Scott Sumner  

The Great Recession, productivity and productivity growth

Henderson on NAFTA and Mercant... The Case Against Education<...

The sharp reduction in productivity growth since 2004 is one of the most notable recent trends in macroeconomics. Not surprisingly, some pundits have suggested that this slowdown is linked to the Great Recession.

In the most simple business cycle models, output returns to the natural rate once wages and prices have adjusted. In that case, demand shocks have only a transitory impact on output. But recessions don't just cause unemployment; they also reduce investment, which leads to a smaller capital stock. In that case, might the Great Recession permanently reduce productivity and output? Probably not, but the impact of a smaller capital stock might be so long lasting that it seems permanent. Even 8 years after the trough of the Great Recession, productivity and real output might be lower than if the Great Recession had never occurred.

Unfortunately, this doesn't really help us understand why recent productivity growth is so anemic. That's because the very same models that allow recessions to have a highly persistent effect on productivity, also predict exactly the opposite effect on productivity growth. Thus if our capital stock was artificially depressed by a low level of investment during the Great Recession, then the marginal product of new capital should increase, and the level of investment should also increase. Our economy is more labor intensive than if there had been no Great Recession, and this should lead to a higher than normal level of investment.

In other words, if the level of productivity was semi-permanently depressed by the Great Recession, then the current growth rate of productivity should be higher than normal. But we observe exactly the opposite---productivity growth is unusually low.

To summarize, we don't know why productivity growth is so low (I suspect it is related to the economy's shift to services, as well as hard to measure stuff on the internet) but it is not caused by the Great Recession. Indeed you'd expect growth to be faster than after a deep slump. And indeed growth was faster than normal after the deep slumps of 1920-21, 1929-33, 1937-38, 1974-75, and 1981-82. This is the only deep slump (in America) followed by slower than normal growth.

In my view, it's a waste of time trying to construct a business cycle model where a deep slump permanently reduces the economy's growth rate. That's because if you "succeed" in explaining 2008-09, then you immediately unexplain what we thought we new about the 5 earlier slumps mentioned above. Why would anyone wish to solve one problem at the cost of creating five brand new problems? Better to treat this recovery as an anomaly, and look for reasons why productivity growth began slowing after 2004.

Some people ask me if I really think it's a "coincidence" that long run trend growth started slowing at roughly the time of the Great Recession. Yes I do. I would have a much harder time accepting 5 improbable coincidences, the hypothesis that deep slumps will normally reduce GDP growth rates, but that this did not occur during the previous 5 deep slumps because each time some sort of mysterious exogenous factor came to the rescue. How likely is that?

PS. Elsewhere I've argued that the slowdown in productivity growth helped to cause the Great Recession. That's because it depressed the Wicksellian natural rate of interest to a level below where the Fed thought it was, and this caused the Fed to tighten policy more than they anticipated.

COMMENTS (17 to date)
E. Harding writes:

The only real parallels I can think of to the present post-Great-Recession global productivity slump is the post-early-1980s crisis productivity slump in Latin America (then again, Argentina boomed after its similar early 2000s crisis). Maybe post-1997 Japan, as well (though that was obviously country-specific). I think some kind of nominal-real-effect interaction has caused the present productivity slump by hindering development of new inventions (the timing is just too parallel), but it's hard to think up of a no-recession counterfactual everyone can agree on.

Cloud Yip writes:

So, does it mean that you are not a supporter of the Hystersis works pursuit by Blanchard and Summers et al. ?

Do you think that Hystersis is an ad hoc model rather than general model?

James Alexander writes:

As I asked at MI, if bad money policy can cause a recession, why not a depresssion? Stop taking about unemployment. Go back to talking about money demand aka anticipated nominal spending aka expected NGDP. Remember NGDPLT?

Todd Kreider writes:

Scott, you wrote:

"The sharp reduction in productivity growth since 2004 is one of the most notable recent trends in macroeconomics."

Is this labor productivity or total factor productivity?

I pointed out that when Tyler Cowen talks about productivity levels when interviewed for his book, they do not coincide at all with productivity. He actually has said different things in different interviews making this even more confusing.

1) There was only a spurt in the mid 1990s.

2) With Ezra Klein he said yesterday: "We've had the internet — and productivity rates have been stagnant or falling."

Neither of these is correct. The productivity from 1995 to 2010, 16 years of the internet rapid expanding in the U.S. and around the world from closer to 1999 or 2000 was 2.7%. After the US financial collapse, productivity has been very weak from 2011 to 2016 at 0.5%.

Cowen replied that I should look at TFP that he uses in his book, so I looked that up on p.89.

