Abstract
This article explores whether the places where people live—and specifically the diversity of incomes where people live—influence views about income inequality. Using a unique survey of New York City that contains geographic identifiers and questions about attitudes toward inequality, coupled with a rich array of Census data, we assess the degree to which the income diversity within spatially customized neighborhood boundaries influences beliefs about inequality. We find consistent evidence that attitudes about inequality are influenced by the places where people live—those who are exposed to more income diversity near their homes perceive larger gaps between the rich and everybody else, and are more likely to believe that the gap should be smaller. Moreover, this effect appears to be especially pronounced among those with lower educational attainment and at either end of the income spectrum.
Rising income inequality in the United States is an empirical reality (Piketty, 2014; Piketty & Saez, 2003). Income and racial inequality has also been imbedded in American cities and neighborhoods for a long time (Samuelson, 2016). While most Americans do appear to have taken notice of this inequality (Page & Jacobs, 2009), explanations of why attitudes vary remain underdeveloped. The existing research has emphasized individual-level explanations such as race, income, self-interest, and partisanship (Franko, Tolbert, & Witko, 2013; Gilens, 1995; Hunt, 2004; Smith, 1985). However, income inequality is a complex issue and opportunities to learn about income distributions and their (potential) consequences may be scarce outside of formal education. With this in mind, we theorize that to understand people’s attitudes about income inequality you also have to understand the places where people live. The places people live in may serve to expose them to different groups (or not), and in doing so may transmit information or alter views about groups (such as the rich or poor). Specifically, we hypothesize that the economic aspects of a person’s neighborhood—namely how diverse the incomes of households are—transmit information about broader economic concerns, and alter perceptions of inequality and attitudes about whether it is a problem.
We use a unique geocoded 2014 survey of New York City residents to test our theory. The survey contains questions about income inequality in both the United States and New York City. By knowing where respondents live, we are able to explore the extent to which neighborhood contexts matter. Our results support the idea that neighborhood income diversity increases people’s perception of and concerns about the income gap. Those who live in neighborhoods where they are exposed to both the rich and poor are more likely to perceive a very large gap in incomes and more likely to believe that the gap should be smaller. Alternatively, those who live in neighborhoods characterized by income homogeneity, who do not receive the same kind of group contact with the rich and poor, perceive of a smaller gap in incomes and are less concerned about the gap.
These results stand against a backdrop in which there is rising income homogeneity in American cities and towns (Reardon & Bischoff, 2011)—the kinds of neighborhoods where we see group contact and changing attitudes about inequality are in decline as Americans are increasingly living near people in similar circumstances to themselves. While the nation is becoming more unequal on the whole, we are less and less likely to actually be exposed to all ends of the income spectrum and to have personal experience with the groups that are central to the discussion. This may help explain why some have argued that attitudes toward inequality have actually become more conservative as the income gap has widened (Luttig, 2013; see C. D. Johnston & Newman, 2016 for an alternative view).
Explaining Perceptions of Income Inequality
Scholarly research that centers on income inequality has traditionally focused on the effects of inequality on a host of societal and democratic outcomes—that is, inequality as an independent variable that is used to explain cross-national differences. Income inequality can shape nonpolitical factors such as happiness (Oishi, Kesebir, & Diener, 2011) and public health (Wilkinson & Pickett, 2006), as well as have a wide range of political consequences. Places (either nations or states) with higher levels of inequality see lower levels of political interest (Solt, 2008), electoral and civic participation (Dahl, 2006; Lancee & Van de Werfhorst, 2012; Solt, 2008, 2010), and social trust (Fairbrother & Martin, 2013). Inequality also heightens the importance of class identification (Andersen & Curtis, 2012), and can reduce tolerance toward minority groups (Andersen & Fetner, 2008). What these findings highlight is that variation in inequality—across both countries and states—can produce differing outcomes for citizens. In many cases, we observe a number of societal “ills” that arise from income inequality. Despite these problems, we have seen Americans exhibit a reluctance to favor government action to address the issue (Shaw & Gaffey, 2012), with some arguing that rising inequality has actually resulted in greater conservatism in the United States (Luttig, 2013).
Individual-level explanations of attitudes about income inequality have garnered much of the explanatory attention with forces such as race (Hunt, 2004), racial attitudes (Gilens, 1995), income (Smith, 1985), economic self-interest (Franko et al., 2013), and partisanship (Franko et al., 2013) offered as the most formative. Furthermore, the way in which people attribute blame to poverty appears to influence the degree to which inequality is seen as a problem and influences beliefs about whether government should take action—those who point to structural or external forces as driving poverty or inequality tend to see inequality as more troubling and favor more governmental action, while those who view individual or internal forces as the primary explanation take the opposing stance (e.g., Bullock, Williams, & Limbert, 2003; Hunt, 1996; Schneider & Castillo, 2015).
Given the expanding body of research that supports a connection between context and attitude formation, we believe that the potential for a spatial context mechanism deserves attention. Across nations, levels of inequality shape policy opinions pertaining to redistribution and inequality (Finseraas, 2009; Lupu & Pontusson, 2011), suggesting that contextual effects of inequality can form citizen attitudes on these topics. Furthermore, there is ample reason to believe that intracountry economic contexts are noticed by people and matter for some attitudes. Objective state-level economic indicators have been shown to influence perceptions of income inequality (Franko, 2017), county-level inequality influences beliefs about meritocracy (Newman, Johnston, & Lown, 2015), county-level median incomes can influence perceptions of maldistribution (Newman, 2015), and local gender-based income inequality influences women’s belief in the “American Dream” (Newman, 2016). Furthermore, local contexts appear to structure the degree to which citizens favor or oppose larger governmental roles in economic policy (C. D. Johnston & Newman, 2016). We know less about how seeing and interacting with economic environments on a daily basis influence or alter attitudes about the income gap, and the degree to which this issue should be addressed.
