This study examines Muslim–non-Muslim disparities in locational attainment. We pool data from the 2004, 2006, and 2008 waves of the Public Health Management Corporation’s Southeastern Pennsylvania Household Survey. These data contain respondents’ religious identities and are geocoded at the census-tract level, allowing us to merge American Community Survey data and examine neighborhood-level outcomes to gauge respondents’ locational attainment. Net of controls, our multivariate analyses reveal that among blacks and nonblacks, Muslims live in neighborhoods that have significantly lower shares of whites and greater representations of blacks. Among blacks, Muslims are significantly less likely than non-Muslims to reside in suburbs. The Muslim disadvantages for blacks and nonblacks in neighborhood poverty and neighborhood median income, however, become insignificant. Our results provide support for the tenets of the spatial assimilation and place stratification models and suggest that Muslim–non-Muslim disparities in locational attainment define a new fault line in residential stratification.
The residential location of racial and ethnic groups has been of long-standing interest to urban scholars primarily because neighborhood characteristics simultaneously reflect and shape individual access to the larger stratification system (Lee et al. 2015; Massey 2008). The literature is replete with studies that compare minority and white access to neighborhoods that are whiter, have greater levels of income, have lower poverty, and are in suburbs (e.g., Alba and Logan 1993; Alba et al. 2014; Logan et al. 1996; Pais et al. 2012; South et al. 2008, 2011, 2016). This research on individuals’ locational attainment compliments the aggregate-level research on residential segregation because it has allowed scholars to document minority access to majority group members’ neighborhoods, but it has also moved beyond macro-level studies of segregation by offering an examination of the quality of the neighborhoods in which minorities and whites reside. The findings of these studies collectively point to a hierarchy of attainment, with whites accessing the whitest and best-quality neighborhoods, blacks occupying the least white and lowest-quality neighborhoods, and Asians and Hispanics falling in between (Alba et al. 2014; Friedman and Rosenbaum 2007; Logan 2014; Pais et al. 2012; South et al. 2008, 2011, 2016). This pattern of attainment is reflective of the shift from a biracial to triracial society identified by Bonilla-Silva (2004: figure 1), composed of whites (e.g., whites, new white immigrants (Russians, Albanians), assimilated white Latinos), honorary whites (e.g., light-skinned Latinos), Asian Indians, Chinese, and Middle Easterners), and the collective black (e.g., blacks, dark-skinned Latinos, and reservation-bound Native Americans).
Surprisingly absent from the literature on locational attainment has been a focus on the residential attainment of Muslims. Since the terrorist attacks on 9/11 and continued intermittent violence perpetrated by Islamic terrorists (e.g., in 2016, in the Pulse nightclub of Orlando, FL), Muslims have increasingly become a racialized group in American society, whereby they have been clearly stigmatized as an out-group that is potentially threatening and demarcated physically by their outward religious symbols (e.g., beard or head covering) and to a lesser degree by their skin color (Braunstein 2017; Byng 2008; Considine 2017; Garner and Selod 2015; Selod 2015). As such, Muslims have been compared with other racial and ethnic out-groups in American society, such as blacks and Hispanics, and to a lesser extent, with other religious groups, like Jews (Edgell et al. 2006, 2016). In 2017, there were 3.35 million Muslims in the United States, accounting for roughly 1 % of the American population, having increased by 43 % from 2.35 million in 2007 (Pew Research Center 2017).1
Consistent with this view of Muslims as a racialized out-group, data from the American Mosaic Survey collected in 2003 and 2014 show an increase in the share of Americans, from 26 % to 45 %, reporting that Muslims “do not agree at all with their vision of American society,” with Muslims surpassing atheists, gays, conservative Christians, recent immigrants, Hispanics, Jews, and African Americans as the most distrusted group (Edgell et al. 2006, 2016). Similarly, the percentage of Americans who do not approve of their child marrying a Muslim increased from about 34 % in 2003 to nearly 50 % in 2014, surpassing all the aforementioned groups as the least desirable in terms of marriage (Edgell et al. 2006, 2016). In addition, in 2014, 22.1 % of Americans viewed Muslims as a threat to public order and safety, relative to 12.6 % of Americans viewing African Americans in the same manner, making Muslims the most feared out-group (Edgell et al. 2016).
Given this animus toward Muslims relative to other racial, ethnic, and religious out-groups, a dearth of literature has examined the extent to which Muslims experience residential inequality relative to non-Muslims. This absence is in no small part due to the lack of large-scale demographic data on religious groups in the United States. Thus, our study’s main contribution is to fill this gap in the literature. Philadelphia provides a good case study for our research for several reasons. Reflecting national trends, Muslims in Philadelphia account for just over 1 % of the population (Pew Research Center 2016), and the metropolitan area ranks fourth in the nation in terms of the number of mosques (with 73 mosques) and seventh in terms of the number of Muslim attendees at mosques (Grammich et al. 2012). The group’s visibility is significant, as reflected by the fact that the Philadelphia School District approved adding two Muslim holidays to its calendar in the 2017–2018 school year, one of the first districts in the nation to do so (Nadolny 2016).
An additional strength of using Philadelphia as a case study is that a majority of Philadelphia’s Muslims are black (Dent 2016). At the national level, in contrast, only 20 % of Muslims are black (Pew Research Center 2017). Given that Muslims have become a racialized group, focusing on Philadelphia can allow us to examine Muslim blacks and nonblacks, and we can begin to observe their location in the racial/ethnic hierarchy of residential attainment relative to non-Muslim blacks and nonblacks. In that way, we can assess the extent to which both race and religion shape residential attainment.
Given that blacks continue to face significant disadvantages in their residential attainment (Sharkey 2013), examining how Muslim blacks fare relative to non-Muslim blacks will allow us to determine whether Muslim blacks are significantly more disadvantaged than non-Muslim blacks and whether they fall at the bottom of the “collective black” triracial category (Bonilla-Silva 2004). We can also compare the residential outcomes of Muslim blacks and nonblacks. Muslim blacks are distinct from Muslim nonblacks in at least two ways. Among native-born black Muslims, who make up the majority of black Muslims in the United States, 96 % report much discrimination against Muslims, a rate higher than among all Muslims (75 %) and blacks (59 %) in the United States (Pew Research Center 2017). Two-thirds of U.S.-born Muslim blacks joined their religion through conversion, compared with 21 % of Muslim nonblacks, which means that Muslim blacks’ roots overlap much more with non-Muslim blacks than with Muslim nonblacks (Pew Research Center 2017). Muslim nonblacks, especially those who are foreign-born, tend to have less experience and understanding about the discrimination faced by blacks in American society (Husain 2017). By comparing Muslim blacks with non-Muslim blacks as well as with Muslim nonblacks, this study will help further contextualize the contemporary racialization of Muslims.
Using a unique data set collected biennially in the Philadelphia metropolitan area—the Public Health Management Corporation’s Southeastern Pennsylvania Household Surveys conducted by the Public Health Management Corporation (PHMC)—our main goals are to (1) examine the extent to which Muslims achieve the same levels of locational attainment (i.e., at the neighborhood level: percentages white, black, and in poverty; median household income; and residence in suburbs) as non-Muslims, among blacks and nonblacks; and (2) if such differences exist, to examine whether they remain after we include socioeconomic and demographic controls. In addition to allowing us to identify whether respondents are Muslim,2 another advantage of using these data is that they include census tract–level identifiers. We pool data from the 2004, 2006, and 2008 waves of the PHMC survey and merge them with census tract-level data from the 2005–2009 American Community Survey (ACS), adhering to the convention of previous research examining the locational attainment of racial and ethnic groups (see, e.g., Firebaugh and Farrell 2016; South et al. 2016). We are unaware of any other data that contain the religious identification, race/ethnicity, and census tract indicators of survey respondents, allowing for the examination of the neighborhood outcomes of Muslim and non-Muslim blacks and nonblacks.
Locational attainment is conceptualized similarly to status attainment (Blau and Duncan 1967). The underlying assumption is that neighborhoods, like occupations, are hierarchically ordered. How individuals translate their individual and household characteristics into their current neighborhood of residence or employment in their current occupations—particularly those neighborhoods or occupations dominated by the majority group in society—is the main question of interest. In the locational attainment literature, two theoretical perspectives—the spatial assimilation and place stratification models—are used to characterize the inequality among racial and ethnic groups in access to the neighborhoods within this hierarchy (Charles 2003). We apply these models to characterize the locational attainment of Muslim and non-Muslim blacks and nonblacks.
The spatial assimilation model maintains that the residential distribution of households across neighborhoods of varying neighborhood quality is influenced by household socioeconomic status (SES), acculturation, and demographic factors (Alba and Logan 1991, 1993; Massey 1985). Those who accrue more human capital and financial resources seek to bring their residential status in line with their improved economic status and therefore tend to live in neighborhoods that are whiter and thereby have more access to majority group members, are of higher SES, and are in suburbs versus neighborhoods with fewer economic resources. Studies have consistently shown that differences in income and education contribute to residential disparities between racial and ethnic minorities and whites (Alba et al. 2014; Logan et al. 1996; Pais et al. 2012; South et al. 2008, 2011). However, for blacks and Hispanics, variation in income and educational background does not completely erase their inability to access neighborhoods that are whiter, higher in SES, and suburban, relative to their white counterparts (Pais et al. 2012; South et al. 2008, 2011).
Nativity status is also a critical factor inherent to the spatial assimilation model. Foreign-born individuals tend to reside in neighborhoods with fewer whites and more coethnics, in neighborhoods with lower SES, and in central cities relative to their native-born counterparts, until such immigrants spend more time in the United States and acquire greater English language proficiency (Alba and Logan 1991, 1993; Charles 2003; Logan et al. 2002; Massey 1985). According to the model, residential parity in locational attainment should be achieved by the time the children of immigrants acquire independent residences. Studies have shown that nativity status shapes locational attainment in the manner predicted by the model (Alba et al. 2014; Cort 2011; Friedman and Rosenbaum 2007; South et al. 2008).
The locational attainment of Muslims and non-Muslims has not been the explicit focus of research to our knowledge, but one study compared the locational attainment of Arabs with whites, blacks, Hispanics, and Asians in four metropolitan areas in 1990 and 2000 (Holsinger 2009). This study found that compared with whites, Arabs (whose national origin is identified via the ancestry question in the decennial census) reside in neighborhoods with significantly lower shares of whites and are more likely to be in suburbs, but no difference exists in terms of the neighborhood median household income. The main problem with this study is that it did not explicitly examine Muslims, and the majority of Arabs are not Muslim (Arab American Institute 2008). In addition, the study did not account for the racial and ethnic diversity of the Muslim population. This research is also limited because the study (1) relied on the decennial census’s ancestry question to identify Arabs, which is prone to measurement error; (2) used data from nearly two decades ago; and (3) defined neighborhoods using public use microdata areas, which are larger than typical neighborhoods.
Another related strand of research has focused on the association between religious affiliation and housing outcomes (Dilmaghani 2017; Keister 2003, 2008), but only Dilmaghani (2017) explicitly examined Muslim–non-Muslim differences in these outcomes. Using Canadian data, Dilmaghani found that Muslims are significantly less likely to own their homes than non-Muslims—a finding attributable to the facts that they don’t speak the native language and are less likely than non-Muslims to use mortgage lending, consistent with the tenets of the spatial assimilation model. However, and contrary to the previous finding of lower homeownership rates for Muslims than for non-Muslims, Dilmaghani found that Muslims who own their homes have housing values that are 8 % higher than those of non-Muslims. Clearly, more research is needed to understand residential inequality between Muslims and non-Muslims.
The tenets of the spatial assimilation model can be applied to explaining Muslim–non-Muslim differences in locational attainment. As Dilmaghani (2017) illustrated, nativity status is likely to be one of the most important factors putting Muslims at a disadvantage in their locational attainment relative to non-Muslims. The Pew Research Center (2017) collected demographic data on Muslims and the general public in the United States. In 2017, 58 % of Muslim respondents reported being born abroad, compared with 18 % of the U.S. general public. Socioeconomic status could also play a role, but it is less clear. Relative to the U.S. general public, Muslims tend to have the same education levels, but a slightly greater share of Muslims report incomes below $30,000 (Pew Research Center 2017). However, all this information is based on Muslims in the United States and not specifically those in Philadelphia. According to the tenets of the spatial assimilation model, we hypothesize that Muslim residential disadvantages, relative to non-Muslims, among blacks and nonblacks observed in Philadelphia, will reflect Muslim–non-Muslim differences in nativity status as well as socioeconomic and other demographic factors. Controlling for such differences should attenuate Muslim–non-Muslim residential disparities among blacks and nonblacks.
The significance of structural constraints in maintaining racial and ethnic inequality in residential location has given rise to a second theoretical model—the place stratification model—to guide the work on minority-majority group residential disparities (Alba and Logan 1991, 1993; Logan and Alba 1993; Logan and Molotch 1987). The model maintains that household access to the best residential opportunities is constrained by the actions of powerful groups and structural factors that allocate housing opportunities unequally on the basis of race and ethnicity or other characteristics, such as religious affiliation, that distinguish subgroups as minorities. Majority groups use their power to maintain social and physical distance from minority groups (Logan and Molotch 1987). This power is often evident in various forms of discriminatory actions, which constrain minority housing choices and cause them to be residentially segregated (Massey and Denton 1993; Turner et al. 2013; Yinger 1995).
American Muslims experience prejudicial treatment and discrimination in the United States. The Pew Research Center (2017) found that 48 % (up from 40 % in 2007) of Muslims reported having experienced at least one of the following incidents of discriminatory treatment: (1) called offensive names; (2) singled out by airport security or other law enforcement officials; (3) physically threatened or attacked; or (4) experienced people acting suspicious of them. A Pew Research Center (2013) survey of U.S. Jews revealed that compared with 47 % of the U.S. general public, 72 % of U.S. Jews believed that Muslims experience much discrimination; at the same time, only 24 % of the general public and 43 % of U.S. Jews reported similar levels of discrimination against Jews. With respect to atheists, Catholics, and evangelicals, 24 %, 17 %, and 30 % (respectively) of the U.S. general public believes that these groups experience much discrimination (Pew Research Center 2013). Thus, Muslims are perceived by the general public as experiencing the most discrimination, above all other religious groups, and even among Jews, who have perceived themselves as a racialized out-group in American society.
Recent civil rights reports published by the Council on American-Islamic Relations (CAIR 2009, 2017) show an upward trend in the number of both civil rights complaints and hate crime complaints filed by Muslims. CAIR (2009, 2017) reported that the number of civil rights complaints filed by Muslims between 2007 and 2016 increased by 62 %, from 2,652 to 4,282 complaints. During the same period, the number of anti-Muslim hate crime complaints nearly doubled, from 135 to 260 complaints (CAIR 2009, 2017). Hate crimes against Muslims are also evident in the Philadelphia area. For example, in December 2015, a pig’s head was discovered in front of the Al-Aqsa mosque in North Philadelphia (Dent 2016).
How have such general sentiments of hate, prejudice, and discrimination translated directly into housing discrimination against Muslims? Only two studies have examined housing discrimination against Muslims in the United States (Carpusor and Loges 2006; Gaddis and Ghoshal 2015). In a field study performed over 10 weeks in 2003, just before and during the beginning of the war in Iraq, landlords in Los Angeles County were three times more likely to discourage an applicant with an Arab-sounding name from visiting their apartment than an applicant with a white, American-sounding name (Carpusor and Loges 2006). Similarly, in a 2014 field study of online roommate ads conducted in four metropolitan areas (Los Angeles, New York, Detroit, and Houston), requests by those with Arab-sounding names were only about 57 % as likely as those with white, American-sounding names to receive a response (Gaddis and Ghoshal 2015). In a study in Toronto, Hogan and Berry (2011) revealed that housing discrimination for Muslims/Arabs was significantly higher than that experienced by whites in the rental market, but Jews did not differ from whites in the treatment they experienced.
Examining data from the National Fair Housing Alliance (NFHA 2017) revealed that although housing discrimination against Muslims exists, few actual complaints are made on the basis of religion. In fiscal year 2017, of the complaints made to private fair housing agencies, the U.S. Department of Housing and Urban Development (HUD), HUD–state equivalent agencies, and the U.S. Department of Justice (DOJ), only 0.8 %, 2.3 %, 2.5 %, and 5.0 %, respectively, alleged discrimination on the basis of religion (NFHA 2017). However, NFHA (2017) reported that complaints of housing-related hate activity or harassment on the basis of religion, race, national origin, or sexual orientation accounted for 23 % of their total complaints, and the share of complaints would likely be even higher because many cases are referred to local or federal enforcement agencies rather than fair housing organizations. In 2016, the most likely place where an anti-Muslim bias incident took place was in a residence or home (CAIR 2017). Such housing-related hate activity and harassment, which appear to have increased since the 2016 presidential election, are prohibited under the federal Fair Housing Act (NFHA 2017). To understand and effectively combat these housing-related challenges faced by the Muslim community, NFHA has recently joined forces with the American-Arab Anti-Discrimination Committee (NFHA 2017).
In sum, given the experiences that Muslims have had with discrimination generally and housing discrimination more specifically, as well as the increase in civil complaints that Muslims have filed and hate crimes against them at their homes, it is likely that prejudice and discrimination against Muslims is a way that powerful groups constrain minority locational attainment relative to non-Muslims. Thus, if residential disparities exist between Muslims and non-Muslims, after demographic and socioeconomic characteristics are controlled for, the tenets of the place stratification model suggest that these differences are due to factors such as discrimination and prejudice that disadvantage Muslims in terms of their locational attainment. We would expect that Muslim blacks would experience a double disadvantage in their locational attainment based on their race and religious background; on the other hand, non-Muslim nonblacks should experience a superior position in the hierarchy of locational attainment.
Data and Methods
This study focuses on the Philadelphia metropolitan area. Our data come from the 2004, 2006, and 2008 waves of the PHMC and the 2005–2009 ACS. The PHMC is a biennial telephone survey that has been administered for more than 30 years in the Philadelphia metropolitan area in the counties of Bucks, Chester, Delaware, Montgomery, and Philadelphia. The PHMC surveys are representative of the Philadelphia metropolitan area and are collected using a stratified sampling frame, via random digit dialing methodology. Data from the PHMC have been comparable with demographic data collected in other surveys of the Philadelphia metropolitan area, including the Behavioral Risk Factor Surveillance System (Yang et al. 2011). Table A1 in the online appendix compares the PHMC in our analytical data set with data from the 2005–2009 ACS, aggregated for the aforementioned five counties that make up the Philadelphia metropolitan area. The racial and ethnic composition in the PHMC is comparable with that in the ACS despite a slight overrepresentation of non-Hispanic whites in the PHMC data. The percentage of foreign-born individuals is similar between the two data sets. The PHMC data slightly overrepresent more-educated individuals (i.e., those with a college degree or more). With respect to the unemployment rate and median income, the two data sets are comparable.
Although the PHMC survey focuses mainly on questions of health, it also asks questions about social, demographic, and economic characteristics. Because Muslims account for a small share of the Philadelphia population, we must pool waves of the PHMC to increase our sample size of Muslims. The respondent’s religion was not asked in the most recent waves of the survey in 2012 and 2014/2015. Therefore, we pool the data from next most recent waves: 2004, 2006, and 2008.
All respondents come with the identifier of their census tract, which allows us to merge data from the ACS with the individual-level data from the PHMC surveys. Following the convention adopted in other studies of locational attainment, we use census tracts as proxies for individuals’ neighborhoods (e.g., South et al. 2011). In all, 956 census tracts within the Philadelphia metropolitan area are covered by the PHMC surveys. The 2004, 2006, and 2008 PHMC data contain census tracts in 2000 geographic boundaries. We merge these data to census tract–level data from the 2005–2009 ACS, which are in the same boundaries.
Missing data are a concern in our data. In particular, 20 % of respondents are missing values on the income variable. Approximately 2.4 % and 1.8 % are missing values on the variables religion and race, respectively, which are the main indicators of interest in our analysis. To avoid the potentially harmful impact of such missing data on multiple variables, we employ multiple imputation procedures, which we discuss more fully later (Rubin 1987).
We examine the locational attainment of blacks and nonblacks separately, regardless of Hispanic origin. Black Muslims account for a large share of the Muslim community in Philadelphia (Dent 2016), and blacks are typically the most disadvantaged group in terms of their locational attainment (Alba et al. 2014; Logan 2014; Pais et al. 2012; South et al. 2008, 2011, 2016). Blacks comprise all blacks; nonblacks comprise whites, Asians, and those who identify as “other” races. Because of the small number of nonblack Muslims in our data, we cannot disaggregate nonblacks into separate races (e.g., whites or Asians).
Our study has several key dependent variables. We include a number of neighborhood characteristics to gauge individual locational attainment, following the convention of previous research (e.g., Pais et al. 2012; South et al. 2008). These measures, from the ACS, include the percentage non-Hispanic white, the percentage non-Hispanic black, the percentage in poverty, and median household income.3 We also use a dummy variable indicating whether the respondent lives in a suburban tract as an outcome given its long-standing role as an indicator for the pinnacle of achievement in residential attainment (Alba et al. 1999). Tracts in the four counties outside of Philadelphia County, which encompasses the city of Philadelphia, make up the suburbs in this study. Although suburbs have experienced racial and ethnic diversity in many places in the United States, in the four counties that comprise the suburbs in our study—Bucks, Chester, Delaware, and Montgomery—non-Hispanic whites account for 89 %, 84 %, 75 %, and 82 % of the populations, respectively. Although census-tract population density and land use may vary within counties demarcated as central city versus suburban, defining suburbs based on tract-level characteristics is beyond the scope of this article. We use the definition that has been traditionally used in previous research (e.g., Alba et al. 2014; Logan et al. 1996), including a recent study finding that suburbs are disproportionately white (Massey and Tannen 2018).
Our key predictors for this study come from the PHMC. Our primary predictor gauges whether one self-identifies as Muslim. To create an indicator of whether the respondent identifies as Muslim, we use the PHMC survey question, “What is your religious affiliation?” We also include controls for demographic and socioeconomic characteristics. Demographic characteristics include age, marital status, gender, number of children in the household, and nativity status. Respondent’s self-reported age is included in the analysis as a continuous variable. Marital status is a dichotomous variable (1 = married, 0 = not married). Gender is assessed as whether one identifies as male. Number of children is included in the analysis as a continuous variable. Nativity status is a dichotomous variable (1 = foreign-born; 0 = native-born). Our data do not contain any information on the birthplace of respondents’ parents, thereby making it impossible to examine generational differences in neighborhood outcomes. This prevents us from fully testing all of the tenets of the spatial assimilation model. However, most studies examining the locational attainment of racial and ethnic groups have not had data to examine generational differences and, like our study, have not controlled for differences in nativity status to ascertain whether group differences remain (see, e.g., Alba and Logan 1993; South et al. 2011).
Socioeconomic status is measured with three variables. First, family income is an ordinal variable gauging total family income in the calendar year before the survey. The four waves of data have 19 levels of income, with similar intervals between the levels across the periods. Second, educational attainment has five levels and is represented by four dummy variables: without a high school diploma (reference group), high school graduate, some college education, a bachelor’s degree, and a graduate degree. Third, employment status is categorized into five groups and is represented by four dummy variables: unemployed (reference group), full-time employment, part-time employment, retired, and other employment status. Key socioeconomic variables not present in this study but worth mentioning are homeownership and duration in the housing unit, recognized as key variables in residential attainment studies (e.g., Logan et al. 1996). These variables unfortunately were not available in the waves of the PHMC used in this study. We recognize that these omissions may inhibit the predictive power of our findings. Finally, we use suburban status as an independent variable in our analyses of the other dependent variables.
We conduct bivariate and multivariate analyses of these data. Bivariate analyses are used to determine how Muslims compare with non-Muslims, among blacks and nonblacks, with respect to locational attainment outcomes as well as individual-level demographic and socioeconomic variables. Our bivariate analyses are presented for the complete data and exclude cases via listwise deletion if they have missing data on any of our dependent or independent variables.
Arbitrary patterns of missing values seen on multiple variables are known to impose a threat to statistical inferences (Allison 2001; Rubin 1987). Inference by multiple imputation is used to allow a full use of the observed data and to formally incorporate uncertainty due to missing data into our inferences. Multiple imputation routines take into account the data structure and multivariate relationships. Specifically, we use the R package, jomo, based on the methodology developed by Quartagno and Carpenter (2019). This package facilitates Markov chain Monte Carlo (MCMC) techniques to draw multiple imputations in multilevel applications for mixtures of variables subject to missing values on multiple observational Level 1 or Level 2 units, which fits our data structure well. The underlying imputation model is a joint model with normal latent variables for the categorical variables. Ten imputed data sets are drawn from the posterior predictive distribution implied by this imputation model. Each of the imputed data sets is then analyzed using the multivariate analyses described in the next paragraph, leading to 10 sets of estimates and standard errors. These standard errors explicitly incorporate the variability in the predictive distribution of missing data (hence uncertainty due to missing data) into the traditional estimate of sampling variability. Using the rules defined by Rubin (1987), we combine these estimates and standard errors, and they are presented later in our multivariate analyses. SAS PROC MIANALYZE is used for this purpose.
We use ordinary least squares (OLS) to identify the association between Muslim religion and the locational attainment of blacks and nonblacks—defined as the census tract–level percentage non-Hispanic white, percentage non-Hispanic black, percentage in poverty, and median household income—while controlling for demographic and socioeconomic variables. Logistic regression is used for the analysis of suburban residence. In both analyses, we use robust standard errors to correct for potential autocorrelation of the results due to clustering of respondents in tracts. We use sampling weights (scaled down to maintain unweighted cell sizes) to correct for sampling design effects and potential under coverage. Note that given the cross-sectional nature of these data, we cannot draw any inferences about the causal relationships between our independent and dependent variables.
How does Muslim religion affect the locational attainment of blacks and nonblacks in the Philadelphia metropolitan area? Table 1 addresses this question, presenting the means for our main dependent variables and focusing on comparisons between Muslim and non-Muslim respondents. Our results show that Muslims experience more residential disadvantages than their non-Muslim counterparts, regardless of race. Muslim–non-Muslim disparities are substantively larger among nonblacks than among blacks, but all differences are statistically significant among both groups. Among blacks, the average percentage of non-Hispanic whites in Muslims’ neighborhoods is 7 percentage points lower than that in non-Muslims’ neighborhoods. Among nonblacks, although the average share of non-Hispanic whites is greater in their neighborhoods, the disparity between Muslims and non-Muslims is roughly 22 percentage points (56 % vs. 78 %, respectively). Not surprisingly, the mean percentage black is greater in blacks’ neighborhoods compared with nonblacks’ neighborhoods. Among blacks, the disparity in the average level percentage black between Muslims and non-Muslims is more than 4 percentage points, compared with 18 percentage points among nonblacks.
