Article Summary
Introduction
Democrats and Republicans increasingly dislike each other. Existing literature, however, has difficulty identifying whether this dislike is caused by partisanship as a social identity or as a signal for policy positions, as both can be inferred from partisan labels. This is even a problem in survey experiments where partisanship and policy can be independently randomized; respondents can infer policy information from partisanship. This means we know more about the effects of affective polarization than its causes, making interventions meant to alleviate partisan disagreement difficult to craft. Orr and Huber (2020) confront this difficulty by attempting to determine the independent effects of partisanship and policy positions on partisan animosity. In doing so, they answer the question of how Americans use and prioritize partisan versus policy information in interpersonal evaluations.
Analytical Approach
Orr and Huber independently randomized characteristics of fictional people in a series of 3 vignette experiments, asking survey respondents to rate how warmly they feel towards the person described in the vignette or, in Study 2, how excited they would be if their daughter married this person. Respondents were assigned a treatment condition wherein they received information about some combination of characteristics, where a randomized attribute is assigned for each characteristic. A full list of studies, treatments, characteristics, and attributes is given in Table 1. For example, a respondent in Study 1 assigned to the PID+Policy treatment could see a vignette about a South Carolinian who enjoys watching sports, is a Democrat, and believes the federal government should make it easier for people to buy a gun.
Table 1: Randomized Attributes of Studies (adapted from Orr and Huber 2020 Table 1)
| Study | Treatments | Characteristic | Attribute |
| Study 1 | PID, Policy, PID + Policy (all with social cues) | PID | Republican |
| Democrat | |||
| Independent | |||
| Not interested | |||
| Policy | Immigration | ||
| Welfare | |||
| Abortion | |||
| Social Cues | State | ||
| Job | |||
| Study 2 | PID, policy, groups, PID + Policy, PID + Groups, PID + Policy + Groups (all with social cues). PID without social cues added for Marriage DV | PID | Republican |
| Democrat | |||
| Independent | |||
| Not interested | |||
| Policy | Immigration | ||
| Welfare | |||
| Guns | |||
| Abortion | |||
| Same-sex adoption | |||
| Groups | Race (White, Black, Hispanic, Asian) | ||
| Religion (Protestant, Catholic, Jewish, Atheist, not religious) | |||
| Social Cues (Feeling Thermometer DV) | Age | ||
| Gender | |||
| Hobby | |||
| Music preference | |||
| Age | |||
| Social Cues (Marriage DV) | Age | ||
| Child-rearing opinion | |||
| Valence (smoke, divorce, exercise) | |||
| Study 3 | PID, Policy, PID + Policy (all with social cues) | PID | Republican |
| Democrat | |||
| Independent | |||
| Policy | Immigration | ||
| Welfare | |||
| Guns | |||
| Abortion | |||
| Same-sex adoption | |||
| Social Cues (pick 2) | State | ||
| Job | |||
| Hobby | |||
| Music preference |
The independent randomization of characteristics and differential access to information across treatment groups allow Orr and Huber to estimate the effects of how additional policy information changes the effect of partisanship on feelings of warmth. If such feelings are driven primarily by social dimensions of partisanship, access to policy information should not attenuate the effect of partisanship. If affect is primarily a function of policy disagreement, however, the opposite should occur, and the addition or removal of party information from vignettes containing policy positions should have minimal effect.
Main Findings
Across all studies, Orr and Huber find evidence for policy-based evaluations of vignettes. For the studies with warmth as the dependent variable, policy information has a greater effect on vignette evaluation than party information when presented in isolation. When presented jointly, the influence of partisan information quickly attenuates (over 50% across all three studies) but the effect of policy information is reduced only slightly (statistically insignificant reductions in studies 1 and 2, a 15% reduction in study 3). The inclusion of additional characteristics, such as race, religion, and character, tend to have effects comparable to the PID + Policy treatment group. The results are almost identical with the fiancé evaluation dependent variable: significant reductions in the effect of PID with the inclusion of policy, but insignificant reductions in the effect of policy with the inclusion of PID.
The effects of the studies are strongest amongst strong partisans and when policy information is incongruent with party. Furthermore, both strong and weak partisans give more weight to policy information than partisanship in study 3 (relied upon due to sample size).
Finally, Orr and Huber followed the vignette experiment in study 2 with an additional experiment asking respondents to assign seats for a friend’s wedding so that everyone has a good time, randomly showing three pieces of information about each guest (partisanship hobby, personality, religion, gender/sexuality, how well they know the couple). Respondents were then assigned randomly to either have policy information included as well. When policy information was presented, demand for partisanship information decreased by 30%.
Implications
These findings suggest affective polarization is moreso a function of policy disagreement than social identity, as the effect sizes for policy disagreement were both larger in isolation and resistant to the inclusion of party identification. The findings also imply the information sought from party identification has more to do with policy positions than social information, as there is a large reduction in demand for such information in the presence of policy information. However, the effect of partisanship does not go to zero, even when presented with additional characteristics like race, religion, and other characteristics, suggesting there is a social component to partisan identification that remains influential in our evaluations of others.
Questions left unanswered
While this study focused on interpersonal dynamics amongst the mass public, it raises questions on if similar effects are present in the evaluations of political candidates. Orr and Huber themselves suggest uniformity of policy stances within party elites means learning of one’s partisanship may be sufficient to know almost all others with a great deal of certainty. The inverse could also be true; upon learning one policy position, respondents could easily identify other positions the candidate is likely to take as well. It is also unclear which interventions to partisan animosity may be more effective given the more ideological underpinnings of interpersonal affect.
Methods and Analysis
Was the study and its analyses pre-registered?: No
Did the study rely on proxy variables to measure polarization?: No
Were standard p-value thresholds used (p<.05 or 95% Confidence Intervals that don’t overlap zero)?: Yes
- Largest p-value presented as significant: 0.05
Were correlational results interpreted with causal language?: No
Limitations / Weaknesses
By independently randomizing partisan and policy characteristics across a variety of treatment conditions, Orr and Huber give compelling evidence that policy and party information have differential effects. While the authors are rightfully wary of the multitude of cues given by party identification, they do not give similar attention to how policy positions can do the same. In the same way that respondents can infer policy positions from partisanship, respondents may also infer partisanship (or, more specifically, partisan type) from policy positions. For example, if a Democratic respondent is presented with a vignette describing a Democrat who wants the federal government to make it easier for people to buy a gun, that respondent may rate this person less warmly not because of disagreement with policy but because they now believe the Democrat to be disloyal to the party. It therefore may not be the policy information itself driving behavior, but the specificity with which respondents can now identify partisan type.
Open Data & Analyses
Does the article make the replication data publicly available?: Yes
Does the article make the replication analysis scripts publicly available?: Yes
Article Citation
Orr, L.V. and Huber, G.A. (2020), The Policy Basis of Measured Partisan Animosity in the United States. American Journal of Political Science, 64: 569-586. https://doi.org/10.1111/ajps.12498
Bibtex
1
2
3
4
5
6
7
8
9
10
11
12
13
@article{https://doi.org/10.1111/ajps.12498,
author = {Orr, Lilla V. and Huber, Gregory A.},
title = {The Policy Basis of Measured Partisan Animosity in the United States},
journal = {American Journal of Political Science},
volume = {64},
number = {3},
pages = {569-586},
doi = {https://doi.org/10.1111/ajps.12498},
url = {https://onlinelibrary.wiley.com/doi/abs/10.1111/ajps.12498},
eprint = {https://onlinelibrary.wiley.com/doi/pdf/10.1111/ajps.12498},
year = {2020}
}