Bryan Caplan and David Henderson  

BLS Data

On Greece and Weimar: Self-Ser... I Win My Unemployment Bet With...
The Bureau of Labor Statistics had not finished releasing the surprisingly good September data that a number of commentators started raising doubts about the truthfulness of the data. There is no doubt that the data were a boost to President Obama, a boost coming at a difficult moment, when the President's lead is faltering. Yet, is this motive sufficient to convict?

 I generally like conspiracy theories, but I find this one highly implausible. Not because statistical offices could not be affected by political pressures. The Greek statistical office lied on Greece public deficit for years. And in Argentina, the official inflation number appears substantially below what most observers believe the true inflation to be. But I refuse to believe this could be happening in the United States of America. Not because politicians in this country are incapable of such acts (Richard Nixon comes to mind), but because I have sufficient trust in the American system to doubt that it would work.
    How many people would have had to be corrupted or bullied for such a number to be distorted? Too many. In particular, too many for the news not to leak out. In a country where even private conversations, whispered at a fund raiser, are broadcasted worldwide, can we really think such pressures would remain secret? How many millions of dollars would a newspaper pay to the first employee of the Bureau of Labor Statistics to blow the whistle?
   It is not a trust in the superior integrity of the individuals, but a trust in the system, that makes me dismiss the claim. The support it received on the blogosphere is a worrisome sign of the increasing level of mistrust Americans have not only for their elected representatives, but for the overall system.

Yet, the most surprising endorsement of this conspiracy theory came from Jack Welch, the former CEO of General Electric. Not only because Welch is a Democrat Republican (corrected thanks to Erik Brynjolfsson)  but because he is a former CEO, who had to release a lot of important statistics every quarter. Welch twitted that "Unbelievable job numbers .. these Chicago guys will do anything .. can't debate so change numbers."

One has to wonder whether this cynicism about data is reflecting his experience in GE. Academic research shows that a person's trust is directly related to his own trustworthiness or, in simpler terms, that one expects others to behave the way he himself behaves. After all, Welch's GE was famous to consistently beat analysts' expectations every quarter.  Does this explain Welch's jumping to conclusions?  

One does not need to resort to a conspiracy theory to explain the September's number. Unfortunately, the news is not so good to be unbelievable. Much of the increase in employment came from temporary jobs. The recovery is progressing, but at a very slow pace. This slowness is partially due to the excessive amount of leverage. But it is also due to the increased    mistrust in in the system. If we want to restart growing at healthy rates, it is this mistrust we have to cure.

Comments and Sharing

COMMENTS (22 to date)

Jack Welch is a Republican.

"Morgan: You're a Republican, right?

Welch: Yes, absolutely."

Greg G writes:

Yup. One of the many things government is really bad at is keeping secrets.

One of the many things GE under Welch was really good at was massaging earnings to bring in the numbers wherever they wanted them to be.

The risk reward tradeoffs for this kind of alleged conspiracy are very different in the public and private sphere as well. It would be devastating for Obama's re-election chances if there was credible evidence of this preposterous conspiracy.

It is hard to imagine that Jack Welch could have lost his job for massaging GE's accounting.

Eduardo writes:
"How many people would have had to be corrupted or bullied for such a number to be distorted? Too many. In particular, too many for the news not to leak out."
Exactly, that's the first thing I think about with any conspiracy theory - fake moon landing, 9/11 inside job, etc. People love to have secrets (they're a form of social capital), and love to let those secrets slip out to the select few. The recent Soderbergh movie "Contagion" had some good bits on that; people tell their top-secret secrets only to their nearest and dearest, but then those people do the same, lather rinse repeat...
MG writes:

Interesting post. I think the real conspiracy here must be trying to goad anyone to speculate about conspiracies. Like you all have said, not only are these theories preposterous, but they distract economists from disgnosing the real information of any release, which ranges from "nothing new to possibly troubling". In the nothing new sategory: payroll growth (~100K- 125K/month) continues to slush along at rates that can not possibly support stable unemployment and a growing labor market for any length of time. If speculation was needed, then how about asking the speculators to speculate on the following. After such a long period of subpar employmeny growth amidst an unprecendented amount of fiscal and monetary stimulus, shouldn't there be some payback? Is this finally that payback, and how sustainable is it? Also, given the Fed's announced commitment to run monetary policy for the benefit of employment, should they see a drop of one full percentage point in the unemployment rate in less than a semester as "mission accomplished"? Or do they know/believe reality to be different? And finally, if sluggish payroll growth can become consistent with a falling unemployment rate, what has changed and what does it say about the economy's ability to generate the kind of GDP growth that we need to generate to pay for our way of living.

