Policy Research Working Paper 9626 Inclusive Statistics Human Development and Disability Indicators in Low- and Middle-Income Countries Sophie Mitra Jaclyn Yap Justine Hervé Wei Chen Development Economics Development Data Group April 2021 Policy Research Working Paper 9626 Abstract This paper disaggregates human development indicators outcomes, poverty, food security, exposure to shocks, living across disability status to assess the situation of persons conditions, and assets. At the same time, not all persons and households with disabilities. The paper uses 24 censuses with functional difficulties experience deprivations. There and general household surveys from 21 low- and middle-in- is a gradient in inequalities associated with the degree of come countries. Disability status is measured through functional difficulty: persons with at least a lot of diffi- self-reports of functional difficulties (for example, seeing culty tend to be worse off than persons with some difficulty, or hearing). There are several findings of interest. First, dis- who themselves tend to be worse off than persons with no ability is not rare in low- and middle-income countries. The difficulty. The results in this paper on the prevalence of median prevalence stands at 10 percent among adults ages functional difficulties and their association with socioeco- 15 and older, and at 23 percent among households. There nomic deprivations show that disability should be central to are consistent inequalities associated with disability and, human development policy, data, and research. More work in particular, with respect to educational attainment, work is needed to curb the inequalities associated with disability. This paper is a product of the Development Data Group, Development Economics. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at mitra@fordham.edu. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Inclusive Statistics: Human Development and Disability Indicators in Low- and Middle-Income Countries Sophie Mitra, Jaclyn Yap, Justine Hervé, Wei Chen Originally published in the Policy Research Working Paper Series on April 2021. This version is updated on May 2021. To obtain the originally published version, please email prwp@worldbank.org. JEL classifications: I32, J14, R20 Keywords: Disability; Functional Difficulties; Human Development; Inequalities; SDGs. Acknowledgments. Wei Chen, Justine Hervé, Sophie Mitra and Jaclyn Yap are affiliated with Fordham University. This paper is a product of the World Bank Development Data Group (DECDG)’s Inclusive Data and Statistics Project (P170339), led by Buyant Erdene Khaltarkhuu and Akiko Sagesaka. Wei Chen, Justine Hervé, Sophie Mitra and Jaclyn Yap were consultants at the World Bank for this project. The authors can be reached at mitra@fordham.edu. This study received funding from the World Bank’s Trust Fund for Statistical Capacity Building (TFSCB) and from the Wellspring Philanthropic Fund. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. We are grateful for insightful comments on an earlier draft from Jose Cuesta, Mari-Helena Koistinen, Mitchell Loeb, Daniel Mont and Deepti Raja. 1 Inclusive Statistics: Human Development and Disability Indicators in Low- and Middle-Income Countries Sophie Mitra, Jaclyn Yap, Justine Hervé and Wei Chen 1. Introduction The Sustainable Development Goals (SDGs) endorsed by the United Nations member states have the mission to “leave no one behind”. Monitoring the SDGs is critical. The Inclusive Data Charter signed by the World Bank and other global data partners in 2018 aims to mobilize political support to improve the quality, quantity, availability and financing of disaggregated data to support the ambition of the 2030 Agenda for Sustainable Development of leaving no one behind. More data are required to highlight different situations of various socioeconomic and geographic groups including persons with disabilities. The lack of data and statistics on disability is in part responsible for the lack of attention to disability in development, but there has been growing awareness on the need for data on disability and improved tools to collect data on disability in an internationally comparable manner. The report by the High-Level Panel on the Post-2015 Development Agenda repeatedly refers to disability, including in the call to end extreme poverty and “leave no one behind” and in the call for a “data revolution”: “data gathered will need to be disaggregated by gender, geography, disability, and other categories, to make sure that no group is left behind.” Disability has been included in several SDG targets and as a cross-cutting issue in the 2030 Agenda for Sustainable Development. As argued in the United Nations report (2019), “efforts need to be stepped up to ensure that the goals and targets will be achieved for persons with disabilities too, in line with the Convention on the Rights of Persons with Disabilities (CRPD)”. Using 24 population censuses and general household surveys for 21 low- and middle-income countries (LMICs) that have internationally comparable disability questions, this paper disaggregates human development indicators across disability status to assess the situation of persons with disabilities and disability related inequalities and facilitate future monitoring. We aim to contribute to SDG target 17.18 “to increase the availability of data disaggregated by disability”. Disability status is measured through self-reports of functional difficulties (e.g. difficulty seeing, hearing) as per the Washington Group Short Set (WGSS) or questions deemed similar. The WGSS is a widely accepted and internationally tested tool (Groce and Mont 2017) that can be used for disaggregation based on disability (Loeb 2015, Washington Group on Disability Statistics 2016). Statistics are comparable across disability status within a country and to a large 2 extent across countries. Globally, the evidence on disability prevalence and on disability gaps in education, employment, living conditions and more has grown in recent decades. Regarding prevalence, as part of the Global Burden of Disease (GBD) study (Murray and Lopez 1996), disability prevalence is regularly estimated: it is inferred from data on health conditions and impairments alone using assumptions on distributions of limitations that may result from health conditions and impairments. The World Health Survey has also been used to develop disability prevalence estimates based on core functional domains (seeing, concentrating, moving around, self-care) (Mitra and Sambamoorthi 2014) or based on an index with information on eight domains (vision, mobility, cognition, self-care, pain, interpersonal relationships, sleep and energy, affect) (WHO- World Bank 2011). Regarding disability gaps in well-being outcomes, while most of the evidence a decade ago pertained to high income countries (HICs), there has been growing evidence that persons with disabilities in LMICs are disproportionately more likely to experience multidimensional poverty, i.e. experience multiple deprivations, are less likely to work and to have ever attended school and more likely to experience morbidity (e.g. Banks et al 2017; Mitra 2018; Pinilla-Roncancio et al 2020; United Nations 2019). For instance, in a systematic review on poverty and disability in LMICs, Banks et al 2017 find that 81% of studies found evidence of a positive association between disability and a poverty marker. This relationship persisted when results were disaggregated by gender, measure of poverty used and impairment types. By age group, the proportion of studies reporting a positive association between disability and poverty was lowest for older adults and highest for working-age adults (69% vs. 86%). However, deriving any definitive conclusion on inequalities across disability status remains problematic in this literature that uses varying measures for disability, well-being indicators, data sources, and methodologies. Some studies measure disability through broad activity limitations (e.g., Mitra, 2008) or impairments (Pinilla-Roncancio & Alkire 2020) while more and more use the WG recommended questions (Mitra 2018, Pinilla-Roncancio et al 2020). Studies using impairments (Pinilla- Roncancio & Alkire 2020) have results on the disability gap in outcomes that are less consistent across countries than those using the WG questions (Mitra 2018; Pinilla-Roncancio et al 2020). Several studies (Mitra et al 2013; WHO-World Bank 2011) rely on the World Health Survey (WHS) that was designed to collect a detailed health and disability profile of individuals but provide only summary measures of economic well-being, for instance on employment and household expenditures. Besides, not every individual in a household in the WHS was interviewed, only one individual per household. Hence, differences across disability status may be underestimated for household well-being indicators. Finally, and more importantly, results vary across well-being dimensions, making the evidence mixed. Studies relying on traditional poverty measures based on consumption expenditures or asset ownership data do not find any consistent significant association between disability and poverty, while studies on multidimensional poverty measures tend to (United Nations 2019). This is surprising given the consistent evidence found in HICs, whatever the measure of poverty (United Nations 2019). More research is therefore needed on inequalities that might be associated with disabilities for LMICs using internationally comparable disability questions. The aim of this paper is to contribute to filling this gap. The rest of the paper is organized as follows: Section 2 presents the 3 data and methods, Section 3 gives the results and Section 4 concludes. 