Abstract
This study focuses on the interaction between disability, chronic poverty and gender in rural Bangladesh, relying on analysis of the Chronic Poverty and Long Term Impact Study conducted between 1997 and 2010. A series of logistic regressions investigate the relationship between disabilities and chronic poverty among women with their employment, education, assistance and household coping strategies. The results indicate that primary schooling is lower among girls compared with boys in chronically poor households, with implications for the intergenerational transmission of poverty. Even where the probability of employment for chronically poor women with disabilities is positive, these women are potentially unlikely to be engaged in work that safeguards their rights or contributes to poverty escapes. Moreover, in the face of shocks, poverty becomes stickier, in the absence of effectively targeted safety nets coupled with adverse coping strategies that prolong poverty. The article concludes with a call for ensuring that intersectionality is more firmly embedded into existing social protection programmes.
Introduction
Disability may cause poverty, but poverty may also precipitate the onset of disability (Banks et al., 2017; Elwan, 1999; Mitra et al., 2013; United Nations [UN], 2019; World Health Organization [WHO], 2011; Yeo and Moore, 2003). Less understood are the nuances of this relationship, such as the relationship between disability and poverty dynamics (Groce et al., 2011), and the role of gender within this nexus. Chronically poor women with disabilities may face a triple disadvantage on account of their disability, gender and prolonged state of poverty. The horizontal inequalities experienced by persons with disabilities, including women in poverty among them, affect a range of outcomes, such as access to education and employment (Hulme and Shepherd, 2003). As a result, these groups are particularly at risk from negative shocks, which can tip households into poverty and reduce the likelihood of poverty escapes.
While Bangladesh is often touted as a success story on poverty reduction, pockets of poverty and vulnerability exist, which continue to be largely sidestepped by government assistance. For example, just under a third of extremely poor household heads with disabilities who live in coastal regions of the country receive government safety net benefits, and only half of these receive the disability benefit (Nokrek et al., 2013). This is despite growing awareness of disability at national levels. For example, the Ministry of Social Welfare, the Department of Social Services and the National Foundation for Development of the Disabled Persons all have programmes or policies targeted to persons with disabilities (Alam, 2009). The National Foundation for the Disabled in Development also allocates government grants to persons with disabilities and nongovernmental organizations (NGOs) working towards improving the situation of persons with disabilities. These are complemented by various stipends and social assistance programmes in place for persons with disabilities (World Bank, 2004).
The objective of this article is to understand the links between disability, gender and poverty in Bangladesh to work towards developing effective poverty reduction measures. It employs a rich, three-wave panel data set, the Chronic Poverty and Long Term Impact Study in Bangladesh, to examine poverty dynamics among chronically poor women with disabilities, and female heads of chronically poor households with persons with disabilities. Regression-based methods are used to compare labour and education outcomes, public transfers and primary coping strategies with negative shocks among these vulnerable populations to their respective male comparator groups.
The rest of the article is organized as follows: the second section begins by reviewing the literature, while the third section provides an overview of the data sets and the empirical model employed in the analysis. The fourth section proceeds to present descriptive statistics of poverty, gender and disability. The fifth section provides further insights into the triple disadvantage hypothesis by exploring chronic poverty, disability and gender in further detail through descriptive and regression-based analysis. The sixth section concludes with policy considerations going forward with an aim to dismantling the triple disadvantages so often endured by chronically poor women with disabilities and their households in rural Bangladesh.
The relationship between poverty, disability and gender
Various studies document the co-dependent relationship and bi-directional pathways between poverty and disability, describing it as mutually reinforcing (Banks et al., 2017; Elwan, 1999; Yeo and Moore, 2003), and also impacted by third factors such as violence or poor quality healthcare that may lead to poverty and disability simultaneously (Grech, 2016; Mitra, 2018). Disaggregating causal channels, Hoogeveen (2005) finds discrimination due to disability, in particular as it manifests in unequal access to food, healthcare, education and work, to affect poverty status. However, results are not always consistent across the literature. Mitra et al. (2013) find no significant difference in the rate of asset deprivation by disability status, in 11 out of 15 countries investigated. Trani and Loeb (2008) examine national household surveys in Afghanistan and Zambia to find that poverty per se, as measured by an asset index, is not statistically different between the group of individuals with disability and those without. In such cases, poverty may be the greater leveller. Their results do demonstrate, however, that people with disabilities suffer from lower access to healthcare, education and the labour market.
