Market updates, Fair's model, and Sahm's rule

Fair's Model

Ray Fair has put up the past forecasts of the unemployment rate alongside actual data on his website (which is laudable!), but they're shown as a table which is hard to read. I've put them graphically in context of the Dynamic Information Equilibrium Model (DIEM) forecast from January 2017:

The context of my tweet about it was that I saw several people discussing Ray Fair's election model (which isn't very good), and wanted to add that his macro model wasn't very good either.

Market Updates

Also on Twitter (follow me @infotranecon!), I posted a couple of graphs of the S&P 500 and interest rates spreads yesterday (after the big fall in the market) which I haven't updated in awhile on the blog (here, here) ...

Here's an update of the 10-year rate forecast from 2015 (originally here). The BCEI means the Blue Chip Economic Indicators (wikipedia) forecast from 2015 which has done considerably worse.

Also, here's a longer run of the S&P 500 showing the previous shocks and overlays the early 2000s recession on the current correlated deviation:

Red bands are the non-equilibrium shocks to the S&P 500. The blue-gray bands are the NBER recessions. The green is the DIEM with in the no-shock counterfactual scenario, while the dashed lines show a counterfactual shock of equivalent size to the early 2000s recession. The blue band shows the AR process error band used to project from the projection data point — which by now matches up with the green model error band (see the discussion in the footnote here for more about this).

Sahm's Rule

Also on Twitter, I found out economist Claudia Sahm had been looking at the unemployment rate as a good indicator of recession for automatic stabilizers and had set up a threshold of about δu ~ 0.5 percentage point. I had put together a "recession detection algorithm" two years ago that used a threshold of δlog u ~ 0.17 (because the DIEM works in log space). Here's what those two thresholds look like on the unemployment rate forecast at the top of this post:

Sahm's rule is the green dashed curve, while the threshold I used is the gray dashed curve. Sahm's rule is closer to δlog u ~ 0.1, but the log difference gets smaller as the unemployment rate gets lower, so it's not exactly commensurate. But here's the threshold at work on the 2008 recession and 2014 mini-boom: