05 Dec 2016

Questioning Scott Sumner’s Numeracy

Scott Sumner 21 Comments

Don’t worry, this isn’t (directly) about Trump.

In a blog post titled, “Innumeracy drives me nuts,” Scott Sumner lists several examples of people (allegedly) being innumerate. I definitely agree with some of his examples, but others are stretching it.

However, one of Scott’s arguments in particular seemed flat out wrong to me. Here’s what Scott wrote:

The claim that greater infrastructure spending would significantly boost US economic growth is absurd. It might boost it, but the US economy is far too large and diverse for a $550 billion infrastructure package to make much difference, especially during a period of 4.6% unemployment and monetary offset. Tax reform and deregulation are more promising, but even here the claims of 4% to 6% RGDP growth are ridiculous, at least over an extended period of time (I suppose one or two quarters are possible.) Trend RGDP growth during the 20th century averaged about 3%, under wildly different policy environments. I’m not saying policy had no impact (I’m a moderate supply-sider), but people tend to overrate the impact.

Now this is a strange argument indeed. Imagine if someone said, “It’s ridiculous to suppose you could sample Americans and get four guys in a row who are over 6’5″, because the average height over the whole population–including people with wildly different genes and diets–is about 5’10”.” The fallacy–dare I say innumeracy?–there would be obvious, and yet that seems to be what Scott’s argument is.

In any event, here’s a chart on annual percentage increases in real GDP, using quarterly data, from FRED:

As you can see just eyeballing the chart, Scott’s apparent claim that 4% to 6% real GDP growth (for longer than a quarter or two) is “ridiculous” is simply wrong.

Indeed, I crunched the numbers and, assuming no Excel mistakes, I found that from 1950 through 1979, the arithmetic average of year/year growth rates in real GDP was 4.0%. If we just look at 1960-1969, the figure rises to 4.5%.

In context, Scott is talking about people who are saying tax reform and deregulation could deliver sustained 4% – 6% real GDP growth. (I found John Taylor saying it here, for example.) I would think that surely, if a new Trump Administration could deliver 4% or higher for four years straight, then that would satisfy these claims. Yet Scott seems to be arguing that history shows us such hoping is ridiculous.

Well, assuming I didn’t make an Excel mistake, if you take the period 1948 – 2016 (3q), and calculate all the arithmetic averages of the annual growth rates in real GDP looking forward 4-years, then 29 percent of the time, that number will be at least 4%.

In other words, if you randomly pick a quarter from the postwar period (as far back as the standard BEA real GDP series goes), there is a 29 percent probability that from that point forward for 4 years, annual GDP growth will exceed 4%.

(For purists: You have to be careful with averaging the individual growth rates etc. [This is why reports of the “average performance” in a stock fund can sometimes be grossly misleading.] So I downloaded the levels of real GDP, and calculated the compounded annualized growth rate from 1950 to 1970: it was 4.2%.)

Now to be sure, we can quibble with the legitimacy of the BEA’s numbers. For example, maybe they are driven by the Korean and Vietnam Wars, and in that respect are bogus. But Scott didn’t seem to be throwing out the “official” numbers, he seemed to be saying the official numbers showed that sustained 4% – 6% growth had never happened in the US.

Yet on the contrary, it did happen, and during a time with very high marginal income tax rates. So I agree with John Taylor that a smart package of economic policies would allow the US economy to grow in this range for several years. (I’m talking in general. As readers know, I have long been warning of a coming crash due to the Fed inflating an asset bubble.)

21 Responses to “Questioning Scott Sumner’s Numeracy”

  1. Transformer says:

    If I were Scott’s lawyer I might say in his defense:

    1.Your data neither proves nor disproves the claim ” Trend RGDP growth during the 20th century averaged about 3%, under wildly different policy environments” as your data set only starts in 1950.

    2. The statement “the claims of 4% to 6% RGDP growth are ridiculous, ” is not based on a random sampling of RGDP from the 20th Century but on the reality as he see it now. RGDP growth appears way lower since about 1990, and I’m guessing there will be few if any period of <4% growth since then.

  2. Andrew_FL says:

    Gee, do you think it might have something to do with the fact that the YoY growth rate of the 16-64 population average .7% in the forty quarters from 2005 Q4 to 2015 Q1 whereas in the previous 197 quarters it average literally double that?

  3. David R. Henderson says:

    I think Scott overstated–it’s hardly innumerate to think we could have 4% growth for a year. But I think you should exclude periods when we’re not recovering from a recession. I don’t know what the numbers look like, but I guarantee that the average for the non-recovery periods will be lower.

    • Andrew_FL says:

      That’s a good point but “non-recovery periods” is not easily defined, either. For what it’s worth if I average the compounded annual QoQ rates of change of RGDP only for quarters where unemployment was below the value of FRED series NROU (which is basically the “NAIRU”) it comes out to about 3.6% for the entire period since 1949. Again, most of that is with much faster growing working age population than we have today.

    • baconbacon says:

      But many of those periods will also include the recession, Scott also thinks that the recovery to the 2008 recession has been to slow and that there is ‘room’ (not his word) to push NGDP above trend to catch back up.

  4. Tel says:

    So if I understand this right, Scott is basing his estimate on AVERAGE RGDP growth but Trump doesn’t care about any long term average, he just needs perhaps two years of good growth, maybe 2018 and 2019 and then Trump can walk into the 2020 election saying, “Believe me, I’m really good at this. Bigly”.

    After re-election in 2020 he just does whatever he feels like for the next 4 years (much like Obama did in his second term) and if growth returns to the long term average then so be it. Speaking of long term average though, if I blur my eyes and stand back from that chart it does look like overall the growth is getting less volatile but also trending down towards zero. This suggests that at least in theory, some policy change might turn that trend around… although such a game changer would be a major development, unlikely but not impossible.

    Is that and error related to “innumeracy” ? Seems to me more like insufficient cynicism when predicting the motives of a public figure (a common problem, although you would think Sumner would not have difficulty being cynical over Trump).

  5. skylien says:

    “Don’t worry, this isn’t (directly) about Trump.”

    I am really disappointed!

Leave a Reply