INTRODUCTION
Brooklyn, New York has been transformed from a gritty borough where minority, blue-collar families work and live to one crowded with white, educated millennials. As the population has grown whiter and wealthier, so has Brooklyn itself. Where bodegas and cuchifritos once stood, coffee shops and craft breweries have emerged. Once run-down brownstones have been restored and re-vitalized, and property values have subsequently soared, in turn marginalizing original residents. Now, Brooklyn struggles with these two identities, as Brooklyn borough president, Eric Adams, explains:
“Our young people coming in need to understand that they are not the modern-day Christopher Columbus[es]: They did not discover Brooklyn. Brooklyn was here long before they set sail, and if anything they need to be part of the greatness of Brooklyn and add their flavor, but not destroy what we are. If we're not careful, gentrification could drive a permanent wedge between us.” (Wolfe Reference Wolfe2017, p. MB1)
On the one hand, the return of young, educated individuals to cities has been championed, as once run-down urban centers are revitalized (Florida, Reference Florida2014). Gentrification often brings improved housing and infrastructure, and with it economic opportunities (Atkinson and Bridge Reference Atkinson and Bridge2004; Byrne Reference Byrne2003; Duany Reference Duany2001). On the other hand, gentrification is known to marginalize original residents, particularly those of color, and ultimately degrades social networks (Abu-Lughod Reference Abu-Lughod1999; Smith Reference Smith1996). Political scientists, for the most part, have yet to contribute their perspectives to this debate. In the field, some scholarship (Hyra Reference Hyra2014; Knotts and Haspel Reference Knotts and Haspel2006; Martin Reference Martin2007) has begun to raise questions exploring the political implications of gentrification, with a particular focus on individual- or organizational-level experiences. Other research suggests that political consequences might apply more broadly, having political and electoral consequences for larger communities or even the city as a whole (Betancur Reference Betancur2002; Owens and Brown Reference Owens and Robert Brown2014; Tighe et al. Reference Tighe, Wright, Renner and Hyra2015; Wilson, Wouters and Grammenos Reference Wilson, Wouters and Grammenos2004). As a consequence, we sought out ways in which to measure these community- and city-level political responses to gentrification.
Here, we employ descriptive representation in order to assess these more aggregated reactions to gentrification. For political scientists, descriptive representation is a particularly useful measure in that it signals the political power of minority groups. Indeed, exploring the effects of gentrification on minority representation in urban areas is important because cities were among the first places where African Americans and Latinos were able to gain political power. From Maynard Jackson's influential role in Atlanta, Harold Washington's achievements in Chicago, to Tom Bradley's ascent in Los Angeles, black mayors have been the pioneers of a generation of black elected officials. Likewise, Henry Cisneros made history when in 1981 he became the first Latino mayor of San Antonio, Texas since 1842. Many of these minority mayors, including Cisneros, began their political careers in city councils. As the first line of minority representation, it is important that we, as political scientists, study how changes in the composition of our nation's cities might affect the representation of minorities in the present day.
Still, gentrification can produce a variety of political responses, ranging from political mobilization to political displacement. For example, Martin (Reference Martin2007) explores how neighborhood associations respond to encroaching gentrification with some organizations mobilizing to protect the political rights of original residents. Wilson, Wouters, and Grammenos (Reference Wilson, Wouters and Grammenos2004) come to a similar conclusion, finding that a Chicago neighborhood was able to successfully mobilize against development and thus keep the forces of gentrification at bay. Yet, the mobilization and resistance seen in Wilson, Wouters, and Grammenos (Reference Wilson, Wouters and Grammenos2004) stand in contrast to other narratives and empirical work showing the hopelessness that residents feel in the wake of gentrification (Hyra Reference Hyra2014; Wyly and Hammel Reference Wyly and Hammel2004). How then can gentrification lead to both political mobilization and political displacement? We seek to explore what factors contribute to these divergent community responses by examining minority descriptive representation in the largest American cities.
In this paper, we argue that political mobilization occurs when the community is still dominated both proportionally and politically by longstanding residents, yet political displacement occurs when the community is close to becoming politically and proportionally controlled by the newly dominant group. These expectations build on Eisinger's (Reference Eisinger1980) work that explores how a once dominant racial group is affected politically when a newly dominant racial group comes to power. We seek to explore a similar process in today's political context, but instead investigate how minority political power is affected by a growing middle-class white population. That is, what happens to the political power of these pre-existing minority residents as middle-class whites flow into their neighborhoods and cities? In line with Eisinger's theory, we expect that a growing white population will increase the chances of electing a white candidate when the minority population hovers near a demographic tipping point (i.e., another race is close to being the dominant demographic group). Conversely, a growing white population will decrease the chances that a minority candidate will be elected when people of color are close to losing their demographic dominance. Yet, we also expect that a growing white population will not necessarily deter minority descriptive representation when the minority population maintains clear proportional dominance. In the early stages of gentrification, a growing white population may politically activate original residents to resist the demographic and economic changes that they foresee coming, which fits with other scholars’ work about the political resistance against gentrification (Hyra Reference Hyra2014; Martin Reference Martin2007).
In order to test political influence, we assess if and to what extent gentrification affects minority descriptive representation—a measure of minority political power. Using an original data collection of the race and ethnicity of city councilmembers in the 40 largest cities and the demographics of the jurisdictions they serve, we find evidence that gentrification negatively impacts black descriptive representation. Gentrification that results in a growing white population negatively affects the election of black councilmembers, and the effect is particularly pronounced when the black population is close to losing its dominance, as we expect. When the community is largely black, though, an increasing white population has a less negative effect on the election of black councilmembers, again as posited. This suggests that gentrification that results in racial and demographic changes has varying effects on descriptive representation, depending on the degree to which the area had already gentrified.
GENTRIFICATION AND POLITICAL DISPLACEMENT
Ruth Glass (Reference Glass1964) initially coined the term, “gentrification”, in her book, London: Aspects of Change:
“One by one, many of the working class quarters of London have been invaded by the middle classes—upper and lower. Shabby, modest mews and cottages—two rooms up and two down—have been taken over, when their leases have expired, and have become elegant, expensive residences…Once this process of “gentrification” starts in a district, it goes on rapidly until all or most of the original working class occupiers are displaced, and the whole social character of the district is changed” (1964, p. xviii).
This original definition provides us with a few key characteristics of gentrification: the influx of a wealthier, more educated group; improvements in properties; and displacement of original, poorer residents. It is this last characteristic we would like to examine—displacement. It is thought that gentrification leads to long-term residents being pushed out through higher property values and rents, but also the changing cultural landscape of their community. And, in some ways, research confirms that physical displacement occurs (Hyra Reference Hyra2008; Podagrosi and Vojnovic Reference Podagrosi and Vojnovic2008). Yet, recent empirical work with more sophisticated measures of displacement has increasingly cast doubt on the notion that gentrification results in the physical displacement of residents (Freeman Reference Freeman2005; Freeman and Braconi Reference Freeman and Braconi2004). Instead, residents may stay, but feel increasingly marginalized by the cultural rise of newer residents. In other words, they feel “politically displaced.”
Martin best defines political displacement as it applies to gentrification: “Political displacement occurs when [longstanding residents] become outvoted or outnumbered by new residents within their organizations or through the creation of organizations dominated by new residents” (Martin Reference Martin2007, p. 605). Indeed, there is a growing body of literature showing how gentrification affects political displacement. Knotts and Haspel (Reference Knotts and Haspel2006) find that gentrification depresses turnout among longstanding residents. In a thorough ethnographic study of a Washington, DC neighborhood, Hyra (Reference Hyra2014) finds that many original residents expressed feelings of no longer belonging. And, of course, low-income individuals of color are those most likely to experience political displacement when a community gentrifies (Wyly and Hammel Reference Wyly and Hammel2004).
Still, other work demonstrates how gentrification may politically displace some residents and not others. For instance, Knotts and Haspel (Reference Knotts and Haspel2006) note that the negative effects of gentrification are largely limited to longstanding residents. Newman, Velez, and Pearson-Merkowitz (Reference Newman, Velez and Pearson-Merkowitz2016) demonstrate that African Americans specifically are politically demobilized through the erosion of social capital in the wake of gentrification. Differential effects are not necessarily limited to racial or ethnic groups, however. In a qualitative examination of a community in New York City, Cahill (Reference Cahill2007) finds that women felt particularly socially and politically displaced in the wake of gentrification. Other scholars have also noted how gentrification might affect women differently, given the gendered distribution of labor within urban areas (Bondi Reference Bondi1999). Michener and Wong (Reference Michener and Wong2018) take up this line of inquiry regarding differential responses and find that gentrification politically displaces longstanding residents, women, and whites. Michener and Wong's work even points to the idea that gentrification mobilizes some current residents, furthering the notion that gentrification can result in varying political responses. Indeed, Martin (Reference Martin2007) finds that some community organizations mobilize in response to gentrification in order to protect the political rights of original residents. Wilson, Wouters, and Grammenos (Reference Wilson, Wouters and Grammenos2004) come to a similar conclusion, finding that a Chicago neighborhood was able to successfully mobilize against development and thus keep the forces of gentrification at bay. Thus, gentrification can both mobilize and politically marginalize some and not others, presenting somewhat of a puzzle.
Much of the pre-existing work has focused on interviews and surveys, exploring organizational and individual-level political responses to gentrification. Yet, other research suggests that political displacement might apply more broadly, having political and electoral consequences for the city as a whole (Betancur Reference Betancur2002; Owens and Brown Reference Owens and Robert Brown2014; Wilson, Wouters and Grammenos Reference Wilson, Wouters and Grammenos2004). Indeed, we opt for a more aggregated approach and examine descriptive representation or the racial and ethnic characteristics of local elected officials as a way to explore how gentrification has broader political implications for cities. Much like Michener and Wong (Reference Michener and Wong2018) posit, there is reason to believe that some communities might be more prone to political displacement or mobilization than others. For instance, influxes of whites in African American communities has resulted in some African American elected officials becoming concerned that these new constituents might not be as electorally supportive of them as their long-standing black constituents and thus result in reduced black descriptive representation (Fraser Reference Fraser2004; Wilson Reference Wilson2012). This suggests that black representatives are electorally vulnerable in the wake of gentrification, yet it is not clear that the same holds for Latino representatives. And given the mixed findings from previous work examining individuals, there is reason to believe that gentrification may affect the descriptive representation of minorities, Latinos, and African Americans differently.
DESCRIPTIVE REPRESENTATION IN URBAN AMERICA
The concept of representation is multifaceted, as scholars have thought about representation in a variety of ways. Descriptive representation involves the notion of being represented by someone of the same race, ethnicity, or gender, while substantive representation involves the idea of one's interests being represented regardless of these demographic characteristics (Pitkin Reference Pitkin1967). The literature on minority representation in political science tends to focus more broadly on the election of minorities to Congress, legislatures, city councils, and school boards (Casellas Reference Casellas2011; Engstrom and McDonald Reference Engstrom and McDonald1981; Marschall and Rutherford Reference Marschall and Rutherford2015; Owens Reference Owens2005; Rocha Reference Rocha2007; Shah, Marschall, and Ruhil Reference Shah, Marschall and Ruhil2013). These studies find, with few exceptions, that greater minority populations lead to more descriptive representation in political institutions. It is also the case that minorities generally are not elected from areas that are not heavily minority (Grofman and Handley Reference Grofman and Handley1989). For the most part, districts with larger white populations will generally elect white city council members, districts with larger Latino populations will generally elect Latino city council members, and the same for African Americans.
Oftentimes, political scientists examined descriptive representation from an institutional perspective. For example, multimember or at large districts have been found to hinder minority representation (Gerber, Morton, and Rietz Reference Gerber, Morton and Rietz1998). Single member districts or “ward”-based districts are found to be more conducive to the election of minorities (Karnig Reference Karnig1976). Welch and Studlar (Reference Welch and Studlar1990) show that African American descriptive representation is more negatively affected by at large districts than Latino representation. Trounstine and Valdini (Reference Trounstine and Valdini2008) argue that context matters such that black descriptive representation increases more from ward-based elections than Latino representation especially because of residential concentration and higher rates of voter turnout among African Americans. Here, we argue that we can also use descriptive representation in order to assess the effects of gentrification on the political power of minority groups.
Why should we care about gentrification and the possible deleterious effects on urban minority descriptive representation? For many minorities, the election of co-ethnics only happens at the city council or state legislative level. They might live in states or congressional districts which elect non-minority candidates at high rates and thus are under-represented at the state and federal levels. Bobo and Gilliam (Reference Bobo and Gilliam1990) have also argued that levels of trust in government more generally increase when minorities are represented by co-ethnics. In addition, Barreto (Reference Barreto2007) finds that Latinos are empowered to turnout to vote when Latino candidates are on the ballot. There is also some evidence that minorities also place higher value on co-ethnic representation (Casellas and Wallace Reference Casellas and Wallace2015). Gay (Reference Gay2001) notes that the potential consequences of fewer minority representatives in office might depress feelings of efficacy among minorities who might sense a potential reversion in political progress.
DESCRIPTIVE REPRESENTATION AND GENTRIFICATION
In this paper, we examine how gentrification affects descriptive representation of minorities generally, but also more specifically Latinos and African Americans, especially given that racial and ethnic groups respond differently to gentrification, per Michener and Wong (Reference Michener and Wong2018). Scholarship is scant on this specific issue with one working paper (see Tighe et al. (Reference Tighe, Wright, Renner and Hyra2015)) identifying gentrified neighborhoods across the United States. Specifically, they examine thirty zip codes that have experienced the greatest increase in their white population and then collect data on the race of the city councilmembers that served these districts from 1990 to 2010. Their analyses compare the characteristics of the districts that switch to having white representatives with districts that continue to have black representatives. They find that, in some cases, gentrification leads to decreased black descriptive representation. Here, we argue that when the white population increases, the effect on descriptive representation is likely conditioned by the current demographics of the district. In other words, a growing white population sometimes has an effect on the election of people of color, depending on the current demographics of the district.
Theoretically, we leverage Eisinger's (Reference Eisinger1980) work on political displacement to help us determine when gentrification begins to affect descriptive representation. He examines how the once dominant racial or ethnic group in Atlanta and Detroit—in those cases whites—responded to demographic transitions that brought about the rise of black politicians. Eisinger (Reference Eisinger1980) finds that white elites largely cooperated with the rise of black politicians and attributes this mostly peaceful transition to two primary causal mechanisms. First, and perhaps most importantly, minority groups composed approximately 50% of the city's population, and they were projected to grow through the foreseeable future, a phenomenon Eisinger terms an ethnoracial transition. As a consequence, white elites sought to maintain positive relations with those who had newly risen to political power. Second, while many white elites had lost political power, they still maintained their social and economic dominance in the city. In fact, white elites saw the necessity of partnering with the new black political coalitions in order to simultaneously secure their own economic power.
In the wake of the back-to-the-city movement, it makes sense that some of Eisinger's conclusions might apply to areas that are experiencing the reverse process. These cities and communities that were once predominantly minority, black, or Latino are now growing slowly whiter as millennials seek the lifestyle and amenities of urban settings. What happens to the political power of these original minority residents as middle-class whites flow into their neighborhoods and cities? Here, we posit that Eisinger's (Reference Eisinger1980) theoretical framework might apply in certain ways. Still, it is important to briefly highlight the differences between a community transitioning into being majority–minority and a community that is gentrifying.
We would like to note that Eisinger's work focused primarily on the reaction of white elites. In this way, it is unclear how much his findings regarding white elites apply to low-income communities and people of color. For instance, he illustrates that the white elites in Atlanta and Detroit may have lost political power, but they still maintained significant social and economic power. When we compare that with gentrifying communities, longstanding residents of historically poor neighborhoods of color likely do not wield the same power and influence as those of white elites detailed by Eisinger. In fact, when white elites lost their political power, they turned to other forms of influence. It is this point specifically where we depart from Eisinger, especially given that much of the pre-existing scholarship examining gentrification details how racial transitions can often lead to political displacement among specific groups or communities as a whole. Nevertheless, we think that Eisinger's general conclusions about when political power might change to be still relevant.
Here, we primarily leverage the demographic portion of Eisinger's framework. That is, we might expect gentrification to have an effect on political representation only under certain demographic conditions. Specifically, he posits that political power will shift when the community begins to tip from having one dominant racial group to another, in this case from white to black. He finds that blacks could assume power when the cities were roughly half African American and half white, and the black population was still growing. Applied to contemporary gentrification, we may imagine that areas that have already experienced significant demographic change due to gentrification will be more likely to elect a person that shares the racial identity of the newly dominant group, in this case whites. Specifically, we expect that a growing white population will increase the chances of electing a white candidate when the minority population hovers near that demographic tipping point and maintains approximately half of the population. Conversely, a growing white population will decrease the chances that a minority candidate will be elected when whites and people of color make up nearly equal proportions of the population. We argue that when faced with this specific demographic composition, original residents will feel politically displaced and, in turn, be unable to support their co-ethnic candidates, which leads to our first hypothesis:
Minority Political Displacement Hypothesis: A growing white population decreases the chances of electing a person of color when the minority population is close to losing its dominance (i.e., it comprises roughly half of the area's population).
Nevertheless, other research points to the political strength, and subsequent opposition, original residents wield when gentrification begins (Boyd Reference Boyd2008; Martin Reference Martin2007). Initial signals that an area is gentrifying, like a growing white population, may politically activate original residents to resist the demographic and economic changes that they foresee coming. Eisinger (Reference Eisinger1980) himself notes that violent interactions and racial responses to the growing African American population occurred earlier in the ethnoracial transition when the demographic inevitability of the city was less clear. These conflicts tended to happen (and resolve) much earlier in the demographic transition, before the newly dominant racial group could mount a credible political challenge. This suggests that at the onset of gentrification, when a community is still largely minority, a growing white population is not necessarily an impediment to electing people of color, and, in fact, might lead current residents to politically mobilize, in turn, electing co-ethnic candidates, leading to the following hypothesis:
Minority Political Mobilization Hypothesis: A growing white population does not affect the chances of electing a person of color when the area's minority population maintains clear proportional dominance.
Here, we have conceptualized “clear proportional dominance” as a racial or ethnic group comprising more than two-thirds of the community. It should be noted, though, that proportional dominance likely varies somewhat by the racial and ethnic composition of the area, given that African Americans vote at higher rates than Latinos.Footnote 1 Thus, you would likely need a slightly greater proportion of Latinos to maintain dominance within the community. Nevertheless, these hypotheses suggest that gentrification and the corresponding demographic changes may empower minorities to elect co-ethnics up until a point. At some juncture in the gentrification process, original residents may feel that they can no longer resist or oppose gentrification, and thus relent to have the area ruled and represented by the now dominant group—middle-class whites.
At the moment, we do not examine the extent to which the city councilmembers are substantively representing their constituents. Scholars have tested substantive representation in Congress, and to a lesser extent, state legislatures, by examining roll call votes, but this is not presently feasible for city councils. And it is also possible that in the face of strong political machines, the election of a minority councilmember may have more to do with the choices of the machine than the preferences of the electorate. Indeed, in some cases, political machines have significant influence over who is elected as seen in Chicago. However, other research does point to how local political machines failed to take hold, particularly in the Southwest (Bridges Reference Bridges1999). While the role of political machines is currently unmeasurable, we do think this is an important point to nevertheless highlight.
Here, we focus on how the changing demographics affect descriptive representation, conceptualized here as the likelihood that a minority councilmember is elected, relative to a white councilmember. Again, we expect that an increasing white population will result in fewer African Americans and Latinos in office when minorities compose roughly half of the area's population. In contrast, a growing white population will not affect the chances of electing a person of color when the area's minority population is clearly and proportionally dominant.
DATA AND METHODS
Our universe of cases is the forty most populous cities as of 2016. The full list can be seen mapped in Figure 1. The 40 cities range in population from 465,000 in Colorado Springs to 8.5 million inhabitants in New York City. These cities’ black populations range from approximately 3% in San Jose to 80% in Detroit. In contrast the Hispanic population ranges from 4% in Baltimore to 81% in El Paso. There is clearly a great deal of variation in terms of population size and demographics among this urban sample. Moreover, this sample is regionally diverse, spanning the South, Midwest, and East and West Coasts.
Figure 1. Map of Sample Cities Included in Analyses.
The unit of analysis is city councilmember districts or seats. The seat itself is an appropriate unit of analysis because the descriptive representation literature often opts for studying the phenomenon at the seat level. Cities that maintained only at-large elections, like Columbus, OH and Portland, OR, were also included in these analyses. Consequently, we included a control variable for at-large seats. This also allowed us to include at-large representatives on mixed councils where some positions are elected at-large while others are elected by district. Still, there are shortcomings with using the district or the city as a whole as the unit of analysis. Gentrification often occurs at the neighborhood level—a finer geographic unit of analysis than the district. Thus, we may be capturing larger demographic shifts in the city than gentrification per se. Still, if we were to use demographic and property changes at the Census tract level (a better unit in which to capture neighborhood change), it would neglect the effect that all other portions of the district have on electing a councilmember.Footnote 2
The dependent variable is the race and ethnicity of every councilperson that represented a district in January and February 2017, leading to a total of 555 observations from forty cities. Categories for race and ethnicity include black, Hispanic, white, Asian, and other. For the purpose of these analyses, we exclude councilmembers that identify as Asian or “Other” because there are so few observations in these categories. Subsequently, we create three dummy variables. The first is a dummy variable for whether the councilmember is a minority, in this case they identify as either black or Latino. If the councilmember is Latino or black, then she is coded as 1, and white councilmembers are coded as 0. Then we disaggregate this minority dummy variable by the race of the candidate to determine if gentrification has varying effects on Latinos and African Americans. Specifically, the second dummy variable for a Hispanic councilmember is coded as 1 if the councilmember is Latino, and 0 if they are white. The third and final dependent variable is for black councilmembers where black councilmembers are coded as 1 and all white councilmembers are coded as 0. Thus, the election of minorities, black, and Latino councilmembers is relative to the election of white councilmembers.
