2.1 Data source
This study used round-I and round-II of the India Human Development Survey (IHDS). IHDS round-I, conducted during 2005, is a nationally representative and multi-topic survey of 41554 households across all the states and union territories of India except for Andaman & Nicobar Islands and Lakshadweep (S. Desai, Vanneman, & National Council Of Applied Economic Research, 2008). IHDS round-II, conducted in 2012, is also a nationally representative and multi-topic survey of 42152 households with geographical coverage similar to round-I (S. Desai & Vanneman, 2015). IHDS round-II re-interviewed 83% of the households from round-I. Both rounds of IHDS adopted a multistage stratified random sampling design. Further details regarding the IHDS sampling frame, data collection procedure and respondent consent can be found elsewhere (S. B. Desai et al., 2010; S. Desai, Dubey, & Vanneman, 2015).
In this study, we refer to persons aged 60 years and above as older adults. During IHDS round-I, there were 17904 individuals and among them, 4736 older adults were not alive and 2641 older adults were lost to follow-up during round-II. Thus, our current study is based on panel data of 10527 older adults. Additionally, there were no records with missing information for all the variables used in our study.
2.2 Outcome variables
We used two binary outcome variables in this study – whether older adults acquired disability and recovered from disability between round-I and round-II, respectively. Both these outcome indicators were obtained from an older adult’s self-reported disability status during both rounds of IHDS. During both rounds, IHDS asked respondents that whether they had difficulty in – “walking 1 km”, “going to the toilet without help”, “dressing without help”, “hearing normal conversation”, “speaking normally”, “seeing distant things” and “seeing near objects such as reading/sewing”. Responses to these seven questions were coded as “0” (No difficulty), “1” (With difficulty) and “2” (Unable to do it). We summed the coded responses for each person to obtain a disability score ranging from 0 to 14 and found the median disability score to be 0 in both rounds. Therefore, older adults with a score of 0 were classified as “not disabled” and with a score above 0 were classified as “disabled”.
In the acquired disability variable (no, yes), older adults who not disabled in 2005 but were disabled during 2012 were categorized as “yes” and otherwise were categorized into “no”. Equivalently, in the recovered from disability variable (no, yes), older adults who disabled in 2005 but were not disabled during 2012 were categorized as “yes” and otherwise were categorized into “no”.
2.3 Control variables
Existing studies have shown several factors, which influences the change in disability status among older adults. We controlled for the confounding effects of the majority of these variables, conditional to their availability in the IHDS dataset. The demographic and social control variables related to the older adults include – gender of individual (male, female), age group (60–69 years, 70–79 years, 80 + years), current marital status (currently married, currently not married), level of education (no formal schooling, less than 5 years of schooling, 6–10 years of schooling, more than 10 years of schooling), working status (working, not working). Additionally, we included self-reported indicators of chronic illness indicating whether an older adult has – cardiovascular diseases (no, yes), hypertension (no, yes), diabetes (no, yes), respiratory illness (no, yes), cataract (no, yes) and any other chronic illness other than the above (no, yes). We also controlled for relevant household socio-economic characteristics – headship status (no, yes), family structure (joint/extended, nuclear, single generation), wealth quintile (poorest, poor, medium, rich, richest), household below poverty line (BPL) status (no, yes), the caste of household (scheduled tribes (ST), scheduled castes (SC), other backward classes (OBC), others), the religion of household (Hindu, Muslim, others), place of residence (urban, rural). Additionally, we also included the country region that a person comes from (northern, north-eastern, central, eastern, western, southern). All the above characteristics were measured for the older adults during round-I.
During round-I IHDS collected information on the marital status of each person and originally categorized them into six categories – “spouse absent”, “married”, “single”, “widowed”, “separated/divorced” and “no gauna”. Owing to the skewed population distribution across each category, we have recoded the original variable into a binary marital status variable (currently married, currently not married). Here, all individuals not included in the “married” category in the original variable were included in the “currently not married” category of the recoded variable.
IHDS obtained the information on whether each person reportedly suffered from – cataract, tuberculosis, hypertension, cardiovascular diseases, diabetes, leprosy, cancer, asthma, polio, paralysis, epilepsy, mental illness, sexually transmitted diseases (STD) and any other chronic disease. If an older adult suffered from the above chronic diseases, then they were coded as “yes” and otherwise they were coded as “no”.
Household family structure was obtained from the information given on the relationship of each household member with the head of the household. Based on this information we classified the family structure into – single generation, nuclear and joint/extended family. The single generation includes a married/cohabiting couple or a single person household. The nuclear family includes married/cohabiting partners along with their dependent and unmarried children. The joint family includes a parent and/or partner along with their children and grandchildren. The extended family is similar to a joint family structure with the exception that it also includes “extended members”, that is, people who are not directly related to the household head by blood.
The household wealth quintile for round-I was calculated using principal component analysis (Filmer & Scott, 2008). We generated a wealth score for each household using the available information on household asset ownership, livestock ownership, building material used in household, household water source, household sanitation facility and the number of rooms. Based on the wealth score we categorized the households into five categories (poorest, poor, medium, rich, and richest).
The country regions during round-I were formed by dividing the erstwhile 33 states and union territories of India into six regions. The northern region includes Chandigarh, Delhi, Haryana, Himachal Pradesh, erstwhile Jammu & Kashmir, Punjab, Uttaranchal and Rajasthan. The north-eastern region includes Assam, Arunachal Pradesh, Manipur, Meghalaya, Mizoram, Nagaland, Tripura and Sikkim. The central region consists of Madhya Pradesh and Chhattisgarh. The eastern region includes Bihar, Jharkhand, Odisha and West Bengal. The western region comprises Dadra & Nagar Haveli, Daman & Diu, Goa, Gujarat and Maharashtra. The southern region comprises erstwhile Andhra Pradesh, Karnataka, Kerala, Tamil Nadu and Pondicherry.