Partisan niche construction: Out-party affect, geographic sorting, and mate selection

Abstract

Partisans in the American electorate are affectively polarized, which coincides with the tendency for partisan geographic sorting. Could mate selection pressures contribute to this geographic tendency, and how might they interact with out-party affect? I propose a model in which an individual’s perception of their mate success in a niche is key. I argue that perceived mate success is a function of a niche’s partisanship and one’s out-party affect, which in turn, incentivizes sorting. The model is partially tested with conjoint experiments on multiple U.S. samples. Results show that partisans perceive a lower probability of mate success in niches with greater shares of out-partisans and that mate success interacts with negative out-party affect. I also replicate findings on political mate choice preferences with a more appropriate method. Lastly, this project links instrumentality and affect, which is a departure from past work. In doing so, it contributes to research on the consequences of mate pressures for political behavior.

Introduction

Partisan masses in the United States are increasingly affectively at odds with one another. That is, Republicans and Democrats dislike each other, and they are growing more distrustful of out-party members. This surge in animosity, also known as affective polarization (Iyengar et al., 2012; Iyengar et al., 2019), traces its roots to the fact that being a “Republican” or a “Democrat” is a potent indicator of who one is—that is, a social identity (Green et al., 2004; Huddy & Bankert, 2017; Huddy et al., 2015). What results from this motivation to belong to a social group in the American political context is “the tendency [for] people identifying as Republicans or Democrats to view opposing partisans negatively and co-partisans positively” (Iyengar & Westwood, 2015, p. 691). Additionally, there seems to be a concurrent dynamic among partisans to cluster with co-partisans and away from out-partisans at multiple geographic levels (Brown & Enos, 2021; Cho et al., 2019), which is in line with work showing increasingly homogeneous districts for members of Congress (Abramowitz et al., 2006; Levendusky et al., 2008; McCarty et al., 2009). Brown and Enos (2021) show the extent of such partisan segregation by creating individual-level spatially weighted measures of partisan segregation using registered voters in the United States. Their results indicate “high partisan segregation across the country, with most voters of both political parties living in partisan bubbles with little exposure to the other party” (Brown & Enos, 2021, p. 1005).

Increasing negative affect across party lines coupled with geographic political segregation should concern those worried about declining democratic norms. While some posit that affective polarization might not be as problematic as it seems (Broockman et al., 2023), the debate is ongoing (Kingzette et al., 2021). Still, it seems that spatial separation fosters intergroup bias and conflict (Enos & Celaya, 2018; Uslaner, 2012; Zhinger & Thomas, 2014). Thus, understanding all the factors that may worsen geographic polarization is vital. In addition, if increasing negative affect and geographic sorting are related in any way, the possibility for self-reinforcement should give us pause, though previous work has undertheorized their connection in the American political context. To fill this gap and to illustrate another potential force behind this increasing social division, this article introduces a mate success model linking geographic sorting and out-partisan affect.

The model proposed here builds on the political mate choice literature, which finds that individuals select and prefer partners with whom they have similar political orientations (Alford et al., 2011; Easton & Holbein, 2021; Huber & Malhotra, 2017; Iyengar et al., 2018; Klofstad et al., 2013; Nicholson et al., 2016; Stoker & Jennings, 2005). Specifically, I argue that this mate preference functions as a selection pressure for individuals that interacts with negative out-party affect, resulting in varying perceptions of mate success across geographic regions, or niches. ¹ Mate success is simply defined here as the probability that one can find a romantic long-term partner. The main claim is that if a niche is composed of individuals mostly of the out-party, and one holds negative affect toward the out-party, an individual’s perception of their probability of mate success will be hindered in that niche. As a result, highly partisan clustered areas repel out-partisan individuals the most, and negative affect magnifies this effect. Individuals are then incentivized to seek out niches with improved prospects or choose such niches when an opportunity arises. Additionally, such a mechanism could create a positive feedback loop between geographic and affective polarization. That is, greater geographic segregation increases perceptions of distinctiveness, which increases negative affect (see Figure 1), pushing individuals toward greater sorting, and so on.

The key innovation here is the contribution to the work on partisan homophily by explicitly incorporating mate choice as an influence on geographic preference/mobility. Other forces could include one’s preference to engage socially with co-partisans independent of mate preferences. Yet, while mate preferences and social preferences are distinct, they are largely complementary influences. The goal of this article is to emphasize the importance of the former, not to argue against the latter. Also, while previous work has considered different potential causes of partisan homophily, such as (1) choice-based homophily, (2) social structure-based homophily, and (3) convergence-based homophily (Huber & Malhotra, 2017, p. 271), the model introduced here presents the possibility of integrating these causes. Future work should expand on this integration (see the Limitations and future work section). Thus, the model here presents a theoretically rich approach to geographic mobility/sorting and the underlying forces that explain who is more or less likely to want to move and when. Empirically, this research seeks first and foremost to link perceptions of mate success to the partisanship of geographic niches and gauge willingness to move.

