Vox has a story that is primarily about the voting habits of the wealthy, but talks briefly of education:

There are a few things we know have become very strong predictors of voting Democratic. One is that nonwhite people tend to support Democrats at higher rates than white people do. Another is that the highly educated have become much more liberalover time, making educational attainment a better predictor of voting for a Democrat.

And over time, the top 4 percent has become much more diverse and much more highly educated.

In the context of an article about the wealthy veering towards the Democratic Party, one could easily read this paragraph and conclude that one of the reasons they are increasingly Democratic is that they are more educated, and educated people tend to vote Democrat. You could come to that conclusion because that’s almost exactly what the article says. Almost. It actually makes two comments pertaining to education. First, that the highly educated are becoming liberal. Second, that educational attainment is a predictor for voting Democrat.

Both of these are more-or-less true. Higher education tends to associate with social liberalism, for example. And educational attainment is a predictor for both Democrats and Republicans. Just not in the way that the article implies. Nor in the way that either link lays it out, either. Because the obvious implication of those two statements, that more education generally leads to Democratic voting habits, isn’t especially true unless you look at the data in very specific ways, or look only at one subset of the population.

educatedvoteThe 2012 data is to the right, and previous years shown below it. You will notice that the one and only group that Mitt Romney wins is college graduates with no post-graduate degrees. This is actually really quite common result. With the exception of 2004, college graduates were the top educational group of Republicans going back all the way to 1988. In 2004, George W Bush did best among “some college” at 54% and then tied among “high school diploma” and “college graduate.” at 52%. For the most part, however, the data to the right is indicative. What it shows is that Democrats do best among the minimally educated and the maximally educated, and Republicans do best in the Great Middle, some college or a bachelor’s degree*.

educatedvotepastSo why does the media continue to portray education as having a linear relationship with education when that relationship simply doesn’t exist? There are two explanations that come to mind, neither of which are especially flattering, but both of which are more flattering than “They partisan hacks are trying to trick you!!!!”

Reporters, on some level, simply believe it to be true. So when they look at something like education level and partisan alignment, the data that’s going to jump at them is the data that confirms their impressions. And when they come across data that confirms their impressions, they’re more likely to believe that ought to be a story. A data that doesn’t confirm their impressions won’t necessarily be dismissed as false, but is more likely to be seen as “complicated” rather than a nice, clean narrative like “We’re the educated ones.”

And yes, I do mean “we.” This is an area where it matters that journalists and editors overwhelmingly lean in a particular direction. I mean, do you know how I know that educational attainment doesn’t line up in a linear fashion? Because I took the time to look it up. Why did I take the time to look it up? Because that sounded like an overly simplistic dynamic that didn’t quite correspond with my impressions. There’s that word again: impressions. They matter. Regardless of why newsrooms lean in the direction that they do, it’s not optimal that you have a lot of shared impressions.

In addition to impressions, there is some data. Not the general data, which states what I point out above, but selective data. The data and anecdata that jumps out at you when it confirms your biases. There are three data points that can confirm biases. First, there is a correllation between education level and social liberalty. It just doesn’t cut as cleanly along partisan lines as you might think. More educated Republicans tend to be social liberals, less educated Democrats tend to be socially conservative. Think Country Club Republicans and The Minority Vote.

Second, which is tied to the first, is that the correllation is true among whites. Which makes it sound like “it’s true if you control for race” but it’s not because the opposite holds true for blacks and Hispanics (the educated ones vote slightly less Democratic). And the data for all of this is relatively murky in any event. Too murky to provide any strong narrative, to be honest.

Third, the relationship is accurate if you simply pretend people who don’t go to college don’t exist. If we talk about “more” as getting a postgraduate degree, and “less educated” as merely having a bachelor’s, well there you go. Democrats are more educated. You’ve tossed out half of the electorate and more than half of the population, but those aren’t the people you know, are they?

Which, in addition to confirmation bias, is the source of at least some of this. And is concerning, in its own way. Newsrooms are disproportionately white, and they’re disproportionately educated. The country is (right now) majority white. So when they think of more educated and less educated, they’re likely to be thinking (consciously or unconsciously) “white.” And when they think about who is more educated and less educated, in their world having a postgraduate degree, or a degree from an elite university, does, confirm the bias. Their reality has a liberal bias. I just don’t think they like the omissions that necessarily get them there.

After 2016, this may all be moot, at least for a cycle. It would not surprise me if Hillary did unusually well among those with college degrees and Trump did more poorly. And this could be where the parties are ultimately headed. But they’re not there yet and the trends do not suggest it’s inevitable.

* – This is, it should be pointed out, increasingly what things are looking like with regard to income, which is what the Vox article touches on. Democrats win among the poor, Republicans do best among the middle rich, and Democrats do better and better among the wealthy.


Category: Statehouse

About the Author


3 Responses to The Persistence of Impressions: Confirmation Bias vs Analysis

  1. greginak says:

    An obvious third reason for poor reporting on stuff with stats is that people in general, and this counts reporters, don’t’ know squat about stats.

    When i’m dictator at least a half of year of stats will be part of a HS education.

    • Michael Cain says:

      Maybe. Will that half year get to the point where things more complicated than linear models are included? When I went back to graduate school to get my masters in public policy, they didn’t make me take the mandatory stats class. Damned good thing, as a number of the other students in my group that I coached through the class kept asking why the hell I wasn’t teaching it instead of the joker that was. When I had a chance, I bitched and moaned that the program was producing a bunch of very, very dangerous people who thought things stopped at simple linear models.

  2. aaron david says:

    excellent article Will.

Leave a Reply

Your email address will not be published. Required fields are marked *

If you are interested in subscribing to new post notifications,
please enter your email address on this page.