Disability may also impose a negative effect on a person’s marriage, education and employment, with women and girls in Bangladesh suffering more than men and boys (Hosain et al., 2002). While persons with disabilities in Bangladesh face vulnerability on multiple fronts, often despite the presence of government programmes that target them, the situation among certain categories of persons with disabilities is even worse. For example, Tareque et al. (2014) find that women and older people are more likely to report having disabilities compared with men and those who are younger. Moreover, Ali (2013) finds that women with disabilities in the country tend to experience severe social exclusion. Women with disabilities in other countries are also more likely to be living on their own, be divorced or experience sexual violence or social categorizations based on their disabilities (Dhungana, 2006; Meekosha, 2004). This issue has been investigated by Hasan et al. (2014), where they found that 84% of a sample of 226 women with disabilities in Bangladesh experienced some form of intimate partner violence at least once during their lives. Unfortunately, these findings are not unique to Bangladesh. On a global scale, disability is compounded by intersecting inequalities, stemming from individual characteristics such as age and gender, to exacerbate poverty outcomes.
Just as women may be victimized due to their disability, children are another especially vulnerable subgroup. Filmer (2008) assesses household surveys in 13 low- and middle-income countries to find that children with disabilities are less likely to start school and ultimately have a lower level of education that contributes to intergenerational poverty. They also face discrimination that can contribute to self-exclusion. During their formative years, children with disabilities encounter ‘biased attitudes, skewed representation, and low expectations’ (Froschl et al., 1999). Even where inclusive education programmes do exist, these may sometimes be weak or ineffective if not complemented with adequate teacher preparation services. However, this is not to say that there have been no positive outcomes from inclusion programmes. For example, children who were provided with hearing aids or wheelchairs in Bangladesh were found to be more likely to complete primary school than those without this assistance (Borg et al., 2012). Overall though, such policies remain limited in scope, and inclusion programmes are often more complex than provision of assistive devices alone.
While the literature on women and children with disabilities is growing, the triple-disadvantage framework has received relatively sparse consideration. The gap in the literature is mirrored in the policy scene, where programmes targeting chronically poor women and children with disabilities remain largely inadequate. The hypothesis in this article is that women with disabilities who live in chronic poverty are likely to endure experiences that are distinctly different from men with disabilities living in chronic poverty. Moreover, it is not just an addition of three factors (gender, disability and chronic poverty); rather, these factors compound and interact with each other in complex ways. The unique panel data set employed in this study with the aid of regression analyses allows for preliminary insights into this triple-disadvantage hypothesis in rural Bangladesh. More on the data set is presented in the next section.
Data set and methods
Data source
This article relies on a three-wave panel data set to inform our understanding of poverty dynamics and disability. The Chronic Poverty and Long Term Impact Study in Bangladesh is among a handful of mixed-methods studies examining poverty dynamics in developing and industrialized countries alike (Davis and Baulch, 2010). The aim of the Impact Study is to improve the understanding of the processes that shape chronic poverty and poverty dynamics in the country (Quisumbing and Baulch, 2010). The survey waves, which span 1994–2010, cover the individual, household and community level, and include information on household consumption expenditures and shocks, as well as education, employment and health of household members. The original round of survey data was conducted by the International Food Policy Research Institute (IFPRI), while subsequent follow-ups were undertaken by IFPRI in conjunction with the Chronic Poverty Research Centre, and Bangladesh’s Data Analysis and Technical Data Ltd. The districts included in the survey were chosen to represent the agro-ecological conditions of rural areas of the country. Thus, though the Impact Study was not nationally representative in a statistical sense, it does ‘broadly characterize the variability of livelihoods found in rural Bangladesh’ (Quisumbing, 2007).
The analysis in this article employs the Micronutrients Gender/ Agricultural Technology (MCG) and Food for Education/Cash for Education (FFE) rounds of this Impact Study, with disability and poverty analysis conducted primarily on the 4231 individuals and 1197 households for which data are available in 1997/2000, 2006 and 2010. In the study, while the last two survey waves were conducted over the same time period in both sites, the first wave of the FFE was conducted in 2000, while the first wave of the MCG was conducted in 1997. These are both combined to provide the first round poverty estimates for the panel analysis.
Model specification
The aim in this article is to explore the relationship between disability and chronic poverty at both the individual and household levels. The hypothesis is that the interactions between disability and poverty are strongly gendered. As such, the article investigates the extent to which disability, chronic poverty and gender may intersect in ways that might limit access to certain activities and opportunities. The analysis does not make claims about causal relationships, but rather is interested in the intersection of these three features and how they may be associated with specific outcomes in education, employment, assistance and coping strategies.
To investigate this relationship, logistic regressions are conducted. Such modelling allows an assessment of the relative strength of various determinants on our outcome variable of interest, as well as testing whether there are any interactions among the predictors. A logistic model is chosen as the outcomes defined are dichotomous; for example, a person is categorized as either a recipient of a social assistance transfer, or a non-recipient. In our regressions, outcomes are defined as binary variables covering a range of daily activities, social and financial assistance, and coping strategies. In the baseline equation
for
where
Poverty representing a household’s chronic poverty status derived from three-wave data;
Sex is a binary equal to 1 if the individual (or household head, respectively, in household-level equations) is female;
H is a vector of individual- and household-specific controls;
X is a vector of regional variables.
