Introduction
Western democracies are diversifying as more people from the Global South settle in, increasing not only ethnic and racial diversity but also religious diversity. In the United States, since 1980, Mainline Protestants and Catholics lost congregants in terms of percentage of the population, while Evangelical Protestants and minority religions significantly grew their membership (Association of Statisticians of American Religious Bodies (ASARB) 2020).Footnote 1 Similar trends are noticeable in Europe where fewer people are religious than in the past, but minority religious groups are growing (Pew Research Center 2012).Footnote 2 These trends are expected to continue (Pew Research Center 2022b). However, surveys show that Americans and Europeans are not as accepting of minority religions as majority ones. Do these preferences affect the fundamental right of religious minorities to exercise their freedom of religion? This paper answers this question by examining the reasons for resistance to new houses of worship in the United States.
The United States is an ideal case for this study. The freedom of religion is especially taken seriously in the United States. It is listed first in the First Amendment of the U.S. Constitution, and 84% of Americans say that freedom of religion is extremely or very important to the U.S.’s identity, with another 11% saying that it is somewhat important (Fields and Deveaux Reference Fields and Deveaux2024). Furthermore, 53% of Americans say religion is very important in their daily life. Less than 20% of Europeans say so in most countries (Pew Research Center 2022a).Footnote 3
Americans should then support the establishment of new houses of worship, but they only sometimes do so. Why? Is it because some religious groups are disliked more? Or is the level of a nuisance the real cause for the opposition, as the not-in-my-backyard (NIMBY) literature suggests? Numerous scholars especially highlight opposition to Muslims and new mosques (Astor Reference Astor2011; Betz and Meret Reference Betz and Meret2009; Gravelle Reference Gravelle2021; Maussen Reference Maussen2004; Oskooii, Dana and Barreto Reference Oskooii, Dana and Barreto2021; Schnabel Reference Schnabel2023a; Verkaaik Reference Verkaaik2020). However, most researchers who study resistance to minority houses of worship do not compare the experiences of different religious groups. Research comparing different houses of worship is limited and shows conflicting results. Some find no evidence of discrimination against religious entities in land use (Hamilton Reference Hamilton2003; Clowney Reference Clowney2007) or bias against some religious groups (Clowney Reference Clowney2007). Others find hostility toward all religious buildings and discrimination against minority religious groups (Laycock and Goodrich Reference Laycock and Goodrich2012), especially Jewish and Muslim congregations (Miller Reference Miller2020).
Furthermore, most arguments against new mosques are not related to Islam. Instead, opponents discuss practical concerns such as lack of parking, increased traffic, noise, or light pollution (Beck Reference Beck, Shadid and van Koningsveld2002; Landman and Wessels Reference Landman and Wessels2006; Schnabel Reference Schnabel2023a; Verkaaik Reference Verkaaik2020). The nuisance level can then explain resistance to houses of worship, as the NIMBY literature suggests (Esaiasson Reference Esaiasson2014). The literature indicates that residents oppose extensive development in residential neighborhoods and facilities that attract crowds because they cause more nuisance than facilities that include open spaces and attract fewer people (Brown and Glanz Reference Brown and Glanz2018). This claim is supported by the frequent opposition to siting houses of worship in residential neighborhoods irrespective of denomination (Luney Reference Luney1957; Wehener Reference Wehener1993).
However, scholars have not systematically tested these explanations or tested them on houses of worship. Most of the NIMBY literature focuses on the siting of facilities that pose health or environmental risks or bring outsiders and undesirable populations such as drug users, people experiencing homelessness, or immigrants into the town. Minimal research has been done on resistance to facilities used by many residents or facilities not posing risks. This research introduces a new type of property to the NIMBY literature. New houses of worship are ideal for studying resistance to public facilities because they are permitted uses in residential areas, are susceptible to causing a nuisance, and, given their status as places of religion, are ideal for testing biases against various religious groups as a struggle over houses of worship reflects power balance in society and acceptance of religious minorities.
I systematically test the explanations from the literature by using a conjoint survey experiment run on 1953 respondents in the United States. Conjoint experiments allow for simultaneous testing of multiple factors, showing which factors affect the respondent’s decision the most (Bansak, et al. Reference Bansak2018). As such, they also reflect real-life decision-making, in which people must balance different attributes. This project employs a paired conjoint design. Respondents evaluated 10 pairs of applications for new houses of worship and supported or rejected each. Each application included information on the religion of the community, the level of nuisance (size, location, new construction), congregants’ outsider or insider status, and the reaction of neighbors and local officials to the application.
The results show that the religion of the new house of worship is the most important predictor of support. People are more likely to oppose facilities associated with minority religious groups irrespective of other factors such as the building’s size, location, or architecture. These results suggest that nuisance is a less important determinant of resistance to public facilities than opponents claim. Instead, people are biased against minority houses of worship, especially mosques, and will oppose them even when they cause a minimal level of nuisance. While respondents supported a Christian church 68.2% of the time, they supported a mosque only 51.6% of the time they had to decide on the application. Other houses of worship were less supported than the Christian church, but none less than the mosque. The level of nuisance, framing, or social influence did not show a significant effect, although respondents slightly preferred smaller houses of worship and facilities that other neighbors did not oppose.
Respondents’ partisanship and ideology also affect support, as Republicans and conservatives are much less accepting of minority houses of worship than Democrats and liberals. For example, while Republicans supported a new Christian church 72.7% of the time, they supported every other house of worship in less than 56% of cases. They supported a new synagogue in 55.6% of cases, a Mormon church in 54.1%, a Hindu temple in 45.8%, and a mosque only in 41% of cases. On the other hand, Democrats supported all houses of worship except for the Mormon church between 60% and 68% of the time. The Mormon church received support only 55.8% of the time.
This finding is important because most new houses of worship belong to minority religious communities. The number of congregations in the United States increased between 1998 and 2012 from 336,000 to 384,000, primarily due to the rise in nondenominational Protestant and non-Christian congregations (Brauer Reference Brauer2017). More specifically, between 2000 and 2020, Mainline Protestants lost 13,195 congregations and 8,779,999 adherents, and the Catholic Church decreased by 2,386 congregations and 176,905 followers. Meanwhile, Muslim congregations rose by 1,562 and added 2,894,614 adherents, Sikhs formed 96 new congregations, and Mormons 3,052 with an increase of 2,497,005 adherents (Association of Statisticians of American Religious Bodies (ASARB) 2020).