2. Data and Methods 2.1 Data Sets Data set selection This study uses general household surveys and censuses rather than data sets that focus on disability, i.e., disability surveys. This stems from the objective to exploit data sets that are collected on a regular basis and may be potentially used to document and track human development for persons with disabilities. Of course, complementary work is needed to use disability-focused data sets, such as the Model Disability Survey developed by the World Bank and the World Health Organization (Cieza et al 2018). We first examined nationally representative general household surveys and population censuses conducted in 136 LMICs (as defined by the World Bank) from 2009 to 2018 (Mitra et al 2021). The questionnaires of general household surveys and censuses were retrieved from online survey databases such as the International Household Survey Network (IHSN) Catalog, or the websites of individual National Statistical Offices. Questions were considered to be similar enough for inclusion in the study if they covered the four domains of functional difficulties recommended by the United Nations (2017, p. 207) for censuses. When functional difficulty questions followed a screener question like “Do you have a disability?” (e.g. Albania LSMS), we did not analyze the data as such screeners lead to under self-reporting. From the data sets reviewed, this paper presents results for 21 LMICs with data sets using the Washington Group Short Set or similar questions where we could access general surveys or census data. We use 24 data sets as follows: eleven censuses (Dominican Republic, Indonesia, Kiribati, Mexico, Panama, Philippines, Rwanda, South Africa, Tanzania, Vanuatu, Vietnam), five Living Standards Measurement Studies (LSMS) (Ethiopia, Malawi, Nigeria, Tanzania, Uganda), four Household Income and Expenditure Surveys (HIES) (Bangladesh, Liberia, Namibia, Papua New Guinea), two Labor Force Surveys (LFS) (The Gambia and Rwanda), and two general surveys (Afghanistan and South Africa). Most data sets are publicly available and are accessible through websites such as those of LSMS or IPUMS International. Some data sets such as Kiribati and Vanuatu’s censuses are not publicly available and were obtained from the respective national statistics offices. The data sets are described in Table 2.1. The data sets have three possible sample designs. Most follow a complex survey design in which case all estimates presented below are weighted. Kiribati and Vanuatu’s censuses have the entire population, while several population censuses have a random 10 percent sample, in which case no weighting was needed. Descriptive statistics are shown in Appendix 1 for the samples of individuals and households. Standard errors are not shown to keep the results legible but are available from the authors upon request. 4 2.2 Conceptual Framework The analysis in this paper needs to be based on concepts that are in line with human rights and sustainable development approaches to disability. Disability is not understood as a purely medical or social phenomenon, let alone through a charity or moral lens (Goodley 2016). Instead, disability is a biopsychosocial phenomenon and is understood as an interactional notion, one that results from an individual with a health condition interacting with structural factors and resources. The analysis in this paper can be framed within the model of the International Classification of Functioning, Disability and Health (ICF) (WHO 2001). Perhaps very apt for a study of human development and disability is the human development model of disability, health, and well-being, an application of Amartya Sen’s capability approach (Mitra 2018). In this framework, human development is framed as an expansion of practical opportunities (capabilities) and achievements (functionings) for all, including for persons with health conditions and functional difficulties. 2.3 Disability Questions and Indicators The United Nations (2017, p. 207) adopted revised guidelines for the collection of disability data in national censuses. It recommends that the following four functional domains be considered essential in determining disability status in a way that can be reasonably measured using a census and that would be appropriate for international comparison: (a) Walking; (b) Seeing; (c) Hearing; (d) Cognition. It also notes that two other domains, self-care and communication, have been identified for inclusion. The data sets under study have questions on functional difficulties from the Washington Group Short Set (WGSS) (Madans et al 2011, Altman 2016) or other functional difficulty questions that meet the UN recommendations for censuses. 5 The WG has recommended a short list of six questions to be included in household surveys or censuses to capture respondents’ self-evaluation. They are presented in Box 1. The questions ask about difficulties in six domains: (a) seeing (even when wearing glasses), (b) hearing (even when wearing a hearing aid), (c) walking/climbing steps, (d) concentrating or remembering things, (e) self-care and (f) communication. The six domains cover four functional difficulties (seeing, hearing, concentrating/remembering, communicating) and two activity limitations (walking and self-care). For simplicity and brevity, we refer to them all as functional difficulties. For each difficulty, individuals respond on a scale of 1 to 4 as follows: 1 - no difficulty, 2 - some difficulty, 3 – a lot of difficulty and 4 - unable to do. The WGSS has the advantage of brevity and international comparability. Albeit cognitively tested in 14 countries (Miller 2016), these questions on functional difficulties are not without limitations. For instance, an understanding of functional difficulties may be limited in a context with limited access to health care (Schneider 2016). This may lead to underreporting in the countries under study. The WGSS covers few domains and may well under-identify people with certain difficulties, including psychological ones. This concern is alleviated in the WGSS-Enhanced, which is the short set plus four questions on anxiety and depression and two on upper body mobility and in the WG Extended Set on Functioning (WG ES-F), which includes questions that address functioning in domains such as upper body (functioning of the arms, hands and fingers), affect (anxiety and depression), pain and fatigue (e.g. Loeb 2016). Box 1: Washington Group Short Set of Questions on Disability The next questions ask about difficulties you may have doing certain activities because of a health problem. (a) Do you have difficulty seeing even if wearing glasses? (b) Do you have difficulty hearing even if using a hearing aid? (c) Do you have difficulty walking or climbing steps? (d) Do you have difficulty remembering or concentrating? (e) Do you have difficulty with self-care such as washing all over or dressing? (f) Using your usual language, do you have difficulty communicating, for example understanding or being understood? For each question, respondents are asked to answer with one of the following options: 1-no difficulty, 2-some difficulty, 3-a lot of difficulty, or 4-unable to do. For a proxy respondent, each of the six questions starts with “does have difficulty…?” Source: http://www.washingtongroup-disability.com/ While the WG recommends that the WGSS be adopted as is, some data sets have included selected questions only, changed the answer scale or the wording of questions. There are thus data sets that have questions somewhat similar to the WGSS. As shown in Table 2.1, 12 data sets under study have the WGSS, while 12 data sets have other functional difficulty questions that fulfill the United Nations principles and guidelines for censuses on disability (United Nations 2017, p. 207) in that they cover at least the four essential domains. The ways in which these 12 6 data sets have questions on functional difficulties that differ from the WGSS are shown in Table 2.1. Six data sets are censuses that do not cover all six domains: Dominican Republic, Panama, Rwanda, Tanzania, Vanuatu, and Vietnam. The questions could have wording that differed from the WGSS formulation. If they do, they are shown in Table 2.1. For instance, in South Africa’s 2010 Census, the walking question is worded differently: "Walking a kilometer or climbing a flight of steps" and the cognitive domain is covered by two separate questions for remembering and concentrating. Another example is that of the 2010 Census in Indonesia, which grouped the cognitive and communication domains in one question: "Do you have difficulty remembering, concentrating, or communicating with others due to a physical or mental condition?" Five censuses have a yes/no answer scale: Dominican Republic, Mexico, Panama, Philippines, Rwanda. 