The Financial Crisis: Junk In, Junk Out

Last Updated Sep 26, 2008 1:31 PM EDT

The Financial Crisis: Junk In, Junk OutThe reasons underlying the US financial crisis seem to be revealed hourly, layer by layer, like a stinking onion. But here's a question I haven't heard adequately explained.

Investment banks and lending institutions have at their disposal some of the most sophisticated computer tracking and modeling programs this side of the Pentagon. Surely in the glow of 31-inch computer displays, analysts were tracking the fundamental heartbeats of the financial industry such as mortgage default ratios, right? Or were the deregulated financial product innovations so complex that no algorithm could paint an accurate picture?

On Harvard Business Publishing, Tom Davenport offers his opinion in the post Is This an Analytics-Driven Financial Crisis? He thinks:

  • Some financial institutions had the right data, but didn't study it enough.
  • Some institutions gamed their own data to produce rosier scenarios.
  • Some models were primed with the assumption that housing prices would continue to increase.
  • Risk analytics are not as powerful as they need to be.
The takeaway:
"Going forward ... financial services organizations need to radically change their analytical focus," says Davenport, a Babson College professor. "They need to incorporate 'model management' -- the systematic capturing and monitoring of analytical models -- into their businesses. They need to be much more explicit and transparent about the assumptions behind models. They -- and their regulators -- need to be skeptical about the ability to model and manage risk in extraordinary circumstances. And financial firm executives need to learn much more about the models that are running their businesses."
This should also be a lesson to executives in all other businesses. We all rely on analytics to one degree or another to gauge the market, assess our own performance, and analyze the strengths and weaknesses of competitors. Are we looking at the right numbers? Do we constantly test assumptions underlying our models?

(Google Analytics image by manop, CC 2.0)


(This post first appeared in BNET's The View From Harvard Business.)

  • Sean Silverthorne

    Sean Silverthorne is the editor of HBS Working Knowledge, which provides a first look at the research and ideas of Harvard Business School faculty. Working Knowledge, which won a Webby award in 2007, currently records 4 million unique visitors a year. He has been with HBS since 2001.

    Silverthorne has 28 years experience in print and online journalism. Before arriving at HBS, he was a senior editor at CNET and executive editor of ZDNET News. While at At Ziff-Davis, Silverthorne also worked on the daily technology TV show The Site, and was a senior editor at PC Week Inside, which chronicled the business of the technology industry. He has held several reporting and editing roles on a variety of newspapers, and was Investor Business Daily's first journalist based in Silicon Valley.