As houses of worship for religious minorities become more common, determining whether these communities face increased burdens when establishing new houses of worship is critical to mitigating discrimination and protecting their religious rights. Establishing a house of worship is a constitutionally protected right. It is also a part of the religious group’s emancipation process and helps members of religious minorities feel included in society. Opponents of new houses of worship often invoke property rights, which are also protected by federal and state constitutions. Nevertheless, this research suggests that residents are motivated more by bias than by the protection of property rights. Such reasons should not affect the zoning process. Thus, this research has important public policy implications and could lead to more equitable decision-making.
This paper starts by reviewing the history of resistance to minority religions in the Western world, with an emphasis on currently widespread Islamophobia. The second section also summarizes the NIMBY and public opinion literature and, from it, derived hypotheses. The following section describes conjoint experiments, focusing on the paired conjoint design used in this research (section III). Then, the paper analyzes the results, moderated by respondents’ partisanship and ideology (section IV), and their meaning and implications for zoning decision-making (section V). Finally, the conclusion (section VI) discusses the limitations of this research and possible avenues for future research, such as further exploration of elites’ impact on people’s decision to resist new houses of worship.
Houses of Worship, NIMBY, Threat, and Elite Discourse
Resistance to minority religions and houses of worship is not new in the Western world. Houses of worship are symbols of group inclusion and exclusion and balance of power, reflecting the broader political situation (Sunier Reference Sunier, Nell and Rath2009). Due to this, Jews, Catholics, and Protestants struggled toward emancipation in 19th-century Europe (Birnbaum and Katznelson Reference Birnbaum and Katznelson2014; Liedtke and Wendehorst Reference Liedtke and Wendehorst1999). For example, it took Jews in Switzerland a century to be accepted, and religious halal and kosher slaughter remains banned (Mayer Reference Mayer2011). Additionally, various houses of worship, such as synagogues, churches of Jehovah’s Witnesses, and other minority churches, were often excluded from residential areas and public life (Luney Reference Luney1957; Zoning Ordinances Affecting Churches 1984). While Protestants, Catholics, and Jews struggled to hold public religious services in the past, Muslims, Buddhists, and Hindus are going through similar struggles to gain equal rights today (Sunier Reference Sunier, Nell and Rath2009).
While Muslims, Buddhists, and Hindus are all currently going through the process of emancipation, Muslims appear to struggle more than other minority religious groups. Public opinion surveys portray Islam as the most disliked religion, arguably because it invokes cultural threats (Doyle and Ahmad Reference Doyle and Ahmad2013; Lajevardi Reference Lajevardi2020; Lenard Reference Lenard2010), while other minority religious groups, such as Sikhs or Buddhists, are perceived as more peaceful (Mayer Reference Mayer2011). Public opinion surveys show that Americans have warmer feelings toward Jews (67°), Catholics (66°), Mainline Protestants (65°), and Evangelical Christians (61°) than Buddhists (60°), Hindus (58°), Mormons (54°), and Muslims (48°) (Pew Research Center 2017).Footnote 4 A newer Pew study shows that the first four groups are viewed favorably rather than unfavorably, while the opposite is true for Mormons and Muslims (Tevington Reference Tevington2023). A YouGov survey also indicates that Christians are viewed the most positively (+34), followed by Protestants (+15), Jews (+11), Buddhists (+10), and Catholics (+10).Footnote 5 Some of the least-liked religious groups are Muslims (–24) and Mormons (–21) (Orth Reference Orth2022).
Similar attitudes are evident in the European Union, where people also view Muslims the least favorably (Wike, et al. Reference Wike2019). For example, when it comes to having a child in a relationship with a person, 81% would feel totally or moderately comfortable with their child being in a relationship with a Jew, 78% with a Buddhist, 92% with a Christian, 85% with an atheist, and only 68% with a Muslim (Kantar 2019).
Such hierarchy among minority religions is also palpable in intergroup interactions and opposition to houses of worship. Conflicts involving Muslims in the West, such as reactions to Rushdie’s Satanic Verses, a fatwa against the verses, the 9/11 terrorist attacks, Huntington’s Clash of Civilizations (Mayer Reference Mayer2011), President Trump’s so-called Muslim ban, and various terrorist attacks, gained significant media attention around the world. Opposition to Sharia law is also regularly discussed (Oskooii, Dana, and Barreto Reference Oskooii, Dana and Barreto2021).
Regarding houses of worship, Buddhist or Hindu houses of worship are rarely at the center of attention, but mosques appear in the media regularly and mostly with negative connotations. Research shows mosques draw more opposition than most other properties (Green Reference Green2010). They are opposed all over the West, especially in Western Europe (Betz and Meret Reference Betz and Meret2009) and the United States (Oskooii, Dana and Barreto Reference Oskooii, Dana and Barreto2021; Schnabel Reference Schnabel2023a; Schnabel Reference Schnabel2023b), although Australians also called for a mosque ban (Gravelle Reference Gravelle2021). Scholars particularly explore mosque conflicts in the Netherlands (Beck Reference Beck, Shadid and van Koningsveld2002; Gravelle, Medeiros and Nai Reference Gravelle, Medeiros and Nai2021; Landman and Wessels Reference Landman and Wessels2006; Maussen Reference Maussen2004; Verkaaik Reference Verkaaik2020), Spain (Astor Reference Astor2011; Astor Reference Astor2012), Switzerland (Mayer Reference Mayer2011), Italy (Green Reference Green2010), United Kingdom (Gale Reference Gale2004), France (Maussen Reference Maussen2009), Austria (Green Reference Green2010; Mayer Reference Mayer2011), and Germany (Green Reference Green2010; Jonker Reference Jonker2005; Schnabel Reference Schnabel2023a).
Scholars argue that opposition to mosques is rooted in Islam-specific measures rather than general ethnocentrism (Oskooii, Dana, and Barreto Reference Oskooii, Dana and Barreto2021). More specifically, mosques are opposed because they symbolize Islam, immigration, social problems, faulty multiculturalism (Astor Reference Astor2012), the threat of Sharia Law, infiltration, radical Islam (Oskooii, Dana, and Barreto Reference Oskooii, Dana and Barreto2021), and foreign culture incompatible with Western civilization and identity (Betz and Meret Reference Betz and Meret2009). In other words, “as symbols of foreignness, oppression, and political power, mosques are a focal point for debate over the place of Islam in modern Europe” (Green Reference Green2010, 631).