I conduct four separate conjoint experiments to test the motivational basis of the model, in total surveying more than 7,400 individuals, resulting in over 61,000 choices made. Overall, this is a more appropriate design to make inferences of preferences over objects with multiple attributes (Hainmueller et al., 2014). This design allows me to estimate average marginal component effects (AMCEs) of levels within attributes (e.g., religious versus nonreligious), average component interaction effects (ACIEs) when I further condition on respondent covariates (e.g., religious versus nonreligious given that the respondent is a Democrat), and marginal means to compare levels of attributes in a more absolute sense across attributes. The first conjoint experiment seeks to replicate findings on political mate preferences. This incorporates previously underexplored attributes such as wealth, race/ethnicity, attractiveness, and other individual attributes. The second and third conjoint experiments link perceptions of mate success to the partisan composition of a niche (at the city level) and the interaction with out-party affect. By randomly altering attribute levels of population size, racial/ethnic demographics, and partisanship (among other attributes), I can estimate the relative and absolute influence of partisan composition on perceptions of mate success. Both surveys also gauge willingness to move. A fourth conjoint (at the neighborhood level) gauges the extent to which willingness to move is driven by additional attributes not included in the second and third surveys. It also looks at mate availability to see whether, among individuals in the market for a partner, this is a key factor for their preferences.

This article makes several key contributions. First, it successfully replicates and extends previous work fortifying our knowledge about political mate preferences using a more appropriate causal method for mate choice. Additionally, the theoretical model connects two key developments in American politics (partisan sorting and affective partisanship) and points us toward an important underexplored variable. Empirically, it establishes a link between perceptions of mate success (and willingness to move) to a niche’s partisan composition for the first time, and the interaction with out-party affect. I also find evidence that willingness to move is pronounced by greater availability of potential mates for individuals who are more likely to be in pursuit of a partner. This lends more support to the idea that prospects for mate acquisition is an additional factor influencing one’s choice of location. The results serve as strong suggestive evidence for the argument that variation in perceptions of mate success is an additional factor in partisan sorting. This study also contributes to research that looks at the effects of mate pressures on political behavior (Dancygier et al., 2022) by exploring how partisan mate selection and out-party affect interact to influence partisan sorting. Moreover, it presents a unique link between affective and instrumental motivations.

Partisan niche construction

Political mate choice

Mate choice is fundamental to all reproductive species because it imposes selection pressures on individuals that may influence the direction of evolution as traits in prospective partners are accepted or rejected (Andersson, 2019; Darwin, 1871). Here, I consider “partisan traits” as a key driving force for mate choice. Specifically, I define partisan traits as a set of traits associated with people who identify with a given political party, which then serves as a heuristic for a worldview, value system, personality, political attitudes/preferences, or lifestyle, and so on (Carney et al., 2008; Hetherington & Weiler, 2018; Johnston et al., 2017).

Some of the first work on political assortative mating concluded that “spouses tend toward like-mindedness because of the selection processes that bring them together in the first place,” but “mutual socialization also works to produce commonalities in political outlook over time” (Stoker & Jennings, 2005, p. 70). Later work also found that married couples have a substantially high degree of similarity in their political orientations, but this seems to be driven primarily by selection (i.e., choice) rather than convergence or assimilation over time (Alford et al., 2011, p. 373). Other research has found that partisan cues influence perceptions of attractiveness (Nicholson et al., 2016), and such results have been replicated under a variety of experimental design choices (Easton & Holbein, 2021). Moreover, research has used both experimental and behavioral data to simultaneously rule out “post-choice convergence, restricted partner markets, and sorting on nonpolitical factors,” concluding that individuals seek out relationships with people who hold similar political identities (Huber & Malhotra, 2017, p. 282). Finally, more recent work has found that political agreement among couples has increased over time in tandem with political polarization and that geographic context has a modest effect on such agreement (Iyengar et al., 2018). The prevailing conclusion is that individuals prefer and select partners conditional on political characteristics or traits, and such congruence has increased over time. Thus, the perceived probability of mate success for individuals belonging to a political minority in regions dominated by political majorities is likely to take a hit. I argue that these partisans have an incentive to seek out niches (or choose niches when an opportunity presents itself) that would improve said prospects.

Such behaviors are not uncommon in nature, as multiple species have been documented seeking out new environments in hopes of improving their prospects of mate success (for a review of migratory mating strategies, see Buchan et al., 2020). Specifically, niche construction theory from evolutionary biology posits such behavior. It argues that individuals can modify their environment via their actions, which can alter selection pressures, including “when organisms relocate in space and encounter different selective environments by doing so” (Odling-Smee, 2012, p. 485; see also Laland et al., 2000). Accordingly, human migration in general, and group-based geographic sorting specifically, can be a strategy to alter one’s selection pressures and thus one’s probability of mate success. This theoretical foundation serves as the basis for the partisan niche construction argument presented here.

Partisan identity and negative affect

Partisan niche construction argues that partisan minorities are incentivized to leave, alter or compensate for partisan traits, or change the niche’s partisan composition. Thus, to the extent that out-partisanship serves as a heuristic for undesirable traits in prospective mates with said partisan identities, areas with high concentrations of in-partisans should repel out-partisans. Individuals of opposing political “phenotypes” ² should be expected to increasingly sort if there exists a perception of lower probability of mate success in out-party-dominated niches. In addition, I argue that this process is compounded by negative affect toward political adversaries, which is an added pressure affecting one’s perception of mate success.

