I’ve been compiling data in a haphazard manner to test whether unemployment rates during business cycle years reflects structural or demand-side problems. Out of a lack of time and, to some extent, expertise, I haven’t built a truly worthy model. But, I’m interested in just how much “structural” arguments explain of current unemployment. So, I took Heritage Foundation data — specifically, from their “labor freedom,” “fiscal freedom,” “business freedom,” and “property rights” indexes) —, IMF unemployment data, and World Bank (broad) money growth rate data for two years (2011 and 2012), for 25 countries (I tried to keep within OECD countries, but a lack of available data for some of these countries forced me to look for comparable non-OECD countries).
The, almost certainly, biased model I test is,
The model probably suffers from an irrelevant and omitted variable bias, and I’m not certain that it’s the right functional form given the variables I use — I was debating introducing a lag for change in money, but I don’t know the literature well enough (and all the data is annual, whereas I thought a lag would be more appropriate if the data were more frequent). The sampling is most likely non-random (common factors may influence the data for various countries sampled), and there is a decent chance that cov(εi,xi) ≠ 0. Finally, if I were using something more professional than Excel I’d use more robust standard errors.
But, this is just for fun. The results,
As you can see, of my independent variables, only “labor freedom” and “change in broad money” are statistically significant at the 5 percent level. Based on these results, I’d say that structural issues, at least those that concern the labor market, do heavily influence cyclical unemployment rates. One problem is that they’ll also influence non-cyclical unemployment, I didn’t adjust the unemployment rate data to take out non-cyclical unemployment (mostly because, from my understanding, recent NAIRU estimates are controversial). Also, the results do suggest that demand-side issues are also heavily involved in determining the current unemployment rate.
Based on those coefficients, I wanted to see how much Greek unemployment would fall if they matched the U.S. in both labor freedom and money growth. The 2012 Greek and U.S. labor freedom ratings are 42.1 and 95.5, respectively; their money growth rates are –4.96 and 4.84, respectively. If Greece were to improve their labor freedom score to 95.5, the model predicts (holding all else constant) a decrease of 5.12 percentage points. If broad money were to grow at the same pace, Greek unemployment would be 2.16 percentage points lower.
Interestingly, one theory my regression doesn’t support is the regime uncertainty theory, if we assume that “business freedom” and “property rights” are good proxies for that.
Hopefully, at the end of December I’ll start having more time for my hobbies and I can look into building a better model, even if it’s still only just for fun. One problem I’ve been having is to find a good variable for aggregate demand shortages. I was suggested to use inventories, but I still haven’t really looked into finding the data to that, mostly out of a lack of time.
Here are the scatterplots for the two statistically significant variables,
My intuition is that structural theories are less relevant for countries like the U.S. and the U.K., which have relatively high labor freedom scores, but highly relevant for countries suffering from excruciatingly high unemployment (e.g. Greece and Spain).