In addition to the baseline specification, we also specify a model with the same controls but using interaction terms between disability, gender and chronic poverty to explore our key hypotheses. The results of interaction terms are presented visually to investigate the overlap of disability, chronic poverty and gender, and marginal effects over modifiers are communicated in Appendix Table 7.
The concept of disability referred to in this article stems from its description in the Convention on the Rights of Persons with Disabilities: ‘Disability results from the interaction between persons with impairments and attitudinal and environmental barriers that hinders their full and effective participation in society on an equal basis with others’. In our article, the set of equations is firmly situated in the International Classification of Functioning, Disability and Health (ICF) model and capabilities approach, as it focuses on the activity limitations or ‘deprivations of functioning’ that individuals with impairments may face. Our identification of persons with disabilities is outlined in the ‘Key variables’ section.
Given the time dimension, random effects are employed, and the model is run using the poverty trajectory constructed from three-wave data, on the sample of respondents in waves 2 and 3. A reliance on the 2006 and 2010 survey years in the model stems from the fact that disability status is estimated in 2006. Including a survey year from a decade earlier might risk categorizing an individual as having a disability in an earlier round when in fact it may have been a more recent phenomenon. Beginning with the year in which disability is captured helps render our measure more precise. In our model, random effects are used instead of fixed effects, as we define disability as a time constant variable (discussed further below) of which we want to investigate causality. Moreover, a Hausman test for the suitability of random versus fixed effects for our household models lends support to our selection. Even so, both random and fixed effects approaches are based on a strict exogeneity assumption that may be questionable. As such, we also present a simple albeit inefficient alternative by using a pooled estimator to investigate whether results substantively differ from our random effects model. In all cases, we cluster standard errors for the individual, and report results in terms of probability metrics. Regression outputs are provided in Appendix 2.
Key variables
The analysis in this article relies on three key explanatory variables: the poverty trajectory of the household, the gender of the individual or household head and the disability status of the individual. To first construct the poverty measure we use, we compare household per capita expenditures of the sample with temporally and spatially adjusted lower and upper poverty lines derived from the Household Income and Expenditure Surveys by the Bangladesh Bureau of Statistics. While poverty is measured at the household level, in this article, we focus on poor individuals as derived from the household in which they live, and specifically individuals in households on different poverty trajectories. This focus on the individual allows comparability with disability status, which is similarly measured at the individual level. Summary statistics of the data set are available in Appendix 1.
To construct the disability variables, we rely on a disability module in the 2006 wave of the survey. In particular, a series of questions asks respondents whether they have physical impairments or experience activity limitations. Physical impairments include a missing or deformed limb, paralysis, use assistance such as glasses or a wheelchair, or experience difficulty hearing/ seeing/ speaking. Activity limitations cover whether an individual is unable to stand after sitting down, sweep the floor, walk 5 kilometres or carry 20 litres of water for 20 minutes. Questions on activity limitations are administered to individuals aged 7 years and above, while those on impairments cover individuals aged 4 years and above. Using these questions, we create an aggregated disability variable for the 2006 survey, which defines a persons with disabilities as one who has ‘generally poor’ or worse sensory capability, can only execute activities with a ‘lot of difficulty’, has a deformed or paralysed limb, or requires a wheelchair or other form of assistance excluding eyewear. This measure may overstate the prevalence of impairments, if the question relating to being able to walk for 5 kilometres or being able to carry 20 litres of water for 20 minutes may otherwise be too heavy for children. Employing only those responses at the upper end of the Likert-type scale of difficulty reduces this possibility. Even so, we also do not include the two responses around walking and carrying 20 litres in our assessment for children who are 14 years old or younger.
Our threshold is, from a conceptual standpoint, loosely derived from the WHO’s (2011) operational threshold of disability. The WHO’s threshold includes a series of scales wherein the ‘average of scores from respondents who reported extreme difficulties or total inability in any of the eight domains of functioning’ is calculated and those above the average were labelled as ‘disabled’ while those below were identified as ‘not disabled’. Our threshold is slightly higher, equivalent to 60%, or responses on the Likert-type scale of at least 3 out of 5. This modification is somewhat arbitrary, but rationally covers responses in the Impact Study’s disability module that may indicate a form of impairment or result in a functional limitation.
There are potential limitations to our approach. First, our categorization of disability is defined from 2006, because the disability module was unique to the 2006 wave. This represents a time where the Washington group question set was only just being developed. Based on the data period, persons with disabilities prior to or after 2006 who may not have had a disability in 2006 are not recognized as persons with disabilities in the other survey years, and vice versa. It could be that for several individuals in our ageing panel, disability onset occurred after the first survey wave. To address this, regressions are conducted on the subset of individuals and households in the balanced panel from the second and third survey waves. In this process, information about the persistence of poverty is retained based on data from all three survey years to examine longer-term chronic poverty across the three-wave survey period.
