Thursday, October 15, 2009
Like Scott Beaulier, I have been asked whether the latest NABE survey that 80% of its respondents think the recession is over is true. And like Beaulier, my first answer is that "it's a survey" which is asking in effect the question "when the NBER business cycle dating committee dates the trough of the Great Recession, what date do you think they'll put on it." Since the survey was done in September, the respondents are saying "we think they'll set it sometime in July or August" (I doubt anyone would say it ended in the second quarter.) We looked at that data last week and you'll know I concluded that "we need a certain move in GDP AND in employment" before we should put a date on it. So I'm in the 20% on that one, but maybe I'm just a lagging indicator :) As I noted, the second derivative issue (or, if you will, the inflection point issue) is a real problem for any forecaster, as Beaulier acknowledges in his third point. We agree there.
What I would resist, however, is some of the lampooning of forecasters that Beaulier engages in:
...many of the "economists" they're surveying are business economists applying flawed forecasting models. Generally speaking, they are not academic economists, who have a stronger sense of unintended consequences and theoretical nuance.And
I view this kind of economics as hackonomics at its worst, and I don't put much stock in the predictions they make.The job of a forecaster is to forecast, to try to get the next number right. Whether I have a "stronger sense of unintended consequences and theoretical nuance" is useful insofar as I get closer to the right number. Beyond that it's only for smug faculty club conversation. I live in a fairly small city; my life gets spent in both the academic world and in Central Minnesota's business community. (I'm not the only forecaster who works in academia, of course.) I educate while I forecast and some "theoretical nuance" hopefully comes out of that. But because I also meet and sometimes share panels with "business economists", I have a view of them that would consider this "hackonomics" smear to just that. It's damn unbecoming.
If the horseshoe works, I use it. I'd like a model that predicted better that didn't use the horseshoe. It'd be swell; I could be confident in both the faculty lounge and the rubber chicken business luncheons I speak at. But there is no tradeoff between theoretical purity and forecast accuracy. Whatever it takes to get an accurate forecast, I'll use.
(BTW, I don't make national forecasts, at least not publicly. I did it as a training exercise for graduate students when I taught forecasting for them. But I retired that model and turned that class over to my younger colleagues a few years ago. So I don't participate any more in any of the national surveys. I don't use any national indicators in forecasting St. Cloud. And I don't have the luxury of "following the herd" because there is no herd forecasting Central MN.)