22 February 2012

Does Increased Productivity Decrease Manufacturing Jobs?

I was intrigued to see an argument being made in a recent Brookings Institute report (Helper et al. here in PDF) that productivity gains do not lead to job losses in manufacturing. This post explains why I think that this argument is not well characterized and probably just wrong.

Helper et al. make the following argument (pp. 9-10):
Some argue that strong productivity growth has caused much of America’s manufacturing job loss, especially in the last decade. This theory, which contends that technology is replacing workers, stems from the observation that apparent productivity gains have coincided with manufacturing job loss in the 1990’s and 2000’s. Yet there is no economic reason why increased productivity must lead to job loss.
The first thing to note is that Brookings is talking about overall productivity gains, and not labor productivity. The different is important. Labor productivity according to the BLS "is the ratio of the output of goods and services to the labor hours devoted to the production of that output." Overall ("multi-factor") productivity "relates output to a combination of inputs used in the production of that output, such as labor and capital or capital, labor, energy, materials, and purchased business services (KLEMS). Capital includes equipment, structures, inventories, and land." The distinction is crucial to understanding employment changes in manufacturing.
There are a lot of moving parts in multifactor productivity beyond labor. So in any analysisa good first place to start, fairly obviously, is the relationship of labor productivity and jobs. The above graph shows the relationship of changes in labor productivity with changes in employment for manufacturing, according to data from the BLS (with 2005 = 100 for each time series).

It shows an extremely close relationship between labor productivity and employment, one that is even stronger on a one-year lagged basis (as companies fully exploit labor productivity gains). This relationship should not be at all surprising since labor productivity is defined in terms of hours worked, which has a direct relationship with employment.

So the first order conclusion necessarily must be that gains in labor productivity in manufacturing necessarily lead to losses in jobs. And it is here that Brookings gets a bit confused with its focus on total productivity:
Even though a productivity increase means that fewer workers are needed to produce a given quantity of output, the productivity increase also allows product prices to be lower, increasing the size of the product market. The bigger market means that firms will need to hire more workers. The additional hiring needed to produce for a bigger product market usually offsets the initial labor-saving impact of the productivity increase. Therefore, the overall impact of a productivity increase is usually to expand employment rather than reduce it.
To see where Brookings has erred, think about the mathematics here -- If a market for a good expands that means that more output is required. Additional labor needed to meet that demand is an input. If the magnitude of the change in the input is equivalent to the magnitude of the change in output, then there has in fact been no increase in labor productivity, which is defined as an ability to produce more with fewer hours on the job. If labor productivity is improving then the inexorable result will in the long run be a loss of jobs, even in an expanding market.  In the short term, market expansion can offset job losses, such as we are currently experiencing in the US, but only temporarily. In short, with respect to jobs, simple math tells us that market expansion cannot defeat labor productivity. Just look at agriculture for an example of this dynamic (e.g., programs such as the Food for Peace effort of the 1950s were designed to expand markets in the name of aid and diplomacy).

Now, with respect to total productivity it is certainly possible that changes in other inputs can work to offset gains in labor productivity, but only at the expense of overall productivity (i.e., if labor productivity improves then KEMS productivity must degrade).  Similarly, total productivity gains can occur even if labor productivity degrades or is held constant (i.e., via improvements in KEMS productivity). Thus, there is nothing inconsistent with overall gains in productivity and constant or even increasing employment, but such an outcome does require that labor productivity gains are smaller than the combined effects of market expansion and degradation in other sources of productivity gains. This is a tough ask: The case of Germany suggests that even aggressive policies to influence labor productivity (which in some cases degrade labor productivity) .

Does increased productivity in the manufacturing sector result in a loss of jobs? If those productivity gains are in labor then the answer must be yes.