February 4, 2005 | WebMemo on Economy
4, 2005, to reflect new employment data.)
Did the economy create jobs during President George W. Bush's first term?
The answer is yes, no matter which employment survey is used to measure jobs. But a lingering controversy over which of the two Bureau of Labor Statistics (BLS) surveys is the better measure continues to cloud the issue. The survey of households, which contacts people directly, reports a net increase of 2.37 million employed Americans since President Bush was sworn in. The payroll survey shows a net gain of 119,000.
Official data were published this morning for January 2005, the final month of Bush's first term. According to preliminary data for January, there were 146,000 new payroll jobs added last month. BLS also added 203,000 additional payroll jobs due to its annual correction of the survey's benchmark. If we also add in the 250,000 to 1,000,000 jobs "lost" to changing turnover rates, then payrolls are solidly in the black for the first term by at least half a million.
The big shock today was the drop in the unemployment rate to 5.2 percent, an excellent indicator of real strength in U.S. labor markets. Real earnings rose as well, and the duration of unemployment spells fell, but the headlines in the mainstream media are likely to highlight the fact that payrolls came in below expectations, which by this time should be recognized as the norm. Payrolls have been disappointing for many years, and one more month of disappointment is just another chink in the payroll survey's reliability, which is why it really should not be the measure of net job creation in this presidency.
The real question policymakers should be asking is not the political one-"Is Bush the new Herbert Hoover?"-but the economic one: "What structural shift caused a statistical earthquake in economic indicators in 2002 and 2003?" The two middle years of the Bush presidency mark the period of lingering controversy in employment data (See Chart 1).
In 2002 and 2003, payrolls were stuck in a "jobless recovery" while the number of working Americans, as measured by the household survey, grew by two million. Last summer, the disparity between the two surveys grew so great that the Labor Department was forced to publicly defend its payroll survey.
The household survey said employment was up 629,000 in July 2004, while payrolls grew a meager 32,000 jobs. Today, BLS says that payrolls actually rose by 83,000 that month. An even better example is the preliminary estimate of 248,000 new jobs last May, which today was cranked up to 419,000. All this revision to the payroll numbers is par for the course. Payroll data are always revised for the first two months after their preliminary release and then again annually to update methodological quirks. Household data are annually revised as well, being especially sensitive to population estimates, which was announced today as well-a miniscule decline of 8,000. But even years of revisions have not changed the long-term picture of divergence between the two surveys.
In retrospect, payrolls began to reflect recovery in September 2003, and since then the two surveys have been generally in sync. But this does not mean all is well and the controversy is over. Something important happened to make the two surveys diverge in the first place.
What is at stake now is not only how half of the President's first term is characterized, but also how the President's economic policies during the first term are evaluated. For example, the 2001 and 2003 tax cuts stimulated the economy, but bad payroll data serves as ammunition for critics who say those tax cuts produced few jobs.
Explanations for the Divergence
In July of last year, BLS began publishing a document every month to try to reconcile the recurring differences between the two surveys. The document essentially peels away workers who are counted in the household survey but not on payrolls, such as farmers and self-employed consultants. Many economists expected that this apple-to-orange reconciliation would resolve the puzzle. But amazingly, when a "payroll-concept" version of household data is "reconciled," the divergence actually increases by roughly 300,000 jobs.
So what happened at the end of 2001 that caused a level shift in payrolls? We can rule out the recession itself as a cause of the household-payroll divergence, because it hit in early 2001 and was brewing long before then. The seismic event that seems most likely to have changed work behavior is the impact of 9/11 and the war mentality that followed.
The idea of phantom jobs in the payroll numbers stems from the theory that 9/11 caused a reduction in job-changing. As explained in the Winter 2005 issue of Public Interest,
Job-changing from one employer to another, which had averaged 3 percent per month in the 1990s, declined by about 0.2 percentage points per year after 2001, settling at 2.4 percent in 2003, where it remains today. This seemingly small change meant…roughly one million fewer workers were being were being double-counted on payrolls, a statistical change that the payroll survey registered as one million 'lost' jobs.
day, the payroll survey mismeasures jobs and job losses by not
correcting for different rates of job turnover in its methodology.
BLS courageously acknowledged the turnover problem in July 2004 and
is working to define its scope. This could have significant
implications. As the BLS reconciliation study published January 7,
2005, put it, "If the rate of job-to-job movement changes
substantially over time, it could impact trends produced from the
One thing that all policymakers should agree on is the need for accurate economic data. Stimulating an economy at the wrong time or failing to see economic weakness until it is too late can wreak avoidable damage on the economy. That is why the recent efforts by Rep. David Dreier to help BLS enhance its methodologies are a step in the right direction. For the time being, though, currently published payroll data should be reported with an asterisk. Americans should not be misled about how strong their economy actually is.
Tim Kane, Ph.D., is Research Fellow in the Center for Data Analysis at The Heritage Foundation.