Average is Over: GDP Edition

The United States hasn’t had a year of above-average growth since 2005.

Addendum: With a smoothed series we haven’t had an above average year of growth in the entire 21st century.


That average is boosted by all the anomalous years after WWII when the economies of Europe and Japan were recovering from heavy damage. Average isn't average.

It's cute how they went back to WW2 and stopped. God forbid they included the Depression years into the "average" or else that would have destroyed their headline, and Alex would have had to post something else today, like great open borders are.

1948 start is probably dictated by the start of robust data collection (a consequence of the need for military strategy and planning -- can we make more tanks than USSR?).

If you include the great depression then might as well include 1920s, and the whole industrial era. The post 2000s stagnation point still stands. We've run out of new ideas for energy use (=GDP) now that the WW1, WW2 and the Cold War ideas (planes, cars, nuclear, internet) are fully deployed.

It wouldn't be a problem if not for the exponential growth of debt contracts whose repayment is predicated on the old higher GDP growth.

Some form of non-linear resolution (war, deflation default, hyperinflation) is likely within our lifetimes.

Nuclear is fully deployed? Solar?

Well seeing that world nuclear consumption peaked in 2006 (at 634m tons of oil equiv, vs 611m in 2018) I'd say yes, empirically. Perhaps the full extent of long-term costs and pricing of tail risks became apparent.

We know from Trump and Bolton and neocons, all nuclear is totally to start a nuclear war and wipe out civilization.

Plus, paying workers always kills jobs.

Eg, Terrapower (leftist Bill Gates green energy project) was forced to stop work building a nuclear power reactor that would produce near zero nuclear waste because the only reason China would pay for Terrapower to built it is to destroy the US.

Trump and Perry claim they are doing more than Obama by increasing funding up to less than 90% of the Obama nuclear power innovation budget, and providing a tiny fraction of the funding the Obama administration did to fund building nuclear reactors using the for profit technology.

Terrapower is effectively non-profit.

Government spending did not increase to produce the data until Democrats took control of Congress in the 30s (and Hoover was a data driven man, as was FDR). The BEA started a series in 1930 by collecting the data, but it wasn't published as an official report. Likely in part due to war secrecy. When the Truman administration and Congress funded publishing the data on an ongoing basis, economists of Milton Friedman's generation refined the data collection and reporting (he wrote about his work collecting and analyzing economic data in the 30s).

This series has been collected and reported consistently since 1947, ie starting with 1946. Job data started a year later. They have been refined at several times, with adjustments made to prior data a few times. But since the 60s or so, "flawed" data collection stays flawed, with new data series created.

The series starting in 1930 is still reported, but it's been replaced by the less flawed series starting in 1947.

See: https://fred.stlouisfed.org/graph/?g=oTdt

A new less less flawed series was created in 2003.

As an engineer (in quality), I learned that you always want your past real time data collection to have been done differently. We engineers do cost-benefit and increase costs to obtain better understanding.

Economists seem to mostly throw up their hands because they see no benefits from anything that increases costs, so ignorance is basically a benefit for most economists. The answers to most economic questions will be rather costly, although the benefits will almost certainly be worth far more, but only if the data is given away for free.

Physicists and geologists spend billions going back in time to create data series spanning millions or billions of years billions of years in the past. But that what creates jobs for scientists, technicians, engineers.

Economists seem not to want more job opportunity, otherwise they would demand lots more funding for data collection, going back in time hundreds of years.

When I was little I sometimes watched a "charting the market" show on UHF. Mainly because I thought the guy was so weird and convinced.

But channeling that guy, the "falling tops line" does start falling right away, from the beginning of the chart.

Anyway, we do need more Tabarrok style growth policies, recognizing also that pollution is real and harmful, and Trump style dirtier air and water aren't the way to get there.

If you don't know what a falling tops line is, note that except for the ~1984 flier no peak exceeds a previous peak.

That's pretty striking.

I used to watch that show too. I thought it was a local (Southern California) show and that must have been in the 80’s?

I think so.

Indeed, this is just away of saying the US economy doesn't grow as fast as it did pre-2000.

And that's not really unexpected, maturing economies have growth plateaus. It's already started with China even.

Throw in declining population growth, and of course GDP will plateau. GDP per capita is more important anyway.

What is a "maturing economy"? It sounds like a tautology.

I suppose it might be a little tautological, but another way to say it is that it's common sense. A poor country will grow faster than a rich one, catch up growth/low hanging fruit etc. The richer countries will struggle to grow as fast as they did before. Hence, they are 'mature'.

Working age population clearly leveling off:


Is that GDP per capita?

That was the first question that came to my mind, what has the GDP per working age adult growth been?

