Thread by @Moshe_Hoffman

A thread on economic models of voting.

This thread is organized into 4 parts, for ease of reading.

Part 1=my problem with the lit.

Part 2=response to one recurring counterargument.

Part 3=some puzzles this leaves us with.

Part 4=some thoughts on how to resolve these puzzles+move the lit forward.

Part 1:

There is a lot of elegent modeling in “political economics” and “social choice theory.”

Such as in this talk by Eric Maskin on the problems with majority voting, and how it gave us Trump:

https://twitter.com/harvardcmsa/status/1070043322838855680?s=21

Live-tweeted here by @ben_golub

https://twitter.com/ben_golub/status/1055205854436306945?s=21

The (Nobel winning) models are elegant. And capture many insights.

For instance

-It is hard to aggregate voters’ preferences.

-Some voting systems do have more problems than others.

-Majority rule (our system) is particularly bad.

-That’s one reason we got stuck w/ trump.

But there seems to be a fundamental flaw in this literature, which I want to expound on:

Namely, these models all assume voters are motivated by preferences over outcomes.

But that’s patently false!

How do we know this?

For one:

Voters have zero influence on outcomes.

Economists know this. They have calculated the odds voters will be “pivotal.”

These odds are infinitesimally small.

For two:

We can check what voter behavior indicate about their motives. (Revealed preferences!)

E.g. Do we vote for 3rd party candidates? Oops. Voting for Nader didn’t help the environment. It gave us Bush.

Can this be explained by preferences over outcome? (Hint: no)

For three:

Voters seem to pay minimal attention to the actual efficacy of proposed policies.

After all, it isn’t *that hard* to check factcheck.org, to know refugees are not terrorists, and electing a con-man president is a bad idea.

Economists see these issues. They know their key assumption fails.

They just shrug their shoulders, and say that requires a behavioral model, and is hard to math-ify, and revert to their assumptions of preferences-over-outcomes.

Thats not how science works.

Part 2 of thread:

Some might retort (as @Jabaluck has):

Figuring this stuff out is *hard.* Beyond the ability of voters.

Who, but an economist, can calculate the odds of being pivotal?
Who can tell which news channel to trust, if they each tell you the other is lying?

My response:

It’s true. People are bad at this stuff.

But is that because they are stupid? And it’s hard? Or because they are strategically ignorant? Or purposefully misinformed? (Correct answer: the latter two)

Let’s look at the evidence.

Fact one:

When presented in contexts where there isn’t a motive to draw one conclusion over another, people don’t make the same logical errors. Or have the same problem discerning truth from fiction.

We have no trouble seeing that a single coin toss in a million is unlikely to change which side will be modal.

And no one is going to trust the “I am not a scientist, but...” talking head on Fox, if instead of assessing climate change, she were assessing the lump on your chest.

Fact two:

Smarter better informed voters are more, not less, prone to make these logical fallacies. Presumably because they have more means to bend the truth.

*See the work of Dan Kahan @cult_cognition, nicely reviewed here:

Fact three:

There is plenty of evidence outside the domain of politics that people purposely remain ignorant/misinformed, when there is a strategic incentive to do so.

See the work of eg @CainDaylian and Jason Dana. Here’s a good example:

Finally:

If people were really unable to discern the truth, then the econ models would still be absurd.

*Albeit not b/c voters wouldn’t have preferences over outcomes. But b/c voters would have no clue how their votes translate into outcomes-another key assumption in the models

Part 3 of thread:

So where does this leave us?

With a few puzzles, for starters.

Puzzle 1:

Why, despite this terrible preference-over-outcomes assumption, do these models still capture some valuable insights, such as the insights from Maskin listed in tweet 4 above.

Puzzle 2:

People often do vote for the outcomes that they would prefer.

Eg black people are less likely to vote for a racist president.

Why?

Puzzle 3:

When we expand the vote, more people’s welfare *is* taken into consideration by the elected representatives.

Why?

It isn’t clear why any of this would be true if our vote didn’t at all correspond to preferences-over- outcomes.

Outcomes do clearly enter our preferences. But why? And when?

For that we need to think about what shapes our political preferences. Part 4 takes a stab at that.

Part 4:

What are the actual, material incentives voters face? The actual consequences to their welfare that come from their own vote?

Let’s think about that. And see if it gives us any bite. (Foreshadowing: it will.)

Any guesses? Here are two of mine:

1) Coalitions of voters that are well managed, will pressure their members to vote a certain way, and that way will typically be what benefits that coalition as a whole, or it’s managers.

2) Voters use their vote to signal their values.

The above list of motives isn’t new (or complete or rigorous). But let’s discuss some of it’s implications.

Implication 1:

Both of these motives will get deeply internalized. And shape our political preferences^.

^That’s the key premise in all my work and threads. Preferences are not exogenous. They are shaped by the actual material incentives, through evolutionary or learning processes. Where else would they come from?

Implication 2:

Now, consider what this tells us about our political preferences:

So long as motive (1) bites, (and the effect of (2) are minimal), we will get Maskin preferences. (And the above puzzles resolved).

But not always. (And that should answer the “when” question).

Of course, this last part of thread is speculative. And not at all rigorous.

Thoughts or suggestions for what other actual material incentives are at play, or how to formalize these, would be very much appreciated.

Or anything else my analysis is missing. Thanks.