Sunday, April 22, 2012
Terminal Velocity? Or Why the Economics Isn't Physics
Recent posts from Evan Soltas on Impulse Functions and Growth Inertia has led to some interesting discussions on whether there are certain consistent correlations that could predict behavior in economic data, much in the way the movement of physical objects could be predicted.
On first pass, this should be a skeptical proposition. Even a weak form EMH would mean that there should be no systematic way to predict the movement of GDP on a quarter to quarter basis. I'm no expert in econometrics, but this core tenet about efficient markets is cause for suspicion.
After looking at the data, it's not clear that there's any meaningful thresholds either. While Evan focused more on logistic growth models or quantitative ways to model the entry into recession, I tried to look at more of what kind of growth can be seen before recessions. I looked at percent NGDP growth from one year ago, percent NGDP growth from one quarter ago, as well as the RGDP equivalents, and tested a variety of traits about those numbers, but found nothing. Among them, the NGDP growth from one year ago seemed the least persnickety with smaller standard deviations relative to the mean, so I use it for this post.
I used the NBER recession classification table to classify all the quarters with a dummy variable, with 1 indicating a recession and 0 indicating regular times. As the interest was in predicting entry into a recession, I focused on the quarters before the start of a recession, and ignored the months within the recession. I looked at the rates one, two, three, four quarters before the recession, the average growth over the four quarters before the recessions, and many others that are listed in the chart below
The last two rows are the average and standard deviation in that column. The four quarter/six quarter OLS measures are kind of like a second derivative: I was looking at the rate change in the "percent change from one year ago" number over the past 4 or 6 quarters. Nothing. Even if some of these confidence intervals are "non-zero", they're not very helpful as they would predict a recession in almost any quarter. Diagnostics aren't very helpful if they keep on returning false positives. Besides, shouldn't these be regime dependent anyways? It really depends on how far governments are willing to let economies go. Perhaps there's stronger indications for other parameters, such as private debt or real wages. But so far, it doesn't look like it's in an economy's growth speed.
P.S. If you want the excel spreadsheet I used, I'd be happy to email it to you. Google docs doesn't do xlsx very well.