Page 11 - IJPS-5-2

Sun R and Wang H

Southwest, and Northwest. Out of 31 provinces and national metropolises, 20 are selected. The survey follows a stratified,
multistage sampling design to randomly interview those 60-year-old or older. This analysis is based on data from two
waves in 2000 to 2010.

2.2. Measurement of Worry about Eldercare
Worry about eldercare is measured by older adults’ response to the question of whether they are worried about getting care
when needed. There are five response options: (1) Not worried at all; (2) almost not worried; (3) so-so; (4) a little worried;
and (5) very worried. It is recoded into two categories: Either worried (4 or 5) or not worried (1, 2, and 3).

2.3. Covariates
Family relations are measured by the number of children, living arrangements, and older adults’ perception of their
children being filial or not. The living arrangements have three categories: Living alone or with spouse only, living with
children, and living with others. As a subjective measure to gauge family relations, older adults were asked in the survey
about whether they had any worry about children being unfilial, which is coded as yes or no.
Older adults’ socio-economic resources mainly refer to income and medical coverage, including both objective and
subjective measures. Two objective measures are whether an older adult receiving any income from public sources
(yes or no), including pension from state insurance programs or enterprises, security income for the poor, and any income
from other public sources; and whether an older adult receiving any public medical care coverage coded as yes or no. Two
subjective measures include whether an older adult having any worry about insufficient income or unaffordable medical
care, coded as yes or no.
There are questions about eight kinds of community services asked in both years: Household work, personal care,
doctors’ home visits, companionship (having someone to chat), older adults’ hotline, meals to home, as company to see
a doctor or to go shopping. For each service, older adults were asked about the availability in the community, their need,
and their use of the service. Since the availability and need for the last four types of services are low in both years (under
10% for all in 2000), we will only include the first four types of services in the analysis. Preliminary results indicate that
the sheer availability of these services is not significant. Instead, we construct an unmet need variable for each service: If
there is a need for a specific service but unavailable in the community, it is coded as yes, otherwise no.
Among the control variables, there are age, sex (male, female), and residence (urban and rural). Since the percentage
of being divorced or never married is <2%, marital status is coded as a dummy variable: Currently married or not.
Education is coded at three levels: Illiterate, elementary school, and middle school or higher. An older adult’s health status
is measured by their difficulty in performing an array of activities of daily living (ADL), which include six basic activities:
Eating, clothing, toileting, getting in/out of bed, bathing, and walking. Each activity is measured by a score at three levels:
No difficulty, a little difficulty, and unable to perform. Since more than 70% of older adults from both years reported no
difficulty in any of the six activities, we recode it into a binary variable: Having any ADL difficulty or no difficulty at all.
Home ownership is a dummy variable: Yes or no.
2.4. Analytical Strategies

In the following analysis, we will first present descriptive statistics to illustrate the changes in older adult population
between 2000 and 2010. Next, we will adopt three logistic regression models for each year to test the two hypotheses.
Model 1 will only include family relations variables in addition to the control variables to show the effect of family
relations without socioeconomic resource variables. Socio-economic resource variables will be added to Model 2, which
will partially test Hypothesis 1, and Hypothesis 2a and b. Community service variables will be added to Model 3 to
completely test all hypotheses.

3. Results

3.1. Description of Older Adults in 2000 and 2010
Table 1 presents descriptive statistics of the two elderly samples in 2000 and 2010, respectively. Although the dependent
variable – worry about eldercare – remained virtually the same between 2000 and 2010, family relations, older adults’
socio-economic resources, and unmet need for community services had undergone dramatic changes during the same
time period.

International Journal of Population Studies | 2019, Volume 5, Issue 2 5