The job market is going to turn around; any day now, which should give the people waiting 5 years for a job some hope, even if delusional.

I wrote earlier about how the Beveridge Curve, used to measure the overall state of the labor market, had shifted to the right. It has caused more than a few analysts to think that a left-shift COULD occur any day now, bringing normalcy to the labor market. There is about as much statistical evidence of that as there is that death rays from Mars COULD kill us all.

Playing with statistical models is a great way to understand social facts. Which is exactly what I did with Beveridge Curve data from the BLS, running regression models to filter out statistical “noise.” Some interesting things pop up.

In order for the Beveridge Curve to even exist, the INDEPENDENT Variable has to be the unemployment rate. In other words, job openings are dependent upon the unemployment rate being low. In a labor market where 16 Million people are looking for 4.5 million jobs (U6), that’s just ridiculous on its face. But I played with that in the regression model anyway:

The black line is the zero intercept point. Both the standard curve and the regression model look pretty much like a Beveridge curve would. At the zero intercept point on both the raw plots and the regression, a near-perfect competitive labor market would have a 4.5% unemployment rate with 4.5 million job openings, which is pretty much where we were in 2006.

However, just based on U6 data, this is a very slack labor market. It’s still slack with U3 numbers. But in no way can the labor market be called “tight,” or “competitive.” So this would reverse the independent variable. In reversing the variables, instead of job openings being dependent upon unemployment rates, unemployment rates are dependent upon job openings; which really makes more sense in a 4:1 worker-to-job ratio, slack labor market. The more jobs that open up, the more people will go back to work. So how’s that look?

It simply looks like a reversed Beveridge curve. Except the zero intercept point, representing a near-perfectly competitive labor market, is completely different (black line). Instead of a 4.5% unemployment rate with 4.5 million job openings being competitive, now a competitive labor market is 2200 job openings at 8.5% U3 Unemployment. That kind of sucks! And here’s the regression, which gets more interesting:

After filtering out for statistical noise, it doesn’t look like a curve at all, but just a circle. The zero intercept point is still 2200/8.5.

Which while counterintuitive, makes sense when you look at t-tails in the distribution:

The Variance (and thus, the SD and SS) are way off for job openings, skewing the t-tail to the left, meaning that there is a large variance for fewer job openings. This will automatically skew the regression for unemployment when the independent variable is “job openings.” So we can look to the MEDIAN for a reference point. The MEDIAN job openings is 3.75 Million and the MEDIAN unemployment rate is 7.5% over the course of 8 years on monthly data.

It’s not that the Beveridge Curve is “wrong,” or even “bad.” It’s more that in a slack labor market, people are reading it wrong. It’s not measuring the overall state of the labor market, it’s measuring the slackness or tightness of the labor market. The curve COULD shift to the left again. Or it may not. No inferences (or assumptions) can be made from Beveridge curve data other than there is a lot of slack in the labor market – 8.72 Million people (U3) looking to get 4.5 Million jobs.