Should We Worry About Trump Fudging Jobs Data?

The unemployment rate this month was 4 per… mmmm let’s make it 3 percent.

Jim Watson/AFP/Getty Images

President Donald Trump does not particularly trust official government statistics. Unfortunately, now that he’s president, the rest of us might need to worry about their reliability, too.

Not to sound like a conspiracist or anything! But the question of whether the White House might try to tamper with federal economic and demographic data in order to suit its political agenda, or try to silence statisticians altogether by defunding their agencies, has definitely been in the air. After all, our new president is a notorious fabulist with an authoritarian streak who refers to the unemployment rate as a “fiction” and whose administration blessed us with phrase “alternative facts” to describe some of its most blatant falsehoods. These are the kinds of things that make people who deal with numbers for a living sweat—just a little. So, recently, we’ve been treated to headlines like these:

From CNN: “Will Trump Team Try to Undermine Official Unemployment Numbers?”

From the Guardian: “Statisticians Fear Trump White House Will Manipulate Figures to Fit Narrative.”

From Science: “Scientists Fear Pending Attack on Federal Statistics Collection.”

How worried should we really be? After talking with a few former government officials, here’s how I’m thinking about the issue right now.

There are three broad ways that a Trump administration could really mess with our official data.

First, there’s the straight-up banana republic approach: Trump could send one of his minions down to the Bureau of Labor Statistics or the Census Bureau, and tell them to doctor some inconvenient numbers. Jack Welch once accused the Obama administration of doing this, which was crazed. Even given our country’s slightly changed circumstances, I do not think that it is at all likely to happen.

Second, there’s the craftier approach: A new agency director could change what questions get asked on key surveys or the way certain statistics are calculated. This is a reasonably serious and immediate threat.

Finally, there’s the brute-force approach: The administration and Congress could simply cut funding and force the Census or BLS to drop certain projects, or just leave their data-gathering capabilities badly degraded. This is also a reasonably serious and immediate threat.

Let’s take each of those in turn.

The banana republic scenario would be a disaster, obviously. I mean, it’s possible to get along in a country where government stats are transparently fabricated (see: China), but I wouldn’t recommend it (see: Argentina). Democracies don’t tend to thrive once the border between statistical truth and fiction has been totally obliterated. U.S. investors and the Federal Reserve also depend on certain key economic indicators, like the unemployment and inflation rates. If word got out that those were being fudged, markets would flip and monetary policymakers would be left flying blind. A panic could ensue if the public discovered that any stats were being tinkered with for that matter—say, the poverty rate or the census’ data on health insurance—since the revelation would inevitably raise questions about the veracity of all of the government’s official numbers. Even if nobody uncovered direct evidence of foul play, any sudden, unexplained changes in monthly or annual data series could leave doubts.

But, again, the idea that Steve Bannon is going to march into the Bureau of Labor Statistics and start dictating unemployment figures is pretty far-fetched. First, Trump’s Cabinet is full of business types who, whatever their shortcomings may be, know the value of reliable econ data. Labor secretary nominee Andy Puzder was a fast-food CEO, and while he may have qualms with the unemployment rate as an economic indicator, he mostly just prefers other figures produced by the BLS (which is fair). Moreover, government statisticians are an intensely independent lot who abide by carefully standardized procedures, and if any political appointees tried to step on their toes, it would almost certainly leak. The number-crunchers at the BLS have an especially healthy sense of paranoia about outsiders interfering or trying to obtain numbers before they’re released. “The few days before the unemployment rate is released, the people who are working on it tape newspapers over the windows so nobody can find out what they’re doing,” Jesse Rothstein, a former chief economist of the Department of Labor and professor at the University of California–Berkeley, told me. That’s caution.

Katharine Abraham, a former BLS commissioner now at the University of Maryland, was similarly reassuring. “I’m not super concerned about somebody coming and trying to force the BLS to produce data that are slanted in some way. I just don’t see an avenue where that could happen,” she said. “There are too many people involved in that process for it to be in any way realistic.”

So ham-fisted manipulation is probably out as a possibility. But there are more sophisticated ways Trump could meddle with government data collection that would still be incredibly harmful.

