Qualitative exploration of factors affecting progress in antipoverty interventions: Experiences from a poverty-reduction program in Bangladesh

Abstract Understanding and addressing the factors that affect progress in antipoverty interventions is central to achieving the Sustainable Development Goals. In Bangladesh this topic has been largely explored through quantitative approaches, and we believe in-depth qualitative analyses of household dynamics in the context of antipoverty interventions is lacking. This article addresses this lacuna. Based on 49 focus group discussions and 15 case studies, we analyse livelihood dynamics of beneficiary households within a national extreme poverty alleviation program. We identify five determining factors to the effectiveness of antipoverty interventions: 1) health shocks, natural hazards, and vulnerabilities; 2) household demography; 3) inappropriate IGA planning, implementation, and monitoring; 4) dependence/inaction; and 5) political and social instability. We argue that livelihood-based antipoverty initiatives often fail because they do not address these five factors, and call for a comprehensive approach that prioritises them in program design.

Abstract: Understanding and addressing the factors that affect progress in antipoverty interventions is central to achieving the Sustainable Development Goals. In Bangladesh this topic has been largely explored through quantitative approaches, and we believe in-depth qualitative analyses of household dynamics in the context of antipoverty interventions is lacking. This article addresses this lacuna. Based on 49 focus group discussions and 15 case studies, we analyse livelihood dynamics of beneficiary households within a national extreme poverty alleviation program. We identify five determining factors to the effectiveness of antipoverty interventions: 1) health shocks, natural hazards, and vulnerabilities; 2) household demography; 3) inappropriate IGA planning, implementation, and monitoring; 4) dependence/inaction; and 5) political and social instability. We argue that livelihood-based Ashraful Kabir ABOUT THE AUTHOR Ashraful Kabir We, the research team, worked with Economic Empowerment of the Poorest/ Stimulating Household Improvements Resulting in Economic Empowerment (EEP/Shiree) program in Bangladesh-one of the largest poverty reduction interventions in Bangladesh. Our research focuses on the exploration of factors that affects the progress of poverty reduction interventions and lead to diverse outcomes among extremely poor households. The research work in this paper addresses whether and how a number of factors determine anti-poverty interventions targeted to extremely poor households and results in differential outcomes which contribute to the information gap in the poverty discourse. Specifically, we are interested in exploring household dynamics and resilience to poverty among low and middle-income households.

PUBLIC INTEREST STATEMENT
In many cases, the project beneficiaries experience significantly different outcomes despite having equitable supports (i.e. incomegenerating assets, and routine follow-up) in Bangladesh. While most households gain the expected outcomes, some fall far behind their counterparts. These individuals or households' condition often fluctuates within the poverty cycle and sometimes falls even deeper into poverty. Understanding and addressing the factors that affect progress in antipoverty interventions is central to achieving sustainable development. This article addresses factors to the effectiveness of antipoverty interventions. Findings indicate that five determining factors affects the progress of poverty reduction interventions and lead to diverse outcomes among the project beneficiaries: 1) health shocks, natural hazards, and vulnerabilities; 2) household demography; 3) inappropriate IGA planning, implementation, and monitoring; 4) dependence/inaction; and 5) political and social instability. We argue that livelihood-based antipoverty initiatives often fail because they do not address these five factors, and call for a comprehensive approach that prioritises them in program design. antipoverty initiatives often fail because they do not address these five factors, and call for a comprehensive approach that prioritises them in program design.

