2.1 Sample and data collection
The data used in the present study was derived from China Health and Retirement Longitudinal Study (CHARLS), a national survey designed to provide micro information on the respondents aged 45 and above. The baseline survey of CHALRS was carried out in 2011 and tracked every two years. CHARLS conducted surveys and interviews in 2011, 2013, 2014 (the special project of "survey on the life course of middle-aged and elderly people in China"), 2015 and 2018 in 450 communities (villages) from 150 counties of 28 provinces, which includes autonomous regions and municipalities directly under the central government. The data included not only demographic information such as gender, year of birth and education, but also physical and mental health, income, family, and retirement information of the respondents. The target population of CHARLS were residents aged 45 and above, their partners/spouses were included in the survey irrespective of age. Adequate information makes CHARLS a good panel data for studying the influence of social activities on depression among the middle-aged and seniors. In this study, we only kept information about respondents aged 45 and above.
The 2011 baseline data included 17,708 respondents, followed by wave 2 investigated in 2013, in which 2,522 respondents were not included in wave 2. Then we got 15,186 records, followed by wave 3 in 2015 and 4 in 2018. We didn’t use any data investigated in 2014, because that survey only includes data about health indicators. Among 15,186 respondents in wave 1 and wave 2, 1,621 residents were not included in wave 3. And 1,577 respondents were lost in contact in wave 4. We finally got 11,988 respondents in all four waves. Since we want to investigate the influence of different types of social activities on depression, we selected 8,610 participants based on the following criteria: 1) Providing social participation information in all four waves. 2) Providing depression scores in all four waves. 3) Aged 45 and above in wave 1. The detailed sampling process was shown in Figure 1.
2.2 Measures
2.2.1 Informal and formal social activities
According to CHARLS, respondents were asked whether they have participated in 10 listed activities in the past month. Those ten activities are “ (1)Interacted with friends; (2) Played Ma-jong, played chess, played cards, or went to community club; (3) Provided help to family, friends, or neighbors who is not living with you and would not pay for the help; (4) Went to a sport, social, or other kind of club; (5) Took part in a community-related organization; (6) Done voluntary or charity work; (7) Cared for a sick or disabled adult who is not living with you and would not pay you for the help; (8) Attended an educational or training course; (9) Stock investment; and (10) Used the Internet;” Seven of them are social activities while the other three, i.e., (3), (9) and (10) are not [27]. Based on the intimacy and intensity of these activities, we classified “(1) Interacted with friends; (2) Played Ma-jong, played chess, played cards, or went to community club; and (4) Went to a sport in park , social, or other kind of club” into informal activities, and the remaining into formal activities.
2.2.2 Depression symptoms
CHARLS survey uses CESD-10 scale to measure depression symptom in middle-aged and senior citizens in China. CESD-10 scale is a shortened scale widely used to predict depression in older adults [28]. The respondents are asked the feeling about experienced things during the last two weeks. The scale includes ten items which are” (1) I was bothered by things that don’t usually bother me; (2) I had trouble keeping my mind on what I was doing; (3) I felt depressed; (4) I felt everything I did was a struggle; (5) I felt hopeful about the future; (6) I felt fearful; (7) My sleep was restless; (8) I was happy; (9) I felt lonely; and (10) I could not live like this anymore” We scored “Rarely or none of the time; some or a little of the time; occasionally or a moderate amount of the time; and Most or all of the time” as 0,1 ,2 and 3 respectively. The question of (5) and (8) are positive items and they were reverse-coded. We added them together and got the CESD score. People with higher score are more likely to be considered as depressed. We checked the internal consistency of the CESD-10 scale in our sample, and the Cronbach’s α=0.792.
2.2.3 Potential confounding variables
Potential confounding variables mainly include demographic variables, financial related variables, children relation related variables and health related variables. Demographic variables include age, marital status, living area, household registration, years of education and year of investigated. Financial related variables include job status and household financial status. Social relation related variables include living status. Health related variables include self-evaluation of health, smoke, alcohol drink and number of chronic diseases. Based on previous research, we used total household income divided by square root of number of people in the household to represent the family’s financial status [7]. We also used an important dummy control variable to indicate whether the respondents had experienced the loss of a child (children). The detailed information about data defining and data processing are shown in Table 1.
Table 1. Process of the potential confounding variables
Variables
|
Process
|
Demographic variables
|
|
|
Gender
|
Female=1; Male=0
|
|
Age
|
Continuous variables
|
|
Marital status
|
partnered=1; Single=0
|
|
Living area
|
Urban=1; rural=0
|
|
Household registration
|
Urban=1; rural=0
|
|
wave
|
“2011”=1; “2013”=2; “2015”=3, “2018”=4
|
|
Education level
|
Illiterate=1; primary school and lower=2; junior middle school=3; senior middle school and higher=4
|
Financial status
|
|
|
Job status
|
Work=1; not work=0
|
|
Household financial status
|
Total household income/square root (#number of people in the house hold)
|
Children relations
|
|
|
Live with children
|
Whether the respondents live with a child of Yes=1; No=0
|
|
Number of living children
|
The number of living children of the respondents
|
|
Ever lost a child (children)
|
Have respondents ever lost a child (children)? Yes=1; No=0
|
Health status
|
|
|
Self-evaluation of health
|
Health=1; unhealthy=0
|
|
Smoke
|
Have respondents ever smoked? Yes=1; N0=0
|
|
Alcohol drink
|
Have respondents ever drinks any alcohol last year? Yes=1; N0=0
|
|
#types of chronic diseases
|
Total of 12 types of chronic diseases including hypertension, diabetes, cancer, lung disease, heart problem, stroke, arthritis, dyslipidemia, liver disease, kidney disease, stomach/digestive disease, asthma.
|
|
memory
|
Self-reported memory: fair, Good or Excellent=1; Poor=0
|
2.3 Analytical strategy
In order to solve the endogenous problem, we used longitudinal fixed effect regression model in our analysis to estimate which type of social activities can decrease the level of depression symptoms. The fixed effect model can solve the problem of omitted variables which are caused by unobservable individual differences. Our model are as follows:
In order to further verify whether our model is reasonable, we made a series of tests. the first analysis we conducted was the F test, the results of F test showed that p<0.001, which indicated the fixed effect model is better than pooled ordinary least square. Then we used Hausman test to test whether fixed effect model is better than random effect model. The results also showed that p<0.001. Thus, we still use fixed effect model.