Second, our measure of disability itself is limited. For example, survey data limitations preclude us from identifying people with psycho-social disabilities and cognitive impairments, and children with developmental disabilities. In Bangladesh, research by Groce et al. (2014) reveals that cognitive and development disabilities are correlated with poor nutrition, which in turn compromises schooling outcomes. However, poor nutrition is not the only cause of these impairments and so other sources remain an omission. In addition, we do not disaggregate people with different types of impairments, such as physical or sensory, or distinguish functional difficulties with other impairments. This may lead to inconsistencies in classifying impairments and distinguishing this with the concept of disability. More generally, since the time of the data set, the ways of identifying and classifying people with different impairments, functional difficulties, or self-identifying as disabled have developed. Even so, our attempt is to provide a first step in exploring the relationships more generally at the aggregated level, where sample sizes of the interactions with gender and poverty dynamics enable more robust empirical analysis.
Finally, the thresholds for disability and poverty affect the magnitude and statistical significance of the relationship. In our article, we find a high share of persons with disabilities likely because the Impact Study constitutes an ageing and rural population among whom the presence of disabilities is generally higher.
Descriptive statistics: poverty, gender and disability
The analytical focus of this section is on the interactions of disability with poverty and well-being trajectories, as it is difficult to know from the data whether poverty causes disability or vice versa. In our data set, around 17% of individuals in the 2006 survey are extremely or moderately poor. Following our operational threshold, around 18% of individuals in our sample are identified to have a disability. This is slightly higher than the global figure of 15% of the world population living with some form of disability (WHO, 2011), though the latter is also beginning to be adapted downwards if considering the subset of individuals who have a lot of difficulty with a core basic activity, rather than some difficulty for core activities (Mont, 2019). In the case of Bangladesh, we may find a slightly higher figure on account of an ageing and rural population included in the panel data set.
Within the subset of individuals living in poverty, half are chronically poor, defined in this study as individuals who come from households that are either extremely or moderately poor for all survey rounds. Overall, we observe a high rate of poverty reduction in the data set relative to national statistics. This variation is likely on account of the purpose behind the creation of the Impact Study; the two data sets employed in our working paper, the MCG and FFE samples, sought to assess development interventions in the form of the introduction of agricultural technologies, and cash transfers to poor households sending their children to school. As such, it is understandable that such interventions may have reaped positive outcomes conducive to poverty reduction in the districts they covered.
We also examine poverty trajectories for individuals present in the 2006 and 2010 rounds alone (Figure 1), to start with the period where disability is measured. From these waves, we observe that persons with disabilities as well as persons without disabilities are improving their well-being at the same rate within individuals beginning the period in extreme poverty and those beginning the period not in poverty. Moreover, among individuals living in extreme poverty in 2006, just under half remain in extreme poverty 4 years later, indicative of a large number of individuals who escape poverty. Together, three-wave and trend data are suggestive of the heightened vulnerability faced by persons with disabilities that may either tip them into poverty or prevent their poverty escape. 1 We must be careful, however, not to associate this trend with causality, especially given that we do not know if these individuals had a disability in the 1990s or in 2010. Rather, there are also many individuals in our data set who were never poor between the first and last survey waves.
Disability-disaggregated poverty transitions by type of poverty.
Individuals beginning the period in moderate poverty, however, display a different trajectory. Over three-fifths of individuals without a disability who began in moderate poverty end up becoming non-poor by 2010. However, the situation is not so sanguine among persons with disabilities within the moderate poor. Here, a 18 percentage point difference emerges between the moderate to extreme poverty descenders over the two survey rounds, when disaggregated by disability status. In other words, there is considerable downwards mobility among the moderate poor in 2006, and especially so among persons with disabilities within this group. For the moderate poor in 2006, their escape out of poverty displays an even larger difference, with one in two persons with disabilities in moderate poverty in 2006 escaping poverty altogether by 2010, compared with two in three among persons without disabilities. The middle two bars of Figure 1 thus indicate that individuals in households relatively close to the poverty line, the moderate poor, experience high volatility in their state of poverty. The majority of these individuals are unlikely to remain moderately poor, but instead transition out or descend into steeper poverty, with the second scenario more likely among persons with disabilities.
We next investigate the presence of disability among women and chronically poor households, relying on the three-wave categorization of poverty dynamics. As noted above, around 18% of individuals are identified to have a disability, with women comprising the majority of the sample (Figure 2). Specifically, 26% of women have a disability, compared with 18% of men. Moreover, among the population of chronically poor individuals, 22% are persons with disabilities. This compares with 18% of persons with disabilities among individuals who are not chronically poor.