1 Meanwhile, Indonesia has only three-level scale: ‘no difficulty’, ‘slight’, and ‘severe’. Tanzania, and Vietnam’s answer scales refer to ‘a little difficulty’ instead of ‘some difficulty’. For the Papua New Guinea HIES, the answer scale is reversed from that in the WGSS as follows: 1. Cannot at all, 2. A lot of difficulty 3. Some difficulty 4. No difficulty. Only in few data sets (e.g. Ethiopia LSMS) does each individual in the household consistently answer about his/her functional difficulties. In the other countries, it is the household respondent who answers on behalf of every individual in the household. Three HIES, two LFS, five LSMS, and two general surveys (Afghanistan and South Africa) under study have adopted the WGSS questions to the letter. Event for this subgroup of countries, caution is needed while comparing results across countries. What people may understand from the questionnaire and how they reply could differ given different languages, 2, cultures, interviewer training and other contextual factors in ways that are beyond the purview of the researchers. Although the functional difficulty questions are worded the same way, there are a few differences in the questionnaires of these five countries. Ethiopia, Malawi, Nigeria, and Tanzania have the WGSS as part of a longer health section in the questionnaire, while Uganda has a separate section titled ‘disability’ after the health section. Tanzania has a somewhat different answer scale including an additional category, no difficulty with assistive devices as follows: 1 - no difficulty, 2 – no difficulty with assistive devices, 3 - some difficulty, 4 – a lot of difficulty and 5 - unable to do. Categories 2 and 3 were collapsed into one category (no difficulty) for comparability with other countries. Although these differences between surveys may seem minor, such minor changes in question wording or in the placement of the question may significantly affect the resulting estimates (Mathiowetz 2001). Finally, while the WGSS was initially developed for use in censuses among those 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 who are 15 years old and older, and their households. 1 The answer scale is in fact: No difficulty, Partial difficulty, Large difficulty, Complete difficulty. However, the IPUMS data set under use in this study only has a yes/no answer. 2 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. 7 Disability indicators In order to determine prevalence or identify a specific ‘functional difficulty status’ group, a threshold needs to be set on the answer scale of functional difficulties. The WG recommends “a lot of difficulty” as the threshold: persons who report “a lot of difficulty” or “unable to do” for at least one domain are considered to have a disability, and persons with ‘no difficulty’ or ‘some difficulty’ to all six questions are deemed as not having a disability. Different thresholds for the WGSS produce vastly different prevalence estimates (e.g. Bourke et al 2020; Mitra 2018). In addition, if persons with some difficulty are potentially more disadvantaged than persons with no difficulty, this categorization will underestimate the extent of inequalities between persons with and without disability. There is mounting evidence that having ‘some difficulty’ is significantly associated with economic and social deprivations with a gradient from no difficulty to some difficulty to at least a lot of difficulty (Clausen and Barrantes 2020; Mitra 2018). Besides, several censuses under study use yes/no answers to the functional difficulty questions, making the breakdown recommended by the WG impossible. Hence, this study categorizes individuals in three ways. First, for all data sets, it groups individuals into two categories: (1) No functional difficulty in all domains; (2) Any functional difficulty in at least one domain (answer Yes for data sets with yes/no answers, or reports at least ‘some’ difficulty for graded scales). Second, following the recommendation of the WG for data sets that have a graded answer scale, we group individuals as follows: (i) A lot of difficulty or unable to do in at least one domain; (ii) No difficulty or ‘some’ difficulty for all domains Third, for data sets with graded scales, we partition individuals into three categories: (a) No difficulty for all domains; (b) Some difficulty in at least one domain but no “a lot of difficulty” or “unable to do”. (c) A lot of difficulty or unable to do in at least one domain. This categorization is more granular and may be able to identify a possible gradient in socioeconomic disadvantage. The second and third categorizations above are both conducted for the subset of 19 data sets that do not have a yes/no answer scale. We use these three categorizations in our analysis and are presented in the tables as panels 1, 2, and 3 respectively. The analysis conducted at the household level categorizes households depending on the functional difficulty status of its members age 15 and older along the three ways of partitioning the population described above. We use basic proportions to calculate the prevalence of functional difficulties in each country based on the first categorization above for all data sets and based on the second and third categorization for countries with an answer scale. Given that the primary objective of this paper is to make comparisons of development indicators across disability status within a country and not across countries, we do not present age and sex standardized disability prevalence for each 8 country. The prevalence rates at the individual and household (HH) levels are calculated using the formulas: ℎ = where c denotes a country. ℎ ℎ ℎ = ℎℎ 2.4 Human Development Indicators We compare human development 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 inequalities associated with disability. Results are presented in graphs and tables, often in appendix tables. Double disaggregation tables by disability and a demographic characteristic (sex, rural/urban, age group) are available from the authors upon request (Appendices A to C). In tables, the disability gap and its statistical significance is typically noted in a separate column. 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, asset ownership) and deprivations (unemployment rate, food insecurity, exposure to shock), positive/negative differences will not consistently reflect a disability gap in achievement/deprivation indicators, respectively. We also use and graph the ratio for a given indicator by dividing the indicator for persons with disability with the indicator for persons with no disability. This is often referred to as a relative rate or as a parity index (e.g. UNESCO 2018, OECD 2003). For instance, for the relative employment rate, a relative employment rate at or near one reflects a similar situation of persons with and without disabilities in terms of the share who are employed. The indicators under study are listed in Table 2.2. To compile this list of indicators, we started from the list of SDG indicators of the Inter-secretariat Working Group on Household Surveys (2018) in a Report to the 49th Session of the UN Statistical Commission (doc CN.3/2018/7) which showed that approximately one third of the SDG indicators are or can be compiled from household surveys. From this list of indicators, we selected those indicators that might be found in general household surveys or censuses, since health or other topic specific surveys are beyond the scope of this study. We selected and were able to produce estimates for those in Table 2.2 (e.g. unemployment rate, out of school rates for children in the household). We then added indicators that are known to be particularly suited to assess the situation of persons with disabilities and could be estimated with some of the data sets above: employment population ratio, economic insecurity (food insecurity, exposure to shocks) and indicators that may reflect the extra costs of living with disabilities for households (poverty status once health expenditures are removed from total consumption expenditures), as well as material well-being 9 indicators (assets, living conditions) that might be affected due to the extra costs of living with disabilities. Some of these indicators were not found in the data sets under consideration and are thus not listed in Table 2.2 (for instance, Household expenditures on care as a share of total household expenditure and Information, Communication, and Technology (ICT) skills among adults). For education, we use “Ever attended school’ as a first indicator for all adults and also for adults age 15 to 29 following UNESCO (2018) 3. For school completion rates, we had to deviate from current international practice focused on specific age cohorts (e.g. secondary completion rate for adolescents age 15 to 17) due to small sample sizes for some surveys (UNESCO 2018). Instead, we group all adults age 15 and older in four mutually exclusive categories: less than primary school, primary school completion, secondary school completion and higher than secondary school. It should be noted that countries may have different thresholds for each type of schooling. For instance, for the completion of primary school, the years primary school requires may vary from five to eight years in the different countries. At the household level, we categorize the educational attainment of the household head according to the four categories above and present the mean educational attainment of household heads across functional status. At the household level, we calculate the percentage of children in the household between the age of six and 14 who are out of school and for data sets with a consumption expenditure module, the share of education expenditures out of total consumption expenditures. For employment, this paper uses six indicators: employment population ratio, unemployment rate, idle rate among youths, shares of workers in the informal sector, in the manufacturing sector and in managerial positions. In four data sets, the unemployment rate could not be calculated as there is no information on the job search status of individuals (Ethiopia, Philippines, Tanzania (LSMS) and Uganda). The idle rate is commonly used for youths age 15 to 24 and reflects the share of youths who are not in school, not working, and not seeking employment. As job search information was not available in all countries, we did not include it in the idle rate. The share of workers in the informal sector was calculated based on information collected in some surveys on individuals’ employment status, whether they have a contract with their employer, and information provided on employers (e.g.business registration status). Disaggregation There may be patterns of intersectional disadvantage that affect subgroups of people with disabilities and their households, such as women, rural residents, or indigenous peoples. Identifying sub-groups within this population is policy-relevant as many LMICs have social policies targeted at the groups that are particularly exposed to vulnerability. For each data set under consideration, we considered further stratifying results at the individual level based on sex, age group, rural/urban residence and at the household level based on rural/urban residence. For data sets with the full population or random sampling, disaggregation is feasible based on sex, age groups, rural/urban. For data sets with complex survey design, disaggregation based on sex, age groups, rural/urban is feasible if sex, age, rural/urban residence 3 An adult who has not been to school at all or only attended nursery school is considered as not having ever attended school. 