The U.S. Big Deal

The advance estimate (rough draft) of real GDP growth for 2020 I came out this morning: the annualized growth rate was -4.8%.

*************************************

OK. Time for some time series econometrics, following the videos on my YouTube playlist for ECON 3020.

I had estimated with logged quarterly real GDP data an AR(1) in differences, which is equivalent to an ARI(1,1) in levels: basically, Case 5 from the Handbook. Here's the rerun with the new piece of data:

(We should not expect a time series model to change too much when we add one new observation. If it does, this is usually indicative of a structural break, and more time fleshing that out in the software. But this one is fairly consistent with the earlier ones).

The important thing to flesh out of this is the size of the residual for 2020:1. This is our estimate of the shock to the economy (remember that the announced growth rate is a composite of the fitted value (capturing the persistent presence of past shocks) plus the residual (the new shock). Also, don't forget that the actual data is not annualized. The estimate of the residual is -1.9%.

Note that this annualizes to -9.3%. In conjunction with the announced growth of -4.8%, this is telling us that we benefited from the persistence of previous positive shocks that were swamped by the new one.

That shock is big, but not enormous: the caveat here is that most of the effect of COVID-19 on the U.S. economy occurred in the last 3 of the 13 weeks in the quarter.

When are the other big shocks? The biggest one is 1958:1 at -2.7%. That was the during a severe recession (the shortness of the 1957-8 recession is what keeps it from being ranked up there with the big 3 — 1973-5, 1981-2, and 2007-9 — since World War II). The other shocks that were bigger than the recent one are (in order from second worst to not quite as horrible) are 1980:2, 2008:4, 1981:4, 1960:4, 1981:2, and 1949:1. Every one of those is the worst part of a particular recession (1981-2 was a double dip), and most recessions don't have a shock as bad as any of these.

Some of those who did the time series portion of the final exam were surprised at the low AR coefficient, indicating that quarterly shocks are not that persistent. Even so, size matters: any persistence of a big shock is going to matter for a while: since the 2009 trough, the current shock is about 6 times the typical size. So, the hangover/aftershock from the persistence of the shock in quarter II will still be -0.7% (that's the -1.9% times the 0.36 coefficient). We haven't had a shock that big in 6 years: so the aftershock will be bigger than all shocks in the last several years ... and we should expect another big negative shock this quarter to be added on top of that. In 2020 III the aftershock from the current one will still be comparable to the most shocks of the last 5 years, and it will be mixed up with the aftershock from 2020 II, and any new shock in III (and who knows how COVID-19 will be doing in late summer???).