Social Participation in Later Years: The Role of Driving Mobility

Abstract

Objectives

I investigate the role of driving mobility for older adults’ formal and informal social participation. I expand the common driving status dichotomy using gradated driving frequency, driving change, and ride receipt to account for the complexity of driving behaviors in later years.

Method

I estimate logistic regression models using the 2011 and 2013 waves of the National Health and Aging Trends Study on a nationally representative sample of 4,359 community-dwelling older adults. I adjust models for demographic, socioeconomic, health, and social activity factors.

Results

Frequent drivers are most likely to visit friends and family, go out for enjoyment, attend religious services, and participate in organized activities compared with occasional drivers, those who ceased driving, and those who never drove. Driving frequency decrease lowers social participation. Participation does not differ between those who ceased driving and those who never drove. Persons with consistent ride access participate more than those never receiving rides. Models using a measure of driving mobility fit data better than models using dichotomous driving status.

Discussion

Both driving frequency and ride receipt matter for older adults’ formal and informal involvement. Facilitating ride-giving and developing flexible transportation options may enhance social participation among older adults who cease or begin ceasing to drive.

The U.S. population is aging, with older persons experiencing increases in both the length and quality of life (Sierra, Hadley, Suzman, & Hodes, 2009). With longer active life expectancy, older adults can remain socially engaged in later years (Annear et al., 2014). Social participation in late life is associated with physical and mental health benefits, such as better self-rated health, lower mortality risk over time (Adams, Leibbrandt, & Moon, 2011), lower rates of depression, dementia, and cognitive impairments (Hao, 2008) and is linked to greater life satisfaction, self-esteem, sense of agency, and positive affect (Adams et al. 2011; Anderson et al. 2014). For these reasons, maintaining “participation in social, economic, cultural, spiritual and civic affairs” (World Health Organization, 2002, p. 12) is part of healthy, active, and successful aging frameworks (McLaughlin, Connell, Heeringa, Li, & Roberts, 2010; Rowe & Kahn, 1997). Transportation mobility—frequently synonymous with driving in car-dependent Western societies—is often crucial for continued social participation and access to services (Fristedt, Dahl, Wretstrand, Björklund, & Falkmer, 2014). However, biopsychosocial factors in later life eventually contribute to a decision to reduce and cease driving (Dugan & Lee, 2014). This may pose difficulties for older adults wishing to remain engaged in social activities.

Driving is older adults’ predominant mode of transportation in the United States. Approximately 85% of individuals aged 60 years and older currently drive (Federal Highway Administration [FHA], 2011), and they do as much as 70% of all their travel by car (O’Hern & Oxley, 2014). Due to retirement, the vast majority of older adults’ travel is nonwork related, and they travel most frequently and farthest for social and recreational purposes (FHA, 2011). However, the proportion of those still driving decreases with age: in 2011, 93% of older adults aged 60 to 69 held a driver’s license, compared with 60% of those aged 85 or over (FHA, 2011). With life expectancy increasing, the period older adults spend not being able to drive on their own is also lengthening. At 70 years, men have a total life expectancy of 18 years, but a driving life expectancy of 11 years, resulting in a 7-year transportation mobility gap. This gap is longer for women at approximately 10 years. In the United States, over 600,000 older individuals yearly must resort to transportation other than driving (Foley, Heimovitz, Guralnik, & Brock, 2002). In a period of their life where they could still be active social participants, older adults today are therefore likely to experience limited driving mobility.

Studies examining the association between driving and social participation indicate that inability to drive poses restrictions for older adults’ engagement in social life. Older persons who have car access are more socially involved (Fristedt et al., 2014; Helliwell & Putnam, 2004). Driving cessation reduces one’s ability to maintain previous social participation levels (Mezuk & Rebok, 2008; Musselwhite & Shergold, 2013). Losing transportation independence is associated with decreases in shopping and entertainment trips, network size, hours volunteering, and general out-of-home activity engagement (Curl, Stowe, Cooney, & Proulx, 2014; Davey, 2006). These findings have implications for older adults’ health and well-being (Adams et al., 2011; Anderson et al. 2014; Hao, 2008).

However, existing studies exploring these associations understand driving narrowly and do not consider the importance of such mobility may vary by social activity type. First, older adults are a social group undergoing driving cessation (Dugan & Lee, 2014). Even reduced levels of driving, not captured in a driving versus not driving binary, may restrict older persons’ social participation. Older adults are also likely to get rides from others, giving them a level of mobility comparable to own driving (Nocon & Pearson, 2000) in accessing participation venues. Previous research using a dichotomous driving indicator did not evaluate the importance of driving frequency or ride receipt for social participation. Second, previous studies did not consider the role of driving mobility may not be uniformly associated with different types of social participation (Utz, Carr, Nesse, & Wortman, 2002). The selection–optimization–compensation (SOC) model (Baltes & Carstensen, 1996) and socioemotional selectivity theory (SST; Carstensen, Isaacowitz, & Charles, 1999) suggest older adults may focus their attention and resources such as driving on social activities they find most important as they age. Activity importance may therefore moderate the association between driving mobility and social participation. This study addresses both concerns.

Older Adults’ Driving Mobility Is More Complex Than Driving Versus Not Driving

The first contribution of this study is examining whether a construct of driving mobility is prospectively associated with social participation in later years. The commonly used binary of driving versus not driving may be insufficient when considering older adults, as it overlooks the variation in driving frequency among the group, and does not consider getting rides as a crucial alternative or complement to driving.

Prior research typically operationalizes driving frequency dichotomously as driving versus not driving. Yet for most older individuals, driving cessation is a process rather than an abrupt decision. It can be conceptualized on a continuum from initial considerations of giving up auto-mobility, through gradual reductions, to eventual driving activity termination (Dellinger, Sehgal, Sleet, & Barrett-Connor, 2001). Older adults stop driving by limiting driving behaviors: they drive less frequently, shorter distances and to fewer destinations, adapt speed, and restrict trips to familiar roads and particular hours (Baldock, Mathias, McLean, & Berndt, 2006). Using a binary measure of driving versus not driving to examine driving behavior may, especially for older individuals, understate or conceal differences in the social participation among those who do or among those who do not drive. The broad and dichotomous “driver” category encompasses both those who actively drive and those in the process of cessation who already made significant adjustments or reductions to their auto-mobility (Burkhardt, 1999). For drivers in the cessation process, such limitations may restrict access to events not occurring during daytime to social venues that are distant and to happenings that require navigating a complex route. Likewise, the “not driving” category encompasses those who previously drove but quit and may be developing coping strategies or have access to rides as well as those who never drove and may have always relied on alternative transportation options to get to social activity venues (Davey, 2006).

Additionally, prior studies limit the definition of driving capacity to the ability to drive oneself. However, although older individuals may be able to undertake independent travel and drive themselves, they are also embedded in social networks and communities and may receive rides (Kerschner & Rousseau, 2008). When one’s ability to drive themselves diminishes in older adulthood, one can therefore still be driven by others. Public and community transportation are often unsuitable for older individuals who are ceasing or have ceased driving due to infrequent nonpeak travel time service, limited service to nonwork destinations, poor accessibility, and low availability (Hjorthol, 2013; Kostyniuk & Shope, 2003). However, family and other contacts providing rides are a key transportation alternative for older individuals during and post-driving cessation (Nocon & Pearson, 2000). Rides represent flexible, supportive, door-to-door service and include assistance and escorting best meeting older individuals’ changed needs (Metz, 2014). In 2011, providing transportation was the most common form of assistance caregivers performed for older adults, with 89% of caregivers indicating they provided it (US Department of Health and Human Services, 2014). In one study, 33% of older respondents relied on contacts to secure rides for all their needs, whereas the rest combined these with other methods: 66% received rides weekly and 20% almost daily (Davey, 2006). High uptake and the flexible, participation-facilitating nature of ride receipt indicates even older nondrivers may still have suitable means for getting around and may not be as restricted in their travel as the broad “nondriver” label may suggest. This warrants the inclusion of ride receipt as a component of driving mobility.

Driving mobility may therefore be conceptualized as having two parts: one’s own driving frequency and receiving rides. Specifically, driving frequency extends along a continuum of driving frequently, occasionally, having ceased, or never driven (Dellinger et al., 2001), rather than simply driving versus not. Additionally, receiving rides is a crucial alternative transportation option for older adults as their own ability to drive declines (Nocon & Pearson, 2000). A view of driving mobility that encompasses both self-driving and being driven may be more suited to understanding the role of transportation for older adults’ social participation than a dichotomous driving indicator.

Previous qualitative studies (Davey, 2006; Musselwhite & Shergold, 2013) and quantitative studies using a dichotomous measure of driving versus not (Curl et al., 2014) suggest transportation independence loss reduces one’s ability to socially participate. To examine how driving mobility is associated with older individuals’ participation, I therefore test Hypothesis 1: older adults with higher driving mobility—conceptualized as both driving frequency and ride receipt—are more likely to participate than their less driving mobile counterparts.

Driving Mobility May Matter Differently for Informal Versus Formal Participation

A second contribution of this study is exploring whether the association between driving and social participation differs by type and importance of participation considered. Formal social participation refers to group activities performed in organizations and done regularly, such as attending religious services and participating in clubs, classes, and other organized activities. Informal social participation refers to activities characterized by individualized and irregular involvement, such as visiting friends and family and outings for enjoyment (Utz et al., 2002). Older adults may be more strongly motivated to participate in one or the other, both equally, or may prefer neither, and driving mobility could therefore matter differently depending on activity type.

The SOC model posits that older adults attempt to use their resources with maximum efficiency when faced with new conditions accompanying aging (Baltes & Carstensen, 1996). Physical changes and worsened health in later life restrict the range of activities one can engage in (Wylde, Livesey, & Blom, 2012) and shrinking social networks provide fewer opportunities and incentives to partake (Rozanova, Keating, & Eales, 2012). Similarly, poor health reduces older adults’ driving ability as a resource, and fewer contacts may mean fewer opportunities to access rides (Dugan & Lee, 2014). As older adults’ health, economic, and social capital decline, their opportunities for social participation—and their ability to access the venues—therefore narrow. The investments required to maintain previous participation levels may become too costly, and older individuals may choose to focus only on select activities in adapting to their new circumstances. In the model’s terms, they perform a loss-based selection, reprioritizing their goals to optimize available resources, such as driving ability or ride access. They resort to other means to acquire the health, well-being, and social benefits previously gained through social participation.

SST (Carstensen et al., 1999) similarly suggests a narrower time horizon prompts older individuals to become more selective regarding what and whom they invest their resources in. They may therefore spend less time on social participation, become selective of the activities they are involved in, and engage in those they find more rewarding or that require fewer resources. In both frameworks, older adults begin to focus on interactions and activities they find meaningful, rather than attempt to maintain their previous levels of involvement at all costs. As individuals age and their resources diminish, we may therefore expect their participation priorities to vary by type and importance they assign to the activity. Older adults may have fewer incentives to overcome transportation barriers for participation in activities they do not prioritize. Conversely, they may strive to adapt to transportation limitations, wanting to continue personally meaningful activities. Driving mobility may therefore become crucial for those finding a social activity very important but be of lesser relevance for those with a weaker interest in such involvement.

To examine how older adults’ driving mobility matters for participation types, this study differentiates between formal and informal participation and considers activity importance as a potential moderator in addition to its stand-alone motivating role for activity engagement. Drawing on the SOC model (Baltes & Carstensen, 1996) and the SST (Carstensen et al., 1999), I evaluate Hypothesis 2: the importance older individuals assign to a social participation activity heightens the association between driving mobility and activity participation (moderation analyses).

Other Influences on Social Participation and Driving Mobility

Older adults’ social participation and driving mobility also have demographic, socioeconomic status, health status, and activity-related correlates. Men are more likely to engage in the formal sphere, whereas women are more likely to engage informally (Bukov, Maas, & Lampert, 2002). Racial and ethnic minority older adults participate less frequently than whites (Li & Cai, 2014). Married, partnered, and widowed individuals engage in more social activities than those divorced or never married (Utz et al., 2002), and older adults with higher socioeconomic status participate more often (Hodgkin, 2011). Poor health restricts participation (Adams et al., 2011). Likewise, women, black or Hispanic, being married, having lower socioeconomic status, and having physical and mental impairments is linked to higher rates of driving cessation (Dugan & Lee, 2014). To avoid confounding an observed association between driving mobility and social participation, I control for these factors in my analyses.

Method

Data

I base the analyses on the 2011 and 2013 waves of the National Health and Aging Trends Study (NHATS, 2011; Kasper & Freedman, 2014). The NHATS is a longitudinal, nationally representative survey monitoring aging changes and their social implications. The first wave of data was collected in 2011, with participants reinterviewed annually. NHATS uses a stratified three-stage sample (Montaquila, Freedman, Edwards, & Kasper, 2012) of Medicare recipients aged over 65, both community dwelling and those in residential care (96% of U.S. older adults are Medicare enrollees). The baseline response rate was 71%. I apply NHATS Wave 1 weights in all analyses to account for the oversampling of blacks and those aged 85 and older, adjusting for differential selection probabilities and nonresponse. The NHATS is ideally suited for assessing the study hypotheses due to its unique combination of information on driving behaviors and social involvement, and due to its multiple waves of data, which allow a prospective exploration of the effects of driving mobility on older adults’ participation.

I limit the analytic sample to older individuals with complete person interviews at both waves, excluding 22.7% (n = 1,315) of the original NHATS sample (N = 8,245) who died or attrited. I limit it to community-dwelling older individuals (n = 4,484), as they face different transportation needs than those in institutional settings. Missing data were rare, not exceeding 1.2% on any variable included in the final models. I use listwise deletion for cases missing data on any variable included, dropping 3% of the sample (n = 125) and conduct the analyses on a 4,359-case final analytic sample. Chi-square tests and multinomial logistic regression predicting attrition and death versus retention show better baseline health, higher education, and white ethnicity increase retention. The analytic sample is thus overrepresentative of healthy, well-educated, white older adults.

Dependent Variable: Social Participation

I operationalize social participation using 4 items. Two capture informal participation: visiting friends and family not living with the respondent and going out for enjoyment such as to dinner, to a movie, to gamble, to hear music, or see a play (hereafter “outings for enjoyment”). Two capture formal participation: attending religious services and participating in clubs, classes, and other organized activities (hereafter participating in organized activities). All questions refer to the last month, allowing for yes/no responses (“In the last month, did you ever …”). I estimate models predicting social participation at Wave 3, controlling for the same items at Wave 1 to better ascertain causal ordering. Exploratory factor analysis indicated the items capture different dimensions of participation, with zero-order correlations ranging from 0.15 to 0.27. Thus, I predict each outcome separately. Outcomes are dichotomous, indicating whether one participated in the predicted activity (1 = participation).

Key Independent Variables: Driving Mobility

I measure driving mobility with driving frequency at Wave 1 and driving and ride receipt change between Waves 1 and 3. Driving frequency has 5 categories: never driven, ceased driving (reference), drives occasionally (<1 time/week to 4 times/week), and drives frequently (>5 times/week). For comparison with the typical measure, I dichotomize it to driving versus not driving (reference). Driving change consists of driving frequency decrease, no change (reference), and driving increase between the two waves. Ride receipt change refers to whether one lost rides from a family member, friend, or someone paid to help to get to places outside their building, received them at neither wave (reference), consistently received them at both waves, or gained receiving rides between the two waves.

Control Variables

Several factors may account for associations between driving and social participation. Demographic controls include gender (=1 if male), age (measured continuously), race (non-Hispanic white = reference, non-Hispanic black, Hispanic, or other), and marital status (married/partnered = reference, separated/divorced, widowed, or never married). Socioeconomic status controls include education (measured as highest level completed; less than high school = reference, high school, some college, or bachelor’s degree or higher), and homeownership (=1 if yes). Health controls include self-rated health (=1 if the self-rated health was very good or excellent), depression risk (measured using Patient Health Questionnaire 2 (PHQ-2), =1 if at risk for depression), and anxiety risk (measured using Generalized Anxiety Disorder scale 2 (GAD-2), =1 if at risk for anxiety). The PHQ-2 and GAD-2 are common, validated 2-item depression and anxiety screeners that consist of two questions reflecting DSM-V core diagnostic criteria; I recode them in accordance with cutoff guidelines (Löwe et al., 2010). Activity controls include previous participation (=1 if respondent participated at Wave 1) and activity importance (=1 if respondent rated the activity as very important).

Analytic Plan

I use Stata 14 to estimate logistic regression models predicting social participation at Wave 3. I employ sequential additions of Wave 1 demographic, socioeconomic, health, and activity controls to examine model fit and changes in the strength of association between driving mobility and social participation, addressing Hypothesis 1. I interact activity importance with driving mobility measures to examine moderation effects on social participation, addressing Hypothesis 2. I run all models using both the driving mobility construct and a dichotomous driving measure. I use the Akaike Information Criterion (AIC; Burnham & Anderson, 2004) to examine whether model fit was superior when using the multi-category versus dichotomous driving mobility measures in predicting participation. Multicollinearity was not problematic for estimates, with variance inflation factors not exceeding 1.7 for any variable. Tables 2 and 3 present the obtained estimates as odds ratios (ORs).

Table 2.

Binary logistic regression results predicting older adults’ informal and formal social participation at Wave 3

		Informal Social Participation				Formal Social Participation
	Outcome Predicted; =1 if yes, =0 if no.	Visited Friends and Family		Went Out for Enjoyment		Attended Religious Services		Participated in Org. Activities
		eb	SE	eb	SE	eb	SE	eb	SE
Driving mobility	Driving frequency (ref = ceased driving)
	Never drove	1.59*	0.36	0.99	0.20	1.04	0.30	0.77	0.19
	Drives occasionally	1.42	0.28	1.39	0.24	1.31	0.27	1.18	0.22
	Drives frequently	3.45***	0.73	2.93***	0.53	2.02***	0.42	2.58***	0.46
	Driving change (ref = no change)
	Driving frequency decrease	0.61***	0.09	0.67**	0.09	0.60***	0.08	0.54***	0.58
	Driving frequency increase	1.38	0.30	1.72**	0.32	1.08	0.18	1.15	0.18
	Ride receipt (ref = never received rides)
	Lost ride receipt	1.26	0.26	0.91	0.16	1.46	0.28	1.01	0.16
	Always received rides	1.72**	0.29	1.57**	0.23	0.97	0.14	1.36*	0.17
	Gained ride receipt	1.53**	0.29	1.10	0.17	1.04	0.15	1.04	0.13
Demographics	Male (=1 if yes)	0.76*	0.10	0.97	0.11	0.73*	0.10	0.67***	0.07
	Race (ref = white)
	Black	0.81	0.12	0.63***	0.08	1.38*	0.21	0.79*	0.93
	Hispanic	0.46***	0.09	0.73	0.15	1.21	0.31	0.39***	0.10
	Other	1.10	0.37	0.85	0.23	1.38	0.48	0.47**	0.13
	Age	0.98*	0.01	0.98**	0.01	0.98*	0.01	0.98*	0.01
	Marital status (ref = married/partnered)
	Divorced or separated	0.94	0.18	0.83	0.14	0.64*	0.12	0.87	0.13
	Widowed	0.82	0.12	0.82	0.11	0.91	0.13	0.97	0.11
	Never married	0.46**	0.14	0.61	0.17	1.11	0.38	1.39	0.35
SES controls	Education (ref = less than high school)
	High school	1.17	0.19	1.49**	0.20	1.64**	0.25	1.43**	0.19
	Some college	1.33	0.23	1.60**	0.23	1.59**	0.26	1.81***	0.25
	BA or higher	1.30	0.24	2.68***	0.44	1.41*	0.22	2.24***	0.32
	Homeownership (=1 if yes)	1.26	0.19	1.24	0.16	1.02	0.15	0.97	0.12
Health	Above average self-rated (=1 if yes)	1.08	0.14	1.37**	0.15	1.07	0.12	1.10	0.10
	At risk for depression (=1 if yes)	0.75	0.13	0.89	0.13	0.76	0.13	0.99	0.15
	At risk for anxiety (=1 if yes)	0.68*	0.12	0.88	0.14	0.96	0.19	0.79	0.13
Activity	Activity is very important (=1 if yes)	1.67***	0.20	1.95***	0.21	6.02***	0.69	2.31***	0.25
	Participated at first wave (=1 if yes)	3.93***	0.55	5.60***	0.62	14.60***	1.66	5.39***	0.57
	Constant	3.83	3.00	1.63	1.11	0.28	0.19	0.33	0.20
	Wald χ2	428.32*** (25)		859.52*** (25)		1261.28*** (25)		889.19*** (25)
	Pseudo log-likelihood	−5795900.3		−7309163.2		−7670457.8		−10056653
	McFadden’s R2	0.17		0.27		0.47		0.29

Note. N = 4,359. OR = odds ratio; SES = socioeconomic status. Exponentiated coefficients (ORs) and robust standard errors are provided.

*p < .05. **p < .01. ***p < .001.

Source. NHATS 2011–2013.

Table 3.

Summary of Logistic Regression Estimates on Ride Receipt and Driving Frequency, Predicting Older Adults’ Social Participation (Wave 3) Using a Dichotomous Driving Frequency Measure Versus Using a Measure of Driving Mobility Encompassing Nuanced Driving Frequency and Ride Receipt

Informal participation	Model	Visited Friends and Family				Went Out for Enjoyment
		Dichotomous Driving		Driving Mobility		Dichotomous Driving		Driving Mobility
		eb	SE	eb	SE	eb	SE	eb	SE
	Drives	1.75**	0.29			1.75***	0.24
	Never driven			1.59*	0.36			0.99	0.20
	Drives occasionally			1.42	0.28			1.39	0.24
	Drives frequently			3.45***	0.73			2.93***	0.53
	Driving decrease	0.68**	0.10	0.61***	0.09	0.75*	0.09	0.67**	0.09
	Driving increase	0.99	0.21	1.38	0.30	1.26	0.22	1.72**	0.32
	Lost ride receipt			1.26	0.26			0.91	0.16
	Always received rides			1.72**	0.29			1.57**	0.23
	Gained ride receipt			1.53**	0.29			1.10	0.17
	Wald χ2 (df); AIC	414.38 (20)***; 2852.44		428.32 (25)***; 2827.96		843.26 (20)***; 3603.66		859.52 (25)***; 3582.35
Formal participation	Model	Attended Religious Services				Participated in Organized Activities
		Dichotomous Driving		Driving Mobility		Dichotomous Driving		Driving Mobility
		eb	SE	eb	SE	eb	SE	eb	SE
	Drives	1.77**	0.30			1.82***	0.26
	Never driven			1.04	0.30			0.77	0.19
	Drives occasionally			1.31	0.27			1.18	0.22
	Drives frequently			2.02***	0.42			2.58***	0.46
	Driving decrease	0.60***	0.08	0.60***	0.08	0.59***	0.06	0.54***	0.58
	Driving increase	0.97	0.16	1.08	0.18	0.90	0.13	1.15	0.18
	Lost ride receipt			1.46	0.28			1.01	0.16
	Always received rides			0.97	0.14			1.36*	0.17
	Gained ride receipt			1.04	0.15			1.04	0.13
	Wald χ2 (df); AIC	1259.71 (20)***; 3301.14		1261.28 (25)***; 3290.96		890.81 (20)***; 4180.94		889.19 (25)***; 4141.59

Note. N = 4,359. OR = odds ratio; df= degree of freedom; AIC = Akaike Information Criterion.ORs and robust standard errors provided. The reference category (OR = 1) for driving frequency in dichotomous driving frequency models is doesn’t drive and ceased driving in driving mobility models. The reference for driving frequency change is no change. The reference for ride receipt change is never received rides. Model with the smaller AIC is preferable, with ΔAIC (AIC_max − AIC_min) > 10 indicating very strong, 6–10 indicating strong, and 2–6 indicating positive preference for the model.

*p < .05. **p < .01; ***p < .001.

Source. NHATS 2011–2013.

Results

Descriptive Statistics

Table 1 shows descriptive statistics for the weighted (unweighted n = 4,359) analytic sample.

Table 1.

Weighted Sample Descriptive Statistics

Variable	%	Mean	SD	Min	Max
Dependent variables: Social participation (Wave 3)
Informal social participation
Visited friends and family (=1 if yes)	89.60
Went out for enjoyment (=1 if yes)	81.61
Formal social participation
Attended religious services (=1 if yes)	56.46
Participated in organized activities (=1 if yes)	39.99
Key independent variables: Driving mobility (Wave 1, change to Wave 3)
Ride receipt
Lost ride receipt W1–W3	11.08
Never received rides (reference)	38.81
Always received rides	32.86
Gained ride receipt W1–W3	17.25
Driving frequency
Never drove	4.65
Ceased driving (reference)	11.82
Drives occasionally (<1 time/week to 4 times/week)	17.49
Drives frequently (5 times/week to daily)	66.05
Driving frequency change
Driving frequency decrease	27.01
No driving frequency change (reference)	61.78
Driving frequency increase	11.21
Control variables (Wave 1)
Demographic variables
Male (=1 if yes)	44.16
Race
White (reference)	82.89
Black	7.85
Hispanic	6.14
Other	3.12
Age		78.72	7.32	67	104
Marital status
Married or partnered (reference)	60.44
Divorced or separated	11.76
Widowed	24.67
Never married	3.13
Socioeconomic status variables
Education
Less than high school (reference)	19.58
High school	26,48
Some college	26.18
Bachelor’s degree or higher	27.76
Homeowner (=1 if yes)	80.92
Health variables (=1 if yes)
Self-reported health “above average” or “excellent”	49.04
Fell in the past month	9.04
At risk for depression (PHQ-2)	11.98
At risk for anxiety (GAD-2)	10.61
Activity variables
Activity is very important (=1 if yes)
Visiting friends and family	65.36
Going out for enjoyment	48.27
Attending religious services	52.96
Participating in organized activities	32.86
Participation at Wave 1 (=1 if yes)
Visited friends and family	89.51
Went out for enjoyment	82.10
Attended religious services	58.87
Participated in organized activities	41.83

Note. Unweighted N = 4,359. PHQ-2 = Patient Health Questionnaire 2; GAD-2 = Generalized Anxiety Disorder scale; NHATS = National Health and Aging Trends Study.

Source: NHATS 2011–2013.

At Wave 3, the majority of the respondents visited friends and family not living with them (90%) and went out for enjoyment (82%) in the past month. More than half attended religious services (56%). Fewer participated in clubs, classes, or other organized activities (40%). These levels of participation appear stable from Wave 1.

Most respondents (66%) drove frequently, over 5 times/week, at Wave 1. Fewer (17%) drove occasionally, or <1 times/week to 4 times/week. Twelve percent previously drove but ceased and only 5% had never driven. By Wave 3, the majority (62%) experienced no change in driving frequency, whereas 27% decreased and 11% increased it. Between the two waves, 11% lost ride receipt from family, friends, or paid help and 17% gained it. Others experienced no change, with 39% receiving rides at neither and 33% receiving rides at both waves. For the majority, visiting friends and family (65%) and attending religious services (53%) were very important, whereas fewer rated outings for enjoyment (48%) or participating in organized activities (33%) as such.

Multivariate Logistic Regression

Table 2 shows ORs from fully adjusted binary logistic regression models predicting older adults’ social participation at Wave 3. Table 3 summarizes logistic regression estimates predicting social participation operationalizing driving mobility two ways: as dichotomous driving frequency and using driving mobility conceptualized as multi-categorical driving frequency, driving change, and ride receipt change.

Driving mobility—as both self-driving and receiving rides—matters for social participation.

Examining the association between driving mobility and social participation, three key findings emerge. Regarding driving oneself, higher driving frequency increases odds of social participation across all predicted outcomes. A decrease in driving frequency lowers them. Regarding being driven, receiving rides at both waves increases the odds of social participation for older adults.

For all four outcomes, frequent drivers have higher odds of partaking in the activity compared with their counterparts who ceased driving. In fully adjusted models (Table 2), frequent drivers have odds 3.45 times as high (p < .001) of visiting friends and family, 2.93 times as high (p < .001) of outings for enjoyment, 2.02 times as high (p < .001) of attending religious services, and 2.58 times as high (p < 0.001) of organized activity participation as those who no longer drive. Occasional drivers do not differ from the latter with respect to participation. Similarly, those who never drove are not distinct from peers who underwent the process of driving cessation, with the exception of visiting friends and family, where those who never drove have odds 1.59 times as high (p < .05) of the outcome as those who ceased driving.

Regardless of initial driving frequency, the odds of social participation for those who reduced their amount of driving, compared with those whose driving frequency did not change between the two waves, were lower across all outcomes. Older adults who began to cease driving between the two waves have 39% lower odds (OR = 0.61, p < .001) of visiting friends and family, 24% lower odds (OR =0.76, p < .01) of going out for enjoyment, 40% lower odds (OR = 0.60, p < .001) of attending religious services, and 46% lower odds (OR = 0.54, p < .001) of participating in organized activities as those who maintained their driving frequency. Conversely, an increase in driving frequency only increased the likelihood of going out for enjoyment: compared with those who maintained driving levels, those who increased them have odds 1.72 times as high (p < .01) of such participation.

Besides driving oneself, being driven increases the odds of social participation for older adults. In full models (Table 2), older adults who consistently (in both waves) received rides from family members, friends, or paid help have odds 1.72 times as high (p < .01) of visiting friends and family, 1.57 times as high (p < .01) of going out for enjoyment, and 1.36 times as high (p < .05) of participating in organized activities as those never receiving such assistance. Additionally, those who gained ride receipt from Wave 1 to Wave 3 have odds 1.53 times as high (p < .01) of visiting friends and family, compared with ride non-recipients. Those who lost ride receipt between the two waves do not appear to differ in their social participation from those never receiving rides.

Finally, compared with a dichotomous measure of driving versus not driving, using the driving mobility measure (encompassing both driving frequency and ride receipt) in predicting older adults’ participation shows fit improvements across the models (Table 3). The AIC (favoring the smaller value model) provides very strong evidence (ΔAIC > 10) in support of using the driving mobility measure in all unadjusted models (data not shown) and in all four fully adjusted ones: visiting friends and family (ΔAIC = 24.48), outings for enjoyment (ΔAIC = 21.31), attending religious services (ΔAIC = 10.18), and organized activity participation (ΔAIC = 39.35). Additionally, the differences in effect magnitude between the dichotomous and driving mobility variables are large and indicate what the simpler measure masks. It over-captures the effects for active drivers, conceals differences between frequent and occasional drivers, and disregards the influence of receiving rides.

Results therefore support Hypothesis 1, showing older adults with higher driving mobility—conceptualized as both driving frequency and ride receipt—are more likely to socially participate socially than their less driving mobile counterparts. Additionally, the construct of driving mobility reveals nuances in the relative importance of driving behaviors for older individuals’ social participation that a simple dichotomous measure of driving frequency masks.

Activity importance is key across outcomes, but does not heighten the association between driving mobility and activity participation

Examining the role of activity importance for older adults’ social participation, three findings emerge. Results show activity importance is a strong explanatory mechanism but does not moderate the association between driving mobility and social participation. Additionally, along with race, age, and education, it is significant across all four activities predicted but not more strongly associated with either formal or informal participation.

Across final models (Table 2), activity importance is a strong predictor of one’s participation in the activity in question. Those considering visiting friends and family very important have odds 1.67 times as high (p < .001) of doing so, those considering outings for enjoyment very important have odds 1.95 times as high (p < .001) of doing so, those considering attending religious services very important have odds 6.02 times as high (p < .001) of doing so, and those considering organized activity participation very important have odds 2.31 times as high (p < .001) of doing so, compared with peers rating the activity in question less important. Similarly to frequent driving, changes in driving frequency, and ride receipt, activity importance is uniformly significant across all four activities predicted, suggesting neither factor is more strongly associated with informal or formal types of social participation.

However, although it appears a strong explanatory mechanism, activity importance does not moderate the driving mobility–participation association, with one exception. Of the possible interactions between the participation types and driving frequency, driving change, and ride receipt change, only three were statistically significant and remained so after a Bonferroni multiple comparison correction (data not shown). Activity importance moderated the association of driving frequency and driving change with attending religious services. Frequent drivers considering such attendance very important have a 91% likelihood of participating in the activity (95% confidence interval [CI] = 0.90–0.93, p < .001), and occasional drivers have an 88% likelihood of attending religious services (95% CI = 0.85–0.92, p < .001), compared with a 77% likelihood (95% CI = 0.71–0.83, p < .001) for those who ceased driving and also consider the activity very important. Those who began driving more and deem religious participation very important have a 90% likelihood of attending such activities (95% CI = 0.88–0.94, p < .001) compared with a 85% likelihood (95% CI = 0.84–0.87, p < .001) of their peers who experienced no driving frequency change between two waves and deem religious participation important. Therefore, among older persons for whom religious participation is important, those with higher driving mobility are more likely to attend religious services than those with lower driving mobility. However, no other moderation effects were significant, the CIs for the religious participation interaction coefficients were wide (data not shown), and the analysis involved multiple comparisons. These findings should be interpreted cautiously.

As noted, activity importance and driving mobility do not indicate a difference between preferences for formal or informal types of participation. Similarly, demographic and education factors are associated with lower participation likelihood regardless of the type. Older men are less likely than women to participate. Men have 24% lower odds (OR = 0.76, p < .05) of visiting friends and family, 27% lower odds (OR = 0.73, p < .05) of attending religious services, and 33% lower odds (OR = 0.67, p < .001) of participating in organized activities than women. Across all four outcomes, the odds of older adults’ social participation decrease slightly (OR = 0.98, p < .05) with each additional year of age. Black, Hispanic, and other ethnic and racial minority older adults generally have lower odds of participation, with the exception of black older adults’ higher odds (OR = 1.38, p < .01) of attending religious services. Finally, compared with those with less than a high school education, older adults with a high school, some college, BA or higher degree have higher odds of social participation across all outcomes, except for visiting friends and family, for which education does not seem to exert an effect.

Results therefore do not support Hypothesis 2. Activity importance is a strong explanatory factor for older adults’ social participation but does not heighten its association with driving mobility. Additionally, neither driving mobility nor control factors appear to be more strongly associated with formal or informal participation types.

Discussion

This study prospectively explored the association between a construct of driving mobility, encompassing both nuanced driving frequency and ride receipt and social participation in later years. With the U.S. population of those over the age of 65 tripling by 2050 (US Census Bureau, 2014) and driving cessation salient for an increasing number of older adults (Foley et al., 2002), the findings have implications for maintaining the latter’s quality of life and for community vibrancy resulting from social participation in later years.

Examining how driving mobility is associated with social participation, I find both driving oneself and receiving rides increase the likelihood of older adults staying involved. First, using a gradated measure of driving frequency reveals differences in participation between frequent versus other drivers. Frequent drivers participate more than those who ceased driving. Occasional drivers and those who never drove do not differ in participation from peers who no longer drive. Irrespective of initial driving frequency, those who reduce their amount of driving are subsequently less likely to participate compared with counterparts whose driving frequency remained unchanged. Second, including ride receipt as a dimension of driving mobility shows receiving such assistance is crucial for older adults’ involvement regardless of their own driving status. Those consistently receiving rides are more likely than non-recipients to participate, whereas those who lost or gained rides do not differ from peers who never had them. Examining how activity importance along with driving mobility matter for participation types, results indicate importance strongly relates to participation, but with the exception of religious activities does not moderate the association or show a pattern of preference for formal versus informal participation types.

In car-reliant Western societies, finding frequent drivers participate more than those who ceased driving, and that those who reduced their driving levels participate less, is unsurprising and consistent with previous research (Curl et al., 2014). However, prospectively exploring the association between driving mobility and participation and using a nuanced measure of driving frequency, this analysis indicates older adults’ needs may differ not only by whether or not they drive actively but also by their level of driving. Occasional drivers still drive themselves but do not differ in participation from peers who already ceased driving. They may be a transitional group already undergoing the driving cessation process. Cessation behaviors such as limiting the times of day when they drive, or restricting their driving to particular areas (Baldock et al., 2006), may narrow their ability to access participation venues. The SOC model suggests such changes require adaptation: occasional drivers may have to adopt new routines in response to their modified driving situation. As they develop them, they may therefore have different transportation needs compared with those who already ceased driving or never drove. Occasional drivers may benefit most from mobility exclusion prevention efforts (Hjorthol, 2013). This finding shows nuances in driving frequency are significant for older adults’ social participation; the distinctive challenges of transitioning out of actively driving are concealed in studies using the driving/not driving dichotomy, where they are combined with frequent drivers. Further research should examine this group’s obstacles and characteristics.

Additionally, the gradated driving frequency measure reveals older persons who never drove do not differ in their social participation from peers who ceased driving. Individuals who never drove have the lowest participation rates net of controls, with comparably low levels for those who ceased driving. This counters SOC (Baltes & Carstensen, 1996) and SST (Carstensen et al., 1999) predictions, where older adults’ resource reallocation when faced with diminishing driving mobility may change their participation. Older individuals who already ceased driving may have made adaptations approximating the lifestyles of those individuals who never drove. The two groups may therefore have similar experiences with respect to social participation (Burkhardt, 1999). The finding lends partial support for using dichotomous driving frequency in examining older adults who no longer drive themselves. Qualitative research may compare the adaptation processes between those who ceased driving and those who never drove.

Receiving rides from others also increases older adults’ likelihood of social participation. Older adults consistently receiving rides (at both waves) prospectively participate more than those who never received rides. Sustained ride-giving support from significant others may therefore prolong social participation in later life. This finding holds implications for individuals providing transportation assistance (Kerschner & Rousseau, 2008). A widening formal–informal care gap, declining fertility, and high rates of migration may mean older individuals have fewer proximate children and smaller networks willing or able to provide rides. Family and friends may be reluctant to assist due to limited time, money, and competing obligations (Bittman, Hill, & Thomson, 2007). Labor, workplace, and care policies need to adapt to changing demands on ride-givers. Paid care leave, flexible work patterns, care work subsidies, and expanding the care workforce may facilitate providing transportation assistance to older adults (Fine, 2011). This study was unable to control for the willingness and capacity to give rides. Future research should explore demographic and psychosocial factors associated with ride provision.

Additionally, activity importance is a strong explanatory but generally not a moderating factor between driving mobility and participation. The exception is religious participation: among older persons for whom religious participation is important, those with higher self-driving mobility (frequent drivers, occasional drivers, and those who increased their driving frequency) are more likely to attend religious services than those with lower driving mobility (those who ceased driving and those who experienced no driving frequency change). Driving mobility may thus facilitate continued engagement in religious activities when older adults find such participation important. The finding is in line with research suggesting religious involvement remains stable until the end of life (Kelley-Moore & Ferraro, 2001; Wang, Kercher, Huang, & Kosloski, 2014) and is relevant in light of religious participation health and well-being benefits (Idler, McLaughlin, & Kasl, 2009; Wang, Kercher, Huang, & Kosloski, 2014). However, activity importance did not moderate the driving mobility–social activity association for any other outcome that indicates religious participation may be a distinct type of participation compared with other activities (Croezen, Avendano, Burdorf, & Lenthe, 2015). This study focused on social participation broadly and did not control for factors such as religious affiliation. Future research taking into account factors specific to religion should further explore whether driving mobility is uniquely relevant to religious participation.

Additionally, even for religious participation, activity importance did not moderate the association between participation and ride receipt as one aspect of driving mobility. This is consistent with qualitative research showing older adults fear being burdensome to others in asking for transportation assistance. Although activity importance appears a key motivator for older adults to socially participate (Hao, 2008), they are not more likely to keep driving or ask for rides even when they feel strongly about an activity. Older adults hesitate to request trip assistance (Connell, Harmon, Janevic, & Kostyniuk, 2013; Davey, 2006), and when they do ask for help, they request rides for basic needs, like doctor visits and grocery shopping, not for leisure ones (Burkhardt, 1999). They compromise trip purposes and destinations considered nonessential, such as those for social activities (Ahern & Hine, 2012). The finding that gaining ride receipt does not increase older adults’ social participation may be explained in this vein. This study does not compare the role of driving mobility for recreational trips to those for more essential purposes, but the importance of gaining ride receipt may be more pronounced for the latter; older adults may be using rides primarily for basic needs. Further research should explore differences in the aspects of driving mobility for instrumental versus noninstrumental purposes predicted here.

Thus, even strong activity importance as a motivating factor to socially participate does not alter older adults’ driving frequency or ride receiving behavior. This finding points to the need for developing transportation alternatives that are accessible and non-stigmatizing (Metz, 2014). Such options could facilitate continued participation without older adults feeling burdensome or risking their safety with prolonged self-driving (Connell et al., 2013). Some states already used policies to establish personal vehicle sharing standards, fund public transportation infrastructure, and implement sustainable senior transportation (Staplin & Freund, 2013). The resulting community and supplemental transportation programs include community buses and minibuses providing fixed and circulator route rides, Dial-a-Ride services and paratransit point-to-point services for those with disabilities, the provision of taxi concessions, and volunteer driver schemes (Kerschner & Rousseau, 2008; Metz, 2014). However, the provision and availability of these services varies by area and is not yet sufficiently developed to represent a viable mobility alternative for many older adults (Davey, 2006). The awareness of these programs among older individuals is low (Hjorthol, 2013), and inadequately trained drivers, inflexible schedules, and accessibility issues may deter older individuals from using such options (Ahern & Hine, 2012). Problems of accessibility, awareness, safety, reliability, and cost commonly characterizing these alternatives further highlight the importance of ride-giving as a flexible arrangement better meeting the changed transportation needs of older individuals (Metz, 2014). Given that due to attrition, the analytic sample of this study was overrepresentative of healthy, well-educated, white older adults—a group that may be less dependent on sustaining driving mobility to draw the benefits social participation provides—the issue of transportation alternatives may be more pronounced of underprivileged older adults. Further research should examine the importance of driving mobility for social participation in such groups.

Finally, the effects of driving mobility are similar for informal and formal participation. SOC (Baltes & Carstensen, 1996) and SST (Carstensen et al., 1999) suggest that as older adults’ driving mobility resources diminish, they may focus on activities they deem most important. However, no clear differences emerged. Frequent driving, consistent ride receipt, and activity importance are strongly predictive of increased participation across all activities. Driving frequency decrease similarly lowered the likelihood of all participation types. A pattern may have gone undetected due to limited detail in participation types examined; analyses capturing more nuances in activity type should examine potential differences. Continuity theory, suggesting older adults continue participation in activities they were involved in earlier in life (Atchley, 1989), may explain the strong effect of activity importance on one’s likelihood of participating in the activity. The present study was only able to control for participation at first wave—with the factor highly predictive of subsequent participation—but further investigations may examine continuity from involvement earlier in the life course.

A key area of future research not explored in this study is a consideration of older adults’ geographic context in relation to their social participation. Many older adults give up the car keys, but age in place, or in rural and suburban areas where access to activity venues requires more frequent use of transportation. These older individuals may be at greater risk of transportation mobility exclusion (Staplin & Freund, 2013). Examining geographic influences may reveal differences in driving mobility and participation across urban, suburban, and rural location; address questions regarding aging in place; and invite comparisons with a sample of older adults living in nursing homes and assisted living facilities. This study also does not address the geographic proximity of potential ride-givers; despite willingness to provide rides, they may not be able to do so due to their distant location.

Despite these limitations, this study demonstrates that older adults’ driving mobility is multifaceted and cannot be adequately captured in a dichotomy indicating whether or not one currently drives. Rather, the “driving” category is heterogeneous, including both occasional and frequent drivers. Likewise, those previously classified as “nondrivers” include both those who ceased driving and those who never drove. Additionally, they may still have access to transportation by receiving rides from friends and family. Crucially, driving mobility matters for older adults’ formal and informal social participation, net of demographic, socioeconomic, health, and activity controls. Facilitating ride-giving and developing flexible transportation options like volunteer driver schemes (Kerschner & Rousseau, 2008) may enable many of the 600,000U.S. older adults yearly who cease driving, and many more in the cessation process wishing to remain involved, to close the transportation mobility gap (Foley et al., 2002) and enjoy a better quality of life.

Funding

National Health and Aging Trends Study (NHATS) is sponsored by the National Institute on Aging (grant number NIA U01AG32947) and was conducted by the Johns Hopkins University.