Daily Diversity Flows: Racial and Ethnic Context Between Home and Work

Abstract

The racial and ethnic diversification of the U.S. population has transformed the demographic makeup of communities and rapidly increased exposure to diversity in American neighborhoods. Although diversity exposure occurs throughout people's daily lives, the conventional approach to describing diversity only at places of residence potentially understates the full extent of this phenomenon. In this study, we explore short-term, within-day changes in the diversity of different neighborhoods by considering U.S. workers’ work and residential locations. Using estimates for daytime and nighttime populations among metropolitan census tracts, our empirical analyses investigate the extent to which the process of daytime mobility for work relates to changes in the racial and ethnic diversity of different spaces. Our results indicate widespread daily shifts toward diversity for most neighborhood types, especially those with residential (nighttime) populations that are predominantly Black, Latino, or Asian. We find that patterns of intraday diversification experienced minor declines across recent decades but are present in most metropolitan areas. Our findings also show that intraday changes in racial and ethnic diversity overlap with nonracial forms of daily diversity change. Further, average within-day changes in diversity are more pronounced in areas with greater residential segregation.

Introduction

Many studies have documented how U.S. neighborhoods have undergone rapid racial and ethnic diversification over recent decades (Ellis et al. 2018; Farrell and Lee 2011; Fowler et al. 2016; Frey 2018; Lee and Sharp 2017; Logan and Zhang 2010; Tach et al. 2019). This demographic research portends a trend toward racial heterogeneity in residential communities driven by rapid immigration to the United States from Latin America and Asia, the suburbanization of non-Whites, new regional patterns of internal migration, and growth in multiethnic populations (Logan and Stults 2022; Logan and Zhang 2010; Sáenz and Morales 2015; Timberlake et al. 2011; Winkler and Johnson 2016; Zhang and Logan 2016). These trends toward diversity are not limited to residential neighborhoods, though, and plausibly extend to the wide range of social spaces (e.g., places of work) that individuals occupy (Browning, Calder, Krivo et al. 2017; Browning and Soller 2014; Cagney et al. 2020; Jones and Pebley 2014). And although the nearly sole focus on residential spaces in sociological and demographic research is understandable, given their connections to environmental exposures, community resources, and public goods, important questions remain about the extent to which individuals’ exposures to diversity change through daily activities, such as work (Sampson et al. 2002; Wodtke et al. 2011).

As individuals move through different spaces as part of their daily activity patterns, they interact with people beyond their residential neighborhoods (Browning et al. 2015; Kelling et al. 2021; Newmeyer et al. 2022). Still, these mobility dynamics are not uniform across the local social contexts of regions such as metropolitan areas. Instead, they interact with patterns of spatial stratification by race, ethnicity, and socioeconomic status. Some individuals’ mobility occurs only between relatively similar types of contexts (e.g., one segregated space to another), creating downstream implications for inequalities in contextual exposures salient to health and social behaviors (Brazil 2022; Graif et al. 2017; Levy et al. 2020; Vachuska 2023).

Research suggests that as individuals go about their daily lives, their many mobility trajectories bubble up into changing neighborhood populations at the meso level. In other words, neighborhoods themselves experience meaningful changes throughout the day, resulting in reductions in aggregate patterns of racial and ethnic segregation during the day compared with at night (Ellis et al. 2004; Hall et al. 2019; Müürisepp et al. 2023; Vallée 2017). Sometimes conceptualizing daytime compositions for a particular local area as a “workhood,” this research raises questions about how much the mobility documented in studies of individual-level commuting and daily activity patterns diversifies some neighborhoods during the day. If substantial meso-level changes in diversity and segregation are observed between day and night, then accounting for how neighborhoods themselves change amid individuals’ daily lives might be essential for research aimed at understanding the wide-ranging consequences of individual-level mobility documented across domains of health, well-being, and social interactions (Browning, Calder, Boettner et al. 2017; Browning, Calder, Soller et al. 2017). Applying a neighborhood focus to research on daily activities can thus inform how widespread and substantial any within-day changes in contextual composition are for broader research based on individual data.

Lower levels of aggregate workplace segregation suggest that daytime populations tend to be more integrated by race and ethnicity than residential populations at night, but other neighborhood and metropolitan characteristics might structure differences between neighborhoods in the extent to which intraday mobility brings diversification. For example, although neighborhoods’ spatial positions are likely important for structuring the potential for daily diversification change, racial and ethnic groups’ relative sizes, neighborhoods’ nighttime compositions, and broader characteristics of metropolitan areas are also potential factors for why some neighborhoods may become more diverse through the day whereas others may remain quite segregated. Differences between metropolitan areas in terms of suburbanization and neighborhood diversity (i.e., among residential compositions) likely shape daily mobility patterns’ tendency to generate a more balanced representation of racial and ethnic groups rather than reify social separation. Similarly, metropolitan contexts dominated by residential segregation at night might be associated with more diversification during the day because segregation reduces starting levels of neighborhood diversity throughout the region.

Finally, growing racial and ethnic diversity during the day creates dependencies in diversification across nonracial forms of heterogeneity, too (Hall et al. 2016; Tach et al. 2019). Thus, these mobility dynamics are also potentially associated with important changes to the social and economic diversity experienced within an area between day and night. A dynamic of daily diversification therefore plausibly relates to within-day changes along other social characteristics, with more integrated workhoods increasing the potential for intergroup contact in places otherwise characterized by residential (nighttime) isolation.

In this research, we investigate the following questions about neighborhood racial and ethnic diversity:

1. To what extent do neighborhoods of different racial and ethnic arrangements become more or less diverse during the day?

2. How do patterns of daytime diversification vary between different metropolitan areas?

3. How much do intraday changes in racial and ethnic diversity overlap with intraday changes to the social and economic heterogeneity in a given area?

Background

U.S. neighborhoods have long been characterized by residential segregation rather than diversity. Throughout the twentieth century, racial and ethnic segregation became entrenched throughout urbanized areas, with redlining, racially restrictive covenants, and steering by real estate actors contributing to heightened residential separation between racial and ethnic groups (Charles 2003; Faber 2020; Iceland 2004b; Rothstein 2017).

However, the United States has witnessed extraordinary diversification over the last several decades, reflected at a high level in population compositions moving toward a majority-minority status (Frey 2018). Further, given that metropolitan regions are where population growth is fastest and immigration tends to be greater, the diversification of the United States can be seen even more so in the growing number of metropolitan regions with significant Asian and Latino representation (Fong and Shibuya 2005). Accordingly, there is strong consensus that the country has undergone a so-called “diversity explosion” over recent decades, transforming not only racial and ethnic compositions but also age structures and rates of natural increase (Frey 2018).

Despite consensus about increased diversity throughout the U.S. population, the implications of this trend for long-standing racialized residential inequalities has been an area of scholarly debate. The growth in diversity has corresponded with steady declines in residential segregation, particularly between Black and White persons, despite residential separation continuing to structure inequality in many metropolitan areas (Elbers 2021; Hess et al. 2019; Logan and Stults 2022; Logan et al. 2004). Population diversification has also parallelled the rapid growth of multiethnic neighborhoods, a substantial decline of entirely White neighborhoods, and growing exposure of White individuals to non-White populations (Ellen 1998; Farrell and Lee 2011; Holloway et al. 2012; Lichter et al. 2017; Logan and Zhang 2020; Zhang and Logan 2016). Despite these trends, evidence on the link between racial diversity and racial segregation is somewhat mixed, with research finding that community diversification does not always translate to more residential integration (e.g., see Ellis et al. 2018; Iceland 2004b). Moreover, diversity in an area can be a liminal and potentially fleeting step amid processes of racial turnover, with White households moving out to further-flung contexts as non-White populations progressively increase in representation (Kye and Halpern-Manners 2023).

Metropolitan differences in the magnitude of intraday population change could be related to several theoretically relevant drivers. First, as populations become more residentially segregated, low-diversity neighborhoods grow in prevalence. In turn, the extent to which workplaces tend to have greater diversity (relative to the level typical among neighborhoods) will structure whether daily activity patterns create diverse social contexts while individuals are at work. Notably, the expected tendency for segregated metropolitan regions to diversify more because of daily mobility can exist while some contexts remain low in diversity and isolated at all times of the day.

In terms of spatial differences, diversification has been common among urban neighborhoods, stemming largely from the historical role of cities as gateways for immigrants (Bader and Warkentein 2016; Fowler et al. 2016; Frey 2018). Although cities have historically been important for the geography of neighborhood diversity in the United States, their contribution has declined in recent decades owing to the broadening of suburbanization across race and ethnicity and the resulting emergence of highly diverse suburban contexts (Fowler et al. 2016; Hall and Lee 2010; Timberlake et al. 2011). Still, some studies have observed mobility patterns among suburban White households that suggest they are likely to move out of diversifying neighborhoods, making segregation of White populations from Black, Latino, and Asian populations between city, suburb, and “fringe” areas an increasingly salient part of residential segregation in U.S. metropolitan areas (Hess 2021; Lichter et al. 2015; Parisi et al. 2019). Overall, these studies lead to an expectation of important variations between urban and suburban areas in their typical degree of diversity and potential for intraday diversity changes due to daytime mobility.

This extensive research on growing diversity among U.S. neighborhoods illustrates that census tract data have been essential for opening up avenues to studying the growth of diverse neighborhoods. Nonetheless, census tracts can limit or diverge from how we would conceptualize and measure neighborhoods in an ideal world. For example, the modifiable areal unit problem refers to substantive differences in analyses of meso-level units, such as neighborhoods, caused by aggregating from one level to another or simply changing how the boundaries divide space (Lee et al. 2008; Wong 2009). Moreover, individuals’ conceptualization of their neighborhood regularly deviates from tract and other administrative unit boundaries (Pinchak et al. 2021).

More recently, studies of micro-level mobility using novel, spatially granular data sources, such as cellphone records, have forced scholars to reckon with the idea that individuals’ daily lives are much more dynamic than a single survey snapshot—not just in terms of long-term periods of years or decades but also in terms of time of day (Cagney et al. 2020). The implications of thinking small temporally have typically been understood from an individual level: if data could track individuals’ mobility throughout the day, it would be clear that many people do not stay in their census tract of residence all day. Instead, individuals’ daily mobility for work, school, and other engagements—including any segregation among these patterns—can be described through the concept of an “activity space” (Browning et al. 2017; Graif et al. 2014; Jones and Pebley 2014; Wong and Shaw 2011). Although activity spaces are naturally conceptualized egocentrically, the implications of the daily mobility underpinning activity spaces potentially extend to the context of census tract data. That is, aggregate data describing a given area, such as a neighborhood, at different points during the day could reveal important differences in how segregated, integrated, or diverse the population is from how individuals’ myriad daily mobility patterns intersect in the space. In studying neighborhood diversity, the activity space perspective thus leads to open questions about whether and how much different neighborhood contexts undergo diversity changes throughout the day.

We conceptualize the complementary aggregate form of these mobility dynamics through intraday diversity change. This concept is important to demographers and social scientists because daytime mobility might bring people from disparate, otherwise-segregated areas under the cosmopolitan canopy (Anderson 2004, 2011). This mobility could bridge across segregated neighborhoods if cross-cutting interactions occur among more diverse workhoods. At the same time, the far-flung spatial locations of some persons might tilt their daily social environments toward less diversity and more homogeneity in the individuals they encounter at home and work.

Our study contributes to the literature on daily activity patterns as well as neighborhood racial and ethnic diversity in a few key ways. First, despite many studies of how individuals traverse different locations during the day (Brazil 2022; Browning, Calder, Krivo et al. 2017), studies of how neighborhood racial and ethnic diversity changes throughout the day are quite limited, and our study stands to be the first national investigation of how the racial and ethnic diversity of different neighborhood areas changes between day and night. To the extent that daily mobility substantially reshapes the population present at different points of the day—creating new opportunities for social interactions across groups but also potential barriers to coordination among residents—intraday population changes might also matter for outcomes such as neighborhoods’ level of social organization or residents’ perceptions of neighborhood conditions.

Second, we investigate metropolitan differences in the extent to which daily mobility for work is associated with changes in the typical diversity of neighborhoods. Beyond our expectation that average levels of neighborhood diversity are strongly associated with average levels of workhood diversity, we also test our expectation that greater multiethnic segregation among residential populations structures greater diversification into metropolitan areas’ daily mobility flows. By linking aggregate segregation to meaningful differences in within-day population dynamics, we highlight how intraday changes in meso-level population dynamics are potentially salient to our understanding of structural inequalities and merit further exploration.

Finally, this research also contributes evidence about how the use of nighttime residential compositions to define the activity spaces that individuals navigate during the day underestimates the extent to which many areas with non-White compositions at night are diverse spaces during the day. Importantly, we demonstrate how changes to the racial and ethnic diversity of a space correspond to changes in the social and economic profile of a neighborhood, such that the intraday change dynamics we observe have a bearing on other neighborhood measures used in activity space research beyond studies of racial diversity. Our perspective is that measuring activity spaces with consideration for time-varying neighborhood compositions can enhance our understanding of the contexts that individuals experience and how these social environments matter for their lives.

Methodology

Data and Sample

Our analyses rely on the Census Transportation Planning Package (CTPP) for three periods (2000, 2006–2010, and 2012–2016) based on special tabulations of the decennial census and the American Community Survey (ACS). These data are publicly available small-area estimates about workers aged 16 or older who reside or work in a given unit, such as a census tract. We use the census tract–level estimates from the CTPP because they are the most spatially granular tabulations available across time. We harmonize the 2000 estimates to 2010 geography with the Longitudinal Tract Database to ensure that these data are as comparable as possible to the census tract boundaries used in the latter two ACS periods (Logan et al. 2014).

We define metropolitan areas using core-based statistical area (CBSA) delineations from 2010. To identify samples of tracts for each metropolitan area, we first spatially intersect tract geodata with CBSA boundaries. We then filter our tract data to include the subset of CBSAs with at least 50,000 persons and 1,000 persons of each racial and ethnic group in 2000, yielding a sample of 181 consistently defined metropolitan areas.¹

Measures

Our study measures tracts’ population composition and diversity at two stylized points during the day.² First, we use counts for workers aged 16 or older who reside in a given census tract to measure and categorize areas’ racial and ethnic composition at night. Second, we use the CTPP estimates for those whose workplace is a given tract to describe workhood or daytime compositions (Hall et al. 2019).³ These neighborhood and workhood composition variables form the basis for categorizing areas by their racial and ethnic representation at different times of the day. The CTPP data provide only a subset of racial and ethnic categories available in traditional census or ACS estimates, so by default, our daytime and nighttime composition estimates capture shares of workers who are Latino, non-Latino Asian (henceforth, Asian), non-Latino Black (Black), non-Latino White (White), and non-Latino other race (other).

We use a modified version of the categories from Farrell and Lee (2011) to delineate the racial and ethnic composition of neighborhoods and transitions between different levels during the day. We use a group share of 75% or greater to denote a predominant composition, a group composition of 50% to 75% to denote a mixed composition, and a tract with no majorities to denote a multiethnic composition. This framework reflects our primary interest in differentiating the least diverse places (predominant) from those with some degree of diversity (mixed) and those with the most diversity (multiethnic).

To measure the relative diversity of different neighborhoods’ racial and ethnic compositions, we compute multiethnic entropy scores for each tract at each time point using counts for persons who are Black, Latino, Asian, other race, and White (Iceland 2004a). In the following equation, i denotes a given census tract, r indexes the five racial and ethnic groups, and $Πri$ is a given r group's share of the ith tract's population:

Because five groups are included in the calculation, these entropy score measures have a maximum of log 5 (roughly 1.6) when each of the five racial and ethnic groups is represented in equal share. In contrast, these entropy scores take a minimum value of 0 when one group composes a given tract's entire population.

We also exhaust the relevant data on tract social and economic composition from the 2012–2016 ACS wave of the CTPP to measure the diversity of tracts’ daytime and nighttime populations in terms of age (16–24, 25–44, 45–64, or 65 or older), sex (male or female), language (English, Spanish, or other), and income (below 100% of the federal poverty line [FPL], 100% to 149% of the FPL, or 150% or more of the FPL) diversity. These measures are used to understand how racial and ethnic and nonracial dimensions of intraday change are associated, so we create intraday change measures by subtracting each tract's nighttime value from the daytime ones. The calculation of these nonracial entropy scores is similar to the previous formula, albeit with different numbers of groups depending on the particular form of social or economic diversity.

Finally, we use segregation measures in our regression analyses to understand how varying degrees of residential segregation correspond with differences in neighborhoods’ daytime diversification. Specifically, we use the multiethnic information theory index (⁠$H$⁠) to assess the extent to which places with greater racial and ethnic segregation differ in their trajectories of intraday diversity change (Reardon and Firebaugh 2002). The following formula is used to calculate H for a given metropolitan area, where E is the metropolitan entropy score, $T$ is the total metropolitan population, and $Ei$ is the entropy score for a given tract i in the metropolitan area:

We use multiethnic $H$ because it captures deviations from an expected level of racial and ethnic diversity and thus aligns well with our other analyses of diversity based on entropy scores. However, results using alternative measures, such as a non-White isolation index, are substantively similar (see section A, online appendix).

Analytic Framework

We draw on the following descriptive analyses to address our three research questions about neighborhood diversification between night and day. First, we outline the mean change in racial and ethnic groups’ tract population proportion between day and night using predictions from simple linear regression models predicting a group's composition change from nighttime to daytime as a function of period-specific associations with different nighttime composition categories. This analysis gives us traction to understand how average trajectories of intraday diversification vary between neighborhoods with different predominant, mixed, and diverse residential compositions at night. By estimating the association of particular nighttime neighborhood types (e.g., multiethnic) with intraday night-to-day change outcomes (e.g., change in the Black population share) on a period-specific basis, we can further assess how intraday diversification patterns have trended since the 2000 census.

Next, we assess how the different magnitudes of composition change manifest in neighborhood differences in the probability of transitioning between neighborhood types between day and night. To understand these patterns of intraday change, we use transition matrices that provide insight into the prevalence of daytime racial and ethnic categories among tracts with different nighttime racial and ethnic categories (i.e., predominant, mixed, or multiethnic). We first present these results using our overall sample. Then, to understand spatial differences in these dynamics, we also disaggregate this analysis by city and suburb location.⁴

Third, we analyze how changes in within-day racial and ethnic composition in different areas also imply changes in their multiethnic diversity. Here, we summarize the mean difference in Shannon entropy scores for multiethnic representation among daytime and nighttime populations on the basis of an area's nighttime neighborhood composition type. Because we are also interested in whether intraday diversity change trended up or down over the span of our data, we present these mean entropy score changes for each nighttime composition at each period of our data.

Fourth, we detail how the largest 50 metropolitan areas by total population (according to the 2012–2016 ACS) differ in their typical neighborhood diversity during the day and at night. Here, we use average neighborhood and workhood entropy scores among these metropolitan areas to understand which metropolitan areas experience relatively larger shifts in their average tract diversity level. These metropolitan differences highlight that a substantial average level of intraday change is prevalent in many U.S. metropolitan regions.

Fifth, to formalize observations from our prior descriptive analysis among metropolitan areas, we use a set of fixed-effect regression models explaining average daytime diversity (i.e., entropy score) for our full set of 181 metropolitan areas. These models predict average daytime diversity for a metropolitan area conditional on period, average nighttime diversity, the extent of residential segregation in the metropolitan area, and time-varying metropolitan population composition covariates. Because models are based on repeated observations of metropolitan areas over three periods, we cluster all standard errors according to metropolitan area.

For our final set of results, we investigate how the magnitude of change that a neighborhood experiences in terms of its racial and ethnic diversity correlates with differences in terms of important changes to other nonracial forms of population diversity. These regression models are essential for documenting that neighborhoods’ diversity changes in multiple, overlapping ways throughout the day. Specifically, we assess how different changes to an area's social and economic diversity predict greater differences in racial and ethnic diversity changes. For all these models, we adjust for metropolitan context through fixed effects for each CBSA. We also adjust for tract differences in tract population and area (in square miles). These models cluster standard errors according to metropolitan area.

Following our main analyses, we discuss several supplementary analyses, available in our online appendix, that we conducted as robustness checks. These alternative specifications include weighting our results by population size at different times of day to rule out the role of total population changes as an explanation for our findings, analyzing neighborhood data to ensure that our aggregated metropolitan-level analysis does not obfuscate important variations in segregation's salience to intraday diversity changes, and omitting very sparsely populated tracts to ensure they do not account for our conclusions from our main results.

Results

Average Intraday Change in Racial and Ethnic Composition

To address our first research question, we start by reviewing mean intraday change for different racial and ethnic groups conditional on tract nighttime racial composition type. This analysis provides insight into the typical magnitudes of difference between tracts’ nighttime (residential) and daytime (workplace) racial and ethnic compositions. We generate these estimates using ordinary least-squares models regressing change scores for a given group on the interaction of nighttime categories and period dummy variables.⁵Figure 1 shows the marginal predicted values for intraday change among neighborhoods with different predominant, mixed, or multiethnic compositions at night. Estimates for intraday change in 2000, 2006–2010, and 2012–2016 are indicated by light gray, dark gray, and black bars, respectively.

The first panel in Figure 1 shows the relative absence of intraday composition change for tracts that are predominantly White in their nighttime residential composition. At no point across our data did predominantly White neighborhoods experience even half the relative magnitude of intraday change in their White share that the other predominant neighborhoods did for their majority group share. Most of the intraday change in the White share for predominantly White neighborhoods is accounted for by modest increases in the Black and Latino shares. However, the results in Figure 1 highlight that these neighborhoods are unique among predominant neighborhood types in having minimal average changes in their racial and ethnic composition related to commuting patterns.

The typical predominantly Black neighborhood shows a substantial decrease in the Black population (>25%) that is mostly accounted for by a comparable increase in the White population (the second panel in the top row). Although these magnitudes have declined somewhat since 2000, the trends over time have not diminished the fact that the average predominantly Black tract (at night) has a magnitude of intraday change that implies a decline to an altogether different category (e.g., non-White–mixed or multiethnic).

The third and fourth panels in the top row of Figure 1 depict similar dynamics for tracts with nighttime compositions that are predominantly Latino and predominantly Asian. Both of these nighttime composition types show that declines of 20% or more for the predominant group are typical, even during the most recent period. White population share increases account for most of these changes. One distinct difference among the three types of tracts with predominantly non-White compositions at night is that predominantly Asian tracts have considerably greater increases in their Black and Latino population shares during the daytime.

The bottom-left panel for White–mixed tracts shows that this neighborhood type has intraday dynamics relatively similar to those observed for predominantly White tracts. None of the changes exceed 10%, with essentially no change in these intraday dynamics over time. In contrast, the panels for non-White–mixed and multiethnic (the middle and right panels in the bottom row) suggest a dynamic somewhere in between the tendencies for substantial change in tracts with predominantly non-White populations and lack of change in tracts with majority-White populations. Tracts with nighttime non-White–mixed and multiethnic compositions show intraday changes in which Black and Latino population shares decline and White shares increase. The average magnitudes of change outstripping 10% indicate some potential for a neighborhood of either type to transition to another type during the daytime because of its worker commuting patterns.

Intraday Transitions in Racial and Ethnic Composition

We further investigate our first research question by using neighborhood transition matrices defined on a daytime/nighttime basis to examine how much the different magnitudes of intraday change we observe correspond to tracts having different composition types between day and night. We present these results for the 2000 census and 2012–2016 ACS to decipher how the changes in intraday dynamics over time documented in our prior analyses led to a shifting of expected daytime/nighttime transitions.

Figure 2 shows the intraday transitions for census tracts among the 181 metropolitan areas under study. Starting with the left-hand matrix for 2000, a few distinct dynamics stand out. First, the relative paucity of intraday changes in predominantly White and White–mixed neighborhoods translates into the very low prevalence of intraday neighborhood type transitions, and more than 80% of predominantly White neighborhoods remained this type day and night in this period. In contrast, predominantly Black, Latino, and Asian neighborhoods were much more likely to transition toward a more integrated non-White–mixed, if not a multiethnic, composition. The share of these three neighborhood types remaining the same type day and night during 2000 ranged from only 6% to 19%. Rather than remain the same during the day, these types of contexts experienced substantial shifts in their racial and ethnic diversity through intraday patterns, and at least one fifth of these tracts even became multiethnic during the daytime.

Among the two mixed composition categories, roughly two thirds of White–mixed neighborhoods remained this type throughout the day in the 2000 period, with a transition to a predominantly White neighborhood the next-most common transition (accounting for roughly one quarter of all White–mixed neighborhoods). The most common trajectory for tracts with non-White–mixed composition types at night was becoming multiethnic through intraday mobility (about 40%)—a trajectory that was rare among White–mixed tracts. The next most common transitions for non-White–mixed tracts were toward a White–mixed or non-White–mixed daytime composition, accounting for approximately 30% and 20% of these tracts, respectively. Finally, neighborhoods that were multiethnic in their nighttime composition were most likely to become White–mixed or stay multiethnic (46% and 42%, respectively).

The right-hand matrix in Figure 2 shows that the relatively strong trajectories of predominantly White tracts remaining the same type all day were somewhat diminished by the 2012–2016 ACS, even though being predominantly White both day and night remains the majority trajectory among tracts with this type. Among predominantly Black, Latino, and Asian neighborhoods, another notable trend in increased transition prevalence is that each type became considerably more likely to retain its predominant population composition day and night. Although this trajectory is still much less common than observed among predominantly White areas, this trend nonetheless appears to be one of the most notable in Figure 2 to show declining diversification through intraday mobility. White–mixed tracts were slightly more likely to become multiethnic during the day, whereas non-White–mixed tracts remained most likely to become multiethnic through intraday mobility. Finally, multiethnic neighborhoods were slightly more likely to stay multiethnic day and night.

Because some of these differences relate to variations in residential composition between urban and suburban areas, Figure 3 presents separate neighborhood transition matrices for urban neighborhoods (i.e., those with centroids within a principal city) and suburban neighborhoods (i.e., those with centroids outside a principal city). Many of the observations about intraday change noted in the overall analysis hold under this relatively simple spatial disaggregation. For example, predominantly White neighborhoods are most likely to have this composition category all day, regardless of whether they are urban or suburban. Similarly, substantial diversification is likely for predominantly Black, Latino, and Asian neighborhoods regardless of their spatial position within a metropolitan area. Finally, multiethnic neighborhoods, whether located in urban or suburban areas, have remarkably stable racial and ethnic diversity: remaining the same type all day is the most likely trajectory for urban and suburban multiethnic tracts by the latest period.

The transition matrices for the urban and suburban subsamples are similar, with a few notable differences. First, in the suburbs, greater shares of predominantly White and White–mixed neighborhoods maintain this composition all day. In contrast, urban neighborhoods are more likely to transition from predominant to mixed and mixed to multiethnic (even though less intraday change is still the most common trajectory). Another important difference is that predominantly Latino and Asian tracts in urban areas are less likely to keep their predominant composition all day relative to comparable suburban tracts, suggesting that some of these suburban contexts are relatively more isolated. Finally, suburban multiethnic tracts are more likely than their urban counterparts to become White–mixed during the day as a result of workers’ commuting patterns—so much so that only 1 percentage point separates the modal and second-most-common trajectories (i.e., remain multiethnic, transition to White–mixed).

Intraday Change in Racial and Ethnic Diversity

We wrap up the evidence for our first research question with analyses directly assessing how the sometimes-substantial observed changes in composition translate into considerable differences in the extent to which areas become more diverse during the day. The bars in Figure 4 show mean changes in multiethnic entropy scores between day and night for each nighttime neighborhood composition type across data periods. Estimates for intraday change in 2000 are visualized with light gray bars; 2006–2010, with dark gray bars; and 2012–2016, with black bars. Regarding the predominant neighborhood composition types, predominantly Black, predominantly Latino, and predominantly Asian neighborhoods all diversify significantly more during the day than the typical predominantly White area: average levels of change for these areas are more than double those observed for predominantly White tracts across all periods. Nevertheless, we observe a clear trend across time for all four types, such that daytime mobility for work corresponds with less diversification in relatively racially isolated neighborhoods (in terms of their nighttime compositions) at the most recent observation.

Among the other mixed and multiethnic neighborhood types shown in Figure 4, we see countervailing patterns in which White–mixed or multiethnic neighborhoods experience modest diversity declines during the day, whereas non-White–mixed neighborhoods experience a comparable magnitude of diversity increase. Over time, these three neighborhood types experienced relatively modest declines, particularly relative to the larger magnitude of change observed among the three predominantly non-White types. Overall, the results related to our first research question highlight a prominent trend of daily mobility for work bringing diversity to spaces where people spend time during the day. The primary exceptions are spaces where diversity is already high at night (i.e., multiethnic) or where Whites are the substantial majority (i.e., predominantly White).

Metropolitan Differences in Daily Diversification

Our next set of results focuses on our second research question about the extent to which metropolitan areas have systematic differences in these dynamics. First, Figure 5 presents mean multiethnic entropy scores based on daytime and nighttime populations for the largest 50 U.S. metropolitan areas (using the 2012–2016 ACS). Metropolitan areas are listed in descending order of average daytime diversity, with average daytime diversity denoted with white circles and average nighttime diversity denoted with black squares.

We note a few important observations from Figure 5 regarding metropolitan differences in the typical racial and ethnic diversity of workhoods and in the extent to which this is markedly different from typical neighborhood diversity. First, typical daytime and nighttime diversity levels are strongly related: these measures have a correlation of .95 within these 50 largest metropolitan areas and .94 among our complete set of metropolitan areas. Despite this apparent relationship between nighttime and daytime diversity, some of the most notable exceptions are among residentially segregated metropolitan areas (e.g., Chicago, Detroit, and Memphis), which show large increases in diversity during the day. These and other metropolitan contexts that typically rank among the most racially segregated areas experience some of the largest diversity shifts between day and night, often leading to substantial changes in their rank between day and night. Finally, although the metropolitan areas in Figure 5 have a general trend of daytime mobility leading to greater typical tract diversity while people are at work, a handful of metropolitan areas (e.g., Seattle, Minneapolis, and Portland) have somewhat lower levels of racial and ethnic diversity during the day than at night.

We now formalize observations from Figure 5 through fixed-effects models exploring the associations of metropolitan characteristics with average workhood diversity. These models, described in Table 1, are based on our full sample of 181 metropolitan areas; they incorporate unit fixed effects to adjust for time-invariant metropolitan characteristics. The focal associations across our nested models are then described by the coefficients for period dummy variables to capture change over time, average neighborhood diversity, and metropolitan residential segregation.

Starting with Model 1, the coefficients for the two period terms show a trend since 2000 of workhoods being more racially and ethnically diverse, on average. Model 2 adjusts for average neighborhood (i.e., nighttime) diversity, indicating how the significant increases in typical workhood diversity since 2000 are largely accounted for by increases in the typical level of racial and ethnic diversity among tracts’ nighttime populations. Although most variation in average workhood diversity is explained by the two-way fixed effects and average neighborhood diversity levels, Model 3 nonetheless shows that metropolitan areas with greater multiethnic racial and ethnic segregation have greater average workhood diversity. Finally, Model 4 shows that average neighborhood diversity and residential segregation continue to have significant positive associations with average workhood diversity even after we adjust for time-varying characteristics of metropolitan areas relevant to intraday mobility. Other covariates in Model 4 provide evidence that the total metropolitan population, the level of Latino representation, and the share of foreign-born persons are also significantly associated with average levels of workhood diversity observed.

Dependencies Between Racial and Ethnic Intraday Change and Other Diversity Forms

Following observations from prior studies that racial and nonracial diversity are systematically linked (Tach et al. 2019), our final results focus on our third research question investigating how intraday racial and ethnic diversity is associated with social and economic diversity changes. Specifically, we use the 2012–2016 ACS and regress measures of each tract's intraday change in racial and ethnic diversity or composition on its respective changes to its age, sex, language, and income diversity to understand how much a neighborhood's intraday changes in social and economic heterogeneity correlate with the expected magnitude of racial and ethnic diversity change. We standardize all diversity measures in this analysis before computing change scores. Table 2 provides regression coefficients and standard errors for these models of intraday change, with each column corresponding to a type of change to a neighborhood's racial and ethnic diversity or composition.

The leftmost column in Table 2 provides descriptive associations between types of intraday change to social and economic diversity and intraday changes to a neighborhood's racial and ethnic diversity. First and foremost, changes in racial and ethnic diversity are significantly and positively associated with changes in age, language, and income. These associations imply overlap between intraday racial and ethnic diversity change and the neighborhoods where work mobility increases the representation of different age groups. Importantly, the associations captured by the first model in Table 1 align with expectations from prior work on long-run changes in racial and ethnic diversity, which found that the largest associations for intercensal changes to racial and ethnic diversity were with changes to a neighborhood's language diversity first and foremost, followed by income and then age (Tach et al. 2019).

The remaining models in Table 2 provide insight into how group population shares differ in the growth or decline between day and night depending on the other types of population diversity change observed. Reflecting racial and ethnic group differences in age structure, neighborhoods with greater intraday change in age diversity show greater increases in the White population shares amid declines in the Black and Latino shares. Reflecting differences in non-English-speaking among different racial and ethnic groups, neighborhoods with a greater intraday change in language diversity have greater declines in their White and Black shares between night and day amid greater increases in their Latino and Asian shares. Finally, reflecting racial and ethnic inequalities in socioeconomic status, neighborhoods with a greater intraday change in income diversity are associated with greater declines in their White shares and greater increases in their Black and Latino shares. Overall, these models show how neighborhoods’ sometimes-large intraday changes are associated with other important shifts in their social and economic profiles, further underscoring that the neighborhood contexts social scientists frequently study are not static throughout the day but rather are dynamic across several associated dimensions.

Robustness Checks

Our main findings highlight the dynamic nature of diversity in social contexts between day and night. To ensure the robustness of our conclusions, we conducted supplementary analyses assessing how differences between areas’ nighttime and daytime population sizes might influence our results. A motivation for these checks is that bedroom communities with large nighttime residential populations might become sparsely populated during the day, whereas central business district areas might have few residents in the evening but be magnets for many individuals during the daytime for work. Some of these robustness analyses also use multilevel modeling to confirm that our metro-level analysis is robust to an alternative approach using tract data.

First, we investigate whether weighting our results to give emphasis to tracts with larger daytime or nighttime populations led to substantively different conclusions compared with our primary unweighted results. The results (presented in the online appendix, section C) remain consistent across these two alternative specifications, with two exceptions in our consideration of dependencies between different forms of intraday diversity change. First, the association between age diversity change and racial and ethnic diversity change shifts from positive to negative when we weight by the daytime population but remains negative when we weight by the nighttime population. Focusing on areas with large daytime populations emphasizes dense urban business districts, where residential populations tend to be younger, leading to a somewhat different association in which rising racial and ethnic diversity comes through declining age diversity. Second, we find that predominantly Asian areas (in terms of their nighttime population) have relatively larger differences in average change across the weighting specifications because these areas are relatively sparse and tend to be proximate to central business districts with the largest daytime populations.

Second, we use multilevel models to ensure that our focal associations between metropolitan residential segregation and average patterns of diversity change remain similar when we use neighborhood-level data. These multilevel models adjust for each neighborhood's population size, nighttime diversity, and other metropolitan area characteristics. We find that our focal association between residential segregation and diversity change remains positive and significant whether we use a multiethnic information theory or non-White isolation to measure residential segregation for each metropolitan area (see the online appendix, section D).

Finally, to ensure that our focal results are not driven by sparsely populated tracts, we estimate additional multilevel models that omit tracts with daytime or nighttime populations of fewer than 1,000 persons across these models. Again, we find that our focal associations between metropolitan segregation and diversity change are positive and significant (see the online appendix, section E), suggesting that these sparsely populated regions do not exert an undue influence on our main findings.

Discussion

By comparing the populations of neighborhoods and workhoods among U.S. metropolitan areas, this study provides novel insights into how the racial and ethnic diversity of metropolitan spaces changes substantially throughout the day. We find that although virtually all neighborhoods undergo intraday diversification, their magnitudes of change differ by underlying racial structures. Majority-White neighborhoods, in particular, differ substantially from neighborhoods with a non-White majority or multiethnic composition. Whether considered through the lens of changing population shares, neighborhood composition categories, or entropy scores, our results illustrate that predominantly Black, Latino, and Asian neighborhoods are the primary sites of daily diversification, whereas White neighborhoods remain stably White throughout the day.