Worrying about identity and change is consistent with threat theory, which posits that majority groups feel threatened by minority groups, especially if the minority group size is large or increasing (Blalock Reference Blalock1967; Hopkins Reference Hopkins2010; Key Reference Key1949). People can fear losing economic and political power and material well-being or experiencing incompatible values and attitudes. Moreover, the threat does not need to be real. People only need to believe that the minority group threatens them (Stephan and Stephan Reference Stephan and Stephan1996). The actual or perceived threat can lead to negative attitudes or anti-minority policies (e.g., Hainmueller, Hopkins, and Yamamoto Reference Hainmueller, Hopkins and Yamamoto2013; Hopkins Reference Hopkins2010; Monogan Reference Monogan2013; Sniderman, Hagendoorn, and Prior Reference Sniderman, Hagendoorn and Prior2004; Ybarra, Sanchez, and Sanchez Reference Ybarra, Sanchez and Sanchez2016).
To gauge bias against various religious groups, the experiment discussed in this paper includes a Christian church representing a majority religion and four houses of worship of minority religious groups to determine whether some are more accepted than others. This literature leads to the first hypothesis:
H1a: People will support a Christian church the most and a mosque and a Mormon church the least, with a synagogue and Hindu temple in between.
Scholars argue that resistance is even more likely when houses of worship of minority religions are highly visible because they draw more attention and change the neighborhood’s character. Nativists want to make Islam invisible, which manifests itself as opposition to headscarves and mosques (Betz and Meret Reference Betz and Meret2009), and residents often argue that an Ottoman or otherwise oriental building would not fit the neighborhood. These arguments persist when a new mosque includes a minaret (Gravelle, Medeiros, and Nai Reference Gravelle, Medeiros and Nai2021; Green Reference Green2010; Verkaaik Reference Verkaaik2020). Opposition to mosques with minarets culminated in a minaret ban in Switzerland in 2009 (Mayer Reference Mayer2011; Pratt Reference Pratt2013). Muslim communities and local officials also “operate on the assumption that the visibility and distinctive presence of mosques in public space is the real problem” and, in many cases, push to reduce the size and visibility of the planned mosques to reduce the chance of conflict (Green Reference Green2010, 627). We can then add to the first hypothesis:
H1b: Respondents will especially reject non-Christian houses of worship with a traditional design for that religion.
If a change in neighborhood culture and character is worrisome for residents, it should also matter whether the congregants of the proposed house of worship already reside in the area. Furthermore, when a minority group is already located in an area, it likely means that the neighborhood has already witnessed a battle over the culture of the neighborhood and minority visibility, and conflict over a new house of worship of that minority group is then less likely (Landman and Wessels Reference Landman and Wessels2006). It then follows:
H1c: Respondents will resist minority houses of worship when the congregants are outsiders more than when they live already in the neighborhood.
Threat and bias are not the only possible explanations for varied responses to new houses of worship. Even in the case of new mosques, opponents often bring up practical concerns, such as traffic congestion, insufficient parking spots, noise, light pollution, and decreasing property values (Beck Reference Beck, Shadid and van Koningsveld2002; Landman and Wessels Reference Landman and Wessels2006; Schnabel Reference Schnabel2023a; Verkaaik Reference Verkaaik2020). Such concerns often come up during discussions of new facilities, whether religious or secular, and are chalked up to not-in-my-backyard (NIMBY) attitudes.
NIMBY literature, which dominates discussions of facility siting, can then also explain resistance to new houses of worship. Central to the NIMBY concept is a claim that people resist any public facility that presents a nuisance (i.e., inconvenience or annoyance), even though such a facility is considered valuable (Esaiasson Reference Esaiasson2014). Residents especially oppose extensive development in residential neighborhoods and facilities that host events and attract crowds because they cause more nuisance than facilities that include open spaces and attract fewer people (Brown and Glanz Reference Brown and Glanz2018). This is supported by the frequent opposition to siting houses of worship in residential neighborhoods irrespective of denominations (Luney Reference Luney1957; Wehener Reference Wehener1993; Zoning Ordinances Affecting Churches 1984) or opposition to houses of worship that are too large (Beck Reference Beck, Shadid and van Koningsveld2002; Schnabel Reference Schnabel2023a). The level of nuisance can then predict the level of resistance. While the public somewhat differentiates between the facility types, the difference in resistance largely disappears as the public becomes more informed about the projects, suggesting that local governments “will likely run into local resistance whenever affected citizens learn about their plans” (Esaiasson Reference Esaiasson2014, 193). This literature leads to the following hypothesis:
H2a: Houses of worship with high nuisance levels will face more opposition than religious houses with low nuisance levels.
Regarding location, there are also other reasons for resistance other than nuisance. Opponents often argue against placing minority houses of worship in central locations or locations significant to the community (Beck Reference Beck, Shadid and van Koningsveld2002). White and/or higher-educated groups tend to be located in the city center, while non-Christian religious centers concentrate on the outskirts of towns (Verkaaik Reference Verkaaik2020). Thus:
H2b: Minority houses of worship will be more supported in business parks than in downtown, and the opposite will be true for the Christian church.
While most scholars agree with the central NIMBY claim, the literature is diverse, explaining opposition to various facilities and coming to different conclusions. NIMBY literature across social sciences examines the reasons why people resist the opening of public facilities in their neighborhoods, such as prisons (Rasmussen Reference Rasmussen1992), renewable energy facilities (Ansolabehere and Konisky Reference Ansolabehere and Konisky2009; Petrova Reference Petrova2016), waste disposal facilities (Hunter and Leyden Reference Hunter and Leyden1995; Rasmussen Reference Rasmussen1992; Lober and Green Reference Lober and Green1994), needle exchange facilities (Davidson and Howe Reference Davidson and Howe2013), FEMA trailer parks (Davis and Bali Reference Davis and Bali2008), group homes (Clingermeyer Reference Clingermeyer1994), housing for people experiencing homelessness (Oakley Reference Oakley2002), and immigrant services (Maney and Abraham Reference Maney and Abraham2008–09). Most of these facilities pose health, environmental, or (perceived) security risks for the neighborhood or bring in undesirable populations such as drug addicts, prisoners, people experiencing homelessness, or immigrants. Only a few scholars explore attitudes toward amenities that most residents could use, such as recreational facilities (Brown and Glanz Reference Brown and Glanz2018).