U.S. GDP per person employed (OECD):

1970s 1.2%
1980s 1.4%
1990s 2.0%
2000s 1.5%
2010s 1.0%

GDP per capita:

1970s 2.5%
1980s 2.1%
1990s 2.0%
2000s 1.4%
2010s 1.0%

Average from 1970 to 2018 = 1.8%
(Alex starts at 1948 so many more post war higher growth years) With respect to per capita growth from 1970, 2015 and 2018 beat the average.


+1, informative

It's unlikely to be per capita when the average shown is more than 3%. Productivity increases around 1% per year if you take the mean of many decades.

"It's unlikely to be per capita": given that GDP itself is a pretty ropey measure of things you might really like to know, and given (if given it be) that no connection is made to demography, then it's likely to be a rather useless graph.

Y-axis: Real GDP Growth Rate (%)

Looks clear to me.

Clear? It hasn't even got the units right. It presumably means "% p.a.".

'hasn’t had a year of above-average growth since 2005'

Sure it has - look at the growth of the federal deficit.

Of course some of that deficit leaves the US borders, and some ends up as savings. So those could answer for some of the discrepancy.

what insight is there to discover starting the graph all the way back at 1948? I can start the graph at literally any date to say whatever I want to say. This is like the "Man Made AGW" school of statistical analysis.

Who cares if we are having individual years of below-average growth?

This is the longest (or 2nd longest) expansion in the NBER's record. Slow and steady growth is probably better for the economy than boom-bust cycles - years of rapid growth followed by recessions.

The chart could use a 5yr moving average.

Is that your chart, Alex? Can you add a 5 and a 10 yr?

I'm looking at the 5-year average chart. Average growth over the past 72 years has been 3.1%. The 5-year line clearly trends downward, and it underscores how much worse the recession that began in 2008 was compared to any previous post-WWII period, with 5-year growth falling below 1% in 2009 for the first time post WWII and remaining there until 2013.

The smoothed chart shows a downtrend as well, but I suppose one that could be part of a random walk. In that scenario the 1960s mountain was as random as the 2000s valley. The human vision system looks for patterns, even when they aren't there, etc.

Still, I am on board for Alex's prescriptions for improvement. They certainly couldn't hurt.

" Slow and steady growth is probably better for the economy than boom-bust cycles - years of rapid growth followed by recessions."

Not according to the data.

10/10/-2 is still better than 5/5/5

Indeed, it's possible you need the recessions to remove the dead wood from the system so it can grow faster. AKA creative destruction.

Much of the New Deal and Great Society macroeconomic thinking was around the idea that "stabilizers" would sacrifice some growth to smooth out the ride. This appears to be what has happened, especially if you look at quarterly (rather than year-on-year) changes in GDP.

Years that saw a 5% annualized contraction in GDP on a quarterly basis since WWII:


Two declining trends in the chart. 1948 to 1980 and 1985 to 2020.

The first trend was temporarily solved with the Nixon Shock. The second trend will be solved by all the MIT economists who figured out 'This time is different' in 1970s.

In one podcast Prof. Cowen said stagnation since the early 1970s might be that we have run out of cheap uses for fossil fuels. This sounds unclear. An innovation that uses a fossil fuel input would experience rapid adoption and a higher output growth rate if the fuel is cheap. Perhaps he mean we haven't started more innovations requiring fossil fuels. In this view, GDP is very energy dependent. That would be a bit of an unexpected take on the situation because the story of recent decades is more GDP from things that do not require much fossil fuel: health care, education, computing, and internet.

The stagnation puzzle is interesting but perhaps we will all stop giving it much attention for the very reason economists always want to point out. That being that people want to maximize what they want to maximize and it is not exactly the market output of all the people within the country's borders. Economists want to point out there is stagnation and at the same time they want to point out that we are not sure if we need to be concerned.

"In one podcast Prof. Cowen said stagnation since the early 1970s might be that we have run out of cheap uses for fossil fuels. "

Except that there hasn't been stagnation from the early 1970s, only for most of this decade.

Whether there is stagnation depends upon the definition. The red line in the chart here implies stagnation is present if the data are below the red line. That's not the only way to demonstrate stagnation. In Tyler Cowen's book on stagnation you can read about a different statistical interpretation that suggests the phenomenon is much older than the past 10 years.

He was wrong there as well.

GDP from things that do not require much fossil fuel: health care
Hospitals are the third largest user of energy. Most of what government does requires energy that we cannot afford, which is why government does it.

