Was Madoff Predictable? Academics Say Yes

Finding Bernie Madoff: Detecting Fraud by Investment Managers by Stephen G. Dimmock and William C. Gerken, a December 2009 paper available in the GARP Digital Library say that yes, Bernie Madoff could have been stopped using the publicly available information of mandatory SEC filings. " ... we find the firm’s fraud risk is on the 95th percentile of all firms in our sample, and the firm’s felony fraud risk is on the 98th percentile. The firm’s economic interests in clients’ transactions, affiliation with a broker/dealer, and its history of regulatory violations all contribute to the high score." That with filings done 11 months before he was charged.

Dimmock and and Gerken found that such public information could have been used to predict over 70% of frauds at 13,592 investment managers serving 20 million investors between 2000 and 2006. The list of things that can predict future fraud:

• past legal and regulatory violations; history of crime, including non-investment crime such as drunk driving, is strong indicator of fraud risk!
  
• conflicts of interest - 1) advisors that have economic interests in securities traded on behalf of clients, 2) accepting soft dollars (when investment advisors direct client trades to a brokerage with relatively high commissions, and in return the broker supplies the advisor with research or other benefits), or, 3) affiliation with brokers/dealers.
• firms with smaller investors - The presence of large institutional clients is negatively related to fraud disclosures (i.e. somebody with clout and capability is actually watching the advisor to keep them in line). They found that the converse is also true - " ... firms with smaller investors have significantly more frauds."
  
• when advisors hide or non-transparently disclose information, as allowed by SEC rules
  

Smaller firms with fewer investment advisors were not associated with higher rates of fraud risk. Another factor that is not associated with future fraud is a history of investor lawsuits.

They also discover that investing money with managers high on their fraud risk scale did not have any payoff in higher returns - there's no payoff for taking the risk of ripoff.

Their methods result in very few predictions of fraud when in fact none occurred. Applying their tests would not unfairly exclude many good investment advisors.

Especially troubling is that it is common practice for firms to remove previously disclosed crimes from subsequent filings (as allowed by the SEC) when such information has high marginal value in predicting fraud.

They offer some suggestions for simple and cheap (as in zero marginal cost to advisors) improvements to disclosure. I wonder if the SEC is listening? That would help, given the less than 100% prediction ability of their methods.

Meantime, US investors can use the Securities and Exchange Commission webpage where there is a link to Check Out Brokers and Advisors. It is Form ADV that provided the basis for Dimmock and Gerken's research.

In Canada, information similar to that used by the study in the USA seems not to be available to investors, judging by the Ontario Securities Commission website for Investors. The only data online seems to be whether the advisor has any actions outstanding against him/her, by which time it is of course too late if fraud is the issue. We must trust the regulator to protect us behind the scenes and attempt to do our own due diligence without assistance.