Introduction
In recent decades, Bangladesh has made remarkable progress in fighting poverty. In 1974, immediately after Bangladesh's independence, US secretary of state Henry Kissinger's aid reportedly referred to the country as a "basket case" (Chowdhury, 2018). Despite the authenticity of that statement, and was widely perceived as a nation characterised by poverty, starvation, and poor social and human capital. After 45 years, the country's economic and social development indicators present a somewhat surprising picture. The poverty rate has significantly declined, and the country has maintained a GDP growth rate averaging 6.2 percentage points over the past decade (Bangladesh Planning Commission, 2015). Moreover, the rates for both poverty and extreme poverty have markedly declined during that time. For example, in 1973-1974 the prevalence of poor people was 71.3% in rural areas and 63.2% among the urban population (Halder & Mosley, 2004). In 1995-1996, the prevalence of poor was reduced to 53% nationally and 57% among the rural population; the corresponding percentages of extreme poor were 36% and 40%, respectively, for the same period (The World Bank, 1998). According to the latest statistics released by the planning commission of Bangladesh in 2015 the prevalence of extreme poor and poor are 12.9% and 24.6%, respectively (Bangladesh Planning CommissionGovernment of the People's Republic of Bangladesh 2015).
These achievements can be attributed to wide-ranging concerted efforts between the Government of Bangladesh (GoB), Non-Governmental Organisations (NGOs), and development agencies. Among those actors, the role of NGOs in designing, implementing, coordinating, and monitoring antipoverty programs is considered vital in supporting government initiatives (Banks & Hulme, 2012;Rahman, 2006;Sajjad, 2004). Over recent years, NGOs have been playing an increasingly prominent role in fighting for poverty through participatory planning and program support. Their role is largely viewed as a structural mechanism for system delivery in poverty-reduction initiatives-namely, input supply, process monitoring, and outcome measurements.
Despite marked advances in poverty-alleviation drives in various NGO interventions-with the support of the government and core development partners-many projects remain beyond performance assessment. It is also evident that even when they receiving equitable incomegenerating assets and routine follow-up support, project beneficiaries experience significantly different outcomes. While most households gain the expected outcomes, some fall far behind their counterparts. These individuals or households' condition often fluctuates within the poverty cycle and sometimes falls even deeper into poverty. The question of why some succeed while others do not, despite receiving similar support, remains unanswered. Understanding these experiences is central to effective poverty alleviation. Much of the existing literature on poverty alleviation from Bangladesh has investigated these aspects primarily through the application of quantitative approaches (e.g. many of the chronic poverty working papers relating to Bangladesh) (Begum & Sen, 2004;Kabeer, 2004;Matin, 2002;Quisumbing, 2007;Sen, 2003) and/ or as a supplementation to quantitative methods (Davis, 2006). Our intention here is then to counterbalance such approaches through a primarily qualitative approach to investigate the different outcomes observed within poverty-reduction interventions among extremely poor households to determine factors of program failure. This study helps fill an information gap in development discourse and poverty studies and contributes toward better programming for povertyreduction initiatives.

Program description
This article is based on research from the 'EEP/SHIREE" programme (Economic Empowerment of the Poorest/Stimulating Household Improvements Resulting in Economic Empowerment) which operated in Bangladesh between 2008 and 2016. EEP/SHIREE was designed as a partnership between the GoB and UK Department for International Development (DFID) (later also receiving funding from the Swiss Agency for Development and Cooperation (SDC)) to support the GoB efforts to reduce extreme poverty in line with the targets of Millennium Development Goal 1. The primary objective of the project was to enable one million poor people to lift themselves out of extreme poverty and achieve sustainable livelihoods. The project included a number of interventions focusing on livelihood improvement, natural disaster and climate vulnerabilities, economic shocks, social exclusion, and health and nutrition, targeted towards the extreme poor. It has then supported those who lack the benefits afforded by the country's economic growth and other social protection mechanisms. To achieve these objectives, EEP/SHIREE collaborated with national and international NGOs covering all of Bangladesh. Partner NGOs (implementing bodies) selected potential beneficiaries (households) based on the selection criteria, delivered resources, trained beneficiaries, monitored activities, and reported outcomes according to a project log frame. These partner NGOs received two separate categories of funds: "scale funds" and "innovation funds". The NGOs with the capacity to facilitate large-scale interventions using tested methods received scale funds, while those that tested methods for improving livelihoods and were ready to scale up their interventions received innovation funds.

Research design
This study deployed an exploratory qualitative research design to develop better understanding beneficiary households experiences of the interventions. Our analysis was guided by a descriptive statistical analysis generated by the program which identified key variables affecting program outcome for beneficiaries (Mascie-Taylor et al., 2016;Mascie-Taylor & Goto, 2015). Our focus is to interrogate these qualitatively, seeking the participants' experiences, opinions and perspectives, attitudes and conceptions, beliefs and perceptions, regarding the factors affecting their capacity to improve their condition within antipoverty interventions. Considering the research objectives we adopted this qualitative approach to uncover these factors and its contextual meaning which are perceived to be significantly important and can support effective interventions (Hammarberg, Kirkman, & De Lacey, 2016).

Study period and setting
This study was conducted between April and August 2016 as part of a "supplementary qualitative field survey for the final research report on lessons for the economic empowerment of the poorest (EEP/SHIREE)". We collected data from 13 NGOs across the country. Of those, six were "scale fund" and seven were "innovative fund" NGOs. The interventions were implemented in five geographical regions: the northwest (especially affected by seasonal hunger), the southern coastal area (most vulnerable to severe climatic hazards), the Chittagong Hill Tracts, the "haor" (large bowl-shaped wetland) regions, and the Dhaka slums. To improve sustainable livelihoods, each NGO designed interventions, applying their respective approaches in consideration of contextual aspects. Most NGOs adopted strategies for forming community-based organisations (CBOs) based on agreedupon written principles. CBOs would generally maintain an executive committee comprising a president, cashier, and general member who would gather at least once a month to organise and facilitate group activities and share views among the members. The beneficiaries maintained a bank account and passbook for savings and loans. Various supplies and supports were provided to each beneficiary according their skills, enthusiasm, local demand, and contextual aspects. The supports and supplies included farm and off-farm assets, such as fisheries, livestock, vegetable cultivation, fishing nets, and grocery shops. NGO field staffs helped identify potential incomegenerating activities (IGAs) that included livestock, small businesses, vegetable cultivation, fishing businesses, shrimp cultivation, and so on. Most beneficiaries maintained regular group meetings on a voluntary basis where they exchanged information regarding their actions, challenges, opportunities, and other emerging issues.

Study population and sample
We conducted 49 focus group discussions (FGDs) and 15 case studies with household members who participated in the EEP/SHIREE interventions, frontline program staffs, and community people ( Table 1). The participants were included in the interviews applying inclusion criteria-aged 18 and above and volunteered to participate. Participants were purposively recruited. In this process, we included individuals who showed a proactive interest to share their experiences, opinions, ideas in the discussion sessions. By applying the principle of data saturation-a point where no new information and/or theme and/or dimension are emerged (Guest, Bunce, & Johnson, 2006). We included 6-8 participants in each FGD which is ideally taken (Krueger Richard A and Casey Mary Anne 2000). We included both male and female participants in same the FGD in order to enhance a dynamic and interactive discussion where participant from different position talked with each other by clarifying, explaining, agreeing, disagreeing, querying, and elaborating with the topics. However, the moderator maintained a good balance of controlling and motivating all participants and thus avoided gender dominance during the discussion. To increase variability, we conducted FGDs from both close locations (close to markets/small townships) and remote locations (distance from markets/small townships).

Data collection
We purposefully selected the study participants to achieve the study objectives. Conversations were conducted in Bangla-the native language of both the interviewers and the participants. We recorded the interviews in audio format. A group of anthropologists and public health researchers moderated the FGDs while an assistant researcher took notes. We used a semi-structured questionnaire that was piloted in other settings (outside the study sites) beforehand. Before starting data collection activities, the research team took time to establish a good rapport with the participants and other community members, and described the purpose of the study. Each FGD took an average of 90-120 minutes. Follow-up visits were made in some cases to obtain missing information and further investigate certain issues.

Data analysis
We used a thematic analysis approach, which is frequently used in qualitative research (Braun & Clarke, 2006). Immediately after interviews, we translated the recordings into English. We manually analyzed data (and did not use any specialist software). Initially, we generated some "codes"meaningful and significant information or ideal dimensions focusing on the research objective. Afterward, we looked for a cluster focusing on the nature of the codes. Once a cluster was formed, we looked for a theme or concept consisting of a few clusters. A group of researchers independently coded the text to increase data validity. To further increase validity, we performed a triangulation of data collection techniques-a methodological mixing of FGDs and case studies (interviews).

Ethical considerations
This study was approved by the ethical review committee of EEP/SHIREE. The committee reviewed issues involving the human subjects and approved the study. We sought to obtain signatures or fingerprints on a paper consent form. Written consent was obtained, and we documented it using audio recording. Before obtaining consent, we explained the study objective, the harms and benefits, the study's importance, confidentiality, and the right to withdraw from the conversation at any stage. We also provided a telephone number for further inquiries. We gathered personal and medical information, including name, age, sex, healthcare costs, income and diseases. We used participant identification (ID) throughout the data analysis, which was removed before reporting the findings.

Results
In this section, we present the socio-demographic characteristics of the study population (Table 2). In the later part, we present the results.

FGD criteria
Case studies: selected in-depth interviews (n)

Sites
Closed location (n)

SSF (n)
Save the Children As shown in Table 2, the mean age of FGD participants was 36.25 ± 7 years. A total of 350 beneficiaries participated in the FGD (56% males and 44% females). The average monthly household income was BDT 5900 ± 750 (considering US$1 = BDT78). More than half (56%) of the participants received first to fifth grade schooling, while 18% of them received no formal education. The majority of our participants were Muslims (66%) followed by Hindu (20%) and Buddhist (14%). Nearly half (44%) of the participants were from nuclear families; while 16% of them were living in extended families. The highest number (32%) of participants was provided small-scale business support such as cloth, shrimp, fishing or furniture. Over one-fifth of the participant (22%) received livestock support including ducks, chickens, cows, and goats. The remaining participants received support such as handicraft (12%), rickshaw/van (12%), vegetable/farming (10%), and grocery shop (6%).
Based on the data, five key themes emerged as determining factors for program outcomes: 1) health shocks, natural hazards, and vulnerability; 2) household demography; 3) inappropriate IGA planning, implementation, and monitoring; 4) a culture of dependence/inaction; and 5) political and social instability. This are explained below in turn.

Health shocks, natural hazards and vulnerabilities
Health shocks, natural hazards, and vulnerability emerged as the most significant causes of program failure. Most participants believed that unsuccessful households were significantly affected by at least one episode of illness, natural hazard, or vulnerability during the project phase. They experienced a number of critical shocks, such as chronic illnesses among family members, epidemics among the animal population, natural disasters, traffic accidents, and political instability within households or in the community. Although the magnitude of such hazards differed by region, and affected IGAs in varying ways, their effects were unanimously reported as having an adverse effect on improving household economic status and wellbeing.
Health shocks: among the various shocks, health shocks were reported as the primary causes of failure in all regions. Such adverse events created a double burden where families would simultaneously lose income and incur healthcare costs. The severity was likely to be higher in families where a person developed a chronic disease. Chronic conditions (such as uterine tumour/infection, chronic respiratory illness, paralysis, post caesarean complication, and hypertension/heart disease) were more likely to cause families to fail. These conditions resulted in prolonged treatment and caused long-term absences from work. Such chronic conditions worsened the situation when income-earning members were afflicted because they led to loss of work days and regular income of families. In addition, women who underwent surgery due to childbirth or uterine tumours/ infections reported paying large sums that were largely beyond the capacity of the families. Households mostly financed healthcare costs in similar ways: they sold their IGAs, borrowed from relatives, or even sold ancestral properties. Almost all participants reported that such adverse shocks exposed them to catastrophic healthcare expenses. The following case reflects the situation: Anwar Sharif (pseudo-name)-a SHIREE beneficiary of Pirojpur (name of place)-received a cow and five chickens three years ago as livelihood support. He was doing well. But his wife got a sudden illness-a respiratory problem. The condition was getting worse as time passed. He brought her to a Kobiraj (traditional healer) and then met a village doctor. Afterward, he visited Pirojpur general hospital, where the doctor referred her to a modern hospital or clinic, which required a lot of money. He had no alternative but to sell his IGA (cow) to bring his wife to a modern clinic at Khulna. He sold the cow for BDT9500 (considering US$1 = BDT78) as his wife's health was deteriorating. After selling the cow, he brought his wife to a private clinic in Khulna with the help of a neighbour who was employed at that clinic. He got his wife admitted and stayed for three months. The total expense was huge-around BDT13400. All the money spent for the treatment came from selling the IGA. Still, she is not fully fit for doing normal household chores. He stated, 'I have witnessed losing my economic and well-being solely due to the sickness of my wife. Unfortunately, we did not receive any medicine or injection from the government hospital'. (A male beneficiary, Pirojpur, remote location, Oxfam).
Similarly, participants from all sites reported that outbreaks among livestock (pigs, and goats) and fish caused them to fail in achieving successful livelihoods. However, disease outbreaks were mostly reported by households who received livestock and fish farming support, especially in the coastal and Chittagong Hill Tract (CHT) districts. Outbreaks on shrimp farms in the coastal district caused beneficiaries to fail in their IGA operations as well. Such experiences were shared by a beneficiary in Satkhira: I am unsure how I can prevent the disease outbreak in the shrimp population. Last year, I had a loss of BDT30000 due to virus attacks, which is a huge cost for me. (A female beneficiary, Satkhira, remote location, UTTARAN).
Livestock epidemics-especially among goats, pigs, ducks, and chickens-most adversely affected IGAs in Rangamati, Bandarbon, Pirojpur, and Bagerhat, possibly due to climatic characteristics. Beneficiaries who received livestock support under such circumstances were exposed to greater losses in their investments and effort. Conditions worsened since very limited government and NGO supports or facilities were available in or around the communities. The condition is illustrated by this statement: My cows and goats were infected, but I had no idea about treatment or prevention measures. I simply bought medicine from the market and fed the cows and goats. Yet, the animals died, and my hard work and money wasted away. (A male beneficiary, Bandarban, close location, Caritas).
Natural hazards and vulnerabilities: apart from health shocks and disease epidemics among animal populations, participants reported that natural disasters and political turmoil also caused them to fail in their efforts. Natural disasters (reported in 16 out of 49 FGDs)-such as floods, river erosion, landslides, tidal surges, and tropical cyclones-were commonly encountered as obstacles for operating small IGAs. However, the types and magnitudes of natural hazards varied among different regions. Among them, river erosion was strong in the northern and haor regions. Meanwhile, flooding was frequent in the south and char area, and heavy rainfall and landslides were found in the CHT regions. Overall, these natural calamities extensively damaged IGAs in most regions. The following accounts reflect these conditions: Very recently, a storm came and left our crops damaged. That storm ate six months of my hard work. (A male beneficiary, Jessore, remote location, Shushilan).
Similarly, some participants in CHT reported the following: Sometimes, the hill collapses after a heavy rain and we lose our production, sometimes even our houses. (A male beneficiary, Rangamati, Green Hill)

Floods washed away fish farming and destroyed IGAs in many areas, as noted below:
Flooding is a big threat for shrimp cultivation. The cost of flooding is very high and detrimental as it takes the shrimp away. Last season, I had a loss of BDT15,000-BDT 19,500due to floods (A male beneficiary, Shatkhira, remote location, UTTARAN)

Household demography
Our data also suggest that widows, elderly, and physically challenged people were unsuccessful beneficiaries since they had little or no capacity to operate IGAs. Sickness and old age often led to ineffective IGA operation, and, as a result, such people showed little progress over time. The situation was explained by a participant as follows: If the earning person in the family is very old or disabled, then they can't work and earn money. They can't take care of the cows either. (A female NGO staff, Kurigram, remote location, MJSKS) Family composition (reported in 24 out of 49 FGDs) was found to be a major factor in failure as well. Participants reported that a high dependency ratio had negative effects; families with several dependents and few income earners (usually just the husband and/or wife) showed little or no progress. A high number of dependents-such as children, parents, or relatives-not involved in cash-earning activities slowed household economic improvement as income went toward feeding dependents. Families with five or more members and a single income earner seemed to obtain little benefit from IGA support. Still, it was reported that supported IGAs were sold to meet essential needs, such as food, education expenses, and healthcare costs. The situation was explained by a beneficiary as follows: The families having large numbers of members but only one income-earning person are struggling. You see, too many family members survive from a single person's income. That's why the whole family fails. (A female beneficiary, Gaibandha, close location, PAB).
We are an eight-member family. But only a single person earns, and the rest survive on his earnings. In such a condition, it is very unlikely our condition will improve. On the other hand, one of my neighbours, having two earning members, is gradually improving their economic condition. (A male beneficiary, Pirojpur, remote location, Oxfam).

Inappropriate IGA planning, implementation, and monitoring
Selecting IGAs without taking local contexts into consideration was also identified as one of the most significant causes of failure. Participants unanimously reported that IGAs that disregarded contextual aspects were much more likely to cause greater losses in operation. Participants from the southern and coastal areas (for example, Pirojpur and Bagerhat) reported that families experienced a loss of livestock-especially goat, duck, and chicken-since they (NGOs) did not consider local contexts for raising backyard poultry. In most cases of families who received ducks as a means of IGA in low-lying areas affected by the flow of big rivers, that setting was found to be a cause of disease outbreaks. Thus, the families faced greater losses, and their livestock would even disappear within days due to outbreaks. Program staff participants also noted that problems arose since ducks and certain chicken breeds were not adapted to the environment. The promoted breed was collected from outside the locality and was much more susceptible to disease. Most of the poultry population was therefore affected by disease during the early monsoon stage, especially with exposure to flooding and an upsurge of river flow. One participant in Pirojpur explained the situation as follows: The goat and duck I received from the project got sick. I didn't exactly know the causes of the sickness. I called a doctor (veterinary) and paid BDT 800for treatment. But I failed to save my goat. What did I have? No goat and a loss of BDT800. How can I improve my condition? (A male beneficiary, Pirojpur, remote location, Oxfam) I have an unhealthy and sick calf; the calf I got was so small that it looked like donkey and was very weak. Despite proper care and feeding, I was not able to keep it alive. (A female beneficiary, Bandarban, Caritas) The data also suggests that beneficiaries from the CHT and the coastal district experienced severe disease outbreaks among animal populations. A beneficiary believed that disease outbreaks occurred because the livestock's adaptability to the local environment was overlooked. Even the NGO staff recognised that they initially failed to take these factors into account, causing losses in the supported IGAs: Approximately 2% of the beneficiaries were provided inappropriate IGA support. (A male NGO staff, Gaibandha, PAB) First, we distributed pigs and goats to the beneficiaries, which were bought from Bandarban town. After a few months, we noticed those pigs and goats were not able to adapt to the local environment, and all of the goats died. (A male NGO staff, Bandarban, Caritas) Moreover, in some cases substandard assets (especially of livestock and poultry) were purchased because of factors such as time pressure, an inadequate understanding of asset quality (by both beneficiaries and PNGO staff), and weak monitoring and follow-up by the frontline PNGO staff.
Inadequate and inappropriate monitoring of IGA operations was also identified as a failure factor. The field staffs were supposed to be consistent and regular in providing support to beneficiaries, including early problem identification, determining causes, planning solutions, and executing interventions. However, that was not always the case. Beneficiary participants reported that they did not always receive optimal or prompt responses from project staff during adverse situations. For example, some participants said they reported disease outbreaks among the poultry population and sought prompt feedback to control the situation but received little in the way of responses from the PNGOs. Meanwhile, poultry died and families incurred great losses in their means of livelihood. Such delayed and inappropriate responses indicated that, in some cases, monitoring mechanisms and feedback provision for certain IGAs were not fully functional. Loose monitoring and a lack of feedback often led to poor IGA management and caused families to experience losses in their investments. NGO staff reported that they lacked full-fledged monitoring and feedback support for reasons such as inadequate manpower, poor logistical support, poor road networks, and the remoteness of locations. Lack of road networks and remoteness were found to be important causes of inadequate monitoring and delayed responses to adverse situations. Those conditions prevailed mostly in the coastal, haor, and CHT regions. Some NGO staff members described the conditions as follows: We do not have enough manpower to visit each household regularly. (A female NGO staff, Rangamati, Green Hill) As the situation is unstable and there are plenty of ethnic conflicts, PNGO staff from minority groups were afraid of visiting some villages, which made it very difficult to carry out efficient Monitoring & Evaluation. (A male NGO worker, Rangamati, Green Hill) There are some remote fields that require long hours to get to. We have to depend on local transport, especially boats and non-motorised vehicles. Boat communication depends on high and low tides. We often remain inconsistent in paying visits and providing feedback support due to remoteness. That might lead to poor performance in their efforts. (A male NGO staff, Bagerhat, SCiBD)

Expectations
High expectations and inaction with regard to IGAs maintenance were also identified as a cause of failure in some cases. Some participants mentioned that frequent and routinely managed NGO interventions, especially grants, caused some beneficiaries to become inactive. They developed the idea that the government and NGOs would continue their non-refundable support and did not take initiatives to use their assets productively. Such perceptions might have developed during the aftermath of the deadly 2007 cyclone "Sidr" in the coastal region. During that time, the GoB and many NGOs provided emergency responses for cyclone victims through large-scale humanitarian support. Such efforts delivered financial and nonfinancial support, regardless of a community's economic status. Since then, some people expect external support. Such misunderstandings led them to be noncompliant with program support and inconsistent in their responses to interventions. Thus, some beneficiaries described others lacking the motivation to work hard and lacking commitment to improving their own condition. One NGO worker described the situation as follows: Some people think that international donors are synonymous with providers of grant support. Such perceptions might have been realised after the deadly cyclone Sidr in 2007. That time, many international, national, and government agencies operated large-scale humanitarian interventions. But some people think such efforts will continue for long periods. Such perceptions misinterpret the true objectives of a particular project. (A male NGO staff, Pirojpur, Oxfam) According to some beneficiaries inactivity and lack of motivation is a cause of poor program outcome. The following statement exemplifies the situation: Laziness is an important cause of failure as well. Some people do not value the asset since they had received it from the project as a grant. Under the circumstance, they seem to be reluctant to work hard. Such negligent behaviour leads to failure in the process of improvement. In the end, they fall behind. (A male beneficiary, Sunamgonj, remote location, Helvitas) This inaction might point to the irrelevance of the selected IGA in some cases, or also be caused by drug addiction in others. As one beneficiary explained how drugs can affect households' economic behaviour: If they could manage one day's meal, they would not think about the next day's plan. It's very tough to work with such types of people. Household male members' drug addiction is also a reason for deterioration because they take money from household savings and sell assets to manage their addiction. (A female beneficiary, Jessore, Shushilan)

