JOLTS day 04-Jan-23

Not much new information, but did clean up the graphs a bit. Plus what happens if we use JOLTS hires to forecast unemployment?

This one is not coming out as an email (don’t want to flood your inbox) but wanted to show the update of the forecasts from the latest JOLTS data. Overall, not much changed.

Job openings returning roughly to the pre-pandemic dynamic equilibrium (gray dotted line). The Monte Carlo (cloud of thin gray lines) tightened up around the mean forecast (red line).

The red dashed line is the post-stimulus equilibrium prior to the recent downturn in the various JOLTS measures (this legend applies to all the graphs).

Here are quits. Overall — same story. This one is a bit lower than the pre-pandemic equilibrium but not by much given the uncertainty in the forecast.

The theme of essentially no change to the expected path follows through to hires:

This one is expected to drop measurably below the pre-pandemic equilibrium. A few years ago, I used the fact that hires, unemployment and wage growth all had similar shock structures to map them all onto each other using a lagged log-linear transformation — essentially scaling and shifting the curves by taking log(y(t)) = a log(x(t+dt)) + b. It turned out that shocks to hires preceded shocks to unemployment by almost two quarters (the lag between hires and unemployment with |dt| ~ 0.4-0.5 y). If we apply these transformation parameters to the Monte Carlo simulation from hires to the expected path of the unemployment rate, we get an expected rise of about 2 pp to 6% by the end of 2023 with an uncertainty of about ± 1 pp:

Current unemployment data appears to be undershooting (or lagging) this estimate so there is some possibility we might see the low end of this. Note that is almost exactly the same size as the 2000s recession (from a low of 3.9% to a peak of 6.3% over a period of about 2 years). In fact, except for the longer duration, the 2000s recession would be mostly consistent with the current projection using the hires data:

(Red is the shifted unemployment data from roughly 2000 to 2005.)