Responding to Rasmus’s response

Jack Rasmus is out with a response to my critique of his analysis of the April U.S. employment numbers. Enlightening Rasmus looks to be a hopeless case, but since there are may be some onlookers who wonder what’s up, here are a few comments. As with yesterday’s post, his comments are quoted and italicized (though the formatting doesn’t show up on an iPhone unless you choose the desktop version—sorry!).

What is significant is that Henwood thinks the CES (Current Employment Survey) is more important and accurate than the CPS (Current Population Survey).

I never said one was more important than the other; they tell you different things. What I did say was that on a monthly basis, the CES (payroll or employer survey) requires smaller moves than the CPS (the household survey) to be statistically significant, since it’s from a much larger sample. And it’s not me who came up with that—it’s the compilers (see question 1 in the FAQs):

“The household survey and establishment survey both produce sample-based estimates

of employment, and both have strengths and limitations. The establishment survey

employment series has a smaller margin of error on the measurement of month-to-

month change than the household survey because of its much larger sample size. 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 500,000.”

But the CES is not really a survey…

It is a survey. See above.

But the CPS is not just a “household survey”; it is also a survey of employment conditions of millions of smaller businesses through the survey of worker households. In fact, it can be argued that, in surveying 110,000 individuals each month, and then rotating and adding more households throughout the year, (roughly doubling the number contacted) the CPS in fact reflects a much larger body of business hiring, layoffs, and thus total employment, than does the CES.

The CES is a survey of jobs; the CPS, of people. They tell you different things. The CES is best for measuring employment, hours, and earnings overall and by industry. The CPS is great for measuring employment and unemployment by age, race, sex, educational attainment, and other demographic characteristics.

Henwood further argues that the CES 142,000 is more accurate because it is checked against the unemployment insurance system. But unemployment insurance has nothing to do with the numbers of employed or unemployed.

The UI system has a lot to do with the numbers of employed. It has less to do with the unemployed; less than a third the officially unemployed are drawing benefits. But the UI system covers almost everyone in formal employment. The quarterly comparisons are part of the Quarterly Census of Employment and Wages (QCEW). Here’s what it’s about:

“The program originated in the 1930s, and was known as the ES-202 program until 2003 when the current QCEW name was adopted. The primary economic product is the tabulation of employment and wages of establishments which report to the Unemployment Insurance (UI) programs of the United States. Employment covered by these UI programs represents about 97% of all wage and salary civilian employment in the country.”

Henwood is confused about the CES and CPS in another important way. There are more than 9 million businesses in the US economy. The 60,000/110,000 CPS survey is a statistically significant survey of employment in those 9 million. The comparison therefore should be 142,000 businesses vs. 9 million businesses. Henwood thus erroneously compares a population (CES 142,000) to a sample (60,000), when the comparison should be a business population (CES 142,000) to a business population (CPS 9 million businesses).

Rasmus’s point here eludes me. Perhaps a few numbers would help him sort things out. There are about 10 million employing establishments in the U.S. A sample of 142,000 of them is 1.4% of the total. There are 128 million households in the U.S.; a sample of 60,000 of them is 0.05% of the total. The CES’s sample is 28 times the share of its total universe as the CPS’s.

But Henwood would have readers believe the CES, with 142,000 businesses, and the 263,000 jobs created last month in that group, is all that matters.

Again, I never said one was more important than the other. In fact, three of the four graphs I included in yesterday’s post were drawn from the CPS.

Like the business press and government politicians, to believe Henwood we should take the 263,000 as the final word of the state of the US job market and forget all the rest.

I never said that either. April’s number is going to be revised next month and again the month after, and yet again next January with the annual benchmarking exercise, in which the monthly survey results are revised in accordance with the near-complete coverage of the employment universe. Revisions are generally small, but the monthly release is not the “final word.”