"But in 1973, TFP declined dramatically, often coming in below 1 percent a year, and for the most part, TFP growth stays low until 1995, with an average of about 0.5 percent....The mid 1990s to early 2000s were a new golden age for TPF, which again at times ran at 2 percent or more."

A FRED Total Factor Productivity graph instead shows a solid increase in TFP going back to 1983 and continues to 2010 when less TFP grow.

Getting back to your opening, labor productivity from 2004 to 2010 was 3.1, 2.1, 0.9, 1.6, 0.8, 3.1, 3.3, for an average of 2.3 which is solid growth.

If you meant Total Factor Productivity, the graph doesn't show a slowdown from 2004 but from around 1999 yet rises at that slower rate to the end of the graph in 2011 despite a dip around 2008.

Finally, you wrote:

"To summarize, we don't know why productivity growth is so low (I suspect it is related to the economy's shift to services, as well as hard to measure stuff on the internet) but it is not caused by the Great Recession"

Yet according to the World Bank, there was hardly any shift to services: In 2002, services accounted for 77.7% of GDP, there was a decline until 2006 when at 76.6%, followed by an uptick to 77.7% in 2009 followed by a decline to 78.0 by 2014. So a 0.3% increase from 2002.

B.B. writes:

I think we might be able to understand the crisis better. It requires expectations and learning.

Asset prices and real interest rates depend, among other things, on expected future growth. If productivity growth has been strong, people might expect it remain strong, especially if it is supported by a Narrative of the "New Economy" and computer / net breakthroughs.

What happens when productivity growth starts to fall? Because it is measured with noise, it may take years for people to catch on that productivity growth has slowed. When people catch on, the result is a downward adjustment to asset prices.

Like a plunge in house prices, a stock market collapse, and real interest rates close to zero. Hmmm....Maybe if people were borrowing on the expectation of strong future income growth, perhaps they start defaulting on their loans.

If the Fed is slow to learn about all this, it might be slow to ease interest rates, allowing a demand shock from lower asset prices to feed through into lower nominal GDP growth.

LK Beland writes:

Real output per hour, annual growth:

2004-2016: 1.25%
1990-2004: 2.37%
1973-1990: 1.22%
1950-1973: 2.72%

2004-2016 looks pretty much like 1973-1990.

Also, interesting fact: output per hour grew 5% in 2009. In other words, productivity exploded during the Great Recession. The US became a lean, mean machine.

Scott Sumner writes:

Cloud, Yes, ad hoc. I think they are confusing supply side problems with long run effects of demand shortfalls.

James, After a few years, wages and prices adjust to a negative demand shock.

Todd, I recall reading that labor productivity growth from 2005 through today was much lower than 1995-2004. I have not looked at the data has closely as you have, so perhaps I overstated the suddenness of a post 2004 slump in growth.

B.B. I think your last point is exactly right.

James Alexander writes:

Whoever cared about NGDP anyway? Or especially that whacky NGDPLT variant.

Let's just talk about "productivity", it's a much simpler concept. And so easy to measure.


Todd Kreider writes:


The BLS site shows

1995 - 1999 2.4%
2000 - 2004 3.4%
2005 - 2010 2.0%
2011 - 2016 0.5%

The slower years leading up to the Great Recession, 2006 and 2007 had labor productivity rising at only 1.3% but even that isn't that low and two years couldn't have possibly caused a meltdown as what happened at the end of 2008.

A very good guess is that part of the recent six year slowdown in productivity growth to 0.5% has been driven by something related to the Great Recession itself. "Driven" may be too strong a word, so maybe a "significant factor."

Andrew_FL writes:

Your problem here is you're trapped into thinking about capital in a Knightian way. If you stop thinking about capital as a homogeneous "stock" it's easy to understand why the growth rate doesn't necessarily increase to compensate for the level decrease.

Jamie writes:

Could the lack of productivity growth be a symptom of chronic AD shortage as Kocherlakota argued? His example would be the 1930s where TFP recovered with the economy.

From watching bossiness in highly cyclical industries, a lot of organizational capital spent and executive time wasted rebuilding companies that were forced to make major layoffs when business dried up in 08-10. Though I suspect that effect is not nearly large enough to explain 08's impact.

Another impact I can suspect is R&D and innovation tend to be risky investments with delayed returns. A rise in risk premia combined with cash strapped firms means R&D is disproportionally cut. I need to check the data to verify... This would explain an exacerbation of the depressed 05- TFP growth.

Scott Sumner writes:

LK, Thanks. BTW, labor force growth was much faster back then, which explains the higher RGDP growth rate during the 1970s and 1980s.

James, What matters is NGDP volatility. When it was volatile, it did a lot of harm. It took a while to recover from that, but we finally have done so. If it becomes volatile again, I'll be all over the issue.