10 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 and not others, especially when some indicators require focusing of 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 unemployment rate, i.e. the share of the unemployed out of the labor force, the idle rate for youths, and the share of workers in manufacturing or managerial positions as the sample sizes for disaggregated samples were often fewer than 100 observations. Results of further disaggregation based on disability and other characteristics in Appendices A to C are available from the authors. 11 12 3. Results and Discussion Disability Prevalence Disability prevalence estimates are shown in Figure 3.1 and Appendix 2. The top panel of Figure 3.1 gives the prevalence of any difficulty at the individual level from the lowest to the highest. The prevalence of any difficulty ranges from a low of 4% in the Philippines to a high of 28% in Papua New Guinea. The median among the 24 data sets stands at 10%. Functional difficulties tend to be more common in rural areas, affect older individuals more, as well as women more often than men (Appendix 6 and 7). 4 Figure 3.1: Prevalence of functional difficulties (%) Among Adults 15+ 30 25 20 15 10 5 0 Any Difficulty (yes/no) At least a lot of difficulty Some difficulty Among Households 60 50 40 30 20 10 0 Any Difficulty (yes/no) At least a lot of difficulty Some difficulty 4 The case of South Africa will require a closer look as results from the Census and the General Household Survey are different in terms of prevalence, especially by rural/urban areas. 13 In the 19 data sets with an answer scale, having “some difficulty” is more common than “at least a lot of difficulty”. The prevalence of at least a lot of difficulty among adults ranges from under 1% in Vanuatu to 6% in Papua New Guinea and the median stands at about 2%. The median prevalence for some difficulty is at about 9%. There is more variation in prevalence when it is measured based on functional difficulty questions other than the WGSS compared to the WGSS. The data sets with other functional difficulty questions tend to be at the lower end of the distribution (e.g. Philippines, Panama, Indonesia) and also at the maximum (Papua New Guinea). Prevalence estimates based on the WGSS range from 7% in Nigeria to 19% in Namibia. The bottom panel of Figure 3.1 gives the prevalence of any difficulty at the household level from the lowest to the highest. The median household level prevalence is at 23%. It ranges from a low of 9% in the Philippines to a high of 55% in Papua New Guinea. Figure 3.2: Distribution of functional difficulties across domains among adults 15+ with any difficulty (%) for datasets with the WGSS 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Seeing Hearing Moving Cognitive Selfcare Communication For countries with two datasets under study (Rwanda (Census and LFS), South Africa (Census and GHS), Tanzania (Census and LSMS)), prevalence rates differ across datasets, perhaps due to different disability questions under use. For instance, in Rwanda, the prevalence rate is 9% with the LFS with the WGSS questionnaire while it stands at 5.5% with the Census which has other functional difficulty questions with yes/no answers. Figure 3.2 gives the distribution of difficulties by domain for countries with the WGSS. Seeing and mobility difficulties tend to be the most common functional difficulties. Appendix 3 also has the distribution across domains broken down for some difficulty in the top panel and at least a lot of difficulty in the bottom panel. 14 Figure 3.3 shows prevalence estimates for any difficulty disaggregated by sex. In most countries, prevalence is higher among women compared to men. Appendix 4 shows that differences in prevalence between men and women do vary by level of difficulty and are larger for some difficulty compared to at least a lot of difficulty. Results from Appendices 4 and 5 suggest that functional difficulties tend to be more common among older age groups and that in most countries functional difficulties are more prevalent in rural areas than in urban areas. Figure 3.3: Prevalence of functional difficulties among men and women age 15+ 15 Discussion The median prevalence of any functional difficulty among adults age 15 and older stands at 10% in the countries under consideration. This result is overall consistent with earlier estimates of prevalence among adults for LMICs (WHO-World Bank 2011; Mitra and Sambamoorthi 2014). The prevalence of at least a lot of difficulty ranges from under 1% in Vanuatu to 6% in Papua New Guinea. The median household level prevalence is at 23%, which is consistent with earlier estimates (e.g. Mitra 2018). Overall, these results show that functional difficulties affect sizeable shares of individuals and households in LMICs. They tend to affect older individuals and people in rural areas more, as well as women more often than men. More research on gender is needed given the higher prevalence found among women in this study and in other studies (WHO-World Bank 2011; Mitra and Sambamoorthi 2014). Functional difficulties are significantly associated with aging in the countries with increasing prevalence for older age groups. More research is also needed on older adults in LMICs, for whom knowledge about their situation is scarce. The prevalence estimates in this paper likely offer a lower bound estimate of prevalence given that only four to six functional difficulties are measured. More data collection efforts are needed for instance using the WG Enhanced short set or the Extended set of questions of the WG to capture psychosocial and mental health-related functional difficulties. Further, the use of the WGSS or similar questions can address issues of prevalence and, through disaggregation, inequality. They do not, however, address questions regarding why such differences might exist. Surveys that can collect detailed information on the environment such as the Model Disability Survey (Cieza et al 2018) would provide information to help understand the determinants of functional difficulties. Because functional difficulties affect sizeable shares of individuals and households in LMICs, a study of the correlation of such difficulties with inequalities is thus warranted and is presented in the rest of this paper. Educational attainment Next, we present several educational attainment indicators. We start with the share of adults who have ever attended school in Figure 4.1. Across the three disability categorizations, results consistently point at significantly lower rates of having ever attended school for adults with functional difficulties. For instance, in Indonesia, 57% of persons with any difficulty have ever attended school compared to 92% of persons with no difficulty. 16 Figure 4.1: Share of individuals who have ever attended school by functional difficulty status Afghanistan Bangladesh Dominican Rep. Ethiopia Gambia Indonesia Kiribati Liberia No Difficulty Malawi Mexico Namibia Any difficulty Nigeria Panama Papua New Guinea Philippines Rwanda (Census) Rwanda (LFS) South Africa (Census) South Africa (GHS) Tanzania (Census) Tanzania (LSMS) Uganda Vanuatu Vietnam 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 17 Afghanistan Bangladesh Ethiopia Gambia Indonesia Kiribati Liberia Malawi Namibia Nigeria Papua New Guinea Rwanda (LFS) South Africa (Census) South Africa (GHS) Tanzania (Census) Tanzania (LSMS) Uganda Vanuatu Vietnam 0 0.2 0.4 0.6 0.8 1 1.2 No or some Difficulty At least a lot of difficulty Afghanistan Bangladesh Ethiopia Gambia Indonesia Kiribati Liberia Malawi Namibia Nigeria Papua New Guinea Rwanda (LFS) South Africa (Census) South Africa (GHS) Tanzania (Census) Tanzania (LSMS) Uganda Vanuatu Vietnam 0 0.2 0.4 0.6 0.8 1 1.2 No Difficulty Some difficulty At least a lot of difficulty 18 For the 19 data sets with graded answer scales, we can compare the extent of the disability gap for the three disability categorizations. Comparing the top and middle graph, the first categorization (No difficulty vs Some difficulty) has less of a disability gap in ever attending school than the second categorization (No or some difficulty vs At least a lot of difficulty). For instance, in Namibia, the disability gap stands at 14 percentage points for No difficulty vs Some difficulty compared to 23 percentage points for No or some difficulty vs At least a lot of difficulty. The bottom graph in fact shows that there is a gradient with persons with at least a lot of difficulty having less often attended school than persons with some difficulty, who in turn have less often attended school than persons with no difficulty. In Figure 4.2 and Appendices 4 and 9, educational attainment is presented in four mutually exclusive categories: 1/ less than primary school completion, 2/ primary school completion, 3/ secondary school completion and 4/ more than secondary school. In all countries, persons with functional difficulties tend to be overrepresented at the lowest educational level, i.e. less than primary school, while they are underrepresented at higher educational levels. Figure 4.2: School completion rates across 'Any functional difficulty' status Less than primary school completion No Difficulty Any difficulty Afghanistan Bangladesh Dominican Rep. Ethiopia Gambia Indonesia Kiribati Liberia Malawi Mexico Namibia Nigeria Panama Papua New Guinea Philippines Rwanda (Census) Rwanda (LFS) South Africa (Census) South Africa (GHS) Tanzania (Census) Tanzania (LSMS) Uganda Vanuatu Vietnam 0 0.2 0.4 0.6 0.8 1 19 Primary School Completion No Difficulty Any difficulty Afghanistan Bangladesh Dominican Rep. Ethiopia Gambia Indonesia Kiribati Liberia Malawi Mexico Namibia Nigeria Panama Papua New Guinea Philippines Rwanda (Census) Rwanda (LFS) South Africa (Census) South Africa (GHS) Tanzania (Census) Tanzania (LSMS) Uganda Vanuatu Vietnam 0 0.2 0.4 0.6 0.8 1 20 Secondary School Completion No Difficulty Any difficulty Afghanistan Bangladesh Dominican Rep. Ethiopia Gambia Indonesia Kiribati Liberia Malawi Mexico Namibia Nigeria Panama Papua New Guinea Philippines Rwanda (Census) Rwanda (LFS) South Africa (Census) South Africa (GHS) Tanzania (Census) Tanzania (LSMS) Uganda Vanuatu Vietnam 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 21 Higher than Secondary School No Difficulty Any difficulty Afghanistan Bangladesh Dominican Rep. Ethiopia Gambia Indonesia Kiribati Liberia Malawi Mexico Namibia Nigeria Panama Papua New Guinea Philippines Rwanda (Census) Rwanda (LFS) South Africa (Census) South Africa (GHS) Tanzania (Census) Tanzania (LSMS) Uganda Vanuatu Vietnam - 0.20 0.40 0.60 0.80 1.00 Table 4.1 presents several educational indicators at the household level for the three disability categorizations. It starts with the share who ever attended school for the household head with and without adults with functional difficulties. Households with functional difficulties have heads with a significantly lower share who ever attended school in all but two countries (Afghanistan and The Gambia). As found earlier for individuals in Figure 4.1, panel 3 of Table 4.1 shows that there is a gradient in the educational attainment of household heads between No difficulty, Some difficulty and At least a lot of difficulty. Results in Table 4.1 also indicate that there is no consistent and significant difference in terms of children out of school rates depending on the household’s functional difficulty status nor in terms of the share of household consumption expenditures dedicated to education. 22 23 Discussion The results from Figures 4.1 and 4.2 and Appendix 4 are consistent with extensive and consistent evidence that adults with disabilities have lower educational attainment in LMICs, as recently reviewed in United Nations (2019). This association consistently found among adults may result from lower school attendance among children with disabilities (Filmer 2008; Mizunoya et al 2018). It could also be due to less education putting people at higher risk of getting a functional difficulty or being correlated with factors making people at risk of functional difficulty such as malnutrition, lack of access to health care, and risky working conditions. More research is needed to investigate the link between low educational attainment and functional difficulties, notably when information is available on age at onset (Mont 2020). The lack of a consistent and significant difference in terms of children out of school rates and education expenditures depending on the household’s functional difficulty status differs from results from a small but growing literature on the negative relationship between parental disability and children’s educational outcomes (e.g. Mont and Nguyen 2013). The data sets used in this study could be used in further research on parental disability and children’s school enrollment and outcomes. Employment Several employment and work related indicators are presented in Table 4.2: the employment population ratio (EPR), the unemployment rate, the idle rate for youths 15 to 24, and several indicators related to the type of work. Disaggregation by sex, age groups and rural/urban areas are in Appendix B. The EPR is lower among persons with functional difficulty in almost all countries in the three panels. In panel 1, it is significantly lower for adults with any difficulty compared to other adults in all countries/datasets except in Papua New Guinea and Vanuatu. The disability gap in the EPR is the largest in Vietnam at 37 percentage points where the EPR stands at 80% and 43% respectively for persons with no difficulty and with any difficulty. Moving to panel 2, the EPR is consistently significantly lower for adults with at least a lot of difficulty compared to adults with no or some difficulties. It is noteworthy that compared to panel 1, the disability gap in the EPR tends to be larger in panel 2 where the employment of persons with at least a lot of difficulty is compared to that of persons with no or some difficulty. In panel 3, the EPR is consistently significantly lower for adults with at least a lot of difficulty compared to adults with no difficulty. In panel 3, comparing persons with no difficulty and persons with some difficulty, results are mixed: persons with some difficulty have significantly lower EPRs in six countries (Afghanistan, Bangladesh, Indonesia, Rwanda, South Africa (GHS), Vietnam) and higher in three countries (South Africa (Census), Papua New Guinea, Vanuatu). Finally, in panel 3, in the six countries where persons with some difficulties and persons with at least a lot of difficulties have significantly lower EPRs, there is a gradient in the difference in employment population ratio with larger gaps for persons with at least a lot of difficulty compared to persons with some difficulty. Within countries, the disability gap in EPRs may vary between females and males, across age groups, and urban/ rural areas. Across countries, there is no regular pattern for sex and area (Appendices B1 and B2 respectively): for instance, in some countries it is higher in rural areas (e.g. Gambia) while in others it is larger in urban areas (e.g. 24 Panama). The disability gap in EPR tends to be larger for older age groups (Appendix B3). The EPR is further analyzed in Figure 4.3, where the relative employment rate is pictured: it is the EPR for persons with functional difficulties divided by the EPR for persons with no difficulty. For instance, in Afghanistan, the relative employment rate is 76%, meaning that the EPR of persons with any difficulty is 76% that of persons with no difficulty. In Figure 4.3, the relative employment rate ranges from 43% in Panama to 100% in Liberia, Vanuatu, and Papua New Guinea. For countries with a graded scale for functional difficulty questions, Figure 4.3 shows that the relative employment rate is consistently lower for persons with at least a lot of difficulty compared to persons with some difficulty. Figure 4.3: Relative employment rates 1.00 0.90 0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 - Any difficulty Some difficulty At least a lot of difficulty Table 4.2 presents other employment related indicators. The unemployment rate could be calculated for countries that had information on job search status. The unemployment rate is significantly higher among persons with any difficulty for six countries (Afghanistan, Dominican Rep., Ethiopia, Mexico, Rwanda (Census), Tanzania (LSMS)) but lower in eight countries (Indonesia, Kiribati, Namibia, Panama, Rwanda (LFSO), South Africa (Census and GHS), Vanuatu and Vietnam. Coupled with often significantly lower employment population ratios described above, this reflects that persons with difficulties are more likely to be both out of work and out of the labor force, and thus not captured in the unemployment rate. Persons with functional difficulties may disproportionately drop out of the labor force by stopping to search for work and hence not be counted as unemployed. The idle rate for youths aged 15 to 24 is significantly higher for youths with functional difficulties in 15 countries, with the largest disability gap in Vietnam. The disability gap is larger for persons with at least a lot of difficulty compared to persons with some difficulty. For Papua New Guinea, results in panel 3 shows that the idle rate for youths is significantly higher for youths with at least a lot of functional difficulty but lower for youths with some difficulty, explaining the lower rate for youth with any difficulty in panel 1. 25 26 For most countries, the type and sector of employment is available and the share of workers in informal work, in the manufacturing sector and in managerial positions are presented in Table 4.2. Considering the results across the three panels, persons with any limitation are found to more often do informal work and less often in manufacturing work and managerial positions. For informal work, the disability gap is statistically significant in 14 countries in panel 1. For manufacturing work and managerial positions, the disability gap is small. It should be noted that the share of workers doing manufacturing work or in a managerial position is generally low in the countries under study. 