By Bloomberg News
November 24, 2006
SourceWASHINGTON -- Inside the U S Federal Reserve headquarters, a small team is testing a forecasting program that does the work of hundreds of economists. Never before has the Fed been able to crunch in real time such a large mountain of data -- as many as 150 indicators -- to divine where the economy is headed.
Chairman Ben S. Bernanke is pushing the "factor model" program -- so named because it reduces everything from home sales to mining capacity into a few weighted averages for making predictions. The Fed could use the help: Its gross domestic product forecasts, which influence its interest rate decisions, have missed the mark by an average of 1 percentage point since 2000.
"It's a powerful tool that can potentially improve the Fed's forecasts," said Richard Clarida, a global strategic adviser at Pacific Investment Management Co. and Columbia University economist who developed a factor model as an assistant treasury secretary in 2001. "Forecasting, especially in real time, is a challenge, and these programs are not swayed by the emotions or the conventional wisdom of the moment."
Bernanke called the models "especially promising" in a speech in 2004. He should know. In 2000, as an economist at Princeton University, Bernanke created one that examined 78 economic indicators. He found that its short-term inflation and unemployment predictions were about as accurate as those produced by some 200 economists at the Fed, according to his published paper.
Bernanke, 52, also found that computers have their limits. As part of his research at Princeton, he ran a program to see what would have happened if a computer had set monetary policy from 1987 to 1998.
Forecasts from Bernanke's factor model were fed into the program, which adjusted rates based on certain rules. The result: Inflation and unemployment fluctuated by larger amounts than in real life, proof that Fed officials are better than software at making calls on interest rates.
"We find this evidence for human superiority comforting," Bernanke wrote.
In February 2006, when Bernanke became Fed chairman, the economy was speeding ahead, expanding at a 5.6 percent annual pace, and inflation was holding steady.
He gained credibility by halting the two-year run of interest rate hikes in August as growth started to slow. Now, he faces a more daunting set of facts: A plummeting housing market has raised the specter of a recession while inflation has nudged up.
Economists aren't providing much clarity. Their predictions for 2007 are sharply divided compared with their earlier calls for 2006. JPMorgan Chase & Co. argues that inflation will spur benchmark rate increases up to 6 percent. Goldman Sachs Group Inc., taking the opposite position, sees the distinct possibility of a recession and rate cuts to 4 percent. According to the median forecast of economists in an October-November Bloomberg News survey, officials will lower rates by half a percentage point to 4.75 percent in 2007.
Factor models may help officials with those twists and turns . "It makes it easier to see margins on which we could improve our reading on the ongoing state of the economy," said Jeffrey Fuhrer, research director at the Federal Reserve Bank of Boston.
The models emulate in a way how Bernanke's predecessor, Alan Greenspan, discerned economic trends from reams of data, says said Princeton's Watson. "Greenspan was someone who had great intuition and understanding about the economy," Watson said.
© Copyright 2006 The New York Times Company