Productivity, or output per hour, is an especially useful metric called upon to answer many questions about an economy. How efficient is production (i.e., how many widgets are we making per hour now versus before), how fast can living standards, wages, and incomes rise without generating inflation, is there evidence of technological advances boosting production? It’s also tricky to measure, especially as we move from manufacturing to services, but the BLS series plotted below is widely accepted as a solid indicator of these sorts of questions.
I’ve plotted the series in annual changes along with a smooth trend in order to help us think broadly about two assertions I hear often these days. First, firms are so profitable—which they are (I’ve got a piece coming out on this on the NYT Economix blog next week)—because they’ve been squeezing more productivity out of their workforce. Second, automation is displacing workers at a faster clip.
Would not both of these show up as faster productivity growth? And yet the trend has clearly decelerated (here’s Dean Baker on the automation question).
I’m not saying this is the final word by a long shot. From the mid-1980s to the mid-90s, economists said the same thing about computerization: if it’s so transformative, why isn’t it showing up in the productivity stats? And then, in the mid-1990s, as you see in the figure, the series did in fact accelerate quite sharply.
But at least first blush, and shaving with Occam’s razor, this decelerating trend in output per hour seems out of sync with those two popular assertions. Which leads me to wonder if folks are just bending themselves into a pretzel with such assertions when the real culprit is just good, I mean “bad,” old-fashioned weak demand.
Source: BLS, nonfarm productivity, quarterly data, yr/yr changes, and HP filter (lamda=1600)
“First, firms are so profitable—which they are …because they’ve been squeezing more productivity out of their workforce. Second, automation is displacing workers at a faster clip.”
Regardless of where the firms “squeeze” the costs out of, we should be seeing the impact of it in prices instead of profits (especially with weak demand and excess capacity). Why isn’t this happening, and why aren’t economists even trying to answer this question? It seems that the leading candidate for an explanation is that economists don’t like the answers. Where does this kind of pricing power come from? Certainly not from competition. And economists are “stumped” by all of this unexplainable inequality as well? Probably just a coincidence right?
“From the mid-1980s to the mid-90s, economists said the same thing about computerization: if it’s so transformative, why isn’t it showing up in the productivity stats? ”
Hmm… Dunno? Maybe it’s because we no longer make what we sell? Just a stab in the dark…
[end of snark]
Look at the massive upswing in productivity 2006-2008. Clearly that was not sustainable. Some day (actually I doubt it), the fed will realize what goes up must come down.
“Look at the massive upswing in productivity 2006-2008.”
I know my eyes are old but either they have really crapped out or I did not see the “snark” tag.
oops. Meant to say 2008-2010
Probably has to do with business squeezing the last juice out of the lemon.
Jared,
I think there are a lot of things going on here. I can agree that good old fashioned demand is what we need right now. We need lots and lots of infrastructure spending and technology investment now.
But I see some other things going on. I think it would be extremely effective in making the case for infrastructure spending if we could flesh out all of these hidden effects. There are always reasons for a shortage of demand. I think I know quite a few of them, and I can see them in the data. If we can make a rock-solid case of identifying these causes, and I think we can, then we can leave every other argument in the dust.
I’m certain of it. I can’t do it my myself though. I would like to help you.
I’m being dense. How do weak demand and productivity growth, or lack of it, relate in the data? Maybe you could do a follow up?
Not dense at all–good Q. Demand=output=numerator of productivity.
Suppose Demand>output or as at present, Demand<output (possible in the short term via inventory growth)? In the first condition we should have modest inflation, wage growth and a tightening labor market. In the second condition we would have slower growth, if any. If demand does not increase, output will fall as inventories are adjusted, suggesting increased slack in the labor market and future demand diminution. The equilibrium argument is simply an unwillingness to deal in more realistic scenarios. Increased demand is the solution to a sluggish economy and increased productivity. Why increased productivity? Increasing output is not as simple as increasing demand. The capital adjustments necessary to increase output would of competitive necessity be spent on some productivity enhancing technology and infrastructure. Where is Helicopter Ben when we really need him?
So is there any way to know (and filter out) how much is of the change is simply due to changes in capacity utilization and the resulting use (or idling) of less productive equipment and facilities? Is it adjusted for overtime rates or does this add more “noise”?
Good Q: for that, and depending on how deep in the weeds you want to go, you have to look at the work of John Fenald from the SF Fed Reserve bank. He’s got various papers on this wherein he tries to develop measures of productivity that control for capacity utilization.
Why try to control for a factor (Capx) of investment in infrastructure which is inherently labor intensive? Productivity increasing Capx would necessarily depend on increased labor, unless we completely disregard the views of Baker and Bernstein on robotic substitution, which we should not do. The long term wealth effects may accrue to the holders of capital (that is another issue), but the more immediate affect would (should) be increased labor demand. Old or new facilities, someone must sweep the floors and wash the windows, so to speak.
