We used the publicly available longitudinal data from the Midlife in the U.S. (MIDUS) study, an ongoing national longitudinal study led by the University of Wisconsin-Madison Institute on Aging. The purpose of MIDUS is to examine the health of the U.S. adult population (43), specifically, the role of behavioral, psychological, and social factors on age-related variations in the health and wellbeing of adults.

The first wave of the MIDUS study (MIDUS 1) was conducted using survey data collected during 1995–1996 from a sample of 7,000 U.S. adults aged 25–74. For the second wave (MIDUS 2), data was collected between 2004 and 2005, with an expanded scope to include cognition data and daily experiences, along with a new sample of Black respondents from Milwaukee, WI. The third wave (MIDUS 3) collected data between 2013 and 2022. MIDUS 3 included a core survey data, and a retention-early warning component aimed at reinstating participants who had previously dropped out. The retention of the sample in the MIDUS longitudinal study remained high. Of the original MIDUS 1 (1995–1996) sample, 69% of the cognitive interview participants returned for MIDUS 2 (2004–2005) and 77% (adjusted for mortality) participated in MIDUS 3 (2013–2014) (43).

We included national respondents who completed all three waves in our study (2,264 White respondents and 76 Black respondents). We used multiple imputation to address missing data by generating several plausible values for each missing observation and creating multiple complete datasets.

We supplemented the MIDUS data to increase the sample size of Black respondents and performed multiple imputations to address for missingness in the data. Given the absence of psychometric tests evaluating the structure of the “Contributions to Community” and “Community Involvement” scales, we conducted a one-factor confirmatory factor analysis (CFA) using the lavaan package in R version 4.3.3 (32, 33). As part of the CFA, we assessed the fit of the hypothesized measurement model for the scales using the Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), and Root Mean Square Error of Approximation (RMSEA). An acceptable model fit was indicated by CFI and TLI values > 0.90 and an RMSEA value < 0.05 (34).

We modeled longitudinal data for MIDUS 1 (1995–1996), MIDUS 2 (2004–2006), and MIDUS 3 (2013–2014) using multilevel linear mixed-effects modeling where measurements Y ti at each time point t is nested in individual i and using the lmer function from the lme4 package in R, which is specialized for fitting linear mixed-effects models (33). We used the fixed effects of race and education as predictors for “Contributions to Community” and “Community Involvement.” For the mixed-effects model, we designated “White” as the reference category for race and “less than high school education” as the reference category for education. This allowed us to compare the impacts of being Black and other education categories against these reference categories. We used a random intercept model to investigate the relationships among these variables. The equation for the model is below: Y ti = π 0 i + π 1 i t i + ε ti

Where: π 0 i = β 00 + β 01 ( Race ) + β 02 ( Edu ) + β 03 ( Time ) + β 04 ( Race × Edu ) + β 05 ( Race × Time ) + β 06 ( Edu × Time ) + β 07 ( Race × Edu × Time ) + u 0 i π 1 i = β 10 + β 11 ( Race ) + β 12 ( Edu ) + β 13 ( Time ) + β 14 ( Race × Edu ) + β 15 ( Race × Time ) + β 16 ( Edu × Time ) + β 17 ( Race × Edu × Time )

Where:

Y ti is the outcome for individual i at time t

One-factor model fit results for the Contributions to Community and Community Involvement scales demonstrated excellent and acceptable fit, respectively. The CFA one-factor model for Contributions to Community produced a CFI value of 0.98, TLI value of 0.98, and an RMSEA value of 0.03. For Community Involvement, the CFA one factor model produced a CFI value of 0.91, a TLI value of 0.90, and an RMSEA value of 0.05.

Figures 1, 2 assess racial differences in “Contributions to Community” and “Community Involvement” across three different time points: MIDUS 1 (1995–1996), MIDUS 2 (2004–2006), and MIDUS 3 (2013–2014). For each racial group, the plot shows the median score (the line inside the box), the interquartile range (the length of the box), and the range of the data (indicated by the whiskers). The presence of outliers is also highlighted beyond the whiskers. From the plots, we can compare the median scores and spread of “Contributions to Community” and “Community Involvement” between the two racial groups. A wider box or longer whiskers for one group suggests a greater variability in scores, and a higher median indicates greater community contributions or higher involvement. Differences in the position of the medians and the spread of the boxes across groups suggest racial disparities in community contributions and involvement. Figure 1 shows that Black respondents had higher median scores than White respondents on all three time points, suggesting that Black respondents contributed more to their community than their White counterparts. Figure 2 shows that Black respondents had higher median scores than White respondents on MIDUS 1 (1995–1996) and MIDUS 3 (2013–2014), pointing to their higher level of community involvement. In short, both Figures 1, 2 show that race-specific “Contributions to Community” and “Community Involvement” were relatively stable over time.

The longitudinal analyses of time in the mixed-effects model (Table 1) revealed that overall “Contributions to Community” remained stable over time (β_{MIDUS 2} = −0.01, p > 0.05; β_{MIDUS 3} = −0.10, p > 0.05), but there was a significant reduction in “Community Involvement” at MIDUS 3 (β_{MIDUS 2} = −0.29, p > 0.05; β_{MIDUS 3} = −1.10, p < 0.001). Similarly, race-specific “Contributions to Community” remained stable over time (β_{Black * MIDUS 2} = −0.05, p > 0.05; β_{Black * MIDUS 3} = −0.15, p > 0.05), but there was a significant increase in “Community Involvement” among Black respondents at MIDUS 3 (β_{Black * MIDUS 2} = 0.51, p > 0.05; β_{Black * MIDUS 3} = 1.92, p = 0.019). Also, time and education did not have any significant interaction effect on “Contributions to Community” at either MIDUS 2 or MIDUS 3 (p > 0.05), suggesting that contributions remained stable across time irrespective of the participant’s educational background. However, the interaction effects of time and education on “Community Involvement” were significant at MIDUS 3 (β_{High school diploma or GED * MIDUS 3} = 0.70, p = 0.008; β_{Some college education * MIDUS 3} = 0.64, p = 0.013; β_{College degree or higher * MIDUS 3} = 1.07, p < 0.001), pointing to a greater involvement in community at higher educational levels at MIDUS 3 (2013–2014) compared with MIDUS 1 (1995–1996).

The longitudinal analyses proved valuable and provided new insight into the stability and evolution of social capital over time. Findings suggest that an individual’s overall capacity or willingness to contribute to their community does not significantly change over time, regardless of racial group or education group. This implies that social capital is shaped by enduring individual factors (e.g., socioeconomic status) and structural factors (e.g., access to resources) rather than temporal shifts. Conversely, findings suggest that the reality of how an individual engages with their community does change over time. This implies that involvement varies as responsibilities (e.g., family or work obligations) and community dynamics (e.g., changes in social networks or local leadership) evolve. Furthermore, these changes in community involvement occurred in the later time period studied (i.e., the last wave of the study), suggesting that physical factors due to aging may also play a role (e.g., reduced mobility or declined health).

Importantly, the results strongly suggest that race is a powerful indicator of how community involvement changes over time, with White involvement significantly declining and Black involvement dramatically increasing in the respondent’s later years. This trend may be driven by the Black community’s generally heightened engagement in culturally specific forms of social capital, such as church activities, grassroots activism, or collective responses to societal challenges. This potentially offers Black people more support and adaptability than White people, who generally have lower engagement in culturally specific forms of social capital.

Race was strongly associated with social capital outcomes, particularly with higher community involvement over time. Black respondents demonstrated greater contributions and involvement than White respondents at most time points. This racial advantage may reflect longstanding cultural, familial, and institutional mechanisms such as church involvement, mutual aid networks, and collective action traditions that foster engagement irrespective of educational attainment.

Our study shows that education is a predictor of greater social capital. Participants with some college education or a college degree reported significantly greater contributions than those with less than a high school education. The significant interaction effects between education and time on community involvement at MIDUS 3 (Wave 3) suggest that the benefits of education on social capital accumulate or become more evident later in life. This demonstrates the role of education in fostering civic engagement and transferable skills that enhance long-term community participation.

The significant interaction effects between time and education on community involvement at the later time period (i.e., higher levels of education were associated with greater increases in involvement) further emphasize the role of education in shaping social capital overall. This implies that education may provide individuals with the resources, networks, and/or skills necessary to participate more actively in their communities. Therefore, structural factors, such as access to education, play an important role in facilitating social capital development over time.

Our study produced inconclusive findings regarding race-by-education interactions. Most interaction effects were not significant, except for Black respondents with some college education, who reported lower community contributions compared to White respondents with less than high school education. This highlights a potential nuanced interaction between race and education that may reflect differing expectations (e.g., collective or public good) or personal experiences (e.g., participating in Black Greek organizations) within this subgroup. Black respondents maintained a social capital advantage across all other educational levels, suggesting that racial identity may play a critical role in shaping social capital. Their community behavior may be deeply rooted in historical and cultural factors, independent of formal education and attainment level.

In our study, Black respondents were disproportionately represented in lower education categories, which underscores the persistent impact of historical and systemic inequities on educational opportunities. Results reflect the pattern of longstanding barriers in predominantly Black communities such as discriminatory policies, underfunded schools, and limited access to higher education resources. These educational disparities are not merely individual outcomes but are deeply rooted in structural inequalities that perpetuate cycles of disadvantage across generations.

Despite this overall educational disadvantage, Black respondents consistently exhibited higher community contributions and higher community involvement compared with their White counterparts. This racial advantage emphasizes the capacity of Black communities in fostering social capital, even in the face of disparities and systemic challenges. Black communities often leverage alternative pathways to develop and sustain social capital, such as strong familial bonds, church involvement, and mutual aid networks (25, 36). These mechanisms may compensate for—or even surpass—the contributions typically associated with higher education, reflecting a culturally rich and community-oriented approach to building and maintaining social capital.

Our findings revealed a racial disparity in community contributions and community involvement, with Black respondents demonstrating an advantage in both outcome variables compared to White respondents. In terms of exposure prevalence, Black respondents had lower education than White respondents. Despite this educational disadvantage, Black respondents consistently maintained a social capital advantage over time, suggesting that their social capital remains robust. Although Black individuals have historically faced systemic barriers to education, greater access to and engagement with education can provide additional opportunities to expand social networks, build transferable skills, and increase civic participation (25). Based on this long-standing evidence, the consistent social capital advantage of Black respondents in the face of persistent educational disadvantages and structural barriers would likely be strengthened and bring greater benefits with higher education attainment.

Despite the stark racial differences in social capital observed in this study, it is important to acknowledge that our operational definitions and measures of social capital outcomes (“Contributions to Community” and “Community Involvement”) may not fully capture the breadth and complexity of social capital as described in the literature. Although not formally validated, these scales to measure social capital were developed and tested as part of the MIDUS study. Still, the use of non-validated, self-developed scales to measure social capital presents important limitations, particularly regarding construct validity and comparability across studies.

Social capital is a multifaceted construct shaped by cultural and systemic factors (25). Consequently, comparing social capital between racial groups requires careful consideration of the historical, systemic, and cultural contexts of each population studied. A notable limitation of the study is the limited racial heterogeneity and unbalanced racial group distribution, which may have affected the reliability of interaction terms and limited the precision of subgroup estimates when assessing differences in social capital trajectories over time. Additionally, our mixed-effects statistical model did not include a random slope, which would have allowed for individual-specific variability in the relationship between time (a categorical predictor) and outcomes (community contributions and involvement). This approach was not feasible due to the limited sample size and unbalanced racial group distribution across four levels of educational achievement. Furthermore, our model adjusted for only a limited set of confounders, which may have resulted in residual confounding and affected the interpretation of racial and educational disparities in social capital over time. Consequently, we assumed that the effect of time on outcomes was consistent across groups and chose a random intercept fixed-effects model. Although this assumption simplifies the model, it may oversimplify the relationship, especially for populations with diverse trajectories over time. However, given the lack of statistical significance for time in many instances and the small size of the Black sample, the random intercept model offered a more stable and parsimonious solution for the data.

