Are the elites more polarized? Yes!

One of the argument from Andrew Gelman's Red State, Blue State, Rich State, Poor State which has percolated into the punditocracy is that the Culture Wars are to a large extent a feature of the upper socioeconomic brackets. Gelman presents data which strongly contradicts Thomas Frank's argument that the less wealthy are voting based on cultural issues as opposed to their economic self-interest. Rather, it seems that the wealthier are voting on social issues because at a particular level of affluence economic concerns have less salience. Mainstream pundits such as Matt Yglesias and Matt Continetti have been repeating Gelman's claim to the point where I think we might finally be able to bury Frankism.

Since Andrew Gelman wrote a whole book on the topic, I assume he knows what he's talking about (he also reports a lot of the data and analysis on his weblog). But I decided to look at political ideology as a function of socioeconomic status in the GSS. Specifically, I broke down the "SEI" (socioeconomic index) into quartiles. I compared it to the "POLVIEWS" variable, which measures political ideology on a 1-7 scale, from extremely liberal, to extremely conservative, with a 4 being moderate. Limiting the sample to whites here is the chart of mean political ideology as a function of socioeconomic quartile (lowest to highest, left to right):

i-06afd05d5dfb4dce770b4efb93dc6a65-richpol.jpg

Though there is a slight shift toward more conservatism in the middle socioeconomic quartiles, it is minimal. On the whole the mean value of political orientation remains pretty much the same across the socioeconomic spectrum. The top 25th are are neither limousine liberals nor conservative fatcats, or, they are, but to an equally greater extent.

Means only give you so much. You want to know how things are distributed across the spectrum, so now let's break down by all 7 political categories:
i-9a9444d1ff9550e5be196f1bf77dc7bb-richvariance.jpg

Now on the Y axis you have the percentages, while on the X axis you still have the socioeconomic quartiles. Though those on the political extremes (those who cop to being an extreme liberal or conservative) don't gain much the socioeconomic scale, but moderation does yield to liberalism and conservatism.

Tags

More like this

If there is one "politics" book you should read this year, it is Red State, Blue State, Rich State, Poor State: Why Americans Vote the Way They Do. Now, this sort of acclamation does need to be tempered by the fact that I myself don't really read "political" books very often. But despite the…
I made a comment to the effect that political moderation tends to correlate with lower intelligence when looking over survey data on attitudes toward unpopular speech. This assertion didn't come out of a vacuum, when mining the GSS there were repeated indications that moderates were somewhat…
Andrew Gelman has a post up, Who are the liberal Democrats and the conservative Republicans?, which shows that conservative Republicans tend toward higher incomes, while conservative Democrats tend toward lower incomes. I decided to see if something similar was discernible in the General Social…
During a conversation with Nick Matzke he asserted that Creationists weren't less intelligent necessarily. I contended that they were less intelligent. I based on this on snooping through the GSS when I was posting about the association between lower educational attainment and intelligence and…

If you thickened the lines a bit the colors would be more evident to people with weak eyes.

By John Emerson (not verified) on 04 Oct 2008 #permalink

Sorry, I meant the lines in the key, not the chart.

By John J Emerson (not verified) on 04 Oct 2008 #permalink

Just FYI, in general when dealing with ordinal variables you don't want to do arithmetic on them. That is, the only thing one knows about level 6 is that it's greater than level 5 -- there aren't 6 physical units of anything.

Which is a longer way of saying that the second graph should be preferred (and ordinal data should not be averaged or used to compute variances, etc.).

That is, the only thing one knows about level 6 is that it's greater than level 5 -- there aren't 6 physical units of anything.

yeah, makes sense.

I find the chart very illuminating. You might as well merge "slight conservative" and "conservative" and "slight liberal" and "liberal". You wouldn't merge the two "verys", even though their numbers are similar. So you'd end up with five lines, extremism constant, moderation declining with income, and both liberalism and conservatism rising.

The high "moderation" of low-income voters tends to make me think that what's measured is low-information voters instead, people don't take an active interest and who answer in harmless generalities. And maybe people who feel totally helpless, as though their input cannot make any difference.

By John Emerson (not verified) on 05 Oct 2008 #permalink

Another way looking at the elite's voting on social issues is that at a certain level of affluence one is insulated from the unpleasant consequences of the policies one has voted for.

What might make it difficult to make a comparison is the shifting nature of what is considered conservativism. Some might say the mainstream veiws of a generation ago are generally considered extreme today.

Elite WASPs are said to be big on enviromental protection is this liberal social policy or economic self -interest in some way.

The high "moderation" of low-income voters tends to make me think that what's measured is low-information voters instead, people don't take an active interest and who answer in harmless generalities.

well, i looked at a few issues, and in general, yes. OTOH, i've seen stuff which shows that low SES voters put more weight to "how well is my family doing economically" type of questions and vote for the party which might make them the better offer.