Article Summary
Introduction
U.S. partisans express increasing levels of disdain for the other side of the aisle, a level of disdain which appears to extend to several areas of social interaction (such as dating decisions). One of the strongest claims in this vein is that partisans have sorted residentially due to political animosity. More precisely, some have argued that partisans move to the neighborhoods they do in order to live in areas with greater shares of co-partisans. This claim, however, is complicated by an empirical contradiction: while many partisans express strong preferences for certain communities, actual moving behavior shows very minimal levels of sorting.
In their paper, Mummolo and Nall argue that this contradiction is due in part to the importance partisans attach to neighborhood social composition, which while different between partisans, pales in comparison to the importance both Democrats and Republicans place on other neighborhood characteristics like crime, location, and affordability. Additionally, even if partisans desired to move, their opportunities to move to neighborhoods on the basis of partisanship may be limited after taking into consideration these other neighborhood features.
By directly engaging with these questions, the authors build on existing literature by simultaneously measuring both stated neighborhood desires and actual moving behavior, whereas previous studies have typically only considered one or the other.
Analytical Approach
The authors conduct a survey of 4,800 Democrats and Republicans through SSI, asking respondents to report moving behavior in the last five years. Additionally, the authors capture neighborhood preferences in the survey through a series of two experiments.
Their first experiment is a forced-choice conjoint design, where respondents are asked to choose between two neighborhoods with randomly generated features. The features (detailed in the table below) encompass multiple neighborhood traits such as crime, commute distance, density, and neighborhood composition (including partisanship). Respondents complete 5 forced-choice tasks, and the authors calculate the effect for any given feature conditional on respondent partisanship, which can be interpreted as the change in probability of selection (relative to a baseline category) when that feature is present. Differences between partisans suggests differences in importance placed on particular factors.
Table 1: Randomized Features in Experiment 1
| Feature | Level (Randomized within Feature) |
|---|---|
| Housing Cost | 15% of pre-tax income |
| 30% of pre-tax income | |
| 40% of pre-tax income | |
| Presidential Vote (2012) | 50% Democrat, 50% Republican |
| 30% Democrat, 70% Republican | |
| 70% Democrat, 30% Republican | |
| Racial Composition | 50% White, 50% Nonwhite |
| 75% White, 25% Nonwhite | |
| 90% White, 10% Nonwhite | |
| 96% white, 4% Nonwhite | |
| School Quality | 5 out of 10 |
| 9 out of 10 | |
| Daily Driving Time | 10 mins |
| 25 mins | |
| 45 mins | |
| 75 mins | |
| Violent Crime Rate | 20% < National Average |
| 20% > National Average | |
| Type of Place | City, more residential area |
| City, downtown area | |
| Rural area | |
| Small town | |
| Suburban, only houses | |
| Suburban, downtown area |
In the second experiment, respondents are given 9 pairs of neighborhood characteristics randomly drawn from a list of 62 and asked which is a more significant factor in deciding where to live. The characteristics are grouped into the following categories: disorder, geography/location, friends/family, neighborhood income, government, transportation, smart growth versus sprawl, children, social life, neighborhood social composition (including partisanship) and attitudes, and neighborhood race. From these choices the authors can determine the proportion of times respondents chose a trait when it was an option.
The authors conclude with two observational analyses. In the first, they conduct a “feasibility” study for potential moves from a respondent’s current zip code. Using data from the 2012 5-year ACS, they apply increasingly stringent constraints on feasible moves, such as spending no more than 3 times their annual income on a house and the proportion of homeowners being at least as high as their current neighborhood. They then calculate the number of neighborhoods with a greater share of copartisans remaining after these constraints are applied. In the second analysis, the authors use the respondents self-reported zip code and compare the partisan composition of that zip code with that of the zip code from which the respondent most recently moved (in the last 5 years). They calculate the conditional difference in means by party across their entire sample and within certain subgroups (such as race, income, strength of partisanship).
Main Findings
While partisans diverge on their preference for neighborhood partisan composition and other features adjacent to partisanship (racial composition and type of place), they are in general agreement about many other features (housing cost, school quality, driving time, and violent crime rate). In fact, many of these features have comparable or greater effects on behavior than the partisan composition of neighborhoods. When examining the direct comparisons of the second experiment, only around 30% of respondents listed neighbors being Democratic or Republican as more important over randomly paired alternatives. More broadly, neighborhood beliefs and values ranked as one of the least prioritized features (along with neighborhood race). Disorder, geography, and family/friends ranked among the most important, being chosen by 60-70% of respondents over randomly paired alternatives.