“…and depending on how deep in the weeds you want to go…”
No better way to keep us nerds happy & busy than to provide some directions to more and deeper weed beds!
Thanks for the pointer.
@Larry Signor
“Why try to control for a factor (Capx)…”
I’m not sure I fully understand your question but if you’re asking why I would “filter” out effects of capacity utilization it is basically to control for what I would call “noise” in the measurement that might obscure the impact of other factors. To illustrate, think of two farmers (equally skilled), one using a modern tractor and the other using a horse & plow. When demand exceeds what the modern tractor can produce, the horse & plow will be used to make up the difference in production and “productivity” will decline sharply to the extent the plow is used; the reverse happens when demand slows and the tractor can produce everything. “Average productivity” increases as the less productive assets are idled without any changes in skills, training, or technical ability. Another example of the effects of the equipment used might be “peaker” plants at electric utilities. They are used to meet high demand conditions but are less efficient to operate per unit of electricity generated. When you turn them off, average efficiency goes up just because you’re not using them, not due to any changes in the “skills” of the operators or “technological abilities.” If we measure output/hours these effects make it appear we’re getting more productive when all we did was stop using the low -productivity equipment (and the reverse when we bring the less productive equipment on line to meet higher than expected demand).
Now what I don’t understand, is what “effects” you propose to filter out with your model. ‘Noise”, “other factors:? Certainly your stylistic examples are dependent on fallacies (very few farmers use horse labor) and peak electric is a moving target, but still a marginal requirement. If I remember Econ101, consumption at the margin, in a fully supplied market, history indicates more demand=more output= increased numerator=increased productivity…the man hour cost in this relationship is irrelevant.
I also question the “obsolete” equipment meme, due to tax externalities which encourage business to invest in newer equipment. In the end, Capx expansion must require labor expansion, how much is the question.
-“Now what I don’t understand, is what “effects” you propose to filter out with your model.”
it depends on the question you’re trying to answer. If I want to know if we’ve gotten better (more productive per hour) as a result of technological advances, I want to know the “state of the art” productivity, not the weighted average of the old and the new. Just using less of the old, less productive equipment doesn’t mean we’ve advanced any technologically, but if we just measure average productivity at high & low capacity utilization, the “measurement” will show we did. The horses and tractors was just an example to illustrate the point, you can use any high vs low productivity equipment and you’ll still get the same result.
This is a little deep for me, but it seems like you want to know how increased technological inputs affect productivity and you want to filter out the affects of obsolete systems. How exactly you would use Capx to do this puzzles me. Measuring Capx at high and low productivity levels might tell us about the relationship i.e. productivity change/unit Capx, but what does it say about the composition of productivity change? I don’t wish to be annoying, but I am truly puzzled by this approach.
-“…but what does it say about the composition of productivity change?”
That composition is what I’d be interested in isolating. For example, If you look at the early 90’s & again in the ’07-’10 period you’ll see large spikes (prod_4) that coincided with recessions. What drove these spikes and returns? Was there some technological change that coincided with each of these? Was it laying off of the less productive providers of “labor,” did employers extract more “unpaid & unaccounted for” hours out of powerless workers, or did we just stop using the using the lower productivity plant and equipment we had been using to meet the higher demand because we no longer needed it? Output/hours throws all of this together as if it didn’t really matter which was more or less important. On average, every American has one breast and one testicle; that fact doesn’t provide much insight into what Americans are like. Data & stats. vs. explanation & understanding.
“What drove these spikes and returns?”…now you have penetrated the fog for me. Thanks. I would have assumed the innate advantage of capital in a contraction, but that would need to be displayed with more data than merely my opinion. Filtering out as many variables as possible would tell us if Capx increased during times of economic contraction coincident with productivity increases, If Capx does not increase, we might be correct in thinking Capx is over compensated during economic contractions. If this is the case, then the business hates labor meme becomes even stronger. A very illuminating discussion.
Seems to be that you can only squeeze so much out of workers. And when labor is cheap there is little reason for less mechanization. There is great reason to replace workers when labor is rare and expensive. Is it any wonder productivity would slow?
It is also worth noting that workers have not gotten any benefit from increased productivity in decades. This is not much of a motivator.
Please analyse the data further to see how much of the trends was from shift in the composition of output/labour force over time and how much was from changes in productivity at industry level (which were the key data we want to see in this discussion).
On the question of the effect of output on productivity, please refer to Verdoorn’s Law.
Thank you.
The calculation of productivity as total_output/total_hours_worked would give us an estimate of the average productivity IF we assume a symmetric distribution of productivity. However, if you have highly skewed productivity distribution, such that significant number of the hours worked was at the lower productivity jobs, as will be evident from the high growth at the low paying jobs in recent years, then the BLS finding of declining productivity is highly influenced by the shift in labor force composition. That will put your answers to the opening questions in completely different light.