Our study also had important strengths, including being one of the first to examine race and education longitudinally in relation to contributions and involvement in the community. We used longitudinal data from waves 1–3 of the MIDUS study, which allowed for a vigorous investigation into evolving patterns and trends over decades and strengthened the reliability of our findings. Second (to our knowledge), no previous study has used a single model to examine longitudinal changes of race and education related to community contributions and involvement. We used a single, mixed-effects statistical model with sophisticated statistical and imputation techniques to explore race and education together. This provided deeper insight into the relationships between the two variables and social capital. Third, we used the Disparity Assessment Framework by Ward et al. for assessing health disparities for a more rigorous approach. We found this to be critical for emphasizing the effect of size and visual plots, rather than relying on statistical significance alone. Unlike statistical significance (p-value), which only measures the likelihood that the results are due to chance, effect size and visual plots offer a more practical understanding of the magnitude of disparities in social capital across time, race, and educational level. This focus allows for a clearer interpretation of the real-world impact of disparities, beyond merely identifying whether differences are statistically significant.

Our finding that Black respondents demonstrated higher levels of social capital than their White counterparts align with existing literature. A few examples: two studies found that Black communities often foster strong communal ties and prioritize collective wellbeing, potentially as a response to systemic inequities (25, 36). Previous studies found that personal experiences, such as attending block parties or participating in Black Greek organizations and Black professional societies, were ways through which Black individuals foster sense of community (25, 36). Other studies indicated that personal experiences, cultural norms, historical factors, and community dynamics were pivotal in shaping the expression and impact of social capital within communities (37, 38). These expressions of social capital are less evident in White communities.

Our effect size as well as statistical significance also aligns with existing literature. Although the racial difference in community involvement was not statistically significant, incorporating effect size indicated that Black respondents may have a substantive advantage in certain dimensions of social capital. This confirms the importance of interpreting effect sizes alongside p-values to better understand the real-world impact of disparities (35). Because our study was one of the first to examine race and education longitudinally in relation to community contribution and involvement, as well as the first to do so using a single model, these findings are pioneering and fill critical gaps in literature.

This study advances our understanding of the complex relationships between race, education, and social capital over time. Longitudinal research is essential to understand these dynamics and address health disparities, especially in marginalized groups, which often face systemic barriers that reduce their social capital. Our findings have important implications for addressing social capital disparities—for both Black people and White people—to the benefit of both racial groups.

The racial and educational disparities we identified indicate a continuing need for policies and interventions to address systemic barriers to education and other resources. Future research should explore the mechanisms underlying the racial and educational disparities observed in this study. There is also a need for future studies to replicate these findings using validated and standardized measures of social capital. Other potential areas for further exploration include racial dynamics and culturally specific forms of social capital. The positive manifestations of social capital we observed should be explored and supported within the Black population.

Moreover, larger and more balanced samples are essential to allow for the use of advanced statistical techniques and ensure that findings are representative. Using random slope mixed-effects modeling would provide a more nuanced understanding of how social capital varies across individuals and groups over time. Incorporating qualitative methods would complement these quantitative findings by providing insights into the lived experiences that drive the patterns revealed in our study. In addition, research that fully integrates psychosocial and biological measures would enrich our understanding of how social capital impacts health, offering a more comprehensive and interdisciplinary perspective on the underlying mechanisms (39). Finally, the large effect sizes observed in the non-significant findings suggest that education’s influence on community involvement merits additional investigation.