Next month, this dynamic information equilibrium model (DIEM) forecast will be 2 years old . It's been pretty much spot on (shown alongside some of the forecasts published by the FRBSF's FedViews):
It's biased a little high with today's 3.7% number, but compared to the competition — the FRBSF forecast had us at 4.7% compared to the DIEM's 3.9 ± 0.2% — the miss distance is 5 times smaller (1.0 pp vs 0.2 pp):
The FOMC forecast a 4.5% unemployment rate average for 2018. With almost all the data (which is why the average point is gray inside a black circle instead of white), it's looking closer to 3.9% — a 0.6 pp difference. The DIEM annual average is 4.1% — a 0.2 pp difference (3 times smaller).
I'm also tracking a couple of forecasts (versus the FOMC and the CBO) based on forecasts made this year. The DIEM is a little lower because the data from 2017 gives us a better indication of the size of the 2014 shock to unemployment (the "mini-boom"). I show some possible counterfactual recessions in the DIEM consistent with the CBO's forecast (i.e. if the CBO forecast is accurate through 2022, then there'd be a recession shock in the DIEM)
As of writing this, the S&P 500 was down almost 2%. I thought I'd update the S&P 500 forecast (also from almost two years ago , and also pretty spot on) to show some more of the recent volatility. The last update was here. We're skirting the edge of the AR process band, but we're still inside the 60-year volatility band of the DIEM model, the lower edge of which (marked by a red dashed line) indicates the need to posit another shock to the DIEM.
The median interest rate spread (which is about the same as the average or the principal component per the original model description) continues to trend downward. Note: this isn't a DIEM model, but rather a simple linear model of yield curve inversion as a leading indicator of recession.
 The model itself was born in a January 10, 2017 blog post when I derived it from an information equilibrium relationship for JOLTS data similar to a matching model.
 I feel like I just have to show the S&P 500 forecast before all those points: