Exploring Factors Associated with Age Segregation by Race Groups in US Counties: A Longitudinal Perspective, 2005–2024

Abstract

Research on age segregation, defined as the spatial separation between older adults (65+) and younger populations, has lagged behind demographic shifts in the United States (US). Existing studies are often outdated, cross-sectional, and narrowly focused on metropolitan areas, with few examining race-specific patterns or structural drivers over time at the national scale. This exploratory study addresses these gaps by analyzing longitudinal trends and determinants of age segregation in US counties from 2005 to 2024, with attention to racial variation (White, Black, and Other). Using county-level data from the American Community Survey and fixed-effects regression models, we estimate how demographic, socioeconomic, industrial, and housing factors relate to changes in age segregation. Results indicate that overall age segregation declined in the mid-2010s but increased in recent years, which may reflect the responses to the economic recession in the late 2000s and baby boomers’ aging. Race-specific analyses show that age segregation is highest among Black populations, with contributing factors largely similar across groups but more pronounced for Black communities. Racial/ethnic diversity is consistently associated with lower age segregation, suggesting that demographic heterogeneity fosters intergenerational integration. Employment in secondary industries (e.g., construction and manufacturing) is positively associated with age segregation, particularly among Black populations, whereas housing market characteristics emerge as strong predictors of both overall and race-specific age segregation. These findings underscore the importance of considering race-specific and structural factors in understanding age segregation as the US population continues to age.

1. Introduction

The United States (US) has undergone rapid population aging, with the number of older adults (aged 65 and over) increasing from 35 million in 2000 to nearly 57 million by 2022 [1]. Despite this significant demographic shift, there remains a lack of research on how residential separation between older and younger populations—commonly referred to as age segregation—has evolved during this period. Even less is known about the factors driving changes in age segregation over time. As noted by Das Gupta and Wong [2], much of the extant literature on age segregation is either outdated [3,4,5] or limited to metropolitan areas [6]. Among the few studies that attempt to examine trends in age segregation [2,7], findings are mixed regarding whether age segregation is increasing. Moreover, these studies often lack a longitudinal perspective, fail to explore contributing factors, and overlook race-specific patterns of age segregation—that is, segregation occurring within specific racial or ethnic groups.

The concept of segregation refers to the extent to which two or more social groups are spatially differentiated across geographic units (e.g., tracts) that comprise a main unit of interest (e.g., county) [8]. Segregation is conceptually and methodologically different from diversity or composition, which is concerned with the relative size of a certain social group in an area [9]. Prior studies have focused on racial/ethnic segregation (also known as residential segregation) due to the concern about the consequence of spatial structure on social order and distribution of resources [5,8]. However, age, as a critical factor for social integration and social expectation at both the individual- and macro-level, has not received sufficient attention in the literature on segregation. In many societies, older adults are isolated from participation in important social institutions [10], such as the labor market and school, and the traditional perspective depicts aging as a process of withdrawing oneself from society, such as diminished social roles and reduced social contacts and community engagement [11]. By contrast, it has been suggested that older adults have the unique potential to facilitate community development, civic engagement, and family dynamics [12,13,14]. Recognizing the diverse roles older adults can play underscores the need to examine age-based segregation as a distinct and consequential dimension of social stratification.

Building on the conceptual features of age outlined by LaGory and colleagues [5], we propose three reasons why older adults may be residentially segregated from younger populations, and why age segregation is likely to shift in response to broader social and demographic processes. First, from the ecological perspective, individuals at different life stages occupy distinct positions in the family and life cycle, leading to differential needs for space and access to desirable locations [4,15]. These differing spatial preferences, combined with disparities in purchasing power within the housing market, can lead to spatial separation between older and younger populations.

Second, people often choose to live near others who share similar backgrounds, values, norms, or life experiences. Such clustering can serve to minimize social friction or manage the cognitive and emotional demands of cultural heterogeneity [5,16]. Given the divergent lifestyles and preferences of older and younger cohorts, this tendency may contribute to age-based residential clustering. For example, older adults in retirement villages have reported strong social capital and mutual trustworthiness with other residents [17], highlighting the possible voluntary separation between older and younger populations and divergent needs for housing, amenities, and services.

Third, the political model—which explores how political and social structures shape the aging process and the experiences of older adults—suggests that age segregation may arise not from competition over space, but from age-based discrimination [5,18]. In this view, younger populations may be reluctant to reside in neighborhoods with high concentrations of older adults, thereby contributing to age segregation. Additionally, individuals in low-income communities may lack the resources to relocate, further reinforcing this pattern [19]. As individuals age and mobility declines, these structural constraints can intensify, leading to more pronounced age segregation. As such, age segregation may reflect broader population dynamics and the symbolic meanings attached to neighborhoods.

While the reasons above suggest that age segregation may increase over time due to the growing older population, the existing evidence for this assertion remains limited and inconclusive. Specifically, more than four decades ago, scholars investigated the variation in age segregation and the determinants of age segregation in US metropolitan areas [3,4,5,20,21]. While it is commonly believed that age segregation has been increasing in urban areas [7,22], a study using nationwide data between 1990 and 2010 reported that age segregation decreased in the US over this time [2], which contradicts prior research. One plausible explanation for the conflicting findings is the inconsistent definitions of “older adults.” For example, Winkler [7] used age 60 to define older adults, whereas Das Gupta and Wong [2] employed age 65 to calculate age segregation, which is the most commonly used definition in official/government statistics and certain social benefits.

The existing knowledge gaps in age segregation research are particularly significant given the aging of the baby boomer cohort—individuals born between 1946 and 1964—who began entering older adulthood in 2011. This demographic shift is often referred to as the “silver tsunami,” highlighting the substantial influx of older adults and its potential societal impacts [23]. For example, the rapid aging process is associated with widening wealth inequality, particularly in high-income countries [24], aggravates the financial burden of the healthcare system, and creates new challenges (e.g., demand for long-term care) for society [25]. Beyond the socioeconomic context, age segregation has far-reaching implications for social cohesion and intergenerational solidarity [26]. When older and younger populations live in separate spaces, opportunities for mutual learning, caregiving, and civic engagement diminish. This spatial divide can generate the dichotomy between “us” and “them,” reinforce ageist stereotypes, and reduce empathy across generations [27,28]. The increasing competition between age groups for limited resources that serve age-specific needs may hence weaken the social fabric of communities [27,29].

The discussion above suggests that the changes in population features, socioeconomic conditions, and local resources and services may be associated with the changes in age segregation. For example, age segregation has been found to be higher in rural areas where the concentrations of older adults are higher and larger minority groups are associated with higher age segregation [30]. However, recent research has not considered all these potential factors simultaneously. The goal of this exploratory study is to fill these gaps by first constructing a longitudinal dataset for US counties and then using fixed-effect modeling to examine potential factors related to the changes in age segregation. Our analysis specifically focuses on within-county, tract-level age segregation, measuring the uneven distribution of younger and older residents across census tracts within each county. Age segregation across counties, states, or regions is not examined in this study.

The analysis will be extended to race-specific age segregation to better understand whether the contributing factors differ by race groups. The race-specific analysis is necessary for the following reasons. First, age and race intersect to structure residential patterns. Examining only overall segregation masks important variation. Second, different racial/ethnic groups experienced distinct patterns of residential segregation and discrimination [31], creating different contexts for age-based patterns. Third, age structures differ by racial/ethnic group [32], which may contribute to age segregation in different ways. Finally, socioeconomic disparities across racial/ethnic groups affect access to age-restricted communities and amenity-rich retirement areas.

2. Materials and Methods

This study draws on county-level data from the US Census Bureau’s American Community Survey (ACS) 5-year estimates, which provide detailed information on population demographics, housing, economic status, and educational attainment [33,34]. Counties represent functional economic units with administrative coherence in policy implementation, making them ideal for studying structural drivers of segregation [35,36]. While previous research has advanced our understanding of age segregation at the metropolitan and block levels, longitudinal analyses at the county level remain limited [6,37]. Therefore, we construct a panel dataset spanning 2005 to 2024 by combining four non-overlapping 5-year estimates: 2005–2009, 2010–2014, 2015–2019, and 2020–2024.

The dependent variable, age segregation, is measured using the Theil information index (H index) [38,39], which quantifies the uneven distribution of age groups across census tracts within a county. It captures residential separation between age groups within a given county but does not reflect age segregation across counties, states, or regions. The Theil information index can be expressed as below:

$$H={\displaystyle \sum _{j=1}^J}{\displaystyle \frac{n_j(E-E_j)}{EN}}$$

(1)

$$E={\displaystyle \sum _{i=1}^I}{\displaystyle \frac{N^i}{N}}\ln {\displaystyle \frac{N}{N^i}}$$

(2)

$$E_j=\sum _{i=1}^I{\displaystyle \frac{n_j^i}{n_j}}\mathit{ln}{\displaystyle \frac{n_j}{n_j^i}}$$

(3)

where $N$ is the total population of the county, $n_j$ is the population of tract $j$ in that county, and $N^i$ is the total population of social group $i$ in the county. $E$ is the global entropy of the county and $E_j$ is the local entropy of tract $j$ in the county. $n_j^i$ is the population of social group $i$ in tract $j$.

To achieve the goal of this study, four indices are calculated. The first captures overall age segregation between the younger population (aged 64 and under) and the older population (aged 65 and over), regardless of racial background. The remaining three indices measure race-specific age segregation for “White”, “Black”, and “Other” populations, respectively. “White” refers to individuals identifying as “White alone”, and “Black” includes those identifying as “Black or African American alone”. The “Other” category encompasses individuals identifying as “American Indian and Alaska Native alone”, “Asian alone”, “Native Hawaiian and Other Pacific Islander alone”, “Some Other Race alone”, or “Two or More Races”. For example, the White-specific index is calculated using only the younger and older White populations within each county. The same approach is applied to compute age segregation within the Black and Other populations.

Choosing age 65 as the cut-off point is based on two reasons. On one hand, age 65 marks Medicare eligibility, traditional retirement age, and housing policy thresholds. On the other hand, our focus on older adult/younger population separation aligns with recent age segregation research [30], making our study comparable with the extant literature.

Our independent variables are grouped into four categories: demographic characteristics, socioeconomic status (SES), industrial structure, and housing market dynamics, all measured at the county level. These categories are commonly used in residential segregation research [40] and adapted to the age segregation context. The demographic characteristics include median age, racial and ethnic composition—specifically, the proportions of non-Hispanic White, non-Hispanic Black, and Hispanic residents—and the sex ratio (males to females). Following the work by Sampson and colleagues [41], we assess socioeconomic status with a composite index derived from the principal component analysis, which mitigates multicollinearity while capturing the latent SES construct through dimensionality reduction in correlated indicators [42]. The index incorporates median household income, the proportion of single-parent households with children under 18, the proportion of adults aged 25 and older with at least a college degree, the proportion of adults aged 18 to 64 living in poverty, and the unemployment rate among civilians aged 16 and older in the labor force. The industrial structure of a county is captured through two variables: the proportion of the workforce employed in primary industries (agriculture, forestry, fishing, hunting, and mining) and the proportion employed in secondary industries (construction and manufacturing). Finally, housing market dynamics are measured by the proportion of occupied housing units built before 2000 and the proportion of housing units where costs exceed 30 percent of household income.

We employ a fixed-effects regression model to estimate the relationship between age segregation and county-level predictors. The model is specified as [43]:

$$y_{it}={{\beta }^{T}X}_{it}+{\mu }_i+{ϵ}_{it}$$

(4)

In this equation, y_it is the age segregation index for county i in time t. β^TX_it includes independent variables X and their coefficients β. μ_i is the county fixed-effects term, which includes all the unobserved factors that do not change with time. ${ϵ}_{it}$ is the error term. The fixed-effect regression approach has been widely used in longitudinal data analysis to yield robust estimates due to several strengths [43,44]. For example, the time-invariant unobserved heterogeneity is considered in the analysis so that it is not necessary to measure all confounders. In addition, fixed-effect modeling focuses on within-unit variation, allowing researchers to better understand how changes in the independent variables are related to the changes in the dependent variable within an area (i.e., a county in this study).

The analytic strategy unfolds in three stages. First, we generate descriptive statistics for all variables discussed above, both for the pooled dataset and separately by each of the four waves of 5-year ACS estimates. This step provides a foundational understanding of how age segregation and related social and demographic characteristics in US counties have changed over the study period. In the second stage, we employ fixed-effects modeling to examine overall age segregation. Four nested models are estimated, sequentially incorporating the four groups of independent variables to assess their individual and combined contributions. The third stage extends the fixed-effects modeling approach to race-specific age segregation, allowing us to explore whether and how the determinants of age segregation differ across racial groups. To construct a balanced panel dataset, we first matched counties across four waves, ensuring that each wave includes the same set of counties. This resulted in a dataset comprising 12,432 observations, with 3108 counties represented in each wave.

