Employment-Population Ratio: war of the graphs

The comments on my last post were particularly good, and pointed to some alternative graphs. And, I think, to the important conclusion, that there is no substitute really for sitting down and doing some economics.

To recap, I passed on Torsten Slok's Graph of employment-population ratio and fraction of the population 25 to 54. Torsten's view is that the lowered trend in employment population ratio comes from older people retiring on their newly flush 401(k) savings.

Mike Nigro answered quickly pointing to a New York Fed study by Samuel Kapon and Joseph Tracy summarized here and producing the next graph.

Samuel and Joseph have the same basic aim: figure out how much the employment-population ratio is falling just due to demographics anyway, and hence is not particularly indicative of the state of the labor market.

Rather than just look at the number of 25-54 year olds They divide people "into 280 different cohorts defined by each individual’s decade of birth, sex, race/ethnicity, and educational attainment," and then estimate age-employment profiles for each bin using a large pre-recession dataset.

Source: Samuel Kapon and Joseph Tracy
Here is their estimated profile for white non-Hispanic men born between 1950 and 1959, in five levels of education. So, if everyone in the economy got older, we push the "normal" employment population ratio to the right. And if demographics shifted in favor of people with less education (or, not shown, women, minorities etc. with lower "normal" employment-population ratios) we would again sum up and decide that the overall "normal" was lower.

But, more sophisticated also means more assumptions one can quibble with.

They estimated profiles using data through 2013, i.e. including the recession. They used a "full set of year effects" so that a shift downwards in everyone's employment doesn't affect the profiles. But the recession didn't hit everyone equally. If the recession hits old people more than young, even with year effects, we will see a recession-induced fall in the profile.

More deeply, "there is no overall intercept for our demographically adjusted E/P ratio—only variations over time. To determine an intercept, we adopt the normalization that over the thirty-one years in our data sample any business-cycle deviations between the actual and the adjusted E/P ratios will average to zero."

In English, looking back at the middle graph, you can slide the blue line up or down to anywhere you want it. Samuel and Joseph put the blue line in the middle of the red lines. But they could just as easily have had the blue line clip the tops of the red lines, and call it "potential" employment in the Keynesian tradition. Or higher still, they could have assumed that the employment of highly educated white men is "normal" and scored everything else as a shortfall. Torsten's graph uses a different scale (right) for the blue line as the red line (left). So there as well, the relative position of blue and red lines is totally arbitrary.

To be clear, all the authors are perfectly clear about this limitation. But if you look at the middle graph and conclude that we're close to some sort of "normal," you're jumping to a big conclusion.

Finally, neither graph by itself says everything is fine in the labor market. Is it "normal" that 40% of 50 year old white men with less than high school educations do not work at all? Is it written in human biology that there is a cliff in employment at 62 and 65? Or do both of these facts reveal people responding to economic incentives, many set up by government programs (in the latter case, medicare and social security rules) that may be less than ideal? Even if the employment-population ratio were to follow the blue line exactly, that does not mean all is well.  Our issue is no longer "recession" but "perpetually low growth."

So far, though, Samuel and Joseph's graph produces about the same result as Torsten's. Actually, it's stronger. Torsten felt that 54-65 year olds were retiring in unusually high numbers on stock market wealth -- something they didn't do in 1999 and 2006 -- while Samuel and Joseph say the decline is all demographics and the usual age earning profile.

Oliver Sherouse chimed in, linking to a graph he created from the St. Louis Fed's excellent Fred database.

"Just a gosh-dern minute here. If you graph the E/P ratio of just the 25-54 year olds, you get basically the same curve."

The blue line is the celebrated overall employment-population ratio. The red line is the employment-population ratio of 25-54 year olds. Good point Oliver! Joseph and Samuel could just as easily produce the time-series of employment in their 280 buckets. (Presuming they haven't done that somewhere else.)

Source: New York Times
A similar point from the New York Times, via Marginal Revolution, at left.

Still, none of these graphs really tell us how much of the overall decline in employment-population ratio is due to these "within-demographic" changes vs. the change in demographics and the "usual" differences in employment across demographics.  It looks like there is a bit of both.

Into the fray, Jim Bullard, St. Louis Fed President, has  a nice speech,  two weeks ago, a refined version of which will appear next week in the St. Louis Fed's economic review. The Fed too is puzzled -- unemployment is down, but the employment-population ratio has not recovered. Still weak "demand?" Now "supply" that the Fed can't do anything about?

Jim reviews a large scholarly literature. What I got out of this is, so far nobody has really clear answers. Well, knowing there isn't a clear answer out there is knowledge.

Echoing my thoughts above Jim says
Much of the literature I have reviewed uses the same basic idea: Certain demographic groups have a certain propensity to participate in market work, and one of the main things we need to do as economists is project the number of people in each of these groups in order to determine a reasonable estimate of the expected (or “normal” or “trend”) labor force participation rate in the U.S. economy. Much of the literature concludes that demographics have contributed substantially to the observed decline in U.S. labor force participation since 2000.
Still, the literature as a whole is a bit hollow. Simply saying that people in certain demographic groups tend to make the participation decision one way or another does not do enough to analyze the incentives of household labor supply decisions. The more we know about the details of the household labor supply choices, including choices to participate in market work, the better we can predict the impact of policy on labor force participation. Furthermore, we would like these decisions to be part of the macroeconomic model, as Erceg and Levin suggest.
Here Jim speculates about the recent, and very worthwhile literature on "household production" that models the decision to work or not. He does not speculate on how much those decisions are distorted by policy. 40% of 40 year old white men are not perpetually choosing to be out of work, facing the full cost of that decision, to be artists, take the kids to soccer practice or even to repaint the living room walls.

The bottom line, I think, is simple. We want to know, is the employment-population ratio low because of generic lack of "demand," "discouraged workers" who have simply given up looking? Is it due to lack of "supply," either due to preferences and demography or due to labor market distortions? Is it due to the "normal" change in desire to work as people get older? What are the economic reasons that minority and less educated people work less than high education white males? It's as offensive to economics as to the people involved to write these employment-population differences as innate features of human beings. Is the decline due to the interaction of demographics with the disincentives provided by existing government programs? Is it due to the disincentives of new government programs -- long-lasting unemployment insurance, expanded food stamps, looming ACA, etc. etc? Is it due to the interaction of a recession with government programs -- people normally work, but once they get hit by a recession, they get stuck in the non-work state? How much is due to resistance, natural and artificial, to moving? How much is the skills mismatch? Job opportunities change from construction in Florida to computer programming in California, and people don't follow. Simple aggregate labor demand and supply can't address that.

Sadly, no simple graph answers those questions.


From John Taylor's blog, Chris Erceg and Andrew Levin's  plot of the employment-population ratio and two projections of demographic effects, made ahead of time.

Post a Comment

Previous Post Next Post