Wednesday, January 18, 2017

Unemployment forecasts

The FRBSF (H/T Mark Thoma) has some unemployment projections coming out that say the unemployment rate is going to flatten out:


I have two different models (one is du/dt = const and the other is d/dt log u = const). In the former, we might see the leading edge of a future recession (however the projection is still very uncertain); in the latter it is much harder to make a case, and less flattening out is expected. Here's the former model:


And here's the latter [counterfactual path with no recession]:


The recession shocks aren't really predictable in either model ‒ they appear to be a random Poisson process with λ = 0.126 y⁻¹ ~ 1/(8 years). You can only tell when there starts to be a deviation from the linear (or log linear) decline.

The interesting thing is that in either model we shouldn't expect the flattening out over the two year period we see in the FRBSF model. We should expect either 1) a recession to start raising unemployment, or 2) a continuing decline (albeit at a slower rate). A constant unemployment rate won't happen, and in fact generally doesn't happen [1]. We might be able to test the various models here [2].

Anyway, this is how science gets done. It's a lot of vague guesswork made under uncertain conditions.

...

Footnotes

[1] The "metastable" states don't appear to last that long.

[2] There is a possibility of a "perfect storm" of a very broad (slow rolling) shock combining with the downward trend yielding a relatively constant unemployment rate. The likelihood that the parameters would work out just so is pretty low, however. Additionally, this perfect storm only thwarts forecasting. After the shock is about half over, we will be able to determine which model was more effective.

No comments:

Post a Comment