For instance, the administration has drafted an executive order that would have the Census Bureau ask Americans about their immigration status for the first time (it currently only asks whether they’re citizens). The order is sort of bizarrely written—it proposes attaching the question to the long-form questionnaire on the decennial census, a supplement that no longer exists (it was replaced by the annual American Community Survey). But the frightening intention is clear enough. From the Washington Post:

“It will drive the response rate down enormously,” said Kenneth Prewitt, a former director of the Census Bureau who is now a professor of public affairs at Columbia University. Immigrants here illegally are unlikely to answer questions about their status, he said, adding that the resulting undercount could have chilling effects.

“If you drive those people out of the Census, the consequence is that they’re not in it,” he said. “It’s a step toward not counting the people you don’t want to count. And that goes very far in redrawing legislative boundaries.”

Technically, the Census Bureau is barred by law from sharing Americans’ personal information. But if you were an undocumented immigrant, or a green card holder worried about being targeted by the new administration, would you take the risk?

Intentionally driving down response rates among immigrants in order to change the counted ethnic and geographical distribution of America would be a fairly dastardly stroke. Less dramatic but still worrying: Agencies could potentially cut off researchers from some specialized data sets. (Will the Trump IRS keep giving Thomas Piketty & co. access to new tax records?) And as the Guardian’s Mona Chalabi noted, political appointees could force agencies to change their press releases on new data publications to bury unflattering statistics or change their methodology for computing some figures (though, again, that might attract some unwanted attention).

Finally, there’s funding. Without money, the statistical agencies can’t go out and properly conduct surveys. And without proper surveys, the quality of our data will suffer. The BLS is already struggling under budget cuts and has had to abandon some ancillary data projects. Further funding reductions could eventually start to hamper its core work. The Census Bureau, meanwhile, needs a substantial budget increase so it can begin its final preparations and dress rehearsals for its 2020 count. Congress hasn’t delivered yet, and if it doesn’t, the agency could be forced to choose between laying crucial groundwork for the decennial census or properly fulfilling some of its other responsibilities, like the Current Population Survey (which gives us the unemployment rate).

“These last Census tests are really, really of utmost importance in ensuring that everything has been planned. That the forms you want to use, the procedures you want to use, are effective and tested,” former Census Bureau Director Steve Murdock, now a professor at Rice University, told me. “I would say you really can’t overestimate the importance of the pre-census activities.”

A less immediate but perhaps scarier possibility is that the Trump administration would attempt to defund the American Community Survey, the annual project that offers a detailed socio-economic picture of the country. The survey is considered essential by social scientists, businesses (retailers often use it to decide where to locate stores), and most sane members of Congress, since the government uses it to distribute hundreds of billions of dollars in federal funding for programs like Medicaid, Section 8 housing, and highway grants. But conservatives have soured on the survey in recent years, supposedly because they find its questions intrusive and dislike that Americans can be fined up to $5,000 if they don’t agree to participate (this is meant to keep response rates higher and reduce the costs for the census, since hassling people over and over again is expensive). In 2012, the House GOP voted to defund the ACS entirely.

One of those in favor was Rep. Mick Mulvaney, Donald Trump’s nominee to lead the White House’s Office of Management and Budget. During a hearing in 2012, the House member from South Carolina repeatedly suggested that the government could ditch the whole survey and let the private sector pick up the slack. “Data has value to it. So what is unique about the stuff on the American Community Survey that you think that no one would want to actually get into this business?” he asked a witness. “I am thinking about doing this after I am out of Congress,” he added later. “If it is in such great demand, it costs us billions of dollars to send out 3 million of these things, I think I might be able to do it better than we do.”

To be clear, it is very unlikely that a private company would replicate the ACS in full; for starters, it covers smaller, poorer communities that aren’t necessarily draws for business, and some of the questions it asks—about family structure, for instance—are more important for social science and public policy than they are commercially valuable. But if Mulvaney sincerely dislikes the survey, he’ll now have a powerful perch at OMB from which to press for its defunding.

So don’t worry that the Trump team will start pulling the unemployment rate out of thin air. It doesn’t need to resort to straight-up supervillainy to erode the data that the world runs on.