Putting Inequality in (a Neighborhood) Context
How do people think about place and inequality? We often talk about inequality at the country level, where it is easy to conceptualize differing income distributions as a function of each country’s economy, tax structure, social support programs, and so on. For example, politicians advocating for more generous social programs in the United States often make cross-national comparisons with other systems with lower levels of inequality. This clarity in thinking about place and inequality at the country level likely becomes more opaque as we move to lower levels of geographic analysis like the state or city where economic contexts are less distinct and government policies are (often) less visible and less varied than they are between nations. That is, thinking about differences in inequality between the United States and Denmark is easier to conceptualize than thinking about differences between Illinois and California or Kansas City and Columbus. Given the challenges with thinking about inequality within countries, how should we expect citizens to form opinions about it?
Citizens do not form opinions in a vacuum. People receive different kinds of information from the places where they live and spend their time, and this contextually supplied information and social contact with other individuals shapes a wide range of political attitudes and behaviors (e.g., Adamczyk & Pitt, 2009; Beck, Dalton, Greene, & Huckfeldt, 2002; Berelson, Lazarsfeld, & McPhee, 1954; Huckfeldt, Plutzer, & Sprague, 1993; Huckfeldt & Sprague, 1995; MacKuen & Brown, 1987). Environmentally supplied influence can come in a variety of forms. Information can come from discussion with people who live nearby or who are encountered in daily life (Huckfeldt & Sprague, 1995; Kenny, 1998), through the adoption of norms and cues from those in similar social positions or surroundings (Burt, 1987; Wald, Owen, & Hill, 1988), or from contact with people from different social groups which can alter beliefs (Allport, 1954). These contextual forces can shape even the most stable and unyielding of attitudes, such as partisanship (Lyons, 2011), and there is a growing body of evidence that economic contexts (rather than the partisan or campaign contexts) are particularly meaningful for political beliefs (C. D. Johnston & Newman, 2016; Newman, 2014, 2015, 2016).
Why should these contextual forces be formative for our attitudes about inequality specifically, and what aspect of our environments should matter? Income inequality is an abstract and (potentially) difficult concept to grasp. Experiences with other citizens can help inform and provide grounding for these attitudes. To understand inequality—especially at more local levels—exposure to it or related concepts may be necessary. While a city could have a high level of income inequality, this inequality may be invisible to much of the citizenry. Those living, working, and recreating in predominantly high-income areas are often not exposed to inequality, nor are those living, working in homogeneous, middle, or lower income environments. For these citizens, the overall distribution of inequality in the city is likely to be of limited visibility. In a sense, economically homogeneous microenvironments are able to shield the individual from the broader inequality that exists in a place. As a result, these citizens may have distinct sets of attitudes about the degree to which income inequality exists in a place and the degree to which the government should intervene. Alternatively, the citizen who lives in a place where she is exposed to a diversity of incomes is much more likely to have a pulse on the nature of income inequality in that place, which (may) result in differing attitudes regarding the issue. The fundamental argument is that the local context can either shield the individual from the inequality in a place or provide a window into it.
What is it about the income distributions in an environment that should help shape attitudes? Inequality is about the widening differences between different income groups. We argue that exposure to these different groups in daily life will inform perceptions and attitudes. In effect, this is an argument about income diversity (homogeneity and heterogeneity). Heterogeneous income contexts expose the citizen to the rich, the poor, and the middle class. They enable people to observe and come into contact with all of the relevant groups in the inequality conversation, opening the door to the individual actually observing different income groups, seeing how they live, and potentially interacting with them.
Clarifying the difference between income inequality and income diversity is important for understanding this theoretical approach. Income inequality (typically measured using a Gini coefficient) is the extent to which the cumulative wealth of a place is concentrated in different levels of earners in a given space. An unequal space is generally understood as one in which the wealthy have much more money than the rest of the population—the corollary of which is that relatively few people control the wealth. Consider a hypothetical neighborhood (Neighborhood A) with 100 residents, where one person makes US$300,000 per year and the other 99 make US$25,000 per year. This would be a highly unequal neighborhood, and a measure such as a Gini coefficient would reflect this lack of income equality. Income diversity is the degree to which multiple earning levels exist in a space. In an economically equal world, there can still be income diversity. And in an economically unequal word, there can be considerable homogeneity (indeed, this is the world many Americans live in). Consider a different hypothetical neighborhood (Neighborhood B) that also has 100 residents: five make more than US$300,000 per year, 10 make US$200,000 per year, 20 make US$75,000 per year, 30 make US$50,000 per year, and 35 make US$25,000 per year. This is a much more income-diverse neighborhood. As we detail below, our contextual theory is that income diversity (not income inequality) informs attitudes about income inequality.
We favor a theory rooted in diversity rather than inequality because of the nature of contextual effects. For contexts to be formative, people have to actually encounter the potential influences. However, the most unequal of contexts—that is, Neighborhood A in the example above—would appear to a resident or observer to just be a poor area. The one wealthy person may not be encountered frequently, be visible, or even be known about. While this would represent high levels of inequality, it would not likely be a situation where social interactions or observations are likely to be delivered to the citizen. Income-diverse contexts—that is, Neighborhood B in the example above—however, have substantial numbers of people in all income groups, meaning that the person who lives in or passes through the neighborhood is able to see and interact with people from a variety of economic backgrounds. We argue that this is what is most likely to deliver contextual influence to the individual. Our expectations are thus rooted in the interactions that take place between citizens of different income groups, and these cross-income interactions are highly unlikely in the hypothetical Neighborhood A, despite the fact that it is a highly unequal neighborhood, while they are much more likely in Neighborhood B, which is income diverse (but actually more equal). While this relationship may seem somewhat counterintuitive, in that these income-diverse places tend to have more equitable distributions of income than the most unequal places, we believe that regular exposure to a diversity of income groups leads to discussion, observation of information/norms, and intergroup contact that inform citizens about the larger issue (both in scope and in space) of income inequality.
Beginning with the discussion mechanism, we know that discussion networks are (at least in part) constrained by where we live (Huckfeldt & Sprague, 1995). Living in a very Republican location will mean that the individual is likely to have a more Republican discussion network compared with someone who lives in a primarily Democratic area. In the case of our exploration of the effects of residential exposure to varying economic contexts, living in some areas means that the individual will talk to people of varying economic situations (both those who have high and low incomes), while others will be in largely homogeneous contexts and will likely have networks that are characterized primarily by one income group. Furthermore, we know that economic information and the economic circumstances of our discussants are relayed through these networks, and influence individual attitudes about topics such as redistribution (Newman, 2014). Putting these pieces together, a picture emerges in which being surrounded by individuals from a diversity of income groups means that the citizen will be likely to know and talk to people across the income spectrum, and that information about their differing experiences will be transmitted. Beyond explicit discussion of either economic circumstance or inequality, there is reason to suspect that citizens are able to observe the relative means of the groups around them. Homelessness, unemployment, poverty, and commercial activity are all visible circumstances, as are affluence and financial security. Where we have evidence that discussion with those who are experiencing financial trouble fosters empathy (Newman, 2014), it stands to reason these kinds of observational mechanisms should have a similar effect.
Finally, a sizable literature in psychology has studied the effects of intergroup contact on attitudes about groups (e.g., Allport, 1954; Pettigrew, Tropp, Wagner, & Christ, 2011). People who are exposed to “out-groups,” whether they are racial groups or groups like the disabled, tend to become less prejudiced about those groups (Pettigrew et al., 2011). Although intergroup contact does not always produce a more tolerant and less prejudiced citizenry (e.g., Paolini, Harwood, & Rubin, 2010), most evidence suggests that it does help dispel myths and build tolerance about these groups that could very well map onto income groups. It may be the case that people are unlikely to believe that the income gap is an issue if they believe that the poor are lazy or undeserving, but if they are exposed to the poor (as well as the rich), prior beliefs about these people may be altered. By reducing such prejudices, we argue that intergroup contact among income groups may shape attitudes about the issue. Thus, those who are in contextual circumstances where intergroup contact among different income groups occurs reap these benefits, whereas those who do not experience this contact are more likely to retain their (potentially prejudiced) priors. Income homogeneity only exposes the citizen to one group, and makes the concept of inequality between groups one that is more distant and less tangible. Homogeneity may provide a plethora of information and cues about poverty or about wealth, but it is poorly situated to transmit information and experience about the relative standing of groups and to show both sides of the coin.
Contextual exposure to income diversity should serve to highlight the potential problems with inequality. 1 Research suggests that people are relatively astute when it comes to discerning how others are performing economically, despite relatively low information about economic factors (Ansolabehere, Meredith, & Snowberg, 2014). The individual who has interpersonal relationships and observations across the income spectrum is going to see some citizens purchasing new houses or cars, going on vacations, and putting their children through the best schools, while others are struggling to pay bills, having to move as rising rents and gentrifying neighborhoods become unaffordable, and watching their children be denied a range of opportunities due to financial constraints. Conversely, those from homogeneous economic environments—whether they are poor, middle class, or affluent—are not exposed to this diversity. The consequence may be that the contrast between groups is harder to grasp, and the concept of income inequality will remain more abstract without a personal component to it. Exposure only to the middle class may create difficulties in grasping the struggles experienced by the poor, as well as extent of the advantages that are presented to the wealthy. Being surrounded only by the affluent may inhibit an understanding of why the poor have not earned higher salaries, may foster stereotypes about them, and may hide the struggles that they face. These expectations are rooted in the literature demonstrating that social contact and discussion with those who are different or who hold different beliefs promote tolerance and understanding (Mutz, 2002, 2006; Newman, 2014).
We posit that daily exposure to income diversity will influence two different attitudinal outcomes pertaining to inequality: The first of these is the perceived gap in incomes between the wealthy and everyone else. Context is, in a sense, an educational mechanism, and we expect diversity is what defines the educative environment that facilitates perceptions about inequality. This leads to our first hypothesis, which we will call the “Perception Hypothesis.”
In addition to believing the gap between the wealthy and everyone else to be larger, we expect contextual exposure to inequality to drive more normative beliefs about the issue. As noted above, such exposure can serve to generate tolerance and allow people to observe the pitfalls of inequality. This leads to our second hypothesis, which we call the “Attitude Hypothesis.”
