Abstract
Conventional narratives of the 2020 presidential election typically overlook the growing importance of voters of color in suburbia, especially in battleground states such as Georgia and Virginia. Using novel precinct-level voting data for twenty-two states for the 2016 and 2020 presidential elections, we analyze the relationship between precinct-level Democratic vote shares and the racial composition of suburban precincts. Predominantly White suburban precincts turned modestly toward the Democratic Party in 2020 relative to 2016, but had the election been decided in the precincts in which most White suburban voters live, Donald Trump would have won both elections handily. The outcome in 2020 depended on turnout in heavily Black suburban precincts, which voted overwhelmingly for Biden, and in Asian and Latinx precincts that also supported Democrats, though less strongly. As the suburbs become increasingly diverse, the new racial demography of suburbs should change conventional understandings of voting behavior in these spaces.
On January 7, 2021, voters in Georgia elected two Democratic senators, giving the newly elected Democratic president a working majority in both houses of Congress. These election results, which came as a surprise to many, were part of a longer-term shift among what had been some solidly Republican states toward the Democratic Party, a movement that became increasingly apparent in election cycles from 2008 onward. How did this happen? More specifically, where did this happen? The media’s answer to this question has often focused on the leftward shift among educated, White suburbanites, particularly among White women (Badger 2020; Badger and Bui 2020; Lerer 2020), bringing renewed attention to the political dynamics of suburbs (Eligon 2020; Pew Research Center 2020; Vozzella et al. 2021). A second narrative emphasized the importance of voters of color, particularly Black voters, in the 2020 elections. In practice, however, these two explanations overlap: voting trends in suburbs coincide with a rapidly changing suburban racial demography, in which a majority of all people of color—not just Whites—now live, and vote (Frey 2018; Lacy 2016). Georgia, a key state in the 2020 elections, is a prime example. In the 2020 Census, 89 percent of the voting-age Black population in greater Atlanta lived in its suburbs. This figure is 94 percent for Asians and 96 percent for Latinxs. These suburban voting-aged populations have increased substantially since 2010: the numbers of voting-age Blacks living in suburban Atlanta increased by 326,000 between 2010 and 2020, of Latinx by 149,000, and of Asian-origin by 112,000 over that decade.1 Media and scholarly narratives often implicitly assume that suburbs and suburban votes are still White, and that suburban voting patterns have changed largely because White people changed their voting behavior.2 However, it could also be that suburban voting patterns changed as suburbs themselves became more racially and ethnically diverse. This article argues that the conversation on voting, race and suburbs in the United States needs to account for the full complexity of demographic shifts in suburbia, especially as racial and ethnic minorities groups continue to suburbanize.
In this analysis, we examine the racial contexts of suburbs in twenty-two states and their relationship to votes cast in the 2016 and 2020 presidential elections to answer the question of how election results vary among suburbs according to their racial composition.3 Did predominantly White suburbs still largely vote Trump, and suburbs primarily composed of people of color vote for Clinton in 2016 and then Biden in 2020? Are college-educated Whites in more diverse suburbs more likely to vote Democratic than their counterparts in mostly White suburbs? Novel precinct-level voting datasets compiled by the University of Florida Voting and Elections Science Team, Tufts University Metric Geometry and Gerrymandering Group, and the New York Times, allow us to go beyond the usual county-level analyses of vote choice. In this analysis of voting in suburban precincts, the median precinct size is 1,700 residents. By contrast, counties, the conventional unit of analysis, are typically much larger, often being home to nearly a million people or more in large metro areas. Matching these voting data with block-group-level census data allow us to better understand the micro-level contexts of voting.
The central question remains: if voting in suburbs has shifted toward Democrats, has it shifted in Whiter suburbs, all suburbs, or largely in suburbs with people of color? To anticipate the findings we present, controlling for critical covariates like education and spatial dependence, our analyses suggest that majority-White suburban precincts did in fact shift modestly toward Biden, the Democratic candidate in 2020, relative to Clinton in 2016. However, the Whitest suburbs—suburban places that are more than 80 percent White, which is where most White suburban residents live—had an average Democratic vote share below 50 percent. The Blackest suburbs showed the highest support for Democrats by far: suburban precincts greater than 75 percent Black showed predicted Democratic support above 65 percent in both election years. Asian and Latinx precincts were both in the middle with more moderate support for Democratic candidates.
Dominant narratives of race and political participation often focus on White suburbanites or people of color as a whole. This analysis shows that to understand the geography of race and voting in the United States requires an understanding of the local suburban contexts of people of color as well. We opened this introduction with the story of Georgia, whose suburbs are increasingly ethnically and racially diverse; but we find the patterns of voting seen in Georgia are mirrored in suburbs across the United States. This analysis provides a foundation for future research that interrogates whether suburban context influences voters of color as well as whether the changing racial structure of suburbs may alter White voting behavior.
THEORIES OF RACE, PLACE, AND VOTING
Folk notions of suburbs (such as the all-White suburb) are largely rooted in mid-twentieth century Black urbanization, segregation, and White Flight. White people moved out of central cities into suburbs to maintain political autonomy of their communities as Blacks moved into cities (Kruse 2007; Trounstine 2018). As suburbs emerged as a distinct residential space in the early twentieth century (Wood 1959; Jackson 1985; Duaney, Plater-Zyberk, and Speck 2000; Archer, Sandul, and Solomonson 2015; Trounstine 2018), it was apparent that they were also become distinctive political spaces: overwhelmingly White and middle class, preoccupied by concerns about the preservation of property values, the provision of services, and lower tax rates. These issues became central to a conservative ideology emerging from the post–World War II suburbs (Jackson 1985; McGirr 2001; Lassiter 2006; Freund 2007; Kruse 2007; Thompson 2012), and to a new Republican coalition that successfully appealed to suburban voters based on pro-growth, low-tax economic policies, a “color-blind” racialized ideology of home ownership and neighborhood schools, and patriotic, “pro-family” values. As a result, suburban voters were seen as safely Republican, counterbalancing the more diverse, urban constituencies more likely to back Democratic candidates (Gainsborough 2001; Williamson 2008; Trende 2021).
By the 1990s, however, signs of stress fractures were evident in what seemed to be this solidly conservative voting bloc. One indication was that after 1992 White suburban female voters began to shift their vote preferences slightly toward the Democratic Party (Gainsborough 2005). Another was that suburbs themselves were becoming less racially homogeneous (Wiese 2004; Cheng 2013; Lassiter and Niedt 2013), the diversification of suburbs accelerating through the 1990s (Frey 2011; Orfield and Luce 2013), driven in part by the passage of antidiscrimination legislation as well as the choice among a majority of new immigrant arrivals to settle directly in suburbs rather than in center cities (Kruse and Sugrue 2006; Singer, Hardwick, and Brettell 2008; Katz et al. 2010). Shifts in the demographics of suburbs, and (not unrelatedly) in the voting preferences of suburban voters have had significant consequences in electoral coalitions and outcomes. By the early 2000s, observers were noting that suburban voters were moving in the direction of the Democratic Party, abandoning their decades-long position as a reliable pillar of the Republican coalition (Lang, Sanchez, and Berube 2008; Trende 2021).
However, exactly which voters in suburbia have shifted toward the Democratic Party remains unclear. Electoral outcomes across recent electoral cycles’ show changes in the preferences of suburban voters, but not precisely among which suburban voters. Sean Trende (2021) and Jack Weisman (2019) largely assume changes in voter preferences are occurring among White voters. Seth McKee and Daron Shaw (2003) look at voting trends in a limited set of localities to argue that the shifts are in part due to demographic changes in suburbs but are occurring among White voters as well. Jeremy Teigen, Daron Shaw, and Seth McKee (2017) look at changes in voter preferences across time in a limited set of suburban locations and observe that these changes are more apparent in urban suburbs—that is, that the shift toward the Democratic Party is correlated with population density. It is still unclear, then, how much of the changes in voting are due to shifts among White voters alone, or to the increasingly diverse ethnic and racial composition of suburbia and its implications for Democratic vote share. Some of the blurriness in the existing analysis is due to limitations in the data. Most studies have yet to examine demographic and political shifts in suburbs across the United States, and many studies examining demographic shifts have relied on county-level demographic data (for an analysis of precinct-level data in a single metropolitan area, see Kinsella, McTague, and Raleigh 2015). Using a novel set of precinct-level data across multiple states, we expect, along the lines of Teigen, Shaw, and McKee (2017), to find that changes in voter preferences in suburbs are due only in part to shifts among White voters, reflecting the demographic changes in the racial and ethnic composition of suburbia as well, as Black, Latinx, and Asian Americans increasingly make the choice to live in suburbs.