Regardless of race, Muslims are more likely than non-Muslims to live in neighborhoods with a lower economic level. Table 1 reveals that the average poverty rate in Muslim blacks’ neighborhoods is 29.2 %, compared with 24.8 % in non-Muslim blacks’ neighborhoods. Among nonblacks, the mean neighborhood poverty level of Muslims is double that of non-Muslims. The average median household income of Muslim blacks is about $4,000 less than that of non-Muslim blacks. Among nonblacks, the disparity in average median household income between Muslims and non-Muslims is about $17,000.
Among both racial groups, Muslims are much less likely than non-Muslims to live in suburbs. Table 1 shows that the Muslim–non-Muslim disparity in suburban residence is larger among nonblacks than among blacks, consistent with the results for the other indicators of locational attainment. Among blacks, 17.8 % of Muslims live in suburbs, compared with 24.8 % of non-Muslims. Among nonblacks, 48.7 % of Muslims and 72.7 % of non-Muslims live in suburbs. Taken together, the results in Table 1 reveal that Muslim religious background is significantly associated with more disadvantaged residential outcomes among blacks and nonblacks.
Variation in demographic and socioeconomic characteristics likely contributes to the residential disparities observed between Muslim and non-Muslim blacks and nonblacks in the Philadelphia metropolitan area. Table 2 reports the results from bivariate analyses of these characteristics. Several differences likely contribute to these residential disparities. Muslim blacks and nonblacks are significantly younger in age than non-Muslim blacks and nonblacks, respectively, averaging 10 to 12 years younger, which could mean that they have less experience in the housing market than non-Muslims. These data are consistent with national-level data on Muslims collected by the Pew Research Center (2017). Regardless of race, Muslims are significantly more likely to have a greater number of children in the home than non-Muslims. Not surprisingly and falling in line with national-level data (Pew Research Center 2017), Muslims are significantly more likely to be born abroad than born in the United States. Among blacks, 14.9 % of Muslims are foreign-born, compared with 6.2 % of non-Muslims; among nonblacks, Muslims are seven times more likely to be foreign-born than non-Muslims. Such differences undoubtedly contribute to differences in their locational attainment because their lack of time in the United States and knowledge about the housing market as well as their limited English proficiency put them at a disadvantage. Muslim blacks are significantly more likely to have a male householder than non-Muslim blacks, but the sex difference is not significant among nonblacks.
In addition to Muslim–non-Muslim differences on a variety of demographic characteristics that are likely important in explaining their residential disparities, Muslims in the Philadelphia metropolitan area tend to have significantly lower levels of SES than non-Muslims among blacks and nonblacks. Muslim blacks and nonblacks have significantly lower levels of income than non-Muslim blacks and nonblacks, respectively. This could affect their ability to move to residences in higher-quality neighborhoods. The pattern in income is somewhat consistent with national-level data finding that Muslims are more likely to report having incomes lower than $30,000 than the U.S. population overall. However, at the national level, the share of Muslims with incomes of at least $100,000 does not differ from the share of the general public (Pew Research Center 2017).
With respect to education, Muslim blacks are significantly less likely to obtain graduate-level education than non-Muslim blacks. The patterns observed in education, particularly for nonblacks, are similar to national-level patterns showing that Muslims overall had relatively equal educational attainment relative to the U.S. population (Pew Research Center 2017). Turning to employment status, regardless of race, Muslims do not significantly differ from non-Muslims in terms of working full-time. Among blacks, Muslims are significantly more likely to work part-time than non-Muslims. Among blacks and nonblacks, Muslims are significantly more likely to be unemployed than non-Muslims, and they are significantly less likely to be retired than non-Muslims, which is consistent with the demographic findings on age discussed earlier.
When we control for the variation in demographic and socioeconomic characteristics, do the disparities in locational attainment between Muslims and non-Muslims remain among blacks and nonblacks in the Philadelphia metropolitan area? Tables 3 and 4 show the results of our multivariate analyses that address this question. Table 3 presents the analyses focused on the racial and ethnic composition of the neighborhoods of Muslim and non-Muslim blacks and nonblacks, which speaks directly to the ability of Muslims to reside near majority group members versus minority group members, relative to non-Muslims. Table 4 presents the results of models that compare the neighborhood SES and suburban location of Muslim and non-Muslim blacks and nonblacks.
The results in Table 3 reveal that after relevant demographic and socioeconomic characteristics are controlled, Muslim blacks and nonblacks have significantly lower shares of white population in their neighborhoods and significantly greater representations of black population relative to non-Muslim blacks and nonblacks, respectively. Column 1 reveals that the percentage white in Muslim blacks’ neighborhoods is, on average, 4.2 percentage points lower than in non-Muslim blacks’ neighborhoods. Column 2 reveals that among nonblacks, the Muslim–non-Muslim gap in average percentage white in the neighborhood is slightly larger, at 5.5 percentage points, net of the effects of other predictors. The results for the mean neighborhood percentage black reveal a similar pattern. When key demographic and socioeconomic variables are controlled, as shown in column 3 of Table 3, Muslim blacks are more likely than non-Muslim blacks to live in neighborhoods with significantly greater representations of blacks—a difference of 3.9 percentage points. The average percentage black in Muslim nonblack neighborhoods is nearly 11 percentage points higher than that in non-Muslim, nonblack neighborhoods, which is also significant net of the effects of the other control variables. In comparing Table 1 with Table 3, it is evident that the gaps in the average neighborhood percentage white and percentage black between Muslim and non-Muslim blacks and nonblacks reduced in size, but significant disparities remain, consistent with hypotheses derived under the place stratification model.
