Wednesday, June 30, 2010

US Productivity Turning Points

Labor productivity is an important variable with respect to several issues that interest me.  On the one hand, we would expect that an approaching machine-intelligence singularity would be signaled by super-exponential increases in productivity.   On the other hand, if the peak-oil-doomer-relocalist axis that thinks we will have to go back to doing everything by hand was correct, we would expect labor productivity to undergo massive drops.

The above graph shows the US data (from the BLS) where I have taken the average growth rate over each ten year period in output/hour for all business.  The underlying quarterly data start in 1947, so 1957 is the earliest date for which I have a ten year growth rate.  I did it this way to smooth out the year-to-year noise.

The narrative I would tell here would be one of competing forces.  On the one hand, the sharp drop in productivity growth in the 1970s seems to me very likely to be associated with the oil shocks, which caused a sharp rise in the price of energy in general, and mobility in particular.  There also seems to be a much less serious turning point in 2005, that it's similarly tempting to associate with the 2005-2008 oil price shock.

On the other hand, the big run up in productivity since the mid 1990s seems very tempting to associate with the growth of the Internet (I have shown the Netscape IPO date above as a marker of the beginning of the Internet as a major commercial force).

Overall, you certainly couldn't claim this is a super-exponential.  On the other hand, it's striking that productivity growth has been accelerating pretty rapidly since the late 1990s.  As soon as the pressure came off oil prices a little with the great recession (oil is still not cheap by historical standards) it started to accelerate sharply again.

Of course, this is a somewhat speculative narrative - firm attribution has got to be extremely difficult here, given the way multiple overlapping waves of innovation are washing through the economy at the same time, all of them affecting productivity.  Still, it will be interesting to track this in the future.


Stefan said...


You've missed the most obvious explanation for that decline in productivity from the late 1960s through the mid 1980s. That period perfectly coincides with the entry of the Baby Boom generation into the workforce. Productivity generally rises as workers become more experienced, so the influx of large numbers of inexperienced new workers into the labor force was bound to depress productivity for a while. Also, younger inexperienced workers are cheaper to hire than are older experienced ones, so the temporary surplus of large numbers of inexperienced young workers during that time period undoubtedly tempted many employers to hire additional workers. An increase in the labor input reduces productivity. (Of course, there was unemployment and downsizings galore during that time period. I am only suggesting that employers hired MORE new employees than they otherwise would have.)

Once the supply of Boomers dried up, the substitution of technology for labor became more profitable and thus started driving productivity up again. The impending retirement of the Boomers can be expected to result in an even further boost in productivity as employers will be forced to substitute even more technology to replace departing workers.

Greg said...

Further to Stefan's points, the 1970s-1990s period was one of structural change in the US economy--the outsourcing of heavy industrial manufacturing. Employment moved into relatively lower-productivity services.

The rising trend in productivity since 1995 is partly due to companies learning how to automate services.

Stuart Staniford said...


That's certainly an interesting point. I don't think the timing quite works for it - the baby boom would have started entering the workforce in sizeable numbers in 1963 when the leading edge hit 18. I don't think there's clear evidence of a decline in productivity growth then - the growth is just kind of choppy without clear trend in the late fifties in sixties. To me, the really clear drop begins in 1973.

However, I do take your point that ideally the numbers should be age corrected, since certainly output/hour would not be expected to be constant with age. However, the BLS data don't provide a basis for doing this.

Anonymous said...

A singularity would be certainly bad for us all as we would be all out of work, see matrix type situation or Terminator, who needs carbon based life forms to disturb ourr routine? say the robots.

But anyway with less cheap energy (Peak Oil) to keep it all going we will need more muscle power which is necessarily less productive in the sense of productivity (GDP) per human unit. If that is important. Essentially we need to produce less at higher qualtiy (long lasting high quality goods) through fine, experienced human craftsmanship. This saves us waste (no more throw away society) and retains the environment. The environemnt remains then for us in its long-term productive state asd we have peace of mind and happiness through enjoyment of nature and our part in it (pleasant walks, etc.).

rks said...

An interesting case is the productivity of a barber. What he does probably hasn't changed much. Yet his income as gone up with average earnings, otherwise no barbers. What proportion of workers are similar? My guess is a lot. So rises in productivity in one place get smeared across the economy. My guess is that the only actual rise in productivity has been in converting energy into useful stuff. We made chairs, then we made machines that make chairs, now we have systems that can quickly create machines to build different chairs or anything else. These systems can go idle when energy itself gets expensive.

Stuart Staniford said...


What you describe has a name - Baumol's disease. The BLS does produce sectoral estimates of productivity, so perhaps I'll explore that data further in future.

There's also an interesting paper here with some discussion of the issues and the literature. As usual, it doesn't seem to have occurred to economists to think that perhaps significant changes in the energy supply to the economy would necessarily cause a significant response in it's behavior, since it's a dissipative thermodynamic system.

rks said...

Thanks for the pointers. Mathematicians say "You don't really understand something until you understand it in two different ways". I recommend this to economists, in which context I suggest trying to understand the economy without talking about money (except as this funny thing that facilitates the whole process without being real).
I have been inspired to compose a couple of messages to our (Australia's) new Prime Minister. While I suspect she doesn't read my blog I may be able to get something published more visibly (they have accepted some earlier stuff, also on my blog, though it hasn't appeared yet). Anyway I'd certainly appreciate your comments, particularly on the 2nd one. My blog is The post is at Or email to @t Of course, other readers of your blog are also welcome to comment.

MisterMoose said...

Speaking of barbers... My girlfriend's Dad was a barber who had a little one-man shop in our neighborhood back in the 1970s. He said business for his profession fell off noticeably after the Beatles became popular, and remained relatively low during the long-hair hippie days into the '70s. So, even though we need barbers to keep cutting our hair, we may need more or fewer depending on prevailing styles and fads.

BTW, "Shave and a haircut - two bits" was the actual price for those services about a hundred years ago, but 25 cents could buy a lot more then than it can now...

Gary said...


Here is a very interesting post on productivity which seems to fit well with your singularity concerns...