EVENTS

Date:	Monday, May 10, 2004
Time:	9:00 AM -- 11:00 AM
Location:	Wohlstetter Conference Center, Twelfth Floor, AEI 1150 Seventeenth Street, N.W., Washington, D.C. 20036

May 2004

How Jobless Is Our Recovery?

Over the past year, stock markets have climbed, productivity and GDP have soared, but job creation appears to have lagged. While job creation took off in March 2004, the cumulative record suggests that this recovery still lags significantly behind past experience. Or does it? In a new paper, Tim Kane of the Heritage Foundation explores possible design flaws in government statistics that may lead to the misinterpretation of job creation. Correcting for these flaws, he finds that the latest recovery is not nearly as "jobless" as it first seemed. Many government officials, however, have disagreed with Kane's analysis and have argued that the negative statistics cited are based on the best possible measure of employment. At a May 10 AEI conference, economic experts considered whether the United States is truly experiencing a jobless recovery or whether the primary method of measuring job creation is the real problem.

Tim Kane
Heritage Foundation

The question of a jobless recovery boils down to the statistical validity of the two applicable surveys: the payroll survey (CES) and the household survey (CPS).  There is a sharp divergence between the two surveys.  It is claimed that we are in a "payroll" jobless recovery, but according to the household survey, there have never been this many working Americans.  And while the CES is riddled with too many measurement problems to be an effective gauge of the workforce within the United States, the CPS talks to workers directly, thereby reducing the risk of any corporate error. 

In measuring the modern economy, the CES often over-counts or miscounts its sample. If you change jobs, the probability that you are on two payrolls in a given month increases, which can lead to a good deal of statistical noise in the survey. Other factors contributing to CES errors include the frequency of Employer-to-Employer flow, a change in pay periods (e.g. from weekly to monthly, monthly to bi-weekly, etc.), and the frequency with which employers update their payrolls.  These all increase the risk of an over-count, thereby devaluing the CES as a data source.

The difficulty of monitoring self-employed workers further reduces the accuracy of the payroll survey. Since the mid-1990s there has been an insurgence in the workforce of consultants and LLC employees. As the payroll survey does not count the growing pool of non-payroll consultants and systematically double-counts people who change jobs, there is an increased need to give added weight to the household survey-a source of data that is less prone to such large misestimates-in order to attain reasonable estimates.

There is no way to truly compare the two surveys given all the conceptual differences.  The household survey has a smaller, direct sample while the payroll survey has a large, indirect sample that over-counts job turnover and miscounts self-employed and consultant workers.  These problems create the illusion of one to two million lost jobs since November of 2001.  The household survey, the better measure of total employment in the United States, shows a record high 138.6 million working Americans and the creation of 2.2 million jobs since the recession ended.  This blind acceptance of the payroll survey as the standard by which to measure the state of the U.S. labor force has masked the increased job creation in the past months.

Tom Nardone
Bureau of Labor Statistics

To be honest, the Bureau of Labor Statistics (BLS) does not have a clear explanation for what is going on because this is so far outside the historical norm.  However, it is not fair to single out one survey as being "wrong" or "erroneous." Both surveys appear to be working as they should, and when the two surveys are adjusted using the "apple to apple" approach by removing agricultural workers and the self-employed, the adjusted CPS employment is in relative sync with that of the CES.

Self-employment is not included in the payroll survey, nor was it ever intended to be. That is not the function of the survey.  We have, however, conducted a contingent work supplement that suggests that overstatement is far less than originally thought.  For that matter, how much has self-employment truly increased over the past four years? Yes, it is up since the end of the recession, but it is simply recovering the ground lost during the recession.

One of the main strengths of the CES is its annual benchmark to a universal count of employment.  Other improvements have recently been added, including an estimate of growth based on a growth/death model.  In fact, the latest benchmark (March 2003) showed that the monthly CES numbers, as well as the quarterly numbers, effectively tracked universal count of employment.

There are noted differences in measured unemployment growth between the two surveys, but by examining the differences, we glean a better understanding of the working of the labor market and the data series.  While complete accounting of the differences is probably impossible, both series provide important information about the state of the labor market.  Both surveys serve a purpose.

David Malpass
Bear Stearns, Inc.

In general, I accept the numbers from each survey.  The ultimate goal is to create an economy with increased employment, and both surveys are showing strong job growth and a relatively full employment expansion.  This follows on the heels of both a period of unsustainable extremes and a recession during which the unemployment rate peaked at 6.3 percent, noticeably lower than the last recession. Thus, we arrive at this strong downward trend in the U.S. unemployment rate over the past forty years.

With an increasing ratio of full-time to part-time workers, not only has the quality of jobs improved, but also the number of jobs.  There have been any number of part-time jobs (296,000 in March 2003) that have falsely increased the number of jobs in the payroll survey.  Yet, in April, net new jobs increased, while the number of part-time jobs decreased drastically.  Also, nominal wages have increased faster than the inflation rate (real-wage growth) even after a recession, which is unlike past post-recession periods.

All of these facts provide evidence in support of a stronger economy than anything the general public has been led to believe in over the past two years.  This durable expansion, which will spread into 2005, makes it extremely difficult to make a case for a jobless recovery, regardless of which data you use.

Bruce Fallick
Federal Reserve Board of Governors

Every time we question the divergence between the CPS and CES numbers, we end up concluding that a great deal of it has to do with problems in the measurements of the CPS.  Explaining what happened between 1998 and 2000 has very little to do with turnover, as does the past four years. More than likely, this stems from a problem measuring population-often an enormous problem in the past.

There is no simple way to look at the two surveys and point to the better one.  Even if there were, that is the wrong way to approach the problem.  The simple answer is to use both surveys for what they were originally designed to measure: the CES excels at measuring the number of jobs as a result of the annual benchmark, and the CPS design lends itself to measuring ratios-the labor force participation rate, the employment-to-population ratio, and the unemployment rate.

The recent divergence between the two surveys probably does not have a simple explanation that allows us to define this as a jobless recovery or not.  In my opinion, however, that is not really the pertinent question.  We should be asking ourselves what has been going on in the labor market over these past three or four years.  To answer that question, we should use each survey in areas in which its individual design performs best.

AEI interns Bo Harrison and Robert Jeffrey prepared this summary.