The Social Construction of Partisanship: Misperceptions About Party Composition and Partisan Identification
Download data and study materials from OSF
Florida State University
Sample size: 2222
Field period: 04/28/2017-05/23/2018
Why are Americans, many of whom lack stable policy preferences and/or practicable political knowledge, so entrenched in their partisan identities? A dominant theory asserts that Americans' partisanship is grounded in their more primal group identities. According to this view, people identify with one of the two major parties to the degree that party's social prototype overlaps with their own group identities. But direct, causally identified evidence is surprisingly lacking. We leverage the tendency of Democrats and Republicans to overestimate the prevalence of stereotype-consistent supporters within their own side's ranks. Since most party stereotypes fit a minority of that party's supporters, these misperceptions are tantamount to overestimating the prevalence of people not like oneself in their own party. We correct these misperceptions for a random subsample. If this particular identity-based theory of partisanship is correct, we ought to observe the randomly-corrected respondents express greater partisanship in response to the treatment.
1. Do citizens' perceptions of their own party's composition affect the degree to which they identify with that party?
2. When people learn that their party is less stereotypical than they previously thought, do their feelings toward the party change?
3. Does the degree to which people hold party-stereotypical identities moderate the treatment (information that the party is less stereotypical than most believe it to be)? That is, do people come to like the party more if they hold fewer party-stereotypical traits but like it less if they hold more?
4. Does the degree to which people misperceive their own party as composed of stereotypical partisans moderate the treatment (information that the party is less stereotypical than most believe it to be)? That is, if party composition matters, theoretically, the treatment should have the greatest effect on those for whom it brings the most correction.
We randomly assigned respondents to one of three experimental conditions:
1. The "ask" condition, in which we elicited Democratic respondents' perceptions of the percentage of Democrats who are nonwhite, atheist/agnostic, or 35 or younger; and in which we elicited Republican respondents' perceptions of the percentage of Republicans who are $250k+/year earners, evangelical, or age 65 or older.
2. The "tell" condition, in which we elicited these perceptions but then provided the correct information.
3. A control condition in which we elicited these perceptions after administering the dependent measures.
1. The traditional seven-point party identification scale, pre-post. All respondents recruited were weak/leaning Democrats and Republicans, so they could move up or down the 7-point scale without "crossing over" to the other party. (We found no movement on this measure, perhaps unsurprisingly.)
2. The expressive partisanship scale introduced by Huddy et al. (2015), which includes four items designed to tap identification with one's political party.
Summary of Results
Republicans behave in a theoretically-consistent manner (at first blush), expressing greater partisanship when they learn their party is more like them (on average)—but Democrats do not exhibit such a response to the treatment, contrary to contemporary wisdom (Grossman and Hopkins 2016). Furthermore, and bearing directly on theories of partisanship, respondents' own identification with party-stereotypical groups does not appear to interact with the treatment. Instead, the degree to which respondents misperceive party composition moderates the treatment but, again, heterogeneously across party: consistent with Ahler and Sood (2018), treatment effects among Republicans are greatest for those with the most stereotype-inflated perceptions of Republican composition. On the other hand, Democrats with the most biased perceptions appear to express weaker partisanship when corrected.