Inside the Mind of Alan Greenspan

Delivering the Good News, and the Bad

Most official government data comes out at 8:30 a.m. or 10 a.m. on a given day. GDP numbers, for example, are issued at the end of each month from the Commerce Departments Bureau of Economic Analysis at 8:30 a.m. Eastern time.

By 8:35 a.m., Cannons office has downloaded the figures from the bureaus Web site. By 8:45 a.m.—9 a.m. at the latest—they are uploaded into FAME. Cannons office then sends an e-mail to let staff know new figures are available.

Fed economists anticipate mornings when marquee numbers come out—such as the GDP or the Consumer Price Index—like baseball fans waiting for the latest box scores.

"Theyre ready and waiting and wanting to know if theres the least little problem [with their economic forecasts], and theyre more than happy to volunteer their assistance to type in numbers if they have to," Cannon laughs.

These days the Fed receives most data electronically, through downloads from Web sites or e-mail. Thats a major change from when Cannon joined the Fed seven years ago. Back then, people, not machines, gathered data, she says.

"We used to send a staff member and a driver to the various agencies to pick up a package that would have some form of media on it—a disk or tape—and then wait for them to drive all the way across town before we could start inputting and processing things," she recalls. Some reports, such as the Advance Monthly Sales for Retail and Food Services, came on paper and would be typed into the database.

This download-upload process of populating FAME is repeated 50 to 60 times per month, whenever new data comes out. When Cannons seven staffers arent shunting information around, they produce tables and charts with the latest figures to distribute across the agency, to the Feds economists, reserve districts, governors and, of course, to Greenspan.

Ingersoll-Rands CTO on how to get a return on technology investments—even during the toughest of times. Click here to find out how.

The chairman likes to get a copy of the actual data and the charts. "Basically, all the data that comes through my office winds up on his desk in one form or another," Cannon says.

Greenspan studies each report, looking for warning signals of inflation or the beginnings of an economic shift. By Feb. 1, for example, Greenspan had been expecting the strong economy, humming along in the latter half of 2004 with gross domestic product growth of 4%, to be creating significant numbers of new jobs. But it hadnt.

Job growth remained in the range of 125,000 new jobs per month in November, December and January, barely half what was expected. He uses the detailed reports to look for clues, such as which industries appear to be lagging in job growth, and asks for more number-crunching to find the answers.

Economic forecasting, though, is both science and art, says David Reifschneider, deputy associate director for the division of research and statistics at the Fed headquarters. The science is the sophisticated computer modeling that examines the data, compares it against past performance and predicts the future.

The art comes from personal insights, knowledge and intuition to deal with the shortcomings of science. Economic statistics can be wrong due to incomplete data or off-kilter sampling that only becomes evident down the road. A GDP number that comes out one month is usually revised the next month and again a month later, as more information comes in, is vetted and refined.

Greenspan has a discriminating feel for how statistics might be off, and weighs that sense against the other data and anecdotes presented to him. Often, he pursues a theory with a question. How could productivity be falling in the service sector, he asked in 1996, when the stories he gathered said that businesses were more productive than ever?

"Theres a degree of forensic science to economics," says Reifschneider, who has been with the Fed for 20 years. "When something unexpected happens, it causes you to pause and say, What was that all about? What were the forces that combined to make that happen?"

Fed economists have their own pet tools for deciphering and modeling the data stored in FAME, such as SAS/Insight from SAS Institute, based in Cary, N.C., and Regression Analysis of Time Series (RATS) from Estima, an economics software firm in Evanston, Ill. Regression analysis is a statistical technique to find correlations between several variables. You can determine, for example, whether and how much a 3- or 8-cent rise in gas prices affected consumer spending during a period when housing prices fell 20%.

However, three homegrown modeling applications do the heavy lifting of understanding the forces at work in the economy: Federal Reserve Board (FRB)/US, FRB/Global and FRB/Multi-Country (for the Group of Seven industrialized nations). Each model pulls on information from FAME and incorporates hundreds of equations to calculate how one factor affects hundreds of others.

Different variables are weighted to perform what-if analyses. How would a $57 barrel of oil impact inflation vs. a $52 barrel? Or, what would a drop in car prices do to consumer savings rates? The Fed would calculate, for example, the combined effect on the GDP of a .5% drop in productivity, a 1% increase in unemployment, a 4% increase in personal consumption expenditures (a measurement of how much of a persons income is spent meeting everyday needs), and a 2% increase in the Producer Price Index. Dozens of other factors might get tossed in.

Today, 70 Sun Solaris servers process that data. But by this summer, the Fed will have replaced the Sun machines with 85 less expensive Intel-based servers running Red Hats Linux operating system. Mike Cringoli, research computing chief, says the Fed was due to upgrade its servers, and smaller servers will allow each section to manage its own computing resources.

The results of the data analysis go into three color-coded books distributed to Greenspan, Fed governors and reserve bank presidents before their meetings.

The Beige Book provides a picture of current economic conditions based on the latest available data and anecdotes collected from the field.

The Green Book predicts how the U.S. economy will perform in the next year, including gross domestic product forecasts and the inflation outlook. Economists run the latest available information in FAME through FRB/US, then adjust the forecasts to reflect their own tacit knowledge.

The Blue Book recommends monetary policy actions. If unemployment is high, the economists will run scenarios using various assumptions, such as lower interest rates and buying back Treasury bills to put more money into the economy.

What Greenspan gets in those books is the combined analysis of almost 500 economists at the Fed and the reserve districts, who spent weeks crunching numbers. It is the collective wisdom of some of the best economic thinkers in the nation.

Story Guide:

Federal Reserve: Stats and Specs

Next: 'ZIFFPAGE TITLEWhat Do You Do ' >>

Discuss Inside the Mind of Alan Greenspan   >>> Be the FIRST to comment on this article!

>>> More IT Management Articles          >>> More By Mel Duvall