While we expect that contextual exposure to income heterogeneity (and the interpersonal connections that such exposure produces) will foster greater recognition of inequality and a normative belief that the income gap should be reduced, the effect may be different for different groups. The kinds of exposures that people tend to receive due to other socioeconomic factors may concentrate the effects of income context among some citizens more than others. People who have diverse life experiences outside of the neighborhood—that is, they are exposed to income diversity at the workplace, in social settings, or places of worship—may be less influenced by the places where they live because they receive the same messages in other settings. However, people who have more homogeneous life experiences outside of the neighborhood—they largely interact with others of like themselves—are the least likely to receive the information and messages pertaining to inequality that the neighborhood environment can provide. While it is hard to fully capture the total heterogeneity or homogeneity of experiences in one’s life, some individual factors may serve as proxies.
Income levels structure many facets in peoples’ lives, ranging from the schools they attend, to the social events they can engage in, to the kinds of places where they work. We argue that the kinds of homogeneous life experiences are most likely to be found on the lower and upper ends of the income spectrum in urban areas. Those with low incomes are likely to be concentrated in certain kinds of jobs around others who are working for low wages. Because of their low incomes, they have less time and fewer resources to pursue a variety of recreational or leisure pursuits. They are likely to shop at different stores and patronize different restaurants. All of these dimensions of their lives suggest that they could have a striking amount of homogeneity of social experience. The same can be said of those on the upper end of the income distribution, but the homogeneity will be largely upper-class homogeneity. For these individuals who experience homogeneity in more aspects of their lives, we expect the effects of neighborhood income diversity to be stronger. The neighborhood represents one of the few venues in their lives that could provide them with information and social ties that would run counter to their social class. In urban areas, those in the middle of the income distribution are more likely to have friends, coworkers, and family on both ends of the income spectrum and less homogeneous lifestyles, without the means to select into certain homogeneity (as the wealthy do).
In addition to these mechanisms pertaining to the diversity of exposure to groups in different facets of life, those in lower income groups may be “activated” by their context. Newman et al. (2015) find evidence that lower income citizens who reside in contexts that are characterized by high levels of inequality tend to reject the notion of meritocracy at significantly higher rates than their peers in more equal contexts. The argument forwarded by Newman et al. (2015) is that those at the bottom of the income distribution are made starkly aware of their disadvantaged status in highly unequal places, and as a result are likely to shift their attitudes more than others who are not galvanized by the information about their relative standing in society.
With both the notions of homogeneity of experience, as well activation by context, we offer the following conditional hypothesis:
Finally, we argue that very similar expectations apply to one’s education levels. Groups that are less likely to learn about income inequality through other means—particularly via formal learning opportunities—ought to be the ones that are most likely to be influenced by the income diversity in their environment. Furthermore, these low education individuals, like those with low incomes, are likely to be most affected by the issue of inequality, and income-diverse contexts may serve impart higher awareness and activation among these less educated citizens (per the logic forwarded by Newman et al., 2015).
Data and Measurement
The 2014 New York City Neighborhood Survey (NYCNS)
The data for this project come primarily from an author-conducted survey in 2014. The 2014 NYCNS was a 20-min internet survey of 1,104 New York City residents that ran in October and November of 2014. The survey was conducted by the survey research firm Qualtrics. A critical feature of the survey was the collection of geographic identifiers for respondents in the form of the nearest cross-streets to the respondent’s home. While providing a location was optional, 920 of the 1,104 respondents provided it. Appendix C (online) provides information on the demographic and spatial representativeness of the survey relative to the population of New York City as well as a discussion of the implications of using a nonprobability sample. The survey is overrepresentative of women and White respondents, and underrepresentative of poor respondents. The consequences of this are discussed in more detail below.
Dependent Variables: Measuring Perceptions and Opinions About Income Inequality
There are two sets of dependent variables: The first set looks at how much income inequality people believe exists (what we call “perceiving the income gap”), and the second set looks at whether income inequality ought to be reduced (what we call “attitudes about the income gap”). For both questions, the survey asks respondents to evaluate for the United States as a whole and for New York City only. Including both allows us to see whether local contexts are tied more closely to attitudes at the local level or national level.
The question we use for our perceiving the income gap variable is as follows:
Would you say the current difference in incomes between rich people and everyone else is very small, small, medium, large, or very large? If the difference is “very large” it means that rich people make a lot more than everyone else. If the difference is “very small” it means that rich people only make a little more than everyone else.
The distribution of responses can be found in Figure 1. Note that few people have identified a very small or small income gap: Most of the variation is between a medium, large, and a very large difference in incomes.

Perceptions of and attitudes about the income gap.
Our attitude about income inequality question is given as follows:
Do you think the current difference in incomes between rich people and everyone else should be larger, about what it should, or should be smaller?
Figure 1 also shows the distributions of this variable for the United States and NYC, and once again the distribution is skewed—this time toward desiring a smaller income gap. However, a little more than 25% of the respondents indicate that the income gap is about what it should be or smaller than it should be. This distribution is similar to that seen in other research about attitudes on income inequality (Page & Jacobs, 2009).
