General Economic Policy

Groundhog Day Economic Forecasts

Happy Groundhog Day IIEP Blog Readers!

This morning I recorded an interview with Marketplace Morning Report, to be aired on February 2nd in honor of Groundhog Day.  The segment was about economic forecasting.  Originally the woman I was corresponding with to set up the interview wanted me to talk about my own forecasts.  Although I do prepare some forecasts for the Survey of Professional Forecasters, I regularly admit that I generally don’t trust economic forecasts, particularly my own.  Her response to my hesitation in sharing my forecasts with 5.9 million listeners (according to her count), was “as long as you’re a little more accurate than Pux Phil, I’m sure you’ll be fine.”   Sadly, I’m concerned that when it counts the most, economic forecasters may not be any more accurate than a groundhog.

You might think, hope even, that economic forecasters should be more reliable than a groundhog.  When it comes to recessions however, even the best forecasters out there seem to fail miserably to predict both when we’re going into one and when we’re coming out of one.  In fact, in an article I wrote with my colleagues Fred Joutz and H.O. Stekler, we found that although the Federal Reserve staff, quite possibly the best forecasters for the US economy overall, know whether or not we are in a recession currently, they cannot predict whether or not we will be in one just one quarter in advance.  This is a big problem since monetary policy actions tend to take at least a couple of quarters to take effect.

Researchers continue to search for ways to improve models that forecast recessions.  One of my favorite models is the one by Marcelle Chauvet and Jeremy Piger which currently suggests only a 2.7% probability of a recession.  But, even this model is just trying to catch a recession as it is going on, not predict it in advance.  Until researchers find a model that better predicts recessions, perhaps we should not expect so much from our forecasters and policymakers.  Maybe recessions are just unpredictable.  Or maybe we just need to find the right groundhog…


Tara Sinclair earned a B.A. in Foreign Languages from Wheaton College in Wheaton, Illinois, and an M.A. and Ph.D. in Economics from Washington University in St. Louis. Before starting graduate school, she worked as a commercial real estate appraiser in Chicago. She was also a visiting scholar at the St. Louis Federal Reserve.


  • Steve, this is a debate that my co-author Herman Stekler and I regularly have. In my view I put an almost zero probability on finding a model that will effectively predict recessions before they start. We may get some improvements to predict (and therefore hopefully prevent) some types of recessions, but then we’ll just get other types of recessions. I think the business cycle is here to stay. Herman, on the other hand, thinks that we could and should be a lot better at forecasting recessions. He’s willing to overpredict recessions – we often tease him of forecasting 15 of the last 10 recessions or something like that. Current approaches greatly penalize false positives, but he has a point that perhaps policymakers should be more concerned with false negatives than false positives. It’s an ongoing debate!

    • Has anyone done a study of forecasting accuracy over the past 15-20 years compared to accuracy during the 60s and 70s? I just wonder if there is any way to measure some improvement in some way. Probably it varies over time horizons, for example 6 month vs. 1 year vs. 5 year. Has there been improvement over any of these horizons?

      • Steve, a number of people have done studies evaluating forecast accuracy over time for the Fed and other forecasters, but the results are contradictory and in general there doesn’t seem to be any definitive evidence of improvement over time. My co-author Herman Stekler has a paper which discusses some of this literature (see pages 2 and 3 here: There may have been some improvement over short horizons, but long horizons are incredibly hard and most evaluations conclude that beyond the third quarter or so most macroeconomic forecasts have little value.

  • I wonder what your thoughts are on the likelihood that economic research will soon find, or even eventually find, a model that better predicts recessions. Over the past few decades, our data collection, computational capacity and econometric techniques have advanced phenomenally. Has our ability to predict recessions advanced to the same degree? If not, do you think there is a chance that we will turn a corner soon and suddenly discover that “model” that can be relied upon? Or do you think recessions are likely to remain unpredictable? What subjective probabilities would you assign to each of these distinct outcomes?

Leave a Comment

Time limit is exhausted. Please reload CAPTCHA.