Moral Bias in Large Elections: Theory and Experimental Evidence

This is a review of Moral Bias in Large Elections: Theory and Experimental Evidence (2009) by Timothy Feddersen, Sean Gailmard, and Alvaro Sandroni. American Political Science Review 103 (May): 175-192. You can find the original in Google Scholar.
Howard Dean with a confederate flag on his truck

Howard Dean and his truck (kidding, kidding)

In late 2003, Howard Dean lamented that southern white guys with confederate flags on their trucks ought to be voting for Democrats1; after all, it’s the Democrats who want to help the working classes. Folks like Dean think that these southern white guys are being duped by wealthy upper-crust Republicans, who trick the southerners into voting for them by tossing them just enough symbolic “red meat” (pro-life, anti-gay, anti-immigrant) to keep them under control.

I just read Feddersen et al’s article, in which the authors attempt to explain exactly why this “red meat” tactic works–that is, why southern evangelicals keep casting a “values vote” for Republicans instead of casting a materially rational vote for Democrats. Okay, in fairness, the authors don’t come out and say it that way, but it sure sounds like that’s what they’re trying to explain. As such, this study is loaded with potential controversy.

I’ll say that again. Feddersen and his colleagues didn’t say a word about Dean or confederate flags. What they authors really say in their analysis is something closer to this:

Some key terms

First, people have instrumental preferences. That is, they recognize that option A will make them materially better off than option B. (Analogy: Working-class southern whites recognize that they would be materially better off with Democrats in power.) These instrumental preferences lead them to prefer option A over option B. Folks know that their personal situation (employment, finances, health care, whatever) will improve under option A.

But second, people also have moral preferences. That is, they recognize option B as the more moral choice than option A.

But here’s a key point: These moral preferences are just “expressive” preferences, not “instrumental” preferences. What’s that mean? People recognize that their personal situation won’t change at all depending on whether option B wins or not; they know that they won’t be visiting an abortion clinic regardless of whether abortion is legal. Thus, these moral choices aren’t “instrumental” choices, since nothing in their life changes either way. So they just want to vote for option B to “express” themselves as being a moral person.2

The point

So, the research question for Feddersen et al is this: Why do people choose to cast an “expressive” (moral) vote for option B rather than an “instrumental” (material, rational) vote for option A? Or, as Dean would put it, why do those southern whites keep voting Republican (a “values vote”) instead of for Democrats (for better health care, Social Security, etc)? Briefly:

As the number of voters grows, an individual voter’s probability of being decisive falls. If there are 11 voters, each voter has a 1/11 “pivot probability”; if there are 101 voters, each one has a 1/101 “pivot probability.” In most American elections, the pivot probability is so tiny that it’s effectively zero.

When voters consider the benefit of casting an instrumental vote (that is, voting for A), they weight the material payoff of doing so against this pivot probability. That is, if I think that I’ll be materially better of with A than with B, but I also know that I’m only one of 100,000,000 voters, then I recognize that my individual vote has little bearing on whether I wind up with A or B in the end. Thus, when the electorate is large, voters don’t find instrumental preferences compelling.

However, the “expressive” payoff of choosing the moral option (option B) does not get reduced by my pivot probability. Regardless of whether A or B wins, I can feel better about myself if I know that I cast the ethically/morally appropriate vote.

Thus, when the electorate is larger than a handful of people (and it always is), we should expect moral preferences to trump instrumental preferences.

The point again, stated differently

When voting for A or B, you get two kinds of “reward” for doing so. The “instrumental” rewards are the outcome-contingent rewards; you only get these rewards if the right candidate wins. The “expressive” rewards are entirely intrinsic; your conscience will feel better if you know you made the ethically “right” choice. Because each individual voter has almost zero influence over the outcome of an election, voters don’t pay much attention to outcome-contingent rewards. But because the “expressive” rewards are not outcome-contingent, voters will pay attention to these sorts of rewards.

Thus, when a voter’s material preferences and moral preferences point to opposite candidates (e.g. Dean or Bush in 2004), expect the moral preferences to dominate.

The authors test this theory with some interesting laboratory experiments. They find that (contrived) instrumental rewards have less and less influence on subject’s “votes” relative to (contrive) expressive rewards as the pivot probability falls from 1 to 1/11. From this, they infer that expressive rewards would dominate in a real election, where pivot probabilities are often smaller than 1/1,000,000.

Some parting thoughts

This is an interesting bit of research. Although the authors never mention Dean, evangelicals, or any other particular group, it’s clear what they’re really trying to explain. They want to explain why southern evangelicals keep voting Republican, despite their better interest (according to Dean). Once the authors’ sterile, academic language is replaced with these specific references to Dean and southern evangelicals, some flaws with the theory become clear, though.

The biggest flaw: Why do we assume that moral preferences are merely expressive and not also instrumental? Sincere believers may well think that divine providence will send greater prosperity, peace, and blessings to the United States if only Americans work to be a moral people. This is a constant theme in the Old Testament; when God’s people follow his will, they prosper.

I’m neither southern nor evangelical, but it’s apparent that the authors of this study, like Dean, do not take evangelicals’ faith seriously. If they did, they would acknowledge that “values votes” may well seem instrumental to the people casting them. And if these moral choices are instrumental, not merely expressive, then the entire theory collapses inward.

Where would that leave us? Perhaps southern white evangelicals preferred Bush over Dean (and Kerry) in 2004 because they actually thought that Bush’s sincere religious faith would be instrumentally good for them and their country.

Possibly related posts:

5 Comments

  1. Matt Unregistered
    Posted September 10, 2009 at 9:12 am | Permalink

    I’m not sure it’s fair to say that the authors were just trying to answer Howard Dean. They never mention Dean in their article or anything like him. Everything they write is completely in the abstract.

  2. Posted September 10, 2009 at 9:26 am | Permalink

    Yes, that’s true, which I tried to make clear in my post.

    But it seems that if their abstract language has any real-world application, the Dean episode would be a perfect example. I used the Dean example in hopes of illustrate more clearly what they were really saying for folks who hadn’t yet read the original article.

  3. CT Unregistered
    Posted September 10, 2009 at 5:10 pm | Permalink

    >> The biggest flaw: Why do we assume that moral preferences are merely expressive and not also instrumental?

    They test for this. It can’t explain why turnout stays high as the electorate grows. Any instrumental voters, selfish or ethical, turnout less when the electorate grows in their theory.

  4. Posted September 16, 2009 at 10:23 am | Permalink

    I could have been clearer, CT.

    They don’t test that moral preferences are expressive and not instrumental. What they test is whether instrumental preferences can explain turnout.

    That’s sort of my point. If moral preferences are instrumental and not expressive, then the experiments do little to explain turnout other than point to the importance of the “D” term generally.

  5. Jack Unregistered
    Posted September 30, 2009 at 11:36 am | Permalink

    Great points at the end.