John Huber and Piero Stanig will be speaking on this paper tomorrow in the comparative politics seminar. This is important stuff. From the paper:

Why do poor people, who would benefit from redistribution, often vote for parties that oppose it? Democratic elections can serve to diminish inequality by enabling a country’s poorest members to receive resources from wealthier ones. The poor can obtain these resources by helping to elect parties that pledge to tax the rich and transfer the proceeds to the poor. But if the poor do tenot vote for parties that are committed to redistribution, elections are unlikely to play this role of raising the economic well-being of the poor. Why, then, would some poor individuals use their vote to advance the electoral success of right-wing parties that oppose redistribution?

Both income and religious attendance are predictive of voting for conservative parties (in the context of this paper, parties that oppose redistributive policies):

The average proportion of right-wing support increases as income increases, regardless of whether we include all voters or classify voters by religiosity. And the average proportion of right-wing support is higher among religious than among secular voters, regardless of whether we include all voters or classify by income. . . . But one interesting interaction emerges from these summaries of aggregate proportions. When we consider income polarization between the middle class and the poor, it is higher among religious than secular voters.

This is consistent with what we found in the U.S.:

But John and Piero go beyond our descriptive analysis by studying how factors of the political system are predictive of stronger or weaker links between income and voting in a country. They find:

The poor have a greater propensity to support right-wing parties in countries that are ethnically heterogeneous, rich, low in urbanization, low in party-system polarization, and that have no parties that are at once left wing on redistribution and right wing on issues related to individual liberty. . . . Income-based voting polarization increases when countries are ethnically homogenous and urbanized, when voters can choose to vote for a left-wing redistributive party that is conservative on individual liberty issues, and when there is no religious tax.

Finally, to get back to the U.S., they say:

Voters in the U.S. are only slightly above average in their level of support for right-wing parties. Three factors should drive U.S. voters to the right: the relatively high level of ethnic heterogeneity, the U.S.’s wealth, and the “forced choice” feature of the party system. Working against these variables, however, is the very high level of party-system polarization, which drives poor and middle-class voters to the left, and the relatively high level of urbanization, which drives poor voters to the left. . . . Although the level of right-wing support is relatively average . . . voting polarization on income is much higher in the U.S. that in most other countries. That is, poor voters are relatively more likely than middle-class voters to support left-wing parties in the U.S. than in other countries.

One thing I’d like to understand is the differences in polarizations in different U.S. states (or, for that matter, different Mexican states). In presidential elections, everybody is voting on the same issues, so the institutional explanations of this Huber and Stanig paper wouldn’t seem to directly answer the question.

Finally, some picky things:

1. I’d be a little more careful about the causal language, for example, “the effect of income” and “the impact of income” on page 4. To the extent that income is an attribute rather than a treatment, the relation between income and voting isn’t necessarily causal.

2. All the tables should be graphs. And when countries are listed, I’d prefer to use some ordering such as avg income or avg religiosity, not alphabetical.

3. Fig 1 should show the country names. (Yes, you can do it, if necessary by breaking into different graphs for different regions of the world.) Also, the x and y-axes should stop at 0 (rather than following the R default and going below zero).

4. Figure 2 shouldn’t be in alphabetical order. Actually, I don’t think Figure 2 has much information, so I’d remove it. Figure 3 should also not be alphabetical, also I think it could be done better. In particular, income polarization should bnot be plotted on the same scale as vote proportions. And something wacky is going on with Belgium and some other countries with the placement of the x-axis. Also, use the full country names rather than tricky 3-letter all-caps abbreviations. Ink is basically free here.

5. FIgs 3,4,5 can be displayed, for example, as a long array of countries, as a 28×3 array, so that each row shows everything that’s going on with a country. I’d really like to see these graphs cleaned up, because to me, they’re the meat of the paper. I’d start by replacing all the barplots with line plots, which will allow multiple lines on a graph. Our own graphs reproduced above (for the 3 states) could be a template here.

6. Figure 6 is pretty good,. I’d put fewer labels on the axes (the y-axis could be just 0, .5, 1), get rid of the horizontal lines (but, yes, given that they’re there, it’s nice that they’re so light), and, again, don’t have the graphs go below 0. I’d also label the two lines directly on each graph, rather than use the key at the bottom of the page.

7. In Figure 7, again, label the lines directly on each graph. This is a bit of work, but once you’ve figured out what you want in the final paper, it would be much easier to follow (as a researcher, not just as a “consumer” of the paper) if the lines are labeled (“secular poor”, “secular middle”, etc.)