Our analysis also explores the extent to which intraday diversification differs between metropolitan regions. Most importantly, we find that although residential segregation by race and ethnicity has long been understood to structure place-based inequalities, greater segregation is associated with higher levels of contextual diversity during the day. Knowing that daily diversity tends to be concentrated in the most racially isolated non-White neighborhoods, this somewhat paradoxical connection between segregation and diversification makes sense: it is the daily mobility patterns into Black, Latino, and Asian neighborhoods that shape daytime contextual diversity the most. Importantly, our observation regarding the linkage between metropolitan segregation and workhood diversity need not be the case, with an alternative expectation being that segregated regions would have relatively less change in diversity because people move from one segregated context to another.

We also find that these patterns of intraday racial change are associated with changes in neighborhood social and economic contexts. Changing representations of persons with different ages, primary languages, and socioeconomic status are directly related to the magnitude of racial and ethnic change observed, suggesting that daily mobility reshapes the social milieu that individuals experience in ways that increase the likelihood of contact not only across racial and ethnic groups but also between persons of different generational cohorts, language backgrounds, and socioeconomic advantage or disadvantage.

Given the far-reaching changes that social contexts experience throughout the day, the implications of this work for future studies accordingly intersect with debates in sociological and demographic literatures. First, this study connects to individual-level research on daily activity patterns by motivating questions about how much difference we might find in a given person's environmental exposures when considering both individual and social contexts to be dynamic throughout the day. The study of neighborhood composition as dynamic can be challenging with conventional data, such as longitudinal surveys. However, novel sources—such as consumer, social media, and GPS trace data—might yield new approaches for measuring the composition of local temporary populations individuals experience based on the time of day they visit an area (Panczak et al. 2020). Mobility research has provided new theories and approaches for studying how individuals’ lives are not confined to where they reside, and the results of our study motivate similar innovations for capturing short-run changes in the social contexts that individuals traverse.

Next, the substantial changes in some neighborhoods potentially affect the behavior and interactions among residents and the individuals whose mobility intersects with the area. To the extent that substantial within-day changes lead to more difficulty in maintaining social ties among residents or exerting social control over public spaces, diverse areas with substantial population changes between day and night might differ in their social organization or outcomes such as fear of crime relative to an otherwise-diverse space that stays stably diverse at all times of the day (Boessen et al. 2017). Future studies could elucidate the downstream outcomes of intraday diversity change and other within-day contextual changes among the individuals whose lives intersect with such areas.

Third, the cosmopolitan canopy where varied people spend time during the day includes neighborhoods that are diverse at all times but also many contexts where Black, Latino, and Asian folks otherwise experience residential segregation at times outside of work (Anderson 2004, 2011). Still, much of the diversification that daily mobility brings is centered on neighborhoods where racial and ethnic minorities are represented prominently in the residential population. Further, many White residential spaces have low levels of racial and ethnic diversity not only in their nighttime population, as previous research has noted (Lichter et al. 2015; Parisi et al. 2019), but also during the day. Our study highlights how the tendency for White individuals’ mobility to favor White neighborhoods demonstrated in activity pattern studies (e.g., Candipan et al. 2021) maps onto patterns of intraday contextual change that further structure different trajectories of intraday contextual change among Whites relative to Black, Hispanic, and Asian persons. These observations motivate additional scholarship assessing whether consequences of short-run population change are also unequally distributed and structured in ways that reflect residential segregation. Such evidence would naturally build on findings from mobility studies using nighttime population data to characterize social contexts during the day (e.g., Chang et al. 2021).

Finally, future research should examine the impact of structural shocks (e.g., the COVID-19 pandemic) on daily mobility patterns because such shocks tend to affect racial and ethnic groups to varying degrees. Whether such changes stem from the downstream consequences of economic downturns (e.g., the Great Recession) on people's mobility owing to job loss or the myriad mobility changes elicited by events that intersect with social and economic activity (e.g., COVID-19), future studies should consider the short-run and lasting implications of these shocks for the potential diversity contact that individuals have through their mobility patterns.

Despite making several theoretical contributions to research on neighborhood diversity, metropolitan segregation, and activity spaces, our study has some limitations. First, people's daily activities extend beyond residing at home and going to work, but CTPP data limitations prevent us from directly addressing how these other forms of daily mobility matter for contextual diversity levels. Second, we cannot explore how educational and occupational differences might structure interactions at work or between those who are present in a given workhood. Either of these factors relevant to daily mobility might operate such that greater exposure to diverse settings reduces perceptions of group difference or greater contact begets greater intergroup conflict. Future efforts to adjudicate which of these theoretical expectations holds—whether in the context of other daily rounds beyond work or across individuals’ socioeconomic status or other dimensions—stand to advance our understanding of how the prevalence and impact of contextual diversity changes throughout the day.

Neighborhoods have long been considered little social worlds, implying that they are dynamic and changing. Further, research is marked by a growing recognition that individuals’ daily lives include mobility to destinations outside of where they live. Nevertheless, we know relatively little about how much the empirical picture provided by snapshots of places’ residential compositions masks important changes in where people spend time and what their communities are like outside of this “nighttime” perspective. By thinking about neighborhood change as occurring over decades but also during a given workday, we can start to properly understand the long- and short-run dynamics of growing contextual diversity across the United States.

Notes

References