Measuring Contextual Income Diversity
One of the key concepts we seek to measure is the neighborhood context of survey respondents. Researchers in myriad disciplines have sought to identify and operationalize neighborhoods for survey respondents using a variety of techniques. For example, Huckfeldt’s (1986) research utilized Census tracts, while others have sought a larger geographic context and used ZIP Code (e.g., C. D. Johnston & Newman, 2016)—the U.S. Census tabulates data for both of these units. The arbitrary nature of imposed neighborhood boundaries such as tracts and ZIP Codes has long been considered a potential problem for appropriate contextual identification. This is the modifiable areal unit problem (MAUP) in which the results of an analysis may be affected by (sometimes small) variations in boundaries for data collection (Anselin, 1988; Openshaw, 1984; Tam Cho & Baer, 2011). Some researchers, such as Wong (2012), have attempted to overcome these problems by having respondents draw their own boundaries on maps. However, Velez and Wong (2017) found that government-drawn boundaries better reflect perceived neighborhood attributes than respondent-drawn boundaries.
New York City clearly has its neighborhoods, but the space any one person defines as “their neighborhood” is a function of precisely where they live. For example, two people can live in the Lower East Side of Manhattan but depending on where they live within the Lower East Side they likely see their neighborhood differently. Recognizing this, our approach to measuring context at the neighborhood level is to aggregate each respondent’s Census tract with all tracts that lie within 1 mile of the centroid of that tract—approximately a 20 min walk in any direction (see Figure 2 for visualization). While we are not aware of any other research that has used precisely the same approach to neighborhood definition, R. Johnston and Pattie (2002, 2005) merge Census data based on the nearest 2,500 people to subject’s home.

Model one-mile spatially customized neighborhoods.
We believe that using the 1-mile radius approach is more appropriate than using solely existing designations such as individual Census tracts (which are too small, particularly in New York City) and ZIP Codes (which more closely reflect the size of a neighborhood but may not actually reflect the respondent’s neighborhood). Likewise, defining neighborhood based on physical distance rather than population more closely conforms to life in New York City. Ultimately, our approach is aimed at placing each respondent at the center of his or her own spatially customized neighborhood while still taking advantage of Census-drawn boundaries.
We acknowledge that using one’s neighborhood as defined by a 1-mile radius from their residence is not a perfect way to capture the totality of a respondent’s exposure to contextual stimuli. For some people who commute long distances from home, it is possible that the neighborhood does not capture the bulk of their daily interactions. However, neighborhoods and the places that are most proximate to our residences are still formative. They (likely) structure the exchanges that we have in many of our nonworking hours. Indeed, some of the most influential scholarship on social influence in the political arena relies in part on neighborhood social influence (e.g., Huckfeldt & Sprague, 1995), finding relatively strong effects on attitudes and behaviors. While it is not a perfect unit of contextual analysis and likely does not apply for all, there is strong reason to believe that neighborhood contextual influence matters on average.
Our choice for measuring neighborhood income diversity within “neighborhoods” is a Herfindahl–Hirschman Index (HHI). HHI is often used to measure market competition (Baker, 2001; Hirschman, 1964). We employ it as a way of looking at the extent to which a space is dominated by a single income group or split up among multiple income groups. For each neighborhood, we identify the percentage of households in the following four groups: (a) less than US$25,000 per year, (b) US$25,000 to US$50,000 per year, (c) US$50,000 to US$100,000 per year, and (d) more than US$100,000 per year. 2 HHI is equal to the sum of the square of those percentages for each neighborhood. We reverse HHI and divide by 100 (rHHI) to ease interpretation. The theoretical minimum rHHI (0) indicates no income diversity—a single income group in the space. The theoretical maximum rHHI (100) indicates perfect income diversity—an equal proportion of all four income groups.
The rHHI measure has some similarities with C. D. Johnston and Newman’s (2016) approach which uses an interaction between the percentage of the population that makes less than US$25,000 per year and the percentage that makes more than US$100,000 per year. The primary difference between the measures is that the rHHI approach takes into account the percentage of households in the middle income groups and treats them the same as the upper and lower income groups, whereas the C. D. Johnston and Newman (2016) approach exclusively uses the percentages of households at the ends of the income spectrum. For example, the rHHI measure treats a neighborhood that is 50% low income and 50% middle income as moderate income diversity, while the Johnston and Newman interaction would treat that same scenario as low-income diversity/inequality. However, both our rHHI measure and C. D. Johnston and Newman’s (2016) measure are testing similar concepts, in that they both pertain to citizen exposure to different facets of the income distribution. We view our approach as complementing (and supporting) theirs, rather than challenging it. In building measures centered on exposure, both our measure and C. D. Johnston and Newman’s (2016) are distinct from a Gini coefficient or other measures of wealth concentration frequently used when discussing inequality. Appendix D (online) expands on the similarities and differences between the measures, and demonstrates how they affect the results.
There is a negative relationship between rHHI and the square of the average tract-level median income for the 1-mile neighborhood (for all of New York City, r = –.3). Areas that have low incomes and high incomes tend to have less diversity, while middle-income areas tend to have more diversity. This is to be expected and speaks to the construct validity of the measure. One of the ways to have mid-level median income is to have income diversity through a mix of income classes in a space. Figure 3 is a map that shows how the income diversity measure is spatially distributed about the five boroughs of New York City. The areas that are the darkest orange are where there is the least income class diversity (i.e., people in the space tend to be making the same amount of money—rich or poor). The darkest pink areas are the spaces that exhibit the most income diversity. The map conforms to what regular observers of New York City would expect. Most of Manhattan below 110th Street (the North end of Central Park) is homogeneous—in this case, homogenously wealthy. The lone exception being the East Village where there is a considerable amount of public housing. Note that even if the blocks where the public housing is homogenously poor, the close proximity to other income groups allows the measure to appropriately capture it as an economically diverse area. A similar dynamic exists throughout the city. Consider the parts of Brooklyn shown in the right-hand map in Figure 3. Here, the homogenously wealthy Park Slope and Brooklyn Heights are yellow; as is the more homogenously lower income Bedford–Stuyvescent area. While Fort Greene trends wealthy, its physical situation between these two areas of Brooklyn means that the people who live there are proximate to other income levels.

