The Jobs Numbers: Somebody Has Some Explaining to Do
The government’s Bureau of Labor Statistics (BLS) reported that payrolls grew by 121,000 in June. With minor revisions to April and May, the total increase in people working was 124,000. The unemployment rate stayed at 4.6%.
The news itself is okay but not great. You would think that at some point if total jobs don’t increase by more than they have during the past few months, the unemployment rate will start creeping back up.
The big question I have is how the ADP report on employment growth that was released Wednesday can differ from the BLS by so much. I noted yesterday that ADP’s total employment numbers were 273,000 higher than BLS’s at the end of May. Adding in the difference between ADP’s reported 368,000 increase in June vs. BLS’s 121,000, the difference is now over a half a million (520,000). BLS says that the average monthly jobs increase so far this year has been 131,000, while the average monthly increase according to ADP has been 218,000 — 66% more than BLS. That’s a ridiculous difference. If BLS is right, employment growth for the first half of 2006 has been okay, but nothing special. If ADP is right, employment growth is on fire.
I believe that ADP and Macroeconomic Advisers, LLC, the company doing the work on ADP’s behalf, carry a heavy burden of proof. I have converted ADP’s cursory explanation of its methodology from PDF to HTML and loaded it at my host, so that those who are interested can take a closer look if they wish. I’d like to see a detailed explanation of the differences at some point. If ADP wants to be taken seriously in the future, they owe us at least that.
__________________________________
UPDATE: Macroeconomic Advisers’ home page has this defense posted:
The ADP National Employment Report for June, reported yesterday, was correctly calculated. There is no truth to rumors that the ADP National Employment Report released yesterday, signaling a June increase of 368 thousand in private nonfarm payrolls, was erroneously calculated. Macroeconomic Advisers stands behind the calculations. Over the 69 months in which the index is available, from 2001 to May 2006, the standard error relative to the “as was reported” BLS preliminary estimate (for private nonfarm employment) is 91 thousand.
Folks, I don’t care whether it was “correctly calculated.” I want to know why you think your measurement of total employment more accurately reflects reality than BLS’s.
If they believe that the BLS numbers will “catch up” to ADP’s in the future, they need to say so, and tell us why.
Waiting ……
______________________________
UPDATE: I am grateful to Chuck Simmins at North Shore Journal for irritating the heck out of me (seriously).
That’s because Chuck points out that the employment numbers used to determine the unemployment rate in June went up by 387,000 (based on a survey of households, the “Household Data”), which is drop-dead great. But the total employment based on calling businesses (the “Establishment Data”) the one that leads the BLS report today, only went up 121,000, which as noted is nothing special.
It gets worse:
Why in the world is there a 9.1 million person difference between the total 144.363 million employment level reported in the Household survey and the 135.230 million in “non-farm employment” in the Establishment survey? (NOTE: See Update 3 for the BLS’s explanation.)
I also just noticed that the ADP-Macroeconomic Advisers total employment numbers are a full 30 million or so below the BLS totals. To say that there’s a mystery here is an understatement…..
Chuck’s central and well-taken point is that George Bush’s record on employment growth is very comparable to the Clinton record, and their comparative unemployment rates are virtually identical through their first 5-1/2 years in office.
___________________________
UPDATE 3: Chuck at North Shore sent me an e-mail with an explanation of the differences between the two BLS reports. I found a probably only slightly updated version at BLS. Here’s their explanation:
….. The household survey provides the information on the labor force, employment, and unemployment that appears in the A tables, marked HOUSEHOLD DATA. It is a sample survey of about 60,000 households conducted by the U.S. Census Bureau for the Bureau of Labor Statistics (BLS).
The establishment survey provides the information on the employment, hours, and earnings of workers on nonfarm payrolls that appears in the B tables, marked ESTABLISHMENT DATA. This information is collected from payroll records by BLS in cooperation with state agencies. The sample includes about 160,000 businesses and government agencies covering approximately 400,000 individual worksites. The active sample includes about one-third of all nonfarm payroll workers. The sample is drawn from a sampling frame of unemployment insurance tax accounts.
For both surveys, the data for a given month relate to a particular week or pay period. In the household survey, the reference week is generally the calendar week that contains the 12th day of the month. In the establishment survey, the reference period is the pay period including the 12th, which may or may not correspond directly to the calendar week.
Differences in employment estimates. The numerous conceptual and methodological differences between the household and establishment surveys result in important distinctions in the employment estimates derived from the surveys. Among these are:
- The household survey includes agricultural workers, the self-employed, unpaid family workers, and private household workers among the employed. These groups are excluded from the establishment survey.
- The household survey includes people on unpaid leave among the employed. The establishment survey does not.
- The household survey is limited to workers 16 years of age and older. The establishment survey is not limited by age.
- The household survey has no duplication of individuals, because individuals are counted only once, even if they hold more than one job. In the establishment survey, employees working at more than one job and thus appearing on more than one payroll would be counted separately for each appearance.
UPDATE 4: Larry Kudlow favors the Household survey, and has a lot of other positive things to say about the economy, because the economy is chugging along pretty nicely in the face of some tough obstacles. His claim that GDP is 20% higher since the Bush tax cuts ignores inflation; in real terms, GDP is up 12.5%.
UPDATE 5: Kudlow also has this interesting para at his related blog post:
Of course, big corporations facing high healthcare costs and globalization pressures to downsize are clearly not hiring as rapidly as small entrepreneurial firms. But the reality is that the jobs picture is much better than these corporate payroll numbers suggest. That is one reason why the economy remains stronger than conventional economists would have us believe.
IIRC from the 2004 debates, the reason is simple: large payroll companies are the traditional employment measure, and the source of the official numbers. Fueled by the internet and an IT consulting boom, there are many more self-employed than in the past.
Comment by TallDave — July 7, 2006 @ 4:05 pm
#1, TallDave, thanks. I was somewhat aware of that going in. One would think that either BLS would tweak the Establishment survey to account for that, or abandon it. 9 million is a lot of people.
You also have to ask why the BLS and the press don’t pay any attention to the Household total employment numbers, and whether this ignoring of that number is consistent with its treatment during the 1990s and before.
Comment by TBlumer — July 7, 2006 @ 4:30 pm
There has always been a difference between the household employment number and the non-farms payroll number. The interesting thing is to see how it has changed over time. During the 80’s and early 90’s the difference was about 9 million. The difference shrank during the later part of the Clinton administration until it was about 4 million. After the 2001 recession and the slow recovery, the jobs number recovered much more slowly than the employment number. The difference has now grown back to about 9 million.
Job growth looks better during the Clinton administration if you look at the jobs number. If you look at the employment number, not so much. The reverse is true for the Bush years.
Various economists (including a couple at the Fed) have looked at the issue and no one has a good explanation. Can you say PhD thesis…
Note something else in the current survey. The number of individuals in the employment number (from the household survey) that have multiple jobs has decreased by about 200K. The jobs number from the establishment survey has no way of knowing if the job holder has multiple jobs or not. If you hold multiple jobs, you get counted once in the employment number, but multiple times in the jobs number.
Comment by Scott — July 7, 2006 @ 7:22 pm
As I read the BLS explanation, it seems clear that the household survey counts every person with a job just once. Since the household survey is used for the unemployment rate, it seems correct to use it to explore job growth as well.
Comment by Chuck Simmins — July 7, 2006 @ 8:23 pm
I held one job, briefly, 33 years, ago which I am certain would have shown up in the Establishment survey. It happens that the industry I worked in at the time no longer exists. If, as I understand, contract work thru staffing firms and employment at firms smaller than 50 employees does not get counted, then over my entire lifetime that was the only job supporting me that exists to the BLS.
Comment by triticale — July 7, 2006 @ 8:30 pm
@3, I think that’s why #4’s point is a good one (and I believe correct), and #5’s point that he has almost never been in the establishment survey is interesting.
Comment by TBlumer — July 7, 2006 @ 8:35 pm
The “other” Scott above me offered a very interesting idea, but I’ll still go with the obvious…
We have an estimated 11-15 million “illegal aliens” currently in America- and between 2/3 and 3/4 are supposedly ‘employed’… How many do you think are on “official payrolls” in a small BLS “sample- space”(especially since the “large” corps tend to use “contractors” to prevent this type of regulation… See Wal-Mart, etc.)?
I’m still not sure about the ‘data’— Do the Feds ever ‘adjust’ their stats considering the large numbers of “illegals” in Texas, Arizona, and California? vs. Do “illegals” respond to ‘household surveys’ in a similar fashion as the rest of the population?
This is much like the difference between “earnings” surveys, and “spending” surveys by household– people seem to spend a lot more than they reported that they earned…
Comment by scott — July 7, 2006 @ 9:03 pm
#7, I agree that at some point you would think that the millions of illegals here would start impacting the employment numbers.
There are even more issues than the very valid ones you noted about illegals. Just one: I wonder if some people tell callers that they’re self-employed when they’re really unemployed out of pride, or if they say they’re “self-employed” because they’ve been sucked into a get-rich scheme by their circumstances and will learn in a couple of months that they’ve been had? That would indicate that the “unemployment” rate is in a sense higher than indicated.
Here’s another: If a person is in the middle of a six-month severance package, does a responding company say he or she is “employed” because they are still on the payroll? I would argue that they ARE employed because they are getting a check, but maybe they aren’t according to survey respondents.
Comment by TBlumer — July 7, 2006 @ 10:54 pm
I’ve seen no hard data but I would imagine that a large # of illegals do indeed show up in the Establishment survey. What is known, although I forget the exact figure, is the dollar value of SS taxes collected on bogus numbers. It would not surprise me in the least to discover that the majority of illegals employed have a valid I-9 (INS employment form) but a bogus, or hijacked, SS #.