Luigi Zingales Author Profile Page writes:

Sorry, my mistake.. You are right

I believe that the model of trust and trustworthiness expressed by Prof. Zingales may be improved with the addition of another dimension. The context of trust is important. A person may be very trusted in one circle, but totally mistrusted in another circle.

So we need to see groupings of people, and the ambitions or needs of these various groupings.

Mikedc writes:

A conspiracy to fudge a fact implies the facts matter.

Why bother with a conspiracy when the starting point prime unemployment indicator is so meaningless in the first place?

Media and politicians routinely report bad things (a falling denominator- people leaving the labor force and a rising numerator due to people vastly underemployed) as good things.

Woerner writes:

What is the 'margin-of-error' in these BLS unemployment statistics ??

All surveys and statistical measures have a margin-of-error.

It is totally impossible to measure the true employment/unemployment rate in the massive U.S. economy ... to the first decimal point. But the BLS claims to do so routinely (though with many oh so convenient fudge-factors).

BLS does not state any margin-of-error in its pronouncements-- though it's obviously there, and they constantly "revise" their official numbers. Why ?

The underlying BLS business/household survey methodology & data is loaded with known sources of error in professional Survey-Research...but no mention of these at all; BLS simply declares its measurement as truth (...accurate to a 10th of a Percent). And most economists somehow accept it without doubt.

Actual BLS margin-of-error must be at least half-a-percent-- and likely much larger. So why agonize over this recent "0.3%"
change(?) in U.S. employment ?

It's all politics... not facts nor economics.

GiT writes:

Woerner writes...

BLS does not state any margin-of-error in its pronouncements

From the BLS:

"For example, the confidence interval for the monthly change in total nonfarm employment from the establishment survey is on the order of plus or minus 100,000"

The September change of +114k jobs would then range from about plus .05 to .55 percentage points. But hey, if you believe the margin of error "must be" at least .5 percent, you must be right. After all, you believe it.

Or, simply check the sub-national data for definite confidence intervals. Regional intervals are about +/-.3:

The underlying BLS business/household survey methodology & data is loaded with known sources of error in professional Survey-Research...but no mention of these at all; BLS simply declares its measurement as truth

That must be why the BLS issues a technical note with a whole section entitled "Reliability of the estimates" which details sources of sampling and non-sampling error.

But, good to see that the paranoid style is alive and well.

Yancey Ward writes:

Is it really true that a lot of people would have to involved? For example, how is the call survey list generated? If I were interested in fudging the numbers, I might try to find a juncture in the process that determines who gets surveyed, and if this just a piece of code somewhere, double-plus good.

I don't believe the conspiracy personally, but I am not sure it actually would have to involve many people, either.

How many millions of dollars would a newspaper pay to the first employee of the Bureau of Labor Statistics to blow the whistle?

My guess would be, zero. Newspapers almost never pay for information.

Better question, do newspapers EVER sit on stories that would disfavor their preferred politicians?

Say, the relationship of Barack Obama and Bill Ayers? Or, the relationship of Barack Obama and Mike Kruglik (of Building One America non-fame)?

R Richard Schweitzer writes:

The point has been raised by others with better insights:

If these BLS "Trends," are fact-based, do they not conflict with the "fact-based" policy decisions of the Federal Reserve Board?

Sieben writes:

... Why is America uniquely insulated against conspiracies? Don't they have the internet and telephones in Greece?

Can our government lie about nothing? Like WMDs in Iraq?

Derek writes:

Professor, Jack Welch is a Republican.

Woerner writes:

... rather unhelpful comment, GiT :

Yesterday's official, complete BLS Commissioner statement on the September Unemployment Rate contains not the slightest hint of any 'margin-of-error' whatsoever -- in the BLS results.

The unemployment rate is flatly stated as "7.8 %"

Even the expanded BLS Technical Addendum fails to mention any specific margin-of-error at all, though it generally admits the existence of significant "sampling and non-sampling error".

WHY does BLS pronounce an exact # for the Unemployment Rate, without clarification -- when it fully knows there is a substantial margin-of-error in that measure ? No reputable survey-research organization would operate this way.

Further, it is technically & practically impossible to measure the national Unemployment Rate to an accuracy of a tenth-of-a-percent... as BLS purports to do routinely.


But GiT, please help us out with your insight into 'sub-national data for definite confidence intervals' -- what is the resulting margin-of-error for that BLS bottom line "7.8 %" ??

Is it "7.8%": +/- 0.001% , 0.1% , 0.2% , 0.5% , 1.0% , 5.0% , 10% , 100% (??)

Mark Crankshaw writes:

Here is a topic that really hits close to home: I am a Mathematical Statistician who works in the Office of Employment and Unemployment Statistics at the Bureau of Labor Statistics. I do not work on the Current Employment Survey whose numbers are being discussed here, but I know personally many who do.