DATA AND METHODS
Diverging from most political-geographic studies, which use counties as the unit of analysis, in this article we examine voter preferences at the precinct level and changes in these preferences from 2016 to 2020. Precincts are the smallest geographic unit for electoral districts, and residents of a precinct vote at a specific polling station. Precincts differ from counties in that they are a much finer grain geography. In this study, for example, in 2020 the median precinct size was 1,693 people, and the median county size for metropolitan counties in the sample approximated a population of 129,000. Precincts allow us a much better understanding of the hyperlocal contexts of suburban voting. These precinct-level voter turnout data as well as geographic shapefiles for 2016 came from the University of Florida’s Voting and Election Science Team (VEST 2018), except data from Ohio, which come from the Tufts University Metric Geometry and Gerrymandering Group (MGGG 2021). Precinct-level data and shapefiles for the 2020 election come from the New York Times (Park et al. 2021). Precinct-level data historically have not been widely publicly available, and we have immense gratitude for these teams and their efforts in compiling these data, often working directly with local governments. Because these data are not yet comprehensive, we limit our sample to states that offer complete metropolitan coverage in both 2016 and 2020. Our sample covers metropolitan areas and their suburbs in twenty-two states: Arizona, California, Colorado, Delaware, Florida, Georgia, Hawaii, Illinois, Maryland, Massachusetts, Minnesota, Nevada, New Mexico, North Carolina, Ohio, Oregon, Pennsylvania, South Carolina, Texas, Utah, Washington, and Wisconsin.
The sample disproportionately covers Democratic-leaning states (figure 1), and data were not available for much of the lower Midwest, Mountain West, and Deep South—states with higher concentrations of White residents who are strongly Republican. As we discuss, this sample also moderately undersamples the White suburban populace. Together, these selection issues suggest that our results likely overreport support for the Democratic Party relative to suburbs overall.
Racial and ethnic composition data as well as other critical demographic characteristics used in this analysis come from the American Community Survey (ACS) 5-year estimates, which we downloaded from the National Historical Geographic Information System (Manson et al. 2021). Our racial and ethnic groups of interest are Latinxs, non-Latinx Black, non-Latinx Asian, and non-Latinx White (Latinx, Black, Asian, and White). We attach 2012–2016 ACS estimates to 2016 presidential election data. Due to data availability, we attach 2015–2019 ACS data to the 2020 presidential election data (2016–2020 data were not available at the time of publication). We attach census data to precinct data by overlaying ACS block-group shapefiles (the finest geographic unit in the census) with precinct data. Block-groups are appropriate for this analysis because they are a fine-grain geography that matches closely with the size of precincts. In our sample, the median precinct size was roughly 1,700 residents and the median block-group size roughly 1,300. We weight estimates for precincts based on areal overlap. For example, if 60 percent of a block-group falls within a precinct, we weight that precincts population estimates by 0.6. On median, precincts in our analysis overlapped with two block-groups when block-groups made up at least 5 percent of the precinct.4 Together, these electoral and census data, matched at the precinct level, allow an examination of diversity in demographics, geography, and politics across a broad range of suburbs in the United States.
For this article we define suburbs as parts of metropolitan areas that are not the central city. Following census definitions, the central city is the largest city by population within a metropolitan area. For example, the Los Angeles metro area consists of Los Angeles and Orange Counties. A suburban precinct of Los Angeles therefore is a precinct either in Orange County or the parts of Los Angeles County not within the boundaries of Los Angeles City. We geocode precinct shapefiles to census shapefiles using precincts’ central longitude-latitude coordinates. We exclude nonmetropolitan precincts because conventional research treats them as rural and not suburban. These inclusion criteria result in 61,599 suburban precincts in 2016 and 59,024 suburban precincts in 2020 with complete data. To provide context, we include central city precincts in some analyses, which a provide an additional 22,036 precincts in 2016 and 20,853 precincts in 2020. Many precincts change boundaries between election years. We associate precincts between election years based on which precincts show the greatest areal overlap. This is imperfect, but we address issues of spatial clustering using a spatial lag variable, discussed in the regression section.
In the sample of suburban precincts, Clinton carried 48.4 percent of the vote (Trump 45.7 percent) in 2016 and Biden carried 51.8 percent of the vote (Trump 46.3 percent). In the sample, populations of color grew over the period and the White population declined: the racial composition of the sample was 8 percent Black, 19 percent Latinx, 7 percent Asian, and 63 percent White in 2016 and 9 percent Black, 20 percent Latinx, 8 percent Asian, and 61 percent White in 2020. This sample moderately oversamples the Latinx population and undersamples Whites relative to the composition of the suburban population overall in 2019: 10 percent Black, 18 percent Latinx, 7 percent Asian, and 63 percent White. Table 1 provides summary precinct-level summary statistics.
Summary Statistics of Racial Composition and Democratic Vote Share
On average the precincts in the sample were 8 percent Black, 17 percent Latinx, 6 percent Asian, and 65 percent White. Overall, in 2020 the Democratic vote share in these precincts was 50.2 percent. Looking at the quantiles shows that roughly 75 percent of precincts were at least near-majority White (the 25th percentile, 47 percent White), but that 10 percent of precincts were at least 22 percent Black, 49 percent Latinx, or 17 percent Asian (90th percentile for the respective groups).5
For our dependent variable, we focus on suburban precinct-level Democratic vote shares. We calculate this variable by dividing the number of Democratic votes by the number of total votes.
Our focal independent variables center on racial and ethnic composition of precincts: Latinx shares and non-Latinx Black, Asian, and White shares.
Because Democratic vote share is a proportion variable arising from count data, we run these models as logistic regressions, running separate regressions for each independent variable, and interacting each independent variable with election year to see whether the effect changes between 2016 and 2020. We include several controls: year to address period specific effects; educational attainment (percentage over twenty-five with a bachelor’s degree) to address how education associates with voting; homeownership as a measure of wealth of the precinct;6 logged population to control for the varying sizes of precincts; and state fixed effects to control for unobserved state-level characteristics that shape local politics.
The models include a spatial lag variable that captures the Democratic vote share among neighboring suburban precincts to parameterize spatial autocorrelation (spatial dependence) (see Anselin 1990; Anselin 1988; for a discussion of spatial dependence and spatial heterogeneity, see Voss, Curtis White, and Hammer 2006). This spatial lag variable is important because voting generally clusters in space in which Democratic-leaning precincts are close to other Democratic-leaning precincts. Cases are not independent with spatial autocorrelation. Put differently, we have some information on a precinct based on having information from its surrounding precincts. Autocorrelation likely biases regression results, which the spatial lag variable addresses. Using a queen’s contiguity matrix, we find all the precincts that touch a precinct (the neighbors) and calculate the overall share of Democratic votes among these neighbors as the spatial lag. Last, we use robust standard errors to address remaining heteroscedasticity.
RESULTS
Our guiding research question is whether voting in suburbs has shifted toward Democrats, and if so whether it has shifted in Whiter suburbs, all suburbs, or largely in suburbs with people of color. We begin our results with figure 2—smoothed means graphs of the Democratic vote share by racial composition, comparing suburban precincts with urban precincts. The dashed line shows data from 2020 and the solid line from 2016. This analysis primarily focuses on suburban precincts, but here we provide data for urban (central city) precincts as well to better contextualize suburban voting. These graphs show two major findings: first, urban precincts generally show larger Democratic turnout relative to suburban precincts, except for predominantly Black precincts. Suburban spaces still show the legacy of conservatism relative to their urban counterparts. Second, patterns of voting vary substantially by the racial composition of precincts.
Smoothed Means of Democratic Vote Share by Racial Composition and Year
Source: Authors’ tabulation.
Note: Racial composition of precincts calculated by associating precincts with block-group-level ACS data. American Community Survey, n = 163,512 suburban-precinct years.
Increasing Black shares associates with dramatic increases in Democratic votes, in which the Blackest precincts show near perfect support for the Democratic Party in both years regardless of urbanicity. In other words, Black spaces turn out strongly for Democratic candidates in both cities and suburbs, and for Clinton and Biden in 2016 and 2020. Increasing Latinx shares associate with greater Democratic vote share as well, though to a more moderate degree; the highest point being roughly 75 percent in suburbs and 80 percent in cities in 2016. Notably, majority-Latinx precincts showed a decrease in Democratic vote share in 2020 relative to 2016, although they still maintained strong support for Biden, at or above 60 percent. Asian composition shows a curvilinear relationship with increasing Democratic vote share that plateaus around 65 percent once a precinct is 25 percent Asian or greater. This result suggests a concentration effect in precincts that are between 0 percent and 25 percent Asian—the precincts in which most Asians in the sample live. High Asian concentrations do not equate to equivalently high Democratic voter share, but a Democratic vote share above 60 percent is still an indication of strong Democratic support. Majority-Asian suburban precincts (n = 720) are not numerous, as shown by the larger standard error bands.
Last, White-majority suburban precincts are much less likely to have supported the Democratic Party, consistent with earlier analyses of White conservatism in suburbia (Jackson 1985; McGirr 2001; Lassiter 2006; Freund 2007; Kruse 2007; Thompson 2012). The Whitest suburban precincts on average show a Democratic vote share of roughly of 30 percent. This changed modestly between the 2016 and 2020 election cycles: the least White precincts became slightly less Democratic in 2020, likely driven by Latinx turnout, and majority-White precincts became slightly more Democratic in 2020 relative to 2016. As a reminder, half of all White people in the sample lived in a precinct that was roughly 80 percent White. Thus, even in our sample, which disproportionately examines Democratic-leaning states, most White people lived in White precincts in which the average Democratic vote share was below 50 percent. Overall, these findings are consistent with the narrative put forward by media and researchers: between 2016 and 2020 a modest shift is evident toward the Democratic Party in majority-White suburbs, which contributed to a majority of all suburban voters in the 2020 election backing Biden, the Democratic presidential candidate. However, the shifting racial demography of suburbs played a substantial role in the changing voting patterns of suburbs, with racial and ethnic minorities—certainly Black voters, but Asian American and Latinx voters as well—significantly more likely to vote for Democratic candidates than their White suburban counterparts.
LOOKING LOCALLY
To provide a visualization of the trends across suburbs reported in the preceding section, and how these play out in specific metropolitan areas with distinct suburban populations, we take a closer look at maps showing the demographics of suburbanization in Atlanta; Columbia, South Carolina; Los Angeles; and Milwaukee. These metropolitan areas vary by region but also, more crucially, in the racial composition of their suburbs. Atlanta shows multiracial suburbanization across all four racial-ethnic groups; Columbia primarily reflects the suburbanization of Whites and Blacks; Los Angeles indicates the suburbanization primarily of Asian and Latinx populations with small Black suburban populations; and, last, Milwaukee provides a picture more akin to mid-twentieth century, with predominantly White suburbanization. Although each of these metropolitan areas has a slightly different narrative, taken together, these maps corroborate the broader narrative illustrated in figure 2: the Blackest precincts in these suburbs show the highest Democratic vote share; precincts with larger Asian and Latinx populations show strong Democratic support, but less dramatically than in Black precincts; and the Whitest precincts show the lowest levels of support for the Democratic Party.
We start with Atlanta. Suburban Atlanta was a critical battleground in the 2020 election, in which Georgia flipped to the Democratic Party for the first time since 1992. Moreover, Greater Atlanta shows high suburban concentrations for all four racial groups examined in this study—more than 90 percent of voters of color in Greater Atlanta live in suburbs. Figure 3 shows the maps of racial composition and Democratic vote share for suburban precincts in greater Atlanta. Lighter shades indicate greater shares of the respective race/ethnic group or Democratic vote for 2020. Comparing the White composition to the Democratic vote map, the share of Democratic vote looks like the inverse of the White composition map. Put differently, the suburban precincts with the largest populations of color also show the largest Democratic vote shares. The Blackest precincts show the brightest spots for the Democratic Party. We see this southwest of the city, as well as in counties south and east of the city. Suburban precincts with greater Asian- or Latinx-origin populations show strong support for Biden, however, this support is more modest than that found in metropolitan Atlanta’s Blackest precincts. This pattern is evident in multiracial counties northeast of the city in DeKalb and Gwinnett counties. (For similar examples, see the map of the Maryland suburbs of Washington, D.C., and Baltimore in figures A.2 and A.3).
Racial Composition and 2020 Democratic Vote in the Inner Ring Suburban Greater Atlanta
Source: IPUMS NHGIS (Manson 2021) and New York Times (Park 2021).
Note: Percent population and Democratic vote are shaded at the precinct-level. For ease of interpretation, we only include counties that border Fulton County, the county in which Atlanta is located. Most people of color in the Atlanta metro area live in these counties. For the full maps of the Atlanta metro area, see figure A.1.
Columbia, South Carolina (figure 4), provides an example of Black suburbanization without the complicating factors of other populations of color. Columbia has a substantial Black suburban population and much smaller Asian and Latinx suburban populations. Further, it is situated in a conservative state which has voted for a Republican presidential candidate for the last eleven elections. The maps presented in figure 4 mirror in many respects the maps of metropolitan Atlanta, indicating high Democratic turnout in Blacker precincts. Despite the state’s strong conservative tilt, the precincts with the highest percentages of Black populations both south and north of the city show the highest Democratic turnout relative to other suburban parts of the metropolitan area.
Map of Racial Composition and 2020 Democratic Vote in Suburban Greater Columbia
Source: IPUMS NHGIS (Manson 2021) and New York Times (Park 2021).
Note: Percentage population and Democratic vote are shaded at the precinct level.
Los Angeles (figure 5), like Atlanta, provides an example of a metropolitan area with highly suburbanized populations of color: people of color make up 70 percent of Los Angeles’ suburban population, but its non-White suburban residents are primarily of Latinx and Asian origin. Majority-Black suburban precincts are relatively small in number, and Black residents make up less than 6 percent of Los Angeles’ suburban population (U.S. Census Bureau 2022). Although suburban Los Angeles leaned toward the Democratic Party even in some White precincts (such as the coastal areas west and south of the city), it has few concentrations of heavily Democratic turnout like those in the Blackest precincts in metro areas such as Atlanta and Columbia. Despite some suburbs in Los Angeles showing few White residents, no suburbs with concentrated Asian or Latinx populations show Democratic voter turnout nearing 100 percent. This point is emphasized when considering the liberal context of southern California. Black precincts in conservative or battleground states show higher Democratic vote shares than non-Black precincts in “solidly blue” California.
Map of Racial Composition and 2020 Democratic Vote in Suburban Greater Los Angeles
Source: IPUMS NHGIS (Manson 2021) and New York Times (Park 2021).
Note: Percentage population and Democratic vote are shaded at the precinct level. For visualization purposes, we exclude Catalina Island.
Last, Milwaukee (figure 6), located in a northern state politically contested in both 2016 and 2020, serves as a contrast to suburbs in Atlanta, Columbia, and Los Angeles. Atlanta shows high diversity and multiracial suburbanization, Columbia shows suburbanization primarily of Blacks and Whites, and Los Angeles shows suburbanization primarily of Latinx, Asian, and White residents. Unlike suburbs in these metropolitan areas, the suburbs of Milwaukee are predominantly White, racial segregation largely occurring between suburbs and the city rather than within suburbs. The demography of Milwaukee’s suburbs is reminiscent of the classic conception of the White postwar suburb of the twentieth century when few people of color lived in suburbs: 99 percent of Milwaukee suburban precincts are at least 49 percent White and 95 percent are at least 68 percent White. Perhaps unsurprisingly, variability is less in Democratic vote share in these suburbs than in the metro areas shown previously. In Milwaukee, 75 percent of metropolitan precincts have a Democratic vote share below 48 percent, White suburbs showing low support for Democrats.
Map of Racial Composition and 2020 Democratic Vote in Suburban Greater Milwaukee
Source: IPUMS NHGIS (Manson 2021) and New York Times (Park 2021).
Note: Percentage population and Democratic vote are shaded at the precinct level.
Despite considerable variation in the composition and degree of non-White suburbanization in Atlanta, Columbia, Los Angeles, and Milwaukee, these four metropolitan areas also exhibit common patterns despite their differences, commonalities reflected in the patterns in the national data presented earlier. Across all four areas, regardless of the state context, voters in Black precincts are the most likely to vote Democratic, voters in more heavily Latino and Asian suburban precincts vote Democratic but less heavily than in Black precincts, and voters in White precincts are the least likely to back Democratic candidates.
REGRESSION ANALYSIS
Figure 7 shows the predictive margins of Democratic vote share by racial composition and urbanicity generated by logistic regressions; dashed lines represent 2020, solid represent 2016, circles represent suburban precincts, and triangles represent urban precincts (for regression parameter estimates, see table A.2). These results control for critical covariates like educational attainment, homeownership, spatial dependence, and state fixed effects. Regression analysis confirms the general trends displayed in the descriptive figure 2: suburban spaces show lower Democratic vote turnout than in urban spaces, and the racial composition of suburban spaces matters substantially for their voting patterns. More heavily White suburban precincts have the lowest voter share for the Democratic Party, consistent with dominant theories of race and voting. Put differently, suburbs that are majority people of color overwhelmingly voted for Democratic candidates in both years. Notably, consistent with media narratives, Whiter suburban precincts show higher shares for Biden in 2020 than Clinton in 2016. For example, the average marginal effect between the two years for suburban precincts that are 95 percent White was substantial: 3.1 percentage points. In 2020 in our sample, the predictive margins suggest that a majority of votes on average went to the Democratic Party in precincts that were up to 80 percent White, whereas in 2016 the same was true for precincts up to only 55 percent White. As suggested by the postelection narrative, the Democratic Party did in fact make inroads in Whiter suburbs in 2020.
Predictive Margins Generated by Logistic Regression of Democratic Vote Share by Racial Composition and Year with Confidence Intervals
Source: Authors’ tabulation.
Note: Precinct-level Election Data, American Community Survey, n = 163,512 precinct years. Precinct-level election data come from the Florida Voting and Election Science Team, Tufts University Metric Geometry and Gerrymandering Group, and the New York Times. Each panel represents a separate regression including an independent variable for the percentage of the respective race group in the precinct. All regressions control for educational attainment, homeownership, the Democratic vote share of neighboring precincts, logged population, and state fixed effects. We use robust standard errors.
Nonetheless, narratives focusing on White suburbanites often elide the importance of people of color living in suburbs. Even as White suburban spaces shifted toward the Democratic Party, suburbs greater than 80 percent White (the kinds of suburbs in which most White people live) still had vote margins of below 50 percent for Biden in 2020. In contrast, majority-Black suburban precincts overwhelmingly voted for the Democratic presidential candidate in both years and showed greater Democratic turnout in 2020 (this difference was not significant in precincts greater than 85 percent Black). This simply underlines the fact that Black suburban voters remain key for the Democratic party.