27 Discussion In an agrarian economy, as is the case in some LMICs under study, many jobs are in the primary sector (agriculture, forestry, mining) and may involve heavy manual labor, which people with physical difficulties may not be able to do. People with hearing or cognition difficulties, on the other hand, may not experience barriers to physical labor. The effect of disability on employment will also depend on the workplace, its accessibility, available accommodations and transport, and whether there may also be discrimination that might prevent access to employment and/or might lead to lower wages. The policy context is also relevant; for instance, vocational rehabilitation, disability insurance or social assistance programs could facilitate, limit, or not affect access to employment for disabled people depending on how they are designed and implemented. Results in this paper are consistent with those from several studies in LMICs finding that persons with disabilities are less likely to be employed and are more often in the informal sector (WHO- World Bank 2011; Mizunoya and Mitra 2013) 5 and that the relative employment rate tends to be higher in low-income countries compared to middle-income countries (Mizunoya and Mitra 2013). People with disabilities may be more often in the informal sector due to barriers they face in the formal labor market. This study further adds to the literature by showing the functional difficulty gradient in the disability gap and the higher idle rate among youths with disabilities. Finally, this study shows that the unemployment rate, although it is part of SDG indicator to be disaggregated by disability status (indicator 8.5.2), should be used with caution and should be complemented with other labor market outcomes such as the employment population ratio. The unemployment rate is significantly higher among persons with any difficulty in six countries but lower in eight countries while the EPR was lower for persons with disabilities in almost all countries. Persons with functional difficulties may disproportionately drop out of the labor force by stopping to search for work and hence are not captured in the unemployment rate. Poverty Table 4.3 gives poverty headcounts using international poverty lines of $1.90, $3.20, $5.50. It also gives the headcount using $1.90 but applying the household consumption expenditures per capita net of health expenditures. Poverty headcounts were calculated based on daily household consumption aggregates divided by household size to obtain local currency and nominal daily per capita consumption expenditures (PCE). PCE were adjusted using PPP (Purchasing Power Parity) conversation rates of the International Comparison Program (2011) for each country. Poverty headcounts are close to published poverty headcounts, except for Uganda where our estimates are higher at more than 60%. 6 Using the international poverty lines of $1.90, $3.20, $5.50, poverty headcounts are significantly higher among persons with functional difficulties in some countries (Ethiopia, Namibia, Nigeria, Tanzania, Uganda). However, there are variations depending on the threshold for the poverty line 5 For a recent review of the literature, see United Nations (2019). 6 For Afghanistan, the international poverty rate is not available. But our computation is larger compared to the national poverty rate. 28 and the results are sensitive to the disability categorization. The differences tend to be larger in panels 2 and 3 compared to panel 1. For some countries, the difference across disability status becomes larger once poverty headcounts are calculated based on consumption expenditures net of health expenditures. For instance, for Tanzania, in panel 2, the difference increases from 7 percentage points to 11 percentage points between persons with no difficulty and persons with at least a lot of difficulty. Discussion In some countries, we find that households with disabilities are more likely to experience consumption poverty with some variation depending on the categorization of disability and consumption poverty. This result is consistent with the mixed results on the association between consumption poverty and disability in the literature in contrast with a more consistent association between multidimensional poverty and disability (Mitra, Posarac and Vick 2013; United Nations 2019). A first cross-country study of LMICs by Filmer (2008) using a variety of data sets with various disability measures finds using a logistic regression model that in eight out of 12 countries, disability in adulthood is associated with a higher probability of being in poverty, where poverty refers to belonging to the lowest two quintiles in terms of household expenditures or asset ownership. Another cross-country study using the World Health Survey data for 15 29 LMICs (Mitra, Posarac and Vick (2013) find a significant difference in household per capita expenditures across disability status in only three countries. This paper’s result needs to be contextualized with the well-established literature on the socioeconomic determinants of health inequalities (Marmot 2005) and with growing evidence of the causal impact of disability on earnings (Mani et al 2018), consumption expenditures and assets (Takasaki 2020). Several issues should also be noted with regard to using household expenditures as a dimension of economic well-being in the context of this study. First, there is a possibility that the intra- household distribution of expenditures is unequal across disability status, which is not captured here. Second, if poverty is measured through overall or non-health per capita expenditures against a poverty line as done in Table 4.3, the comparison of households with a disability to other households may be biased because households with disabilities may have additional expenditures due to the disability (for example, transportation, personal assistance). Consumption-based poverty indicators used above do not account for extra costs related to disability in order to provide a more accurate assessment of poverty among persons with disabilities (Mont and Nguyen 2017). More research is needed on the links between disability and monetary poverty, including on extra costs of living with a disability and the dynamics of household expenditures, which may be affected by different patterns of household formation or mortality in relation to disability (Mitra 2018). Social Protection, Economic Insecurity, Living Conditions and Assets Table 4.4 presents several household indicators for households with and without any difficulty. Appendix 8 gives results on the same indicators but at the individual level, i.e. whether persons with and without any difficulty live in households with specific household level outcomes such as receiving social protection or having safely managed drinking water. It should be noted that in Appendix 8 a given household will be counted as many times as it has adult household members with or without functional difficulty. Results are overall consistent with the household level results in Table 4.4 but sometimes vary for some country/indicator pairs. Results in Table 4.4 in the three panels indicate that households with functional difficulties more often receive social protection benefits in eight out of nine countries, with no statistically significant difference in Nigeria. In the nine countries with social protection information (Figure 4.4 and Appendix C1), the difference across disability status in the receipt of social protection is larger for men in Bangladesh, Liberia, Malawi, Namibia, Nigeria, and South Africa. For instance, in Bangladesh, the share receiving social protection is 12 percentage points higher for males with functional difficulties compared to males without, and nine percentage points higher for females with functional difficulties compared to women without. 30 Figure 4.4: Share of households receiving social protection No difficulty Some difficulty At least a lot of difficulty in one domain 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 Bangladesh Ethiopia Liberia Malawi Namibia Nigeria Papua New South Africa Tanzania Guinea (GHS) (LSMS) In the countries with information on food insecurity, except for Afghanistan, households with functional difficulties are more prone to food insecurity, no matter the level of functional difficulty. For instance, in Malawi, the share of food insecure households respectively stands at 52%, 57% and 67% for households with no, some or at least a lot of difficulty. Similarly, for Tanzania, the share of food insecure households stands at 46%, 59% and 69% for households with no, some or least a lot of difficulty. Food insecurity tends to affect the households where women with disabilities live even more than those of men with disabilities Appendix C1. Likewise, in all the countries with information on shocks, persons with disabilities are more often in households that have been exposed to shocks recently. The differences in terms of exposure to shocks are consistently sizeable and significant. Similarly, in all countries with information on health expenditures, households with functional difficulties have a higher share of consumption expenditures dedicated to health expenditures across the three panels. For instance, in Tanzania, households with any difficulty spend 5% of total household expenditures on health compared to 3% for households with no difficulty. Considering living conditions one by one, i.e. water, electricity, clean fuel, floor/roof/wall, the difference in the share of households with a satisfactory living condition is statistically significant across disability status in most countries with a lot of variation in the size of the disability gap. Sanitation is a living condition for which results are mixed. In several countries in Table 4.4 and Appendix 8, persons with functional difficulties have a significantly higher share of access to improved sanitation services In most countries, households with functional difficulties have a significantly lower share owning assets. In four countries (Afghanistan, Dominican Republic, Philippines, South Africa 31 GHS), households with functional difficulties have a higher share who own assets. However, this result does not hold for Dominican Republic and the Philippines in Appendix 8 when the analysis is done at the individual level and for South Africa, inconsistent results are reached across datasets (GHS and Census). For cell phone ownership, households with any difficulty have a significantly lower rate of ownership except in Papua New Guinea. Finally, in six countries where transportation expenditures are available, households with functional difficulties are not found to have a higher share of transportation expenditures out of total expenditures. 32 33 34 35 Discussion Although households with functional difficulties more often receive social protection benefits in eight out of nine countries, households with functional difficulties are disadvantaged in terms of food security, exposure to shocks, living conditions and assets. This result contributes to the growing literature on the association between disability and economic deprivations (United Nations 2019). With respect to asset ownership, this literature had somewhat mixed results so far. Several studies show a higher prevalence of disability among households with fewer assets (Hosseinpoor et al 2013) and that households with disabilities have fewer assets and worse living conditions compared to other households (e.g., Mitra 2018 (Ethiopia, Malawi, Tanzania, Uganda)); Palmer et al. 2012 (Vietnam); United Nations 2019; World Bank 2009 (India)). At the same time, Trani and Loeb 2010 find no significant difference in Afghanistan and Zambia, and Mitra, Posarac and Vick (2013) find a significant difference in the rate of asset deprivation in only four of 15 LMICs. Our results add to this literature by showing for most countries a significant correlation between having a functional difficulty and having fewer assets and worse living conditions. More research is needed on what may explain such differences, and notably if they result from extra costs of living with a disability and resulting challenges in accumulating assets. 4. Conclusion Disability and its socio-economic implications have received very limited attention in the context of LMICs. This paper adds to the literature with several main findings and implications for policy, further research and data collection and monitoring. Main findings In the countries under study, disability is not rare among adults and their households. Difficulties seeing and walking are the most common functional difficulties. While individuals experience various degrees of functional difficulties, having some difficulties is more common than having at least a lot of difficulty. There is a significant and consistent association between functional difficulties and socioeconomic deprivations. Among adults, functional difficulties are significantly associated with lower educational attainment, lower employment population ratios, a higher youth idle rate and a higher share of workers in the informal sector. Households with functional difficulties are consistently more likely to be economically insecure (food insecurity and being exposed to shocks). They also tend to have worse living conditions and fewer assets. Results tend to hold no matter how degrees of functional difficulties are split: (1) No difficulty vs Any difficulty, (2) No and some difficulty vs At least a lot of difficulty and (3) No difficulty vs Some difficulty vs At least a lot of difficulty. The magnitude of the disability gap varies considerably across socioeconomic indicators and countries. 36 For countries with two datasets under study (Rwanda (Census and LFS), South Africa (Census and GHS), Tanzania (Census and LSMS)), results were similar on education and work indicators. However, results varied in terms of prevalence rates and disparities for some living conditions, perhaps due to different disability questions under use or other differences across datasets. Implications Results overall suggest that disability matters when it comes to human development policy. Policy work is needed to curb the stark inequalities across functional status shown in this paper. In the context of calls to “leave no one behind” in the SDGs, this paper shows some of the gaps that need to be addressed and closed. The microdata used in this study are rich and yet have limitations for the purpose of this study. They focus on a small set of development indicators, often economic in nature. For instance, there is no information on individual subjective well-being, nor on social connections. Besides, one limitation of the data sets under consideration is that they only cover the household population in each country. They exclude the homeless and the institutionalized population such as people in nursing homes or psychiatric hospitals. This is problematic as functional difficulties may make individuals less likely to live in a household. While institutionalization among adults is suspected to be low in the countries under study, no data could be found to confirm this. Homelessness may be a more significant problem in the countries under study and could affect the results. This study uses general household surveys and censuses and not data sets that focus on disability. This stems from the objective to exploit data sets that are collected on a regular basis and may be potentially used to document and track human development for persons with disabilities. Complementary research is needed to use disability focused datasets, such as the Model Disability Survey developed by the World Bank and the World Health Organization. This work is needed notably to highlight the structural factors (e.g. social norms, attitudes, physical environment) that persons with disabilities may face in the workplace or in school settings for instance. In addition, information is lacking considerably, and more data are needed on health conditions that may lead to functional difficulties and/or socioeconomic deprivations. Data collection efforts that collect information on health conditions such as the Study of global AGEing and adult health (SAGE) are steps in this direction. Despite the limitations of the data sets under use, this study has several implications for further research and data collection and monitoring. A measure of functional difficulties such as the WGSS should be included as a standard correlate in studies of inequalities. As it would be inconceivable not to include age or gender as correlates, applied researchers should at least include a measure of disability based on the WGSS or similar questions as a potential correlate of socioeconomic outcomes. Social protection programs were found to disproportionately reach persons with functional difficulties in some countries and yet they do not do away with inequalities. Policies and programs that some LMICs have adopted after ratifying the CRPD need to be assessed with 37 respect to their impact on people with disabilities overall and by sex, age group, residence and other potential factors of intersectional disadvantage. More research is needed on program or policy evaluations. Not all such assessments need to be quantitative and large scale in nature. Qualitative, mixed methods and participatory studies are necessary on social protection programs and more broadly to complement the statistics produced in this paper by trying to understand the results in their complex contexts and by listening to voices and perceptions (e.g. Chaudhry 2019, Diaz et al (2015), Kuper et al 2016). Consistent with earlier research (Washington Group on Disability Statistics (2016)), this paper shows that data sets with the WGSS and similar questions can be used to monitor the disability related targets of the 2030 Agenda. The Washington Group’s recommendation is to consider those who have at least a lot of difficulty in at least one functional domain as a threshold for disability. Our results suggest that analyses should try to incorporate the degree of functional difficulties through different degree categories or a functional score (Mitra 2018; Loeb 2020). Analyses that focus on persons experiencing a lot of difficulty leave out persons with some difficulty who, as shown in this paper, are also at risk of deprivations. In addition, the range of development indicators on the radar screen of international and national organizations to monitor the SDGs for persons with disabilities should be expanded to more comprehensively monitor disability inequalities. Results in this paper suggest that considering indicators in addition to the SDG indicators such as the employment population ratio is necessary to monitor disability inequalities. Further, there is a need to include information on environmental barriers that could answer the question of why disparities exist. Finally, for progress to continue in research and data analysis, internationally comparable disability questions need to become standard in censuses and general household surveys in LMICs, as well as in the monitoring systems of NGOs and governments. Results in this paper stress the need to produce disability prevalence and disability disaggregated statistics in LMICs. 