Today the renowned political scientist Arthur Lupia visited GMU.  His mission: Attack the quality of academic research on voter competence.  His arguments changed my mind, but in the opposite of the intended direction.  Given Lupia’s intelligence, expertise, and effort, his presentation was shockingly weak.  Highlights:

1. Lupia spent about half of his talk arguing that ability to correctly answer standard political knowledge questions is neither necessary nor sufficient for voter competence.  For example, voters who know nothing about politicians’ voting records might “vote correctly” (i.e., vote “as if” they were fully informed) simply by voting on party lines.

My reply: Every serious researcher on this topic has long known this.  They almost never talk about knowledge that voters must have to “vote correctly.”  They almost never talk about knowledge that guarantees that voters will vote correctly.  Instead, they talk about knowledge that raises the probability of voting correctly. 

2. Lupia complains that voter competence research ignores the crucial question: Would voters actually change their positions if they gained additional knowledge?  Again, this claim is bizarre.  This is the whole point of the massive empirical political science literature on “enlightened preferences.”  Scott Althaus has written an exhaustive survey of research that estimates how much knowledge affects policy preferences.  The punchline is that knowledge has big systematic effects on policy preferences – and by any plausible standard does indeed raise the probability of voting correctly.

3. When pressed, Lupia is of course well-aware of the enlightened preference literature, but it
was hard to get him to explain why it was not a satisfactory answer to
his complaints.  Eventually, though, he started airing some of the dirty laundry of political knowledge research.  Since he’s the Principal Investigator of the American National Election Studies, Lupia has many entertaining anecdotes of sloppy scholarship.  It turns out, for example, that free response questions are often poorly scored.  Take the question, “Who is Tony Blair?”  Respondents were marked wrong for answering “The Prime Minister of the United Kingdom.”

This is all useful to know.  Higher-quality data is a good thing.  Yet Lupia neglects a crucial statistical point: Higher measurement error leads to attenuation bias.  The noisier your measure of political knowledge, the more likely you are to falsely conclude that political knowledge doesn’t matter.  The fact that researchers find large knowledge effects with low-quality measures implies that the true effect is probably even larger.

As a good Bayesian, I try not to overuse the cliche, “I’m even more convinced of my position than I was before.”  Lupia’s talk leaves me no choice but to state the cliche.  Most of his arguments attacked views that no serious researcher believes.  The rest actually showed that his critics are understating their case. 

O ye defenders of voter competence – is this all you’ve got?!