Smallholder poultry model for poverty alleviation in Bangladesh: a review of evidence on impact

S M Fakhrul Islam and M A Jabbar*

Bangabandhu Shaikh Mujibur Rahman Agricultural University, Salna, Gazipur, Bangladesh.
*International Livestock Research Institute, PO Box 5689, Addis Ababa
m.jabbar@cgiar.org

Abstract

Smallholder poultry as a tool for poverty alleviation has been developed and widely applied in Bangladesh. In this paper, the evolution of the model has been summarized and studies conducted to assess the impact of the poultry model at various stages of its evolution have been critically reviewed.

The results indicate that the project has reached the poor though not always ultra poor, the main target of the model, and that participants have benefited positively in terms of income, consumption and nutrition, empowerment of women. However, the results need to be assessed with a high degree of caution because of several methodological limitations of the impact studies. These limitations have been illustrated and suggestions have made for more objective assessment of impact of past projects and for conducting additional research for supporting smallholder poultry as a tool for poverty alleviation.

Key words: Bangladesh, poultry, poverty, scavenging, smallholder, women

Background

Scavenging poultry is a common enterprise in the rural areas of developing countries. In the 1960s and 70s, several unsuccessful attempts were made by donor and national agencies to develop the system through backyard poultry projects and cockerel exchange programmes. These initiatives did not produce sustainable technical and institutional mechanisms to support scavenging poultry development. Since many of the poor, especially ultra poor throughout the world either do not have any livestock or have a few chicken, poverty reduction programmes continued to search for ways in which chicken or other small animals could be used as a vehicle to assist them get out of poverty and ensure food security. Other reasons for the appeal of the concept are:

1. there is a dire shortage at donor project or programme levels for effective interventions to address poverty,
   
   
2. gender is a major focus in development and poultry is an obvious starting point to reach poor women, and
   
   
3. it is appealing to livestock extension and research workers as using this concept they can reach out (complying a and b above) to a much larger population than when they confine themselves to cattle or practically any other animal (Ashley et al 1999; Dolberg 2001).

Smallholder poultry as a tool for poverty alleviation has been developed and widely applied in Bangladesh, one of the poorest countries in the world with over 40% of the population lying below the poverty line. Parallel development of the concept has taken place in a number of countries and adaptation of the Bangladesh model is also underway in a number of other countries with support from various donors including IFAD, FAO, DANIDA, Asian Development Bank (ADB). A number of rapid or extensive survey based assessment of different poultry projects in Bangladesh indicate that the project participants have benefited positively in terms of income, consumption, nutrition and empowerment of women.

However, the real extent of benefits and the degree of success of the model under various projects in Bangladesh remain very fuzzy. For example, the Bangladesh Rural Advancement Committee (BRAC), the non-government organization (NGO) which played the leading role in the development and execution of the model in Bangladesh, reported in 2000 that the semi-scavenging poultry model was being practiced in 380 of the 460 thanas (sub-district or police station) of Bangladesh and that by 1997 1.27 million women were involved in small-scale poultry production under BRAC's poultry programme alone. Moreover, BRAC was supplying one million day old chicks per month, representing 60% of the total day old chick production in the country (Saleque 2000). These figures are much larger than the targeted number of thanas (195) and number of beneficiary women (873,000) under three major development projects, which replicated the model. On the other hand, an IFAD Project Completion Review of these projects in 2002 expressed concern about sustainability of the semi-scavenging poultry production system as the mission found that 35-40% of the beneficiaries in one of the projects dropped out even before the project was completed, and about 50% of beneficiaries in another project seemed to have dropped out after the project ended (Anonymous 2002).

In this paper, the evolution of the semi-scavenging poultry model and its application in Bangladesh is summarized, the evidence on the effectiveness and impact of the poultry model on poverty alleviation and food security is reviewed, the limitations of the previous assessments in terms of methodology, geographical coverage, findings and conclusions are identified, and knowledge gaps and needs for further research to support poultry as a tool for poverty alleviation are suggested.

In section 2, a brief description on the evolution of the model is given. In section 3, impact assessments done by various studies are critically reviewed. In section 4, knowledge gaps and research needs to support poultry for poverty alleviation are described.

Evolution of the semi-scavenging poultry model in Bangladesh

After its independence in 1971, Bangladesh had to deal with widespread and rising poverty and malnutrition exacerbated by several natural calamities. During the 1970s, food for works, food aid and relief were major sources of food security for millions of rural households. Many NGOs emerged and developed at this time alongside established international NGOs undertaking relief and food for works programmes. By the late 1970s some of these national NGOs were trying to combine relief work with development work as a vehicle to create long-term sustainable livelihood opportunities for the poor. BRAC, already a large and highly regarded NGO by that time, was one such organization which pioneered activities for income generation for vulnerable groups of households implemented through its rural development programme.

