6.3.1 Model
A probit regression model
is used to investigate the probability of participating in volunteer activity
in Eq. (
6.1).
$$\mathrm{Pr}\left({Y}_{i}=1\right)=\mathrm{Pr}({a}_{0}+{\beta }_{ofe}{Fe}_{i}+{\beta }_{0X}{X}_{i}+{v}_{i}>0)$$
(6.1)
In Eq. (
6.1),
\(\mathrm{Pr}({Y}_{i}=1)\) denotes the probability of participating in volunteer activity.
i indicates individuals,
\(Fe\) is a female dummy variable, and
X represents the control variables. The coefficients of variables are denoted by
\({\beta }_{0fe}\) and
\({\beta }_{0X}\). In addition,
a is a constant, and
\({v}\) is the error term. When
\({\beta }_{0fe}\) is statistically significant, it indicates that a gender gap remains in the probability of participating in volunteer activity.
Two econometric problems must be addressed in Eq. (
6.1). First, there may be an endogeneity problem
between market work and volunteer activity. Instrumental variables (IV) methods
are used to address the endogeneity problem. IV methods are expressed as Eqs. (
6.2.1) ~ (
6.2.5).
$$\mathrm{Pr}\left({Work}_{i}=1\right)=\mathrm{Pr}({a}_{1}+{\beta }_{1Z}{Z}_{i}+{{\beta }_{1fe}{Fe}_{i}+\beta }_{1X}{X}_{i}+{u}_{i}>0)$$
(6.2.1)
$$\mathrm{Pr}\left( {\widehat{Work}}_{i}=1\right)=\mathrm{Pr}({a}_{1}+{\beta }_{1Z}{Z}_{i}+{\beta }_{1fe}{Fe}_{i}+{\beta }_{1X}{X}_{i}>0)$$
(6.2.2)
$${Resi}_{i}=Pr{Work}_{i}-\mathrm{P}r{\widehat{Work}}_{i}$$
(6.2.3)
$$\mathrm{Pr}({Y}_{i}=1)=\mathrm{Pr}({a}_{2}+{\beta }_{2Work}{\widehat{Work}}_{i}+{{\beta }_{2fe}{Fe}_{i}+\beta }_{2X}{X}_{i}+{\varepsilon }_{i}>0)$$
(6.2.4)
$$\mathrm{Pr}({Y}_{i}=1)=\mathrm{Pr}({a}_{2}+{\beta }_{2Work}{Work}_{i}+{{\beta }_{2Resi}{Resi}_{i}+{\beta }_{2fe}{Fe}_{i}+\beta }_{2X}{X}_{i}+{ \varepsilon }_{i}>0)$$
(6.2.5)
$$corr\left(Z,\varepsilon \right)=0\; \mathrm{and }\;corr\left(Z,u\right) \ne 0$$
Based on Eqs. (
6.2.1)–(
6.2.5), a two-step procedure is used for estimates: (1) In the first step, we employ the probit regression model
(Eq. (
6.2.1)), and then calculate the imputed value of probability of participating in market work
\(\widehat{Work}\) (Eq. (
6.2.2)) and the residual items
\(Resi\) (Eq. (
6.2.3)). (2) In the second step, we use the imputed value of
\(\widehat{Work}\) or residual items
\(Resi\) as an explanatory variable and estimate the probability of participating in volunteer activity (Eqs. (
6.2.4), (
6.2.5)). Whether the estimates are unbiased hinges critically on the validity of instrumental variable
(
Z); that is,
Z needs to be correlated with
Work while satisfying the conditions of
\(corr\left(Z,\varepsilon \right)=0\), and needs not to be correlated with
\(\mathrm{Pr}({Y}_{i}=1)\) (
\(corr\left(Z,u\right) \ne 0\)). We use the formal retirement experience dummy variable as the IV. IV methods
include the two-stage least squares (2SLS)
, the two-stage predictor substitution (2SPS)
, and the two-stage residual inclusion (2SRI)
model. Considering that the dependent variable of participation in volunteer activity is a binary variable, 2SPS and the 2SRI models are used in this study.