The BLS publishes a household employment series that’s adjusted to match the payroll concept (mainly by subtracting agricultural and family workers and the self-employed). It’s much more volatile than the establishment count from month to month, but over the long term they tell the same story. Here’s a graph of the two; they’ve only been publishing the adjusted household series since 1994, which is why the graph begins in 1995. Note that while the two lines generally move in tandem, the adjusted household one is far bouncier—even on an annual basis, which smooths out some of the monthly volatility.

[image error]

Instead we get jobs created by large businesses (CES) and unemployment from the 9 million population of all businesses. This problem leads to often conflicting data reported by the two sources, CES and CPS.

Everything in these two sentences is wrong. It’s not a matter of opinion or interpretation—they’re just wrong.

Henwood further assumes the role of slavish apologist of government stats by defending the U-3 unemployment rate as the best and final word on the state of the US labor market. He does refer to the U-6 unemployment rate, but unquestionably accepts the government’s current (and chronic) low estimates for the U-6.

Where did I say the U-3 rate is “best and final word”? And how does Rasmus know the U-6 figures are “chronically” low? Does he have a problem with the definition or the method of collecting the raw numbers? Given how little he seems to know about the construction of the employment stats, I’m curious how he knows how they’re wrong.

The U-6 picks up ‘involuntary part time’ employment. (U-4 and U-5 reflect what’s called ‘marginally attached’ and ‘discouraged’. These latter numbers too are grossly underestimated in the official stats).

Again, how does he know? The CPS survey asks people who aren’t working if they’d like to work. Depending on their answers, they’re counted as “discouraged” or “marginally attached.” Here are the official definitions of those two categories (see note to table A-15): “Persons marginally attached to the labor force are those who currently are neither working nor looking for work but indicate that they want and are available for a job and have looked for work sometime in the past 12 months. Discouraged workers, a subset of the marginally attached, have given a job-market related reason for not currently looking for work.” That job-market reason is typically “there’s no work out there.” It doesn’t seem too strict to me to exclude people who haven’t looked for a job for over a year from these categories; others might disagree.

Henwood disputes my claim that the U-3 is essentially an estimate of ‘full time’ jobs and says “No, it refers to work of any kind, not just full-time”. But if that were true, why add on ‘part time’ as the U-6 category separately? If there were part time unemployed in the U-3 and part time in the U-6 there would be likely ‘double counting’ of part time unemployed.

Again, everything here is wrong. People who are working part-time who want full-time work are counted as employed because they are employed—just not as much as they’d like to be. Adding unwilling part-timers to the U-6 measure is a way of accounting for their partial unemployment.

How do you square the 108,000 full time jobs created in the CES with the 191,000 full time jobs lost in the CPS, Doug? What’s your explanation?

I explained that yesterday, and here I am explaining it again.

Given the CPS number showing full time job decline (191,000), and the otherwise CPS rise in part time jobs last month (155,000), in my prior article I suspected that there are more workers taking on second and third jobs. Henwood pooh poohs this and trusts the government numbers on ‘multiple job holders’ showing little change. Once again, trust the government numbers!

Where do you think the inaccuracy arises? The survey of 60,000 is poorly designed? Government economists can’t add? They’re just making shit up because they’re the gummint? The affinity of this line of thinking to right-wing know-nothingism is impressive.

Henwood provides charts that show that Temp jobs (almost always part time) have not been changing for at least the past two decades.  As he says, temp jobs have been steady as a percent of the total work force for the past two decades, peaking at 2% of total jobs. “It’s barely changed for five years.” Sure, Doug. No one’s been hiring attempts except through agencies.

Again, not what I said. What I said: “This graph only shows employment arranged through temp firms. For other forms of contingent labor, like contract and on-call work, see here, here, and here. Spoiler: it’s nowhere near as big a deal as the likes of Rasmus would have you believe.”

The Labor Dept. has been covering up the growth of temp jobs since the 1990s. It produced three one-off reports, then George Bush stopped it. Too volatile.