New York City neighborhood income diversity (rHHI) maps.
Results
We are interested in citizen perceptions of the size of the income gap (in both the U.S. and New York City) and a normative question about whether the gap should be larger or smaller (in both the United States and New York City). These four measures will be the dependent variables for the remainder of the article. We use ordered logit estimators in all models as we have ordered dependent variables with either three or five values. 3 While the exploration of contextual factors may suggest that we should be using a hierarchical model, the nature of the Level 2 variable (rHHI) measured using the 1-mile radius for each respondent means that almost all respondents have a unique “context.” Thus, a hierarchical modeling approach is not appropriate.
Baseline Analysis
Our analysis began by estimating baseline models that test individual-level explanations before adding measures of economic context. Table 1 presents these models which contain only the individual-level demographic variables, including age, gender, education, income, and race, as well as attitudinal variables such as social and economic ideology, and interest in politics (see Appendix A for variable measurement details). The results presented are fairly straightforward and conform to expectations. While the results vary a bit between the U.S. and NYC models, we find that more educated, older, non-White, socially liberal, and economically liberal people are more likely to perceive income inequality. The general pattern is that variables that predict a perception of more income inequality also predict a belief that the gap should be smaller: More educated, non-White, socially liberal, economically liberal, and more politically interested people are more likely to believe that the income gap should be reduced. 4
Baseline Models—Effects of Individual Characteristics.
Note. Models display ordered logit coefficients, standard errors in parentheses. NYC = New York City.
p < .1. *p < .05.
The Effect of Income Diversity Context
Table 2 presents results where we add the individual’s economic context to the models using our measure of income diversity, as well as the percentage non-White in the respondent’s 1-mile radius neighborhood, and a question asking for how they perceive of their income compared with the incomes of their neighbors. Starting with the model looking at the effects of economic context on perceptions of the income gap in the United States, the objective measure of income diversity is not statistically significant. The contextual factor that appears to be more formative here is the racial context. As the percentage of individuals who are non-White increases in one’s neighborhood, we see that perceptions of the income gap in the United States increase. Turning to the second model, where we explore perceptions of the income gap in New York City, a different story emerges with respect to the effects of residential context. Here, our measure of contextual income diversity is significant and consistent with our expectation in the Perception Hypothesis: Those who experience more income diversity as measured by the rHHI perceive larger gaps between the rich and everyone else in New York City.
Effects of Context—Income Diversity in NYC and the United States.
Note. Models display ordered logit coefficients, standard errors in parentheses. NYC = New York City; rHHI = reverse Herfindahl–Hirschman Index.
p < .1. *p < .05.
The magnitudes of the effects are consequential: They predict sizable changes in the likelihood that people perceive very large gaps between the rich and everyone else in their local surroundings. Across the full range of the income diversity measure, the predicted probability of perceiving a very large gap rises from .50 in the most homogeneous contexts to .65 in the places where people are exposed to the greatest amount of income diversity. While education (.25 increase in probability) and social ideology (.23 increase) have larger magnitude effects than the context measures, the effect of age (.16 increase) is comparable. In sum, it appears that individual perceptions of the magnitude of inequality in society are at least in part structured by their surroundings.
The next question is whether these local environments can shape normative attitudes about inequality. We begin by looking at whether people believe that the gap between the rich and everyone else should be smaller in the United States (the third and fourth models in Table 2). The rHHI measure of income diversity is significant and in the direction predicted by the Attitude Hypothesis—contextual exposure to income diversity makes it more likely that people believe the gap between rich and everyone else should be smaller in the United States. The predicted probability of believing that the income gap should be smaller rises from .57 for those in the most homogeneous contexts to .72 (.15 increase) in the most income-diverse areas. The magnitude of this effect is comparable with education (.17 increase), economic ideology (.24 increase), social ideology (.15 increase), and interest in politics (.15 increase). Thus, we find support for the Attitude Hypothesis as it pertains to attitudes about inequality in the United States.
In our model for attitudes about inequality in New York City, we once again see significant results for the context measure. As income diversity increases in the respondent’s neighborhood, respondents are more likely to say that inequality should be smaller in New York City. The predicted probability of believing that the gap between rich and everyone else should be smaller in New York City rises from .63 for those in the most homogeneous income context to .75 (.12 increase) in the most diverse income context. This effect is smaller than those observed for education (.22 increase), economic ideology (.19 increase), and social ideology (.16 increase) but still of a consequential magnitude.
There is a stronger connection between local context and views about inequality in New York City than there is between local context and views about the United States as a whole—neighborhood income diversity is significantly associated with both perceptions and attitudes in New York City, but is only associated with attitudes regarding the United States. This is what we should expect to see if context is driving broader perspectives on inequality: People should be using information from the local level and drawing stronger conclusions about matters at that local level than they should about the national level. For example, state-level contextual factors appear to drive knowledge about politics at the state level but not knowledge at the national level (Lyons, Jaeger, & Wolak, 2013). In our case, it is likely that citizens are (relatively) aware that New York City is not representative of the nation at large, and thus a linkage does not exist between local income diversity and perceptions of the size of the income gap in the nation at large, but it exists for perceptions of inequality in the city.