The results in Table 4 reveal that the Muslim–non-Muslim disparity in the average SES of neighborhoods becomes insignificant after relevant individual-level demographic and socioeconomic predictors are controlled. Column 5 of Table 4 reveals that the Muslim–non-Muslim disparity persists for blacks in their access to suburbs, after relevant predictors are controlled. The odds of living in suburbs are roughly 0.7 times greater for Muslim blacks than for non-Muslim blacks. However, the Muslim–non-Muslim disparity in suburbanization is eliminated among nonblacks (column 6).
How do the demographic and socioeconomic characteristics used as control variables in our multivariate analyses relate to the locational attainment of Muslim and non-Muslim blacks and nonblacks? In general, the results are consistent with hypotheses derived from the spatial assimilation model and suggest that demographic and especially socioeconomic characteristics are associated with individual locational attainment in the Philadelphia metropolitan area. For example, in almost all the models in Tables 3 and 4, marital status is significantly associated with neighborhood outcomes, with other factors controlled. Table 3 shows that among nonblacks, married couples are more likely than nonmarried couples to live in neighborhoods with significantly greater shares of whites and lower average shares of blacks in their neighborhoods. Table 4 reveals that among blacks and nonblacks, married couples generally have significantly lower levels of poverty in their neighborhoods, have higher levels of neighborhood median income, and are more likely to live in suburbs than their nonmarried counterparts, net of the effects of other predictors.
The family income and educational attainment variables are significantly associated with locational attainment in all models in Tables 3 and 4. Among blacks and nonblacks, the coefficients for family income reveal that the effects of family income are positively associated with neighborhood percentage white, median household income, and location in suburbs; they are negatively associated with neighborhood percentages black (among nonblacks) and in poverty. The effects of having a bachelor’s degree and a graduate degree are significantly and positively associated with the representations of whites in the neighborhood, neighborhood median income, and whether respondents live in suburbs, with other predictors controlled. However, the effects of these educational variables are significantly and negatively associated with the neighborhood representations of blacks and neighborhood levels of poverty.
Nativity status is another demographic variable that exhibits significant associations with most of the locational attainment outcomes in Tables 3 and 4. However, the direction of the relationship between nativity status and locational attainment varies between blacks and nonblacks, which is consistent with tenets of the segmented assimilation model (Portes and Rumbaut 1996; Rosenbaum and Friedman 2007; South et al. 2005). Among blacks and net of the effects of other predictors, foreign-born individuals live in neighborhoods with significantly lower shares of blacks and lower levels of poverty, and are more likely to live in suburbs than native-born individuals, which is consistent with other research (e.g., Friedman and Rosenbaum 2007; Rosenbaum and Friedman 2007). However, among nonblacks, foreign-born individuals live in neighborhoods with lower representations of whites, greater representations of blacks, higher poverty levels, lower median household income; and are less likely to live in suburbs than native-born individuals.
Discussion and Conclusions
The primary goal of our study was to determine whether Muslims experience residential disadvantages, relative to non-Muslims, among blacks and nonblacks in the Philadelphia metropolitan area. Muslims have become a racialized group in American society, but little research has examined Muslim–non-Muslim differences in residential outcomes. Our study addresses this gap by pooling data from the 2004, 2006, and 2008 PHMC merged with census-tract level data from the ACS. Our descriptive analyses reveal that significant Muslim–non-Muslim residential disparities exist in the Philadelphia metropolitan area. Both black and nonblack Muslims are significantly more likely than their non-Muslim counterparts, respectively, to live in neighborhoods with lower shares of whites, greater representations of blacks, higher levels of poverty, and lower median income, and they are less likely to be in suburbs. Our multivariate analyses show that among blacks and nonblacks, significant Muslim–non-Muslim differences persist in the neighborhood percentages white and black; among blacks, Muslims and non-Muslims differ significantly in access to suburbs. However, the Muslim–non-Muslim disparities in neighborhood poverty and median income are not significant.
Taken together, the results provide support for tenets of the spatial assimilation model. We find that Muslim–non-Muslim residential disparities are attenuated with the inclusion of demographic and socioeconomic variables; for blacks and nonblacks, the differences between Muslims and non-Muslims in terms of neighborhood poverty and median household income are eliminated. Both of these findings suggest that in the Philadelphia metropolitan area, Muslim–non-Muslim disparities in terms of neighborhood poverty level and median income are likely explained, in part, by the lower SES of Muslims (relative to non-Muslims) and by the greater share of Muslims who are immigrants (relative to non-Muslims), particularly in the case of nonblacks. These findings are similar to those found in previous research focusing on white–nonwhite disparities in residential outcomes (Alba et al. 2014; Logan et al. 1996; Pais et al. 2012; South et al. 2008, 2011).
With respect to the association between individual-level variables and neighborhood residential attainment, the findings for the socioeconomic characteristics and one of the demographic variables are consistent with hypotheses under the spatial assimilation model, although the findings for nativity status are not. In general, higher levels of SES, measured by family income and higher levels of education, are significantly associated with more advantaged residential outcomes for individuals, similar to results in other studies of locational attainment (e.g., Alba et al. 2014; Pais et al. 2012; South et al. 2008, 2011). With respect to the demographic variables, only marital status is consistently related to the residential outcomes adhering to patterns expected under the spatial assimilation model (Alba and Logan 1991, 1993; Charles 2003; Massey 1985).
