Data and sample
The data used in this study were based on the Survey of Texas Adults 2004, which contains a statewide representative sample of 1504 community-dwelling adults aged 18 and over in Texas [13]. The sampling procedure was based on a modified random digit dialing design. A household-level cooperation rate of 37% and a respondent-level cooperation rate of 89% were obtained in the data-collection process. The survey was mainly conducted in English. Translation of survey instruments into Spanish and administration by a Spanish-speaking interviewer were applied when needed. Each computer-assisted telephone interview lasted about 30–35 min. Due to the overrepresentation of women, older adults, non-Hispanic Whites, and highly educated respondents in the original sample, the data were weighted to match the characteristics of the sample to the 2000 Texas population census estimates. The Survey of Texas Adults 2004 provided fruitful information on people’s participation in various types of voluntary services and health outcomes. Detailed socio-demographic data available in the survey can help adjust for confounding from the relevant background characteristics. The socio-demographic variables included in this study are gender, age, race/ethnicity, education, citizenship, marital status, number of children at home, employment status and family income. These background characteristics have been found influential on both volunteering and health outcomes in past research [1,2,3, 5, 9, 14].
Measures
All measures employed for analysis in this study were self-reported by the adult participants. The information about these measures was mainly drawn from the sections of volunteering, physical health, mental health and demographic characteristics in the survey.
Health outcomes
Mental health
In the Survey of Texas Adults 2004, a question was used to measure the mental health of the adult participants: “Overall, how would you rate your mental health at the present time? Would you say it is excellent, very good, good, fair, or poor?” The ratings are based on a 5-point scale (1 = excellent, 2 = very good, 3 = good, 4 = fair, and 5 = poor). For easy interpretation, the scale was reversely coded, meaning that higher scores represent better mental health.
Physical health
The adult participants in the survey were asked to respond to the question: “How would you rate your physical health at the present time? Would you say it is excellent, very good, good, fair, or poor?” The answers were used as a general indicator of physical health. The responses are also based on 5-point scale from 1 = excellent to 5 = poor. Again for better interpretation, the scale was reversely coded to indicate that higher scores represent better physical health.
Life satisfaction
The survey question used to rate participants’ levels of life satisfaction was: “How satisfied you are with your life overall?” The question was rated on a 4-point scale (1 = very satisfied, 2 = somewhat satisfied, 3 = not too satisfied, and 4 = not at all satisfied). The scale was reversely recoded, meaning that higher scores indicate better life satisfaction.
Depression
Five question items in the survey were used to measure participants’ depression in the past 30 days: “feeling sad”, “feeling hopeless”, “feeling everything was an effort”, “feeling worthless” and “had trouble breathing”. These items have been used to indicate depressive symptoms in prior research [7, 15]. They were rated on a 5-point scale from never (1) to several times a day (5), and a composite score was gathered by summing up the items. Cronbach’s α was .739 in this study.
Social well-being
Two question items in the survey were used to measure social well-being: “I am lacking companionship” and “I feel isolated from others”. These items are indicative of social integration and social acceptance and satisfaction about relationships with others [16, 17], which has been found influential on health [16]. The items were rated from strongly agree (1) to strongly disagree (5). The items were summed up to form a composite score. Higher scores imply better social well-being. Cronbach’s α was .717.
Volunteering
Other-oriented volunteering
Recent philanthropic research indicates that volunteers simultaneously engage in various types of voluntary services; therefore, simply dichotomizing participants into volunteers and non-volunteers is inadequate [7, 14]. In this study, volunteering is classified into the two broad forms of other-oriented and self-oriented volunteering. Prior research indicates that volunteers of other-oriented motivation were more likely to volunteer in health, social, religious and other philanthropic services [11]. Hence, other-oriented volunteering was measured by summing up participation in the past 12 months in the voluntary services of health, education, religious groups, human services, public/social benefits, and youth development. In fact, these voluntary services explicitly bear other-regarding and altruistic features that, by their nature, show concern and care for the needs of others [2, 11]. The scores of this volunteering form range from 0 to 6, higher scores indicative of more participation in other-oriented volunteering.
Self-oriented volunteering
In this study, participation in the voluntary services of recreation, arts or culture, environment or animal welfare, work-related service, political campaign or movement, and other service simultaneously in the past 12 months was summed up to form self-oriented volunteering, the types of voluntary services bearing features of self-actualization and development or self-serving [10,11,12]. Concordantly, prior research has reported that volunteers of self-oriented motivation were fond of volunteering in culture/recreation, environment, law/politics, and business or professional services [11]. Participation in these voluntary services shows that those who volunteer may actually emphasize reciprocation of volunteering to benefit and enhance themselves, e.g. increased social network and ties, understanding of self, evasion of personal problems, acquisition of new skills, and career development [2, 11, 12]. The scores of this form of volunteering also range from 0 to 6, higher scores indicative of more participation in self-oriented volunteering.
Control variables
The socio-demographic variables adjusted in this study include gender (1 = female, 0 = male), age in years, race/ethnicity, education (1 = none, 2 = high school, 3 = GED, 4 = associates degree, 5 = bachelor’s degree, 6 = graduate degree, 7 = doctorate), citizenship (1 = US citizen, 0 = other), marital status (1 = currently married, 0 = other), number of children at home, annual family income (1 = $0 to $14,999, 2 = $15,000 to $34,999, 3 = $35,000 to $49,999, 4 = 50,000 to $64,999, 5 = $65,000 to $84,999, and 6 = $85,000 or more), and employment status (1 = employed, 0 = other). Three dummy variables were constructed for race/ethnicity, in which African American (Black), Hispanic/Mexican American, and other races or ethnicities were the contrast groups, and non-Hispanic White was the reference category. Number of children was coded 0 to 4 or more children. As family income has the missing values of 34.9%, Expectation Maximization Imputation (EM) was used to replace the missing values rather than mean substitution, which was applied in previous relevant research [18]. Mean substitution will bias the mean distribution and restrict variance. EM, however, may set off these problems by using a two-step iterative process that involves regression analysis and maximum likelihood procedures to allow all available pertinent variables as predictors for imputing missing data [19]. A dummy variable was created to indicate whether the participant had missing information on income (1 = missing, 0 = other), to preclude confounding.
Statistical analyses
Due to the multi-correlated nature of the health outcomes, the current study employed multivariate linear regression to analyze the results.Footnote 1 This modeling approach has the advantage of reducing multi-collinearity and problems of Type I errors when there are significant correlations among the outcome variables [22]. Then, all five health outcomes are concurrently regressed on the predictors of volunteering and pertinent socio-demographic covariates. For easy interpretation of the regression results, the predictors of volunteering and socio-demographic covariates were standardized into z-scores, so it is possible to calculate percentages increased in the health outcomes by one unit increase in volunteering (e.g. additional participation in the voluntary services) through likelihood ratio= e
β − 1. In addition, the Wald test of parameters equivalence constraint was used to ascertain whether other-oriented volunteering had stronger health effects than did self-oriented volunteering. The statistical analyses were performed by Mplus 7.11.