Turning to the observational results, the authors find partisans are severely constrained in their ability to move to more Democratic/Republican neighborhoods after taking into consideration affordability, neighborhood quality, and geographic location. 38% of Democrats and 28% of Republicans would have less than 5% of the national housing stock to choose from if they wished to avoid financial hardship. The additional constraint of neighborhood quality means majorities of both parties would be able to choose from at most 10% of the national housing stock. Finally, with the geographic constraint, 90% of both parties would be able to choose from 5% of the housing stock. Put simply, there are not many feasible options for partisans wishing to move to more co-partisan neighborhoods.
Finally, when analyzing movers in their sample, the authors find no difference in the partisan composition of neighborhoods that Democrats and Republicans have moved to. If anything, both Democrats and Republicans have both moved to neighborhoods about 2% more Republican than their previous neighborhoods. While there is some difference between the moving patterns of strong partisans, it is not in the expected direction, as strong Democrats seem more likely to move to more Republican neighborhoods.
Implications
In aggregate, the results of this paper suggest partisans are only mildly influenced, if at all, by partisan homophily or partisan discrimination. While partisans differ in their stated preferences for the partisan composition of neighborhoods, they share stronger preferences for more mundane features such as safety and commute time. This is not to say partisans don’t deeply dislike one another, just that such disdain is not enough to drive residential sorting.
The authors make a significant contribution to the literature by providing for the first time individual-level attitudes toward moving linked with self-reported moving behavior, whereas previous studies have relied either exclusively on macro-level movements or individual-level desires. The study re-emphasizes the need to link attitudes with behavior, as the two often do not perfectly align.
Additionally, while neighborhoods are still broadly heterogeneous with regard to partisanship, they are not necessarily heterogeneous in other social features such as race or income. Even more, just because partisans live in the same area does not mean they are not polarized. Indeed, the most “purple” neighborhoods may contribute to greater affective polarization.
Questions left unanswered
While the authors give a wealth of evidence on the primary considerations partisans hold when moving, there is still an open question on the information they have access to in a non-experimental setting. By construction, the experiments in the paper give prospective movers all the information they researchers are interested in, but the information-search process may be different from the decision-making process described here. We also do not know anything about moving behavior beyond the 5-year timeframe provided by the paper, so more long-term trends may point toward more partisan sorting. Finally, the paper does not tackle the issue of partisan updating within a neighborhood. For example, if someone moved to a primarily Republican zip code as a Democrat, do they remain a Democrat 5 years later? This paper is specifically interested in the moving process, but other questions arise on post-move attitudes.
Methods and Analysis
Was the study and its analyses pre-registered?: No
Did the study rely on proxy variables to measure polarization?: Yes
Affective polarization was not the main quantity of analysis in this study, so the main variables of interest were related to choices between neighborhood features and self-reported moving history.
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
This paper is incredibly thorough in its questions, design, and implementation. There is a wealth of information in a very concise article, with the experimental and observational design complementing each other nicely. One limitation imposed by the reliance on survey data, however, is the use of zip codes as stand-ins for neighborhoods. While many zip codes do encompass areas one would consider a singular community, others are much more diverse in their compositions, especially at the micro-level. As Brown and Enos (2021) find when using voter file data with individual addresses, “a large proportion of voters live with virtually no exposure to voters from the other party in their residential environment.” So while sorting may not occur at the zip code level, there does seem to be evidence it occurs at the micro level. This raises a question as to whether the same decision considerations hold, or if differences in preferences across other categories preempt the need for explicit consideration of partisanship. Additionally, the authors make an assumption in their analysis of self-reported moves that the reported partisanship of the respondent at the time of the survey is the same as it was when they moved. Recent (unpublished) work by Jacob Brown suggests recent transplants are not immune to being swayed ideologically in one way or another by their neighbors. If that is the case, residential sorting may not be a function of mover choice but of post-move influence.
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
Mummolo, J., & Nall, C. (2017). Why Partisans do not sort: The constraints on political segregation. The Journal of Politics, 79(1), 45–59. https://doi.org/10.1086/687569
Bibtex
1
2
3
4
5
6
7
8
9
10
11
12
13
@article{doi:10.1086/687569,
author = {Mummolo, Jonathan and Nall, Clayton},
title = {Why Partisans Do Not Sort: The Constraints on Political Segregation},
journal = {The Journal of Politics},
volume = {79},
number = {1},
pages = {45-59},
year = {2017},
doi = {10.1086/687569},
URL = { https://doi.org/10.1086/687569},
eprint = { https://doi.org/10.1086/687569}
}