Bryan Caplan  

Plus Noise!

PRINT
Are German schoolchildren taug... Some Replies to Me...
One lovable thing about Weeden and Kurzban's The Hidden Agenda of the Political Mind: When they criticize me, they also quietly hand readers my rebuttal.  Them:
Economist Bryan Caplan in his book, The Myth of the Rational Voter, asserted (with his tongue in its usual non-cheeky place): "There are countless issues that people care about, from gun control and abortion to government spending and the environment...  If you know a person's position on one, you can predict his view on the rest to a surprising degree.  In formal statistical terms, political opinions look one-dimensional.  They boil down to roughly one big opinion, plus noise."

[...]

If we take the assertions from Caplan and Pinker (not to mention Stewart) seriously, we should be able to take people's view on one of these issues and know their views on the other issue...

The data are decidedly less tidy.
Just one problem: I'm well-aware that the data aren't tidy.  Accurately predicting individual opinions is hard.  I deliberately included the words "plus noise" to ensure that readers knew I was not claiming great predictive powers.  

I should add, however, that opinion has more ideological coherence than Weeden and Kurzban claim.  Yes, using one specific issue position to predict an unrelated specific issue position is almost fruitless.  In the General Social Survey, the correlation between gun control and abortion positions is a mere .02.  But using self-identified ideology to predict specific issue positions is noticeably more rewarding.  In the GSS, ideology correlates at .11 with gun control views and .22 with abortion views. 

Why would this be?  Simple.  Specific issue views, like individual items on the SAT, are very noisy.  Ideology, like overall SAT score, is not so noisy.  When you correlate two noisy things with each other, you get a really tiny correlation.  When you correlate one noisy thing with a not-so-noisy thing, you get a moderate correlation.  If Weeden and Kurzban really wanted to dispute the one-dimensionality of political opinion, they should have been correlating specific issue views with ideology, not specific issue views with each other.


Comments and Sharing






COMMENTS (5 to date)
Nathan W writes:

Informally, I estimate that opinions on gun control are probably one of the strongest single predictors of views of political views. In particular, people who endorse gun control seem willing to blindly accept abstruse reasoning on all manner of other issues. However, I have met some people from rural areas where hunting is legitimately a part of their life, and many of these people broadly support other aspects of the welfare state such as health care and social security. So, I would say that people who actually use guns in their lives for practical reasons probably defend access to guns for legitimate reasons, and most of the rest have been fully eye wool pulled (aka quasi brainwashed, a side effect of watching news of such high quality that it must remind us hundreds of times a day that it is unbiased (Fox)).

Indeed, it is useful to welcome critique of our intellectual ideas. We are more likely to learn that way.

Abortion, I think less so, I think because there are many people who are very broadly compassionate, and this, not ideological purity, is what drives them to wish that people would not find themselves in situations where it seemed like the right answer. (It's not my body, and of that I have no doubt.)

Jason Weeden writes:

This is just getting weird. You explicitly said of individual issues that "If you know a person's position on one, you can predict his view on the rest to a surprising degree." And we ran a test of that proposition, showing that it can be pretty weak stuff, depending on the issue pair. (I've got a blog post showing similar points here: http://www.pleeps.org/2014/06/30/if-being-routinely-liberal-or-conservative-is-a-human-universal-why-is-it-true-only-of-recent-college-educated-whites/)

But now you're saying we should have known we were testing wrong thing! We shouldn't have tested the thing we quoted you as having said, but, instead (obviously!) we should have tested something else, namely, whether people's views on individual issues correlate with whether they call themselves "liberal" or "conservative." On this (different) question, I'd say: Of course they do (to different degrees for different issues); we never say that they don't.

To our minds, the more interesting questions about lib-con measures go deeper: (1) To what extent is calling oneself "liberal" vs. "conservative" typically a cause rather than an effect of one's issue attitudes? (2) Even though there are only two major labels people use these days, to what extent do various demographic groups nonetheless show patterns of combining liberal views on some issues and conservative views on others?

BC writes:

"If Weeden and Kurzban really wanted to dispute the one-dimensionality of political opinion, they should have been correlating specific issue views with ideology, not specific issue views with each other."

Or, more generally, perhaps they should have tried to find the principal components of the correlation matrix to identify the "ideology component", the component with the largest eigenvalue that identifies the mix of views that best defines ideology. It would be interesting to see what fraction of views is determined by the largest eigenvalue. I suspect that there are probably at least two significant eigenvalues (social and economic ideology) and possibly three.

Jason Weeden writes:
ScottA writes:

BC preempted my first thought on this.

A long second thought: not persuaded that we can't solve this with regressions. Weeden argues that running models predicting positions with ideology/income/various demographics is misleading, as we're not getting at whether ideology causes positions or vice-versa.

The solution is probably to find situations where ideology obviously existed prior to an issue position. Take cases where a new issue position became prominent, or, when a person knew nothing about an issue, then learned about it, and see which predicts the positions that emerge better: self-interest or existing ideology.

Related: not buying that it's enough to say that people obviously have a race/gender/education/income prior to picking ideology, so they come first in the causal chain. We can look at whether people change ideology as these variables change (this doesn't tend to happen for education/income in any systematic way). Race and gender are pretty fundamental, but why not run some multistage models predicting ideology with self-interest, then predicting issue-positions with predicted ideology and self-interest?

It's fair to defend these decisions, but it's also always fair to be skeptical of people who resist controlling for confounding factors that they know impact their results.

Comments for this entry have been closed
Return to top