Fake ID’s needed to fill out an I-9 are easy to come by. Pick any restaurant in LA County at random and do a search on all the employees I-9s, I’d bet on avg. as many as 25% are bogus, if not more.
At any rate I would put more stock in the Household survey since it is a random, representative sample, although I’m sure it undersamples illegals (advertisers have long lamented that Hispanics are similarly undercounted in reported ratings).
The Establishment survey is, as far as my (very) limited knowledge of statistics less reliable in that regard.
Comment by burt — July 8, 2006 @ 12:06 am
The Establishment survey is just that, a survey. I have participated. I used to handle payroll at a contractor that handled refinery shutdowns. We had 535 people on the payroll on the 12th of September one year (they were mostly on board for about two weeks)and I had to supply that number versus another period – we had about 150 people at that time. God only knows how that got extrapolated. There is no box to check that says “This is an exceptional 12th of the month.”
Statistically, when you read about a survey with a margin of error of 4%, here is what it means: one time out of every 20 the error is more than 4%. Read it that way and consider how many surveys you look at. Is this one off by maybe 10% – yes it could be – or 20% or 30%. Unusual things happen.
Comment by T J Sawyer — July 8, 2006 @ 2:05 am
@TJ: The idea is that if the sample size is sufficiently large, that will balance out. For every guy like you that was exceptionally high, there was another exceptionally low. That’s why they asked you to report payroll for that day.
Likewise, while people do certainly under-report, this should have no effect on relative comparisons since people’s rate of under-reporting shouldn’t change too much over time.
I am reminded of John Cowperthwaite, the finance minister of Hong Kong, in the 1960s. Asked by Milton Friedman why the HK government collected so few macroeconomic statistics, Cowperthwaite replied, “If I let them compute those statistics, they’ll want to use them for planning!”
Comment by Colin Kingsbury — July 8, 2006 @ 8:23 am
Reporting on Jobs Report
An interesting topic being covered in this week’s newsletter – the difference between the Household and Establishment number for the jobs report. The Wall Street Journal runs a couple of articles on the report of 121,000 jobs added in June, below expe…
Trackback by Recruiting.com — July 10, 2006 @ 1:32 pm
If the household survey is touted as reliable attempt to PROVE IT by searching the IRS self-employment taxes paid and compare it to the supposed number of jobs created.
Comment by gordon — July 15, 2006 @ 12:52 pm
#13, that is a good idea within limits which go in both directions:
- your check against self-employment taxes should really be a check against SE returns, because a person can be legitimately self-employed fulltime (i.e., not dabbling or hobbying) and lose money in the first year or two. That would pick up “everyone,” but….
- I don’t know how you would distinguish someone who is really going at as a real business vs. dabbers and hobbyists. Part of it might relate to whether they get a lot of employment or other income from elsewhere, but that’s not a definitive screen.
It wouldn’t be easy, but it would seem possible.
But I think we also have to ask how ADP is coming up with significantly higher numbers in its EMPLOYER survey than BLS is. It could be that ADP is capturing a more accurate cross-section of the economy with better small-business representation, and even representation of self-employed people who use ADP as a convenience, than the BLS.
Comment by TBlumer — July 15, 2006 @ 1:03 pm
If they’re self-employed and not making any money then they’re statistically unemployed. Or do we call the Household Survey the Underground Survey? Cheaters? How many of those 60,000 proxys go out to the same sample of deadbeats who claim they are employed or fall into the above “employed” catagory? I have’nt seen L.Krudlow back up his claims with SE data.He could get it.I’ve emailed him a couple times, no answer.
Comment by gordon — July 15, 2006 @ 2:49 pm
#15, all of this descends into a tunnel of murkiness that is difficult to resolve.
I don’t agree, but I can’t prove, that people starting a business that isn’t making money yet are considered “unemployed.” And I’m not talking about undergrounds, I’m talking about people who fill out sch. Cs who have more expenses than receipts (but who ARE getting at least SOME receipts). The fact that they are getting receipts would I believe qualify them as employed.
Kudlow is going with his gut; I’m sure he doesn’t know any more than BLS does. The household number is more stable, but the Employment numbers seem to vary a lot more than you would expect, esp in comparison to the household numbers. Comment 3 noted how this difference shrank significantly during the 1990s and widened again after that, and there’s really no supportable answer as to why.
And there’s still the ADP survey that is correlating more closely to the Household survey, which the developer, as expected of course, is stridently defending.
Comment by TBlumer — July 15, 2006 @ 3:01 pm
Economics and Socail Policy IV
The fourth edition of Economics and Social Policy is now up and links to this post.
Trackback by The Boring Made Dull — July 16, 2006 @ 5:05 pm