Thursday, May 31, 2012

Maybe he had a point

In writing my recent posts on inflation, I recalled an obscure article I read when I was in graduate school (Tjerk Huppes, "Anomie and Inflation," in Economics and Sociology:  Towards an Integration, edited by Tjerk Huppes, 1976).  The article argued that contemporary (at the time) inflation was the result of a breakdown of norms ("anomie").  Basically, the idea was that traditional standards that people had accepted in various areas of life were breaking down, resulting in conflict, and conflict over wages and economic rewards produced inflation.  His evidence was that several "indicators" of anomie (suicide, homicide, divorce) were strongly correlated with rates of inflation over the period 1960-74.  Having taken a couple of courses in quantitative methods, I spotted the obvious statistical problem--all of the variables had a strong trend over the time period, so of course they had a high correlation.  I recall adding some more recent data, trying it again, and discovering, much to my surprise, that the apparent relationship still was there.

Now there's been a lot more time for data to accumulate, so I looked up data on homicide, divorce, and inflation from 1950 until the latest year available (2009).  Suicide is the classic measure of anomie, but it was harder to find annual data, so I left it out.  I constructed an index of "anomie" that combined homicide and divorce, and regressed inflation on anomie and unemployment.  With a correction for first-order autocorrelation (basically, inertia in everything else that affects inflation, here are the results:

constant  2.52

anomie    1.31

unemp    -0.51

rho       0.61

So there is very strong evidence of a relationship between anomie and inflation (plus pretty strong evidence of a traditional Phillips curve relationship between unemployment and inflation).  Of course, this is a simplistic model, but when you have an apparent relationship that's this strong you have to suspect there's really something there. 

No comments:

Post a Comment