Recession Probabilities for the United States

Jeremy Piger is an Assistant Professor of Economics at the University of Oregon. He has put together a model that presents recession probabilities for the United States "obtained from a dynamic-factor markov-switching model applied to four monthly coincident variables: non-farm payroll employment, the index of industrial production, real personal income (excluding transfer payments), and real manufacturing & trade sales.

It is a simple but robust model with a good track record of providing a modest advanced warning of recessions. Its updated quarterly.

Here is the prior Recession Probabilities for the United States. Note that this model gave a short warning prior to actual recessions beginning:

>

Historical U.S. Recession Probabilities June 1967 – February 2008
Historical_us_recession_probabiliti

Graph courtesy of Jeremy Piger

 

>

Note: The main weakness in the model is the reliance on government data (which tends to be revised downwards eventually. See this chart of NFP vs B/D as an example.

Print Friendly, PDF & Email

What's been said:

Discussions found on the web:
  1. Marcus Aurelius commented on May 6

    They’ll need to go past 100% probability for this one. Kinda’ like the amps in Spinal Tap.

    These go to 11.

  2. Pool Shark commented on May 6

    OT, but can anyone explain to me how Fannie can post losses far greater than even the most pessimistic forecast, and yet have their stock up 4% on the news???

    Talk about ‘disconnects’…

  3. Banker commented on May 6

    Recession is just a technical term as we all know, so we may not be technically in a recession, but to me we have been for a few months.

    Banker

  4. Jim Haygood commented on May 6

    Now THAT’S what I was talkin’ about in the previous post about defining a factor model for recession. The DEFINITION of “recession” should be based on Piger’s model, or something similar to it.

    For example:

    Model reading > 20% = in recession
    Model reading < 20% = in recovery What I am saying, to be perfectly blunt about it, is that the useless wise men of the NBER could be replaced by Piger's computer. At most, the NBER should be in charge of maintaining and updating a published quantitative recession model -- not holding their fingers up to the wind and announcing authoritatively (and long after the fact), "It was a recession!"

  5. michael schumacher commented on May 6

    Probably means that FNM just got put at the head of the line for the next acronym creation program that,again, let’s us all know it’s just fine.

    and FNM goes over 30//

    can’t make this shit up…….

    Ciao
    MS

  6. Myr commented on May 6

    It would make much more sense to show the graph using the initial unrevised data. Otherwise, we’re comparing apples and oranges.

  7. John Borchers commented on May 6

    Of course current is a recession and it’s going to be a bad one.

    Especially since the Fed is back stopping the market on every make believe news statement.

    This is going to bite everyone in the market. They are teaching people that bad is good and like Pavlov’s dog they are being trained.

    Unfortunately the people that are going to lose in the market are the ones that need the money the most.

    I guess Yahoo can get those treasuries now?

  8. worow commented on May 6

    Is it possible to estimate what the recession probability would be if the gov wasn’t fiddling with the B/D adjustment? How sensitive is this model to unemployment?

  9. John commented on May 6

    Piger’s model uses “four monthly coincident variables: non-farm payroll employment, the index of industrial production, real personal income (excluding transfer payments), and real manufacturing & trade sales”.

    From the NBER’s web site (http://www.nber.org/cycles/recessions.html): “The committee places particular emphasis on two monthly measures of activity across the entire economy: (1) personal income less transfer payments, in real terms and (2) employment. In addition, the committee refers to two indicators with coverage primarily of manufacturing and goods: (3) industrial production and (4) the volume of sales of the manufacturing and wholesale-retail sectors adjusted for price changes.”

    Are these not the same? Why the vitriol directed at the NBER?

    And why is it important to know in real time whether or not the country is in recession?

  10. Jim Haygood commented on May 6

    One reason Piger’s chart starts in 1967 is that most economic data series showed a sharp contraction in late 1966. But allegedly due to pressure from President Lyndon Johnson, the GNP was jiggered upward, and no recession was declared by the NBER.

    I’m guessing that Piger’s model CORRECTLY indicated a recession. But since everyone’s in thrall to the NBER’s wise men who declined to annoint the event as a “recession,” conventional thinkers would call it a “miss” by the model. I wouldn’t.

  11. Jim Haygood commented on May 6

    “And why is it important to know in real time whether or not the country is in recession?”

    Are you kidding? Why is any kind of timely data important? Because business decisions, government policy, and consumer behavior are based on incoming data.

    The NBER does its job badly, because its recession determinations would be sounder if they were quantified, and more timely to boot. Including a discretinary factor into its recession determinations does not add value; it subtracts it.

  12. Richard commented on May 6

    So the Fed waits until after the next administration takes office, then raises rates ( to finally stamp out inflation) and we get the recession we should have had 18 months prior. By then it will be too late and it will take a global recession to kill the beast.

  13. AGG commented on May 6

    Government Data. Why don’t we just abreviate that to G D?

  14. Anderson commented on May 6

    Per Barry’s comment on the model’s weakness: I read through Piger’s paper and his model’s inputs are based on the original government statistics. Therefore, it does not reflect adjustments made after the initial data release.

  15. michael schumacher commented on May 6

    at least five times today the “buyers” just show up and buy everything in sight regardless of price.

    No sane buyer long on anything buys that way…..ever

    The perma bulls will also be emboldened by data that present the “worst is over-scenario” that’s been spouted as the catalyst for “second half growth”….

    I guess data that doesn’t get worse is great if you can’t see the forest through the trees.

    Ciao
    MS

  16. MarkTX commented on May 6

    Todays action in the market and in the past few weeks reminds of that “magical” 2006 last quarter run-up when HP took over at the Treasury. Alchemy at its best.

    Since most of the Market is in control of the few wealthy,
    I guess the new thinking is that recession is only for the little people….

    $2 beer night here I come :)

  17. Pool Shark commented on May 6

    FNM up nearly 9% on the day???!!!

    Temporary Permanent suspension of disbelief…

  18. Our man in Helsinki commented on May 6

    One interesting point in this model is that there seems to be only one direction for the economy to go after the recession propability has reached 10% or so.
    Maybe there’s a bias towards low propabilities – even during a recession the model can sometimes give a modest recession propability.

  19. John commented on May 6

    Jim Haygood,

    You are absolutely right that “business decisions, government policy, and consumer behavior are based on incoming data”. The DATA is what is important, not the determination of a recession’s start and end dates. In other words, “business decisions, government policy, and consumer behavior” aren’t different because growth is -0.2% instead of +0.2%.

  20. Josh commented on May 7

    Note: The main weakness in the model is the reliance on government data (which tends to be revised downwards eventually. See this chart of NFP vs B/D as an example.

    Barry, as Anderson noted, Chauvet and Piger (2008) use real-time data. The model does a good job identifying recessions, which the NBER declares after the fact, in real-time – using real-time data. The model doesn’t benefit from any data revisions.

  21. gianluca carrera commented on May 7

    very interesting… I will use it on my blog too, thanks. your blog is amazing!

Posted Under