Hospitals are not health care. Some of hospital electricity is hydroelectric, nuclear, etc. I doubt if you really want to be arguing that fossil fuel cost is a big contributing factor to hospital contribution to GDP. Granted without electricity hospitals would produce very little GDP but that doesn't mean the electricity company has much pricing power because this is a regulated industry. Doctors, nurses, unions, and pharma companies would seem to have more of an effect on the value of hospital output.

The numbers are biased by post World War II and 1930s economic depression and because it is not "per capita".

Come now, this is facetious.

The other thing I would not about this misleading graph is that all of the peaks are only one-two years and all follow after major slowdowns. With real world frictions, I suspect that a lot of these peaks are just the bump that comes from restarting the use of idle resources.

For instance, 1949 had a recession where unemployment reached 7.9%. In 1950 that had fallen all the way down to 4.3%. Back then wages were over 90% of GDP (today they are somewhere in the 80s). 3.6% of the 1950s GDP growth, a priori, is just the workers absent from 1949 going back into the labor force. Add in whatever value they add beyond their wages and we are pushing whatever endogenous growth that occurred in 1950 down to right around average.

Mathematically this sort of volatility further distorts the average. If the economy goes from 100 to 106 in two years equally it is an average of 3% growth. If it goes from 100 down to 97 then immediately up to 106 you have a year of -3% growth and one of 9.3%. This averages out to 3.15%. Using growth rates is always prone to distortions like this and we should not expect a highly volatile period to be informative of a non-volatile period.

Part of the reason why we have so few above average years recently is that we have had a total of one year of economic contraction in about 30. Certainly our politicians and central bankers have become much more active in combating GDP contraction; and there certainly was something anomalous about the 2008 recovery ... but I suspect a lot of this is like anything else, decreasing volatility limits downside risk but leads to less over all growth. Certainly a government backstop makes it that much harder to realign malinvestment (e.g. as seen in Japan or China).

To whit, of course the average was higher back when we had higher volatility, it mathematically has to be and lest we forget the volatility of markets has actual utility. Add in the demographic and time period posts and I am decidedly unimpressed.

Back then wages were over 90% of GDP (today they are somewhere in the 80s)

They weren't and aren't. Since 1929, total employee compensation has fluctuated between 60% and 70% of gross domestic product.

Sorry, it appears that the quick Google search failed me. Thank you for the correction. I believe the point still stands, when you have a recession a large amount of the "above average" growth is just resuming production with idle resources.

It seems normal for the GDP growth rate to slow down as the level of GDP gets higher. The countries with the fastest GDP growth rates are all very poor; their growth rates inevitably slow down as they get richer.

The U.S. and the U.K. have been at or near the lead of the technological frontier for over 200 years and their GDP per capita growth rates largely accelerated until fairly recently. China could easily see growth at 6% for much longer than assumed, that is, about over and will drop to 3% soon, due to much faster US and EU growth rates.

Pikkety and Gordon will be way off with respect to their near zero growth predictions for after 2070. Cowen will be way off with his 2009 prediction that stagnation will continue until the 2040s.

"It seems normal for the GDP growth rate to slow down as the level of GDP gets higher. "

Except that the US had the highest GDP from 1948 on. That entire graph represents both the US having both a high GDP and a high GDP per capita. So your explanation doesn't explain the question being asked.

The lost century.

This is great news! We're bound to return to trend sometime soon - I mean we can't possibly be so compentently incompetent that we'lll always be under trend.

Exactly, and Ivanka will be much better than her father.

Maybe central banks getting good at controlling inflation is the problem with recent times. Standard Austrian framework says that artificial low interest rates encourages too much investment, so likewise high real interest rates will cause under investment. With low inflation and the Zero bound, real interest rates are too high.

Not enough demand in the economy.

Wages have been flat or worse for the bottom 4/5 of Americans for a long time. And those are the ones who would spend.

Has it?
Data from BLS suggests this is not true:

The "real" wages appear to already account for CPI changes and it does not show flat or worse.

Not exactly flat, but according to the BLS, real wage growth over the last 10 years is 0.43% per year.


"Launching The Innovation Renaissance: A New Path to Bring Smart Ideas to Market Fast"

https://www.amazon.com/gp/product/B006C1HX24/ref=as_li_tf_tl">$3 at Amazon

Here's the formula for GDP: C + I + G + (X – M). X means exports and M means imports. Let's see, America began outsourcing intermediate goods in the 1970s but it accelerated in the 1980s and beyond. On the other hand, manufacturing output has been rising during this same period, dropping during the great recession but recovering since. Why is GDP flat but manufacturing output rising? Duh.

Isn't it becoming more average over time rather than less? Seems the variability is flattening, not increasing.

Service Sector vs Goods Producing Sector.

The service sector is more stable than the goods producing sector in terms of annual output.

