Are the payroll data overstating job growth in the U.S.?

Some pundits say the numbers are.

Over the past three months, the government's establishment survey has shown that 947,000 jobs have been created -- the most since March to May 2000. The gains, along with signs of rising inflation, have helped to boost the odds for a quarter-point interest rate hike by the


in June.

But some economists say the payroll numbers don't reflect reality, and they cite the Bureau of Labor Statistics' own words to bolster their case.

On its Web site, the BLS says its methodology to compute payroll data is "likely to have some difficulty producing reliable estimates at economic turning points or during periods when there are sudden changes in trend."

The main complaint about the payroll data centers around the birth/death model used by the BLS to estimate the growth of new businesses. The government samples about 400,000 individual worksites to compile the payroll data, but the numbers are adjusted each month to account for employment growth generated by new business formations.

Because there is a lag between an establishment opening for business and its appearance on the sample, the BLS uses a model to estimate this growth. The model is relatively new, having been phased in over recent years and only fully implemented in June 2003. The birth/death method replaced the BLS' "bias adjustment" model, which was considered less accurate.

In May, the BLS estimated that 195,000 new business jobs were created while total nonfarm payrolls came in at 248,000. In April and March, the government estimated that 270,000 and 153,000 new business jobs were created, respectively. Total nonfarm payrolls reportedly rose 346,000 in April and 353,000 in March.

According to these numbers, 65% of the new jobs created in the last three months have come from the birth/death model estimations.

The BLS says it is unfair to compare the birth/death estimates with total nonfarm payrolls because total payrolls are seasonally adjusted and the birth/death figures are not. But Doug Henwood, co-editor of the

Liscio Report

, said the results are practically unchanged when compared on an apples-to-apples basis.

"A matter of some concern is that a very large portion of employment growth in recent months has been the result of imputations from this birth/death model," he said. "So if they're not right, if this model isn't any good, then a lot of what we've seen as acceleration in employment may not have happened."

Of course, the survey could just as easily be underestimating the strength in the job market. When payrolls were poor at the start of the year, bullish economists and Republican politicians suggested this was indeed the case.

Still, Lacy Hunt, chief economist at Hoisington Investment Management, doesn't believe that's true today. He noted that the household survey has been much weaker than the establishment survey in recent months, and said the divergence may have something to do with the BLS' calculations.

The household survey, which is a sample of about 60,000 homes, was hailed as the more accurate measure of employment earlier this year because it reflects the growth in new small businesses, without any guesses having to be made.

Over the past three months, this survey has shown that 471,000 jobs have been created, far short of the 947,000 new jobs reported in the establishment survey.

"We know discrepancies emerge from time to time," Hunt said. "Payrolls were lagging the household

survey prior to March and some of it could be catch-up."

Still, he said the discrepancy is extremely wide and suggests that payrolls might be inflated. "When you see that difference, you cannot help but wonder whether some of it is originating in the birth/death model," he said.

The details of the household survey have certainly been less encouraging than those in the establishment survey.

In May, the household survey showed that employment rose by 196,000 but 137,000 workers were classified as self-employed. Hunt said many of those workers are self-employed because they can't find a payroll job and they won't have health or retirement benefits.

Full-time employment actually fell during May while part-time employment increased, the household survey showed. Those working part-time because they couldn't find a full-time job rose by 134,000. Meanwhile, 39,000 more people were out of work in May compared to April, and both the average and median duration of unemployment increased last month.

In a speech at the International Monetary Conference in London, Fed Chariman Alan Greenspan noted that "the proportion of increases in temporary workers relative to total employment gains has been unusually large, suggesting that business caution remains a feature of the economic landscape."

Raymond James' chief investment strategist, Jeffrey Saut, believes the payroll numbers might be exaggerating the strength in the labor market.

The payroll data showed a broad-based recovery in the job market last month, with factory employment recording its biggest gain in almost six years. Aggregate hours worked rose a solid 0.3% and the factory workweek rose to 41.1 hours, the most since October 2000.

"While we don't really understand most of the formulas the government uses to calculate various economic numbers, we do have a sense that there is a measurement problem in some of their techniques," Saut said.

Mizuho Securities' director of research, William Quan, said the BLS birth/death model has drawbacks because it assumes that historical patterns will continue in the future.

The BLS uses data on business deaths to derive a figure for the number of business births occurring during a particular month. Business births and deaths have tended to move together over time. Then the government uses historical data to estimate the residual impact of net births not accounted for in the first calculation.

Although this model can be imprecise when there are sudden shifts in the trend, Quan believes that it has proven to be more accurate than its predecessor and said the model has significantly reduced annual revisions to total nonfarm payrolls.

"While only time will tell the appropriateness of the BLS' new model, as of right now it appears that the BLS is taking large steps in the right direction to create more accurate and reliable labor market data," he said.