Arnold Kling  

Labor Market Surveys Diverge

PRINT
A Bad Month for Libertarians?... Growth Inhibitors in Europe...

The Bureau of Labor Statistics uses two different surveys to measure job growth. The latest report shows that for the last two months, the household survey shows an increase of just over one million jobs. The payroll survey shows an increase of just under 200,000 jobs.

If you believe the household survey, then the party is getting going and the Fed needs to think about taking away the punch bowl. If you believe the payroll survey, then the recovery is still anemic.

As of a few weeks ago, economists at the Fed were not worried about the discrepancy between the two surveys, because they both showed a labor market with essentially zero job growth--one survey showing slight gains and the other showing slight declines. My guess is that the current discrepancy is too wide to brush off, and now the economists are under pressure to get the real story.

UPDATE: for more background, see this report from Congress' Joint Economic Committee, done in October.

For Discussion. What other data can be used to assess the state of the economy?


Comments and Sharing


CATEGORIES: Macroeconomics



COMMENTS (21 to date)
Eric Krieg writes:

Economists MUST have an opinion as to which survey is most accurate!

I fail to see how a survey of only large corporations could be seen as accurate. Small business is the engine of job creation! The household survey has got to be more accurate.

Steve writes:

I fail to see how surveying 65,000 households is more accurate than asking a bunch of companies how many employees they have! Either one is as good as liking your finger and sticking it in the air to determine temperature and wind direction.

Maybe we can have a workfare system where all the unemployed white collar workers go door to door selling magazines, and in the meantime taking household surveys? We'd reach a lot more than 65,000 households, and it would give them hope that the country gives a darn about them.

The IRS might be able to give us a better indication of how many people are employed, don't you think? They get money from us every two weeks (unless you're rich, then it's every quarter, but they are such a small percentage, who cares about them?) Maybe you could survey gas stations to ask how many full fill-ups vs how many under $10.00 fillups occurred? When I'm feeling flush with cash, I always fill 'er up.

Steve writes:

Another thought for you pro-market guys who wanted terrorism futures:

Just watch the stocks of the beer manufacturers.

When the Miller subsidiary of SAB is doing well, clearly something is wrong because people are buying the cheap nasty beer. When BUD is doing well, things are improving and people are buying good beer. Since nobody outside of America buys their beers, this would be a perfect indicator.

Or perhaps you prefer WalMart sales vs. Target? The first thing people do when they graduate from college is stop going to WalMart and start going to Target. When people claw their way into the middle class, WalMart is the first to leave their weekly agenda. This is why WalMart donates to the GOP, but that's another story all together.

Eric Krieg writes:

>>When the Miller subsidiary of SAB is doing well, clearly something is wrong because people are buying the cheap nasty beer. When BUD is doing well, things are improving and people are buying good beer

Jesus, add to all your other hangups, you like BUD over Lite?!?

A survey can be as accurate as anything social scientists do.

Steve writes:

Bud is more expensive and, hence, not an "inferior good" as defined by economists.

...and it is brewed by some of the finest people, in the best city on the planet!...

triticale writes:

In Milwaukee, increased sales of Budwasser would be a sign of an improved economy as a second order effect; more tourism. Locals drink Miller, or, if better off, micro-brews.

BTW, after 35 years of employment, I have yet to work for a company large enough to make the payroll survey, other than as an outside contractor.

Steve writes:

tritcale--

Good point! :)

How large is large enough for the survey? 15,000? 25,000?

Boonton writes:

"How large is large enough for the survey? 15,000? 25,000?"

I seem to recall from my old statistics classes that a sample of just 1,000 is enough to get a pretty accurate picture of a population of any site (even infinite). The fine print, though, is that the thousand selected have to be truely random!

The corporate figures are probably more accurate. How many employees a corporation has is roughly an unemotional number. "Is your husband working or not?" is an emotional question for the wife who wonders if she should count the $20 her husband made for fixing the neighbors car as work. Eric is right that small business may be the 'engine of job creation' but it is also the 'engine of job destruction' as well as the engine of semi-jobs, temp jobs, marginal jobs & unemployed people with pretensions about their over-glorified hobbies.

So the corporate survey probably misses spurts of job creation from the small business sector and the household survey probably understates unemployment from the prideful.

Perhaps what is needed is a method that combines both techniques as well as incorporating some other data such as from payroll processing companies (who will often do the payrolls for the 2-100 employee businesses), IRS returns and more.

John Thacker writes:

Boonton is right about sampling. The margin of error depends only on the sample size, not on the size of the population being measured. (So long as the population is significantly larger; obviously if population == sample size, then it's different.)

