Dwaine van Vuuren

Long only, long-term horizon, research analyst, macro
Dwaine van Vuuren
Long only, long-term horizon, research analyst, macro
Contributor since: 2012
Company: RecessionAlert.com
Sorry..my first sentence should read : "Piger has real-time vintages only *from* August 2006".
Piger has real-time vintages only until August 2006, so the assertion that your MIN/MAX analysis runs all the way until 1960 and offers a sound inference is quite flawed. Prior to Aug 2006 you are showing the MAX of revised data since Aug 2006. There will be huge maxima posted prior to Aug 2006 not captured by your analysis. Also you make comment that the series is subject to significant revision due to the revisions of the underlying data which is incorrect. Although the revisions to underlying data do of course impact the analysis the bulk of the revisions to this index come from the model inferences as discussed here : http://bit.ly/QqXSj0 and here http://bit.ly/TDmHIc
Chauvet is a she, not a he. Her model and Pigers' are identical in mechanics except they use seperate data for employment. The main point is you cant use their models for "never before" type analysis as the past history on the charts is subject to significant revisions (due to the Markov estimation process). You have to use rules like >50% for 3 months. The model should drop to around 10% next month, but also watch how the current months 18% will also plunge next next month. To use a model that deploys the same 4 NBER indicators in "never before" type analysis, go download the sample probability report from http://bit.ly/ZIUqkm
The 100% probability of recession interpretation of this model is grossly incorrect and debunked over here http://bit.ly/QqXSj0
People are not interpreting this as the authors originally intended.
Thanks for putting this to bed Marcelle. I suggest Jeremy and yourself get the FED to publish both smoothed and unsmoothed to avoid the gross misinterpretation the inuagural listing of this indicator had on FRED. Thousands of people were scared into hiding because of misinterpration of the data. Tha chart should cleary state SMOOTHED PROBABILITIES to make the distinction. These things may seem obvious to us who work with these models all day but not the people viewing them for the first time.
The public figures Jeff uses is 1 month delayed to the figures presented to paying clients. Although we present probabilities of recession we use specific thresholds of the underlying indexes we compile to make hard recession calls. The probabilities are academic.
An astute observation OilFinder, one that escaped me. Thanks for pointing it out.
I think that is the point I am making- although there is no showing of immediate recession, the tepidness of whatever expansion there is leaves us vulnerbale to all manner of external risks. Anyone not a little bit concerned about the future would have their heads in the sand.
The Shadow WLI project is just a fun project to predict in advance what ECRI will publish in 4 days time.Because so many people follow the WLI. It also includes 4 growth metrics and a dynamic factor probability model upon which you can make hard recession calls (when probability rises above 50%). It is, from what my research can determine, the best use of the ECRI WLI assuming you are using it on its own. It is imperfect and subject to false positivess (but our methodology of determining probability reduces the false positives somewhat.) The article is quite clear that relying on this single model to forecast recession is dangerous. For market timing, we prefer to use the Recession Forecasting Ensemble (RFE) which is described on our site.
Yes, we track some other more accurate growth variants of the WLI which Doug Short displays every week in his WLI update. Go check it out.
nice article james. Which p,q and d parameters did you eventualy end up with and was there any seasonality in your model?
You are quite correct on "US growth seems resilient", but in fact this is not a "recent" thing, the US economy, despite a slowdown is still far from recessionary levels as far as I am concerned. (Ive posted about curent recession probability on my instablog)
I sure hope nobody shorted the markets based on this info. A recession is not even close at this time. I have posted on this in my instablog.
The ADS has a logistic probability model NBER recession correlation of 0.87 versus 0.66 for the ECRIsince 1960. As such it is a far better predictor of recessions with half the false positives of the ECRI