Delivering the Data

But, of course, Greenspan doesn't make decisions on anecdotal information alone. Inside the marble Fed building, 225 Ph.D. economists are dedicated to the single-minded task of understanding the U.S. economy and its relation to world economies.

The Fed's research division captures and monitors information from hundreds of corporate, government and university sources. Housing starts. Jobs created. Gas prices. Measurements of mood, such as the University of Michigan Consumer Sentiment Survey, which examines consumers' confidence in the economy and their likelihood to spend money on cars, TVs and clothes.

The Fed also buys mortgage and credit card information from banks and lenders to study how and how much consumer spending is being financed.

All the data must be consolidated, analyzed and delivered to Greenspan and the other members of the board. They meet formally eight times a year, but receive reports daily.

Sandra Cannon, Greenspan's chief of economic information management, makes sure reports on the GDP, consumer prices and other economic indicators are gathered as soon as they're available and loaded into the Fed's Forecasting, Analysis and Modeling Environment (FAME) database system.

Is the economy leaving your tech department behind? Click here to find out.

A number of top-tier banks, investment houses and energy trading firms such as Credit Suisse First Boston and Morgan Stanley use FAME for financial analysis. Disaster recovery specialist SunGard Data Systems acquired the system in 2003 from FAME Information Services.

Once numbers are uploaded to FAME, economists at the Fed begin their calculations.

Unlike a typical database, which stores data in categories such as date in one column and price in another, the system builds time into each piece of information. It can store, for example, that the rate on a five-year variable mortgage was 5% on Feb. 1, 2005, and 5.25% on Feb. 2, after Greenspan's tweak.

The payoff for financial organizations, says SunGard FAME product specialist Kenneth Kunin, is that time-based calculations can be done faster and with less programming. Time is already attached to the information, making it easier to figure out, for example, how much interest might be earned on a bank's deposits from September 12 to March 16.