Because economists can’t seem to measure much these days using their traditional macro models, someone decided to dig up, and dust off an old 1958 Model of the labor market called the Beveridge curve. It’s gotten a lot of attention on the “inter-webs.”
The model is amazingly simple: as unemployment rises, job vacancies drop. As job vacancies increase, unemployment drops. Ordinarily, under 1958 rules, it could tell us a lot. However, I suspected that it isn’t measuring much in 2014, so I downloaded some data, and ran my own model, comparing it to the Beveridge Curve. I was right; it isn’t measuring much, and the traditional explanations for why isn’t matching the numbers.
First is current data for an actual Beveridge curve. In this model, I use both the traditional U3 rates, and compare it to the U6 rates (those only marginally employed + unemployment). This is where things get funky:
While U6 is also not an accurate measure of real unemployment, it’s way more accurate than U3, especially in our part-time/low wage economy where more labor is likely to be marginalized. The curve is not just pushed way more to the right, but also shows more volatility. It also indicates a wider spread between those who are marginalized workers, and those who are not. This would indicate that the increase in open jobs corresponds mainly to marginalized workers; low wage/Part-time.
The traditional argument is that this curve indicates that there is a skill mismatch, because the same number of job openings is garnering a higher unemployment rate now than in 2005 (a shift to the right).
So then I ran the data for a Beveridge Curve that uses more accurate unemployment data: the Labor Force Participation Rate. These are people of working age, who have a job in relation to those of working age that do not have a job. Reasons may vary, but the fact is that the LFPR has been dropping since the Great Recession, and at its lowest levels in 30 years. How does that relate in Beveridge terms?
It’s looking more funky, if not downright weird. And there’s some points to this:
- If the traditional curve was because of a skills mismatch, then we should not see any change in the LFPR; yet we do.
- The curve also no longer shifts to the right; it goes into the unknown abyss.
- If all of these people are leaving the labor force, then we should see a tighten of the labor market, spurring an increase in median wages. Yet wages have been flat.
- If more jobs opened up (magically), then there should be more people re-entering the workforce (for all of those awesome wages), and yet we see no movement in the low LFPR.
It only gets weirder from there. I then decided to compare (in Beveridge-esque fashion) the Labor Force Participation Rate (LFPR), U3, U6, and the Quit rate. The quit rate is from the Bureau of Labor Statistics that shows the number of job openings caused by people who quit. That’s been hovering around 1.8% since the Great Recession, from its pre-recession high of 2.5%.
While the correlations are not a smoking gun, they are statistically significant.
- The relationship between U6 and the LFPR is stronger
- The trendlines are disturbing, as it shows unemployment increasing with lower LFPR. We expect to see the opposite in a tight labor market.
- Under current trends (trendlines notwithstanding), in order to get to a pre-recession unemployment level (on either U3 or U6) we have to drop the LFPR to roughly 60% (or lower). That means about 4 Million working-age people need to leave the labor force in order to get unemployment down to 4.4%
- Less people are quitting their job with lower LFPR. We expect to see the opposite inder traditional “assumptions.” If LFPR rate is dropping, the traditional assumption is that it is because lots of people are quitting their jobs, or retiring. However, as LFPR drops, less people quit, indicating that people don’t want to loose what ever job they have.
- More people tend to be more marginally employed over time. The spread between U3 and U6 has grown over time.
And none of these data represents changes in population, increases and decreases in the labor force, or socioeconomic pressures (such as a collective $1.2 TRILLION student loan debt) that could easily change the labor market landscape.
In the realm of inaccurate measures and models, one explanation could be that the JOLTS job openings data themselves are not accurate. There are 4 Million Job openings and over 16 Million people looking for jobs; according to the U6 numbers, which include people looking for better jobs. That’s a 4:1 ratio. Historically, unemployment rates on both U3 and U6 have been much lower with these number of job openings. It’s hard to imagine that out of 16 Million people looking for jobs, that HR managers can’t find anyone. This raises some basic methodological questions:
- The Jobs Opening data is gathered from surveys, which are largely self-reported. Could HR managers simply not be filling open positions in an effort to lower marginal cost?
- Could they simply be fudging the numbers?
- Is there some level of discrimination going on here?
- What exactly are HR managers and CEOs thinking when trying to fill alleged openings?
- Also consider that those who are unemployed longer take longer to find their next job, while people who are unemployed for the shortest time find jobs quicker. What socioeconomic forces are in play here in determining who is unemployed longer?
The JOLT Job Openings data clearly doesn’t tell the whole story, and if it doesn’t tell the whole story, then how accurate can the measure be?
While these data may not tell the whole story either, it tells a much better story than curving two groups of data where neither tells the whole story. LFPR is a better measure, U6 is a better measure, and looking at quit rates shows collective social behavior, rather than assuming it.
When looking at more and more accurate data to figure out what’s going on in the labor market, two things become clear: first is that more research into people’s collective behavior needs to be done, and second is that a lot of cold water just got poured onto labor from the Beveridge Curve.