7 The 2SPS model is expressed by Eq. (
6.2.4), and the 2SRI model is expressed by Eq. (
6.2.5).
The second econometric problem for Eq. (
6.1) is the heterogeneity problem
.
\({v}_{i}\) in Eq. (
6.1) includes the unobserved individual-specific time-invariant
effect,
\({\mu }_{i},\) and true error,
\({\delta }_{it}\) (
\({v}_{it}\)=
\({\mu }_{i}\)+
\({\delta }_{it}\)). When the unobserved individual-specific time-invariant
effect is not addressed, a bias may exist in the results. To address this problem, a random effects model
is used. It is expressed by Eq. (
6.3).
$$\mathrm{Pr}({Y}_{it}=1)=\mathrm{Pr}({a}_{3}+{\beta }_{3fe}{Fe}_{it}+{\beta }_{3X}{X}_{it}+ {\mu }_{i}+{\delta }_{it}>0)$$
(6.3)
Finally, to address the heterogeneity
and other endogeneity problems
simultaneously, a random effects (RE)
probit regression model
and IV methods is used. It is expressed by Eq. (
6.4).
$$\mathrm{Pr}\left({Y}_{it}=1\right)=\mathrm{Pr}({a}_{4}+{\beta }_{4Work}{Work}_{it}+{\beta }_{4Resi}{Resi}_{i}+{\beta }_{4fe}{Fe}_{it}+{\beta }_{4X}{X}_{it}+ {\mu }_{i}+{\delta }_{it})>0$$
(6.4)
The Cragg-Donald Wald test is used as the weak instrumental variables test, and the Durbin-Wu-Hausman test is used to check for endogeneity between participation in market work and participation in volunteer activity.
To compare the differences in determinants of volunteer activity participation, we also run these models using female and male samples.
6.3.3 Variable Setting
Two dependent variables are constructed as follows: first, a binary variable for work is equal to 1 when an individual is working and equal to 0 when the individual is not working. It is used in the first stage of the IV method
. In China, the individuals experienced formal retirement are those who worked in the public sector (e.g., government organization, state-owned enterprises) before retriement, majority of them can receive higher public pension benefits after mandatory retirement. It can be predicted that the probability of participation work are lower for the individuals experienced formal retirement than the counterpart (who exited labor market without formal retirement experience, for example, the self-employed). On the other hand, for the individuals exiting labor market, when the other factors are constant, because the leisure hours (time constraints) may be similar regardness of formal retirement experience, the influnence of formal retirement experience on participation in volunteer activity may small. Therefore, the dummy variable of formal retirement experience (1 = has experienced the formal retriement, 0 = otherwise) is used as the IV in this study. The results of the first stage of the IV method are shown in the Appendix Table
6.4. Second, a binary variable for participation in volunteer activity is equal to 1 and equal to 0 when the individual is not participating in volunteer activity. Based on the CHARLS questionnaire item “Have you done any of these activities in the last month,” when the respondent chooses “done volunteer or charity work,” “cared for a sick or disabled adult not coresident with you,” or “provided help to family, friends, or neighbors not coresident with you,” it qualifies as participation in volunteer activity.
8
The main independent variables are conducted as follows. First, a female dummy variable is constructed to estimate the gender gap in volunteer activity participation.
Second, a set of variables of six factors are used as follows:
Market work
Based on the individual utility function in neoclassic economics theory, there is a trade-off relationship between market work
and volunteer activity.
4 It is expected that the probability of participating in volunteer activity
is lower for the working group than for the non-working group because of time constraints. Carlin (
2001) found that time spent on volunteer activity decreases as working hours increase. For Japan, Atoda et al. (
1999), Atoda and Fukushige (
2000), Yamauchi (
2001), Ono (
2006), and Moriyama (
2007) reported that the probability of participating in volunteer activity is inversely proportionate to the householder’s working hours
and working days. Although the labor force participation rate of women and men is higher for China than for other countries, the labor force participation rate for middle-aged women has decreased during the current period. It is assumed that there remains a trade-off relationship between labor market work and volunteer activity. The dummy variable of work is equal to 1 when an individual is working and equal to 0 when the individual is not working;
Education
Some research has demonstrated that educational background has a strong influence on participation in volunteer activity, and that persons with a higher level of education are more likely to volunteer (Vaillancourt
1994; Freeman
1997; Ma and Ono
2013; Wu et al.
2018; Tong et al.
2018; Lin
2019). This can be explained by the human capital theory
(Becker
1964). It may also be that social contribution consciousness differs among different educational attainment groups. For example, Carlin (
2001) found that the higher the market wage
(higher educational attainment) of married women, the more they volunteer (probability of participating in volunteer activity, volunteer activity hours). It also can be expected that educational attainment may affect volunteer activity. There educational attainment level dummy variables—low, middle, and high education levels
9 were used.
Income factors
Menchik and Weisbrod (
1987) employed an empirical study to investigate the mechanism of participation in volunteer activity based on the consumption model
. Menchik and Weisbrod (
1987), Vaillancourt (
1994), Ma and Ono (
2013), Tong et al. (
2018), and Lin (
2019) found that the higher the unearned income
, the longer the time spent on volunteer activities; conversely, the higher the market wage rate, the shorter the volunteer activity hours. Based on the consumption model
advocated by Menchik and Weisbrod (
1987) and previous studies, it is assumed that the probability of participating in volunteer activity may be higher for the high-unearned income group than for the low-unearned income group in China.
Three income factor variables were constructed as: (i) the logarithmic value of the annual household consumption; (ii) the pension receipt dummy variable, which is equal to 1 when a woman is receiving a pension and equal to 0 when she is not; and (iii) the transfer to children dummy variable, which is equal to 1 when a women transfers her pension income to her children and equal to 0 when she does not.
Family care
Menchik and Weisbrod (
1987), Morgan et al. (
1977), and Vaillancourt (
1994) found that the probability of volunteer activity tends to be lower and the hours of volunteer activity tend to be less when caring for infants.
5 In a study of married women
, Carlin (
2001)
6 found that the probability of participating in volunteer activity increases but the hours of volunteer activity decrease as the number of children
increases. In Japan, Atoda et al. (
1999), Atoda and Fukushige (
2000), Yamauchi (
2001), Ono (
2006), Moriyama (
2007), and Ma and Ono (
2013) reported that the duties of family care
(child care
, parent care
) and the number of children decrease the probability of participating in volunteer activity
and the hours of volunteer activity.
Regarding family care in China, because the one-child policy had been in effect since 1979, most middle- and older aged women aged 45 and older in the survey years from 2011 to 2015 had one or two children, and most children were more than 14 years old. It can be assumed that, with an increase in the number of children, parents have more hours of work at home, which may decrease the probability of participating in volunteer activity. In contrast, family caregivers may have stronger altruistic values and greater motivation to volunteer to help others than those who are not family caregivers. It is predicted that the number of children may decrease the probability of participating in volunteer activity, but the availability of family care may increase the probability of participating in volunteer activity.
Three family factor variables were used as: (i) the number of children; (ii) the grandchildren caregiving dummy variable, which is equal to 1 when a women is caring for grandchildren and equal to 0 when she is not; and (iii) the child care giving variable, which is equal to 1 when a women is caring for children and equal to 0 when she is not.
Age
Both the probability of participating in volunteer activities
and the time spent on such activities tend to change with age (Menchik and Weisbrod
1987; Vaillancourt
1994; Wu et al.
2018; Tong et al.
