The May Jobs Report was released today, and on the face of it, it indicates that 431,000 new jobs were created last month. Although this number was somewhat of a disappointment, as it did not show as much growth as had been estimated and hoped for, the result of the report was a flurry of media articles trumpeting that “May brings 5th straight month of jobs growth” and “U.S. adds jobs in May”, although, to be fair, the exuberance of these articles was tempered somewhat with somber commentary on the modest increases in private sector employment.
But a closer look at the May jobs report shows a situation deeply more concerning than what the media is describing. (Alas, the stock market isn’t falling for the spin either, it seems.)
Breaking the establishment report down, which is the source of the BLS’s number proclaiming job growth, we see the following:
Total Increase in Jobs: 431,000.
Total Government Job Increase: 390,000
Total Private Sector Increase: 41,000
But wait. We added a total of 411,000 census jobs in May — which means we actually lost 21,000 non-census government jobs. (Which means, not counting census jobs, there was only a net increase of 20,000 jobs total.)
And it gets worse. In the private sector, the largest growth was found in a category called “temporary help services,” with an increase of 31,000 jobs. Perhaps this is merely reflecting a May increase in summer jobs that will be gone again in August, such as camp counselors and resort employees? At any rate, it is not a strong indication of job market stability.
On the other hand, the unemployment rate at least has been reported as being modestly optimistic, falling from 9.9% to 9.7%, as shown in the household survey. It went from 15,260,000 unemployed in April to 14,973,000 in May, for a total swing of -287,000 unemployed persons. That’s good news, right?
Not so fast.
These two statistics — the decrease in unemployment and the increase in jobs — come from entirely different surveys. The figure showing a growth in jobs was taken from the establishment survey, while the May Job Report’s conclusion that there was a decrease in the unemployment rate comes from the household survey.
But while the household survey does show a decrease in unemployment, but it also shows a decrease in total employment as well. In April, the household survey found 139,455,000 employed persons. In May, it found 139,420,000, i.e., it found a decrease in total employment of 35,000 persons. How is it possible for the survey to show both a decrease in jobs and decrease in employment, you might ask? Because it also found a decrease of 322,000 in the overall civilian labor force.
Looks like there’s been some statistical cherry-picking here. Favorable numbers from each survey have been reported, and the unfavorable ignored.
The discrepancy between the two reports can be explained in part because the two use different methods, and count slightly different types of data. For instance, the household survey looks at employed persons, while the establishment survey looks at number of employment positions on the payrolls, so a person with two jobs shows up once on the first survey, but gets counted twice on the latter.
But the discrepancy between the two surveys is better explained by the fact that the alleged “increase” in jobs or “decrease” in unemployment are in fact meaningless, nothing more than statistical phantoms. As the BLS report states,
BLS analyses are generally conducted at the 90-
percent level of confidence.
For example, the confidence interval for the monthly
change in total nonfarm employment from the
establishment survey is on the order of plus or minus
100,000. Suppose the estimate of nonfarm employment
increases by 50,000 from one month to the next. The 90-
percent confidence interval on the monthly change would
range from -50,000 to +150,000 (50,000 +/- 100,000).
These figures do not mean that the sample results are off by
these magnitudes, but rather that there is about a 90-percent
chance that the “true” over-the-month change lies within
this interval. Since this range includes values of less than
zero, we could not say with confidence that nonfarm
employment had, in fact, increased that month.
At an unemployment rate of around 5.5
percent, the 90-percent confidence interval for the monthly
change in unemployment as measured by the household
survey is about +/- 280,000, and for the monthly change in
the unemployment rate it is about +/- 0.19 percentage point.
Full disclosure: I haven’t taken a course that required me to do math since freshman year of college, let alone taken statistics. So I can’t comment too deeply on what’s going on here.
But it looks to me like the May Job Report is largely a wash. Most of the increases or decreases are just a chance fluctuation, not a reflection on any actual underlying change in the data.
The BLS report also states that “An over-the-month employment change of about 100,000 is statistically significant in the establishment survey, while the threshold for a statistically significant change in the household survey is about 400,000.” [Note: I’m assuming by statistically significant they mean the odds of it being meaningless are the usual 1-in-20 figure? Which is why these numbers don’t quite match up with the previous quote, which used a 1-in-10 figure.]
So remember that net change in the employment rate of 287,000? Yep, turns out that doesn’t mean anything. There has been no actual decline in unemployment, as far as is shown by the BLS’s May Jobs Report — and ignoring the minor blip upwards in unemployment rates that happened in April, we’ve had essentially stagnant unemployment rates for all of 2010 so far.
As for the establishment survey, true, the increase in jobs that it found is significant. But remember — 411,000 of the new jobs were from census positions, which will evaporate once the census is over. And that 41,000 uptick in private sector jobs? Illusory. This survey does not demonstrate any net increase in private sector jobs; the most positive thing that can be said about it is at least it doesn’t necessarily mean there’s been any decrease, either.
Moral of the Story: If it were not coincidentally that time of the decade for the census, we would have zero job growth right now. Take out the census, and we only “gained” 20,000 jobs in May — a statistically meaningless number, under either of the BLS’s surveys.