Todd, You said:

"A very good guess is that part of the recent six year slowdown in productivity growth to 0.5% has been driven by something related to the Great Recession itself. "Driven" may be too strong a word, so maybe a "significant factor.""

I don't agree, for reasons I explained in this post.

Jamie, I don't see how the 1930s supports the claim that the current problems are related to the Great Recession. Productivity growth was very high in the 1930s, despite massive unemployment. So what's the causal explanation?

Matthew Waters writes:

This paper seems to be the best source about the question of the Great Recession. Productivity growth moved from procyclical to countercyclical around 1980.

It's hard to put in a model, but it seems simple to me why productivity growth was pro-cyclical before 1982 and countercyclical afterward. Employers would be more reluctant to let go of employees before 1982, obviously hurting productivity.

After 1982, employers were quicker to lay off employees. Lay offs could be productivity neutral, but more cyclical industries tended to be less productive (i.e. construction vs. healthcare).

IMO, the productivity "slowdown" 2011-2015 has been the slow burn recovery where these less productive workers were slowly added back. So the Great Recession caused measured productivity to go down during the recovery, though it's far from clear it hurt the course of long-term productivity.

Todd Kreider writes:

"I don't agree, for reasons I explained in this post."

OK, but the data you based this conclusion on was incorrect. I'm not sure how 1.3% productivity the two years before the Great Recession could have caused it as two years in a row of low productivity increases have occurred without triggering a recession. The mid 1990s and late 1990s are especially interesting

Keeping in mind that the low productivity in 2006-2007 averaged 1.3%:

1967-1968 2.7% right before the 1969 recession
1969-1970 0.8%
1971-1973 3.3% right before the 1974 recession
1974 0.6%
1979-1980 -0.2% right before the 1981/82 recession
1988-1989 1.3% right before the relatively mild 1990 recession
1993-1995 0.7% average over three years but no recession

1997-1999 3.0%
2000 3.0% Four years of very strong productivity right before the 2001 recession
2001 2.7%

JustinD writes:

--"IMO, the productivity "slowdown" 2011-2015 has been the slow burn recovery where these less productive workers were slowly added back. So the Great Recession caused measured productivity to go down during the recovery, though it's far from clear it hurt the course of long-term productivity."--

I think that's probably the case.

Productivity growth from 2004Q1-2008Q4 averaged 1.26%, and 2009Q1-2016Q4 averaged 1.33%.

Gab Cormier writes:


Ive looked up FRED's data, you're right about the comment Tyler made, TFP grew steadily from the dip in 1982 and not only from the mid 90s as he suggested (I calculated about 1,15% a year 82-95)

but TFP did flatten around 2005 all the way through the recession (it was actually lower in 08-09 than 05) spiked in 2009-2010 and grew slowly since, supporting what Scott said.

Then I looked up Labor productivity, I guess this all depends on how you break the data down, but my result is
2.88%/yr for 95-04 and 1.34%/year 05-14
If I use 5 years cycles I get
1,74% for 05-09 and 0.94% for 2010-2014

In any event, Scott's argument was that the slowdown in productivity HELPED to cause the recession, not that it was the fundamental reason for it, He's always argued that volatility of NGDP was the fundamental factor.

to conclude, my main objection is towards that comment
''but the data you based this conclusion on was incorrect.''
After checking the facts, the slowdown in productivity did start some time before the Great Recession and not after as you seemed to insinuate, if it was exacerbated by the recession? I don't know, it doesn't seem impossible to me. As Scott said though it is not what we observed in other recessions, but other recessions usually had sharper recoveries as well, maybe there are some dots to connect here.

Anyways thx to both, this all made me think and learn some stuff as well.


Private Sector Non-Farm labor productivity

Jose writes:

I think the theme of this post is the one that will in the future explain the macro events of the last 20 years. Unfortunately the focus on macro variables may hide the cause of the problem.

My hypothesis is that a certain negative supply shock slowly hit the economy from the turn of the century on, lowering potential growth, so slowly that standard macro (new keynesian) models didn't capture it. Most market monetarists analysis of the macro events from 2003 on seem to at least not disprove this idea. David Beckworth at times even argued that the Fed had a too dovish stance in 2003-2005, mostly focusing on reaching a certain NAIRU target, having in mind the low unemployment of the late nineties, after unemployment went up during 2001-2003. Then, of course, all market monetarists think that the Fed was too hawkish from 2007 on.

Summarizing: a lower potential growth from 2000 on may have led the Fed to a series of monetary policy errors that led up to the critical mistakes of 2007-2008. Figuring out productivity slow down is critical. My guess is that excess regulation has something to do with it. And it just occurred to me, Prof. Sumner, people working around regulatory issues in companies and the government, they all fit the bill of workers in the "service sector"...


Return to top