Partisan identity in the United States is a broad social identity, and it is additionally associated with a host of other social affiliations that have become aligned with partisanship and more differentiated from each other (Green et al., 2004; Mason & Wronski, 2018). This within-party convergence/between-party divergence has coincided with the growing dislike and distrust of out-partisans (Iyengar et al., 2012; Iyengar et al., 2019) and of social groups associated with the out-party (Robison & Moskowitz, 2019). Since it has been shown that groups that are perceived as being more different are viewed more negatively (van Osch & Breugelmans, 2012), this should raise further concerns when geographic separation is incorporated, since, as I show next, may increase these differences. This is independent of the negative affect that leaders in respective niches can generate toward the out-group.

Partisan geographic boundaries

Literature on human cultural evolution views group and geographic boundaries as antecedents to the diversity that exists across human cultures around the world today (Foley & Mirazón Lahr, 2011). The implication is that when people sort into different geographic spaces, they begin to develop distinct norms, customs, and ways of living, which increase a group’s distinctiveness. “Human communities reproduce themselves over generations, and in doing so fission … [T]he rate of fissioning is strongly influenced by ecological and geographical factors … The result is the formation of boundaries between human communities; cultures are the consequence of these group boundaries” (Foley & Mirazón Lahr, 2011, p. 1080). I extrapolate this logic from kin-based groups to American partisan groups, which seem to be following not just a fissioning geographic pattern but one that extends to some lifestyle domains (Carney et al., 2008; Hetherington & Weiler, 2018; see Praet et al., 2022 for limitations) and, more importantly, fissioning on issues of morality, sexuality, and reproduction (Haidt & Hersh, 2001; Hatemi et al., 2017; Lakoff 2010; Petersen 2018). Thus, given the decreasing levels of out-party exposure and increasing in-party isolation across geographic levels in the United States from the state down to the residential area (Brown & Enos, 2021, p. 1002), it is not hard to see how a self-reinforcing process might ensue. Perceptions of distinctiveness increase negative affect (van Osch & Breugelmans, 2012), and this negative affect interacts with partisan composition, which I argue influences perceptions of one’s mate success and thus creates pressures for migration and sorting. As groups sort, they develop even more distinct cultures or ways of life, which further increases the distinctiveness, again increasing/sustaining negativity toward the out-group, and so on (see Figure 1). It would not be hard to imagine how incorporating leaders could further exacerbate the situation if and when they have incentives to do so.

Figure 1. Group sorting and reinforcing negative affect.

In the next section, I provide a model to illustrate the relationship between niche partisan composition and out-party affect on perceptions of mate success. Other factors that influence an individual’s perception of their probability of mate success are also illustrated in the model. Additionally, while this mate success pressure is not argued to be the predominant factor for sorting, it can be a substantial influence on the decision to move for certain individuals and an influence on where specifically one decides to move when options arise.

A model for perception of mate success

According to census data, the top reasons for moving within the United States are job related (16%–19% of movers), family related (26%–30%), or housing related (48%–51%) (Ihrke, 2014). Thus, moving is a function of both economic and noneconomic factors. The model that I provide does not argue that perception of mate success is necessarily the most significant factor in the decision to move. Instead, it argues that perception of mate success is an additional influence on the decision to move under specific conditions and on the specific location a mover chooses when opportunities arise. As mentioned earlier, the innovation here is going beyond social partisan homophily (e.g., “I like being near co-partisans”) to include political mate preferences more explicitly in potential geographic relocation.

The model

Let an individual’s out-party affect be defined as $ {a}_i\in \left[-\frac{1}{2},\frac{1}{2}\right] $ and their partisan compatibility in a niche as $ {c}_{ij}\in \left[0,1\right] $ . Affect can be negative, neutral, or positive, and compatibility gauges an individual’s partisan compatibility with the average prospective partner in a given niche. In niches where most individuals are from the in-party, an individual is closer to 1, and where out-party individuals dominate, an individual is closer to 0. Every individual has a certain political “phenotype,” which I assume is unidimensional $ \theta \in \left[0,1\right] $ . Thus, $ {c}_{ij}=1-\mid {\theta}_i-{\theta}_j\mid $ , where $ {\theta}_i $ is individual $ i $ ’s political phenotype and $ {\theta}_j $ represents the average political phenotype of prospective partner $ j $ in their niche. The further away an average prospective partner $ j $ is from individual $ i $ , the lower the compatibility.

Other relevant individual-level attributes that affect one’s mate success are captured by $ \left({\gamma}_i\right) $ which can include factors such as wealth, attractiveness, personality, and so on. In addition, $ \left({\pi}_i\right) $ represents an individual’s perceptual error. I define $ {\pi}_i\in \left[-\frac{1}{2},\frac{1}{2}\right] $ and $ {\gamma}_i $ $ \in \left[0,1\right] $ . Incorporating these together, I get the model for a partisan’s perceived probability of mate success in a given niche as $ pr\left({M}_{ij}=1\right)=\Phi \left({\pi}_i+{a}_i+{c}_{ij}+{\gamma}_i\right) $ . ³

The implications illustrated by Figure 2, ignoring other variables for now, are straightforward. Individuals in the dark bottom-left region of the mate success space within their niche are those who would gain by migrating to more compatible niches. On the other hand, those in the lighter regions are not as incentivized to move. An assumption to note is that the effect of compatibility and out-party affect are interchangeable with respect to their influence on perceived probability of mate success. That is, an individual who is mostly compatible with their niche but holds intense negative out-party affect has comparable perceived probability of mate success to someone low in compatibility with high positive out-party affect, all else being equal.

Figure 2. Mate success space.

Lastly, this model presents additional parameters that may help us explain other phenomena. While the story posited here is a niche push/pull story, the model may tell us why some individuals do not move or how they might behave. That is, individuals in darker regions might be most compelled to further improve $ \gamma $ , which serves to counteract decreased mate success probability from the two main variables, a and c. For example, individuals in the dark regions may be more likely to pursue greater wealth (and boost their $ \gamma $ ) to compensate for reduced partisan compatibility if wealth is a boost in mate success. One might then expect Republican individuals in large Democratic urban centers to be more financially ambitious and stay regardless of the overrepresentation of out-partisans. Thus, the model presented here could be applied to a variety of political questions to explain not just why people move, but why they might stay, and how they might behave in certain areas. Nonetheless, the key contribution for this project is that it makes clear what factors matter for a partisan’s perceived probability of mate success and under what conditions they are most incentivized to relocate.

Main expectations

To test some immediate implications of the model, I conduct conjoint experiments, which allows me to (1) replicate the findings on political mate choice preferences, (2) link perceptions of mate success to a niche’s partisan characteristics, and (3) gauge willingness to move. This would present suggestive evidence for partisan mate selection pressures as a force for some of the variation in partisan segregation. The six main expectations are as follows:

• • E1 (Out-party mate aversion): The probability of choosing someone as a mate when they hold an out-partisan identity is less than when they hold a co-partisan identity.

• • E2 (Out-party mate aversion × negative affect): For individuals who hold negative affect toward the out-party, the expected probability from E1 is further reduced.

• • E3 (Out-party niche decreases perceived mate success): One’s perceived probability of mate success in a niche decreases as out-partisan share increases.

• • E4 (Out-party niche decreases perceived mate success × negative affect): For individuals who hold negative affect toward the out-party, the expected probability from E3 is further reduced.

• • E5 (Out-party niche decreases willingness to move): The probability of saying one would move to a niche decreases as out-partisan share increases.

• • E6 (Out-party niche decreases willingness to move × negative affect): For individuals who hold negative affect toward the out-party, the expected probability from E5 is further reduced.

Design and methods

Four separate surveys were conducted, comprising just over 7,400 respondents who collectively made about 61,000 choices. Choosing among prospective neighborhoods, cities, and romantic partners are tasks in which individuals likely consider multiple attribute dimensions. Thus, implementing a choice-based conjoint design gives us a better chance of uncovering what matters to people when making such decisions. Some critics might correctly point out the artificiality of such choices in surveys, so I take some steps to attenuate this issue, though not all of them.

For the mate choice survey (Study 1), I do not create “dating app”–style profiles since dating profiles that mimic real dating apps, which include photos, would not allow for the sort of randomization of certain attribute values that I would like. For example, even validated images of attractiveness across different racial groups would be susceptible to bias. Instead, presenting numeric scores constrains the respondents to accept that such a profile is very attractive (e.g., 10/10) or not very attractive (4/10).

In the niche (neighborhood and city choice) surveys (Studies 2, 3, and 4), I give respondents a vignette that presents the niche choice as part of a decision in a job hiring process, which more closely resembles a real-world scenario. Specifically, respondents are told to imagine that they have recently been hired by a company that has multiple locations across niches (i.e., cities or neighborhoods), and they have to indicate which one (across multiple pairs) they would prefer. Studies 2 and 3 are at the city level and gauge both perception of mate success and willingness to move. Study 4 looks at willingness to move at the neighborhood level.

Study 1: Mate survey

The mate preference conjoint survey seeks to replicate existing findings on political mate choice but incorporates individual attributes that have been underexplored in previous studies. I ran the survey on an MTurk sample (N = 203; choices = 3,248) composed mostly of adults aged 18–45 (see the Supplementary Material). This is good for two reasons: first, the differences between MTurk samples and larger samples such as those from the Cooperative Congressional Election Study decrease for younger subsets of respondents (Huff & Tingley, 2015, p. 7), and second, this age range largely overlaps the active mating age range outlined in other work (Dancygier et al., 2022). As an additional measure, I incorporated the protocol outlined by Kennedy and colleagues (2020) to block non-U.S. respondents and those using a virtual private server attempting to mask their location. This is an effective way to block bots and bad-quality responses.

For the tasks themselves, my survey assigns respondents to eight pairs of hypothetical prospective partners whose attribute values vary uniformly, as suggested by Hainmueller and colleagues (2014). Respondents must choose one individual from each pair. Specifically, they are asked, “Which of the following two people do you find MORE DESIRABLE as a long-term romantic partner?” These hypothetical individuals have eight attributes (see Table 1; the full list of values is presented in the Supplementary Material). I limited the number of levels that each attribute can take for all surveys to three to four, since Schuessler and Freitag (2020) show that this increases power in conjoint experiments. Lastly, to prevent priming partisanship prior to the tasks, respondent covariates were collected after the tasks were completed. The potential for post-treatment bias is recognized (Montgomery et al., 2018), though without the opportunity to conduct a panel study, this is the best alternative. In this case, the dangers of priming by collecting covariates pre-treatment may outweigh post-treatment dangers since having a respondent indicate their affect toward the out-party before they choose from among prospective partners could inflate results.

Table 1. Example: Mate choice pair

Study 2: Niche survey (city level)

In the niche preference conjoint, the city attributes are altered such that the main independent variable of interest is partisan composition as proxied by previous election vote share. In constructing the survey, I follow a similar, although not identical, approach to Mummolo and Nall (2017), who used communities. Here, individuals are presented with a vignette in which they are told they have been hired by a company with multiple locations across many cities. They are then tasked with choosing which location they prefer, and they repeat this task multiple times. Rather than presenting people with cities that they could hypothetically move to without context, framing the task as a potential job location enhances the realism of the choice and increases the ecological validity (Morton & Williams, 2010, p. 265). This survey uses a separate MTurk sample (N = 206; choices = 2,060) in which respondents (1) indicate in which city they perceive to have a better chance at finding a long-term romantic partner and (2) choose the city they would prefer to move to for their job. Regarding the first question, they are specifically asked, “In which of the following two cities do you think you have a BETTER chance at finding a long-term romantic partner?” Respondents choose one city from each pair of a set of five pairs. Each city has seven attributes, as shown in Table 2. As with the previous survey, respondent covariates are collected after all choices are made.

Table 2. Example: City choice pair

Study 3: Niche survey (city level)

The third survey is a replication of Study 2 with a more representative sample of the United States that was implemented via Lucid (N = 6,018; choices = 48,144). Additionally, here I follow the design outlined by Acharya and colleagues (2018), where a given attribute of interest is present in the conjoint tasks for half the respondents and not presented in the conjoint tasks for the other half.

As with the previous study, the main interest is in estimating the effect of partisanship on perceptions of mate success (and willingness to move). However, I can further disentangle the mechanism by experimentally manipulating a potential mediator: racial/ethnic demographics. Specifically, by withholding the racial/ethnic demographics of a city for half the respondents and presenting it to the other half, I can estimate the “controlled direct effect” of partisanship on the outcomes of interest (Acharya et al., 2018, p. 372). In the absence of racial/ethnic information, partisan information contained in the profiles may activate respondents’ presuppositions about the specific niche’s racial/ethnic characteristics. Thus, by manipulating the exclusion of this attribute, I can estimate the “eliminated effect,” which represents how much the city’s partisan composition interacts with or is mediated by the racial/ethnic demographics of a city on the outcomes of interest. This quantity would then let us conclude whether racial/ethnic information plays a role in the effect of a niche’s partisan composition on perceptions of mate success or willingness to move. A negative coefficient on the interaction term of percent co-partisan and a binary indicator of whether racial/ethnic information was included would indicate the portion of the effect of partisanship that is “eliminated” because of inferred racial/ethnic characteristics indirectly or via an interaction. In other words, the larger the negative coefficient on the interaction term, the more it would indicate that partisanship has less of an impact on the outcomes when racial/ethnic information is included.

Study 4: Niche survey (neighborhood level)

The final survey is an altered version of Studies 2 and 3 in which niche preference is considered at the neighborhood level. I also incorporate other attributes such as commute time to work, community political engagement, school quality, religiosity, and the amount of single people in a neighborhood. This survey was also fielded via Lucid (N = 1,010; choices = 8,080). The point of this survey is to see how other attributes alter willingness to move with respect to a niche’s partisanship, without inducing mate perceptions that could inflate willingness to move by inducing a sort of consistency bias. Lastly, I test whether these effects are stronger for individuals with characteristics that indicate higher mate pursuit (i.e., single men; Clark et al., 1999; Hooper et al., 2011; Roney & Gettler, 2015) ⁴ by looking at the intercation with the amount of single people in a neighborhood and its partisanship. Namely, is an out-partisan neighborhood with mostly people in relationships the least desireable to individuals more likely to be in pursuit of a partner? This would indicate stronger evidence that mate acquisition is something that is, in and of itself, considered in opportunities for relocation.

Results

Studies 1 and 2 are similar regarding partisan identity, racial demographics, age, gender, and education. Studies 3 and 4 improve with respect to partisan makeup and non-White respondents (see Table 3). For a more complete summary of the descriptive statistics, see the Supplementary Material.

Table 3. Descriptive statistics

Additionally, before getting into the results, it is important to note that the estimated effects should not be interpreted as conventional effect sizes such as Cohen’s d, where, for example, an effect of 0.2 is considered a “small effect size.” Such an effect in a conjoint experiment represents a change of 20 percentage points in the probability of selection, which for binary choices is a Cohen’s d of 0.4 (Schuessler & Freitag, 2020, p. 6). Estimates for each attribute level represent its effect on the respondent’s probability of selecting a mate/niche with a given attribute level relative to the baseline level, after controlling for combinations of other attributes. Most of the analyses were conducted using the cregg software package (Leeper, 2020).

Study 1: Political mate preferences

As illustrated in Figure 3, the first main expectation is mostly supported. Being a Republican has a strong negative effect on the probability that a Democrat respondent will choose that prospective mate, and vice versa. For independents, both Republican and Democrat prospective mates are less likely to be chosen relative to an independent. Figure 4 shows marginal means of all attribute levels for better a comparison on a common scale. Here, one can more clearly see the relative negative and positive effect of party identity on mate choice. These effects are comparable to levels of attractiveness and salary (and race for independents/Republicans). An interesting and unexpected finding was the substantial preference overlap, for a Democrat or Republican respondent, when a prospective mate is a co-partisan versus when they are an independent. This indicates that independents are not negatively as affected by out-partisan selection pressures.

Figure 3. ACIEs by party identity.

Figure 4. Marginal means by party identity.

Regarding the second set of expectations for Study 1, I also find support for the interactive effect of negative affect. When I condition on a respondent having negative affect toward the out-party, a drastic difference arises from those without negative affect as shown in Figures 5 and 6. In fact, for the Democrat sample, the effects of different levels of party identity on mate choice overlap substantially for respondents without negative affect toward Republicans. For the Republican sample a similar pattern arises, though a slight preference for co-partisanship even among those who report having no negative affect toward Democrats emerges. In all, these results replicate previous findings, but disentangling by affect presents a nuanced picture in which negative affect toward the out-party largely drives partisan mate preferences, and for non-independents, there are similar levels of mate preferences between a co-partisan and independents.

Figure 5. Democrat ACIEs by affect toward Republicans.

Figure 6. Republican ACIEs by affect toward Democrats.

Study 2: Political niche preferences (city level)

In the second study, I find that Democrats are both averse to greater out-partisan composition and find greater in-party composition more appealing in prospective cities. This holds both for willingness to move (see the Supplementary Material) and where they perceive to have a better chance at finding a long-term romantic partner (Figure 7). For the Republican sample, there is a similar pattern, albeit weaker. Still, they are clearly averse to a greater out-party share. Independents seem to display similar preferences as Republicans for perceptions of mate success—that is, they perceive lower probability of finding a long-term romantic partner in cities with an 80% Democratic vote share. Interestingly, though, independents are largely indifferent with respect to partisan composition when choosing a city to move to for a job offer (see the Supplementary Material). ⁵

Figure 7. ACIEs by party identity.

Looking more closely at the Democrat and Republican subsamples with the baseline level as the out-party, the previous results are much clearer. For Democrats, I get an estimate of –0.24 (SE = 0.03, p <.001) for a city with a 80% Republican share relative to Democrat majority level. For the Republican sample, there is a comparable estimate of –0.20 (SE = 0.07, p < .005) for the 80% Democrat vote share attribute level. These results strongly illustrate that partisans perceive a lower probability of mate success in cities dominated by the out-party. When I further condition on out-party affect, it seems these effects are largely driven by those respondents who report negative affect toward the out-party, as illustrated in Figures 8 and 9. Thus, partisans seem to be much less willing to move to and perceive a lower probability of finding a long-term romantic partner in locations with higher out-partisan compositions. As with the findings from the mate choice study, the effect is stronger for respondents with negative out-party affect. In all, the main results from the first two studies present a novel and more nuanced picture of political mate/niche choice by considering variation along out-party affect. This also establishes for the first time a link between niche partisanship and perceptions of mate success.

Figure 8. Democrat ACIEs by affect toward Republicans.

Figure 9. Republican ACIEs by affect toward Democrats.

Study 3: Political niche preferences (city level)

The strongest evidence, however, comes from the larger sample in the third study. This replication saw several improvements in addition to the fact that the sample was larger and of better quality. For example, rather than presenting attribute levels as below, at, or above the national average for crime and housing prices, it presents specified quantities that represent low, average, or high rates, respectively. Specifying incidences of crime per 100 residents and housing price as a share of monthly income presents a more accurate representation of crime and housing costs. Additionally, I incorporate an experimental design that allows me to investigate the causal mechanism of partisanship, specifically with respect to the mediating or interactive effect of race/ethnicity.

The first set of results align with the results from Study 2, where Democrats and Republicans perceive a lower probability of mate success in high out-partisan niches, and this is magnified if the respondent has negative affect toward out-partisans (Figures 10 and 11). I also get similar patterns for willingness to move for both partisan subsets (see the Supplementary Material). Additionally, for the Republican sample, the effect of high non-White demographics has a significant negative effect on perceptions of mate success and willingness to move regardless of out-party affect. This is not found for the Democrat subset.

Figure 10. Democrat ACIEs by affect toward Republicans.

Figure 11. Republican ACIEs by affect toward Democrats.

Lastly, by investigating the “controlled direct effect” mentioned earlier, it is evident that providing racial information alters the effect of partisanship on mate success. This indicates that partisanship has an indirect or an interactive effect with race/ethnicity on perceptions of mate success. As shown in Table 4, high co-partisanship has a clear positive effect, but when racial/ethnic information about a city is included, partisanship becomes “less important,” as indicated by the negative coefficient on the interactive term. This is consistent with the interpretation that beliefs about a city’s racial/ethnic demographics to some extent mediate the effect of partisanship when race is not specified. Still, partisanship is clearly the predominant factor overall. In all, it is clear from the results that partisanship matters for perceptions of mate success and willingness to move, and the extent to which racial/ethnic characteristics mediate this effect.

Table 4. Eliminated effects

Notes: 80% out-partisan is the baseline. * p <.01.

Study 4: Political niche preferences (neighborhood level)

In this last study, I consider willingness to move at the neighborhood level and include a host of additional variables that could influence niche preference. Here, I reduced the levels of high out-partisanship from 80% to 70% and constrained non-White racial/ethnic demographics to not exceed 50%. This improves the extent to which the possible combinations of randomly generated profiles match real-world cases. For example, this gets rid of the possibility that a profile is 80% Republican and 80% non-White, which is highly unlikely in the United States. Additionally, this study isolates willingness to move to attenuate the possibility that willingness to move and perceived mate success are highly correlated due to consistency bias in the previous studies. In all, the results continue to indicate that partisan characteristics drive willingness to move, and that this willingness is stronger for partisans with negative affect toward the out-party (Figures 12 and 13).

Figure 12. Democrat ACIEs by affect toward Republicans.

Figure 13. Republican ACIEs by affect toward Democrats.

Finally, I look at the effects for individuals likely to be in more active pursuit of a partner—that is, single men. Specifically, I look at the interaction between being a single male with niche partisanship and number of available mates, to see whether greater availability of mates attenuates the effect of high out-partisanship. As shown in Figure 14, the effect of high out-partisanship remains the same across levels of mates available for nonsingle males, but for single males, more available mates in out-partisan dominated niches are preferred to those with people mostly in relationships. In other words, individuals most likely in pursuit of a partner view high out-partisan dominated niches with less available single individuals as the least desireable to move to. This indicates that the prospect of finding a mate, apart from a variety of other factors, influences individuals’ preferences across locales in and of itself. A similar analysis was done using data from Study 3, where I find that the effects of partisanship on perceptions of mate success are stronger for single men compared to nonsingle men, which serves as additional evidence (see the Supplementary Material).

Figure 14. Partisanship × Mate availability × Single male.

Are mate selection pressures enough for partisan segregation?

The foregoing evidence strongly indicates that the partisan composition of niches drives perceptions of mate success and willingness to move. Additionally, it is clear that these effects are larger for those who hold negative affect toward out-partisans. Yet, recent work looking at the extent to which co-partisan preferences more generally are the main driver of the current geographic segregation seem to indicate that other factors are more influential. Namely, research by Martin and Webster (2020) looks at the extent to which having politically congruent neighbors is the key driving force behind the partisan segregation pattern observed in the United States. Using voter registration data, the authors find that this pattern is partly due to sorting on nonpolitical neighborhood characteristics that covary with partisan preferences (also see Mummolo & Nall, 2017). Additionally, they run a simulation in which they find that the political preferences are five times too small to explain the current configuration. While seemingly contradictory to part of the story outlined earlier, this is not necessarily incompatible with the argument and evidence presented here.

More specifically, it has not been argued that the conditions for mobility are exclusive to pressures from partisan traits. Other factors, as illustrated by the census data and the conjoint tasks presented earlier, make it clear that there are other economic and noneconomic factors that encompass a variety of reasons for moving within the United States. Rather, the argument here underscores an underexplored but fundamental variable that adds pressure to relocate, the characteristics of whom would be most susceptible to this pressure, and under what conditions. That is, I make clear that when a prospective niche is composed largely of out-partisans, especially for those with negative out-party affect, perceptions of mate success are negatively impacted and willingness to move there less likely.

While this argument might thus seem narrow in scope (i.e., high out-partisan niches and for individuals with negative out-party affect), people with such characteristics can have substantial influence their political environments. Huddy and colleagues (2015) show that strength of partisan affect is linked to greater levels of current and future campaign activity. Thus, while “strong” partisans do not constitute a majority of partisans, they are more politically active, and thus potentially more influential in their local political environments. The argument here is not that partisan mate selection pressures (nor general partisan preferences) are entirely responsible for the extent of current partisan geographic homogeneity. Rather, partisan selection is an additional pressure, and individuals with the strongest incentives to move based on these pressures (who have proven characteristics for greater political action) are those most willing/likely to move.

Limitations and future work

While the evidence is consistent across a variety of samples, the studies discussed here only present support for the motivational basis of the theoretical model. An ideal test could incorporate “real-world” behavioral data that captures individuals’ partisan characteristics and their actual geographic movement. While I took some steps to improve the ecological validity of the studies here, the lack of behavioral data is the prominent limitation of this research. Future studies could seek out partnerships with dating app companies, as was done by Huber and Malhotra (2017), and incorporate geospatial data at multiple points in time. The researcher could investigate a host of additional outcomes such as “likes,” “matches,” and so on, to see whether partisans are actually “less successful” in out-partisan dominated niches while simultaneously measuring relocation patterns. Of course, such data (without experimental manipulation) would be riddled with issues that would complicate causal identification. Nonetheless, it would serve as an important complement to the work here which provides more control and clearer identification of the factors that influence the outcomes of interest.

