Bad News on Jobs

The economy added a shockingly low total of only 32,000 jobs in July, far below expectations. The May and June job totals were also revised downward.
The unemployment rate, which is computed from a different survey, fell to 5.5%, which raises again the question of how accurate the payroll survey is. That’s an academic question, however; from a political standpoint, this news is grim for President Bush.
BIG TRUNK adds: Reader Norman Nescio writes somewhat angrily:

Get it right, will you? Those are 32,000 PAYROLL jobs (establishment survey), which is NOT the full story. The full story is here (PDF file). Table A-1 (household data) shows TOTAL June employment of 139,031,000 and TOTAL July employment of 139,660,000, an increase of 629,000.

HINDROCKET responds: Chill, Norman. As we have repeatedly pointed out, the Department of Labor has two different surveys that track employment: the household survey and the “establishment” or “payroll” survey, which gathers information from employers. As we have noted more than once, there has been considerable divergence between the two surveys in recent years, leading to speculation–by us, along with many others–that the payroll survey may be providing a misleading employment picture. (Among other things, it fails to take any note of self-employment.) Note that in my post this morning, I pointed out that the fall in the unemployment rate, coinciding with the very anemic July jobs number, “raises again the question of how accurate the payroll survey is.”
But the fact is that the payroll survey is regarded as very important by economists, investors and politicians. Economists and sophisticated investors recognize that payroll jobs are not the whole employment picture; nevertheless, they are an important part of the picture. Note that the Dow is down over 100 points today in response, mainly, to the weak jobs data. The Bush administration has also taken credit for the very strong jobs growth, as reflected in the payroll survey, that occurred earlier this year. So it’s a little late in the day to try to educate the voting public about the relative merits of the two employment surveys. Hence my conclusion: the weak July jobs number, which is universally being described as such in the press, is very bad news for President Bush. See, e.g., CBS Marketwatch: “The major indexes all hit new lows for the year Friday after the Labor Department shocked Wall Street with a July jobs report that showed growth below expectations.”
BIG TRUNK notes: I think the market went down because of the anticipated political impact of the jobs report rather than its intrinsic economic importance. Reader Gerald Hanner provides a link to Econopundit with some suggestive supporting information.
University of Michigan Business School Professor of Business Economics and Public Policy Scott Masten writes:

May I respectfully suggest that, rather than supporting the relevance of the payroll figures, the fact that the Dow went down on the payroll news is further indication of the weakness of the stock market as an indicator of economic performance in an economy where much of the activity is going on in small, entrepreneurial businesses not traded on the exchanges (which the Labor Dept. data show)? Sorry, Hindrocket; I’m with Norman on this one.

Foremost among the impressive credentials that Professor Masten wields, at least in our eyes, is his status as a Dartmouth alum.
On the other hand, Dartmouth economics Professor Andrew Samwick, just off his one-year stint with the President’s Council of Economic Advisers, writes as follows:

I applaud you for sticking to your guns on using the payroll survey. The household survey tells us something, but nothing particularly reliable month-to-month about the number of jobs. In that survey, the unemployment rate is much more reliably measured month-to-month, since the errors or unusual factors that tend to boost the household jobs number (in the numerator) also tend to boost the labor force number (the denominator).
Here are two other issues about the monthly employment numbers:
1) Watch the workweek.
The average workweek is about 33.7 hours. So a change of 0.1 hour is equivalent, in terms of labor input, to about 0.3% of a worker.
The average workweek applies to about 80 million production workers. So a change of 0.1 hour in the average workweek is equivalent, in terms of to about 240,000 jobs in the payroll survey.
So if the jobs number disappoints, but the workweek ticks up by a tenth (as it did in July), then that’s not quite as bad. To be fair, though, the workweek ticked down two tenths in June, and so factoring in changes in the workweek doesn’t really improve our assessment of the last two months together.
2) The summer is messy.
There are a lot of things that happen in the summer months that are seasonal. Auto companies reduce their production schedules. Public sector employment–particularly in education–also changes quite a bit. The employment data are seasonally adjusted, meaning that the *average* changes in June, July, and August are accounted for. But if the current summer is a bit out of the ordinary, this can make it difficult to interpret the changes month-to-month. (Seasonal adjustment accounts for the mean effect, not the variance of the effect.) I haven’t been following the key sectors closely enough to have a view on how important this may be right now. But during the summer especially, it is good to be patient and wait for a longer time series of data to come in.

I note only that, one way or another, Dartmouth rules.
MORE: Reader James Lacey writes:

The payroll survey is a prediction based on past data, and current data, where as the household survey afaik is more of a traditional survey.
I’ve been reading up on how the BLS comes up with their estimate for job growth in the CSE (commonly known as the payroll survey), and one thing that seems to be key in their model is the CSE Net Birth/Death Model. It’s also a time-series model, that they use to estimate the net birth/death of jobs that were due to companies being created as well as going out of business.
The historical data for this model is here (they have only been using this model since Apr ’03 apparently).
According to their FAQ (link to it is on the above page) these estimates are not seasonally adjusted, but do show seasonal trends such as lots of new jobs being created in retail, right before xmas.
It also shows lots of jobs being destroyed in Jan, when retail jobs dry-up after the xmas shopping season. If you look at the historical data for 2003 and 2004, you will also see a large negative number, indicating the destruction of jobs, in July. Why? All the kiddies and college students are going back to school.
The point is that the BLS always predicts lack-luster job growth in July. So no one should be surprised by today’s report. In fact, based on the 2003 data I predict good job growth in August, and lackluster job growth in Sept. and Oct.
Now, as you guys know, the real question is which is the more accurate predictor of real job growth — the payroll survey or the household survey?

I’m not sure “predictor” is the right word, but the information is appreciated.
STILL MORE: Reader Stan Brown advises:

Please read Crudele (“Why Wall Street is guessing wrong on jobs growth”). As linked by Mickey Kaus, Crudele wrote before the number came out that it would be low. The Labor Dept makes an ASSUMPTION that lots of small businesses go bust in July. So they take a huge chunk out of the number every year in July.
Bottom line – the number doesn’t reflect what really happened this year in July 2004. It only reflects a Labor Dept assumption based on past years. This number tells us NOTHING about the current economy!

HINDROCKET concludes: Well, I learned a few things today. But the bottom line, I think, remains what I said in my original post. Notwithstanding all of the defects and uncertainties in the payroll jobs number, the fact is that most economists were predicting more than 200,000 such jobs to appear in July. They didn’t. There is no way that can be anything but bad news. My real point, however, was not that the economy is in the tank, but that the political implications are bad. That, too, seems indisputable. The Bush administration has never tried to educate the public about the various employment numbers (I don’t mean to suggest that they should have), and the administration has, understandably, trumpeted the payroll job numbers when they were good. This is nothing new. The Clinton administration did the same thing, and the Trunk and I got our start in punditry a decade ago, in part, by publicizing the job growth that took place during the Reagan administration, based on the same payroll numbers. Under the circumstances, the public perception that July’s job growth was anemic is both reasonable and inevitable.