Cross-Sectional Association Between Social Activity Participation, Age, and Cognition: Evidence from the Health and Retirement Study

Abstract

Objectives:

To examine whether social activity participation moderates the association between age and cognitive function in U.S. adults aged 50 and older.

Methods:

We analyzed data from the 2016 and 2018 waves of the HRS (N = 10,985). Multivariable linear regression models estimated the main and interaction effects of age and social activity participation on cognitive function, adjusting for covariates, and clustering into households.

Results:

The sample was predominantly female (55.6%) and White/Caucasian (80.3%), with an average age of 66 years. Age was negatively associated with cognitive performance (β = −.089), while social activity participation was positively associated (β = .127). A significant interaction (β = .004, p < .05) suggests that the strength of the age–cognition association varies by level of social engagement. Activity-specific analyses showed that charity work, educational activities, and non-religious meetings were the strongest individual predictors of better cognitive scores.

Discussion:

This large, population-based study suggests that greater social activity is associated with higher cognitive performance and may help explain variation the inverse relationship between age and cognition. While causality cannot be inferred, results highlight the relevance of social engagement in understanding cognitive patterns in adults. Future research should examine these findings persist longitudinal designs.

Introduction

Cognitive health is a crucial component of healthy aging, because it enables older adults to maintain independence, engage in daily activities, and sustain a high quality of life (National Institute on Aging, 2024). Age-related differences in cognition often reflected in variations in memory, executive functioning, and processing speed are associated with substantial individual and societal burdens, including increased healthcare utilization and reduced autonomy (Alzheimer’s Association, 2025; Harada et al., 2013; Prince et al., 2013; Salthouse, 2010). As the population ages, identifying factors that support cognitive functioning has become a critical focus of public health research and policy (Kelly et al., 2017; Livingston et al., 2024).

The Cognitive Reserve Hypothesis offers a compelling framework for understanding individual differences in cognitive aging (Stern, 2021; Stern & Barulli, 2019). It posits that cognitively enriching experiences accumulated across the life course such as formal education, complex occupational roles, and sustained engagement in stimulating environments enhance the brain’s efficiency, capacity, and flexibility, enabling individuals to maintain cognitive performance despite age-related neural changes (Cabeza et al., 2002; Coleman et al., 2023; Gamble et al., 2025; Manly et al., 2003; Park & Bischof, 2013; Scarmeas et al., 2003; Schaie, 1984; Wilson et al., 2002). While education and occupation are typically viewed as early- and mid-life proxies of reserve, the theory also recognizes that cognitively stimulating experiences in later life can contribute to or sustain reserve (Cabeza et al., 2016; Stern, 2009). In this context, social engagement has been conceptualized as an ongoing form of cognitive stimulation that may help preserve cognitive function through the maintenance of neural adaptability and compensatory processes (Fratiglioni et al., 2004; Stern, 2021).

Building on this theoretical foundation, engagement in cognitively and socially stimulating activities in later life is increasingly recognized as a pathway to sustaining cognitive performance. Among these, social activity participation defined as regular involvement in structured activities that facilitate interaction with others, such as volunteering, club membership, community group meetings, and educational programs has emerged as a promising behavioral target for cognitive health promotion (Barnes et al., 2004; Bassuk et al., 1999; Fratiglioni et al., 2004; James et al., 2011; Levasseur et al., 2022; Livingston et al., 2024; Zhou et al., 2020). These activities often offer a combination of cognitive stimulation, emotional engagement, and social network maintenance mechanisms believed to underlie their positive association with cognitive functioning (Ertel et al., 2008; Fowler Davis et al., 2024; Li et al., 2025; Waite, 2018; Zimmer et al., 1995). Consistent with the Cognitive Reserve framework, it is plausible that social participation not only supports general cognitive performance but may also moderate age-related differences in cognition by reinforcing adaptive neural processes.

Despite a growing body of international evidence linking social engagement with cognitive health, relatively few studies have examined whether social activity moderates the age–cognition association in U.S. populations (Oh et al., 2021; Sirven & Debrand, 2008; Tomioka et al., 2016). Cultural and structural factors such as differences in social policy, institutional support, and community infrastructure may influence opportunities for social participation and its cognitive effects (Hamlin et al., 2022). A recent meta-analysis by Cunha et al. (2024) identified only two U.S.-based longitudinal studies eligible for quantitative synthesis, both narrowly focused on volunteering. Broader forms of engagement (e.g., clubs, educational activities, community groups) and their potential to mitigate age-related cognitive differences remain underexplored.

To address this gap, the present study examines whether social activity participation moderates the association between age and cognitive performance in a large, nationally representative sample of older adults in the United States. We hypothesize that greater engagement in social activities attenuates the negative association between age and cognition, such that older adults who are more socially active will demonstrate higher cognitive performance than their less engaged peers. We also explore whether specific types of engagement such as club attendance, volunteering, and participation in non-religious organizations are differentially associated with cognitive outcomes. By identifying activity-specific associations and their relative strength, this study contributes to the growing literature on potentially modifiable behavioral strategies that support cognitive resilience and healthy aging.

Methods

Study Design and Population

This study uses a pooled cross-sectional design using data from the 2016 and 2018 waves of the Health and Retirement Study (HRS; N = 10,985). The HRS is funded by the National Institute on Aging. Initiated in 1992, this longitudinal study collects detailed economic, health, and daily activity information from middle-aged and older adults in the United States. Data collection occurs biennially and involves a nationally representative sample of households, with at least one member aged 50 or older. The data is publicly available, de-identified data from the Health and Retirement Study. As such, this study was exempt from institutional review board review under U.S. federal regulations.

Because social activity participation data were collected from only half of the core HRS respondents in each wave, we combined data from the 2016 and 2018 waves to create a single dataset. This approach allowed us to increase the sample size and enhance the robustness of our findings. Each participant was included only once, ensuring the cross-sectional nature of the analytic sample was maintained. In addition, because the HRS includes both focal respondents and their spouses, the final sample could contain multiple eligible individuals from the same household.

In this study, the inclusion criteria encompassed individuals aged over 50 years and participants with no severe cognitive deficits according to the Langa-Weir classification (Langa et al., 2020). The exclusion criteria included individuals with missing data on key independent variables such as age, gender, education, socioeconomic status, work status, physical and mental health conditions, and sensory impairments (vision and hearing), as well as missing data on social activity participation. We applied listwise deletion to maintain consistency in moderation modeling and transparency in analytic reporting. A detailed flow chart of the sample selection and exclusions is provided in Figure 1.

Figure 1.

Participant selection flowchart for 2016 and 2018 age and cognitive function analysis.

Note. Both focal respondents and spouses were retained.

Household-level clustering was addressed in all regression models using cluster-robust standard errors.

Measures

Cognitive Function

The dependent variable, cognitive function, was assessed using the Langa-Weir Classification of Cognitive Function, a validated scale developed from core HRS data (Crimmins et al., 2011). This classification combines performance-based cognitive tests and proxy assessments administered via face-to-face or telephone interviews. The 27-point scale includes the following components: immediate and delayed 10-noun free recall (memory; 0–20 points), serial sevens subtraction (working memory; 0–5 points), and counting backward (processing speed; 0–2 points; Langa et al., 2020). This composite measure has demonstrated high internal consistency (Cronbach’s alpha = .83). The continuous cognitive score was used in all regression models.

Social Activity Participation

Social activity participation was measured using the SoPart-10 scale, a validated index developed for use with HRS data (Howrey & Hand, 2018; Koopman et al., 2022). This scale includes five social activities: volunteering with children or young people, attending an educational or training course, participating in charity work, attending social clubs (e.g., sport or hobby groups), and involvement in non-religious organizations. Each item was scored from 0 to 2, where 0 = never, 1 = monthly to once a week, and 2 = several times a week. Scores were summed to produce a continuous composite ranging from 0 to 10, with higher values indicating more frequent and diverse engagement (Cronbach’s alpha = .74). Example items include: “How often do you do any volunteer or charity work?” and “How often do you go to a sport, social, or other club?” For descriptive purposes, social activity participation was also categorized as low, medium, or high. The SoPart-10 scale has demonstrated favorable psychometric properties in graded response modeling and Mokken scale analysis, supporting its use in population-based studies to assess modifiable behavioral engagement (Howrey & Hand, 2019; Koopman et al., 2022).

Covariates

We adjusted for a range of sociodemographic and health-related covariates, including age, gender, education, race/ethnicity, marital status, household income, and employment status. Health-related variables encompassed limitations in activities of daily living (ADL) and instrumental activities of daily living (IADL), number of chronic conditions (arthritis, cancer, diabetes, heart disease, hypertension, lung disease, and stroke), depressive symptoms (assessed using the CES-D 8-item scale), hearing impairment, self-rated health (excellent/very good, good, fair/poor), physical activity level, smoking status, and alcohol use. These variables were coded following established HRS protocols (Lee et al., 2011; Steffick, 2000; Turvey et al., 1999; Lahiri & Li, 2020; Sonnega et al., 2014) to ensure comparability with existing research.