Despite media attention to growing Trump support and conservativism among Latinx communities in places such as South Florida and South Texas, predictive margins suggest that majority-Latinx precincts across our sample turned out in greater numbers for Biden than for Clinton. Moreover, the margins are above 50 percent. In 2016, Latinx suburbs came out in higher rates for Clinton than non-Latinx suburbs did; but this trend is not found in 2020 because of increased Democratic support in suburbs with small Latinx populations. Like other groups, Asian American suburbs generally showed increased turnout for Biden relative to Clinton. Suburbs up to 80 percent Asian American showed a significantly higher Democratic turnout in 2020 versus 2016. The larger confidence intervals at the higher end of the distribution point to the small number of suburban precincts that are predominantly Asian American. In 2020, roughly five hundred suburban precincts were greater than 50 percent Asian American, but fourteen thousand between 17 percent and 50 percent Asian. The 2020 trend shows a downward slope—indicating that suburban precincts with very high shares of Asians vote less Democratic—however, this downward trend is nonsignificant, reflecting the relatively small decline in Democratic support as a precinct’s Asian American population increases and the fact that relatively few suburban precincts anywhere in the country are majority–Asian American.
We also explored how these patterns of racial/ethnic suburban voting might differ by state. Figure A.1 provides the predictive margins for regressions with a state by race interaction. We conducted these regressions to see how the overall patterns might change by the state in which suburban precincts are nested. In general, the results suggest the same trends in each state as in the overall regression (figure 7) except for idiosyncratic states or states with small populations of the respective race group living in suburbs (see figure A.3). For example, Hawaii is idiosyncratic for Black and Latinx populations, which likely are disproportionately associated with the military presence in the state. States like New Mexico have very small Asian and Black suburban populations, which is reflected in the wide 95 percent confidence intervals. Some conclusions worth noting from these state analyses are that, first, the negative trend for White composition is consistent across states—that is, Whiter suburbs are less likely to vote for the Democratic candidate. Second, Black voting patterns are consistent: suburban precincts with a higher percentage of Black residents are significantly more Democratic. Third, most of the variation across states is in Asian and Latinx voter trends. In every state, these suburban precincts are more likely to vote Democratic than corresponding White suburban precincts, but in some states as precincts become more Asian or Latinx, the percent Democratic vote plateaus or declines.
DISCUSSION AND CONCLUSION
Since the 1950s, suburbs in the United States have had a distinctive political profile, reflecting their equally distinctive demographics. Residing in racially homogenous, middle-class enclaves, White suburban voters embraced a set of policy positions that perpetuated their racial and class position. Since the 1990s, however, the demographics of suburbs have been changing, with consequent political shifts. In 2020, for instance, suburban voters were more likely to back Biden, the Democratic candidate, than his Republican counterpart Trump. The question we explore in this article is whether the shift in voting in suburbs in recent elections is occurring because of partisan shifts among all suburbs, among White suburbs, or largely because of the suburbanization of people of color?
Our analysis of suburban precinct data across twenty-two states indicates that, as emphasized in some media narratives, White suburban precincts showed greater support for Biden in 2020 than for Clinton in 2016. Our analysis also indicates, however, that if all suburban voters had voted like White suburbanite precincts, Trump would have carried metropolitan suburbs in 2020. These precincts on average voted for Trump in both election cycles. After controlling for critical covariates such as education, homeownership, and the spatially lagged variable for Democratic vote shares, we still see declining Democratic turnout with the increasing share of White residents. The precinct data indicate that the Blackest suburban precincts showed the highest voter turnout for the Democratic Party in both 2016 and 2020, followed by precincts with Latinxs and Asian Americans. All in all, although White suburban voters did indeed shift somewhat toward the Democrats between 2016 and 2020, Democrats would likely not have won a majority of suburban voters in the absence of the increasingly racial and ethnic diversification of suburbia.
This analysis likely underestimates the probability that Democrats would have lost the suburbs in 2020 had they relied on White voters. The sample analyzed here disproportionately covers Democratic-leaning states that have higher concentrations of suburban people of color. As a result, the findings of this analysis are favorably biased in its estimates of Democratic Party voter turnout relative to suburbs across the nation overall. Put differently, because of the availability of data, the analysis presented misses substantial parts of the White suburban electorate in more conservative states across the Midwest, Mountain West, and South. Trump carried the suburban precincts in which most Whites lived, but, given our sample, this finding likely understates the strength of White suburban conservativism.
