About this Blog
About The Author
Email Me

RealClearPolitics HorseRaceBlog

By Jay Cost

« A Review of Obama's Voting Coalition, Part 1 | HorseRaceBlog Home Page | A Review of Obama's Voting Coalition, Part III »

A Review of Obama's Voting Coalition, Part II

At the conclusion of yesterday's essay, we noted an interesting trend in Obama's share among white males: it seems to depend in part upon socioeconomic status. That is, as socioeconomic status rises, so also does Obama's performance among voters in this demographic group.

We examined white males in detail because we had noticed that Obama has sometimes won this group, sometimes lost it. Perhaps the next question we should ask is: why must this trend be limited to white males? Could it not also apply to group that he consistently wins (e.g. young white voters) or consistently loses (women)? Of course it could. The margin of victory or defeat might vary systematically with socioeconomic status - even if the fact that he wins or loses does not.

Let's look at this in more detail. To start, we'll conduct a visual investigation of his performance among several key groups across regions. We'll examine Obama's performance among white women, white young voters (ages 18-29), and African Americans. Since we are interested in whether Obama's performance among them varies according to socioeconomic status, and we know that the white male vote varies by socioeconomic status, let's include Obama's performance among white males as a kind of control. Our expectation is that, if socioeonomic status is having an effect on his performance with white women, white youth, and African Americans - we will see his margins with them rise or fall as his margins with white males rises or falls.

We'll start east of the Mississippi River.

Obama's Performance Among Select Groups, East of Mississippi.gif

This indicates that white females and white youths seem to vary systematically with the white male vote. The difference is that Obama always does better with white youths than with white males, always does worse with white females than with white males. Still, the three groups move in tandem. As one rises, the others rise. As one falls, the others fall.

The same cannot be said for the African American vote. Clinton did relatively well with these voters in the northeast, but otherwise the African American vote seems to vary without regard to the other groups. Actually, it does not seem to vary much at all, instead remaining constant around 85% outside of the northeast.

Now, let's look west of the Mississippi.

Obama's Performance Among Select Groups, West of Mississippi.gif

Again, we see the same essential pattern among white groups in the west - although it is not as well defined. In particular, the Upper Midwest seems to be an exception. However, recall from yesterday that we have only one state (Missouri) in the Upper Midwest region with usable exit poll data. So, while the Upper Midwest does not seem to fit perfectly well, this is not really a problematic piece of evidence.

Once again we see that the African American vote seems largely unresponsive to whatever is altering the other three groups.

So, both east and west of the Mississippi River, we find that the white male, white female, and white youth vote vary together, while the African American vote does not follow along. We know from yesterday that the white male vote varies according to socioeconomic status. Thus, it stands to reason that the white female and white youth vote do as well.

We can confirm this, as we did yesterday, by regressing the white female vote and the white youth vote on our metrics for socioeconomic status, which are the number of voters earning more than $100,000 and the number of college educated voters. When we do this, we find that socioeconomic status is a statistically significant predictor of both the white female and white youth votes, though it does a better job for white youths.

Given what we saw in the above charts, our expectation is that socioeconomic status should have no effect on the African American vote. Indeed, statistical testing confirms that it probably does not. To date, the only significant factor affecting the African American vote that I have been able to identify is time. Obama has done better with these voters as time has passed.

What is the implication of this? It is that, among white voters, socioeconomic status permeates the Obama v. Clinton contest. It seems that one's inclination to vote for a candidate does not depend simply upon age and gender, but age and gender in the context of socioeconomic status. These factors interact with one another to produce (ultimately) a vote choice. White youth are more likely to vote for Obama than white women or men of all ages, but the particular likelihood that a white youth will vote for Obama also depends upon his or her socioeconomic status. Ditto white females. They are less likely to vote for Obama than white males or white youths, but the likelihood increases with socioeconomic status.

All in all, Obama's coalition seems to depend in large part upon African Americans, white youths, and upscale whites generally. The follow-up question is: just how much does his voting coalition depend upon these groups?

We cannot answer this question directly because the exit polls are simply not comprehensive enough. However, we can put together a rough estimate if we follow a simple, intuitive strategy.

First, we take the states that have useful exit polls. Our goal will be to determine what proportion of Obama's coalition in these states has been composed of the groups we just described - African Americans, white youths, and upscale whites. Our metrics for these groups should be as separate from each other as possible. For instance, we can easily count white youths and African Americans because these two groups are by definition independent of each other. The trick is that we believe that socioeconomic status is having a great effect among whites - but the exit polls do not sort socioeconomic status by race. Thus, we need to select a socioeconomic variable that overlaps with white youths and African Americans as little as possible. The best one available is probably voters who make more than $100,000. While it is obvious that some young whites and African Americans make more than $100,000 - in all likelihood most of these income earners are older than 30 and not black. Thus, we will have some overlap, but a tolerable amount.

So, this gives us a target. For each region, we'll develop an estimate of the proportion of Obama voters who are African American, whites under 30, or wealthy voters. Because we know they are wholly independent of each other, we'll combine white youths and African Americans into a single bar on our graphs. We'll use a solid line to separate the two groups, with white youths placed atop African Americans.

We'll against start east of the Mississippi.

Estimated Share of Obama's Vote Coalition East of Mississippi.gif

In New England, these groups probably fail to account for a majority of Obama's voting coalition - in large measure because his victories in Connecticut and Vermont were broadly based. However, outside of New England, these three groups account for a solid majority of his total vote. If we assume, for instance, that 3/4ths of voters who make more than $100,000 are not black and over the age of 30, we have accounted for anywhere between 55% and 82% of his total vote in the Mid-Atlantic, the South Atlantic, the North Central, and the South Central regions. If we assume that a larger proportion of $100k earners is white and over 30 - these estimates would increase. If we assume that a smaller proportion is white and over 30 - they would decrease.

Generally, we can conclude that for the part of the country that is east of the Mississippi River and south of New England - African Americans, young whites, and wealthy whites account for a sizable majority of Obama's total vote.

What about west of the Mississippi River?

Estimated Share of Obama's Vote Coalition West of Mississippi.gif

Clearly, things look very different here. Take the same assumption that 3/4ths of voters who make more than $100,000 are white and over the age of 30. In that scenario, these voters made up about 54% of his coalition in the Upper Midwest (i.e. Missouri), and 65% in the Lower Midwest. That is about what we found east of the Mississippi River. However, as we move farther west, these voters are less integral to Obama's coalition, making up just 2 out of every 5 Obama voters in both the Mountain West and the Pacific West.

What inferences can we draw from this?

I would suggest that, east of the Mississippi River, Obama's coalition has been relatively narrow. It has not been built upon a broad cross-section of the electorate, but rather the intense support of a few, loyal demographic groups. There are exceptions, of course. But by and large, these three groups account for a very large segment of his voting coalition. When we go west, we see that matters are different. Obama's coalition is somewhat broader - drawing less heavily upon these groups.

In a sense, he has been like two different candidates. In the west, his candidacy has been broadly based and relatively diverse (though he has lagged behind with Hispanics). In the east, it has been more narrow, largely failing to build a real cross-section of the electorate, at least outside New England.

The next question we might ask is how Obama has performed in the parts of the country that have swung the last three presidential elections? We'll investigate this tomorrow.

-Jay Cost