For both academics and practitioners, there is no consensus on how to operationalize gentrification (see Zuk et al. (Reference Zuk, Bierbaum, Chapple, Gorska, Loukaitou-Sideris, Ong and Thomas2015) for a review). More traditional methods of measuring gentrification involve Census indicators. In one of the older and more popular methodologies, Hammel and Wyly (Reference Hammel and Wyly1996) used many Census indicators, including characteristics of the neighborhood, such as median household income, percentage of workers in managerial, professional, or technical occupations, percentage of people 25+ with 4 or more years of education, median contract rent, and median cost of housing, as well as percentage change measures of these indicators, to capture gentrification. Other research has opted to frame the socio-economic and demographic measures as being relative to the city itself (Chapple Reference Chapple2009; Freeman Reference Freeman2005; Laniyonu Reference Laniyonu2017). For instance, Freeman (Reference Freeman2005) employs a variety of Census measures in relation to city averages, like whether a Census tract has a household income below the city's 40th percentile of household income. In these analyses, we are unable to use the relative measures for cities employed by Freeman because some of our observations are at-large seats, and we would be unable to compare the at-large seat's gentrification measures with the city's median measures because they would take on the same values.
More recent research (Papachristos et al. Reference Papachristos, Smith, Scherer and Fugiero2011) has opted to include demographic measures as well, such as percent Hispanic, black, and white. This is because gentrification, at least in the American context, is often the dominance of upper-middle class whites in a neighborhood that was once majority–minority. Since the mid-20th century, minority groups have been the primary residents of urban cores, yet as upper-middle class whites returned to city centers, blacks and Latinos have become increasingly marginalized either by displacement or proportionality. It is this specific aspect of gentrification—the demographic and political displacement of original minority residents—that we are theoretically most interested in. Nevertheless, we also include other aspects of gentrification, like changes in education, income, rent, and employment, to determine if other aspects of the phenomena affect descriptive representation of people of color. Here, we hew closely to Hammel and Wyly's (Reference Hammel and Wyly1996) conceptualization and include their five Census measures, measures of race and ethnicity, and corresponding percentage change and percentage point change measures, which include the following measures, as can be seen in Table 1.
Table 1. List of Census Gentrification Indicators
We calculate how much these measures have changed from 2000 to 2015 because gentrification is better conceptualized as neighborhood change, and this is a common technique in the literature (Hammel and Wyly Reference Hammel and Wyly1996; Knotts and Haspel Reference Knotts and Haspel2006). To measure gentrification as the static percentage of the population that is white in a given neighborhood at a given time would be inappropriate. Take, for instance, two communities that currently have similar demographic compositions, but one community has experienced significant gentrification over the past 10 years while the other has remained relatively stable. These two neighborhoods would look virtually identical in an empirical analysis using percent white in 2015 to measure gentrification.
In order to do so, we overlaid maps of city council districts in ArcGIS with maps of Census block groups in order to obtain demographic measures for the district in 2000 and 2015. In situations where the block group does not neatly fit within the district, we weighted the population estimate by the area that falls within the district. We subsequently sum all the weighted block groups to create estimates for total population, Hispanic population, black population, non-Hispanic white population, total houses, percent over 25 with a college education, and percent employed in “gentry” professions at the district level in both 2000 and 2015.Footnote 3 These latter two measures regarding employment and education have been used primarily (Sharp Reference Sharp2014) and secondarily (Laniyonu Reference Laniyonu2017) as measures of gentrification, with gentry professions being defined as those employed in professional, scientific, or management positions. We employ a similar weighting technique to obtain median household income, rent, and property values. Instead of weighting these measures by geographic area, we weight them by percent of the district population (which was calculated in the previous step) and percent of district houses, respectively.Footnote 4
We also include a number of controls, such as whether the councilmember is an incumbent and whether the seat serves a district or the entire city. We employ probit models with random effects for the city to account for heterogeneity across cities not captured in our models thus far.Footnote 5 We also include controls for region as defined by the Census. Regions include the South, Northeast, Midwest, and West.
In order to test our theoretical expectations, we create interactions between the percentage point change in the white population from 2000–2015 (hereafter referred to as white influx) and the current minority composition of the district. Specifically, we multiply the white influx variable with the percent minority population as of 2015 to create the interaction term for the first model. In the second and third models, we multiply the white influx by the percentage of the population that is Latino and black, respectively, in order to determine the effect of a growing white population on the election of Latino and black candidates. Again, we expect that a growing white population will have a negative effect on the election of people of color when the area's population is more evenly divided between minorities and whites. In contrast, a growing white population will not have any effect on the election of people of color when the area's population is largely comprised of Latinos and blacks and is still in the midst of preliminary stages of gentrification. Because we employ non-linear models, interpreting interaction terms differs from traditional explanations. As a consequence, we heed Golder's (Reference Golder2013) advice for “marginal effects” of interactions in probit models with random effects. Specifically, we examine how a one-unit increase in the white population (in this case a one percentage point increase) affects the probability of electing a candidate of color, conditional on the current minority population of the area (Table 2).
Table 2. Descriptive Statistics
RESULTS
In the first analysis, seen in Table 3, our results show the effect that current demographic and socioeconomic controls and percentage change measures have on electing candidates of color. First, though, we examine our control variables. The regional dummies are all insignificant in predicting the election of minority, Latino, and African American candidates. This is likely attributable to the fact that we have random effects included for the city, which helps explain some of this geographic variation. Being an incumbent, though, does seem to have a positive and statistically significant effect on the election of minority candidates generally, but, more specifically, black candidates.
Table 3. The Effect of Gentrification on Electing Minority Councilmembers
t statistics in parentheses.
*p < .05, **p < .01, ***p < .001.
Turning to the gentrification percentage change measures, many of these are statistically insignificant, with the exception of percent change measure of “gentry” professions. Areas where those employed in “gentry” professions increased proportionally were less likely to elect Latinos. However, the effect size is, again, relatively minimal, with areas that maintained the same proportion of those employed in “gentry” professions having a 9% predicted probability of electing a Latino councilmember. Compare this with a similar area where the proportion of residents in “gentry” professions increased by five percentage points, and the predicted probability is approximately 6%. Ultimately, the social and economic percentage change measures were largely insignificant in explaining the election of minority candidates. However, the racial and ethnic percentage change measures were consistently significant across models.