But, Boonton, it's not necessarily so obvious to a corporation who is on payroll. Is an independent contractor, freelancer, or temp worker considered on payroll? Perhaps, perhaps not, at various companies.

Eric Krieg writes:

>>Eric is right that small business may be the 'engine of job creation' but it is also the 'engine of job destruction' as well as the engine of semi-jobs, temp jobs, marginal jobs & unemployed people with pretensions about their over-glorified hobbies.

There is an adage in the auto industry that quality is whatever the customer says it is.

Economists should have a similar adage: a job is whatever the jobholder says it is. If someone says that they are employed, they are employed.

Boonton writes:

Errr no we should not follow that adage.

Steve writes:

Boonton--

I believe it also requires that you know something about your population. Such as, you need to know whether it is normally distributed or not. If what you are studying is normally distributed, 1,000 random sample should be pleanty as you said.

What I mwant was, how big does the employer have to be in order to make the survey today?

Lawrance George Lux writes:

Employment statistics will never be a good vehicle, as collation of data is difficult, and it tells little about Payroll dispersion. The IRS should be made to publish monthly quotas on Payroll deductions. You can get such information if you try hard enough, but it is always dated and tells little. Used as an Economic evaluation tool, though, it could be the most effective Indicator of them all. lgl

Monte writes:

It would be interesting to see how the national crime rate relates to the prosperity index. Controlling for prohibition, I'm reasonably certain the overall crime rate was low during the roaring 20s and high during the Great Depression? The crime rate from the early 80s through the prosperous Clinton years, a period of unparalled economic growth and prosperity, also showed a precipitous decline.

Any takers?

Boonton writes:

I believe crime increased during the 80's and fell during the 90's even though both periods were releatively prosperous. I don't have stats but I have heard that crime did not increase in the 30's (often voiced as an argument against the left-wing assertion that poverty causes crime).

In fact, I remember reading one theory that crime actually increases with the economy. The thinking is that periods of prosperity creates resentment towards the successful as well as the need that one needs material goods to 'fit in'.

Monte writes:

Oh well, I knew it was a stretch. Intuitively, you’d think things like the crime rate and divorce are inversely related to prosperity, while marriage and the birth rate would go up. I suppose no single demographic is reliable enough to work as a leading indicator. These things are more likely to lag good economic news rather than signal it.

How about the food service industry as a major economic force and key indicator of shifts in consumer attitudes and spending? What better way to celebrate good times than getting out and ordering steak and/or lobster at your favorite restaurant?

Scott writes:

In the past hasn’t the sale of men’s underwear been a reliable indicator of the direction of the economy. Anyone know what the direction has been recently?

Monte writes:

"In the past hasn’t the sale of men’s underwear been a reliable indicator of the direction of the economy. Anyone know what the direction has been recently?"

That's ridiculous! Any economist worth his weight in guns or butter knows women account for most underwear sales, and they're likely to spend as much in good or bad times.

Boonton writes:

"How about the food service industry as a major economic force and key indicator of shifts in consumer attitudes and spending? What better way to celebrate good times than getting out and ordering steak and/or lobster at your favorite restaurant? "

But what better way to finish a day where you got laid off than by going to your favorite watering hole and getting totally, totally smashed!

How about strippers? If there was some way we could survey their weekly cash tips I think that would be a a good measure of economic health. People tipping in $5 clips would indicate good times while $1 tips would be bad times.....if the strippers report they need to wear change purses to accept $0.25 tips then we know we are in real trouble!

Monte writes:

"How about strippers? If there was some way we could survey their weekly cash tips I think that would be a a good measure of economic health."

I'm sure that wouldn't work. They're not likely to report all their tips, particularly when it comes to table dances. However, strippers are a luxury item. During recessions, consumers generally settle for cheaper substitutes (magazines, videos, etc.), so we can reasonably conclude that when sales of these items are up, economic health is poor.

Dave Sheridan writes:

A good back-of-an-envelope indicator would be something like same-store sales at Starbuck's. Designer coffees are a small luxury whose sales, if you could abstract away from factors like changing tastes and competitors, should fluctuate pretty reliably with changes in disposable income.

As to the divergence between employer and household surveys, clearly something is going on, and I think the same something or things are behind the incredible reported productivity numbers. Something other than labor productivity is behind the GDP/hours worked phenomenon Outsourcing and freelance work strong candidates. Alan Meltzer suggested as much here "http://www.mosler.org/wwwboard/messages/518.html" in October.

boonton -- the reason the household sample is so large is that the survey is also trying to reliably survey specific geographies, age groups, racial groups and occupations. Sampling enough different geographic areas with enough observations to be reliable in several dimensions at the local level means a lot of observations at the national level.

Comments for this entry have been closed
Return to top