Another avenue for future work could be the further integration of the causes of partisan homophily that have been identified in previous work by situating them in a mate success framework. The three causes are (1) choice-based homophily, as demonstrated by selection for (or against) certain traits; (2) social structure-based homophily, which arises from structural constraints on whom one is likely to come in contact with; and (3) convergence-based homophily, which is the tendency of partners to become more similar after meeting. The parameters introduced by the mate success model gives us a way to illustrate how all three processes may be occurring. Clearly, the results indicate that choice is key to homophily, as co-partisans are preferred to out-partisans, but in particular for those with negative out-party affect. If this selection pressure contributes (though not exclusively so) to partisan sorting, then structural constraints increase, which compounds with choice pressures, resulting in even more homophily. Additionally, convergence itself cannot be ruled out either, as the results here indicate that this aversion of out-partisans mostly holds for those with negative out-party affect. So, for individuals without negative affect toward the out-party, and given that other mate characteristics also matter (e.g., wealth), I can imagine how these alternative attributes may sometimes take precedence. It would not be inconceivable for changes in partisanship to follow if one finds a partner with appealing alternative traits. ⁶ Thus, by considering the role of out-party affect, and appealing nonpartisan mate traits, it is possible to integrate all three causes, as opposed to viewing any one of these as the cause to partisan homophily.

Lastly, in thinking beyond the U.S. partisan context, we could consider topics like human migration from a mate success perspective. That is, we could incorporate the role of mate selection pressures in the motivation for migrants to move, as well as the extent to which natives will resist newcomers. Prior work has illustrated that nativist attitudes toward immigration are influenced by prospective migrant attributes, such as language spoken, occupation, and gender, among others (Hainmueller & Hopkins, 2015). Additionally, recent work has found that in areas where the male-to-female ratio is greater, there is a greater probability of observing hate crimes toward refugees where native men are more willing to support such crimes (Dancygier et al., 2022). Thus, we could begin to theorize how, for example, men’s attitudes toward prospective single women migrants are less hostile than their attitudes toward women migrating with children, women migrating with partners, and other men. From the migrant’s perspective, a mate success model could help explain previous patterns, such as why, “historically, labor migrants were typically working-age men” and why single individuals are more likely to migrate relative to married individuals (Riosmena, 2009; Simpson, 2022). In all, human migration is an area whose existing (economic and noneconomic) theories can be more coherently integrated by considering nativist and migrant motivations within a mate success framework.

Discussion

These results illustrate strong empirical evidence for the motivational basis of the partisan niche construction argument. Study 1 replicates the findings from the literature on political mate choice using a more robust causal method and extends them. I find that out-partisanship has a negative effect on selection, that this effect is greater for respondents with negative affect toward the out-party, and that being an independent does not hurt one’s probability of selection. The last two findings are novel, though the latter unexpected. For Study 2, I find that partisans perceive a lower probability of mate success and are less willing to move to a city with higher shares of out-partisans. As expected, this effect is stronger for those with negative out-party affect. Study 3 successfully replicates the results from Study 2 with a better, more representative sample, and improved design implementation. Additionally, I disentangle the mechanism by estimating the mediating/interactive effect of race/ethnicity, showing that partisanship has a strong direct effect and a smaller indirect effect (via race/ethnicity) on perceptions of mate success. The final study isolates willingness to move at the neighborhood level having included additional attributes, where I find that partisanship is one of the predominant features driving willingness to move. I also look at the effect of mate availability and find that for single men this plays a key role, and using data from Study 3, I also find stronger effects for single men. This gives greater credence to the notion that prospects of mate acquisition is an additional factor in niche preference for those in pursuit of a partner. In all, the main expectations are supported, which indicates that partisanship is a key driver of perceptions of mate success and willingness to move. This is especially evident among individuals with negative affect toward the out-party. All of this points to the strong possibility of it being an additional factor in partisan sorting in the United States.

This project contributes to work on the consequences of mate pressures on political behavior. In doing so, it bridges instrumental and affective motivational elements. Previous work has pitted the affective (or expressive) motivational basis of political behavior against the instrumental (or rational) basis. Yet, the results here clearly draw a link between affect and instrumentality. While the instrumental approach has largely been characterized by individuals maximizing their utility function given specific policy preferences, economic evaluations, candidate/party performance, or other such factors, I argue that it can also encompass other domains. That is, the objective function need not be solely composed of economic preferences or policy ideal points in the strict sense, but should incorporate fundamental drives and consider a variety of other politically relevant evolutionary domains.

Instrumental ends could be something like mate prospects, which, as suggested by the work here and by Dancygier and colleagues (2022), can have clear political consequences. While tying psychological approaches to political science with more formal analytic tools is not a novel suggestion (Bendor et al., 2011; Diermeier & Li, 2019; Little et al., 2022; Wilson, 2011), it is certainly an underexplored one that I hope to have contributed to with this project. Affective or expressively motivated behaviors serve ends that might be maximizing some utility function, and a worthwhile goal might be to posit and investigate them. By grounding my argument in something as instrumental as mate selection and linking it directly to out-partisan affect, I hope to have taken a step in that direction. In all, this project introduces an important set of factors to consider given the partisan segregation and general social division that pervades the United States. To better understand this political division, I believe this fundamental dynamic (i.e., partisan mate selection) cannot be overlooked.