In areas where the minority, Hispanic, and black populations increased (see terms Minority Pct. Diff., Hispanic Pct. Diff., and Black Pct. Diff.) there is a negative and statistically significant effect on the election of minority, Hispanic, and black candidates. For instance, when the minority population remains the same in an area (i.e., Minority Pct. Difference = 0), the district has about a 51% probability of electing a minority candidate. Compare this with a district where the minority population increased by 10 percentage points (i.e., Minority Pct. Difference = 10), the predicted probability of electing a minority candidate declines to 15%. Indeed, similar patterns hold when examining the election Latino and African councilmembers. While, at first, this might seem counterintuitive, we attribute this to the racial threat hypothesis, to be further discussed and explored in later sections.
Next, we assess how the gentrification measures at their current levels performed in the model. It should be noted that many of the other socioeconomic measures of gentrification were largely insignificant, like household income and property values. Still, the percent of those with a college education has a negative and statistically significant effect on the election of Latino candidates, though the effect size is relatively minimal. Additionally, areas with higher rents are more likely to elect Latino candidates. Specifically, areas where the median rent is around $1,200 have about a 2% predicted probability of having a Latino councilmember, all else equal. Compare this with an area where the median rent is $1,400, and the predicted probability of having a Latino councilmember increases to about 10%.
With regard to current racial composition of the district, we find results consistent with the descriptive representation literature (Casellas Reference Casellas2011; Owens Reference Owens2005; Shah, Marschall, and Ruhil Reference Shah, Marschall and Ruhil2013). That is, a greater proportion of a specific racial and ethnic group leads to greater co-ethnic representation. Also, a greater proportion of whites in a district has a negative effect on the election of black representatives, but not Latino representatives. Specifically, percent Hispanic has a positive and statistically significant effect on electing a Latino candidate, as seen in column two, and percent black has a positive and statistically significant effect on electing a black candidate, as seen in column three. Still, we should be cautious with interpreting these terms because they are conditional. As a consequence, we must plot the marginal effect of electing minority, Hispanic, and black candidates in order to more easily determine the effect that a white influx has on a neighborhood of color.
Figure 2 shows the “marginal effect” plot of an increasing white population, conditional on the percent minority population. Because marginal effects are somewhat different in probit models than a traditional ordinary least squares model, we heed Golder's (Reference Golder2013) recommendations regarding interpretation of interactions in probit models with random effects. Specifically, Figure 2 depicts how a one-unit change in the white influx variable affects the probability of electing a candidate of color, conditional on the current demographic composition of the district. The solid line is the effect that a growing white population has on the probability that a candidate of color will get elected. When the solid line and dashed lines lie below the y = 0 line, then a growing white population has a negative effect on the predicted probability of electing a candidate of color. The x-axis is the percent minority population, and, thus, the plot demonstrates how a growing white population affects the probability that a minority candidate is elected, conditional on the area's current demographic composition. An influx of whites has a negative effect on the probability of electing a minority candidate when the community is largely minority. That is, areas that are pre-dominantly minority will be less likely to elect candidates of color when the white population is increasing. This finding runs somewhat counter to our expectations. We disaggregate these models by the race and ethnicity of the candidate to ensure that the effect holds across both Latino and black candidates because it is possible that the demographic changes that result from gentrification have varying effects on electing Latino and black candidates, as can be seen in Figures 3 and 4, respectively.
Figure 2. The Effect of a Growing White Population on Electing a Minority Councilmember, Conditional on Current Minority Population.
Figure 3. The Effect of a Growing White Population on Electing a Hispanic Councilmember, Conditional on Current Hispanic Population.
Figure 4. The Effect of a Growing White Population on Electing a Black Councilmember, Conditional on Current Black Population.
Figure 3 depicts how a one-unit increase in the white influx variable (in this case, the white population increased by one percentage point) affects the probability of electing a Hispanic councilmember, conditional on the current Latino composition of the district. When the solid line and dashed lines encompass the y = 0 line, a growing white population has a statistically insignificant effect on electing a Latino candidate. In the case of Latino candidates, a growing white population does not necessarily affect the chances that a Latino candidate is elected. The solid line shifts in ways that are congruent with our hypotheses, where a growing white population has the greatest negative effect when the area is about 65% Latino. When the population is between 65 and 100% Latino, a growing white population has a less negative effect on electing a Latino candidate. Nevertheless, the confidence intervals encompass the y = 0 line, suggesting that these results are statistically insignificant.
In Figure 4, we demonstrate how a one-unit increase in the white influx variable affects the probability of electing a black councilmember, conditional on the current black composition of the district. A growing white population has the greatest negative effect on electing a black candidate when black residents comprise approximately 50% of the area's population, congruent with our Political Displacement Hypothesis. As the population moves from approximately 50% black to 100% black, a growing white population has a less negative effect on electing a black candidate, which supports our Political Mobilization Hypothesis. That is, a growing white population still has a negative effect on electing black candidates, but it is less negative and even approaches statistical insignificance as the community becomes 100% black. Take, for instance, a community where 90% of residents are African American, and the white population has grown by about five percentage points. The probability of electing a black councilmember to represent that community is 66%. However, if the community was instead 50% African American, but still experienced the same inflow of whites, the probability of electing a black councilmember declines to 29%. Compare these figures with communities that have not experienced inflows of whites, and they are more likely to elect co-ethnic representatives. For instance, communities that are 50 and 90% African American with no real changes in the proportions of whites maintain predicted probabilities of 46 and 91%, respectively. This suggests that a growing white population has a significant effect on the election of African American candidates.
In sum, we find support for our Political Displacement Hypothesis and some support for our Political Mobilization Hypothesis—a growing white population has the greatest negative effect on electing a black candidate when the area is approximately half black and has less of a negative effect as the community becomes more black proportionally. However, results for Latinos, while in the hypothesized direction, were statistically insignificant. We explore these findings further in the Discussion section.
