Analysis of Microdata: Method

Go Back

The rest of this report uses 45 national datasets for 41 countries to estimate indicators disaggregated across disability status for countries with census or household survey data that have internationally comparable disability questions. The analysis was conducted using Stata 16.0. Results are estimates and are nationally representative. Standard errors are not presented for conciseness. For datasets with a complex design, estimates are weighted.

Disability Measurement

Disability is notoriously difficult to define (Appendix 3 Method brief #1). In fact, in some cultures, the word and the notion do not exist. In this report, disability is defined as an interactional notion, one that results from an individual with a health condition interacting with the environment. Disability is measured through self-reported functional difficulties (e.g. difficulty seeing, hearing, walking) (Appendix 3 Method brief #2). Functional difficulties vary in terms of their type (e.g. seeing, hearing), degree (from mild to extreme) and age at onset (from birth to old age).

The datasets under study all have questions on functional difficulties that meet at least the United Nations (2017) Principles and Recommendations for Population and Housing Censuses. They are listed in Table 4.1. Twenty-five datasets have the questions of the internationally tested Washington Group (WG) Short Set (WGSS) covering six domains (seeing, hearing, walking, cognition, self-care, and communication) (Altman 2016)^[5],[6].

The other 20 datasets include some datasets that have questions similar to the WGSS but with modifications of either the wording of questions or the answer scale, and other datasets that do not adopt the WGSS but have functional difficulty questions in four to six domains as per United Nations (2017)^[7].

Disability is measured solely based on selected functional difficulties, and thus does not capture all persons with disabilities, in particular persons with psychosocial and mental health disabilities. Given the very incomplete nature of the measure under use, we refer to “persons with functional difficulties” and not to persons with disabilities, a broader group.

To identify a specific ‘functional difficulty status’ group, a threshold needs to be set on the answer scale of functional difficulties. Recognizing that identification and categorization lead to imperfect results and potentially varying ones depending on the threshold, the disability data initiative uses three ways to partition individuals based on functional status (Appendix 3 Method brief #3).

A. First, for all datasets, individuals are in two categories:

• No functional difficulty.
• Any functional difficulty in at least one domain (respondents answer Yes for datasets with Yes/No answers, or reports at least Some difficulty for graded scales).

B. For datasets with graded answer scales, individuals are in three categories:

• No difficulty for all domains.
• Some difficulty in at least one domain but no A lot of difficulty or Unable to do responses across all domains.
• A lot of difficulty or Unable to do in at least one domain.

C. Finally, and again only for datasets with graded answer scales, following the recommendation of the WG, individuals are grouped as follows:

• No difficulty or Some difficulty for all domains
• A lot of difficulty or Unable to do in at least one domain.

Data tables available on the disability data initiative website include results for the three categorizations above.

While the WGSS was initially developed for use in censuses for individuals 5 years of age and older, the six domains may not be adequate to capture disability among children (Loeb et al 2018).  We therefore calculate disability indicators only for adults 15 years and older and their households.

The analysis conducted at the household level categorizes households depending on the functional difficulty status of its members aged 15 and older along the three ways of partitioning the population described above.

Table 4.1: Microdatasets Under Study

Indicators

This report uses various indicators to capture the rights and human development situation of persons with disabilities. The indicators are in Table 4.2 and described in Appendix 3 Method Brief #5. The list of indicators was developed by reviewing the questionnaires of datasets in light of the provisions of the CRPD and the SDGs that they could inform (IWGHS 2018; OHCHR 2021). Indicators known to be particularly suited to assess the situation of persons with disabilities were included (United Nations 2019, Mizunoya et al 2013): employment population ratio, economic insecurity (using proxy variables food insecurity, exposure to shocks) and indicators that may reflect the extra costs of living with disabilities for households (health expenditures as a share of total consumption expenditures), as well as material wellbeing indicators (asset ownership, living conditions) that might be affected due to the extra costs of living with disabilities.

This report and Results Data Tables on the ddi website compare indicators across groups by functional difficulty status to establish the size of the gap that may be associated with disability, i.e. the disability gap or inequality associated with disability. For each dataset and indicator, we set 100 observations as the minimum requirement to produce estimates disaggregated across functional difficulty status. Results are presented in tables. The difference across functional difficulty status and its statistical significance is noted in a separate column. Statistical significance is based on a t-test (*, **, and *** at the 10%, 5% and 1% levels respectively). We use the term disability gap to refer to a difference that is statistically significant and refers to a disadvantage for persons with functional difficulties.

There may be patterns of disadvantage that affect subgroups of persons with disabilities and their households, such as women and rural residents. Data tables on the ddi website give a disability disaggregation of subgroups of the population by sex, rural/urban and age when the sample size is above 100.

Table 4.2: Indicators Under Study

Scope of the Study and Limitations

The countries and territories under study are described in Appendix 2. They were picked due to the availability of national survey or census data that at least meet the United Nations (2017) Principles and Recommendations for Population and Housing Censuses. They vary greatly in terms of life expectancy at birth, Gross National Income per capita and human development as measured by the human development index. They are also heterogeneous with regards to their legislative and policy backgrounds with respect to disability as shown in Appendix 2. All but four have ratified the CRPD, 13 countries have constitutional guarantees on the rights of persons with disabilities and 26 countries have anti-discrimination legislations in the workplace.

This analysis has some limitations. Results in this report are based on censuses and household surveys that do not include population members that are not in a household, such as the institutionalized and the homeless population. The data under use are affected by a mortality bias, as adults with functional difficulties may be disproportionately affected by premature mortality. It does not identify persons with a variety of disabilities, including psychosocial and mental health ones, which are counted under persons with no difficulty. For indicators that are captured at the household level in a survey (e.g. assets, food insecurity), there is no information on how resources are distributed within the household. It is possible that disability, as well as sex and age, impact resource allocation across members of the household. No information is available on this in the datasets under study. For these reasons, this analysis may lead to underestimates of disability inequalities.

Additionally, results come from a variety of surveys and censuses that may use different questions to capture indicators such as food insecurity or literacy, leading to results that are not entirely comparable across countries. For functional difficulty questions, what persons may understand from the questionnaire and how they reply can differ given different languages^[8], cultures, interviewer training and other contextual factors in ways that are beyond the purview of the researchers.

Only three factors that may contribute to intersectional disadvantages with disability are considered (sex, rurality, age), while others are not covered (e.g. immigration status, ethnicity, indigeneity).

At the same time, results from this study contribute to a growing international literature on disability inequalities. This report uses a variety of datasets from 2009 to 2018 with functional difficulty information on all adults in a household and detailed information that indicate whether equal rights have been respected and human development has been achieved. It is intended to add to a literature that notably relied on the 2002-2004 World Health Survey (WHO-World Bank 2011; Mitra and Sambamoorthi 2013) and for some countries, it provides a comprehensive assessment of the situation of persons with disabilities shortly before the COVID-19 pandemic broke out.

[5] More information is available at

[6] For four countries (Nigeria, Senegal, South Africa, Uganda), results are from two datasets: the DHS and another dataset. The DHS was used only to estimate indicators related to family planning and domestic violence in the four countries and to cell phone ownership for Uganda.

[7] Details on how datasets with other functional difficulty questions differ from the WGSS are in Table 4.1.

[8] The WG has a translation protocol to help preserve the meaning of the questions, but it is not known whether that protocol was used for the data sets with the WGSS.