Boris pointed me to this paper by Edward Glaeser, Giacomo Ponzetto, and Jesse Shapiro. Here’s the abstract:

Party platforms differ sharply from one another, especially on issues with religious content, such as abortion or gay marriage. Religious extremism in the U.S. appears to be strategically targeted to win elections, since party platforms diverge significantly, while policy outcomes like abortion rates are not affected by changes in the governing party. Given the high returns from attracting the median voter, why do vote-maximizing politicians veer off into extremism? In this paper, we find that strategic extremism depends on an important intensive margin where politicians want to induce their core constituents to vote (or make donations) and the ability to target political messages towards those core constituents. Our model predicts that the political relevance of religious issues is highest when around one-half of the voting population attends church regularly. Using data from across the world and within the U.S., we indeed find a non-monotonic relationship between religious extremism and religious attendance.

And here are my thoughts:

1. First off, to refer back to my own recent research with Jeff Cai, it’s not necessarily optimal for politicians to move toward the median–even setting aside issues of core consituents–if voters observe candidate positions with error. In that case, it can make sense for a party that is more centrist (in recent U.S. political terms, that would be the Democrats on economic policy, and the Republicans on social policy) to keep some daylight between themselves and the opposition, so that voters clearly see the distinction. This can be modeled theoretically and also be seen in opinion poll data, where voters do vary quite a bit in their perceptions of candidates’ positions.

Thus, I don’t quite agree with the last sentence on page 2 of their paper (that continues on to page 3), since it assumes that voters agree on where the candidates stand and measure these positions with no error.

2. Getting back to the Glaeser, Ponzetto, and Shapiro paper, the next step in their analysis of income, religious attendance, and voting is to look at the interaction between income and religious attendance as it predicts vote preference. As we’ve found, the two factors interact highly, with income being a much stronger predictor of Republican voting among religious attenders (and, conversely, religious attendance being a much stronger predictor of Republican voting among high-income people). See here.

In particular, Figure 4 is not so interpretable, given that in recent years, the interaction is as high as the main effects. (Actually, I’m surprised they stopped the graph at 1992. It really starts to get interesting after that.) Also, since I’m on Figure 4, I’d suggest a cleaner y-axis (for example, labels at 0, 0.1, and 0.2)

3. They also allude to the red-state, blue-state stuff (see Figure 3), but I don’t really see how it fits into their story.

4. The actual model is pretty complicated, but I guess I follow the basic idea that politicians should be able to better convey their stances to their supporters than to the other side. Still, I wonder whether (at least in the U.S. case) the story ultimately comes down to more specific circumstances, involving the Republicans’ increasing strength in the South since 1950 (arising from racial and, more generally, social issues more than religion, I’d think), which then put them in a position of representing a bunch of voters who were more strongly religious. And then some other story involving the connection of organized religion to abortion and other issues.

The comparative analysis is intriguing partly because it sidesteps these U.S.-specific concerns. And I certainly agree that there’s room for some good explanations of why the parties aren’t hugging the median voter (which, as the paper points out, they’re not doing on economic issues either). The summary on page 27 is interesting:

Economic policy divergence is accompanied by platform uniformity. This suggests that economic extremism is driven by political preferences not strategic considerations. Conversely, divergence of platforms on abortion is accompanied by no observable difference in abortion rates during the years of Democratic and Republican presidents. This combination is consistent with the view that abortion extremism is driven by strategic considerations, not political preferences.

I wonder whether part of this is that it’s harder to alter national policy on abortion (short of Supreme Court appointments, which changes things slowly). In contrast, economic policy has lots of levers.

5. On page 37-38, they write that the U.S. coefficient of income on “right-wing status” is “quite low” compared to other countries. But Huber and Stanig find the U.S. coefficient to be relatively high! What’s going on here??

6. I don’t know if I buy Section 5.3, the state-year analysis. The trouble is that things changed a lot in the 1990s (see our red-state, blue-state paper, for example), so I’d rather see a separate analysis in each year.

7. The conclusion of the paper states, “if geography creates a natural ability for politicians to target their messages, i.e. if people in your own area are more aware of your policies than people elsewhere, then national systems are always at risk of politicians proposing extreme regional policies. One way to counter this tendency is to have voting systems, like the U.S. electoral college, that limit the value of getting extra votes in one particular geographic area.” Maybe so, but this seems like a funny thing to say given the historically (and continuing) highly geographical nature of U.S. politics. One might argue the opposite, that the focus of attention on a few “swing states” might tend to reduce parties’ efforts to compete in each others’ home regions.

8. I’d like the tables to be graphs. Or, if not that, please do some rounding, for example 0.4829 +/- 0.1383 is really 0.48 +/- 0.14. (Actually, 0.5 +/- 0.1 is fine, but I know that bothers some people.)

In summary, this is an interesting study full of data and theories. I can’t say I buy all the theories here; on the other hand I don’t really have anything better to offer, so this seems to be a good place to start!