The public views most of these facilities as necessary but resists their placement in their neighborhoods. The level of opposition depends on several individual and community-level factors. At the community level, the public is more likely to oppose the project if it poses a risk and the community does not benefit economically (Ansolabehere and Konisky Reference Ansolabehere and Konisky2009; Mansfield, Van Houtven, and Huber Reference Mansfield, Van Houtven and Huber2001; Rasmussen Reference Rasmussen1992). The public is also more likely to oppose the facility when it changes the neighborhood identity by bringing in drug addicts, prisoners, the homeless, or immigrants (Davidson and Howe Reference Davidson and Howe2013; Ben-Moshe Reference Ben-Moshe and Ben-Moshe2020; Maney and Abraham Reference Maney and Abraham2008–09).
Individual-level variables, such as distance from the facility, demographic characteristics, and personal views, also influence people’s responses to new facility siting. While the distance from the facility is one of the predictors of opposition in the NIMBY literature (Brown and Glanz Reference Brown and Glanz2018; Lober and Green Reference Lober and Green1994; Mitchell and Carson 1986), demographic variables can better predict resistance (Hunter and Leyden Reference Hunter and Leyden1995; Subiza-Pérez et al. Reference Subiza-Pérez2000). Female, older, more educated, and wealthier residents are more likely to oppose public facility sitting (Hunter and Leyden Reference Hunter and Leyden1995; Schively Reference Schively2007). More educated, older, and affluent people are also surer of their opposition and more likely to participate actively in the permit approval process (Mansfield, Van Houtven, and Huber Reference Mansfield, Van Houtven and Huber2001). People’s ideology and political views also influence opinions on zoning decision-making (Hunter and Leyden Reference Hunter and Leyden1995).
Regarding this research, political views are particularly relevant predictors of support. In the West, parties on the right and their supporters are less accepting of minorities in general and Muslims in particular. In the United States, Republicans are less sympathetic to minority groups, including religious ones, as they “pursue outsized power and benefits for dominant social groups” and “envision a Christian theocracy” (Kalmoe and Mason Reference Kalmoe and Mason2022, 2). Conservatives are also more likely to view Muslims negatively than liberals and more likely to oppose mosques (Oskooii, Dana, and Barreto Reference Oskooii, Dana and Barreto2021). Similar patterns are distinguishable in other Western democracies, where the parties on the right view Muslims more negatively than those on the left (Betz and Meret Reference Betz and Meret2009; Gravelle Reference Gravelle2021; Gravelle, Medeiros, and Nai Reference Gravelle, Medeiros and Nai2021). To illustrate, the Swiss conservative and right-wing party SVP/UDC campaigned against officially recognizing new religious groups, and the party also supported the 2009 minarets ban (Mayer Reference Mayer2011). Parties on the right often use Islam in their campaigns to gain electoral votes (Betz and Meret Reference Betz and Meret2009) and take on a crusade against mosques (Verkaaik Reference Verkaaik2020). Officials from right-wing parties also tend to support new mosques in the zoning process less than left-leaning officials (Landman and Wessels Reference Landman and Wessels2006; Schnabel Reference Schnabel2023a).
Public opinion also reflects these partisan differences. In the United States, Republicans feel much more affinity toward Christians than members of minority religions, and they differentiate among different religions much more than Democrats. To illustrate, a Pew survey from 2017 showed that Democrats feel the warmest about Jews (66°) and the coolest about Mormons (52°), a 14-point difference. On the other hand, Republicans feel the warmest about Evangelical Christians (71°) and the coolest about Muslims (39°), a 32-point difference. The following hypotheses reflect the partisan differences in the United States as well as the more broadly applicable effect of ideology:
H3a: Republicans will reject minority houses of worship at a higher rate than Democratic respondents.
H3b: Conservatives will reject minority houses of worship at a higher rate than liberals.
Relatedly, people do not make decisions or form opinions in a vacuum; elites and peers influence them. While the NIMBY literature acknowledges that trust in local officials and experts and longstanding community divisions affect opposition to facilities (Hunter and Leyden Reference Hunter and Leyden1995; Schively Reference Schively2007), it does not say much more about the impact of elites on community opposition. However, the public opinion literature shows that individuals form their opinions and act partly because elites and social groups influence them (e.g., Allport Reference Allport1954; Druckman Reference Druckman2004; Schattschneider Reference Schattschneider1957; Zaller Reference Zaller1992). Officials play a crucial role because they can enable and participate in conflict, or, on the other hand, they can prevent conflict (Landman and Wessels Reference Landman and Wessels2006; Schnabel Reference Schnabel2023a). Formulating a situation as a problem and shining a light on it activates opposition (Allport Reference Allport1954). Instead, “when elites uphold a clear picture of what should be done, the public tends to see events from that point of view” (Zaller Reference Zaller1992, 8). The public is predisposed to listen to elites who share their views (Druckman Reference Druckman2012; Zaller Reference Zaller1992). Frames thus enable and constrain what is possible in policymaking, including regarding the construction of new houses of worship (Maussen Reference Maussen2004). However, some find that negative information increases opposition to new mosques, but positive frames do not significantly lower opposition (Oskooii, Dana, and Barreto Reference Oskooii, Dana and Barreto2021).
Peers can also influence people’s opinions and behavior through information and peer pressure (Jost, Baldassarri, and Druckman Reference Jost, Baldassarri and Druckman2022; Sinclair, Reference Sinclair2012). Therefore, local officials’ discourse and behavior and neighbors’ attitudes might also affect residents’ attitudes toward new houses of worship, which leads to the last set of hypotheses:
H4a: Respondents will resist new houses of worship framed negatively at a higher rate than houses of worship framed positively.
H4b: Local officials will influence respondents more if respondents and local officials share partisanship.
H4c: Respondents will resist new houses of worship more often when their neighbors resist the new facility.
Experimental Design and Data
This research employs a conjoint survey experiment to explain why people resist new houses of worship. Conjoint experiments allow researchers to test several causal hypotheses simultaneously and thus mimic the decision-making processes in the real world, where people balance different attributes at the same time (Hainmueller, Hopkins, and Yamamoto Reference Hainmueller, Hopkins and Yamamoto2013). Conjoint experiments then provide more realistic results than experiments focusing on the effect of a single variable. As people evaluate multiple packages with randomly assigned attributes, they show their preferences for each attribute and the impact of each feature on their choice. Since the conjoint experiment uncovers the effects of individual variables on the same outcome, researchers can compare the impact of each variable on the outcome and, thus, evaluate the explanatory power of different theories (Hainmueller, Hopkins, and Yamamoto Reference Hainmueller, Hopkins and Yamamoto2013). Hainmueller, Hopkins, and Yamamoto (Reference Hainmueller, Hopkins and Yamamoto2013) mention another advantage of conjoint experiments that is especially pertinent to this research. When people balance different attributes, they do not have to be concerned about providing socially desirable answers since they have multiple ways to justify their decisions.
Conjoint experiments have different designs. Researchers can either force respondents to choose among two or more packages or ask respondents to rate the packages. Respondents rate either single or paired packages. This project employs a paired conjoint design because of its closeness to real-life results and lower levels of satisficing than in other conjoint designs (Hainmueller, Hangartner, and Yamamoto Reference Hainmueller, Hangartner and Yamamoto2015). In the paired conjoint design, respondents evaluate a pair of applications on each screen and decide whether to support or reject each application individually instead of choosing which application to support, as in the forced-choice design. The respondents in this experiment could then decide to support both applications, only one in each pair or neither. The outcome variable can be binary (e.g., support or reject) or have multiple values indicating the level of support. The dependent variable in this research is binary.
Lucid fielded this experiment in fall 2021. A pretest was run in September 2021 on 100 respondents, and the final survey was fielded in December 2021. After removing respondents who disagreed with the consent, did not finish the survey, or did not pass an attention check question, the final survey sample contains 1953 respondents. Detailed information about the survey, attention check question, characteristics of survey respondents, and wording of survey questions is located in Appendix A. The experiment was preregistered with AsPredicted, hosted by the University of Pennsylvania in May 2021.
In this project, respondents evaluated ten pairs of proposals for new houses of worship and decided whether to support or oppose each project individually (binary outcome, support = 1).Footnote 6 Figure A6 in the appendix provides an example of one of the screens with paired proposals. Each proposal presented to survey takers contained one level of each of the following features: name of the proposed house of worship indicating religion, congregants’ place of residence, location, size, and the architecture of the house of worship, local government control (i.e., partisanship), local officials’ attitudes, and neighbors’ reaction. All levels and features are listed in Table A7 in the appendix. All projects’ features were the same, but levels were fully randomized. Each feature had between three and eight levels within a recommended range (Qualtrics n.d.).
The features tested the previously mentioned hypotheses derived from the NIMBY, prejudice, and public opinion works of literature. The first three features tested whether people are prejudiced against outsiders and minority religious groups. The name of the new house of worship told respondents what religious group was planning to move to their neighborhood. To make the experiment more realistic, the houses of worship were named; they were Al-Salam Mosques, Shree Swaminarayan Hindu Temple, Synagogue Beth Shalom, St. John Christian Church, and the Mormon Church of Jesus Christ of Latter-day Saints. The congregants’ residence identified the religious group members as insiders or outsiders. The feature architecture could trigger the threat of foreign architecture when a non-Christian religious group planned to build a traditional house of worship. The house of worship could have been a converted residential property, a former church, or a traditional architecture for the religion.
Along with location and size, the architecture feature also tested whether a nuisance level affects resistance, as the NIMBY literature suggests. The architecture feature indicated the nuisance level as it told respondents whether the facility was an existing structure with proper infrastructure, an existing structure without the infrastructure to support a house of worship, or a new development. The house of worship could have been planned for a residential area, a residential area bordering a business district, a business park, or a downtown location, and it could have been small (fitting 100 worshippers), medium (500 worshippers), or large (1000 worshippers).
Finally, the experiment also contained three features testing the influence of local officials and community attitudes. First, the feature local government control told respondents what party controlled the local government, which mediated the effect of local officials’ attitudes. The feature local officials contained eight levels including positive, neutral, and negative frames. The neighbors’ reaction feature informed the respondents whether neighbors opposed the application, supported it, or did not get involved.
Results
The conjoint attributes were independently randomized since no combinations of attributes were excluded, and Figure A1 in the appendix shows that all levels were balanced in the experiment. The results are based on the analysis of responses from 1,953 respondents who completed the survey. Ten pages of two profiles on each page, which respondents evaluated separately, produced 39,060 observations (1953 × 10 × 2 = 39060). Due to the random attribute’s assignment, coefficients could be estimated in a simple linear regression with robust standard errors clustered on the respondent (Hainmueller, Hopkins, and Yamamoto Reference Hainmueller, Hopkins and Yamamoto2013).
The results are reported as marginal means.Footnote 7 Marginal means show the level of support for each attribute, that is, how likely people are to support a new house of worship with each attribute. Marginal means’ values range between 0 (a package with that attribute was never supported) and 1 (supported every time). The middle value of 0.5 indicates that a package has the same chance of support and rejection. Values below 0.5 imply that respondents intentionally chose to reject houses of worship with that attribute, more so than packages with attributes whose marginal means are above 0.5. Since this experiment employs a paired conjoint design in which respondents decide whether to support or reject each house of worship, the marginal means indicate an absolute level of support, not a preference between attributes.
Marginal means are also the preferred quantity of interest when analyzing subgroup differences (Leeper, Hobolt, and Tilley Reference Leeper, Hobolt and Tilley2020). Average marginal component effects (AMCEs), typically reported in conjoint experiments, cannot be used to compare subgroups because they change with different base categories, and subgroups might not have the same value of the base category. Thus, comparisons of subgroups’ AMCEs can lead to false conclusions. In other words, “while AMCEs do provide insight into the descriptive variation in preferences within-group and across-features, and conditional AMCEs do estimate the size of causal effects of features within groups, AMCEs cannot provide direct insight into the pattern of preferences between groups because they do not provide information about absolute levels of favorability toward profiles with each feature (or combination of features)” (Leeper, Hobolt, and Tilley Reference Leeper, Hobolt and Tilley2020).
Following conjoint experiment conventions (Hainmueller, Hopkins, and Yamamoto Reference Hainmueller, Hopkins and Yamamoto2013), the paper conveys results in figures. The appendix includes tables with coefficients and standard errors corrected for within-respondent clustering. Coefficients are denoted in the figures as points with 95% confidence intervals appearing as lines going to the sides of each point. The marginal mean of each attribute informs us how often respondents supported a new house of worship with that attribute on a scale from 0 (never) to 1 (every time). The midpoint (0.5) denotes indifference regarding the attribute.
Figure 1 displays the marginal means of all conjoint variables. Appendix B includes a table (B8) with coefficients and standard errors for the model. Since all attributes have coefficients higher than 0.5, respondents approved more applications than resisted regarding all attributes except for the mosque. The mosque’s confidence interval crosses the 0.5 line, meaning a mosque has an equal chance of being supported or rejected.
Marginal means of all attributes.
The approval level varies across the houses of worship and their attributes. In conjoint experiments, the factors with the most considerable differences in support are the factors that affect respondents’ decisions the most. Figure 1 shows that the religion of the group establishing a new house of worship is the most important predictor of support. Respondents supported a new Christian house of worship 68.2% of the time (SE = 0.01, p < 0.01), but a new mosque only half of the time (β = 0.52, SE = 0.01, p = 0.15). Respondents also showed lower support for the other houses of worship than the Christian church, but none less than the mosque. Respondents supported the Mormon church 53.8% (SE = 0.01, p < 0.01), the synagogue 58.8% (SE = 0.01, p < 0.01), and the Hindu temple 53.1% (SE = 0.01, p < 0.01) of the time they evaluated them.
Respondents also slightly preferred smaller houses of worship located downtown and projects that other neighbors did not oppose. The respondents supported small houses of worship 58.4% (SE = 0.01, p < 0.01) of the time and large ones 55.9% (SE = 0.01, p < 0.01), for a difference of 2.5 points. Respondents also preferred a downtown location (58.3%) compared to all other locations (ranging between 56.6 and 56.8% and both statistically significant with p < 0.01). Support or disengagement of other neighbors was also vital in reducing opposition. When neighbors actively supported the new facility, that had a similar effect on the support of a new house of worship as when no neighbors were involved in the approval process. However, when others in the neighborhood opposed the facility, respondents were likelier to oppose it. Respondents preferred houses of worship that other neighbors supported (57.7%, p < 0.01) or did not oppose (57.9%, p < 0.01) compared to when neighbors opposed the new house of worship (55.5%, p < 0.01).
The first figure also includes features of government control (GovControl) and the feature informing of local officials’ attitudes toward the new house of worship (Officials). While most of the Officials’ feature levels go in the expected direction, that is, people support houses of worship more when they are either supported or framed positively by the local officials, the differences are not statistically significant. One of the reasons for the lack of significant differences could be that the experiment also informed of the government’s partisanship, and the two features interact with each other and the respondent’s partisanship. However, even the interaction results (Figure C2 in the appendix) do not show a clear relationship between local officials’ attitudes and support for new houses of worship. They are, thus, discussed only in the appendix (see Appendix C).
The difference in support for new houses of worship grows even more prominent when we look separately at responses from Democratic and Republican respondents as well as conservatives and liberals. Figure 2 shows that Democrats are more likely to approve a house of worship with any attribute (estimates average 61.7%) than Republican respondents (estimates average 53.9%), except for the Christian church. Figure 3 shows that the differences between Democratic and Republican respondents are statistically significant in most cases. Republicans supported the Christian church more often than Democrats, and both supported the Mormon church about the same. However, Republican respondents supported a house of worship with any other attribute less than Democratic respondents.
Marginal means of nongovernmental attributes by partisanship.
Differences between Democrats and Republicans.
Thus, Figure 2 shows that Democratic respondents supported a Mormon church the least (55.8%, p < 0.01) and a Christian church the most (67.9%, p < 0.01), with a significant gap of 12.1 points. Meanwhile, Republican respondents approved the Christian church 72.7 % (p < 0.01) of the time regardless of the other attributes. On the other hand, Republican support for a mosque was only 41% (p < 0.01), a difference of 31.7 points. Thus, respondents of both partisanships were likelier to support a Christian church than other houses of worship. However, the differences in support of other houses of worship were starker for Republican respondents. The slight preference for smaller houses of worship downtown and not opposed by other neighbors remained true for both partisans.
Figure B3 in the appendix shows a similar divide between conservatives and liberals, although the difference in support is even more prominent.Footnote 8 Conservatives indicated almost identical support for houses of worship with each attribute as Republicans. However, liberals’ support for houses of worship with most of the attributes was higher than Democrats’ support. Like Republicans, conservatives averaged 53.9% of support across the attributes, while liberals averaged 65.1% compared to Democrats’ 61.7%. Responses also varied depending on the strength of ideological conviction, as Figure 4 shows. Slight liberals and slight conservatives showed similar support for houses of worship with most attributes, except the mosque, where the difference was almost ten percentage points. In fact, the preference differences between slight liberals and slight conservatives are much smaller than the differences between extreme and slight liberals and extreme and slight conservatives. The differences are especially substantial regarding the minority houses of worship, such as the mosque or Hindu temple. While slight liberals supported the mosque 58.8% of the time and slight conservatives 48.9%, extreme liberals showed much higher support (74.3%) and extreme conservatives much lower support (29.4%). The difference in support of a new mosque between extreme liberals and extreme conservatives was a whopping 44.9 points.
Marginal means of nongovernmental attributes by ideology.
So far, the religion of the house of worship was the most important predictor of support. Size, architecture, and location, the variables signaling the level of nuisance, were less important predictors of support for new houses of worship. However, it is possible that respondents cared about the nuisance level but tolerated nuisance differently regarding each house of worship. For this reason, the next set of figures displays interactions between the house of worship and each variable signaling the nuisance level.
The results of all three interactions in Figures 5–7 convey a similar story: size, architecture, and location do not significantly affect support for each house of worship. While Figure 5 shows a slight preference for smaller houses of worship that look like churches, the differences are not statistically significant, as conveyed by mostly overlapping points and confidence lines. Thus, Republicans and Democrats supported the Christian church the most, and the level of support was similar across sizes, locations, and architectures.
Marginal means of size*house of worship interaction by partisanship.