As more of the workforce moves into services, we should expect to see less variability in output.



Rayward -- be careful, real GDP has not been flat, it is just growing at a slower rate. So is industrial production. Moreover,the composition of industrial production has changed significantly.

Manufacturing output in the US: https://fred.stlouisfed.org/series/OUTMS

"Flat" to describe GDP wasn't the best choice, stagnant is a better choice.

If you look at real per capita GDP growth you get quite different results.

From 1850 to 1950 the exponential trend growth for per capita growth was only1.6% while from 1950 to 2000 it was some 2.1%.
Even compared to the 2% long run trend the only periods when per capita real GDP growth exceeded the trend was from 1880 to 1910.

Much of the rapid growth from 1850 to 1940 was due largely to population growth rather than improved output per capita. Ending the comparison in 1925 to avoid the depression distorting the comparison does not make any difference. While from 1940 to 2005 it was constantly above the 2% long run trend.


Manufacturing output is always more volatile than services. Changes in demand for physical goods shifts up and down faster than changes in demand for services like education or health care.

As the economy shifts from manufacturing to services, we should expect to see more macroeconomic stability. No?

This guy gets it.

Three recessions under Eisenhower? Not a successful presidency.

Presidents don't affect the business cycle very much, so it's mainly luck if they will be dealing with a recession (Bush I) or two (Bush II) or three (Eisenhower). One reason people think Clinton was a good president is he had the luck to get in right after a recession and get out right before the next one.

Presidents get the blame and credit though.

+1, yes, it really is silly partisan politics. But it is what it is.

Since 1980, recessions have been a boundary condition for presidencies, each president get one and only one. Generally caused by the previous presidency. The mechanism is simple, one group pushes the expenses onto the next group. The politics alternatie repub and dem presidency, the cycle is generally in phase, coincident, has a public policy associated with it, and actually planned by the parties. Look at the pattern since 1980, the only exception is the Alzeimeirs president who got a recession right at the start.

The Swamp is both a cause and a synchronized. The usually mechanism is commodity prices jump and the economy stops. Congress the one fouling up demand for commodities.

Plain to see in the blue bars appearing at a regular pattern in our FRED charts.

No thermonuclear wars; It's a great record.

Way, way better than Truman's record.

Since 2000 the US has been conducting expensive non stop wars and has imported large numbers of low skilled third world migrants, what could go wrong?

I’m pretty sure the 40’s, 50’s and 60’s had their own wars which, although not non-stop, were plenty impactful.

Non-stop expensive wars?

Like the Vietnam War and the Korean War? https://fas.org/sgp/crs/natsec/RS22926.pdf

For the entire 1950 to 1970 period, US military spending never dropped below 8% of GDP.

For the entire period of 1991 to now, US military spending has never exceeded 5% of GDP.

Defense spending is higher than it was in the late 1990s but it's hardly high by historical standards. The War On Terror hasn't cost nearly as much as the Vietnam War.

Conceptually, isn't this the opposite of 'average is over'?

Variance is decreasing, not increasing! Average is very much here. It's just the average is a bit lower.

We got one or two growth spurts from the Nixon Shock, that is all for one entire generation. Pathetic, and they tell me we are getting stupider about this, not smarter.

What happened? Ignorant politicians, especially the small states, not enough brains to make two good senators. Hence we end up with cattle ranchers in Montana starting wars, and the smarter folks in California refusing to pay for them. Or visa verso, we blow a trillion on Obamacare, then decide to redo it ten years later.

About 8T of that debt was pure loss, starting with the Alzheimer's president bailing out the Bush kid in the Texas S?L. We rolled that over four times, and the interest charges way beyn anything useful from the bailout. Texas Repub still organize mass shootings! They haven;t learnt a thing and ow the Swamp about 2 trillion for that bailout.

Then we had two trillion blown on new bomber technoloy which never worked, and our maintstay remains the rusty B52. But worse, we spent another two trillion of F22 and F35s only to find them useless and defenseless in the age of drones.

No Child Left Behnd was another volatile disaster, cost California about 10 billion in excess pensions as we moved teachers around to meet federal standards.

Now we are broke, once again, with interest charges growing at 10% a year, unsustainable and likely we will have to do another 'sudden' Nixon shock to dump the excess debt accumulated.

The US hasn't had a year of average real per capita GDP growth since some time in 2017-2018.

Of course all growth since 1970 that exceeded averages was ephemeral, especially the 2002-2005 timeframe.

After reading every comment I have come to this conclusion: This article is a well written opinion, backed by well managed data, designed with the purpose to expose or define everyone else's opinion. It serves to remind everyone the importance of managing the data.

Comments for this post are closed