Friday, April 5, 2013

Who cares about the median voter? Not the politicians, maybe

Most implementations of political economy in economic modeling assume that the median voter is the determinant voter. Some studies may take into account that the likelihood of voting is not the same for everyone, but the basic principle is that politicians adjust their proposals according to some sort of median in a universe where every man and woman has roughly equal weight.

Pablo Torija tests this by looking at which parties were in power in OECD countries. Assuming that voter preferences are quadratic in income and depend on the distance of the individual's policy preference to the one of the party in power, one can then determine for which household this policy would be optimal, and then see where this household is in the empirical income distribution. The actual implementation is a bit more involved than that, as parties' ideology is coded on a linear scale, and this is interacted with income. All this is then regressed on happiness indicators from surveys to proxy for utility.

The results are striking. Right-wing parties' median voter is at the 99th percentile. For left-wing parties, it is at the 95th percentile. That is for 2009. Extrapolating the empirical model, it is much closer to the median in the 1970s. But are these results believable? For one, the policy preferences of the individual somehow vanish between the simple theoretical model and the empirical implementation. And more generally, static preferences that only depend on income and ideology seem rather crude. Is this income before or after taxes? That can have a tremendous impact. And people may care about other things, see for example the recent debate about the importance of leisure and working conditions in US and French tire plants. I suspect that results are biased upwards by all this.

1 comment:

Pablo Torija said...

Thanks for your interests in my research. I think that debates on the quality of democracies are crucial now-a-days, thanks for contributing to it.

I would like to comment two things about your concerns on the paper. First the theoretical background has been criticised often, I was a bit concern in using a more complex one as all that I needed was the function of the citizens U = f(y) + f(x,y). Where y is income and x the position of politicians. This utility function is rather standard in economic theory, but as I said it is a general critique on the paper.

Second, the income scale is relative, not absolute. It is indifferent if it is after of before taxes (as long as taxes do not take that much from an individual that he is better off by earning less, ie. marginal taxation below 100).

Finally a potential upward bias (I am not sure if it exists, anyway) would make as maximum the social democrats work in 2009 for the 95th percentile in the best model (the one which claims that politicians work for the 1%) as the position of the politicians must be in R+...

Anyway, thanks again for your interest, and congratulations for your great blog.