Dean Baker & Heather Boushey: Misleading Job Numbers from the Heritage Foundation

To the Editor:

We are writing because we object to your decision to publish Tim Kane's column ("Labor's Lost Jobs," 4-7-04), which implied that the Bureau of Labor Statistics (BLS) payroll survey seriously understates job growth. This piece seems intended to create public confusion about basic economic data, when there is no foundation for such skepticism. Economists, regardless of their political leanings, would agree that virtually every claim in this article is either false or hugely exaggerated.

In his first claim, Mr. Kane argues that 1 million jobs have been lost in the establishment survey since 2001 because the rate of monthly job turnover has fallen by roughly 1 million, and that people are typically counted twice in the establishment survey when they change jobs. Presumably Mr. Kane knows that the survey asks employers about the number of workers on their payrolls during a reference week, not an entire month. This means that workers will only be double counted in the survey if they worked in two different establishments in the second week of the month.

Accepting that the rate of monthly churning has fallen by 1 million, this implies that the weekly rate of churning would have declined by approximately 250,000. Estimating conservatively that half of these job changers either take some time off between jobs, or at least work until the end of the week, the recession-induced reduction in churning would lead to a decline of no more than 125,000 in the number of jobs reported in the establishment survey since 2001.

Next, Mr. Kane points to other data in the BLS household survey suggesting a more robust labor market. He notes that the unemployment rate is low and then observes that “real earnings rose by 3 percent over the last three years. Jobless claims are 10 percent below their historic average, and that's without adjusting for population.” While the average real hourly wage rose by 2.6 percent over the last three years, this is primarily due to strong wage growth coming out of the late nineties boom. More recently, real wage growth has slowed sharply -- exactly as would be expected in a weak labor market. In the last two years, average real hourly wage has grown at just a 0.2 percent annual rate. In the last six months, average real wages have actually been declining.

As Mr. Kane notes, jobless claims have fallen, which does suggest a recent pick-up in the labor market. Lower initial unemployment insurance claims are likely due to fewer lay-offs. However, continuing claims tell a somewhat different story. The share of unemployment insurance beneficiaries who have exhausted their benefits -- that is, they are no longer eligible because their time has run out -- was around 33 percent in 2001; it rose to 43 percent late in 2002 and has not come back down. This rise in the percentage of unemployed workers who have exhausted their benefits is not evidence of a strong labor market.

The reason that current claims are below their historic average -- another point made by Mr. Kane -- is that there has been a long-term trend of declining eligibility due to tighter restrictions. This tightening of eligibility requirements would ensure that unemployment claims are below their historic average except in the most severe labor market conditions.

Mr. Kane's discussion of the current unemployment rate is also misleading. The BLS household survey shows that the number of people who say that they are looking for work but cannot find jobs is 5.7 percent of the workforce -- a relatively low level by the standards of the last three decades. However, this survey also shows that the share of the working-age population that is working has declined by 2.1 percentage points, a sharp drop of a magnitude that is only seen during periods of severe labor market weakness. The data in the household survey suggest that close to 4 million workers have simply given up looking for work, and the share of American men with a job is lower than it has been since 1988, both indicating anything but a bright labor market picture.

Mr. Kane then argues that the household survey shows an increase of 600,000 jobs since President Bush took office -- in contrast to the decline of 1.84 million jobs shown in the establishment survey. Surely Mr. Kane knows that this job growth is attributable to the fact that the BLS put these jobs into its count in the household data when it corrected for undercounting in the nineties (see http://www.bls.gov/cps/cpspopsm.pdf). The employment numbers in the household survey are entirely dependent on BLS assumptions about underlying population growth; the survey does not have an independent source for checking total employment numbers.

In contrast, each year, the establishment survey is benchmarked to an independent data source: the tax filings for state unemployment insurance. Since over 99 percent of employees nationwide are covered by this system, the benchmarking provides a virtual census of the U.S. workforce, guaranteeing that the establishment data cannot veer far from the true level of employment in the economy.

Next Mr. Kane focuses on the fact that the establishment data does not count self-employed workers. He argues that the modern economy provides for a wide variety of work relationships that would not be picked up as jobs in the establishment survey. The quality of these non-standard work situations can be debated, but the number is reasonably well measured in the self-employed category of the household survey. This category shows a drop of 164,000 between March of 2001 and March of 2004, a number that should be added to the job-loss shown in the establishment survey.

Finally, Mr. Kane does correctly note that the establishment survey can be in error when the economy reaches a turning point. (This is due to the fact that it does a poor job of picking up job creation in new firms.) It substantially overstated job growth as the economy slipped into a downturn in 2001 and again in 2002, and it understated job growth as the economy began to recover in the early nineties. However, the 900,000 understatement in 1992 cited by Mr. Kane was primarily attributable to a one-time technical error in the method used for counting workers at temporary employment agencies (since corrected), not an inherent flaw in the survey design.

If the economy is in fact recovering strongly, then it is likely that we will see some upward revision to the number of jobs, but based on past history, the revision is more likely to be in the neighborhood of 200,000 to 400,000 for the year through March of 2004, not enough to change the basic picture. And contrary to what Mr. Kane asserts in the article, we will not have to wait until next January; BLS should provide this data in August -- plenty of time for us to know the employment situation by the November election.

Many people have difficulty understanding economic concepts and data. The New York Times does not serve the public when it publishes a column that seems deliberately designed to increase this confusion.

Sincerely,

Dean Baker, Heather Boushey

Co-Director Economist

202-293-5380 x229 (W) 202-293-5380 x212 (W)

202-332-5218 (H) 202-483-4794 (H)

cc: Louis Uchitelle