I mentioned the nonfarm payrolls report in my last column on the market's slavish reliance on unscientific economic data. A special shout-out to Real Money readers Scott Mills, Paul Joseph Williams and Angel Rodriguez for reaching out based on that column and requesting my analysis of the composition on the nonfarm payrolls data. I always appreciate the feedback.
The first thing to understand about the jobs report is that it is really two reports, each based on a survey of a different segment of the population.
The household survey -- CPS in Bureau of Labor Statistics (BLS) terminology -- is a monthly survey of about 60,000 U.S. households. It is from this data that the oft-quoted unemployment rate is calculated.
The establishment survey (CES) is a monthly survey of about 146,000 U.S. businesses and government agencies. It is from this data that hours worked and wage data are derived and, of course, the headline figure for nonfarm payrolls created, a.k.a. the jobs number itself.
Both surveys suffer from low sample sizes. According to the U.S. Census Bureau, there were 115,610,216 households in the U.S. at last count. So the households that are being surveyed by the BLS represent a whopping .039% of the total in the U.S. It is hard to imagine that such a sample could accurately indicate the number of people in the civilian labor force.
The establishment survey represents a much higher proportion of its target population. The 146,000 establishments in the survey base would represent 1.9% of all nonfarm establishments that employ at least one worker, according to Census Bureau data. The survey base also includes government agencies, so, on the private side, the sample size is probably closer to 1%. That would be a more representative survey, except it excludes so many workers.
From the BLS website:
"The CES employment series are estimates of nonfarm wage and salary jobs, not an estimate of employed persons; an individual with two jobs is counted twice by the payroll survey. The CES employment series excludes employees in agriculture, private households and the self-employed."
So the establishment survey double-counts some workers and excludes so many others that don't work in traditional punch-the-clock industries. Ugh.
But this is what we live for. The first Friday of every month at 8:30 a.m. ET ... it's on! Clearly, the market is getting excited about numbers that aren't particularly relevant, but are the survey methods at least reliable?
In a word, no.
The household survey is somewhat bizarre, and perhaps inspired by Ed McMahon and his famous Publishers Clearing House home visits.
The BLS generates a list of 72,000 households from Census Bureau data covering 824 sample areas, and then its operatives hit the bricks. The first interview is preferably conducted in person. So during the week that includes the 19th of the month, the BLS employee will contact the target families and inquire about employment status during the week that includes the 12th.
Subsequent interviews are usually conducted via telephone, but really, who are these at-home folks who are more than willing initially to allow random government employees into their homes? These are the people on whom we want to base our nation's economic policies?
The methodology of the establishment survey is somewhat more robust, but the first five months of a business' inclusion relies on computer-assisted telephone interviews. I run a small business and I often get robo-calls. On the rarest of rare occasions when I miss one via screening, I hang up immediately. I guess some people don't.
Unsurprisingly given the methodology, some of the numbers in the two nonfarm payrolls surveys are inconsistent and confusing, and the two surveys themselves often disagree wildly. I'll break down those numbers in my next column.