Productivity is a somewhat mysterious concept, and I love untangling a good productivity mystery. As Brad de Long has observed in a string of recent posts, US productivity has behaved quite mysteriously in the last few years. The key facts noted by Brad are
- Labour productivity (output per hour worked) usually falls during recessions/slowdowns
- Over the last three years, US output per labour hour has risen, at an accelerating rate
- Over the same period, hours worked have fallen at a rate consistent with a deep recession
Brad’s hypothesis is that the increase in productivity is primarily due to technological progress in information and communications technology, and that the decline in hours worked is caused by the combination of rapidly growing productivity and inadequate demand. I don’t think this analysis can be sustained.
An obvious weakness in Brad’s hypothesis is that it’s implicitly using a closed-economy model. If US productivity is rising faster than in the rest of the world , this should be reflected in some combination of rising exports, declining imports and an appreciating exchange rate. At least for the last year, all these variables have gone in the ‘wrong’ direction. Growth in US demand has been met by imports rather than by domestic production, even though the US dollar is depreciating against most trading partners (China is the notable exception).
In a recent post, I observed another point that complicates the puzzle. While labour productivity (output per hour worked) has been rising at near-record rates, US performance on a broader productivity measure, multifactor productivity, has been much weaker. MFP actually fell between 2000 and 2001 (the most recent year for which MFP data has been published).
The most important feature of the MFP measure is that it takes account of inputs of capital (including land) as well as labor. But before moving on to that, it’s worth clearing up some issues relating specifically to labor productivity. I’ll begin with a fact that may be surprising to those who haven’t followed this debate closely. As Brad himself noted not long ago, the world leader in labor productivity per hour is Belgium. (The US is about equal to France and Germany on this measure). Belgium’s high productivity is accompanied by a low level of labor input (hours worked per person), due to a combination of shorter hours and a very low employment/population ratio due mainly to high levels of hidden and overt unemployment.
This gives Belgium three sources of high productivity, leaving aside any advantages or disadvantages with respect to technology and work organisation
- High capital-labour ratio
- Composition effects – the least skilled workers are most likely to be unemployed
- More speculatively, marginal productivity effects – beyond some point additional hours per person produce less additional output
I discussed the first two sources of higher productivity recently, and argued that productivity gains from these sources are not sustainable and, in welfare terms, essentially illusory. The productivity effects of shorter working hours are trickier and I will try to discuss them in a later post.
An obvious implication of the Belgian example is that productivity will increase when a country becomes more like Belgium. This has clearly happened in the United States in the past few years, as both the employment/population ratio and hours worked per person have fallen. Particularly noteworthy in relation to labor composition is the big decline in teenage employment over the last few years. The US was until recently, a striking exception to the general trend of declining teenage employment, reflecting both cultural factors and the strength of the labor market in the 1990s boom, but teenage employment has fallen sharply. Since teenagers are normally employed in low-productivity jobs, this decline tends to increase output per our.
Another factor improving labor composition is that older workers whose investments lost value in the stock market crash have been delaying retirement or, in some cases, returning to the labor force.
The Bureau of Labor Statistics Multifactor Productivity series (unlike Australia’s) takes explicit account of labour composition. For the period from 2000 to 2001, the BLS estimated that hours worked fell by 1.4 per cent, but labor input fell by only 0.6 per cent, with the difference reflecting changes in labor composition.
It seems likely, judging by data on things like teenage employment, that the change in labor composition has continued since 2001. So about one percentage point of the annual increase in US labor productivity can be attributed to the ‘Belgium effect’.
In a sense, of course, this is getting things backwards. Relative to countries like Belgium, US productivity growth was understated in the 1980s and 1990s, leading to the spurious conclusion that Belgium had overtaken the US. Now this artificial gap is being wound back.