38 References Altman, B. M. (ed.). 2016. International measurement of disability: Purpose, method and application, the work of the Washington Group. Social indicators research series 61. Switzerland: Springer International Publishing. Banks, L. M., H. Kuper and S. Polack. 2017. Poverty and disability in low- and middle-income countries: A systematic review. PLoS ONE 12(12): e0189996. Barnartt, S.N. and B.M. Altman (eds.). 2001. Exploring theories and expanding methodologies: where we are and were we need to go. Research in Social Science and Disability Volume 2. JAI Elsevier science. Chaudhry, V. 2019. Centering embodiment in disability research through performance ethnography. Qualitative Social Work. Vol. 18 (5), pp. 754-771. Cieza A, Sabariego C, Bickenbach J, Chatterji S. 2018. Rethinking Disability. BMC Med. Jan 26;16(1):14. doi: 10.1186/s12916-017-1002-6. PMID: 29370847; PMCID: PMC5785824. Clausen, J., and Barrantes, N. 2020. Implementing a Group-Specific Multidimensional Poverty Measure: The Case of Persons with Disabilities in Peru, Journal of Human Development and Capabilities, DOI: 10.1080/19452829.2020.1828316 Díaz Ruiz A., N. Sánchez Durán and A. Palá. 2015. An analysis of the intentions of a Chilean disability policy through the lens of the capability approach. Journal of Human Development and Capabilities. Vol 6(4), 483-500. Duerto Valero, S. (2019) Gender data and multi-level disaggregation: an LNOB perspective to SDG monitoring. United Nations Women Accessed Dec. 20th 2020 at: https://unstats.un.org/sdgs/files/meetings/sdg-inter-workshop-jan- 2019/Session%208.a_UNWomen_Gender%20data%20and%20multi%20level%20disaggregatio n.pdf Filmer, D. 2008. Disability, poverty and schooling in developing countries: results from 14 household surveys. The World Bank Economic Review, 22(1), 141–163. Goodley, D. 2016. Disability Studies: an Interdisciplinary Introduction. London: Sage, 2nd edition. Groce, N. and Mont. D. 2017. Counting Disability: Emerging Consensus on the Washington Group Questionnaires. Lancet Global Health Vol 5(7), pe649-650 Hosseinpoor, A.R., J.A. Stewart Williams, J. Gautam, A. Posarac, A. Officer, E. Verdes and S. Chatterji. 2013. Socioeconomic Inequality in Disability Among Adults: A Multicountry Study Using the World Health Survey. American Journal of Public Health, 103(7), 1278–1286. Intersecretariat Working Group on Household Surveys. 2018. Report to the 49th Session of the UN Statistical Commission (doc CN.3/2018/7). 39 Kuper, H., Walsham, M, Myamba, F., Mesaki, S., Mactaggart, I., Banks, M., and Blanchet, K. 2016. Social protection for people with disabilities in Tanzania: a mixed methods study. Oxford Development Studies, Vol. 44 (4), pp. 441-457. Loeb, M., Mont, D., Cappa, C., De Palma, E., Madans, J., and Crialesi, R. 2018. The development and testing of a module on child functioning for identifying children with disabilities on surveys. I: Background. Disability and health journal, 11(4), 495-501. Loeb, M. 2020. Creating Disability Severity Indicators Using the WG-SS. Washington Group on Disability Statistics. Accessed 01/12/2021 at: https://www.washingtongroup- disability.com/fileadmin/uploads/wg/Documents/WG_Document__5E_- _Analytic_Guidelines_for_the_WG-SS__Severity_Indicators___1_.pdf Loeb, M. 2016. International Census/Survey Data and the Short Set of Disability Questions Developed by the Washington Group on Disability Statistics. In Altman (2016) pp. 255-304. Loeb, M. 2015. Disaggregation by disability: a way forward. Accessed Nov. 1st 2020 at: https://www.cbm.org/news/news/news-2015/disaggregation-by-disability-a-way-forward/ Mani, S., Mitra, S. and Sambamoorthi, U. 2018. Transitions in Health and Employment: New Evidence from Indonesia. World Development Vol 104, pp. 297-30. Marmot, M. 2005. Social determinants of health inequalities. Lancet, Vol. 365, 1099-1104. Mathiowetz, N.A. 2001. Methodological Issues in the Measurement of Persons with Disabilities. In Barnartt and Altman (eds). McClain-Nhlapo, C., V. Costa and M. Tiberti. 2020. Let’s get counting: the Disability Measurement in Household Surveys Guidebook is out! Accessed 01/12/2021 at: https://blogs.worldbank.org/opendata/lets-get-counting-disability-measurement-household- surveys-guidebook-out Miller, K. 2016. Summary of Washington Group question evaluation studies. pp. 69–84. In Altman, B. M. (Ed.). International measurement of disability: Purpose, method and application, the work of the Washington group, pp 69-84. Social indicators research series 61. Springer: Cham. Mitra, S. 2018. Disability, Health and Human Development. New York Palgrave Macmillan US. Mitra, S., Chen, W., Hervé, J., Pirozzi, S., and Yap, J. 2021. Invisible or Mainsteam? Disability in Surveys and Censuses in Low- and Middle-Income Countries. World Bank Policy Research Working paper. Mitra, S., Posarac, A. and Vick, B. 2013. Disability and Poverty in Developing Countries: a 40 Multidimensional Study. World Development Vol. 41; pp.1-18. Mitra, S., and U. Sambamoorthi. 2014. Disability prevalence among adults: Estimates for 54 countries and progress toward a global estimate, Disability and Rehabilitation, 36, 940–947. Mizunoya, S. and Mitra, S. 2013. Is there a Disability Gap in Employment Rates in Developing Countries? World Development Vol. 42; pp.28-43. Mizunoya, S., Mitra, S. and Yamasaki, I. 2018. Disability and school attendance in 15 low- and middle-income countries. World Development. Vol 104, pp. 388-403 (earlier version is Innocenti Working Paper No.2016-03, UNICEF. Mont, D. 2020. When does disability begin? Identifying the age of onset. Washington Group on Disability Statistics. Accessed 12/18/ 2020 at: https://www.washingtongroup-disability.com/wg-blog/when-does-disability-begin-identifying- the-age-of-onset-288/ Mont, D. and C. Nguyen. 2013. Does parental disability matter to child education? Evidence from Vietnam. World Development, 48, 88–107. https://doi.org/10.1016/j.worlddev.2013.04.001 Mont, D. and C. Nguyen. 2017. Spatial Variation in the Poverty Gap Between People With and Without Disabilities: Evidence from Vietnam, Social Indicators Research. DOI 10.1007/s11205- 017-1619-z. Murray, C. J. L., & Lopez, A. D. (Eds.). (1996). The global burden of disease: A comprehensive assessment of mortality and disability from diseases, injuries and risk factors in 1990 and projected to 2020 (1st ed.). Cambridge: Harvard University Press. NBS. (2008). Disability survey report. Tanzania National Bureau of Statistics. OECD 2003. Transforming Disability into Ability Policies to Promote Work and Income Security for Disabled People. Organisation for Economic Cooperation and Development. Palmer, M., Thuy, N., Quyen, Q., Duy, D., Huynh, H., & Berry, H. 2012. Disability measures as an indicator of poverty: a case study from Viet Nam. Journal of International Development. Vol 24(S1), pp. S53-S68. Pinilla-Roncancio, M., Mactaggart, I., Kuper, H., Dionicio, C., Naber, J., Murthy, G., and Polack, S. 2020. Multidimensional poverty and disability: A case control study in India, Cameroon, and Guatemala. SSM - population health, 11, 100591. https://doi.org/10.1016/j.ssmph.2020.100591 Pinilla-Roncancio, M., & Alkire, S. 2020. How Poor Are People With Disabilities? Evidence Based on the Global Multidimensional Poverty Index. Journal of Disability Policy Studies. https://doi.org/10.1177/1044207320919942 Schneider, M. 2016. Cross-national issues in disability data collection. In Altman, B. M. 41 (Ed.). International measurement of disability: Purpose, method and application, the work of the Washington group, pp 15-28. Social indicators research series 61. Springer: Cham. Takasaki, Y. 2020. Impacts of disability on poverty: Quasi-experimental evidence from landmine amputees in Cambodia, Journal of Economic Behavior & Organization, Vol.180, pp 85-107. Tiberti, M. and V. Costa 2020. Disability Measurement in Household Surveys A Guidebook for Designing Household Survey Questionnaires. LSMS guidebook. Washington, D.C.: World Bank. Accessed on 06/24/2020 at: http://documents.worldbank.org/curated/en/456131578985058020/pdf/Disability-Measurement- in-Household-Surveys-A-Guidebook-for-Designing-Household-Survey-Questionnaires.pdf Trani, J. and Loeb, M. 2010. Poverty and disability: a vicious circle? Evidence from Afghanistan and Zambia. Journal of International Development. UNESCO 2018. Education and Disability: Analysis of Data from 49 Countries. Information paper N.49. United Nations Education Scientific and Cultural Organization. United Nations 2015. Report on the forty-sixth session United Nations, Statistical Commission, (New York, 3-6 March 2015). Available from: https://unstats.un.org/unsd/statcom/46th- session/documents/statcom-2015-46th-report-E.pdf. United Nations. 2017. Principles and Recommendations for Population and Housing Censuses. United Nations Department of Social and Economic Affairs. ST/ESA/STAT/SER.M/67/Rev.3 United Nations. 2019. Disability and Development Report 2019. United Nations. https://social.un.org/publications/UN-Flagship-Report-Disability-Final.pdf Washington Group on Disability Statistics. 2016. Report of Ability of Countries to Disaggregate SDG Indicators by Disability. Accessed on Nov. 16th 2020 at: https://www.washingtongroup- disability.com/fileadmin/uploads/wg/Documents/WG_Implementation_Document__10_- _SDG.pdf WHO. 2001. The International Classification of Functioning, Disability and Health. World Health Organization: Geneva. WHO -World Bank. 2011. World Report on Disability. World Health Organization: Geneva. World Bank. 2009. People with disabilities in India: from commitments to outcomes. Washington, D.C.: World Bank. 42 43 44 45 46 47 48 49 50 51 52 53 54 55