January 23, 2013

Structlical Problems?

Unemployment in the United States is still high: it has been recovering, but very slowly. Compounding the unemployment problem is the increasing wage inequality that has been ongoing since the early 1980s. There are a number of theories floating around that try to address these issues; the simplest way to frame the unemployment debate is as a question of structural versus cyclical unemployment. Structural unemployment is when jobs cannot be filled because people do not have the right skills or access to jobs for other reasons (such as location). Cyclical unemployment is due to the business cycle and businesses laying off workers because the macroeconomy is doing poorly and businesses are not investing.

Last week I linked to several papers that attempted to address these issues:

~ NBER working paper arguing that unemployment is currently cyclical.

~ EPI paper arguing that changes in wages were not due to skill-biased technological change (SBTC); instead they were due to government policies.

In the past two weeks both of these papers have gotten a bit of a reaction. Most notable is the back-and-forth between Washington Post columnist Dylan Matthews and one of the EPI paper's authors, John Schmitt. Matthews argues that the SBTC is a bigger deal than the paper makes it out to be. Schmitt comes replies with a half-critique, half-reframing of the issue. Schmitt's main argument is that in the last decade SBTC has been almost non-existent according to some of the original data that prompted the paper that the EPI paper was mostly critiquing. Both articles are a good read.

I also came across a rather harsh critique of the NBER paper by Rob Atkinson of The Innovation Files/ITIF. Atkinson argues that the large (continued but increasing) slide in manufacturing employment over the past decade is clear evidence of structural change. On the other hand, EPI's Economic Snapshot of the day looks a lot like cyclical unemployment, with a graph showing how the number of job seekers dwarfs the number of job openings in every industry.

There are two points to be made.

The first is to point out the somewhat tangled political allegiances and the politcal conclusions attendant to the sides of the economic debate. If unemployment is cyclical, government stimulus (fiscal or monetary) can have an impact by prompting businesses to invest more and thus hire more. If unemployment is structural, government spending intended to stimulate the economy in the Keynesian sense will not encourage businesses to invest.

It's interesting to see who is arguing what here, however. The left of the political spectrum is generally pro-stimulus, and for policies that are explicitly pro-poor (welfare, etc...). Arguments for stimulus depend on cyclical unemployment being an important factor. But a focus on cyclical unemployment arguments detract from longer-term structural arguments. An example of this complexity is the way that Atkinson's post manages, somehow, to the attack the liberal wing of the neoclassical economic (basically neokeynesian) consensus, from a centrist pro-business standpoint, using arguments many leftists would be comfortable with.

This argumental jujitsu leads to my other point, which is that politics has a hard time with economic complexity. Most academic economists would support a wide range of policies from a wide range of the political spectrum. Economic findings, if economists are lucky enough to find any quantitatively sound conclusions, often make claims like "30% of the increase in X can be explained by Y". Such claims are underwhelming even if extremely important because having two causes is far less convincing than having just one. We like to think in absolutes--structural or cyclical--because it's simpler for us and for anyone we are trying to get to listen to us.


  1. I'd be curious to see if there is any discussion about the turning point where the marginal unemployed worker goes from being structurally unemployed to cyclically unemployed, and what policies (training?) are suggested for converting structurally unemployed workers to cyclically unemployed. Or can you go straight from structurally unemployed to unemployed?

  2. Err, straight from structurally unemployed to employed, that is.