In the context of a discussion of rich and poor voters in the U.S. and other countries, Matthew Yglesias posted this graph from our Red State, Blue State book:

The commenters raised several issues that I’d like to clarify here. (In particular, it looks like we miscoded some of the GDP per capita numbers, which doesn’t affect our conclusions but is a bit embarrassing.)

1. The meaning of the graph

In most countries for which we have data, richer voters are more likely than poor voters to vote for conservative parties. The gap is usually less than 10%, though, meaning that voters in the upper third of income vote about 10 percentage points more on the conservative side than do voters in the lower third of income.

In the U.S. for the particular election included in this dataset, the Republicans did about 10 percentage points better among the voters in the upper third than among the voters in the lower third of income.

The rich-poor voting gap in the U.S. was larger than about 80% of the countries in our dataset and in particular was as large or large than just about all the countries in western Europe.

This is interesting because we often think of European politics as having strong left-right divisions. Actually, though, in most European countries, the left and right groupings aren’t far apart on economic issues. This is a point made more thoroughly by John Huber and Piero Stanig in their article, “Why do the poor support right-wing parties? A cross-national analysis.”

Some people commented that some of the rich-poor difference in the U.S. is also a white-black difference. That’s true, but recall that other countries have ethnic divisions too. In a European country, there can be big ethnic differences between people who would all be classified as “white” in the U.S. (This is not to say that we shouldn’t look at ethnic breakdowns of votes, just that it’s not necessarily appropriate to compare whites in the U.S. with whites in Europe.)

2. Presentation

Some of Yglesias’s commenters were unhappy with the presentation of the graph. I think the main complaint is that the graph and its caption aren’t completely self-contained. In the book where the graph appears, we include explanatory material in the Notes section at the end of the book so as not to clutter the main narrative. The graph is on page 106, and turning to the Notes for page 106, we see: “Within each country, we collapse the five income categories of the Comparative Study of Electoral Systems into three and then take the difference of conservative vote in the highest and lowest of the three income categories.”

I have to admit, though, that we didn’t present the details of which parties we classified as “conservative,” and, in fact, is was my coauthor, Jeronimo Cortina, who did this. What I can tell you here is that we categorized each party in each country as left, right, or neither, and what we called the conservative vote is simply the aggregate for all the parties on the right within any country. Thus, “right” includes center-right. Our goal in making this graph was to compare the other countries to the U.S., and so we tried as best we could to break up the votes into two large groupings corresponding roughly to the two large groupings we have in our country.

One commenter writes: “This graph represents everything I hate about political ‘science.’ It does not appear to correct or account for . . .” I’d like to respond with something sarcastic here, but actually I do want to communicate with people–that’s why I emailed the graph to Matt in the first place–and so I’d like to defend myself by saying that, yes, this graph is not in itself a complete analysis, but it can be helpful, I think, to lay out the data clearly. Before I saw this graph (or, more precisely, before I saw Huber and Stanig’s graphs upon which ours were derived), I just assumed that European countries had big rich-poor divides in voting. But, no, that’s not the case. I learned something, and Jeronimo and I made this graph in order to convey this information to others. If, upon seeing this graph, you’d like to know more, and to account for differences between countries in other ways, go for it! The graph is intended to be a starting point (and, more specifically, to illuminate Matt’s earlier discussion of income and voting behavior).

3. Data issues

As noted above, our definitions of “rich” and “poor” weren’t easily available to readers of the graph. Beyond this, there seems to be some problem with our GDP per capita axis, as one of the commenters noted. In particular, several of the eastern European countries have GDP per capita values that are much higher than shown on the graph.

For example, we have Russia’s per-capita GDP at $2600 per capita, but according to the World Bank it was $7560 in 2007 (the most recent year for which we had data, as the book came out in 2008), or $14400 with purchasing power adjustment. Clearly there’s something wrong here.

First off, I want to say that I don’t think this problem affects our substantive conclusion–all it does is move the points horizontally on the graph. That said, we’d like to figure out what’s going on.

I went back and looked at our data, and I’m not quite sure where all our numbers came from. Here’s a portion of our dataset:

            Country WB06GDPPerCap
1         Australia         23372
2           Belgium         24541
3            Brazil          4055
. . .
27          Romania          2443
28           Russia          2621
. . .
33          Ukraine          1040
34   United Kingdom         27582
35    United States         38165

So it looks like we thought these were World Bank 2006 numbers. But when I went to the World Bank site, I didn’t see anything like this.

Then we have another dataset with what appears to be IMF numbers. This other dataset includes some but not all the countries listed above:

31         Brazil      4055       6841.600
19         Russia      2621       8611.672
21        Ukraine      1040       2829.701
36        Britain     27582      45301.059
28  United States     38165      45593.855

A bunch of things appear to be going on here, some of which involve changes in the exchange rates between the dollar and European currencies. But I can’t figure out what was going on with those low figures in Russia, Ukraine, and the other countries of Eastern Europe. We’ll have to track this down; even though it won’t make any difference in our findings–we were merely using per-capita GDP as a convenient x-axis for plotting the data–it’s still embarrassing to make a mistake. So I thank Matt’s commenters for their eagle eyes (even while I respectfully disagree with their position that “the author of the book wanted to prove a theory he had so he slapped together a crappy graph” . . .

The funny thing is, I remember that when we were putting together the book we spent some time staring at these GDP numbers trying to figure out which ones to use. Maybe Jeronimo will remember what’s going on here.

P.S. Mitch remembered some things; see here.