The Corner

Politics & Policy

Understanding the Data: The American Working-Class and the Economy

(Jonathan Alcorn/REUTERS)

In response to Re: The Roots of American Working-Class Discontent

In my opening salvo of this exchange, I argued that the problems of non-college-educated Americans are not primarily due to economic decline (because there has been no economic decline). In his reply, Oren Cass reassuringly begins by saying, “we generally agree on the data itself.” While I’d like to agree with him, it seems clear to me that the whole crux of the debate is that we actually don’t agree on the data at all — not how to interpret it and not how to measure the things that we care about in the most relevant way.

Oren compares me to a defense attorney marshaling selective evidence to save my client. Shockingly, I would put it differently. I once had very different ideas about the state of the economy than the ones I hold today. You could say that in this trial, I was on the prosecution team with Oren. (I even dabbled in Marxism once upon a time. Hell, I worked for ACORN.) But in the course of evaluating the evidence in the case, I came to change my view of the accused. If I have to be the defense attorney in this metaphor, I am one who switched sides because I went where the evidence led me and who would drop his client like a bad habit if new evidence of his guilt were introduced. (I should also say that if I really wanted to cook the books, I could do a lot better than showing that male pay at the 20th percentile has been stagnant!)

Oren’s suggestion that I am offering a biased take on economic trends takes me back a decade, to the time when I was a New Democrat facing the same criticisms from liberal analysts. In fact, it is the declensionist drift on the right — of both populist Trumpism and the reform conservatism I now consider to be my ideological home — that has me worried. I can think of little worse than a politics in which both left and right have fundamentally misdiagnosed the nature of our problems.

This post will be even more in the weeds than my opening one, unfortunately, but it’s necessary to get the measurement and interpretation of data right, which is necessary to help the non-college-educated Americans about whom Oren and I both care. Oren accuses me of only making adjustments to the data that produce stronger wage-growth estimates, suggesting alternative adjustments that would make for more “balanced” estimates. But for the most part, those suggestions just reflect the sort of misperception about “decline” that I have long criticized.

For instance, Oren says I should factor in longer commuting times. But between 1980 (the earliest year for which data are available) and 2015, median commuting time rose by three minutes, up from 22 minutes. That’s less than one song on the satellite radio. The share of workers living and working in different counties was 26 percent in 1970, while today it has risen to . . . 27 percent. Between 1969 and 2009, the share of car trips and of miles driven constituted by commuting declined, so to some extent, these modest increases in the home-to-work drive just reflect the fact that we’re driving more in general.

Another suggestion he gives is to factor in rising income volatility, but I wrote my dissertation on that topic (read it on my ugly, broken, and outdated personal website!). The  increase in volatility has also been modest. The share of working-age adults who experienced a one-year income drop of 25 percent or more rose from about 7 percent at the end of the 1960s to 9 percent on the eve of the Great Recession. And because some volatility is anticipated — even planned, as when women temporarily leave the labor force to have a child — it’s not clear that we should worry about even a small rise in volatility. Earnings volatility actually fell among women workers. These questions are far from settled because of measurement issues, but it’s hardly the case that one has to strain to find research that suggests little change in volatility (e.g., from the Congressional Budget Office).

My estimates already take into account another suggested adjustment offered by Oren — changes in the value of health insurance. That said, the government economists who measure healthcare inflation think that the inflation adjustment I use understates the rising value of health insurance, contrary to the assumption of Oren and many liberal interlocutors of mine over the years. (I’d ask what kind of defense attorney declines to seize on such an opportunity to make his case stronger?)

Oren suggests that I adjust for rising payroll taxes, but that only makes sense to the extent that worker preference for cash over social insurance has changed (open question, as far as I know) or to the extent that social insurance benefits have gotten stingier over time (they have not).

He also recommends that I account for the rising share of men who no longer work or look for work (men I call “inactive”). Lower-earning men at the margin were more likely to work in 1973 than their counterparts today, meaning that we observe wages for more of them in the data. Those wages pull down the earlier wage estimates more than the wages of marginal men do today (fewer of whom have wages we can see in the data).

