Friday, February 24, 2012

Country size and the unemployment rate

When firms merge, they mention economies of scale and sharing of fix costs. They also sometimes imply increased market power. When firms shed business lines to shrink, they put forward concentrating on core business. From this, it is not very clear where it is an advantage to have a large or small firm. What about countries? Are large countries more successful? One way to look at it, is to compare the unemployment rate to country size. Why not.

Norbert Berthold and Klaus Gründler did that and find that small is beautiful. Indeed they find a positive correlation of the unemployment rate with population and with area, looking at subcontinents, countries and regions within countries. How could that be? They first offer a bizarre explanation, using a Cobb-Douglas production function where they substitute labor by population times the employment rate, then isolating the unemployment rate. Of course, if you leave the other factors constant, increasing population will increase the unemployment rate. The problem is that the other factors are not constant. The paper later ties the correlations in a parabolic way to collective bargaining and the size of the government, the reasoning being rather obscure.

What bugs me more about the paper, though, is first the underlying assumption that each data point (a region) is treated like a closed economy. Regions trade with each other, and this implies that labor markets are linked. This is especially the case when you look at Europe today where borders have much less significance than just ten years ago. Second, the analysis is performed for 2010, not long term averages. All what is said is thus valid for a particular point in the business cycle. Finally, nowhere in the text it is mentioned that populous regions may have different characteristics that could matter for unemployment, say they are more urban, have more manufacturing, are more diversified. After all, these regions are accidents of geography and history, and both lend some economic characteristics to them. And these are the ones that matter. Otherwise, you would get the idea that you just need to administratively split up countries, and immediately a labor market miracle happens.

1 comment:

Anonymous said...

I searched their paper for the word "cluster" and nothing came up. The tables contain no footnotes about geographic clustering of their standard errors.

Seems like a finding, if accurate, would be really interesting. But only interesting if it holds up over the long haul, and with careful statistical work. Specifically, I'd want to see a between regression on all post-war data with continent dummies and sub-continent regional clusters. Small country = Europe dummy, in general. And then I'd want out-of-sample predictive power for the pre-war data...