Disability status according to key characteristics.
We finally disaggregate results by gender to see whether the triple disadvantage hypothesis is supported. Restricting the analysis to the sub-sample of individuals who are either women, have a disability, or are chronically poor over the two survey years, we find that on average around 8% of this specific subsample are chronically poor women, while one in five are women with disabilities (Figure 3). More specifically, if we begin by focusing instead on the group of persons with disabilities who are also chronically poor in 2006, we find that 63% are women. This highlights a strong gender dimension among the persons with disabilities, lending impetus to testing our triple disadvantage hypothesis.
Interactions between poverty dynamics, disability and gender.
This section introduced poverty trajectories and some gender- and disability-disaggregated descriptive statistics derived from the data set. A key conclusion from this analysis is that persons with disabilities have varied poverty trajectories, but that within these trajectories lies a strong gendered dimension. To better understand the effect of this dynamic on outcomes of daily living, we next present results of our empirical analysis.
Results and discussion
This section presents results of the regressions and descriptive analysis of the triple disadvantage hypothesis explored in this study. The results are focused on employment outcomes, school enrolment, receipt of assistance and coping strategies to shocks.
Employment outcomes
The results indicate that disability is associated with a reduced probability of employment (Table 3, column 1). When focusing on the subsample of persons with disabilities (Table 3, columns 2 and 3), women have a lower probability of employment compared with men. However, when focusing on the interactions (Figure 4), we observe that women with disabilities have a higher probability of employment if living in chronic poverty compared with women with disabilities who are not chronically poor, though at slightly higher p-values (p = 0.138). Among men with disabilities, however, the opposite trend is observed where chronic poverty is associated with a lower probability of employment at significant levels. These gendered effects result in a reduced gender differential in employment as observed by the converging probabilities between men and women in Figure 4. However, this must be interpreted with caution given the need to work and precarity of employment among chronically poor women with disabilities.
Employment outcomes (subsample of individuals with disabilities), predicted probabilities.
Employment in the context of this study and chronically poor individuals equates with economic activity in a broad sense, and does not necessarily refer to formal labour market inclusion or salaried work. Instead, most chronically poor individuals in our rural data set engage in agriculture, while some also partake in non-farm activities. Of women with disabilities, there is a large difference in rates of non-farm work between those in chronic poverty and others (Figure 5), in favour of individuals in households that are not chronically poor, though these differences are only significant at higher than conventional p values (p = 0.1438). In contrast, among men with disabilities, rates are largely comparable for men in chronically poor households compared with other households. The concern with this difference is not on account of the type of activity, but rather its returns. Indeed, in the data set in 2010, the monthly income from the primary activity of individuals working in agriculture (2306 taka) was less than half of those working in non-farm activities (4940 taka). Even among chronically poor individuals, income from agriculture (1853 taka) was much lower than for non-farm work (2972 taka). Finally, when comparing women to men with disabilities, these differences are even more severe, with income from women’s primary activities at a mere 1561 taka per month, compared with 4021 taka on average for men in 2010.
Employment type by gender and poverty among persons with disabilities, pooled data.
Our regression results indicating higher probability of vulnerable women’s employment could be a result of programmes in the country which target poor women and lead to their self-employment. For example, a third of the beneficiaries in the Employment Generation Programme for the Poorest (EGPP) are women. This quota increases the likelihood that female heads engage in employment, on community sub-projects, that may allow their households to cope with vulnerability (Kumar, 2013). To this end, it could be such opportunities for self-employment relative to wage and other forms of employment, as well as gendered norms behind these, that is driving our result. Given that some studies find that employment may be an important means to escape chronic poverty, economic activity among chronically poor women with disabilities could potentially act as a pathway out of poverty (Diwakar and Shepherd, 2021).
However, the positive association with employment among chronically poor women must also be viewed with caution. If women are disproportionately concentrated in poor quality jobs, this does not bode well for promoting the agency required to escape poverty (World Bank, 2014). Indeed, many individuals partake in informal work such as rickshaw/van pulling or other small self-employment labour, where ‘poor quality of that work, including its low-pay, dangerous conditions and insecurity or irregularity means that this work frequently maintains people in poverty rather than enabling poverty escapes’ (Bulla et al., 2013). It could be that chronically poor women have few options but to seek employment even if on very poor terms, or to engage in livelihoods in ways that are unable to effectively contribute to pathways out of poverty. In instances where work is of poor quality and high insecurity or vulnerability in terms of pay, benefits, and opportunities, employment may sometimes act as a perpetrator of rather than pathway for escaping poverty (CPAN, 2013; Wood, 2003). As such, we must be careful not to assume that the status of employment alone is a certain pathway out of poverty; rather, poor women especially are likely to be subsumed in wage traps if their labour income is close to subsistence (Bulla et al., 2013).
Education outcomes
Employment is generally more effective if workers have first acquired, through education, basic skills to engage in gainful employment. In rural Bangladesh, in the last survey wave, education levels remained low among people in chronic poverty. Adults aged 18 years and older living in chronically poor households had on average 1.8 years of school, compared with 4 years of adults in other households. There was also an intergenerational persistence in these trends, with just 68% of children aged 6–18 years in chronically poor households enrolled, compared with 79% in other households in our data set. Over this same age group, girls were on average more likely to enrol compared with boys (83% of girls, compared with 73% of boys enrolled) in the latest survey year.
Regression analysis corroborates these descriptive findings, indicating that children in chronically poor households have a lower probability of enrolment, compared with children in households that are not chronically poor (Table 4, column 1). Girls have a higher probability of enrolment compared with boys, likely on account of the wider age range in the analysis. This might reflect boys dropping out of school for employment, particularly into and beyond secondary school levels. For girls and boys with disabilities, the experience of chronic poverty is associated with a lower probability of enrolment, though only statistically significant for girls (Figure 6, left; Table 7.2B). Overlapping disadvantages moreover contribute to a gender penalty in education for children in households with disabilities. Specifically, girls in chronically poor households with persons with disabilities have a lower probability of enrolment in school, compared with girls in other households with persons with disabilities (Figure 6, right). Moreover, this decline is enough to eliminate any significant gender differential in favour of girl’s education in households with persons with disabilities that are not in chronic poverty. This might imply a trade-off in the poorest households, where caring responsibilities in resource-constrained environments may fall on girls in ways that lead to withdrawal from school. The lower probability of education attendance among these girls stifles the girl effect and could contribute to a vicious cycle that perpetuates the intergenerational transmission of poverty.
School enrolment (subsample of individuals with disabilities = left, households with persons with disabilities = right), predicted probabilities.
The analysis on enrolment controls for stipends, which is one way of promoting school access. In Bangladesh, the value of these stipends allocated per person with disability increases per level of education (ILO, n.d.). In our data set, slightly more children with disabilities receive education stipends (45%) relative to children without disabilities (38%) in the latest two survey years (Figure 7), though differences were not statistically significant. There was a statistical difference though in receipt of stipend by gender, where 48% of girls and 32% of boys received these stipends. Similarly, there was a slight difference in receipt of stipend for children living in chronically poor households (46% compared with 37% otherwise), though at slightly higher levels of significance (p = 0.1294). Econometric results indicate the relationship between receipt of stipends and enrolment is positive. However, removing the stipend variably only minimally alters the relationship between gender and enrolment, suggesting that the presence of stipends on its own is inadequate in driving education access for girls facing overlapping disadvantages.
Stipend coverage by disability, gender and poverty, 2006–2010.
While education has the potential to act as a ‘portable asset’ that contributes to building resilience and sustaining poverty escapes (Bird et al., 2010), children in chronically poor households are less likely to be in school (Ssewanyana and Kasirye, 2012). As such, they are less likely to develop the technical or life skills necessary to succeed in formal employment, or to build the networks through which employment opportunities may arise (Hossain et al., 2012). The presence of education stipends alone is moreover insufficient in prompting education uptake among children facing compounded disadvantages. Indeed, these stipends form just one type of public transfer available to Bangladeshis. The next section explores public transfers at large for female heads of chronically poor households with persons with disabilities.
Coping with prolonged conditions of poverty
A focus on female heads of chronically poor households with persons with disabilities as a unit of analysis is important, because it is often the case that persons with disabilities may be carried along by the fortunes of their households, or that households are affected by the presence of persons with disabilities. To this end, targeting social transfers not just to persons with disabilities but also to female heads of their households is likely to have a substantial effect on well-being trajectories among poor population and persons with disabilities alike. Exploring the triple disadvantages descriptively, around 65% of chronically poor women lived in households that received a transfer over the last two survey years, compared with 56% of chronically poor men, with the difference statistically significant. Rates of persons with disabilities living in households receiving transfers, however, were relatively comparable to people without disabilities (43–45%), suggesting that this targeting could be improved.
Regression results, however, promisingly indicate that chronically poor households have a higher probability of receiving assistance compared with households not in chronic poverty (Table 5, column 1). This relationship also holds when focused on the subsample of households with persons with disabilities (Table 5, columns 2 and 3). Even so, when focusing on interactions, the difference by gender in the receipt of transfers among chronically poor households with people with disabilities is not statistically significant (Table 7.4A). This result may reflect wider country level targeting over a similar period. Even though there are various social safety net programmes by Bangladesh’s Ministry of Social Welfare that attempt to ensure gender parity, the number of female recipients remains lower relative to male (Ministry of Finance, 2015). These results point to an area where social protection programmes could be further improved. Mitra et al. (2015) argue that female-headed households should be included in the legislated social beneficiary groups in Vietnam, given their precarious welfare situation following health shocks. Our study disaggregates this one unit further in Bangladesh to advocate that policies need to better target female heads of chronically poor households with disabilities, a group who is often among the poorest of the poor (Figure 8).
Receipt of assistance, predicted probabilities.
The presence of social policies can influence the impact that disability has on well-being. For example, regular social transfers can help insure the chronically poor by smoothing consumption trends. In some cases, assistance promote poverty escapes if they enable people to take business risks that result in positive economic outcomes enough to propel households across the poverty line. In addition, in the face of a household shock or increased vulnerability due to the onset of disability, assistance can help provide economic resilience and so prevent further impoverishment. Regular assistance over time could also lead to a host of human capital development outcomes, through improved nutrition, health and education (Shepherd et al., 2011). This pathway would potentially be even more beneficial for poor persons with disabilities, and other members of households with disabilities, on account of their state of high vulnerability stemming from both poverty and activity limitations, as well as restrictions on participation.
Regular social protection transfers and schemes are useful as a type of insurance for the chronically poor. However, equally important are transfers in times of distress. Health shocks, such as the death of a member or illness or injury of a household member were the most common shocks among households in the latest survey year (Figure 9). Environment or agriculture-related shocks, followed by family-related shocks such as divorce or abandonment or marriage-related expenses were the second and third most common types of shock in the latest survey year, respectively. The prevalence of shocks was generally comparable between households with and without persons with disabilities. Two exceptions were for agricultural shocks and family shocks. Of households experiencing shocks in the latest survey year, 28% of households with persons with disabilities recorded agricultural shocks, compared with 37% of other households. Family-related shocks were in contrast more common for households with persons with disabilities (23%) compared with other households (13%), though in both instances differences are not statistically significant. These two shock types were also more common among female heads of households. In contrast, by poverty status, health shocks and family shocks were more commonly reported by chronically poor households compared with other households.
Types of primary negative shocks, 2010.
In the absence of safety nets in times of crises, households might be propelled into poverty or worsen their state of well-being. A negative shock in female headed, chronically poor households with persons with disabilities may have a multiplier effect, rendering households especially vulnerable in the aftermath. This is because disability may itself be perceived as a factor that reduces income and well-being or entails increased expenses (e.g. care costs, medical costs, special diet, special transport), and that accordingly aggravates poverty dynamics or even offsets the gains accrued through positive assets. In an effort to cope with a potential multiplier effect stemming from a negative shock, households may adopt a variety of coping strategies (Figure 10). The main differences in coping strategies to negative shocks depending on whether the household had a person with disability was found in asset sales and consumption-related coping strategies. Households with persons with disabilities were more likely to reduce consumption of food, or engage in a distress sale of productive assets, though only the difference on assets is statistically significant.
Primary coping strategy adopted to household shocks, 2006 and 2010 pooled.
Regression results reflect these differences, indicating that when faced with a shock, chronically poor households are less likely to respond by selling an asset, relative to the reference group of other households not in chronic poverty (Table 6, column 1). 2 Regarding the presence of overlapping disadvantages, both men- and women-headed chronically poor households have a higher probability of asset sales in households with persons with disabilities, compared with households without persons with disabilities (Figure 11). Among women-headed households, moreover, even those not in chronic poverty have a higher probability of engaging in a distress sale of assets when living in households with persons with disabilities. That our poverty variables are measured at the household level, but disability at the individual level, lends credence to our earlier hypothesis that persons with disabilities are likely to be carried along by households. There may be things happening in the household which have an effect on the person with disability, and vice versa. Equally, as we see here, the presence of a person with a disability may differentially affect coping strategies depending on the gender of the household head. Relatedly, caring roles including looking after disabled family members is often undertaken by women and so might affect what other coping strategies are possible for them, regardless of poverty status.
Coping strategies for households recording any shock, predicted probabilities.
More generally, distress sales of assets are especially harmful as they can initiate or hasten a household’s decline into extreme poverty, or even push non-poor households under the poverty line as observed in rural Bangladesh (Davis, 2011). Effective policy interventions that seek to assist the chronically poor should take into account not just the prevalence or severity of shocks, but also the composition of households susceptible to such shocks and attitudes towards disability more generally.
Conclusion and looking forward
This article focuses on the interaction between disability, poverty and gender, and how these influence employment, education and assistance at both the individual and household levels in rural Bangladesh. As noted earlier, it relies on a classification of disability based on a data set from the early 2000s, though the Washington group questions and associated identification methodologies have since been developed. Future research to understand the relationships using more disaggregated measures of disability is needed to further nuance this nexus. Relatedly, a merely cumulative approach of the gender, poverty and disability nexus is necessary but not sufficient. As the data show, these combinations are not straightforward given the complexities of this relationship. Finally, it should be highlighted that the results of this article is specific to Bangladesh, and so generalizations beyond its context and to other settings with varied national policies should be viewed with caution.