I looked at this adjustment five years ago, in response to wildly inappropriate claims that men’s median earnings had fallen by 28 percent between 1969 and 2009 (a paper I’m grateful Oren cites in his book). A proper accounting does not much change the picture over the long run. Wage trends should be confined to the able-bodied, and over half of prime-age inactive men report being disabled. Some of them could and should work, but even in the 1960s, when receipt of federal disability benefits was relatively uncommon, most inactive men were disabled.

Furthermore, among the able-bodied, we cannot assume that all of these inactive men would have below-median — and especially below-20th-percentile — wages if they worked. About one-third of inactive men report either being in school, a homemaker, a caregiver, or retired. It is far from obvious that these men have lower wages than the typical working prime-age man. Combine these men with the disabled, and just 10 percent of inactive men remain. And some of these men live in higher-wage areas of the country and might be above the median or 20th percentile if they worked. (Geographic cost-of-living adjustment, not mentioned by Oren, is probably the single issue with the potential to change my mind about wage trends, but we don’t really know how to implement that yet.)

Finally, at least some of the impact of rising inactivity on male wage trends is reflected in the data. Men who start a spell of inactivity in, say, May (perhaps a long or permanent spell) still report their wages in April. To the extent their absence from the data in May makes wage trends too rosy, their presence in the April data makes up for it. About one-third of the long-term rise in prime-age male inactivity comes from men who only temporarily leave the workforce, and about half of that rise can be explained by the fact that the partners of married men have become richer.

In the end, declining labor-force participation ends up being yet another false sign of economic deterioration. If fewer men are in the labor force over time, that can reflect a shift in labor demand — a weakening labor market — or a shift in labor supply (less interest in employment among workers). Everyone assumes it reflects a shift in demand, as Oren does when he says that this trend reflects more and more men “giving up on work,” but the evidence for this view is pretty weak. Labor-force participation has been falling among prime-age men since the 1930s — through the strong labor markets of the 1940s, 1950s, and 1960s. It has fallen among college-educated male workers too. And I’ll note that, as was the case with wages, women’s labor-force participation trends look much, much better than men’s. They aren’t giving up on work.

The Bureau of Labor Statistics has three measures of joblessness that are intended to capture weak labor demand in a better way than the official unemployment rate. All three include “discouraged workers” who are available for work and want a job, but have given up looking out of economic frustration. All three show the same trend as the unemployment rate (which shows historically low joblessness). Well under 10 percent of prime-age men out of the labor force meet the criteria to be designated “discouraged workers,” and that was also true in the early 1970s (Figure 6).

Federal surveys ask inactive adults whether they want a job. Only one in four inactive prime-age men do, and they account for only a quarter of the rise in inactivity within the group. Put another way, three quarters of the decline in prime-age male labor-force participation has involved men who say they do not want a job.

How to interpret this fact? If male pay has declined by a lot, then we might assign most of the blame for rising inactivity to the labor market and perhaps to federal economic policy, as Oren does. But as I showed in my first post, male pay has not declined. That strengthens the case that the drop in labor-force participation among men is more about labor supply. It is easier today to get by without work than it was in the past, due to the expansion of the federal-safety net, greater work among wives and girlfriends, and fewer obligations toward children.

I’ve gone on too long again, so I’ll save my discussion of Oren’s other data points for the next round. (Spoiler: I have issues.) A meta-question that Oren raises is why the official and most widely used economic indicators give unduly negative impressions of how we’re doing. The only answer I’ve given in the past, which is by no means fully developed, is that everyone has incentives to accentuate the negative — researchers, journalists, policymakers, funders, and activists. (Scroll to the end of this scintillating piece on inflation adjustment, starting with “There is also the reality . . . ”)

The indicators we use today reflect the past decisions of generators and consumers of data, decision-makers who have their own, generally unrecognized, biases. When academically trained researchers and journalists are drawn from overwhelmingly liberal ranks, for instance, we should not be surprised if many economic indicators have evolved to incorporate adjustments that reinforce the view that our economy is unjust, serves only the affluent, and should be backstopped by a more generous safety net — and that they do not incorporate adjustments that point away from those conclusions. To be clear, I am not alleging any explicit conspiracy; in the presence of a homogeneous group, motivated reasoning, ideological bubbles, and confirmation bias are enough.

No further questions for this witness, Your Honor. Back over to you, Oren!

Scott Winship — Mr. Winship is a resident scholar at the American Enterprise Institute and its director of poverty studies.