Open source research in finance

Eighteen months ago, Bill Ackman of the Pershing Square hedge fund released his “Open Source Model” providing detailed data on 30,499 tranches of 534 CDOs to which the big US bond insurers (MBIA and Ambac) had exposure. On the basis of this model, he claimed that MBIA and Ambac were insolvent. At the time, many regarded it as a publicity stunt; after all Ackman was heavily short these insurers and was merely talking his book.

While many read Ackman’s letter, few bothered to download the actual data. This was partly because the data was really huge (110 megabytes) and the letter warned that each recalculation of the model took about half an hour on a typical workstation. Moreover, the yousendit link where the data was uploaded expired within a few days. In any case, after the near-collapse of the bond insurers, people lost interest in the model.

However, last month Benmelech and Dlugosz published a paper on what they called “The Credit Rating Crisis” which relies partly on this data to document the failures of the rating agencies. Among other things, Benmelech and Dlugosz show that CDOs rated by only one rating agency were more likely to be downgraded than those rated by two or more agencies. They also confirm what was well known anecdotally about one particular rating agency being worse than the others.

This suggests that open source research can work in finance. One way to get transparency about the balance sheet of financial institutions might be for the regulators to create large detailed databases which anybody can analyse. I think several gigabytes of data would do the job, but even if the data grows to a terabyte or more, it would be well worth the effort.

Updated: Changed “collapse of the bond insurers” to “near-collapse of the bond insurers”

Posted at 12:09 pm IST on Wed, 29 Jul 2009         permanent link

Comments