In the context of rural Bangladesh, across work and education, financial assistance and coping strategies, our results indicate that there exist a range of outcomes when both disaggregating disability by gender and poverty status, and interacting the three together. These results substantiate the presence of triple disadvantages in rural Bangladesh. Even where the probability of a daily activity for chronically poor women with disabilities is positive, such as with employment, our results and the wider literature suggests that these women are unlikely to be engaged in employment that safeguards their rights or contributes to poverty escapes. The article further demonstrates that primary schooling is reduced among girls in chronically poor households relative to the reference group, within households with persons with disabilities. The negative education outcomes for this subset may perpetuate the intergenerational transmission of poverty. Moreover, in the face of shocks, poverty can be prolonged, particularly in the absence of effective safety nets. Specifically, the marginal effect of gender on the receipt of public transfer among chronically poor households with persons with disabilities is positive but not statistically significant in regression results, which suggests that there could be better targeting. The precarity of the relationship emerges not in the least because in the event of a shock, these households are also very likely to engage in a distress sale of assets and worsen their condition of chronic poverty.
The results of this analysis provide evidence for the presence of triple disadvantages facing chronically poor women with disabilities in rural Bangladesh. Disadvantages in education access and vulnerable employment exist despite a range of policies and programmes in place to cater to specific vulnerabilities and risk factors in the country. Existing programmes like the EGPP in Bangladesh need to be complemented with policies and legislation geared towards improving the conditions and status of poor women in informal employment. Inclusive education should also be on the agenda of policy-makers, through developing proper infrastructure, ensuring quality of teaching, and incentivizing families to send girls with disabilities in chronically poor households to school. Education stipends can also help, but need to specifically target girls with disabilities in chronically poor households, and be coupled with quality inclusive education programmes if they are to translate into positive education outcomes and agency, contribute to gainful employment and reap lasting outcomes conducive to sustained poverty escapes.
Other forms of social protection are also needed at the intersections. Examining the policies in the country that frame public transfers for persons with disabilities, we see as early as 1995 the adoption of a National Policy on Disability in Bangladesh, which outlines ‘prevention, identification, education, rehabilitation, research and management’ of national programmes for the persons with disabilities. A 5-year National Action Plan on Disability was launched in 2006 to better align with the stated goals of the policy (Sultana, 2010). The same year saw the introduction of the Allowance for Financially Insolvent Persons with Disabilities, a cash transfer programme prioritized to the ‘elderly, homeless, women with multiple disabilities and poor intellectually impaired children in economically disadvantaged areas’ of Bangladesh (ILO, n.d.). This programme thus acknowledges key disadvantaged groups. What is now needed is for a convergence of policies and programmes to ease the triple disadvantage among the vulnerable populations, including those which consider household-level demographics. While regression analysis is one way in which to contribute to our understanding of these realities, triangulation through mixed methods, in particular using life histories, can provide a more detailed and rich understanding (see, for example, Shepherd, 2016).
Rural Bangladesh already has social transfers in place and shock-responsive social assistance, many of which do attempt to target vulnerable populations. Other low- and middle-income contexts also have varying types and coverage of social protection. What is needed in these contexts is for a linking of the vulnerable populations addressed in country and area-specific programmes, or a weaving together of the myriad of social protection schemes already in place to this end. Specifically, protection schemes need to acknowledge the compounded disadvantages faced by chronically poor women with disabilities as well as female heads of chronically poor households with persons with disabilities. Moreover, beyond cash transfers, other forms of social provision and support need to be strengthened to provide poor women with disabilities with the financial capacity to engage in gainful employment, send their children to school, and so contribute to sustained poverty escapes. Only by ensuring that social protection schemes better target poor women with disabilities as well as female heads of chronically poor households with persons with disabilities can policy-makers hope to effectively dismantle the triple disadvantages and reap lasting benefits in rural Bangladesh and beyond.
Footnotes
Appendix 1
Appendix 2
Acknowledgements
This article has benefitted from invaluable discussions with and comments from Andrew Shepherd (ODI), Sophie Mitra (Fordham University), Daniel Mont (UCL), Jean-Francois Trani (Washington University in St. Louis), Alexandre Cote (IDA), Sylvie Cordier (IDA), Mosharraf Hossain (ADD), Peter Davis (University of Bath) and Binayak Sen (BIDS). The author is also grateful for feedback from two anonymous peer reviewers. Responsibility for the content rests entirely with the writer.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship and/or publication of this article: This article draws on research funded by the International Disability Alliance. The views expressed in this publication do not necessarily reflect the views of funding bodies. It was led by the Chronic Poverty Advisory Network, ODI.
