Shares of job growth aren’t the best metric of labor market health for less-educated workers.

There’s little doubt that workers with only a high school diploma — or less — have it tough in the American labor market. Their unemployment rates are higher than those for workers with even just some college, and their wages are lower. There was plenty of suffering from the Great Recession to go around, but less-educated workers were hit particularly hard, and have recovered more slowly. Looking forward, as the US economy shifts towards a greater emphasis on services and cognitive tasks, it’s unlikely that this disadvantage will reverse itself anytime soon, if ever. And the story gets even less optimistic when you look at noneconomic metrics like health.

But a recent report from the Georgetown University Center on Education and the Workforce found a way to raise this already pessimistic assessment to stratospheric levels. The report itself is an interesting and robust deep dive into government data on the relationship between education and jobs since the recession. The authors, however, lead with the headlining-grabbing conclusion that 99 percent of jobs added during the recovery have gone to people with at least some college education.

It’s tempting to read this result as nothing short of a catastrophe for the labor market prospects of those with a high school diploma or less. But the truth is that while, yes, these workers have faced huge hurdles in the wake of the financial crisis, their recovery has not been as dire as this bullet point might suggest.

Briefly, what the authors appear to have done to arrive at this result is straightforward. They dated the trough in employment [1] after the recession as happening in January 2010, and so defined that month as the starting line of the recovery. Between January 2010 and January 2016, they find that the level of employment of all workers 18 and over rose by 11.6 million, while the level of employment of all workers 18 and over with at least some college rose 11.5 million. Hence the 99 percent.

The problem with this approach lies not with any error: the authors cut the data in a couple of mildly nonstandard but justifiable ways that prevent quick verification, but using official summary tables as the basis for a spot check, the calculation appears correct as far as it goes.

Instead, the problem is that over an extended period of time, this approach overstates the weakness of the labor market for high school-only graduates because it lends itself to the interpretation that these shifts in employment are entirely due to labor demand — the qualifications that employers want to see in the people they hire. But a good chunk of this story is one of changes in labor supply — the educational composition of the people who have or want work. And a lot of this shift is in turn being driven by the ageing of the US population, a long-understood phenomenon that pre-dates the recession and would have occurred even in its absence.

Consider that by the authors’ own definition, the recovery is now in its seventh year. During this time, the size of the labor force 25+ years old with a high school degree or less — that is, people with or actively looking for a job — has fallen by 3.5 million, even though the overall population has risen. This makes sense since the US population is aging overall thanks to the demographic dominance of the Baby Boomer generation. The young people coming into the labor force now are far more likely to have a college degree than the Baby Boomers who, now reaching their mid-60s, are leaving the labor force for retirement. This trend towards more education among incoming young workers predates the Great Recession. It’s also plausible that workers further along in their career, sensing a rising advantage to having a college degree, went back to school during the recession and initial recovery and earned their bachelors or associates.

The result of these dynamics is that even if the Great Recession hadn’t happened, we would have expected a shift in the labor supply towards more college-educated workers. Just as a hot bath gradually cools down if you turn on the cold water and pull the drain plug, so too was the trend of higher educations inflows bound to make employment in America more dominated by college graduates. That takes away a lot of the meaning of the 99 percent number from the Georgetown report.

So rather than ask how much of the aggregate change in jobs went to those with a high school diploma or less, a slightly different but better question is how successful these folks have been at finding jobs. Framed this way, we can answer that question in a way that won’t be as affected by the composition of the labor force.

The graphs below show the job finding rates for these workers: the probability that, given you’re without work in one month and have a high school education or less, you’re employed the next month.The data show that workers with high school-only or below took a deep hit during the recession, but their success at finding jobs has gotten gradually stronger.

The first graph just looks at all of these workers without any adjustments for underlying demographics. It shows that in 2007, someone with a high school education or less had roughly a 36 percent chance of landing a job the next month. By 2013, this had fallen to 33 percent. This may sound like only a small change, but it had devastating ripple effects on the labor market outcomes of these workers. And while this approach shows that job finding has gotten better for people with a high school diploma or less — recovering to about 34 percent over the last 12 months — the rate is still well below where it was before the crisis. So while the conclusion that there’s been no improvement is clearly wrong, the raw unadjusted job finding rates for these workers don’t appear to be healthy yet.

However, remember that people with a high school diploma or less skew older, and older people are less likely to want a job in the first place [2]. So we would reasonably expect this vanilla job finding rate to fall over time as an ever — greater share of nonemployed people with a high school education or less are retired with no intention of getting another job. In this case, then, when looking at job finding rates over a long period of time in the face of secular demographic trends, one year’s rate is not strictly comparable to another year’s, particularly when they’re six years or more apart.

To account for this, the graph below calculates the same job finding rate but with the added constraint that the proportion of the population in each age and sex group stays constant at their 2007 levels. So the percent of the population that is, say, male and 57 stays flat rather than rises. This makes the job finding rate demographically-comparable over time since it prevents the mix of older and younger people from affecting the results.

What it shows is that once you account for the ageing of the population, the job finding prospects for workers with a high school education or less has healed considerably more than the prior graph suggests. In fact, on average, job finding rates for these workers are back within their pre-crisis range and not far removed from their 2007 highs.

Again, the point here is not to argue that the recovery has been gang-busters for high school graduates, or that they don’t still face real hurdles, or that their outlook going forward is overwhelmingly positive. You’ll notice that I haven’t dug into the types of jobs these workers have taken, or their wages, and I cannot emphasize enough that these are serious and valid concerns.

But on the narrow point of employment, the finding that “99 percent of jobs added in the recovery went to workers with some college” implies that less-educated workers have had virtually no success in finding jobs since 2010. This interpretation however isn’t accurate. Among the shrinking share of the population with a high school education or less, they have had low but growing success at find work since the Great Recession. And when we account for the fact that they are ageing, their success is far more apparent and is even approaching something like a full recovery.

[1] The authors are using the measure of employment in the government’s Current Population Survey (CPS), commonly called “household employment”. This is the broader of the two official measures of employment and is a count of all people who tell the government they have any job, regardless of which industry they’re in, whether they’re self-employed, or if they’re working for pay or not. People with multiple jobs are only counted once. The other official measure, taken from the Current Employment Statistics (CES) and often called “payroll employment”, is a measure of the number of jobs that firms tell the government were on their payrolls and received pay. This measure excludes the self-employed, private household workers, and agricultural workers, but in principle counts each job held by workers with multiple jobs separately.

The basic trade-off for a researcher is that the CES has a much larger sample and so is a more reliable estimate of month-to-month changes in employment (as it defines it), but the CES hardly includes any demographic data at all and the government doesn’t make its underlying raw “microdata” available to the public, so researchers are stuck with the summary tables that the government releases. The CPS is far more volatile and unpredictable from month-to-month, but asks a rich array of questions about demographics, education, and labor status, and the individual responses are made available to the public.

[2] One question that logically follows is why not look at job finding rates among just the unemployed (those who are actively looking for work) rather than anyone without a job regardless of their job intentions. The answer is that I chose to broaden the focus to all nonemployed because job finding among people not in the labor force is a meaningful part of the labor market story in any recovery including this one, even among people who say they don’t want a job. Also, beyond downward demographic effects, the dynamics of labor force participation in this recovery are as of yet unclear and beyond the consensus of economists.

One clap, two clap, three clap, forty?

By clapping more or less, you can signal to us which stories really stand out.