Statistical Analyses

Descriptive statistics and survey-weighted estimates were computed in SAS 9.4 using PROC SURVEYFREQ and PROC SURVEYMEANS, which account for the HRS’s stratified multistage design. We applied respondent-level weights suitable for pooled cross-sectional analysis (rwtresp), along with stratification (raestrat) and clustering (raehsamp) variables, as recommended in HRS documentation. Taylor series linearization was used for variance estimation. Given that some households contributed more than one eligible participant (e.g., focal respondent and spouse), we accounted for within-household correlation by estimating random-effects linear regression models (xtreg, re in Stata), with household ID specified as the level-2 unit. This approach models shared variance at the household level and addresses non-independence of observations.

Multivariable linear regression analyses were conducted in Stata/BE 17.0. To examine whether social activity moderated the association between age and cognitive function, we estimated a random-effects model using xtreg, re, with household ID (hhidnum) as the grouping variable. The primary model included a mean-centered interaction term between age and social activity participation. All continuous predictors were centered prior to analysis to facilitate interpretation and reduce multicollinearity.

Regression assumptions were assessed via standard diagnostics. Variance inflation factors (VIFs) were below 5 for most predictors, with the exception of the three self-rated health variables (e.g., self-rated health excellent/very good = 655), retained due to theoretical relevance. Residuals approximated normality based on visual inspection of histograms and kernel density plots. Linearity and homoscedasticity were confirmed using residual-versus-fitted plots.

Results

Demographic and Health Characteristics

The final analytic sample included 10,985 older adults with a mean age of 66 years. Most were female (55.6%) and identified as White (80.3%), followed by Black (7.5%), Hispanic (7.2%), and Other (4.6%). Participants had, on average, 13.75 years of education. The majority were married or partnered (68.1%), and 44.9% were currently working for pay. The average number of chronic conditions was approximately 2, with 13.3% reporting difficulty with at least one activity of daily living (ADL) and 9.6% reporting difficulty with instrumental activities of daily living (IADL). About 34.6% had hearing impairment and 42.3% had vision impairment. The average depressive symptom score was 1.17 on the CES-D 8 scale. Approximately 45.1% of respondents rated their health as excellent or very good, and 38.7% reported high levels of physical activity (Refer to Table 1).

Table 1.

Social Activity Participation Levels by Socio-Demographic and Health Factors (n = 10,985).

Variable	Total	Level of social activity participation
Socio-demographics:		Low (n = 2,729)	Medium (n = 4,404)	High (n = 3,852)
Age (years), M ± SE	66.2 ± 0.18	69.4 ± 11.0	68.3 ± 10.4	66.3 ± 9.8
Education, years, M ± SE	13.75 ± 0.05	11.9 ± 3.0	13.1 ± 2.7	14.4 ± 2.6
Female (%)	55.6	26.2	39.2	34.5
Race/ethnicity (%)
White/Caucasian	80.3	22.5	41.1	36.2
Black/African American	7.5	25.2	38.1	36.6
Hispanic	7.2	38.1	37.9	23.9
Other	4.6	4.6	3.6	4.3
Marital status (%)
Married	68.1	22.0	40.2	37.0
Separated, divorced, or never married	19.8	27.4	39.0	33.5
Widowed	11.9	33.2	40.0	26.6
Income, M ± SD	5.83 ± 0.02	5.1 ± 1.2	5.6 ± 1.2	5.9 ± 1.1
Currently working for pay (%)	44.9	15.3	39.9	44.7
Health and functional status	15.3	41.8	36.5	21.6
Difficulty with an ADL (%)	13.3	26.8	15.0	20.6
Difficulty with an IADL (%)	9.6	46.0	36.9	17.0
Number of chronic conditions (0–8), M ± SE	2 ± 0.21	18.6	41.3	40.0
Depressive symptoms, M ± SD	1.17 ± 0.02	10.4	39.2	50.2
Hearing deficit (%)	34.6	23.37	38.16	49.12
Vision deficit (%)	42.3	29.4	40.4	30.1
Self-rated health status (%)
Excellent/very good	45.1	16.2	38.8	44.8
Good	33.9	22.8	43.4	33.6
Poor	20.8	41.7	37.1	21.1
Level of physical activity (%)
Low	39.5	21.9	38.0	39.9
Moderate	21.6	25.0	41.9	33.0
High	38.7	41.7	37.1	21.1
Smoking status (%)
Never	47	21.9	38.0	39.9
Past smoker	41.8	25.0	41.9	33.0
Current smoker	11	37.8	41.2	20.9
Alcohol use (%)
Abstainer	34.9	33.0	38.0	28.8
Past	18.9	23.9	39.9	36.1
Moderate	37.7	16.7	41.6	41.6
Heavy	8.3	21.8	43.9	34.2

Note. ADL = activities of daily living; IADL = instrumental activities of daily living.

Social Activity Participation

Table 1 displays the distribution of social activity participation by sociodemographic and health characteristics. Individuals with high levels of social activity were younger, more educated, and more likely to be married, employed, and in better health. A graded pattern was observed across levels of engagement, with high participation associated with fewer functional limitations, fewer chronic conditions, lower depressive symptoms, and more favorable health behaviors.

Multivariate Analysis and Moderation Effects

Table 2 presents multivariable linear regression model predicting cognitive performance. Older age was significantly associated with lower cognitive scores (β = –0.089, p < .001), whereas higher social activity participation was associated with better cognition (β = .127, p < .001). A significant interaction between age and social activity (β = .004, p = .001) indicated that the negative effect of age on cognition was weaker for individuals with higher social engagement. Standardized estimates for age, social activity, and their interaction in to facilitate comparison of effect sizes across predictors.

Table 2.

Multivariate Analysis of Factors Influencing Cognition (n = 10,985).

Variables	β	95% CI
Constant	10.891***	[7.994, 13.787]
Age	−.089***	[−0.097, −0.082]
Social activity participation	.127***	[0.099, 0.156]
Age × social activities	.004**	[0.002, 0.007]
Female	.785***	[0.668, 0.901]
Education, years	.289***	[0.266, 0.312]
Race (ref: White)
Black	−1.884***	[−2.047, −1.722]
Hispanic	−.905***	[−1.102, −0.708]
Other	−1.066***	[−1.350, −0.783]
Marital status (ref: Separated, divorced, never married)
Married or partnered	−.107	[−0.260, 0.047]
Widowed	−.130	[−0.305, 0.046]
Average household income	.294***	[0.237, 0.351]
Currently working for pay	.095	[−0.046, 0.237]
Difficulty with an ADL	−.073	[−0.252, 0.105]
Difficulty with an IADL	−.699***	[−0.893, −0.505]
Number of chronic conditions	−.04	[−0.085, 0.004]
Depressive symptoms	−.096***	[−0.130, −0.062]
Hearing loss	−.102***	[−0.160, −0.044]
Vision loss	−.142***	[−0.208, −0.075]
Self-rated health status (ref: poor/fair)
Good	.325	[−2.535, 3.186]
Excellent	.28	[−2.580, 3.141]
Level of physical activity (ref: low)
Moderate	.213***	[0.058, 0.367]
High	.065	[−0.070, 0.200]
Smoke status (ref: never)
Past smoker	−.114	[−0.236, 0.008]
Current smoker	−.586***	[−0.779, −0.393]
Alcohol use (ref: heavy)
Abstainer	.465***	[0.309, 0.623]
Past	.416***	[0.278, 0.555]
Moderate	.475***	[0.243, 0.709]

Note. Significant values are indicated with an asterisk (*p < .05, **p < .01, ***p < .001). CI = confidence intervals; ADL = activities of daily living; IADL = instrumental activities of daily living.

Several covariates were also significantly associated with cognition. Higher education (β = .289, p < .001) and income (β = .294, p < .001) were linked to better cognitive performance. Being female (β = .785, p < .001) was positively associated with cognition. Conversely, racial and ethnic minority status was associated with lower scores (e.g., Black: β = –1.884, p < .001), as were IADL difficulty, depressive symptoms, hearing and vision loss, and current smoking.

Exploratory Analyses

Lastly, as part of our exploratory analysis the item-level models were conducted to identify which social activities were most strongly associated with cognition. Attending education or training courses (β = .073, 95% CI [0.056, 0.090], p < .001, sr² ≈ 0.006), charity work (β = .076, 95% CI [0.059, 0.092], p < .001, sr² ≈ 0.007), and participation in non-religious organizations (β = .047, 95% CI [0.030, 0.063], p < .001, sr² ≈ 0.003) were consistently significant predictors of higher cognitive scores. Volunteering with younger adults (β = .040, 95% CI [0.024, 0.056], p < .001, sr² ≈ 0.002) and attending sports clubs (β = .032, 95% CI [0.015, 0.049], p < .001, sr² ≈ 0.001) also showed significant associations, though of smaller magnitude. These findings are detailed in Supplemental Tables 1–6 and Supplemental Figure 1.

Figure 2 displays the interaction between age and composite social activity participation, based on the full regression model. Higher levels of social engagement are associated with attenuated age-related differences in cognition, consistent with the hypothesized moderation effect.

Figure 2.

Visualizing the moderating effect of social activity participation in 3D.

Note: Predicted cognition scores across age and levels of composite social activity participation, based on the full regression model including the interaction term (age × composite activity score).

Discussion

In this first large-scale, cross-sectional study of middle-aged and older adults in the U.S., we found that social participation significantly moderated the relationship between age and cognitive function. Specifically, greater social activity was linked to a weaker negative association between age and cognition. A one-standard-deviation increase in social participation (about 2.18 points on the SoPart-10 scale) corresponded to a 0.277-point higher cognitive score, equivalent to the difference observed over approximately 3.1 years of aging in this sample. While these differences do not indicate causality, they highlight a cross-sectional pattern of variation that could have public health implications. This highlights that even small increases in social participation may help support cognitive health in older adults, with important implications for maintaining independence and quality of life (Kelly et al., 2017).

Our findings build on a robust body of evidence linking social participation to cognitive health in later life (Lydon et al., 2022), while advancing this literature in key ways. Numerous cross-sectional and longitudinal studies have shown that greater social engagement is associated with better cognitive outcomes. For example, Clare et al. (2017), in a neuroimaging study of 67 community-dwelling older adults, found that higher engagement in social and intellectual activities was associated with better cognitive performance and greater global brain volume (r = .37, p < .01). Zhou et al. (2020), analyzing data from 7,973 Chinese adults aged 45 and older, reported that regular participation in social activities (≥1/week) was significantly associated with improved cognitive scores (β = .604, p < .001). Longitudinally, Barnes et al. (2004) found that greater social engagement was associated with a slower rate of cognitive decline over 5.3 years in 3,899 older adults from the Chicago Health and Aging Project. Similarly, Krueger et al. (2009) reported that frequent participation in social activities was positively associated with baseline global cognition (β = .160, p < .001) in 838 older adults without dementia. However, none of these studies directly tested whether the association between age and cognition differs by levels of social engagement. In addition, our exploratory analyses, the strongest associations were observed for education or training courses, charity work, and participation in non-religious organizations, aligning with prior evidence that such activities foster mental engagement, purpose, and social connection (Bae et al., 2019; Guiney & Machado, 2018; Hsieh et al., 2025; Nie et al., 2021; C. Zhang et al., 2022; J. Zhang et al., 2024). More research is needed to compare activity domains across and clarify which forms of participation may be most influential for cognitive health.

Our study addresses this gap by demonstrating a significant moderating effect of social participation on the age–cognition relationship in a large, nationally representative sample of U.S. adults. Specifically, we found that the negative association between age and cognitive performance was attenuated among individuals who reported higher levels of social engagement. This moderation effect was reflected in a 0.004-point reduction in cognitive decline per year of age among those with greater social participation. While the mechanisms underlying this effect remain to be fully understood, they may involve cognitive reserve, stress-buffering pathways, and enhanced social and emotional support (Kelly et al., 2017; Stern & Barulli, 2019).

From a Public health perspective, our results underscore the value of developing social participation opportunities for older adults (C. Zhang et al., 2022). Community-based programs, such as group activities, clubs, or events could be particularly beneficial in supporting cognitive health. Policymakers might consider strategies to reduce barriers to social participation, including funding for community programs and public awareness campaigns about the benefits of social engagement. For healthcare providers, social participation should be considered as part of a holistic care plan, encouraging older adults to remain engaged in social activities where feasible.

The findings should be interpreted considering certain limitations. First, the cross-sectional design precludes conclusions about causality or directionality. While causality cannot be inferred, this significant moderation effect suggests a meaningful cross-sectional difference that could be tested in more rigorous design and future longitudinal studies. A strength of this work is the use of a large, nationally representative sample, including minority populations, which supports the generalizability of the findings. However, self-reported data on social activity participation may be subject to recall bias, though this is likely minimal given the short recall window and use of dichotomous variables. Finally, while we used a validated index of social activity participation, this prevents us from disentangling the unique contributions of each social activity beyond those explored separately in the Supplemental Tables.

Conclusion

In conclusion, this study provides new insights into the moderating role of social activity participation in cognitive health among older adults. Longitudinal research is needed to determine the direction and causality of these associations. Such work will help clarify whether social engagement can help preserve cognitive health, and if so, what types of social activities are most effective. Considering these findings, public health efforts should prioritize accessible, culturally appropriate social participation programs that foster engagement in later life. Future research should build on these cross-sectional insights to inform interventions that promote social connectedness and cognitive health at the community and policy levels.