Data

This study utilizes data from the 2020 China Longitudinal Aging Social Survey (CLASS2020), a nationally representative longitudinal study tracking social and economic challenges among older adults. Conducted by the China Social Survey Network, the survey employs a multistage sampling approach: first selecting households via Secondary Sampling Unit mapping, then randomly choosing one resident aged 60 + per household. The 2020 wave includes 11,398 initial respondents.

To assess digital literacy, we first created an “internet use” variable based on survey responses to the question: “How often do you go online? 1. Daily; 2. Weekly; 3. Monthly; 4. A few times yearly; 5. Never.” Respondents answering “Never” (n = 3118, 27.36% of sample) were coded 0, while all others were coded 1. For internet-using respondents, we developed a composite digital literacy index encompassing five dimensions: device operation, information acquisition, social Literacy, application skills, and security awareness. Non-internet users were assigned a baseline value of 0 due to lack of observable digital behaviors.

Variables description

Explanatory variable

The dependent variable is community-based home care service utilization, measured using CLASS survey data. The survey categorizes services into nine types:1. Home visits;2. Elderly care hotline;3. Medical appointment accompaniment; 4. Daily shopping assistance;5. Legal aid;6. Housekeeping services;7. Meal delivery/elderly dining services;8. Day care centers;9. Psychological counseling. Respondents were coded as “1” if they used any of these services and “0” if they used none.

Core explanatory variable

The core explanatory variable is digital literacy, building upon existing multidimensional measurement frameworks developed for older adults. Prior research includes Huang et al.‘s (2021) [44] five-dimension media and information literacy assessment (awareness/knowledge, access/needs, evaluation/understanding, application/management, and ethics/security) and Wu et al.‘s (2023) three-dimension self-assessment scale (digital practice skills, learning awareness, and payment awareness) [45]. We use the following indicators, combined with the questions related to digital literacy in the CLASS2020 questionnaire, to measure the digital literacy of the older adults (Table 1). Using the factor analysis method, we subjected the digital literacy measurement data to Kaiser-Meyer-Olkin (KMO) and Bartlett’s test of sphericity. The results showed that the KMO was 0.710 and that the chi-square value of the Bartlett’s test of sphericity was 4669.21 (P < 0.001, degrees of freedom = 66). These results reject the original hypothesis that the variables are not related to each other, indicating that the data are suitable for factor analysis.

Table 1 Digital literacy measurement indicators

Control variables

Referring to studies such as Hu et al. (2023) and Luo et al. (2023) [11, 46], this paper controls for variables related to personal characteristics of the older adults (age, gender, ethnicity, education level, marital status, and self-care ability), household characteristics (number of co-residing family members, number of housing units, and household income), and community characteristics (community location and type). Variable names and descriptions are presented in Table 2.

Table 2 Variable name and variable description

Mechanism variables

Social support and family support

We used the frequency of contact with children and friends as proxies for family and social support, respectively. According to the questionnaire item, “How many friends do you see or contact at least once a month? Answer options: 0 = None; 1 = 1; 2 = 2; 3 = 3–4; 4 = 5–8; 5 = 9 and above.” According to the respondents’ answers, the higher the value, the stronger the social support is. According to the questionnaire question: “In the past 12 months, how often have you been in contact with this child (including various means of communication, such as telephone or WeChat)?1. Almost every day;2. At least once a week;3. At least once a month;4. A few times a year;5. Almost never;6. No need to be in contact.” The highest frequency of contact between the older adult and all children was generated by transforming the data into an ordered categorical variable with four levels:3 = Strong support: At least one child is contacted almost every day; 2. Moderate support = At least one child is contacted weekly, but not every day; 1 = Weak support: All children are contacted monthly or less; 0. No support/independence: All children do not need to be contacted.

Self-efficacy

Referring to Kong and Yan (2023) [47] and the questions in the CLASS2020 questionnaire based on the respondents’ evaluation of themselves: “do you feel that the following descriptions are consistent with your current reality? 1. I would be happy to take part in some work in the village/neighborhood committees if I have the opportunity; 2. I often want to do something for the society again; 3. I Like to study at present; 4. I feel that I am still a useful person to the society”, we assign values to the responses: not at all = 1, a little = 2, average = 3, more in line = 4, fully in line with the table = 5. Self-efficacy was calculated by summing up the scores of the respondents’ answers to the four questions, with higher the score indicating a higher sense of self-efficacy.