DISCUSSION
The findings of the previous analyses show that the demographic changes resulting from gentrification have varying effects on descriptive representation. That is, as districts become whiter, they are less likely to elect black councilmembers. This is especially true when the population is approximately half black, offering evidence for the Political Displacement Hypothesis. In contrast, a growing white population has less of a negative effect when the community is largely black, providing some support for the Political Mobilization Hypothesis. The effect is negative and significant when the population is nearly 100% black, but not necessarily insignificant, as originally posited. The findings echo other research and anecdotal evidence that as gentrification is well underway, longstanding residents are increasingly marginalized politically. For instance, Hern (Reference Hern2016) details the effects of gentrification in a historically black neighborhood in Northeast Portland, OR. Hern describes how the local African American community has increasingly felt excluded from decisions made by the Portland Development Commission (PDC), a local government agency. As the PDC has sought to develop the area with middle-class shops and services, black residents have felt politically marginalized, as John Washington of North Northeast Business Association, explains in an interview with Hern:
“We are so precarious in this neighborhood. We have no anchors, no cornerstones. How can we dig in deep enough that they can't remove us?” (Hern Reference Hern2016, p. 62)
This interview illustrates that when gentrification has fully taken hold, original residents—even those in positions of power—feel marginalized. In gentrifying and gentrified neighborhoods, local governments shift to serve the newly dominant constituents, and thus, it is no surprise then that original residents and their representatives, in this case black residents and their representatives, feel marginalized.
Our findings regarding Latino representatives follow the hypothesized pattern, but are nevertheless insignificant. These null results coupled with the significant ones for African Americans provide more support for the notion that different racial and ethnic groups have differential responses to gentrification, per Michener and Wong (Reference Michener and Wong2018). In addition, context matters so that the dynamics of racial coalitions vary across cities and which candidates are supported by the political establishment will be conditional on many factors. We may expect that a community that is growing whiter may not be opposed to electing minority candidates, especially in communities that are predominantly Latino. And for Latinos, the literature increasingly suggests that ethnic-based voting is not the primary reason that Hispanics gain office, as Latinos have been able to increasingly win in state legislative and congressional districts without Latino majorities (Casellas Reference Casellas2011). We also know that the lack of Latino candidates running for office is more of a hindrance to increased Latino representation in legislatures rather than ethnic-based voting (Juenke Reference Juenke2014).
In areas where the minority, Hispanic, and black populations increased, there is a negative and statistically significant effect on the election of minority, Hispanic, and black candidates. For instance, when the minority population increases by 10 percentage points, the predicted probability of electing a minority councilmember declines from 51 to 14%. We witness similar patterns with African Americans as well. When the African American population increases by 10 percentage points, the predicted probability of electing a black councilmember declines from 20 to 2%. We also sought out alternative specifications to explore if and how this effect might vary depending on the proportion of non-Hispanic whites as seen in Appendix B. We find that when the white population comprises more than 50% of the population, an influx of minorities has a negative effect on the probability of electing a person of color, which suggests that there are thresholds at which the racial threat is triggered. For instance, an influx of minorities into a predominantly minority community does not decrease the probability of a minority being elected to the local council. But an influx of minorities into a predominantly white community does, in fact, decrease the probability of electing a minority.
Ultimately, in localities where the minority population is growing, the probability of electing candidates of color decreases, suggesting that perhaps feelings of racial threat (Blalock Reference Blalock1967; Giles and Hertz Reference Giles and Hertz1994; Key and Heard Reference Key and Heard1949) are at play. That is, original residents likely feel threatened by a growing racial or ethnic group and consequently vote in ways to deter their political power, in this case not electing candidates of color. This speaks to an expansive body of literature, including V.O. Key's foundational Southern Politics in State and Nation (1949) where he finds that larger populations of African Americans breed fear and contempt among whites, which in turn leads to a more cohesive and racially conservative voting bloc. Indeed, more recent scholarship, like that of Shah (Reference Shah2017), finds a similar dynamic when examining mayoral races in American cities. Minority residents are therefore presented with complex choices—move to a new area and perhaps be politically marginalized or stay in their current gentrifying neighborhood and risk political displacement. Either option will not politically empower minorities and is largely congruent with the helplessness and lack of political power that longstanding residents expressed in other research (Hyra Reference Hyra2014; Knotts and Haspel Reference Knotts and Haspel2006).
CONCLUSION
We find evidence that gentrification negatively impacts minority descriptive representation, specifically black descriptive representation. Gentrification that results in a growing white population negatively affects the election of black councilmembers, and the effect is particularly pronounced when the black population is close to losing its dominance (i.e., they comprise roughly half of the area's population). When the community is largely black, though, an increasing white population has a less negative effect on the election of black councilmembers. This suggests that gentrification, here conceptualized as a growing white population, has varying effects on descriptive representation, depending on the degree to which the area had already gentrified. Yet, the election of Latino councilmembers did not seem to be affected by a growing white population, as the relationship was largely insignificant.
With regard to the null Latino findings, we would like to point to a few strands of literature to help shine light on these somewhat puzzling results. First, Latino political preferences are quite diverse, particularly with regard to national or regional origin (De la Garza et al. Reference De la Garza, Desipio, Garcia, Garcia and Falcon1992; Griffin and Newman Reference Griffin and Newman2007; Leal Reference Leal, Espino, Leal and Meier2008; Uhlaner and Garcia Reference Uhlaner and Garcia2002). Thus, it is possible that this within-group diversity lends itself to differential political responses to gentrification and, in turn, the null findings we see here. It is also possible that Latinos would accept a growing white population (and their subsequent white representatives) as a way to maintain some sense of power, per Eisinger (Reference Eisinger1980), especially because of evidence that Latino/as see themselves as having more in common with whites than other minority groups, although this is dependent on many factors (Wilkinson Reference Wilkinson2014). In any case, future research should explore Latino representation in the face of gentrification further.