Political and social instability
Political turmoil also affected IGA operations in some areas. For example, one participant in Pirojpur reported that a clash between two political parties took place at the market place, and they (political activists) set fire to a beneficiary's rickshaw. Such instability was found to be particularly prevalent in urban areas more than in rural areas. Beneficiaries in urban slums reported that their IGAs and incomes were adversely affected by on-going political unrest, especially during the first quarter of 2015.
Dowry and social disputes were also found to hinder beneficiary households' improvement. Dowry-related disputes (reported in 10 out of 49 FGDs), frequently reported as a cause of failure in north-western part of the country, especially in the char areas. The following statement helps explain the situation: I got a dairy cow through which my economic situation improved a lot. However, my daughter grew up, and I had to arrange her marriage. I had to sell my cow to meet the wedding expenses and more importantly to pay the dowry. Now, my economic condition is deteriorating. (A female beneficiary, Gaibandha, close location, PAB) Social disputes were also reported as causes of failure. These were more frequently found in the northern and southern regions. In the southern part, many beneficiaries got involved in intracommunal conflicts over the possession of Khas land. Due to the competition for ownership, many became involved in fighting and prolonged and costly faced legal procedures. These adversely affected households' conditions: My family suffered greatly to get access to Khas land. We received nothing but false claims. We are falsely accused of crimes, and we have to defend ourselves alone. (A male beneficiary, Satkhira, UTTARAN)

Discussion
Adopting a qualitative approach, this study aimed to explore the factors affecting progress in antipoverty interventions among extreme poor households in Bangladesh despite receiving inputs designed to do so. We identified several determining factors that contribute to explain program to improve beneficiary households' economic status and well-being. These include health shocks, natural calamities, inappropriate planning for IGA selection, weak monitoring of implementation, household demography, and sociocultural norms and practices. Our findings align with those of international studies conducted in similar settings which argue that each year a higher proportion of member of poor households are forced into poverty due to health shocks related expenditure which is largely financed by the out-of-pocket mechanism (Bhojani et al., 2012;Garg & Karan, 2009;McIntyre, Thiede, Dahlgren, & Whitehead, 2006). The majority of these households met these healthcare cost by selling their means of livelihoods, marketing physical assets, and borrowing money. A study focused in Bangladesh showed that approximately 4-5 million people in Bangladesh are forced into poverty due to the cost borne from healthcare cost (Soor et al., 2015). Although these studies focused on the health shocks related cost of all people, we assume that the vulnerability of extreme poor households in this study will be worse because the poor has relativity lower level of resilience to health shocks. Existing literature suggests that health shocks and illness are more likely to be predominant among extremely poor households, resulting in a loss of income, greater treatment tariffs, and high opportunity costs (Meessen et al., 2003). The literature suggests that members of poor households have less access to healthcare services and receive poor healthcare services (Kabir & Maitrot, 2017;Kabir, Maitrot, Ali, Farhana, & Criel, 2018). Other recent studies have indicated that epidemiological transition is underway in Bangladesh. This has resulted in a higher prevalence of chronic conditions in rural settings (Gupta et al., 2003;Reddy et al., 2007;Turin et al., 2013) across South Asian countries. Given the associated catastrophic out-of-pocket (OOP) healthcare costs, such chronic conditions heavily influence household economic status and well-being.
Our study suggests that most households managed to improve their economic status to some extent but could not sustain it due to family illnesses. Such situations deteriorated when illnesses were chronic (Quisumbing, 2007). Our results indicate that extreme poor households' experiences of chronic illness episodes (such as uterus tumors, cardio vascular, renal, and reproductive health) lead to financial hardship due to their reliance on prolonged facility-based treatment and familybased care. The extreme poor households in this study were found to receive very little healthcare support from the government facilities and sought care from private facilities which entailed greater costs. Our findings are supported by the 2011 Bangladesh Demography & Health Survey (BDHS) report showing that per capita healthcare expenditures were US$27, two-thirds of which were financed OOP (Hamid, Ahsan, & Begum, 2014). Such OOP healthcare costs gradually increased between 1997 and 2007 (Huda, Khan, Ahsan, Jamil, & El, 2014).
Our findings showed that a number of beneficiaries who received chicken, duck, pig, and goat experienced a high incidence of disease among those animal populations. Notably, greater losses were experienced by families with ducks in coastal and lowland areas (haor, for example) and pigs in CHT. Similar disease outbreaks have been reported in other studies. For example, Hoque et al. (Hoque et al., 2011) reported an overall duck mortality of 15% in a coastal district. Another study (Haider et al., 2017) reported higher mortality rates among poultry populations (chicken 61%, duck 47%, geese 73%) caused by avian influenza (H5N1) infection during an outbreak in a northern district. Inadequate knowledge about poultry rearing might have resulted in higher mortality rates (Shanta et al. 2012). Disease outbreaks among pig and goat populations are also reported in other studies (Chowdhury et al., 2014;Kulkarni, Tosh, Venkatesh, & Senthil, 2013). A beneficiary household might be affected by the high mobility and mortality resulting in the loss of IGAs.