Controlling for demographic and socioeconomic variables, we find that black and nonblack Muslims are significantly more likely than their non-Muslim counterparts to live in neighborhoods with greater shares of blacks and lower shares of whites; black Muslims are significantly more likely than black non-Muslims to live in central cities. These findings suggest that the place stratification model is also important in characterizing Muslim–non-Muslim locational attainment. More powerful groups are likely using their privileged position in society to maintain their social and spatial distance from Muslims (Alba and Logan 1991, 1993; Logan and Alba 1993; Logan and Molotch 1987; Massey and Denton 1993). What is particularly troubling about our findings is that even among blacks—a group that continues to experience significant disadvantages in locational attainment—Muslim–non-Muslim residential disparities exist in terms of neighborhood racial composition and in their access to suburban location. Thus, the racialization of Muslims puts blacks at even more of a residential disadvantage than they experience by their race alone, suggesting that they fall at the bottom of the “collective black” category (Bonilla-Silva 2004).
In the multivariate analyses, the findings that Muslim–non-Muslim disparities in neighborhood racial composition persist but that disparities in neighborhood median household income and poverty do not may also point to the salience of the tenets of the place stratification model. The segregation of blacks from whites in the Philadelphia metropolitan area is characterized as hypersegregated and has been hypersegregated since 1970 (Massey and Tannen 2015; personal e-mail communication with D. Massey on October 22, 2018). An examination of the income segregation of whites and of blacks in the Philadelphia metropolitan area between 2007 and 2011 produces an h index of .118 for whites and .112 for blacks, indicating low levels of income segregation for both groups (Reardon et al. 2018). Whites and blacks of varying incomes live in the same neighborhoods, which could yield fewer differences in the median income and poverty levels between whites and blacks in comparisons of their locational attainment. Thus, the significance of race in the residential stratification system in Philadelphia seems to outweigh the significance of income.
The same discriminatory forces that perpetuate hypersegregation between blacks and whites, as implied by the place stratification model, could contribute to the Muslim–non-Muslim disparities in the neighborhood racial composition and the lack of differences between these groups in terms of neighborhood median income and poverty. In our sample data among nonblacks, although 45 % of Muslims are white, 35 % are Asian, and 19 % are other races, Muslims are significantly more likely than non-Muslims to live in blacker neighborhoods and less likely than their non-Muslim counterparts to live in whiter neighborhoods, suggesting the operation of discrimination. Our study suggests that the anti-Muslim sentiment that has been documented elsewhere is likely impacting Muslims’ residential choices, particularly the racial composition of their neighborhoods. This finding is consistent with the limited research focusing on housing discrimination against individuals with Arab names (Carpusor and Loges 2006; Edgell et al. 2006, 2016; Gaddis and Ghoshal 2015; Hogan and Berry 2011) and the growing trends in hate crimes against Muslims, especially those occurring at people’s homes (CAIR 2017; NFHA 2017).
Three lines of inquiry would be useful for future research to pursue in order to better theorize about this new fault line in residential inequality. First, scholars should examine the locational attainment of Muslims and non-Muslims outside the Philadelphia metropolitan area. The majority of Philadelphia’s Muslim population is black. Our case study cannot be generalized to the experiences of Muslims living elsewhere in the United States.
Second, more research should be done to examine housing discrimination against Muslims. The two studies on this topic in the United States are limited to online correspondence tests in Los Angeles, New York, Detroit, and Houston. To understand the core tenets of how the place stratification model applies to explaining Muslim–non-Muslim disparities in locational attainment, future research should conduct nationally based, online, and in-person audit tests.
Finally, future research should examine the residential preferences of Muslims and non-Muslims. No research exists on this topic, but it is necessary to shed light on who Muslims and non-Muslims identify as in-groups and out-groups in their residential contexts, shaping who they want as their neighbors. Research on the residential preferences of blacks has been instrumental in clarifying the role that preferences play in impacting their locational attainment and has shown that although blacks ideally would like to live in whiter neighborhoods, they often do not because of fears of white hostility stemming from the prejudice and discrimination that they have experienced in whiter neighborhoods (Krysan and Farley 2002).
By 2050, the Muslim population is expected to more than double in size (Pew Research Center 2015). Therefore, it is imperative that urban scholars shift their focus to this new fault line of stratification in the American housing market. It is likely that the inequalities documented here will persist or even magnify, particularly because of the anti-Muslim sentiment that exists and appears to be increasing (Edgell et al. 2006, 2016). Because residence is inextricably linked to other dimensions of the American stratification system, such as education and employment, as well as to individual health, these inequalities will no doubt have implications for disparities experienced by future generations of Muslims.
Support for this research was provided by a grant to the Center for Social and Demographic Analysis at the University at Albany from NICHD (R24 HD044943). We thank Emily Rosenbaum, Tse-Chuan Yang, and the anonymous reviewers for their helpful comments, and Ruby Wang and Hui-Shien Tsao for their programming assistance.
Pew Research Center (2015:46–49) used a multistep process of identifying Muslims through a combination of self-reported data on religious affiliation, census data on place of birth for foreign-born individuals, and data on religious composition by country.
To create an indicator of whether the respondent identifies as a Muslim, we use the PHMC survey question, “What is your religious affiliation?”
Similar to other research on locational attainment, we include only measures gauging the percentages non-Hispanic white and non-Hispanic black in the neighborhood, which gauge each end of the racial and ethnic hierarchy in residential attainment (e.g., Alba and Logan 1993; Logan et al. 1996; South et al. 2011).
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.