The Scandal is What's Legal...

Last week, the Trump administration nominated E. J. Antoni, chief economist of the Heritage Foundation, to be the new commissioner of the Bureau of Labor Statistics. If confirmed, he would replace Erika McEntarfer, who was fired by Trump on the ridiculous assertion that she and the BLS had purposefully distorted the jobs report to harm the administration politically. Antoni is a longtime critic of the BLS. Among other things, he has suggested in the past that until the data and statistics can be assembled more accurately in real time, the monthly jobs report should be made quarterly, and the entire process should be reviewed and potentially reformed, dramatically.

That has, predictably, led to howls of protest that his appointment, on the heels of the firing of the last head, was a step toward the funhouse distortion of reality that has long characterized authoritarian regimes, the ‘data is what I say it is’ approach of dictatorships. “E.J. Antoni is completely unqualified to be BLS Commissioner,” Harvard University economist Jason Furman, who worked for the Obama administration, wrote on social media. “He is an extreme partisan and does not have any relevant experience.”

And yet, qualified or not, Antoni is not entirely wrong. In fact, he is largely correct, at least insofar as the way that the BLS collects data and the way it tries to gauge the unemployment rate is woefully archaic, increasingly problematic, and long overdue for an overhaul.

It is more than ever the case that we must ask the basic question: “If Trump or his partisans say it, does that mean it’s not true?” In the case of the flaws in our macroeconomic data, the answer is unequivocally that our data is flawed, our methods out of date, and the entire question of what we are trying to measure needs to be revisited. No, there has not been any politicization of the data or the stats, but yes, Houston, we have a problem.

More than ten years ago, I wrote a book called The Leading Indicators which told the story of how our modern suite of economic statistics came to be. Statistics are stories we tell about data, and both the data that underlie our leading indicators and the stats themselves are all a product of the New Deal era, give or take a decade. Before that, there were no consistent, reliable government statistics about what we now call “the economy” anywhere in the world.

All our stats – unemployment, inflation, GDP above all – were invented to measure 20^th century industrial economies. The methods that they depend on – often surveys – were well-suited to a time when people faithfully responded to questions that the government asked. They were also well-suited to an economy that was primarily about how much stuff was made (or in the case of farm output, grown). They were never designed to be a complete picture.

GDP, for instance, was invented largely by the economist Simon Kuznets in the 1930s as a set of national income accounts that would provide the Roosevelt administration with a dashboard and snapshot to assess whether the programs of the New Deal were working as intended. One major decision was not to include volunteer work or “domestic work,” namely, the work predominantly done by women to maintain a household. That decision was not meant to suggest that these did not contribute to national output or prosperity, simply that capturing them statistically was too arduous at the time. Kuznets himself famously warned about generalizing too much from GDP, “The welfare of a nation can scarcely be inferred from a measure of national income.” Yet, today, we do that routinely, over-relying on GDP numbers as the only gauge of whether “the economy” is thriving.

Unemployment statistics started to be assembled by U.S. states in the late 19th century, but national numbers were a product of the 1920s and 1930s, spurred first by Herbert Hoover as Commerce Secretary and then as president. It may come as a surprise to learn that one thing that first bedeviled the early statisticians was how to define “unemployment.” Even today, that category is not as simple as it sounds. It is not a binary case of either you have a job and are employed or you don’t and are unemployed. To be unemployed, you have to be looking for a job. If you get fired and haven’t looked for a job in the last four weeks, you aren’t unemployed. You just aren’t in the labor force. To get an unemployment rate, you have to have the size of the labor force, because the unemployment rate is a percentage, not a raw number. And if you say in a survey that you have looked for a job in four weeks but also drive for Uber or do some gig work, you might be unemployed statistically but also earn more money than if you had a job at McDonald’s.

Read The Zombie Numbers That Rule the U.S. Economy (2014) by Zachary Karabell

And that raises a further issue: unemployment statistics treat any job as an inherent plus, and any unemployment as an inherent negative. But having a minimum wage job in many parts of the country places a person barely at the poverty line if that person is a primary breadwinner in a family and often earning less than they would get from unemployment benefits, Medicaid and SNAP (food stamps). Many jobs are so crappy that they in no way “feel” like a good thing, which may explain the disconnect between very low unemployment rates (other than the Covid anomaly) over the past decade and ever-higher levels of public discontent with “the economy.”

Even getting the data is becoming harder and harder. Unemployment stats, like inflation numbers, are highly dependent on surveys. In the case of the unemployment report that comes out monthly, it relies in part on a survey of 60,000 people, and finding people to answer those monthly surveys is becoming harder, making it that much more challenging for the BLS to get reliable information. The reports from businesses about their payrolls are more consistent, but one of the reasons for the huge revisions of month-to-month numbers is the increasing difficulty of getting timely and consistent survey responses.

None of these issues should be partisan. There is a high degree of consensus that good data and consistent statistics are a vital tool of economic decision-making. It’s also evident that in a world where so much data is digital and increasing amounts of work are as well, our statistical universe should be more informed by the digital sphere, and soon enough aided by AI. But the BLS, though staffed by thousands of diligent individuals who are dedicated public servants driven only by a passion for data and getting the statistical story right, has not had a major analytical overhaul in decades. Yes, its methods have improved, and the array of statistics it creates, from regional variations to employment by age, race, education and a host of other parameters, has grown. But the core methodologies remain largely a product of the mid-20^th century.

That is no longer the world we inhabit. Manufacturing jobs as a percentage of the total have gone down substantially. In 1950, manufacturing jobs were about 30% of all employment. Now they are less than 8%. Small businesses now have a larger share, as does gig work. It is far easier to capture statistics of a manufacturing workforce, rooted in one place producing tangible output, than it is to assess an increasingly service-oriented, digital economy. Our economic matrix has changed far more than our statistical one.

These issues are well known and understood within the wonky worlds of academia and government. What has been lacking is the will and the money to revise how we measure what we measure. The result is that we are using a statistical dashboard designed for a different world, and we are thus chronically mismeasuring our reality. That is one reason for the increasing chasm between what our numbers tell us and what people experience. Much of the focus is on the question of why so many are negative about “the economy” when the numbers tell a better story, but there is also the equally pressing issue of why there is so much wealth creation and economic resilience in the face of headwinds such as inflation and tariffs. We aren’t flying blind, exactly, but we aren’t seeing the whole picture.

Perhaps the upheaval caused by the precipitous firing of the BLS commissioner will lead to a wider examination of where we should go from here. Perhaps it will generate some of the political will to begin the arduous process of real reform. Given how quickly everything becomes bitterly partisan, that seems a high bar. But we shouldn’t fool ourselves into thinking that the status quo has been working when it comes to how we measure our economic world.