Marginal means of architecture*house of worship interaction by partisanship.
Marginal means of location*house of worship interaction by partisanship.
Discussion
The conjoint survey experiment helped to determine whether people resist new houses of worship because of the level of nuisance or bias. The experiment also tested the influence of elites and peers on people’s support of new houses of worship. The results clearly show that the religion of the group establishing the new house of worship is the most important determinant of support. The survey respondents, especially Republicans, displayed bias against minority houses of worship. This section evaluates the collected data against the hypotheses from section II.
The first hypothesis concerned bias against minority religious groups, foreign architecture, and outsiders. The first of the three hypotheses (H1a) stated that people would support a Christian church the most and a mosque and a Mormon church the least, with a synagogue and Hindu temple in between. The data supports this hypothesis, although the Hindu temple received the second lowest level of support instead of the Mormon church. As expected, respondents supported the Christian church more than other houses of worship, especially the mosque, Hindu temple, or Mormon church. Figure 1 shows that respondents supported the mosque only 51.6% of the time (p = 0.15), while they supported the Christian church 68.2% (p < 0.01) of the time it appeared in a package.
The data also shows that respondents’ partisanship and ideology affect the approval of different houses of worship. Providing support for hypothesis H3a, Republican respondents showed higher bias against minority houses of worship than Democrats. While all respondents strongly supported the Christian church, Democrats and Republicans differed significantly in supporting the other houses of worship. Republican respondents exhibited the most significant gap between support for the mosque and the Christian church. They endorsed the Christian church 72.7% (p < 0.01) of the time and the mosque only 41% (p < 0.01) of the time. Democratic respondents showed a smaller gap, supporting the Christian church 67.9% (p < 0.01) of the time and the Mormon church 55.8% (p < 0.01) of the time. The differences were similar for conservatives and liberals, supporting hypothesis H3b.
Furthermore, hypothesis 1b stated that resistance to minority houses of worship would grow when the non-Christian houses of worship presented a traditional architecture for that denomination, and hypothesis 1c stated that resistance would increase when congregants were outsiders to the neighborhood. The data does not support these parts of the first hypothesis. Respondents showed similar support across different architectures and did not differentiate between insider and outsider communities.
Thus, it appears that the name of the house of worship is sufficient to cue minority and outsider status, and it does not matter what the facility looks like or whether the religious community is already established in the town. This finding is significant because opponents often use both factors to justify resistance. For example, opponents frequently argue that Muslim groups should not establish new mosques when no members live in the neighborhood. They also do not want houses of worship that would not architecturally fit. However, this experiment suggests that insider status or architecture that would blend into the neighborhood might not prevent resistance because many people will oppose a mosque or another minority house of worship no matter what it looks like or whether the community members already live there.
The second set of hypotheses concerned the level of nuisance and preferred location for the new house of worship. Based on the NIMBY literature, respondents should have supported houses of worship that are large, new constructions, and located in a residential area less because such facilities are more disturbing than smaller existing facilities in business parks. The conjoint results partially support this hypothesis since all respondents supported small houses of worship slightly more often than large houses of worship. The percentage difference was 1.9 for Democratic respondents and 1.8 for Republicans. Respondents also supported new houses of worship in existing churches more often than in converted residences (difference for Republicans 3.7, Democrats 1.4), further supporting the idea that increased nuisance increases resistance. However, the differences in support among sizes or architectures are small, indicating that the attributes are less important predictors of resistance than the religion of the group establishing the new house of worship.
Regarding the location, I expected the highest support for houses of worship in business parks and the lowest support for houses of worship in residential areas because residents often worry about increased traffic, parking issues, and noise and light pollution near their homes. Minority worship houses should have also been more supported in business parks than downtown, and the opposite should have been true for the Christian church (H2b). However, Democrats showed almost the same support for houses of worship in residential and business zones, and Republican respondents indicated the lowest support for new houses of worship in business parks. Both partisans preferred the downtown location, possibly due to convenience. Since many Americans visit houses of worship regularly, people may want these facilities to be centrally located and convenient for most residents. Furthermore, respondents showed similar support for locations across the houses of worship, which means that respondents did not show higher support for convenient locations for preferred houses of worship, such as churches.
Finally, the last three conjoint features tested the impact of local officials’ and residents’ attitudes on respondents’ decision to support the new facility. Literature shows that people alter their opinions based on other people’s attitudes and discourses. Hypothesis 4a focused on elite influence, stating that respondents would resist new houses of worship framed negatively at a higher rate than houses of worship framed positively. Hypothesis 4b further specified that people would be especially responsive to cues from elites sharing respondents’ partisanship.
The influence of elites is complicated. The features specifying partisan control of the local government and local officials’ attitudes cannot be interpreted independently, and they are meaningful only when interacting with the respondent’s partisanship. Expected results would show that whenever the local official of the same party affiliation as the respondent supports the house of worship, the respondent is more likely to support the establishment and vice versa. However, the results in Figure C3 do not display such a relationship or any obvious pattern, as discussed in the appendix. The support for this hypothesis could be limited because only a few respondents saw the specific combinations of features or because the relationship also depends on what house of worship the application concerns. Based on the previously mentioned results and theory, people may take elite cues into account only when the cues do not conflict with people’s previously formed opinions. For example, Republicans may listen to their Republican local officials only when the officials reject a mosque or support a Christian church. It is also possible that people do not take partisanship into account in local politics as much as they do regarding federal issues. The impact of local officials on residents’ involvement in zoning issues will be explored more in future research.
In addition to elite influence, the literature suggests that people respond to their peers. Thus, the experiment also tested hypothesis 4c, which asserted that respondents would resist new houses of worship more often when their neighbors resisted the new facility. The experiment results show an apparent effect of neighbors’ response to the application. While neighbors’ support affected support of new houses of worship similarly to no involvement, respondents’ support for a house of worship decreased when neighbors opposed the new house of worship. The difference in percentages is similar for Republican and Democratic respondents, as well as conservatives and liberals. This finding suggests that people are susceptible to signs of concern and are more likely to resist whenever others actively oppose the establishment of a new facility.
