Data Underestimates New Jobless Claims Reuters Tuesday November 18, 4:49 pm ET By Andrea Hopkins WASHINGTON (Reuters) - The U.S. Labor Department, at the urging of Federal Reserve Chairman Alan Greenspan, is working to fix a statistical quirk that causes the government to routinely underestimate the number of newly unemployed Americans each week, a department official said on Tuesday. Each Thursday, Labor issues its count of Americans filing initial claims for state unemployment benefits in the latest week. In 51 of the past 52 weeks, it has also revised the previous week's number upward, making the picture less rosy than it originally appeared. The weekly jobless claims figures are closely watched by financial markets as the most timely measure of U.S. employment. The data's importance has climbed in recent months as concerns about a "jobless recovery" from the 2001 recession threatened to undermine economic growth and the Fed warned that unemployment could dampen vital consumer demand. But claims have recently dipped below the 400,000 level that economists see as the divide between a deteriorating job market and an improving one, suggesting employment may finally be picking up. While the weekly revisions are small -- averaging about 4,700 additional claims a week, or a little over 1 percent of the total -- the official said the Labor Department is worried about the problem. "Obviously we're aware of this and we're actively working with the states to identify the source of the difference and correct it," the Labor official, who asked not to be identified, told Reuters, adding that Greenspan "is aware of this bias and wants it fixed." The official said high unemployment and the government's decision when the economy was weak to extend the duration of its benefits program have taxed Labor's efforts to fix the statistical glitch. "Probably we haven't addressed this situation as quickly as we might have liked to," the official said. "We're making progress." A spokesman for the Federal Reserve declined to comment on whether the Fed was pressuring Labor to fix the glitch. PINPOINTING THE PROBLEM The problem with the data, the official said, is linked to a process Labor uses to avoid overlapping state claims, and can be blamed in part on the speed at which it publishes the figures -- just five days after the end of each week. The so-called "interstate adjustment" is designed to weed out any duplicate claims, such as when a person is laid off from a job in New York but lives in New Jersey. "Conceptually they should cancel each other out. There are two states involved, one reports to us for the advance number and one should report to us when we revise it," the official said. "But when the interstate sorting takes place, we believe that this revision pattern has established itself. So that's where we're focusing." James Glassman, senior U.S. economist at J.P. Morgan Chase, said Labor could take the easy way out by simply adjusting the data by a small amount each week to lessen the size of the revision. "Apparently Greenspan has been pressing these guys to fix this problem, so they'll either do it by figuring out what the real problem is, or, if there is a systematic bias, they'll just build in a certain bias for it," Glassman said. The Labor official said statisticians have considered this but would prefer to solve the underlying issue. Despite the glitch, Glassman said economists would rather have the weekly number in a timely way than to have a delay for more accuracy. He said the pattern of upward revisions is so consistent that economists mentally increase the number themselves when it is released. "I think it is useful to have it published frequently even if it is off a few thousand, because really we're not interested in (a difference) of three or four or five thousand one way or another. We're interested in the bigger wave that is coming ... and the (employment) trend has still been decidedly better even with the revisions." http://biz.yahoo.com/rb/031118/economy_jobless_3.html