It is important to understand what these precinct-level results do, and do not, tell us, and not to confuse contextual, precinct-level analyses with an analysis of individual-level voting. One important limitation of this study is that we cannot make claims about individual Black, Latinx, Asian, or White voters. One implication of this limitation is that it could be, for example, that as suburbs diversify ethnically and racially that White voters in diverse suburban precincts are “micro-segregated” (Lichter, Parisi, and Taquino 2017) and remain more conservative than their neighbors. Perhaps equally plausibly, White residents who remain or choose to move to more racially and ethnically diverse suburbs may be sorting themselves ideologically (Tam Cho, Gimpel, and Hui 2013) from White suburban residents who live in White-majority suburbs. More diverse suburbs may be voting more for Democratic candidates because all their residents, regardless of race, are voting more Democratic. Access to geographically coded individual-level voter files may provide a path forward to understand how local contexts shape individual voting behavior in suburbs. For example, future research may investigate whether White people are more Democratic leaning in racially diverse contexts or ask whether Whites become more Democratic as their context becomes more diverse over time. The answers to these questions will provide a richer understanding of the relationship between local racial context and political choice.
Despite these caveats, the overall picture remains clear: even if White suburban precincts shifted modestly toward Biden in 2020 relative to Clinton in 2016, they have not become bastions of liberalism. Democrats carried metropolitan suburbs in 2020 because of suburban voters of color. The 2020 election results suburban underscore the fact that voting cannot be understood without taking suburban residents of color (that is, most people of color) into account, especially as the United States grows increasingly racially and ethnically diverse.
Appendices
Composition of the Suburban Precinct for the Median Individual for the Respective Race Group
Logistic Regression Estimates Predicting Democratic Vote Share
Racial Composition and 2020 Democratic Vote in Greater Atlanta
Source: IPUMS NHGIS (Manson 2021) and New York Times (Park 2021).
Racial Composition and 2020 Democratic Vote in Maryland Suburbs and Baltimore
Source: IPUMS NHGIS (Manson 2021) and New York Times (Park 2021).
Predictive Margins Generated by Logistic Regression of Democratic Vote Share by Racial Composition and Year with Confidence Intervals
Source: Authors’ tabulation.
Note: American Community Survey, n = 120,623 suburban-precinct years. These regressions only focus on suburban precincts, setting aside the urban precincts which were included in the regressions in the main body of the article.
FOOTNOTES
↵1. We calculated these numbers by growth in the metropolitan area less growth in Atlanta city.
↵2. For exceptions, see Herndon 2020; Tavernise and Gebeloff 2019.
↵3. Suburbanization varies greatly across U.S. metro areas and regions and, currently, the literature lacks consensus on what spaces constitute suburbs (Lacy 2016). But generally, the term suburb refers to urbanized spaces in metropolitan areas that are neither the central city nor rural. By central city, we mean the largest city by population within a metropolitan area, after which metropolitan areas are typically named (for example, Philadelphia is the central city of the Philadelphia metropolitan area). We follow the census and consider nonmetropolitan areas to be rural. Thus, in this analysis, we consider suburbs to be parts of metropolitan areas that are not the central city.
↵4. Due to idiosyncratic boundaries, it was common for block-groups to share minor overlap with precincts and cover less than 1 percent of the precinct.
↵5. This is a precinct-level analysis, so we present precinct-level descriptive statistics here (for the composition of the precinct where the median individual of the respective race group lived, see table A.1). The median Asian person lived in a precinct that was 18 percent Asian in 2016 and 19 percent Asian in 2020. These statistics can be interpreted as half of Asian Americans in the sample lived in a precinct that was at least 19 percent Asian in 2020. The median Black person lived in a precinct that was 25 percent Black in 2016 and 24 percent Black in 2020; the precinct for the median Latinx person was 39 percent Latinx in 2016 and 40 percent Latinx in 2020; last, the precinct for the median White person was 81 percent White in 2016 and 79 percent White in 2020.
↵6. We conducted additional regressions examining median household income and median house values as alternative measures of precinct-level socioeconomic status. We separated these variables into different regressions due to issues of multicollinearity. These regressions produced results that were substantively identical to those presented in this analysis.
- © 2023 Russell Sage Foundation. Rastogi, Ankit, and Michael Jones-Correa 2023. “Not Just White Soccer Moms: Voting in Suburbia in the 2016 and 2020 Elections.” RSF: The Russell Sage Foundation Journal of the Social Sciences 9(2): 184–203. DOI: 10.7758/RSF.2023.9.2.08. Direct correspondence to: Ankit Rastogi, at ankitxrastogi{at}gmail.com, Department of Political Science, The Ronald O. Perelman Center for Political Science and Economics, 133 S. 36th Street, Philadelphia, PA 19104-6215, United States; Michael Jones-Correa, at mjcorrea{at}sas.upenn.edu, Department of Political Science, The Ronald O. Perelman Center for Political Science and Economics, 133 S. 36th Street, Philadelphia, PA 19104-6215, United States.
Open Access Policy: RSF: The Russell Sage Foundation Journal of the Social Sciences is an open access journal. This article is published under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.