3. Disability Disaggregation
Go Back 3.1 Disaggregation Based on Disability
We compare indicators across groups by disability status. Disaggregating an indicator (e.g. ever attended school rate) by disability status aims to establish the size of the gap that may be associated with disability, i.e. the disability gap or inequalities associated with disability.
Disability is measured by functional difficulty questions and measures (Method brief 2). When functional difficulty questions have yes/no answers, disaggregation is done for persons with no difficulty vs any difficulty (disaggregation A). When functional difficulty questions have a graded answer scale, disaggregation is done in two additional ways: persons with no difficulty vs some difficulty vs at least a lot of difficulty (disaggregation B); persons with no difficulty and some difficulty vs at least a lot of difficulty (disaggregation C).
In tables, the difference between groups and its statistical significance is typically noted in a separate column. A disability gap represents a statistically significant disadvantage for persons with functional difficulties compared to persons with no functional difficulty. Statistical significance is based on a t-test (*, **, and *** at the 10%, 5% and 1% levels respectively). As indicators reflect achievements (e.g. employment population ratios) or deprivations (food insecurity, exposure to shock), a disability gap may be reflected in a positive or a negative difference.
This study uses national household surveys and censuses. Censuses typically include all people in a country, irrespective of their disability status. In contrast, household surveys are constructed out of sampling from censuses with complex sampling design. It should be noted that none of the household surveys under study is sampled to be representative of persons with disabilities. Censuses are thus better able to represent the situation of persons with disabilities than household surveys, which may not be representative of all persons with disabilities due to their sampling. At the same time, interestingly this study finds patterns in the results on disability gaps with census data and with survey data.
3.2 Disaggregation Based on Disability and Demographic Characteristics
There may be patterns of intersectional disadvantage that affect subgroups of people with disabilities and their households, such as women or rural residents. For each data set under consideration, we tried to disaggregate results at the individual level based on disability as well as sex, age group, rural/urban residence and at the household level based on rural/urban residence. Double disaggregation tables by disability and a demographic characteristic (sex, rural/urban, age group) are available in Results Tables.
For data sets with the full population or random sampling, disaggregation is feasible based on sex, age groups, rural/urban as long as information on sex, age and rural/urban residence is available. For data sets with complex survey design, disaggregation based on sex, age groups, rural/urban is feasible if sex, age, rural/urban residence were used as part of the stratification of the survey.
Besides, for each data set and indicator, we set 100 observations as the minimum required to produce estimates for subgroups following common practice (e.g. Duerto Valero 2019). Hence, for a given data set, disaggregation may be possible for some indicators but not others, especially when some indicators are constructed particularly for subsamples: for instance, for employment, for men and women separately, we were able to disaggregate the employment population ratio across both disability and sex, while this was not feasible for the idle rate for youths (individuals ages 15-24) as the sample sizes for disaggregated samples were often fewer than 100 observations.
References
Duerto Valero, S. (2019) Gender data and multi-level disaggregation: an LNOB perspective to SDG monitoring. United Nations Women Accessed May 20th 2021 at: 