Real people. Not government statistics.

Photographer: Patrick T. Fallon/Bloomberg

Government Statistics May Be Wrong, But They're Not Manipulated

Justin Fox is a Bloomberg View columnist. He was the editorial director of Harvard Business Review and wrote for Time, Fortune and American Banker. He is the author of “The Myth of the Rational Market.”
Read More.
a | A

I can think of lots of reasons to question or criticize the unemployment rate, gross domestic product and other economic numbers that the U.S. government puts out. Manipulation by politicians looking to win elections is not one of them.

Why don’t I worry about this? First, because I know a little bit about the people who put together our nation’s economic statistics. The Bureau of Labor Statistics, Bureau of Economic Analysis and Census Bureau are run on a day-to-day basis by career employees, not political appointees. Even the appointees are often career staffers who get promoted, and many have served under multiple administrations. When top statistics-agency officials do leave government, it’s often for jobs in academia. Credibility with peers is generally of far more value (economic and otherwise) to these people than anything a politician could do for them.

That’s not to say that some politician might not try to influence them -- which brings me to the second reason I’m pretty confident overt political manipulation isn’t happening. Word would get out. As Republican-leaning economist Douglas Holtz-Eakin told Marketplace last week:

If political superiors were to attempt to force the BLS to come up with the quote “right answer,” lower the unemployment rate, the career staff would be calling you and leaking this to every whistleblowing outfit under the sun.

As best I can tell, the worst case that has come to light of the White House meddling with a statistical agency involved Richard Nixon, who was convinced that a “Jewish cabal” at the BLS was trying to undermine him.  Nixon was frustrated with the BLS chiefly for characterizing several drops in the unemployment rate in 1971 as statistical quirks rather than real job market improvements. He succeeded in getting the BLS official responsible reassigned to a less prominent role, and after the 1972 election pushed out BLS Commissioner Geoffrey Moore, to some consternation in statistical circles. But Moore’s replacement, Julius Shiskin, was a lifelong government statistician who ended up being reappointed by Jimmy Carter, and there’s no indication that Nixon tried to alter the actual unemployment numbers.

More recently, New York Post columnist John Crudele claimed to have unearthed a couple of scandals in government data gathering, but they really didn’t amount to much. Crudele is a veteran business journalist with a great nose for stories but a less-than-rudimentary understanding of economic statistics.  In 2010 he asserted that the Census Bureau was hiring and firing and rehiring workers because each time it rehired a worker it could “report the creation of a new job to the Labor Department.” But the BLS doesn’t count “new” jobs, it just counts jobs. If there are more jobs one month than the last, news reports often describe the increase as “new jobs.” Job churn of the sort Crudele described would have no impact on this -- not to mention that Census Bureau hiring during the decennial census is generally discounted by economy watchers.

There’s more meat to Crudele’s second charge, first made in 2013, that a Census Bureau employee had been caught fabricating data that goes into the monthly employment report.  In fact a field representative in the Census Bureau’s Philadelphia region was found to have made up responses on eight different surveys in 2010 and 2011, including three employment-report surveys. Follow-up investigations by the Commerce Department’s inspector general and the Republican-led House Committee on Oversight and Government Reform and the Joint Economic Committee found no indication that the falsifications were politically motivated or had any impact on the reported unemployment rate.

The House report did conclude that the Census Bureau was putting lots of pressure on employees to improve response rates, and not doing enough to root out potential falsifications. Low response rates are a real problem for government statistics agencies -- as they are for private pollsters and market researchers. A 2014 paper by economists at Princeton University and the Congressional Budget Office found that rising nonresponse rates were making unemployment numbers in the U.S. less reliable.

So that’s a good reason to be dubious of government economic statistics! But those who rail against the statistics gatherers may be making it worse. According to that 2014 paper, there was a “dramatic acceleration” in the nonresponse rate for the Current Population Survey (from which the unemployment rate is calculated) after 2009, which “coincided with publicity surrounding the 2010 Census, and calls by some irresponsible officials that individuals should not participate in the Census.”

What happened in 2010 was that, first, some guy in Southern California made a video calling for a boycott of the census as an violation of the Fourth Amendment ban on unreasonable search, and it went viral. There was no noticeable decline in participation in the census that year, but then Republican officials began criticizing the bureau’s annual American Community Survey, with the Republican National Committee calling it a “dangerous invasion of privacy.”

That’s been followed by a stream of accusations from the political right -- most notably former General Electric Chief Executive Officer Jack Welch with his October 2012 tweet accusing the “Chicago guys” (aka the Obama administration) of making up September’s unemployment numbers -- that government economic statistics are tainted. And yes, there do appear to have been similar complaints from the left during the Bush years, although I don’t think they got nearly as much traction.

Again, none of this is to say that our official economic statistics are perfect. There are lots of other criticisms that do have at least some validity, and I hope to take up some of them in subsequent columns. But the claim that they are political fictions not only doesn’t hold water, but is probably making it harder for the statistical agencies to do their jobs.

  1. John Williams, who publishes the Shadow Government Statistics newsletter, writes that “if Lyndon Johnson didn't like the growth that was going to be reported in the GNP [gross national product], he sent it back to the Commerce Department, and he kept doing so until Commerce got it right.” That does sound like the kind of thing LBJ might have tried to do, and Williams says he got the story from a former Commerce official. But I have not seen this allegation from any other source, so I guess we’ll just have to wait for the fifth volume of Robert Caro’s Johnson biography.

  2. I’d put my understanding level right at rudimentary.

  3. Census Bureau field representatives do in-person and telephone interviews of 60,000 households every month for the Current Population Survey, and the Bureau of Labor Statistics crunches the numbers and reports the results. The BLS conducts a separate establishment survey of 146,000 businesses and government agencies to get the nonfarm payroll jobs numbers that are also included in the monthly employment report.

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

To contact the author of this story:
Justin Fox at

To contact the editor responsible for this story:
Stacey Shick at