The 2% per capita growth rule

Common adage: that the U.S. grows at 2% per capita, from the year 1870 to today. The economy rarely deviates from that trend growth line by more than 4% (a standard business cycle magnitude). As such, gravity pulls us back to a 2% trend growth line, remarkably.

My suspicion: this is revisionist history. We focus on the start and end of this graph and forget how big the deviations are in the middle, since “the economy went back to trend”. And then we find a bunch of reasons to justify why those outliers are large (wars! immigration! participation!), and are comforted by the fact that they disappeared.

But do they disappear by construction? What if we had instead taken the growth rule that we would have come up with in 1970, 1990, and projected out growth, what would we have forecast the size of the U.S. economy to be today?

I take the U.S. per capita series from 1947 to today (where it begins from the BEA), and then run a linear regression on log gdppc. I do that for the period ending in 1970, 1980, 1990, 2000, 2007 (to avoid the crisis) and see what the projection is.

The usual fact

This method gets you the standard result, that the economy grew at 2% per capita from 1947 to 2007, and projecting that growth level gets you essentially to where the economy was before the crisis.

The crisis puzzle

And the puzzle: that the crisis ruined it. We never returned to this trend growth path after 2007. Even the (relatively speaking) boom of the last 5 years of growth has taken us further from the trend. (A lot of the boom in growth recently must have been population).

But is it?

My contention: that this 2% rule is convenient and we invented it sometime around 1990. The only interesting fact is that there was 15 years of stable growth of around 2%. But we never had this “return to x%” rule before. What if we had stopped the model in 1980, and projected growth out?

The rule was never that robust. If you had projected out growth from 1980 to today, you would have expected the economy would be a quarter larger than it is.

Even BEFORE the crisis, you would have had this issue. Looking at the period 1990 to 2007, you always had a gap from trend of at least -5%, and usually around -8%. By 2007k the gap is already double digits. If you had projected from 1980 you would have had the same puzzle that we’ve had over the last decade for 30 years.

True for any decade.

Basicall any out-of-sample forecast gets you to a materially different endpoint for 2020 gdp.