On the Median Voter
A David Brooks column inspires some thoughts.
David Brooks had a column the other day about the mid-term results versus recent polling concerning Trump v. Biden. It is entitled Democrats: You Can Chill Out Now! I am not going to comment on the whole thing, but the following leaped out at me (such is the way in which the clearing of tabs can take one into a lot of words).
So, here we go. From Brooks:
The median voter rule still applies. The median voter rule says parties win when they stay close to the center of the electorate. It’s one of the most boring rules in all of politics, and sometimes people on the left and the right pretend they can ignore it, but they usually end up paying a price.
The Democrats’ strong showing in elections across the country this week proves how powerful the median voter rule is, especially when it comes to the abortion issue.
Ok, there is actually a lot to unpack here and there is no way to cover it all in one post, but let me note something that I think is really important: the median voter “rule” (really, what is usually called the median voter theorem–although it is really isn’t a theorem, but more a model or theory) is highly dependent on the distribution of voters, and therefore to what the median of that distribution is*. Journalists, and often people in general who talk about such things, tend to speak as if we always have a normal distribution of voters in terms of ideology (so that the centrists are the big peak in the curve) and that is how we should understand election outcomes.
Basically, the simple model looks something like the following and assumes a vast middle with the true left and right being far smaller. Yes, this is a simplified model but it should give everyone an idea of what I am talking about. The higher the curve, the more voters at the given ideological point under that part of the curve. In this distribution there are a lot of centrists and, indeed, the center of the distribution (the median, wherein half the voters are to the right of that point and half to the left) is the Most Centrist of Centrists (which the space at which many, many, many pundits pretend like they exist and the position from which many a column pining for a Third Party Savior has been based). Logically, a far-left extremist would lose to a moderate, or even a conservative, and vice versa.
So, sure in a distribution like that, the candidate who is more centrist out of two is likely to win. And assuming that the US electorate looks like this is why columnists and third-party activists pretend like the way forward is centrism. This was the logic of Ross Perot and, to a degree, people like Joe Manchin and his No Labels (yes, it’s a label) buddies.
And frequently this is the way that people talk about American democracy as if it is just a big relatively fair contest and the competition is in (or should be in) the middle.
However, one of the various reasons that I argue that the US has a democratic deficit is because, no, it doesn’t work that way.
First, if we think about the US presidential race, the reason (well, one of them) that the Electoral College is a problem is that it skews the outcome away from whatever the median preference of all of the voters is. A national popular voter ought, in fact, to drive electoral competition to roughly wherever the actual median of the electorate is (which is not the perfect bell curve above, but for the sake of argument I will say that it isn’t radically different).** But, instead, the process drives the competition to a limited number of competitive states.
I would note that one of the reasons that national polling for the president is a problem is because, as we all know, there are actually 56 individual contests (50 states plus DC plus the two Maine and three Nebraska district-level elections). The issue becomes, therefore, not what the national median voter is, but what the median voters are in each of those elections. (By extension this is true in the 50 Senate races and 435 House races).
Second, this is why single-seat districts are a problem (which is made worse by gerrymandering). The median voter in state X might in fact be more moderate, but when you can slice state X into, say, seven segments, you can make damn sure to skew the outcome as far away from the state median towards your party as possible. And even if the state’s districts are drawn in a non-partisan way, they are very likely to skew away from the state’s overall median preference because of population distributions.
Further, most US congressional districts would have curves that would peak either substantially farther to the left or to the right than the curve above.
I would note that both the EC and the usage of single-seat districts result in where lines are drawn ending up to be more important than the actual democratic preferences of the voters. This is why they diminish democratic quality and representativeness. Again, the location of the lines ends up being more important than the broader distribution of voter preferences.
Third, I would add that primaries further skew the issue of whether the system is really driving competition to the median. The median primary voter almost certainly has a different ideological profile than the median co-partisan at the district level. But, the median district-level voter is going to vote for their party’s nominee because they are closer ideologically to them than they are to their opponent’s nominee (or, at least, their partisan ID tells them this is the case even if their district is a case of an MTG-like GOP candidate and moderate Democrat).
On that last point, I would note that partisan identity itself as a general matter influences how voters perceive candidates, meaning that it is possible that a given candidate might actually be closer to the median ideologically in a given district but that voter perception is swayed by party labels and the signals they are perceived to be sending. This is a large part of the reason why even if, say, in a heavily Republican district a moderate or even center-right Democrat who might be closer to the ideological median of the district is going to lose, despite pundit (and sometimes by OTB commenter) fantasies to the contrary.
To Brook’s point, sure, it is clear that the median position in Ohio is pro-choice, but this isn’t a good illustration of the median voter concept, because a referendum of this type is about voter preferences on one single issue, not a guess by voters of whether candidate X or Y is closer to the ideological median of a given district. Not only is that just one issue, which makes it easier to determine where median preferences lie, but the mechanism of a simple yes/no choice on a statewide ballot definitionally finds the median preference (more or less)–at least it identifies roughly where the median preference is on the exact question being asked.
Indeed, if the statewide contest for, say, Senate, had been a pro-choice Democrat and a pro-life Republican, the outcome of that contest would have been a bit different, even if abortion was THE topic of that race (but it still would only but one variable in a far more complex stew of issues and views). That illustrates in simple terms the difference between a single-issue choice and a race between candidates even with the exact same pool of voters.****
After all, Ohio was roughly +8 R in 2020 and J.D. Vance won his Senate seat in 2022 by +6.1. The abortion reference passed +13.2. As such, what this likely tells us is a lot more about abortion than it does about the appropriate application of the “median voter rule.” (Or at, least, the potential limitations of using it to find political clarity).
A side note to conclude, which opens up another can of worms. It is worth considering the degree to which we should be concerned with just having the median views of the country represented, or whether we should prefer a system that proportionally represents the entire distribution of voters. The entire discussion above more or less assumes only two choices and either plurality or majority decision rules.
*If you are at a party and 40% of the people like pepperoni pizza, 20% like sausage, 20% like a combo of both, 10% want Canadian Bacon, and 10% are vegetarian and there is a vote, it sure looks pepperoni to me. But if the distribution is 70% vegetarian, 20% pepperoni, and 10% “can’t we have hamburgers” the vote is going to be vastly different.
**This is where I note that I am writing this for funsies on a weekend and am not writing a real academic treatise. Were I doing that the figures would be much better!
***And I will risk being accused by some readers of being too pie in the sky and note that an electoral system with proportional representation, even modest PR, could solve this. This is doable for the House by legislation, depending on the system deployed. Yes, it is a political non-starter at the moment, so I probably should not even mention it!
****Note I said “pool of voters” and not just “voters” since turnout affects who the voters are, even if they all come from the same pool.