New draft paper: "Does Affective Polarization Undermine Democratic Norms or Accountability? Maybe Not" (w/ @j_kalla & @seanjwestwood)

Would love comments! Draft is here: https://osf.io/9btsq/ .

For those who want a quick summary, a thread... (1/n)

First, what is affective polarization? The increasing dislike people have for people who identify with the other political party (the "outparty").

There is *a lot* of research that worries/speculates/asserts that increasing affective polarization spells doom for our democracy:

We consider five worries:
1) affpol leads voters to ignore their party's shortcomings
2) affpol leads voters to adopt their party's issue views
3) affpol leads voters to oppose bipartisanship
4) affpol reduces support for democratic norms
5) affpol distorts perceptions of reality

Many assume we know all this. We don't. In fact, it turns out, we don't know much at all about the political consequences of affective polarization.

Now, a lot of research *has* found affpol permeates everyday life: eg, people don't want to date or hire outpartisans now.

But...

..it's very difficult to understand its political effects.

This is because more affectively polarized people are different in lots of ways. If they behave differently, we don't know if it's because of affective polarization or something else.

Our solution: randomly reduce/increase affective polarization, then measure consequences.

We manipulate affpol w/ a trust game: respondents think they are playing with an outpartisan who is either generous or unfair to them. (We debrief later.)

This creates a difference equivalent to >30 years of rising affpol. We rewind ~32 years of affpol, holding all else constant.

This has the predicted effects on interpersonal measures like social distance: affpol causes people to want to avoid interactions w/ outpartisans.

But when it comes to political outcomes, we see no evidence for all 5 of the worries I mentioned:

1) We tell respondents who actually represents them in Congress & how their MCs actually voted on recent bills.

First, increasing affpol doesn't increase approval of copartisan MCs

Second, it doesn't lead people to ignore how their representatives vote:

People whose affpol we increased react just as negatively (if not more so) to news their representative cast votes they disagree with as people whose affpol is reduced.

I.e., accountability effects null.

2) Now that people have learned how politicians of their/the other party vote, does increasing affective polarization mean they are more likely to adopt their party's issue views? We find no: another null.

3) What about bipartisanship? We adapt a vignette Laurel Harbridge-Yong & @namalhotra developed that measures preferences for elite bipartisanship.

People on average prefer bipartisan to partisan elite behavior - and the increased affpol group does so just as much. Another null.

4) We also measure support for a *lot* of different democratic norms, with both yes/no questions and vignettes, like a vignette asking if one's party should override an election of an outpartisan.

Nulls across the board.

5) Last, we see no effects on stated perceptions of objective conditions. Ds & Rs report different perceptions, but increasing/reducing their affpol does not affect these at all.

Another null.

Just in case some of those coefficients close to p < 0.05 are tempting, we also show all our t-stats. The number of t-stats near 1.96 is essentially what you'd expect from chance.

And, as we discuss, all the results that are near significance don't replicate.

Why all these nulls?

Social identity theory makes clear predictions for interpersonal interactions, but political decisions are different: they involve a lot more considerations than whether one dislikes outpartisans, and our results suggest these are more important to voters.

As a last exercise, we compare what our estimates say about the causal effect of affpol to what a naive regression would say.

In the social distance domain, the two estimates are essentially identical. That is, the correlations scholars have seen are probably causal. Good news!

But in political domain, endogenous estimates are almost always positive despite the causal nulls & 3/4 are larger than the causal estimates.

What researchers may at first glance think is the power of affective polarization appears to be the power of omitted variable bias.

Why does this matter? To improve American democracy, we need to understand what is & is not undermining it. Our results suggest affpol may not be a major factor.

Our results also suggest where affpol may matter: when politics & interpersonal interactions are inherently linked.

I'll repeat what I said at the start: this is a first draft, and we would love to hear comments!

Full paper here: https://osf.io/9btsq/ . (n/n)