2018; Lin
2019). For example, for developed countries, Menchik and Weisbrod (
1987) pointed out that, while the time spent on volunteer activity increases with age to a point, it decreases after the age of 43. Vaillancourt (
1994) found that the most likely age for volunteering for both men and women is from age15 to 19, and that while men are more likely to participate in volunteer activities from age 25 to 54, this probability decreases between age 55 and 69. For women, the probability of participating clearly decreases at age 70 and older. In America and Canada, participation in volunteer activities during one’s student years is utilized as one aspect of socio-cultural background that may influence the employment and wages of an individual, which may reinforce an individual’s participation in volunteer activities; this is known as the human capital
investment theory. In China, Wu et al. (
2018) and Lin (
2019) reported that the probability of participating in volunteer activity decreases with age for those aged 18 and older. Tong et al. (
2018) found this was also true for people aged 60 and older; the probability of participating in volunteer activity decreases with age. To consider the lifestyle and situation of middle-aged and older individuals in China, regarding the decrease in health status and social contribution efforts with age, it is assumed that the probability of participating in volunteer activity decreases with age. Regarding the retirement eligibility age for women is 50 years for workers and 55 years for cadres, we constructed three age dummy variables—age 45–49, 50–59, and 60–69 years old.
Health status
Health status
as a part of human capital can influence both market work and volunteer activity. Grossman (
1972) advocated a health investment model to emphasize the importance of health status for an individual. It is expected that health status may enhance participation in volunteer activity. Atoda et al. (
1999) and Ma and Ono (
2013) found that a healthy status increases the probability of participating in volunteer activity in Japan. In China, Tong et al. (
2018) and Wu et al. (
2018) found that poor health status decreases the possibility of participating in volunteer activity. Based on the human capital theory
(Becker
1964) and the results of previous empirical studies, it can be assumed that the probability of participating in volunteer activity
is higher for the healthy group than for the group with poor health in China. The subjective health status (poor, fair, good, very good, excellent) dummy variables were used as the indices of health status.
Other factors
Gender, marital status, environmental changes (i.e., earthquakes, natural disasters), and community factors, have also been shown to possibly affect participation in volunteer activity (Schram and Dunsing
1981; Menchik and Weisbrod
1987; Vaillancourt
1994; Atoda et al.
1999; Atoda and Fukushige
2000; Yamauchi
2001; Ono
2006; Moriyama
2007; Ma and Ono
2013). Therefore, the other factors such as gender, marital status, and regional disparity variables are also constructed as follows. (1) A female dummy variable is used to investigate the gender gap in volunteer activity participation. (2) According to the labor market segmentation hypothesis (Piore
1970), the behaviors of workers are shaped by the characteristics of the labor market. As is well known, the Chinese labor market is segmented by the population registration (
Hukou) system. In 1958, the
Hukou system was implemented by the government. Under the planned economy period from 1949 to 1977, migration from rural regions to urban regions was prohibited. Since the 1980s, the
Hukou system has been deregulated, great differences in the labor market remain between rural and urban areas. For example, the social security systems (i.e., public pension, medical insurance schemes) differ under the
Hukou system. Therefore, a urban
Hukou dummy variable is used to control the influence of segmentation, which is equal to 1 when a woman has urban
Hukou and equal to 0 when the individual has rural
Hukou. (3) A spouse dummy variable is conducted, which is equal to 1 when a woman has a husband and equal to 0 when she is single. (4) The culture and the economic development level may change for different periods; therefore, the year dummy variables (2011, 2013, 2015) are included in the control variables. Although previous studies (Wu et al.
2018; Lin
2019) reported that social capital may affect volunteer activity in China, because we cannot obtain the information from the CHARLS, this must be a future research issue. The gender gap of volunteer activity is discussed in the following section.