3. Results

Table 1 presents the descriptive statistics for all variables. The average overall age segregation index between younger and older populations from 2005 to 2024 was 0.0167 (SD = 0.020). The overall age segregation decreases from 0.0170 in 2005–2009 to 0.0151 in 2015–2019. However, the overall age segregation increases to 0.0191 in the 2020–2024 wave. When examining race-specific age segregation, the Theil indices were consistently higher than the overall age segregation. The Black age segregation was the highest (mean = 0.1348), and the White and Other age segregation scores were comparable (approximately at 0.018). Temporally, White and Other groups exhibit a pattern consistent with the overall age segregation, decreasing in the first three waves and increasing in the 2020–2024 wave. In contrast, age segregation among Blacks first decreased in the 2010–2014 wave (0.1237) then increased to 0.1506 in the 2020–2024 wave. The geographic distribution of the four dependent variables can be found in Figures S1–S4 in the Supplementary Files.

Table 1. Descriptive statistics of variables in this study.

	Total Sample	2005–2009	2010–2014	2015–2019	2020–2024
	Mean (S.D.)	Mean (S.D.)	Mean (S.D.)	Mean (S.D.)	Mean (S.D.)
Dependent Variables
Overall Age segregation	0.0167 (0.020)	0.0170 (0.019)	0.0155 (0.020)	0.0151 (0.020)	0.0191 (0.021)
Age segregation within Whites	0.0183 (0.022)	0.0187 (0.021)	0.0164 (0.021)	0.0161 (0.021)	0.0220 (0.025)
Age segregation within Blacks	0.1348 (0.128)	0.1317 (0.128)	0.1237 (0.120)	0.1332 (0.128)	0.1506 (0.136)
Age segregation within other groups	0.0178 (0.021)	0.0182 (0.020)	0.0163 (0.021)	0.0160 (0.021)	0.0206 (0.023)
Independent Variables
Demographic Characteristics
Median age	40.889 (5.313)	39.473 (4.936)	40.749 (5.201)	41.464 (5.392)	41.869 (5.397)
Proportion of Non-Hispanic White	0.770 (0.197)	0.796 (0.190)	0.779 (0.196)	0.766 (0.199)	0.741 (0.197)
Proportion of Non-Hispanic Black	0.089 (0.144)	0.088 (0.144)	0.090 (0.145)	0.090 (0.145)	0.086 (0.141)
Proportion of Hispanic	0.090 (0.136)	0.076 (0.128)	0.087 (0.135)	0.095 (0.139)	0.103 (0.140)
Sex ratio	1.007 (0.118)	0.995 (0.109)	1.006 (0.115)	1.007 (0.111)	1.022 (0.133)
Socioeconomic Status
SES score	−0.083 (0.858)	−0.083 (0.867)	−0.080 (0.874)	−0.093 (0.817)	−0.078 (0.873)
Industrial Structure
Proportion of employed in primary industry	0.066 (0.073)	0.069 (0.075)	0.070 (0.075)	0.065 (0.073)	0.059 (0.068)
Proportion of employed in secondary industry	0.201 (0.072)	0.213 (0.075)	0.193 (0.070)	0.198 (0.071)	0.198 (0.068)
Housing Market
Proportion of occupied housing units built before 2000	0.838 (0.099)	0.909 (0.060)	0.853 (0.077)	0.821 (0.088)	0.770 (0.108)
Proportion of housing cost higher than 30% in income	0.256 (0.066)	0.273 (0.069)	0.272 (0.069)	0.239 (0.059)	0.241 (0.059)
N	12,432	3108	3108	3108	3108

Note: Proportions can be interpreted as percentages by multiplying by 100.

Regarding demographic characteristics, the average median age in US counties over the study period is around 41 years overall (SD = 5.313) and increased from 2005 to 2024. Racial/ethnic composition largely suggests that the proportion of the non-Hispanic White population decreased by roughly 5 percentage points and the proportion of the Hispanic population grew by slightly more than 2 percentage points. Sex ratios generally exceeded 1, which indicates more males than females, across all periods except 2005–2009. The SES score averaged around −0.083 for all counties in four periods. Employment patterns showed that almost 7 percent of working populations were employed in the primary industries, while roughly 20 percent worked in the secondary industries. Housing market indicators suggested that most counties contained older housing stock, with over 82% of occupied units built before 2000 during earlier periods, declining to 77.0% in 2020–2024. Housing cost burdens average 25% of household income overall, showing a modest decline from 2005 to 2009 (27.3%) to 2020–2024 (24.1%).

We also summarize descriptive statistics for the Northeast, Midwest, South, and West from 2005 to 2024 (Table 2). Overall age segregation is highest in the Northeast (0.0223) and lowest in the Midwest (0.0126), with the South (0.0173) and West (0.0220) in between. For age segregation within Whites, the West ranks highest and the Midwest lowest. In contrast, age segregation within Blacks is substantially higher across all regions, with the Northeast (0.2022) nearly doubling the South (0.1038). Age segregation among other groups is similar to the overall pattern.

Table 2. Descriptive statistics of variables by region.

	Northeast	Midwest	South	West
	Mean (S.D.)	Mean (S.D.)	Mean (S.D.)	Mean (S.D.)
Dependent Variables
Overall Age segregation	0.0223 (0.018)	0.0126 (0.014)	0.0173 (0.021)	0.0220 (0.026)
Age segregation within Whites	0.0227 (0.021)	0.0125 (0.015)	0.0206 (0.025)	0.0229 (0.027)
Age segregation within Blacks	0.2022 (0.120)	0.1471 (0.131)	0.1038 (0.111)	0.1746 (0.151)
Age segregation within other groups	0.0223 (0.019)	0.0124 (0.014)	0.0198 (0.024)	0.0221 (0.026)
Independent Variables
Demographic Characteristics
Median age	42.126 (3.850)	41.606 (5.093)	40.150 (4.981)	40.949 (6.991)
Proportion of Non-Hispanic White	0.835 (0.156)	0.883 (0.119)	0.690 (0.204)	0.727 (0.202)
Proportion of Non-Hispanic Black	0.048 (0.062)	0.024 (0.045)	0.165 (0.179)	0.012 (0.019)
Proportion of Hispanic	0.065 (0.080)	0.044 (0.059)	0.103 (0.156)	0.178 (0.172)
Sex ratio	0.988 (0.106)	1.008 (0.073)	1.000 (0.139)	1.041 (0.129)
Socioeconomic Status
SES score	−0.069 (0.527)	−0.467 (0.830)	0.205 (0.825)	−0.103 (0.794)
Industrial Structure
Proportion of employed in primary industry	0.021 (0.019)	0.077 (0.077)	0.054 (0.060)	0.101 (0.094)
Proportion of employed in secondary industry	0.186 (0.052)	0.217 (0.075)	0.206 (0.070)	0.148 (0.049)
Housing Market
Proportion of occupied housing units built before 2000	0.892 (0.052)	0.870 (0.078)	0.811 (0.107)	0.824 (0.102)
Proportion of housing cost higher than 30% in income	0.311 (0.067)	0.231 (0.053)	0.257 (0.061)	0.288 (0.078)
N	868	4220	5688	1656

Note: Proportions can be interpreted as percentages by multiplying by 100.

In terms of demographic characteristics, the Northeast has the highest median age, while the Midwest has the highest proportion of non-Hispanic Whites, the South the highest proportion of non-Hispanic Blacks, and the West the largest Hispanic population. Sex ratios are relatively balanced, though the Northeast skews slightly female, and the Midwest and West slightly male. Only the South shows a positive SES score, while other regions are negative. The West has the highest employment in primary industries, whereas the Midwest displays the largest proportion employed in the secondary sector (21.7%). Across all regions, most housing units are built before 2000, especially in the Northeast. Housing cost burden is lowest in the Midwest, which may indicate relatively greater affordability.

The fixed-effect modeling results of overall age segregation were presented in Table 3. Model 1 considers demographic characteristics and the results suggested that racial/ethnic composition plays an important role in reducing overall age segregation. For example, when the non-Hispanic White population increased by 10 percent (i.e., 0.1 in proportion), the overall age segregation decreased by 0.005 (i.e., −0.052 × 0.1 = −0.0052). While this number seems to be small, it is almost one-third of the mean (i.e., 0.0167 in Table 1) of overall segregation. The estimated relationship of Hispanic ($\hat{\beta }$ = −0.046) and overall age segregation is comparable with that of non-Hispanic White, while the non-Hispanic Black group ($\hat{\beta }$ = −0.074) seems to have a relatively stronger relationship. County-level sex ratio was not related to age segregation. The SES score was included in Model 2, and it was not statistically related to the change in overall age segregation.

Table 3. Fixed effect modeling results of overall age segregation, 2005–2024 (N = 12,432).

	Model 1	Model 2	Model 3	Model 4
	Estimates	Estimates	Estimates	Estimates
Demographic Characteristics
Median age	0.0002 ** (0.0000, 0.0003)	0.0002 ** (0.0000, 0.0003)	0.0002 ** (0.0000, 0.0003)	0.0001 (−0.0000, 0.0003)
Proportion of non-Hispanic White	−0.0522 *** (−0.0656, −0.0387)	−0.0524 *** (−0.0659, −0.0390)	−0.0522 *** (−0.0658, −0.0387)	−0.0406 *** (−0.0574, −0.0238)
Proportion of non-Hispanic Black	−0.0740 *** (−0.1037, −0.0443)	−0.0740 *** (−0.1037, −0.0444)	−0.0738 *** (−0.1037, −0.0440)	−0.0635 *** (−0.0933, −0.0338)
Proportion of Hispanic	−0.0460 *** (−0.0664, −0.0256)	−0.0461 *** (−0.0665, −0.0257)	−0.0457 *** (−0.0664, −0.0250)	−0.0428 *** (−0.0638, −0.0217)
Sex ratio	0.0016 (−0.0037, 0.0069)	0.0015 (−0.0038, 0.0068)	0.0015 (−0.0039, 0.0069)	0.0013 (−0.0040, 0.0066)
Socioeconomic Status
SES score		−0.0002 (−0.0007, 0.0004)	−0.0002 (−0.0007, 0.0004)	−0.0003 (−0.0008, 0.0003)
Industrial Structure
Proportion of employed in primary industry			−0.0024 (−0.0117, 0.0070)	0.0011 (−0.0086, 0.0108)
Proportion of employed in secondary industry			0.0039 (−0.0037, 0.0115)	0.0073 (−0.0006, 0.0151)
Housing Market
Proportion of occupied housing units built before 2000				−0.0099 ** (−0.0159, −0.0038)
Proportion of housing cost higher than 30% in income				0.0073 (−0.0004, 0.0150)
Intercept	0.0584 *** (0.0430, 0.0738)	0.0588 *** (0.0433, 0.0743)	0.0577 *** (0.0420, 0.0734)	0.0562 *** (0.0404, 0.0721)
AIC	−88,112.6	−88,111.2	−88,109.9	−88,146.2
BIC	−88,075.5	−88,066.6	−88,050.5	−88,072.0

Note: *** p < 0.001, ** p < 0.01. Numbers in parentheses refer to 95% confidence intervals.

With respect to industrial structure (Model 3), the results do not suggest that the population employed in the primary or secondary industries matters. When the housing market is considered (Model 4), higher proportions of occupied housing units built before 2000 are related to lower overall age segregation ($\hat{\beta }$ = −0.0099).

Table 4 reports the fixed-effects results for race-specific age segregation, controlling for all independent variables. Notably, the patterns for race-specific segregation closely resemble those observed for overall age segregation (Model 4 in Table 3). Racial/ethnic composition variables, along with employment in secondary industries and occupied housing units built before 2000, show significant associations with changes in age segregation among White, Black, and Other groups. However, age segregation among Black populations appears more sensitive to these factors than among the other groups. For instance, the racial composition variables have significant negative effects for White and Other groups, but these effects are not observed for Black age segregation. The proportions of individuals employed in secondary industries and of occupied housing units built before 2000 significantly influence age segregation across all three groups. However, the magnitudes of these effects are substantially larger for Blacks than for Whites and Others. Specifically, the coefficient for secondary industry employment is 0.0181 for Whites and 0.0088 for Others, compared to 0.0884 for Blacks. Similarly, the coefficients for housing characteristics are −0.0167 for Whites and −0.0086 for Others, whereas the effect for Blacks is much stronger at −0.1358.

Table 4. Fixed effect modeling results of age segregation by race/ethnicity, 2005–2024.

	Whites (N = 12,432)	Blacks (N = 12,432)	Other (N = 12,432)
	Estimates	Estimates	Estimates
Demographic Characteristics
Median age	0.0001 (−0.0000, 0.0003)	−0.0005 (−0.0016, 0.0005)	0.0001 (−0.0000, 0.0003)
Proportion of Non-Hispanic White	−0.0528 *** (−0.0748, −0.0309)	−0.0914 (−0.2258, 0.0431)	−0.0495 *** (−0.0669, −0.0321)
Proportion of Non-Hispanic Black	−0.0904 *** (−0.1238, −0.0570)	0.1609 (−0.0084, 0.3302)	−0.1083 *** (−0.1443, −0.0722)
Proportion of Hispanic	−0.0557 *** (−0.0868, −0.0246)	−0.1665 (−0.3492, 0.0161)	−0.0440 *** (−0.0665, −0.0215)
Sex ratio	−0.0048 (−0.0101, 0.0005)	0.0136 (−0.0116, 0.0388)	−0.0017 (−0.0068, 0.0034)
Socioeconomic Status
SES score	−0.0002 (−0.0010, 0.0006)	0.0028 (−0.0021, 0.0077)	−0.0004 (−0.0010, 0.0003)
Industrial Structure
Proportion of employed in primary industry	−0.0041 (−0.0160, 0.0078)	−0.0532 (−0.1445, 0.0382)	−0.0048 (−0.0152, 0.0056)
Proportion of employed in secondary industry	0.0181 *** (0.0080, 0.0282)	0.0884 * (0.0120, 0.1647)	0.0088 * (0.0001, 0.0175)
Housing Market
Proportion of occupied housing units built before 2000	−0.0167 *** (−0.0232, −0.0101)	−0.1358 *** (−0.1789, −0.0928)	−0.0086 ** (−0.0150, −0.0023)
Proportion of housing cost higher than 30% in income	0.0084 (−0.0006, 0.0174)	−0.0503 (−0.1145, 0.0140)	0.0077 (−0.0009, 0.0163)
Intercept	0.0802 *** (0.0586, 0.1019)	0.3259 *** (0.1928, 0.4590)	0.0705 *** (0.0537, 0.0873)
AIC	−82,585.0	−27,220.1	−85,956.0
BIC	−82,510.7	−27,145.8	−85,881.7

Note: *** p < 0.001, ** p < 0.01, * p < 0.05. Numbers in parentheses refer to 95% confidence intervals.

4. Discussion

This exploratory study provides a longitudinal analysis of age segregation in US counties from 2005 to 2024, with a novel focus on race-specific patterns. Our findings contribute to the limited but growing body of the literature on age segregation by offering updated empirical evidence and identifying key structural and demographic factors associated with its changes over time. As discussed previously, prior research has tended to focus on urban areas, and a nationwide examination of how age segregation evolves during the rapid aging process in the US remains desired [2]. Importantly, race-specific age segregation has been understudied, particularly from a longitudinal perspective. Below we summarize key results that advance our understanding of age segregation in the US over roughly the last 20 years.

First, we find that overall age segregation has fluctuated over the study period, with a decline from 2005 to 2019 followed by an increase from 2020 to 2024. The decline may reflect the increase in local moves around 2010, such as within-county moves [45] and economic pressures limiting young adults’ residential choices [46]. Specifically, around 2010, it was reported that young Americans tend to move locally (within-county), which may lower age segregation. As for economic recession, limited economic opportunities may undermine housing affordability and make young adults choose to live with parents or age-mixed communities. These findings align with concerns raised by Freedman and Stamp [22], who argue that the US is becoming increasingly segregated by age, particularly in urban and suburban contexts. Another possible explanation for the rise in age segregation between 2020 and 2024 is the COVID-19 pandemic, which restricted internal migration and may have constrained residential mobility, thereby increasing separation between younger and older populations. As Petersen and colleagues [47] noted, there is little evidence of widespread migration to rural areas during the pandemic, and out-migration from rural areas remained low even in the post-pandemic period. This lack of migration may, at least partially, account for the recent increase in overall age segregation.

Another reason why age segregation may have increased recently may be the rise in remote work and digital connectivity during the pandemic [48], which reinforced socioeconomic advantages and spatial separation of work and residence that favored some groups over others and may have long-term implications for age segregation. On one hand, remote work could enable younger populations to relocate to traditionally older or rural communities, potentially fostering greater age integration. On the other hand, digital divides—especially among older adults—may reinforce spatial and social isolation [49]. Our data do not allow us to further investigate which argument is more likely to happen, but the descriptive findings seem to suggest the latter.

Second, racial and ethnic composition emerged as a consistent predictor of overall and all race-specific age segregations. Counties with higher proportions of non-Hispanic Black and Hispanic residents experienced lower levels of age segregation, suggesting that racial diversity may facilitate intergenerational integration. Though the role of racial/ethnic diversity in shaping residential segregation has been noted [9,50], it has not been linked to the changes in age segregation, both overall and race-specific. Increasing racial/ethnic diversity is likely to facilitate intergroup contact and shared community experiences [51], which can reduce social distance and foster integration across demographics, including age structure. It should be noted that Black age segregation appears to be more sensitive to racial/ethnic composition than other groups, highlighting the potential intertwined relationship between age structure and racial/ethnic composition in US counties.

Third, employment in secondary industries was positively associated with race-specific segregation with Black age segregation having the strongest relationship. This may reflect the spatial clustering of secondary industrial employment in areas with limited housing diversity or age-specific amenities, which could disproportionately affect older adults [52].

The proportion of occupied housing units built before 2000 was found to be negatively related to overall and race-specific age segregation. The significantly negative association between older housing stock and age segregation suggests that neighborhoods with a high proportion of homes built before 2000 tend to foster greater age integration. Older housing is often more affordable and located in established communities with mixed amenities, making these areas attractive to both younger households seeking lower-cost options and older adults aging in place [53,54]. These patterns highlight the role of housing market dynamics in shaping age-based spatial arrangements and suggest that affordability pressures and the availability of older housing stock can counteract tendencies toward age segregation [22].

The fixed-effects approach does not allow for the estimation of time-invariant factors, such as region of the county. We conducted a sensitivity analysis by census region (Northeast, Midwest, South, and West) to explore potential regional differences. The results (see Supplementary Files) indicate that counties in the Midwest and South are more responsive to changes in racial/ethnic composition, employment in secondary industries, and housing market conditions. These findings suggest that regional variation warrants further attention in future research. Similarly, we replicated the analyses using the dissimilarity index, a widely used segregation measure in prior studies [8,9], and found largely consistent results (see Supplementary Files), reinforcing the robustness of our findings.

This study is subject to several limitations. First, our analysis measures age segregation as the uneven distribution of younger and older residents across census tracts within counties, capturing only within-county, tract-level patterns. Age segregation at higher geographic levels, including between-county, between-state, and between-region variation, falls outside the scope of this study. These higher-level patterns may be particularly important given retirement migration trends and the concentration of older adults in certain regions. A multi-level decomposition of the Theil H index, which would allow quantification of within-county versus between-county versus between-state and between-region components over time, represents a natural extension of this work. The use of county-level data may obscure finer-grained patterns of segregation observable at the neighborhood or tract level, and the conclusions may be altered [55]. It is also worth noting that county boundaries are administrative rather than ecological and that county size varies considerably across the US, raising concerns about the Modifiable Areal Unit Problem (MAUP) [55,56]. Although our H index is computed from census-tract-level data and thus incorporates within-county local variation, our fixed-effects estimates reflect county-level patterns and may differ from analyses conducted at the tract level. Future studies should explore age segregation at smaller geographic scales and examine whether age segregation is sensitive to the concentration of population in institutional group quarters (e.g., prisons and mental hospitals). Second, spatially extended versions of the H index should be considered in future research, which incorporate contiguity functions, density functions, or kernel functions to account for the geographic arrangement of populations across tracts (e.g., the work of Reardon and colleagues [57,58]). Similarly, given the exploratory nature of this study, other dimensions of age segregation in a county, such as concentration or centralization [8], are not considered. Third, our findings may not fully reflect the lived experiences of inter-generational interaction, and as such, studies that incorporate qualitative data to better understand the mechanisms behind residential choices would be useful. Fourth, while fixed-effects models control for time-invariant unobserved heterogeneity, they cannot account for all potential confounders. For example, land use policies, such as zoning and housing constructions/development, may change relatively fast compared with other social and demographic changes [59], which may contribute to age segregation. This type of variable should be considered in future research. Fifth, future studies should examine whether the factors associated with age segregation differ across dimensions of segregation [8,35]. Finally, there is heterogeneity among young populations (e.g., young adults vs. middle-aged), and our approach likely masks such heterogeneity. Future research should examine multiple age groups simultaneously.

5. Conclusions

Several implications can be drawn from the findings. First, counties with a strong secondary industrial base and old housing stock may need targeted interventions, such as mixed-use zoning, accessible public spaces, and age-inclusive housing, to counteract the isolating effects of economic clustering on age segregation for all racial groups. Second, the results underscore the importance of considering race-specific dynamics, especially among the Black population, as a driving force underlying the process of age segregation. Third, given the influence of racial/ethnic diversity on age segregation, demographic heterogeneity may contribute to environments where boundaries between age groups are more fluid.

To sum up, this exploratory study underscores the dynamic and multifaceted nature of age segregation in US counties between 2005 and 2024. By incorporating a longitudinal and race-specific lens, we offer some new insights into how demographic, economic, and structural factors shape intergenerational spatial patterns. As the US population continues to age, understanding and addressing age segregation will be critical for fostering desirable residential environments.