Friday, January 04, 2008

1 Myth in 5 Myths about American Voters

via Mark Thoma I read this interesting article by Bryan Caplan in The Washington Post

His point is that US voters are ignorant and irrational and it matters. However, he overstates his case in two ways

1. "Surely Americans want serious change on Iraq, you say? True, some 60 percent of Americans now say the war was a mistake. But given the available options, voters are still getting what they want. If Iraq were a stable and enthusiastic ally, we'd like to leave today, but that's not on the menu. Most Americans now favor a timetable for withdrawal, but how many would want to stick to a schedule if that means handing Iraq over to radical Islamists? In a few years, the majority may be ready for "peace at any price" -- but not yet."

This is nonsense. Most Americans want a to impose a timetable for withdrawal. They are not getting what they want. Caplan argues that they might change their minds if they are convinced that the policy they currently support will lead to a worse outcome than he imagines they currently imagine. With this sort of argument, he can prove that people don't support any policy no matter how strongly they support it.

For some reason, Caplan has decided to ignore the fact that the clear wish of the majority of Americans on the most important issue of the day is not being obeyed. His argument to support his refusal to accept this clear fact would enable him to ignore any fact at all about public opinion. Caplan has decided that he knows better than the American people what they really want and, if they disagree with him, it just shows that they are too ignorant to understand the issue. This in an article allegedly about public opinion.

Caplan is similarly dismissive of protectionists. I agree with him on that issue, but don't think I know enough to be sure we are right and I'm an economist.

2.. "there is only the tiniest correlation between income and party. " This is a quantitative argument. However, Caplan does not say anything about how large is the allegedly tiny correlation. Let's ask the Google. First link here.

"Upper Income Quintile (Annual household income above approx. $92,000, in 2005 dollars): In 2005, the GOP edge over Democrats among people in this income bracket is 38%-27%"

"Lower Income Quintile (Annual household income below approx. $19,000): Republicans continue to trail by sizable margins in this income bracket. Currently the Democrats enjoy a 42%-20% advantage overall, comparable to their 43%-18% edge in 1992. Among whites, the Democratic edge is 37%-24% now and had been 37%-22% back in 1992."

That does not seem to me to be a tiny correlation.

In any case, Caplan really should not make a claim about the magnitude of a correlation without giving numbers. People (including I would say Caplan) have a lot of trouble with the idea that correlations are not all 1 0 or -1, that is, tend to assume that if a is positively correlated with b and b is positively correlated with c then a must be positively correlated with c, that is, they interpret "positive correlation" to mean "correlation 1" then when this isn't true (red states are poorer than blue states on average) they conclude that the correlations must be so small as to not matter (so positive correlation becomes zero correlation).

I agree with Caplan that the voters are ignorant and irrational, however I also agree with Churchill that Democracy is the worst possible system of government except for all of the other ones which have been tried. Caplan convinces me that journalists such as Caplan are even less qualified to decide policy questions than the electoral process, because voters, at least, do not all share the same fact resistant views on critical issues.

No comments: