Labor market statistics can be confusing to nearly anyone who has not had a good introductory macroeconomics course. Take, for instance, the regular disconnect between the number of jobs created (or lost) and the up-and-down fluctuations of the unemployment rate over the same period of time.
So whenever the Bureau of Labor Statistics releases monthly numbers that don’t seem to make sense, econ profs get bombarded with questions. That was true when January’s unemployment rate and employment numbers were released Feb. 7.
The number of “employees on non-farm payrolls” was up a paltry 113,000 over December. This was substantially less than most private-sector gurus had predicted. The U.S. unemployment rate dropped to 6.6 percent from 6.7 percent. In general, markets and the media interpreted the report as one of weak recovery. There was much speculation about whether this would lead the Federal Reserve to hold off on its tentative steps at tightening monetary policy.
However, as observant readers noted, there seems to be no consistent relationship between the number of “new jobs” and a change in the unemployment rate. Over the past several years, one can find instances when a high job growth number didn’t drop the unemployment rate at all. What gives?
There are several factors in this.
First, a quick explanation of key terms in these employment indicators. All deal with people who are civilian, non-institutionalized and age 16 and over. So anyone in the military is excluded, as is anyone in jail, a mental institution and the like. If you’re in this group and you work for someone else for pay or in your own business, you are “employed”; if you don’t have such employment (or work) but are actively looking, you are “unemployed.” Add these two together and you have the “labor force.”
If you don’t have a job and are not trying to get one, you are “out of the labor force.” Most people not in the labor force are retirees, students or parents taking care of children in the home. But all non-institutionalized civilians past their 16th birthdays have to fall into one of these three categories: employed, unemployed or out of labor force.
The widely reported “unemployment rate” is the number of unemployed people divided by the total number in the labor force. In turn, the labor force divided by the civilian, non-institutionalized population age 16 or older gives us the “labor force participation rate.” And there also is an “employment-to-population ratio.”
The unemployment rate can drop if unemployed people get jobs. That reduces the numerator of the computation. But the rate can also fall if some people stop actively looking for work and thus “drop out” of the labor force. This reduces both the numerator and denominator, with the net effect of a fall in the percentage rate. That has happened frequently in recent years. Coincidentally, it also lowers the labor force participation rate.
But this did not happen in January. The LFPR actually rose slightly, and this was considered good news.
The source data also causes confusion. First, note that two surveys are conducted: a survey of households for the labor force and unemployment rate data and a survey of employers — officially called “establishments” — for the jobs-created data.
Because we cannot do a complete census every month to ascertain the work status of every individual in the country, we rely on statistical surveys in which we check a small fraction of the population to make an inference about the population as a whole. Such surveys never can be exact, and the smaller the sample, the less confident one can be.
For example, even though the survey used as the source of the non-farm payrolls number covers over 140,000 employers in some 550,000 distinct workplaces, the resulting data vary because we are comparing an estimate from one month to an estimate from another. For the 113,000 increase estimated for January, the Bureau of Labor Statistics is 90 percent confident that the true number, if we measured the whole population, would fall between about 33,000 and 203,000.
Because this “establishment” sample is much larger than the “household survey,” it gives more reliable data about jobs by sector. But it is not exactly the same as the “employment” number from the household survey. It does not include any farm workers and is less reliable in covering the self-employed or very small businesses. Over time, the two data series do track each other closely. But in any given month, they can diverge greatly.
Meantime, the labor force survey covers about 60,000 households out of the more than 120 million in a U.S. population of 317 million. That is one-twentieth of 1 percent. This household survey showed the number of “employed” people increasing by 638,000 and the labor force by 523,000. The number of “unemployed” dropped by 115,000 and the unemployment rate to 6.6 percent from 6.7 percent.
In general, one should not pay too much attention to month-to-month or quarter-to-quarter changes in any economic indicator. Trends of several months are more important, as are “year-over-year” comparisons to 12 months earlier. But because important decisions, such as private-sector business plans or Fed policy questions, hang on new trends, people naturally consider new data every month. But they should do it with caution.
There are further complications too technical to explain here, including how to account for regular seasonal patterns such as summer jobs for high school and college students or temporary retail employment before Christmas. This involves “seasonal adjustment.” Most data series are published in “seasonally adjusted” and “non-seasonally adjusted” tables.
It would be nice to have more certain data. But that is expensive. One of the first acts of the Reagan administration was to cut the sample size of the monthly household survey to save money. It did, but it also reduced the usefulness of the data. And there is little public call to “expand government” so as to make estimates of indicators such as these more reliable.