They analysed analysts' response to companies that had committed an accounting fraud over the period from 1995 to 2002 as evidenced by the SEC's issuance of an Accounting and Auditing Enforcement release. They found that analysts were more likely to drop coverage of firms of with overvalued asset frauds but they didn't really seem to care about related party transactions. Analysts were more likely to revise down for companies that had understated their income or assets, but less likely to do it for those that had overstated their assets or had engaged in related party transaction frauds.

Also, there was no evidence that they would revise downwards or drop coverage of companies that had created fictitious revenues or recognised revenues before the money came in.

"Given that these frauds are the most common type of fraud, we should expect analysts to be more sensitive to their occurrence," they write.

They say there are a few reasons why analysts are not that proactive in the area.

"One reason is the cost of detecting accounting fraud ... Additional reasons include a lack of ability or experience and incentives to please management and increase personal compensation. In the case of fraudulent reporting, we expect that observations of no revisions or no dropped coverage are more likely to represent lack of ability rather than incentives."

The results of this study are not surprising. But it's an indictment nonetheless of analysts as the market's gatekeepers.