Self-related health

According to the questionnaire question: How do you feel about your current health? (1) very healthy (2) relatively healthy (3) average (4) relatively unhealthy (5) very unhealthy 9. unable to answer, a self-rated health variable was set. We reverse coded based on responses, with higher values indicating better self-rated health.

Mental health

The variable was set based on a set of questionnaire questions, which ask respondents to rate their mood in the last week. This questions include the following: 1.“Did you feel in good spirits?“;2.“Did you feel lonely?“;3.“Did you feel sad?“;4.“Did you feel your life was going well?“;5.“Did you have poor appetite?“;6.“Did you have trouble sleeping?“;7.“Did you feel useless?“;8.“Did you feel you had nothing to do?“;9.“Did you find life enjoyable (with many interesting things)?” The nine dimensions were assigned the following values: no = 1, sometimes = 2, often = 3. The responses for dimensions 1, 4, and 9 were reverse-adjusted, and the nine dimensions were summed to calculate mental health, with higher scores indicating lower levels of mental health.

Descriptive statistics

Table 3 shows descriptive statistics for some variables. Only 10.9% of the older adults have used CHCS, indicating a low utilization rate. The older adults’ digital Literacy has a mean of 1.055 and a maximum of 2.226, indicating a low overall digital literacy level among the older adults.

Table 3 Descriptive statistics

Empirical models

Probit model

This paper utilizes Probit modeling to estimate the effect of digital literacy on the utilization of CHCS for the older adults.

$$\:\text{P}\text{r}\left({\text{Y}}_{ic}\right)={\alpha\:}_0+{\alpha\:}_1{digital}_{ic}+{\alpha\:}_2C_{ic}+\lambda{}_c+{\epsilon\:}_{ic}$$

(1)

In Eq. (1), \(\:{\text{Y}}_{ic}\) is the binary variable indicating whether older adult i in county c utilizes community – based home care services. \(\:{digital}_{ic}\) represents the level of digital literacy, and \(\:{C}_{ic}\) denotes the control variables. \(\:{}_{c}\)represents county fixed effects, capturing time-invariant characteristics at the county level. \(\:{\epsilon\:}_{ic}\)stand for error term, respectively. The parameter \(\:{\alpha\:}_{1}\) captures the total effect of digital literacy on CHCS utilization. A significantly positive \(\:{\alpha\:}_{1}\) suggests that digital literacy considerably promotes such utilization among the older adults.

Heckman two-stage model

Since digital literacy is only calculable for internet – using older adults, selecting such a sample for regression may cause sample selection bias. This paper employs a Heckman model to correct for this bias. The estimation of digital literacy’s effect unfolds in two steps: first, constructing a selection equation to identify factors influencing elderly internet use, estimating this probability via Probit model, and computing the inverse Mills ratio; then, in the second stage, adding the inverse Mills ratio as a control variable to Eq. (3) and estimating it using the Probit model.

Therefore, this paper sets up the two – stage Heckman estimation model as follows:

$$\:\text{P}\text{r}\left({\text{I}\text{n}\text{t}\text{e}\text{r}\text{n}\text{e}\text{t}\_\text{u}\text{s}\text{e}}_{ic}\right)={\beta\:}_{0}+{\beta\:}_{1}{C}_{ic}+{\beta\:}_{2}{tele\_num}_{d}+{\epsilon\:}_{ic}$$

(2)

$$\:\text{P}\text{r}\left({\text{Y}}_{ic}\right)={\gamma\:}_{0}+{\gamma\:}_{1}{digital}_{ic}+{\gamma\:}_{2}{C}_{ic}+{\gamma\:}_{3}{mills}_{ic}+{\epsilon\:}_{ic}$$

(3)

Where, Eq. (2) is Heckman’s first – stage choice model for estimating the probability of Internet use among the older adults. \(\:{\text{I}\text{n}\text{t}\text{e}\text{r}\text{n}\text{e}\text{t}\_\text{u}\text{s}\text{e}}_{ic}\) is a binary variable of whether they use the Internet, 1 = yes, 0 = no. \(\:{C}_{ic}\) is a series of control variables affecting the older adults’ Internet use, and \(\:{tele\_num}_{d}\) is the number of landline telephones in the city in 1984, serving as an exclusionary variable.

Source link