Our analysis showed how natural calamities exacerbate beneficiary households' vulnerability, which is concordant with other observations (Quisumbing, 2007;Shahabuddin & Ali, 2006). Extreme disasters-such as floods, river erosion, tidal surges, and tropical cyclones-were frequently reported as causing households to fail in achieving sustainable economic development. Bangladesh is highly affected by the negative consequences of changing climatic features because of its topography and geophysical location. Other studies have shown that natural disasters notably damage lives and livelihoods in rural settings, thus exposing poor households to greater vulnerability (Cook, 2010;Heltberg, Siegel, & Jorgensen, 2009;Henry, 2011).
This study found that household size influenced antipoverty efforts. The larger the size of the beneficiary's household, the more negative the effect on economic growth and well-being. Other studies found that larger family sizes contributed to poverty incidence and poverty endurance in the Philippines and Sub-Saharan African countries (Maralani, 2008a;Orbeta Jr 2005;Virola et al. 2007). However, a study in Indonesia found that the relationships among family size, children's educational attainment, and socioeconomic development differed according to context, varying from positive to neutral to negative (Maralani, 2008b). In line with the abovementioned literature, this study found that larger families (> 5 members) with a single income earner failed to achieve optimal improvements as a result.
Inappropriate IGA selection, planning, implementation, and monitoring emerged as an important factor negatively affecting progress in many households. Avoiding contextual factors in IGA selection, personal enthusiasm, and technical skills exposed beneficiaries to greater losses in IGA operation. Such income-generating projects might be designed and implemented under institutional contexts where multilevel stakeholders and donor agencies interact in complex circumstances (Sanyal, 1991), likely causing inappropriate project planning and execution.
Our data also suggests that high expectations and some form of dependency have in instances a profound influence on participants' motivation and economic behaviour, which are key determinant of program success. In the southern and coastal regions in particular, participants viewed antipoverty responses as regular supports to be continued in the future. Therefore, a negative form of dependency emerged, resulting in a loss of enthusiasm for investing in improving one's condition (Bradshaw, 2007). This contradictors findings from studies conducted elsewhere-for example in the African context-that have found social assistance and antipoverty responses increase poor people's capacity to avail themselves of social capital and natural resources that will in turn enable them to improve their circumstances (Shepherd, Wadugodapitiya, & Evans, 2011).
Social instability-such as political violence and sociocultural norms and practices (dowry, for example)-leads to a loss of IGA and drags beneficiaries into poverty. In Bangladesh, dowry is a common and often socio-culturally accepted phenomenon, especially in rural settings (Quisumbing, 2007). Our data suggest that even if they have the skill and motivation to improve conditions, some families fail because of dowry. In particular, participants in northern Bangladesh faced high dowry demands that contributed to failure and even the loss of IGAs. Participants have reported that they accepted dowry demands to avoid domestic violence (Baulch & Davis, 2008;Davis, 2007;Khan, Rob, & Hossain, 2000). Our findings further indicate that political violence can hinder the livelihoods of those involved in small business and farming activities. Such events have a profound influence in urban settings, especially in the Dhaka slums.

Limitation of the study
The findings of this study were based on a relatively small sample of extreme poor beneficiaries (qualitative study) in Bangladesh. Due to the contextual characteristics, the generalizability of the findings to other areas might be limited. Nevertheless, considering the triangulation of methods and participants, we believe that this study provides an in-depth focus for contextualizing the peoples; lives and livelihoods in understanding the failure factors for antipoverty interventions in Bangladesh.

Conclusion and policy implications
This study revealed several factors that are the prime causes of failure to improve economic conditions among extremely poor households. These include illnesses among family members, disease outbreaks among poultry and cattle populations, ineffective IGA selection, weak supervision from providers, adverse natural events, and negative forms of dependency. Although program support was provided in line with the program design, households slipped into poverty because of limited or non-existent resilience to such adverse events. Most households failed to continue IGAs since they met their immediate needs by selling IGAs and household physical assets, or by relying on credit. Subsequently, households became trapped in a poverty cycle. We argue that livelihood support alone, either through cash transfer or assistance delivery cannot be effective in achieving optimal antipoverty outcomes. A greater emphasis on healthcare services, disease control among animal populations, process monitoring, and effective feedback is needed to promote sustainable livelihoods. These issues should be addressed in future poverty-alleviation programming.