Is Rasmus referring to the contingent worker survey? The latest version was released in June 2018 after a long delay. I wrote it up here. Despite claiming it hasn’t been produced in decades, Rasmus then cites it, but not very accurately.

How the government purposely underestimates labor stats that are embarrassing to it was clearly revealed, yet again, last year in its report on ‘precarious jobs’ (meaning temp, part time, gig, otherwise contingent, etc.). I and others have dissected that official report which claimed the gig economy was insignificant. But it turned out what the report defined as ‘gig’ was only full time uber/lyft drivers. Drivers as second and third jobbers were left out.

Completely untrue. Quoting the report on “electronically mediated employment”: “Electronically mediated workers were more likely than workers overall to work part time…. By industry, workers whose main job was in transportation and utilities were the most likely to have done electronically mediated work, with 5 percent of workers in this industry having done such work. Those employed in professional and business services, information, and other services on their main job were also more likely to do electronically mediated work, at 3 percent, 2 percent, and 2 percent, respectively.”

And from the FAQ: “This work can be done either in person or entirely online. The work could be done as a main job, a second job, or additional work for pay. For example, some people use their own cars to transport others from place to place, having obtained customers through a mobile app (such as Uber or Lyft) that also facilitates payment. Others do household chores or yardwork after finding clients through a website (such as TaskRabbit or Handy). Those who do this work entirely online could take surveys or add descriptive keywords to photos or documents through a platform (such as Amazon Mechanical Turk or Clickworker).”

As far as my suggestion that the April jobs numbers may reflect hiring of census workers, it is true the government to date has not indicated how many hired.

It has indicated how many it has hired, on a page I linked to in yesterday’s post. In case Rasmus missed it, here it is again: 1,000 in May 2018, who were laid off in July, and another 1,000 in February 2019, who are still on the job.

Henwood does give a backhanded concession to me that maybe my point of the 646,000 ‘Not in the Labor Force’ reported number indicates something is going on with the government data underestimating the total actually unemployed by having left the labor force in recent years.

Here’s what I said, after analyzing the persistently low participation rate: “That translates into a shortfall of about 2.5 million jobs, which is equal to the past year’s job growth. So there are more people on the sidelines than a 3.6% unemployment rate suggests, which is almost certainly a lingering effect of the Great Recession, whose damage persists more than a decade after its onset.”

Henwood accepts as accurate the Labor Dept’s estimate of discouraged workers (U-5) as accurate. In November 2007 just prior to the great recession the discouraged workers category represented only .2 of 1% of the labor force. Given the 10 million increase in the labor force since then, it is today still .2 of 1%. Can it be true that the percentage of discouraged workers has not risen at all in the intervening years–given the impact of the great recession, lagging economic recovery for years, and the fact of 7 million have dropped out of the labor force? It makes no sense that there should not be a corresponding increase in the percent of discouraged workers given the changed conditions of the last decade. The government data must be underestimating the discouraged worker category of unemployed (defined as out of work but having given up looking for the past year).

What a mess this is. The share of the population categorized as “discouraged” was around 0.16% just before the Great Recession. It peaked at 0.55% in 2010 and was 0.18% in April 2019. (See below.) Recent levels are low compared to that miserable slump, but are still slightly higher than they were in 2007, before the recession began, or 2000–2001, the strongest job market in recent decades.

The BLS estimates these things by asking lots of people lots of questions. Rasmus, by contrast, just has some feelings.

Also, they have to have looked for work at some time in the past year, not “having given up looking for the past year.”

[image error]

You help them maintain the fiction that the economy is doing great, that jobs are plentiful and well-paid, and we’re all better off than we think.

What I actually said: “An awful lot of Americans have a hard time getting by. Many are working dull, even dangerous, jobs, for stagnant pay and disappearing benefits. There’s strong support for that story. But unless you want to discredit yourself, there’s no point in making a lot of less-than-half-baked claims.” Put this one back in the oven, Jack.