Our research also seemed to touch on a longstanding issue about political responses to growing minority populations, specifically that growing minority, Hispanic, and Latino populations seemed to deter the election of co-ethnic representatives. We attribute this to Blalock's (Reference Blalock1967) racial threat hypothesis. Still, subsequent scholarship would be served in exploring how original inhabitants of gentrifying areas feel about these residential choices they are faced with and how these choices affect their sense of political power.
In addition, scholars should continue to explore the effects of gentrification on minority substantive representation, through descriptive representation or perhaps other mechanisms as well. Some evidence suggests that Latino growth in particular can have differing impacts on substantive representation where the growth of Latino citizen populations leads to increased substantive representation while the growth of non-citizen populations leads to decreased substantive representation (Matsubayashi and Rocha Reference Matsubayashi and Rocha2012). As the process of gentrification in American cities continues at a fast pace, future scholarship should continue to monitor the effects of this phenomenon on the representation of minorities at all levels of government.
Indeed, much research on gentrification focuses on the phenomena at the community or neighborhood level (Hyra Reference Hyra2014; Knotts and Haspel Reference Knotts and Haspel2006; Martin Reference Martin2007) whereas our work takes a more aggregated approach, examining the effects on the district, ward, and city as a whole. Future research would be well served in examining gentrification at even higher levels of geography. Take, for instance, Alexandria Ocasio-Cortez, who was elected to New York's 14th Congressional district in 2018.Footnote 6 And while we do not weigh in on the validity of this specific argument, we think that it is important that the effects of gentrification on political power are examined from higher levels of geography, including state legislatures and the US Congress.
We also earlier made note of the role that political machines play in who gets elected in local elections. And while incorporating the strength of political machines into empirical models seems unfeasible currently, how political machines react to the changing demographics that come with gentrification is certainly a topic ripe for analysis. Michael Jones-Correa (Reference Jones-Correa, Yvette Marie and Lawrence2000) already explores how political machines help (or hinder) the political incorporation of outsiders and immigrants in years past, but little work explores how machines currently react to the influx of an already dominant racial group. Machines likely see these changing demographics that accompany gentrification as a threat, but rather than attempting to suppress these racial and ethnic shifts, they might attempt to co-opt them. In either case, this would be a topic worthy of further exploration.
While we have focused primarily on the experiences of original residents, scholars would be well served in exploring the political preferences and behaviors of gentrifiers, particularly those who seemed to have the most effect in the analyses here—white gentrifiers. It is possible that some whites who move into majority–minority areas are more supportive of candidates of color, given that they are comfortable living in a predominantly minority area. This comports with Hajnal's (Reference Hajnal2006) research where some whites may be more open to supporting minority candidates, given their previous experience with black leadership. Specifically, Hajnal (Reference Hajnal2006) finds that whites will electorally support black candidates when whites have had positive experiences with black representation and less supportive when they have little information regarding black officials. In the case of gentrification then, when a new white resident moves to an area and experiences positive minority leadership, do they electorally support that minority representative? Indeed, we should be exploring the political preferences and behavior of these gentrifiers and how it, in turn, affects original residents, but also political outcomes for the community as a whole.
Appendix A
Table A1 and Figures A1–A3.
Table A1. The Effect of Gentrification on Electing Minority Councilmembers in Districts
Figure A1. The Effect of a Growing White Population on Electing a Minority Councilmember, Conditional on Current Minority Population.
Figure A2. The Effect of a Growing White Population on Electing a Hispanic Councilmember, Conditional on Current Hispanic Population.
Figure A3. The Effect of a Growing White Population on Electing a Black Councilmember, Conditional on Current Black Population.
Appendix B
Here, we have interacted the minority Percent Difference measures with the current demographics. We would expect that when the minority, Latino, and black populations grow proportionally in an area for them to conditionally affect whether a person of color is elected to office. Specifically, when the minority, Latino, and black populations have grown in an area that is majority–minority, the probability of electing a person of color will increase. In contrast, when the minority, Latino, and black populations have grown in an area that is predominantly white, we expect the probability of electing a person of color will decrease, thus providing support for the racial threat hypothesis. We find evidence that a growing minority population only has a negative effect when the non-Hispanic white population maintains slightly more than half of the population (Table B1).
In Figures B1–B3, we have plotted the marginal effects of a growing minority, Latino, and black community and how it varies by the current demographic composition of the electoral district. Specifically, Figures B1–B3 show the effect of a growing minority population accompanied by an equal decline in the non-Hispanic white population. Figure B1 demonstrates that a growing minority population has a negative effect on the probability of electing a minority candidate when non-Hispanics currently comprise more than 50% of the population. Figure A2 has a similar relationship where a growing Hispanic population has a negative effect on the election of Latino candidates, and the effect is particularly pronounced when the white population is greater than 50%. Finally, Figure A3 depicts a very similar relationship where a growing African American population has a positive effect on the probability of electing a black candidate when the non-Hispanic whites comprise less than 50% of the population. Conversely, a growing African American population has a negative effect on the probability of electing a black candidate when the non-Hispanic whites comprise more than 50% of the population.
Table B1. The Effect of Gentrification on Electing Minority Councilmembers in Districts
Figure B1. The Effect of a Growing Minority Population on Electing a Minority Councilmember, Conditional on Current White Population.
Figure B2. The Effect of a Growing Hispanic Population on Electing a Hispanic Councilmember, Conditional on Current White Population.
Figure B3. The Effect of a Growing Black Population on Electing a Black Councilmember, Conditional on Current White Population.