I despise the Obama administration and would like to see it defeated in November. However, I can state unequivocally that CES and CPS unemployment/employment estimates are not subject to political manipulation. This is a conspiracy "theory" that simply has no basis in reality. If there were the slightest hint of evidence of political manipulation at the BLS that I even remotely came across that was damaging to the Democratic Party, I wouldn't need even $1 to whistle-blow...

GiT writes:

Actually, my comment was entirely helpful. You just appear to be rather innumerate.

The technical addendum does mention a margin of error:

"For example, the confidence interval for the monthly change in total nonfarm employment from the establishment survey is on the order of plus or minus 100,000."

As I said, this translates to a margin of error of less than +/-.3. One could easily infer this from the sub-national data, because the regional margin of error is +/-.3. A larger n means a smaller margin of error, ceteris paribus.

This should all be obvious to anyone who actually understands what they're saying when they utter the words "margin of error."

As to your claims about the impossibility of measuring employment to within tenths of a percent, the only thing they demonstrate is a lack of understanding of statistics.

john hare writes:

Possibly not related to this post, I would like to see employment listed as "Employment rises from 91.9% in August to 92.2% in September." This would say the same thing as unemployment drops from 8.1% to 7.8% in a positive manner rather than the doom and gloom of tracking unemployment. Of course the employed number would be somewhere in the eighties or even seventies.

I think this would send a positive message to people that "can't find a job because nobody is hiring." If eighty something percent of people are working, then obviously there are jobs out there.

Harun writes:

The obvious way to test this theory that it would take many, many people would be to find out just how Greece and Argentina cooked their books and why no whistle blowers appeared in Greece until it was too late. In Argentina, you do have people being threatened with jail who dispute the numbers. Did those people exist in Greece?

By the way, if we had some amount of press interest in the fast and the furious, or any of the other scandals that would normally have been front page for months, then less people would believe these conspiracies.

When the media became the palace guard, of course people start to suspect we might have other aspects of a banana republic in place.

Ari Tai writes:

I had a conversation about some of these issues with a friend in the Commerce department 5-6 years ago - a good friend's daughter was looking for an internship that involved statistics. We scheduled a conference call - she explained her interest and he suggested that she'd be more challenged and at the forefront of her discipline if she pursued an opportunity with one of the major pharma, bio, or agriculture firms - since (he believed) the Department didn't really do statistics - they (had to) follow the law, irrespective of the state-of-the-art and reliability and applicability of their data and analysis. And he didn't expect the law to change given it was the result of a deliberate political process that would require both parties to agree. Which meant there were no regular red-teams, no double-blind studies, no auditor beyond the collection process, no independent verification (or publication) of what is and is not being measured. And since the last thing a Congress wants to do is take responsibility – they tend to demand fidelity to a process – rather than an application of intellect and judgment and all the risks to the establishment this entails.

Needless to say this U of Colorado math graduate student found other employment - that was initially distasteful to her – bioengineering plants, certainly not “green” or “organic” – but she dug into Mr. Borlaug’s story and now knows she is part of efforts that will bring a full belly if not salvation to the least-of-us.

The best economic data is in the Treasury departments of the Fortune / Global 100. It’s a pity that it’s so closely held – but then again, it’s the pulse of their business – and comes at a significant cost.

egd writes:

As Yancey Ward asked: how do we know that the conspiracy would be too large?

I think everyone agrees that 800k+ jobs were not created last month. Conspiracy is probably the wrong answer, but how to explain numbers so far from the norm?

david nh writes:

Never worked in government, have you? I have.

It's perhaps best to remember that there is usually a good deal of subjectivity, judgment and indeed estimation that goes in to the production of so-called data. Each of these stages is a point in which bias, whether conscious or not, could be injected. A change in "methodology" would be the smoking gun. A one-off tweak to bias the numbers for political reasons might be hard to bury/spin even within the organization. On the other hand, it's best not to underestimate the role of informal culture within an organization - being a "team player", knowing what the boss wants to see without having to be told, etc.

Academic research shows that a person's trust is directly related to his own trustworthiness or, in simpler terms, that one expects others to behave the way he himself behaves.

On the other hand, common sense tells us that intelligent and effective people modify their expectations of others (which may well in the first instance be based on their own ethics) based on life experience and, for example, an understanding of the institutional framework (and the incentives or world view it might imply) in which the other person operates. Lots of fundamentally well-inclined people end up doing bad things in dysfunctional institutional settings, e.g., USSR. In fact, I would have thought that much of economics was based on the latter notion.

I think the experience of many is that the US government has done things, over the last 5-10 years, that they thought they would never live to see and that perhaps this has prompted a closer inspection of things done in the past, resulting in a re-appraisal of the state.

Comments for this entry have been closed
Return to top