A new Labor Department survey shows that the U.S. labor market, described in recent months by many economists and policymakers as stagnant, has instead been a churning cauldron of change, with millions of people each month getting hired and fired.

In August, according to the department's closely watched monthly survey of business payrolls, the number of payroll jobs rose by 96,000, a small net change when the number of workers on private and government payrolls was 130.9 million. The self-employed and some other workers are not on payrolls.

But last week the department's Bureau of Labor Statistics released the August results of its new Job Openings and Labor Turnover Survey, which painted a far more dynamic picture. According to the survey, 3.7 percent of those at work at the end of August were hired that month. At the same time, 4.1 percent of those at work on the last day of July left their jobs during August. Of those who left, three-fifths quit and only one-third were discharged or laid off for more than a week, with the remaining fraction departing because of retirement, disability, death or transfer to another location.

The survey showed that 4.85 million people were hired and 5.37 million "separated" from their jobs, which includes the 3.22 million who quit and 1.78 million who were discharged or put on extended layoff.

Viewing the numbers is "like looking at the motion picture and not the snapshot," said BLS statistician Thomas L. Nardone.

The figures may even understate the amount of change, according to an analysis by two Federal Reserve economists of other BLS data collected each month in the survey of 55,000 households on which total employment and unemployment figures are based. The survey is based on a nationwide sample of 16,000 businesses that includes virtually all large companies.

The Fed economists, Bruce C. Fallick and Charles A. Fleischman, found that labor market turnover was so great over a seven-year period ended in 2001 that, on average, 6.7 percent of all workers employed one month were not employed by the same employer the following month. In 2000, for example, with employment around 135 million, that meant that roughly 9 million employees left their jobs in an average month, with most of them moving from one job to another.

Rick Clayton, program manager for the survey, said there are a variety of reasons that the estimates of labor turnover are different, including that self-employed, agricultural and household employees are not covered by the new survey but are covered by the household survey used by the Fed economists.

Whichever approach is closer to the mark, the churning of the labor market belies the impression of little month-to-month change. For instance, the BLS reported that its household survey found that 8.14 million people were looking for work in August but unable to find it. In September, the ranks of the unemployed dipped to 8.09 million, down 50,000 from the previous month, but then rose again last month to 8.21 million.

But given all the hiring and firing going on, it seems that the more than 8 million jobless in September were not necessarily the same 8 million unemployed in August or October. According to their study, the Fed economists concluded that in an average month, slightly less than half those counted as unemployed one month were still unemployed the following month. Their data suggest that in September, about 4 million workers who could not find employment in August either got a job or dropped out of the labor force, while a different 4 million joined the ranks of the unemployed.

Even with all the churning, however, a substantial number of people are remaining unemployed for months. Last month, 1.66 million unemployed people had been jobless for 27 weeks or more. And 355,000 people had stopped looking for work because they believed none was available, the Labor Department estimated.

Labor market experts understand that the regular monthly reports on employment and unemployment imply far less change than actually occurs each month. All the hiring, quitting and firing are added together, and the net result often shows little change from one month to the next. It is that net change that is seen in the snapshot.

The survey also offers one figure that does not appear in the regular monthly statistics: the number of job openings.

In August there were 3.49 million job openings for which employers were actively searching for workers. The services industry had the largest number of openings, with 1.5 million. Retailers had 642,000 openings. Manufacturers, which have been shedding employees for more than two years, trailed with just 295,000 openings. Governments, both federal and local, had 501,000 openings.

The industry with the greatest turnover was retail; 5.9 percent of employees left during August, 4.3 percent of whom quit. Meanwhile, 5.7 percent of retail employees at the end of August were hired during the month, according to the survey.

Claytonsaid it is too early for the survey results to tell a full story about changes in the labor market. The data go back only to December 2000 and cannot be seasonally adjusted. That means the August numbers cannot be compared meaningfully with those of preceding months to spot trends.

By sometime next year, enough data should be in hand to estimate seasonal adjustments and to allow the survey, which is still billed as "developmental," to become an official BLS statistical series. At that point, Clayton said, the survey will contain "very useful numbers" that will help economists better understand what is happening in the economy.