Table
6.1 presents descriptive statistics for (a) the total sample, (b) participation in a volunteer activity group (PVA), and (c) non-participation in a volunteer activity group (non-PVA) of individuals aged 45–69. It is observed that (1) in China, the proportion of participation in volunteer activity for individuals aged 45–69 is 15.8%. (2) To compare the employment rate between PVA group and non-PVA group, it is 7.6% points higher for the PVA group than for the non-PVA group. (3) The proportion of middle-and high-level education groups is higher for the PVA group than for the non-PVA group. (4) The income level is higher for the PVA group than for the non-PVA group. In addition, the proportion of income transferred to children is higher for the PVA group than for the non-PVA group, while the proportion of pension recipients is lower for the PVA group than for the non-PVA group. (5) The number of children is less for the PVA group than for the non-PVA group, but the proportions of both caring for grandchildren and for parents are higher for the PVA group than for the non-PVA group. (6) The proportion of women with poor health status
is 7.0% points lower for the PVA group than for the non-PVA group, while the proportion of healthy women is higher for the PVA group. The results show that individual attributes, family structure, and income factors differ between the PVA group and the non-PVA group. These factors should be controlled in estimations.
Table 6.1
Descriptive statistics of variables
Volunteer activity participation | 0.158 | 0.365 | | | | | |
Female | 0.483 | 0.500 | 0.439 | 0.496 | 0.492 | 0.500 | −0.053 |
Work | 0.765 | 0.424 | 0.829 | 0.377 | 0.753 | 0.431 | 0.076 |
Education |
Low-education | 0.860 | 0.347 | 0.796 | 0.403 | 0.872 | 0.335 | −0.076 |
Middle-education | 0.118 | 0.323 | 0.160 | 0.367 | 0.110 | 0.313 | 0.050 |
High-education | 0.022 | 0.147 | 0.044 | 0.205 | 0.018 | 0.133 | 0.026 |
Consumption | 8.907 | 0.911 | 9.139 | 0.925 | 8.864 | 0.902 | 0.275 |
Pensioner | 0.285 | 0.452 | 0.267 | 0.443 | 0.289 | 0.453 | −0.022 |
Transfer to children | 0.379 | 0.485 | 0.507 | 0.500 | 0.355 | 0.479 | 0.152 |
Number of children | 2.361 | 1.191 | 2.186 | 1.119 | 2.394 | 1.201 | −0.208 |
Grandchildren care | 0.402 | 0.490 | 0.412 | 0.492 | 0.401 | 0.490 | 0.011 |
Parent care | 0.093 | 0.291 | 0.125 | 0.331 | 0.087 | 0.282 | 0.038 |
Age | 56.186 | 6.852 | 54.689 | 6.667 | 56.467 | 6.850 | −1.778 |
Health status |
Poor | 0.206 | 0.404 | 0.147 | 0.355 | 0.217 | 0.412 | −0.070 |
Fair | 0.523 | 0.500 | 0.532 | 0.499 | 0.521 | 0.500 | 0.011 |
Good | 0.148 | 0.355 | 0.162 | 0.369 | 0.145 | 0.352 | 0.017 |
Very good | 0.111 | 0.314 | 0.138 | 0.345 | 0.106 | 0.307 | 0.032 |
Excellent | 0.013 | 0.113 | 0.020 | 0.139 | 0.012 | 0.107 | 0.008 |
Urban | 0.211 | 0.408 | 0.261 | 0.439 | 0.201 | 0.401 | 0.060 |
Having spouse | 0.924 | 0.265 | 0.931 | 0.253 | 0.923 | 0.267 | 0.008 |
Survey year |
y2011 | 0.305 | 0.460 | 0.161 | 0.368 | 0.332 | 0.471 | −0.171 |
y2013 | 0.340 | 0.474 | 0.364 | 0.481 | 0.335 | 0.472 | 0.029 |
y2015 | 0.355 | 0.479 | 0.475 | 0.499 | 0.333 | 0.471 | 0.142 |
Observations | 28,463 | | 4,491 | | 23,972 | | |