What if income diversity is actually proxying for racial diversity in the models, as the two types of contexts are often interrelated (e.g., Intrator, Tannen, & Massey, 2016)? Are people using race to make inferences about income diversity? To examine this possibility, we rerun the models in Table 2 adding a measure of racial diversity (again using a rHHI where higher numbers indicate more diversity). For the whole of New York City, this measure is correlated with income rHHI at r = .45. The results using racial diversity can be found in Appendix B. In short, we find that racial diversity is not a complete substitute for income diversity. Racial diversity is only significant in the NYC perceptions model, but it does appear to reduce the importance of income diversity in this one case. This suggests that residents may pick up clues about the size of the income gap within the City from racial diversity, but this racial diversity does not drive attitudes in the same way that income diversity does. 5
Appendix D (online) contains a robust discussion of alternative measurement approaches—with respect to both operationalization of economic context and neighborhood definition. In short, we find similar results using the measure of inequality presented by C. D. Johnston and Newman (2016) but different (and mostly insignificant) results when using Gini coefficient. Results are also similar when the diversity and inequality variables are measured at the ZIP Code level.
Respondent Education and Neighborhood Income Diversity
Having shown that income diversity appears to affect perceptions and attitudes about inequality, we explore how these effects vary across different types of individuals. First, we look to the ways in which education structures the citizen’s susceptibility to contextual influence. To present these results, we split the sample between those who have completed less than a bachelor’s degree and those who have completed at least a bachelor’s degree. 6 We estimate two models for each dependent variable with different samples (Table 3). We split the sample rather than using an interaction term as we have reason to think that multiple variables in the model are interactive depending upon education, rather than just the income diversity measure. The results that we present appear to support this idea, as a number of variables perform differently across education levels. Because we find that local context is tied most closely to local perceptions and attitudes, we focus the remaining analyses on the New York City dependent variables.
Effects of Context—Income Diversity in NYC by Respondent Education.
Note. Models display ordered logit coefficients, standard errors in parentheses. NYC = New York City; rHHI = reverse Herfindahl–Hirschman Index.
p < .1. *p < .05.
Beginning with our perceptions-dependent variable across both education groups, we see that some differences emerge: Looking first to the income diversity measure, we see that there is an effect of income diversity for those with less than a bachelor’s degree (p < .1) but no effect for the more educated sample. For the less educated, as income diversity increases, perceptions tilt toward a greater gap existing between the rich and poor. For the educated, there are no significant effects for local economic context. Across the full range of the income diversity measure, we see a change in the predicted probability of perceiving a very large gap rising from .4 to .58 (.18 increase). These results align with our Conditional Education Hypothesis where we predict that those who may lack higher stocks of information that can be gained through education are informed by their contexts, while those who have had more education and (may) have higher levels of knowledge about such issues are less influenced.
Turning to the models of attitudes, we again see that the effects of the income diversity measure are concentrated among those who are less educated. As local income diversity increases, less educated respondents are more likely to say that the income gap should be smaller, while local income diversity does not influence the more educated. From the most homogeneous to most heterogeneous environments, the predicted probability of believing that the income gap should be smaller increases by .29—a sizable impact. When compared with the other significant variables in the model, we see that local income diversity has the largest magnitude. In sum, individual education levels appear to be a factor that leads some citizens to respond to local income contexts with altered perceptions and attitudes, while others are unaffected by these contexts.
Respondent Income and Neighborhood Income Diversity
Finally, we turn to the ways in which reactions to income diversity vary by respondent income. We use the same approach of splitting the sample that we employed previously—not only because multiple variables may be interacting with income diversity but also to explore nonlinearity in the relationship. Three household income groups were used: less than US$50,000 per year, between US$50,000 and US$100,000, and more than US$100,000. Results appear in Table 4 and they are not sensitive to small changes in these cutoffs.
Effects of Context—Income Diversity in NYC by Respondent Income.
Note. Models display ordered logit coefficients, standard errors in parentheses. NYC = New York City; rHHI = reverse Herfindahl–Hirschman Index.
p < .1. *p < .05.
Looking first to the lowest income group, we see that income diversity is positive and significant for both of our dependent variables. Among those with lower incomes, being exposed to income diversity increases perceptions of a very large income gap and is associated with believing that the gap should be smaller. For the perceptions-dependent variable, going from the most homogeneous income environment to the most heterogeneous income environment predicts an increase in the predicted probability of about the same magnitude as the social ideology effect and greater than the education effect. And once again, we find a similar relationship for the attitudes-dependent variable. Among those with lower incomes, exposure to income diversity appears to have the largest influence on attitudes about inequality.
For the middle-income group, we observe that income diversity is no longer a significant predictor of attitudes toward inequality. Rather, social ideology and education appear to be more consequential, depending upon the model. Looking finally to our two models with those who earn more than US$100,000 per year, we see that income diversity approaches significance again (p < .104) for attitudes about inequality but not for perceptions of inequality. Focusing on the attitudes model where income diversity is very near significant at the p < .1 level, we see that like those with lower incomes, exposure to income diversity among those with high incomes increases the individual’s likelihood of believing that gap should be smaller. While the change in effect sizes is smaller than those that we saw among the lower income groups, a .16 shift predicted probability is not trivial. Overall, these results support the Conditional Income Hypothesis.
It thus appears that individual reactions to contextual income diversity depend, in part, on personal income levels. For those in the lowest income group, there is a sizable effect of income diversity (this finding is consistent with that offered by Newman et al., 2015). For these citizens (who we have argued are likely to encounter primarily income homogeneity in other facets of their lives), neighborhood income diversity may provide exposure to those who are better off financially than they are, likely drawing the differences in their relative outcomes (and the issue) into focus. For those in the middle of the distribution, there is no effect of contextual income diversity. We have argued that this is likely due to the heterogeneity of experiences that these people receive outside of the neighborhood, making the neighborhood less formative. Alternatively, it may be the case that because inequality is fundamentally an issue about the extremes of the distribution, that those in the middle of the income distribution are not as responsive as it perceived by them to be a less relavent issue.
Perhaps most notably, exposure to income diversity appears to influence those on the upper end of the distribution as well. That is, intergroup contact and observation with those in worse financial standing increase the likelihood that those on the upper end of the income distribution believe that the income gap should be smaller—a position that could be viewed as antithetical to their self-interest. When exposed to income diversity, the attitudes of those on the upper and lower end of the income spectrum appear to converge on the belief that the gap should be smaller. Absent this contextual exposure, both groups are much less likely to take such stances.
Conclusion
The evidence that we have presented here supports the theory that our surroundings—the places where we live and the influences that they provide—structure our attitudes about income inequality. People who are exposed to income diversity in daily life, seeing and coming to know various parts of the income spectrum, are more likely to perceive a larger gap between the rich and poor, and are more likely to believe that the gap should be smaller. However, people who are exposed primarily to income homogeneity and do not see the range of outcomes that are associated with inequality are less likely to believe that the gap between rich and poor is very large, and are less likely to believe that the gap should be smaller. Finally, we have shown that the effect of contextual exposure to income diversity differs depending upon the education and income levels of the respondents. We argue that this individual nuance is driven by the opportunities that people have to learn about or experience inequality and income diversity, as well as the activation that can occur when one’s circumstances are contrasted with those who surround them (e.g., Newman et al., 2015). Some learn about inequality during the course of their education, while others learn about it through exposure in various life settings such as the workplace. For these two groups, the information provided by their context is not likely to be as formative. However, for groups that have not had the opportunity to learn or be exposed to the matter (less educated individuals and those with more homogeneous experiences such as the lower and upper classes), the neighborhoods that we live in can step in to inform them.
What are the implications of these findings? Looking at our social environments and the influences that they provide is important for understanding how changes in aggregate inequality (could) produce changes in citizens’ attitudes. Some have argued that the increases in inequality that we have seen over the last several decades have not been met with increasing recognition of the issue among the public or more interventionist attitudes about it (Luttig, 2013). That is, changes in national circumstance have not been met with changes in attitudes. One of the secondary findings of this article is that neighborhood Gini coefficient does not affect perceptions of income inequality. We suspect this is because income inequality—in the wealth concentration sense—is difficult to discern in smaller spaces like neighborhoods. The suggestion from the primary findings here is that the separate but related concept of contextual income diversity can produce changes in citizen beliefs about the issue in part because it is discernable. However, this ability for contextual income diversity to alter beliefs is mitigated by rising residential income homogeneity (Reardon & Bischoff, 2011), so there may be fewer and fewer people who actually live in the kinds of (income-diverse) places where we see these effects.
We conclude by noting that there are many questions unanswered and several shortcomings for our analysis: First, whenever we are dealing with the effects of residential location, issues of selection are a concern. We concede that citizens who already believe that the income gap is a sizable problem and that those who believe it should be smaller may be more willing to move to and reside in diverse income areas. However, the reality of citizen moving decisions (especially in highly competitive housing markets such as New York City) is that they are constrained by many factors such as price, commuting distances, and schools, potentially reducing the degree to which people are able or willing to select residential location based on metrics that would lead to serious inferential problems for our analyses. Second, while we have offered several theoretical mechanisms for contextual influence, we are unable to explicitly test them or evaluate one against the other. A large body of research demonstrates that places are able to exert influence through the discussion and provision of information channels that we have posited, but it is possible that one is more formative for attitudes on inequality, and we are unable to test these distinctions. Finally, while the survey that we have used is uniquely situated to address questions of economic context and inequality in an urban setting, we recognize that conclusions drawn from New York City may not generalize as well to parts of the country with different racial, economic, or partisan geographies. New York City is a decidedly liberal location, meaning that the results we have presented here are driven in large part by the ways in which liberals respond to income diversity. Were these analyses to be replicated in a metro area with more conservatives, different average effects may emerge. With this disclaimer in mind, we have explored whether there is an interactive effect between income diversity and partisanship to see if Democrats respond differently than Republicans, and find that Democrats appear to be more responsive to income diversity in establishing their perceptions of inequality in New York City but not for any of the other outcomes explored (see Appendix E [online]). Despite these potential sample limitations, New York City offers a great deal of variation in socioeconomic environments and, as a result, is a good case for testing these theories.
Footnotes
Appendix A
Appendix B
Acknowledgements
The authors would like to thank Barnard College for its support conducting the 2014 New York City Neighborhood Survey (NYCNS). We would also like to thank Peter Enns, Anand Sokhey, Justin Vaughn, Stephen Utych, Jaclyn Kettler, Lori Hausegger, and our anonymous reviewers for their valuable feedback.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
Supplementary Material
Supplementary material is available for this article online.
Notes
Author Biographies
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
