ven if the economy is on the road to recovery," Alan Greenspan warned last week, "the unemployment rate, in typical cyclical fashion, may resume its increase for a time." Mr. Greenspan, the Federal Reserve chairman, who will appear before the Senate Banking Committee today, did not mention that the lingering effects of high unemployment early in a recovery tend to be concentrated among the unskilled and minorities. This is true even though recessions are becoming more egalitarian.

Such a pattern was evident in the early 1990's. The recession officially ended in March 1991, with the unemployment rate at 6.8 percent. But unemployment continued to rise for 15 months and did not settle below 6.8 percent again until the end of 1993. Moreover, the rate rose from 12.3 to 13.5 percent for high school dropouts in the year after the recession ended, while for college graduates it stayed at 2.9 percent. The "jobless recovery" mainly involved the less skilled.

In part, unemployment remains high after growth resumes simply because more people actively look for work if they think their prospects are better. Having more active job seekers and fewer discouraged workers is not a sign of weakness. For this reason, economists often prefer to focus on the employment-to-population rate — the fraction of the population that is working.

Even with this measure, however, jobs often keep sliding after a recession ends. The employment rate did not hit bottom until nine months after the 1991 recession, and the slide was greater and lasted longer for less-skilled workers. The fraction of high school dropouts who were employed fell for three full years after the recession ended.

Jobs are slow to recover at the end of a recession because employers are not sure if improved conditions will persist, so they expand work hours rather than hire new employees. Many employers also "hoard" skilled workers during a downturn because it would be costly to hire and train replacements when conditions improve. Neither reason, however, accounts for why job growth is particularly sluggish for the less skilled when the economy picks up.

The economist Melvin Reder suggested a reason in 1955: in a downturn, many employers raise skill requirements for a given job, rather than cut pay. The less skilled thus find their job options even more limited until demand picks up smartly, while skilled workers take positions further down the job ladder.

Lawrence Katz of Harvard suggests another reason. "Think of a recession as a time when firms reorganize," he said. Reorganization tends to increase demand for skilled workers, who are more flexible. Furthermore, when companies introduce new technology as part of a reorganization, they tend to hire skilled workers to operate the equipment and release unskilled workers whose jobs are made redundant.

To some extent, the lingering pattern of unemployment after growth resumes, especially among the less skilled, is inadvertently reinforced by the major policy tool used to lift the economy out of recession — interest rate cuts by the Federal Reserve. With a lag, rate cuts stimulate demand for new capital and consumer durables. But this is a two-edged sword for workers. On the one hand, a rise in economic activity increases the demand for all factors of production, including workers. On the other hand, because machinery is cheaper, many companies replace some workers with machines, or hire fewer workers, because the machines can do the work at lower cost.

What's more, capital — especially high- technology equipment — and skilled labor are complementary factors of production, while capital and unskilled workers can be substituted for each other. In other words, highly skilled workers are hired to operate and service new machines, while less- skilled workers are let go because the machines can do their work.

The current recession — which many forecasters think may have already ended — is shaping up as if it consists of two distinct phases distinguished by how skilled and unskilled workers fared. In the first phase, which lasted from March to July 2001, the employment-to-population rate fell for those with at least some college training, from 74.9 to 73.9 percent, while it actually increased slightly, from 55.1 to 55.3 percent, for those with a high school education or lower.

In the second phase — the six months after July 2001 — the opposite happened: the employment rate stopped falling for those with at least some college education, and dropped 1.7 percentage points for those with a high school education or lower.

A similar picture holds by race and ethnicity. At the beginning of the recession, the employment rate fell more for whites than for blacks and about the same for whites and Hispanics. Since July, however, employment has fallen more for blacks and Hispanics. Likewise, unemployment rose for whites in the first phase and held steady for blacks and Hispanics, while in the second phase unemployment is growing around twice as fast for blacks and Hispanics.

The egalitarian nature of the first phase came as a surprise but is consistent with the egalitarian tilt of the early 1990's recession. This phenomenon probably resulted from the plunge in capital investment that apparently precipitated the downturn, the implosion of many dot-coms and the fact that the cyclically sensitive manufacturing sector is much more skill intensive than it was 20 years ago.

It appears, however, that the egalitarian phase is over. If recent trends hold, job opportunities will rise or remain steady for the better educated and continue to contract for some time for the less educated and minorities.

The message to policy makers should be clear: to avoid another jobless recovery, stimulate demand for less-skilled workers. This suggests that policies that spur investment, like an accelerated capital depreciation allowance, may prove to be counterproductive; lower capital costs would increase demand for skilled workers but reduce demand for less-skilled workers.

Senator Pete V. Domenici's proposal for a temporary cut in the payroll tax, which is borne disproportionately by low-paid workers and their employers, would do much more to increase job growth.

Although many have argued that the opportunity or desirability for fiscal stimulus has ended, if the immediate goal is to avoid another jobless recovery, a temporary cut in the payroll tax still has much to offer, especially versus initiatives that would subsidize capital investment.