Why I’m Concerned About the Independence of U.S. Statistical Agencies

Source: iStock/wildpixel

Are the White House and the Bureau of Labor Statistics headed for a train wreck? The most recent BLS statistics for December showed 7.5 million unemployed persons and an unemployment rate of 4.7 percent. But throughout his campaign, President Trump repeatedly described the official BLS unemployment statistics as “phony numbers,” saying that the “real” unemployment rate could be as high as 42 percent. His story hasn’t changed since the election; at his most recent (January 11) press conference he said that there “are 96 million really wanting a job … they can’t get.” Since then, Trump’s nominee for Treasury Secretary, Steven Mnuchin, said that the official “unemployment rate is not real.” And yesterday’s White House press briefing, press secretary Sean Spicer avoided answering the direct question, “What is national unemployment rate?”

To be sure, the BLS acknowledges that there are alternative ways to measure labor underutilization, and the monthly employment report includes five alternative measures in Table A-15. The broadest alternative measure from BLS is U-6, which includes “marginally attached” people (those who want and are available for a job and have looked for work sometime in the last 12 months) and people who are employed part time “for economic reasons” (those working part time though they want and are available for full-time work). In December, the U-6 rate was 9.2 percent, or nearly double the official rate. When fact checkers have compared the BLS statistics with Trump’s purported “real” unemployment rate, however, they note that to get a rate as high as 42 percent, one would need to include millions of people who say they do not want a job, such as retirees, students, stay-at-home parents, and the disabled.

How will this difference get resolved? Traditionally, the career professional staff at U.S. statistical agencies have been given wide latitude to make technical decisions about statistics. This latitude is described by the National Research Council as “independence from political and other undue external influence.” They note that “the credibility that comes from independence is essential for users to maintain confidence in the accuracy and objectivity of a statistical agency’s data and for data providers to be willing to cooperate with agency requests.” Similar guidance is contained in the “Fundamental Principles of Official Statistics” promulgated by the United Nations.

While the existing norms allow the U.S. statistical agencies to make statistical decisions independently of political influence, the U.S. has surprisingly little actual legal protection of that independence. The main rules that protect independence are several regulations from OMB, including guidelines on information quality and four statistical policy directives. (Although they don’t appear to currently be available on the new White House website, these guidelines and directives remain in effect unless the new administration explicitly decides to change them.)

For example, Statistical Policy Directive Number 3, which covers the principal economic indicators, requires that the statistics be released as promptly as possible and according to a pre-announced schedule. Changes in methodology must be announced in advance. Executive Branch employees (other than those from the agency that issued the report) are required to wait at least an hour after the release before commenting on the data. The directive prohibits prerelease dissemination except to the President (through the Council of Economic Advisers), who generally receives the indicators the afternoon before they are released. That means, for example, that the Secretary of Commerce does not receive BEA or Census economic indicators prior to their public release.

These rules protect the statistics from direct manipulation, but don’t address all my concerns about independence. For example, a Cabinet secretary could still order the statistical agency to drop certain statistics or to change methodologies in ways that seem politically expedient.

Some statistical agencies are directed by political appointees, while others are directed by career senior executives who report to political appointees, and having a qualified, professional political appointee in place can support an agency’s independence. While a new administration may criticize career civil servants, it’s harder to criticize someone the president has appointed. The agency directors typically have support within the administration, so they can serve as intermediaries between the career staff and the political part of the administration. Of course, it will be several months until many of these appointments are made and confirmed, but we should check to see that qualified individuals are named to these posts.

How will this play out for BLS and the unemployment statistics? It seems like there are number of different ways the conflict could play out, and I have no idea which one will happen.

One possibility is that someone convinces Trump that he should quietly drop the issue, the way he dropped his threat to pursue criminal charges against Hillary Clinton. At some point, Trump is going to own the BLS statistics, and there seems little long-term political advantage for him to continue to claim that they are phony. A similar scenario would have him continue to criticize the numbers until he has his own BLS Commissioner in place, and then drop the criticism.

Another possibility would be for the White House to ask the Secretary of Labor to order BLS to add another broader unemployment statistic to its existing menu of measures of labor underutilization. I’d hope that someone would convince Trump that his ratio of persons without jobs to the total population is too expansive, and with baby boomers retiring, it is not likely to show the improvements that he expects. But something like the prime-age employment population ratio (recently promoted by Jordan Weissman of Slate) could address concerns conveyed not only by Trump, but also by progressive economists, that the traditional unemployment statistics miss those who might want to work but have long since given up hope.

A worst-case scenario would have the administration continuing to attack the credibility of BLS and ultimately driving away many of their best professional staff. It shouldn’t be allowed to happen. My own experience has been that political appointees who’ve questioned technical decisions have been willing to engage in constructive dialogue with the professional staff toward the goal of producing objective, nonpartisan statistics.

The closest I’ve personally come to enduring the sort of criticism that BLS is currently receiving came at the time of the 1990s Boskin Commission, when I was at BLS overseeing research on biases in the consumer price index. While BLS was pressed by both Congress and the administration to eliminate biases, the Clinton administration’s economic policy makers did allow BLS to make the ultimate technical decisions on what changes to make and whether to incorporate them in the featured CPI or in an alternative index, like the chained CPI. The BLS Commissioner at the time, Katharine Abraham, was able to oversee the resolution of the technical issues and deal with the political pressures in a way that preserved the independence and credibility of BLS statistics.

While I’m concerned about threats to statistical agency independence, perhaps the biggest threat I currently see to statistics is coming from likely budgets cuts, as well as indirect cuts to resources such as the recently announced hiring freeze.

Update (1/26): A recent post by Kevin Drum points out that rather than the prime-age employment-population ratio, the unemployment measure  should be its inverse. That measure is about 21.8% and remains above the 1990–2007 average level of 20.2%, suggesting there’s room for it to come down some more.


  1. James West

    When you have a Chief Executive who was forced to settle fraud charges civilly to avoid criminal indictment, the credibility associated with all offices that are derivative from the presidency is nil. At least, that is the perception among those of us who don’t reside in America.

  2. François Lequiller

    Dear Brent,
    Very interesting comments on the US case. We had also in France a high-tempered president who tended to despise statisticians. He loathed revisions, considering them as a proof of the incapacity of statisticians. In such context, there is a need to have strong directors of statistics, such as K. Abraham. The case of the US is interesting as you have, if I understand well, under the spoil system, a political nominee as head of statistics. It would be useful for the US administration to reflect on the recent Code of Good Statistical Practice of the OECD which contains detailed recommendations about the nomination and dismissal of such a key person for the independence of statistical agencies.

  3. Katherine Wallman


    I am a bit confused by your statement, “The case of the US is interesting as you have, if I understand well, under the spoil system, a political nominee as head of statistics”. Are you referring to the US Chief Statistician? That position, to the best of my knowledge, has always been filled (since its creation in the 1930’s) by a career member of the Civil Service. Or are you referring to the head of a particular agency in our decentralized statistical system? A few (5) of those, including the Census Bureau and the Bureau of Labor Statistics, are “political appointees” though 3 of those 5 once appointed have fixed terms ranging from 4 to 6 years with the possibility of reappointment. The heads of our 8 other principal statistical agencies, including the Bureau of Economic Analysis, are career Civil Servants.

    Best, Katherine


Leave a Reply

Your email address will not be published. Required fields are marked *