The Data Reside in the Field

In the New York Times' “Needed: A Clearer Crystal Ball” (Sunday Business pg. 4) professor Robert Shiller claims that if we sharpen our risk measurement tools we will better understand the risk of another financial shock. He argues that improved data collection can substantially increase the predictive power of our financial models.

As mathematicians we must agree with his sentiment: more complete data is always useful. But as market participants we wonder if professor Shiller has missed out on a valuable lesson to be learnt from this financial crisis.

One key lesson was that the source of the failure was neither data-driven nor model-driven, but rather a direct result of the expected behavior of poorly incentivized parties.

In fact, one could argue that for the large part the data were comprehensive. The models were highly sophisticated, perhaps too sophisticated. But what caused the crisis was that originating parties were financially rewarded for structuring and selling low quality mortgage loans. The incentive was clear and by mid-2005 the FBI was already commenting on the pervasive and growing nature of mortgage fraud.

Misrepresented financial documentation skews the data and cannot be spotted simply by poring over ever more abundant realms of data: you have to go into the field itself to follow the incentives.

The financial downturn was made worse by financial institutions’ lack of confidence in the creditworthiness of their counterparts. Absent a level of certainty as to the true nature of others’ balance sheets, a lending freeze precipitated an illiquidity crisis. But a more thorough examination of data won’t tell you what resides off-balance sheet: you have to understand the prevailing accounting environment (that led to the mechanical reproduction of the negative basis trade) and the fundamental nature of the opaque shadow banking system.

It is a concern that intellectuals and academics risk being lulled into a false sense of security based on their access to copious amounts of statistical data.

Copious analysis of imperfect data is unlikely, alone, to help our regulators hone in on an inevitable crisis – much less prevent it. Let's not strive to build complex economic models whose success hinges on sensitive data. Let us rather encourage a keener appreciation of the limitations of data, the intentional proliferation of informational asymmetries, and the incentives that can cause a meltdown.