Overall, the support for new houses of worship is a partisan issue and has much more to do with the religious group establishing the new house of worship than the characteristics of the house of worship, such as size, architecture, or location. Even cues from local officials and neighbors do not matter as much as what religious group is trying to establish the house of worship. Republicans are significantly more likely to support a Christian church than any other house of worship, substantially more than a mosque averaged over the joint distribution of the other features. Democrats do not show as tremendous difference in support among the different houses of worship as Republicans but still support the Christian church the most. The same is true for conservatives and liberals.
The experiment results help us better understand the reasons for facility siting opposition. Literature explaining resistance to new facilitates primarily focuses on the level of nuisance (Esaiasson Reference Esaiasson2014). Some authors mention changing neighborhood identity as a reason for resistance (Ben-Moshe Reference Ben-Moshe and Ben-Moshe2020; Davidson and Howe Reference Davidson and Howe2013; Maney and Abraham Reference Maney and Abraham2008–09) but do not compare it to other factors in single research. This research contributes to the literature by systematically testing possible explanations from different works of literature and including them all in one conjoint experiment. The results allow us to order the factors in terms of their impact on the decision to oppose new houses of worship.
The results show that bias is a much more potent explanation of resistance than any other factor in the experiment, including the level of nuisance. Furthermore, peers appear to have a greater impact on people’s decision-making than elites regarding facility siting. The results also show that results vary depending on the partisanship and ideology of the respondents. As expected, partisanship and ideology affect people’s attitudes toward facility siting (Hunter and Leyden Reference Hunter and Leyden1995; Landman and Wessels Reference Landman and Wessels2006; Oskooii, Dana, and Barreto Reference Oskooii, Dana and Barreto2021; Schnabel Reference Schnabel2023a), and Republican and conservative respondents are less supportive of minority houses of worship than Democrats and liberals. Republicans and conservatives also support any house of worship other than a Christian church less than Democrats and liberals.
The results also have practical implications for the zoning relief application process and local decision-making. Residents significantly impact local decision-making, especially regarding land use, since federal and state constitutions protect property owners’ rights and allow residents to participate in the zoning relief application process. Additionally, research suggests that local elected officials take constituents’ concerns and objections seriously. To illustrate, Esaiasson (Reference Esaiasson2014) shows that 23% of residents-protested projects in Gothenburg, Sweden, were denied permits by local officials. On the other hand, officials denied permits to only 3% of those applications with no resident opposition. Data from DuPage County in Illinois shows an even starker difference. Among applications from businesses and institutions, the DuPage County Board denied 41% of resisted applications but only 3.5% with no opposition from residents (Schnabel Reference Schnabel2023b).
Thus, since local officials listen to the public when deciding on zoning relief applications, knowing whether the public resists the applications for reasons protected by the law, such as protecting property values or living conditions, is crucial. This research provides evidence that the public resists minority group entrance into their communities and is much more likely to support a Christian church than a house of worship of a minority religious group. Local officials need to be aware that the public does not resist houses of worship only to protect their properties but also due to bias. Thus, when deciding on a zoning relief application from a minority group, local officials should discount public opposition and focus on the application itself. This research should guide local officials toward more equitable decision-making regarding all types of zoning applications involving minorities.
Conclusion
This paper focuses on public response to zoning relief applications. Using a conjoint experiment to understand the support for new houses of worship, I tested theories found in the NIMBY and public opinion literature to determine whether people resist applications from minority groups or whether nuisance motivates resistance. The results show that bias is a more important predictor of support than any other factor, including the level of nuisance. While the size and architecture of the house of worship and the reaction of other residents also matter, each attribute changes the chance of support only slightly. Thus, even large churches resisted by other residents have a lower chance of resistance than small mosques accepted by neighbors. Local officials need to consider these results when deciding on a zoning relief application from a minority group and focus more on the application and its compatibility with law and zoning ordinances instead of public opinion. Focusing on the application benefits the officials by reducing the amounts spent on legal challenges to their illegal or illegally influenced decisions.
While this paper shows a bias toward minority groups in local decision-making, it leaves some important questions for future research. One of the puzzling conclusions from the conjoint experiment is limited support for the theory that elites influence people’s opinions. The public opinion literature suggests that elites significantly impact people’s views, including in zoning decision-making. However, the experiment did not show a significant impact of local officials on people’s decision to support new houses of worship. The lack of support for the theory could be because respondents had to balance several attributes and did not find the attitude of local officials as important for their decisions as the information regarding the house of worship. However, it is also possible that local officials affect the process in other ways, or their influence varies depending on other factors. Thus, future research needs to focus on the impact of local officials in different situations.
Furthermore, the conjoint experiment includes responses only from Americans. However, other Western democracies also grapple with increasing religious diversity and applications for new houses of worship from religious minorities. Thus, the conclusions from this experiment should also be tested in other Western countries. Based on the literature, the findings should hold. At the same time, most Europeans do not view religion as an essential part of daily life and prefer to keep religion away from public space. Thus, all houses of worship may face more resistance in Europe. Future research should test this idea.
The results presented in this paper are also limited because they come from an experiment and might not correspond to people’s actions. For example, even if people claim they would not support the new house of worship, it is possible that they would not actively resist it if such a facility was proposed in their neighborhood. On the other hand, people may say that they would not oppose a new house of worship but may end up actively resisting it. Future research should test the conclusions from this experiment in the field and determine what attributes are more likely to transform attitudes into action.
Finally, local officials should consider people’s biases and focus on whether the applications align with law and zoning ordinances instead of listening to people’s complaints. However, that argument is based on the proposition that local officials are not biased and make unbiased decisions. Unfortunately, as previous research shows (Schnabel Reference Schnabel2023b), that is not always the case. Thus, future research should also more systematically examine whether local officials consider public complaints when deciding on zoning relief applications or whether their biases are more likely to affect them.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/rep.2024.22.
Acknowledgments
I am thankful for the support and valuable feedback on earlier versions of this manuscript from Alexandra Filindra, Amanda Sahar d’Urso, Amy Lerman, Geneva Cole, Arpita Ghosh, Stephanie Davis, Leanne Powner, E.J. Fagan, and others at the University of Illinois at Chicago that participated in various workshops where I shared parts of this work. I also thank fellow panelists and the audiences at the 2022 and 2023 Midwest Political Science Association meetings, the JREP editor, and three anonymous reviewers for providing constructive feedback.
Funding statement
The Institute for Research on Race and Public Policy at the University of Illinois at Chicago funded the conjoint survey experiment.
Competing interests
The author declares none.
