Professor Cole's Finance Blog

On Friday, the Bureau of Labor Statistics (“BLS”) released its monthly report “The Employment Situation,” which provides a summary and statistics on employment and unemployment during the previous month. The  headlines numbers for January were a loss of 20,000 payroll jobs but a decline in the unemployment rate from 10.0% in December 2009 to 9.7% in January 2010. Most people will scratch their head and ask how unemployment can decline when 20,000 jobs were lost.

The simple answer is that these contradictory numbers come from two different surveys conducted by the BLS each month. The payroll employment number comes from the “establishment survey,” which is a monthly survey of large businesses that employ most workers. The unemployment number comes from the “household survey,” which is a smaller monthly survey of households. The larger establishment survey is more accurate because of a larger sample size, but it cannot provide information on unemployment because it is a survey of employers, not workers. The smaller household survey is less accurate because of a smaller sample size, but enables the BLS to estimate how many workers are unemployed.

During January, the household survey showed that unemployment declined by 430,000 workers and that employment actually rose by an astounding 541,000 workers at the same time that the establishment survey showed employment falling by 20,000. One reason that the two surveys can be going in opposite directions is coverage. The establishment survey only covers workers on the payrolls of large established businesses, so it misses the self-employed, employees of small and newly established firms. This is critically important because research has shown that about two-thirds of employment growth takes place at small businesses, especially at newly established entrepreneurial firms. The half-million new jobs identified by the household survey may indicate that the economy has finally turned the corner and that small firms are on a hiring spree. I say “may” because of the large sampling error inherent in the design of the household survey. Just one month earlier, the household survey for December 2009 showed employment falling by 589,000, so we are still looking at a net loss of 48,000 jobs over two months. We can be hopeful, however, and await the February report for more information on which way the trend is really moving.

Now, more about the January report. Twenty thousand workers lose their jobs, but the BLS report states that “payroll employment was essentially unchanged.” Are these bureaucrats tone deaf or what? Also, the employment reports for November and December were revised, with November going from +4,000 to +64,000, but December going from -85,000 to -150,000. These revisions emphasize that the initial numbers reported each month are preliminary because not all establishments get their surveys in by the monthly deadlines. Instead, they get counted later and then the BLS releases revised numbers.

Also in January, construction lost 75,000 jobs but temporary workers gained 52,000 jobs. Now that’s change we can believe in, isn’t it! Of course, the federal government grew, adding 33,000 jobs. The average work week increased by 0.1 hour to 33.9  hours. This is another good sign, but increasing the number of hours worked by existing employees enables firms to avoid hiring new workers, keeping unemployment at elevated rates for a longer period. Average hourly earnings rose by 4 cents, up 2% over the past year.

U-6, the broadest measure of unemployment provided by the BLS, declined from 17.3% in Dec to 16.5% in Jan. on a seasonally adjusted basis. The 9.7% number is the U-3 measure of unemployment, which only includes unemployed members of the labor force. To be counted in the labor force, a worker has to actively look for a job during the previous four weeks; otherwise, they “drop out” of the labor force. Also excluded are employees working part-time even though they want to work full time. U-6 includes these discouraged and underemployed workers, but even U-6 excluded discouraged workers who have not actively looked for a job for more than one year. Some estimate that unemployment actually exceeds 20% when these long-term discouraged workers are counted as unemployed.

All of these numbers are “seasonally adjusted” by BLS staff using a statistical model that “smooths” out the month-to-month variations in employment numbers. For example, we know that employment jumps every June when high-school and college students look for summer jobs and then declines every September when they go back to school. The problem is that these “seasonal adjustments” allow for considerable mischief should unscrupulous or politically motivated staffers tweak their models to make things look better than the real numbers show. For example, U-6 actually rose from 17.1% in December to 18.0% in January on an unadjusted basis, so that the January unadjusted U-6 was of 9% higher than the seasonally adjusted U-6. Unadjusted, the U-3 measure of the unemployment rate would be 10.5%, not 9.7%. Such a large adjustment factor raised questions about the validity of the seasonally adjusted series. The January employment summary notifies us that the BLS just updated its seasonal adjustment model and its net birth/death model that is used to account for its undercoverage of small and newly established firms. Hmm.

Also in the January report is the result of its annual benchmarking process, whereby it reconciles the results from the establishment survey with actually unemployment records, which are not available on a timely basis, but are far more accurate. As a result of this process, payroll employment for March 2009 was revised downward by 930,000 and December 2009 employment was revised downward by 1,363,000!!! The numbers as originally reported and as revised cleary show that the establishment survey was consistently biased in the direction of higher employment. They appear in Table A of the BLS report.

The January report also notifies us that the household survey also was adjusted to reflect new population estimates from the U.S. Census and that this adjustment renders comparisons across time unreliable. So, how are we to ascertain if the newly announced decline in unemployment rate from 10.0% in Dec 2009 to 9.7% in Jan 2010 is actually an improvement, as it would appear.