Formative and development stage

According to Saleque (2000), BRAC considered poultry as a potential candidate activity for income generation among the landless, particularly destitute women, many of whom owned a few chicken. In Bangladesh, poultry is kept by 70-90% of the households, while fewer households keep goats and cattle. Households owning no land or less than 0.5 acre of land own more than 50% of the total poultry population. Poultry is sometimes used as the first investment for a livestock ladder (in the sense that one can move from poultry to goat/sheep to cattle etc) to increase income and get out of poverty. There were almost no job opportunities for the landless, disadvantaged women in the country, who were BRAC's targets for relief and development work at that time. So it was conceived that poultry rearing in which these women were already engaged but at a miniscule level, could be an income earning activity for a large number of landless, poor women. This decision to target poor women was a major factor in the future course of development of this initiative. The belief that relief dependent ultra poor could be helped to undertake some income earning activities starting with a few chicken to gradually move away from relief to self-sustained livelihood activities was the basic foundation of the poultry model that eventually became a major development innovation.

In its beginning, it has some similarity with the founding principles of the Grameen Bank, which started by challenging conventional wisdom that poor had no credit worthiness because they had no collateral to provide as security, they were risky as clients as they would be unable to generate enough income to repay the loan. Working with a few ultra poor households in a village in Chittagong in the mid 1970s, the founder of the Grameen Bank was able to show that those hypotheses were wrong, that material security was not needed for providing loan to the poor and that poor had the knowledge and ability to use credit as a vehicle to get out of poverty, and with additional technical assistance they could do even better. Thus, belief in the poors' ability to escape poverty was a fundamental element in both Grameen's development as an innovative micro-finance institution and BRAC's development of the poultry model

High mortality and low productivity were major problems with scavenging poultry. So BRAC started participatory action research in the late 1970s in Manikgonj district involving poor households aimed at increasing productivity of scavenging poultry. Initially efforts were made to increase the productivity of local breeds through cockerel exchange, i.e. giving improved breeds of cockerels for breeding and facilitating exchange among neighbours. This initiative largely failed because the supply of HYV cockerels was limited, some farmers sold the high value HYV cockerels rather than using those for breeding, and mortality among hybrid off-springs remained high. In order to reduce bird mortality, a trial was introduced wherein poultry birds were vaccinated regularly in five action research villages for one year. The positive results in terms of reducing mortality rate and increase in bird population led to realize that vaccination must be an integral part of any intervention to promote poultry rearing as an income earning activity.

Between 1978 and 1982 the BRAC poultry programme had no model or design, it was being done on an ad hoc basis. The programme included supply of improved chicks, common disease prevention and training in improved rearing under scavenging conditions. It was then decided to involve women in the vaccination work and let them vaccinate for a fee as a source of income, using vaccines supplied free of cost from the local office of the Department of Livestock Services (DLS). However, it was observed that under this programme the government and other firms were supplying inadequate number of pullets for delivery to the participating households and such pullets also faced high mortality in the scavenging rearing system. It was therefore decided to buy day old chicks from the government farms and let selected, trained and supervised rural women rear the day old chicks for two months in confinement in houses built in their homestead plots and thereafter sell the chicks to other women for rearing. The advantage was that the chicks would become better adapted to the rural environment.

By about 1985, these initiatives led to the development of a prototype semi-scavenging poultry model especially targeted to poor women for rural poultry development. The model was a supply chain consisting of the following beneficiaries (Ahmed 2000):

• Model Breeder-Small low cost parent farms with a breeding stock of 54 Fayoumi hens and the requisite number (6) of RIR cocks received either from the project site or directly from government poultry farms. The birds are raised under a semi-scavenging system with balanced rations for producing quality fertile eggs to be used for hatching. The fertile eggs were to be sold to Mini Hatcheries but a substantial amount of the fertile eggs would be sold to the Key Rearers who would hatch them under local broody hens.
  
  
• Mini Hatchery-Small low cost hatcheries operated with solar energy and kerosene stove. Black pillows filled with rice husk were heated in the sun or by means of kerosene and the eggs were placed into a cylinder between two pillows for hatching. Each hatchery had a capacity to hatch 1000 chickens per month. The day old chicks were sold to the Chick Rearers but Key Rearers also would purchase day old chicks to be reared by the broody hens.
  
  
• Chick Rearer-Small rearing farms, each with a capacity of 200-300 chickens per batch and 4 batches per year. The chickens were reared in low cost houses from day-old to 8 week of age. The chickens were fed with balanced feed. The 8 week-old birds were mainly to be sold to the Key Rearers within the same village development committee.
  
  
• Key Rearers-Small farms with only around 5 crossbreed layers for the production of table eggs. The hens were kept under semi-scavenging conditions with 30-70% supplementary feed. Additionally 4 local hens were kept to hatch eggs preferably from Model Breeders and rear chick from Mini Hatcheries.
  
  
• Poultry workers-A numbers of poultry workers were trained to vaccinate the birds to control diseases. The vaccine was supplied free by the DLS through the Area Office of BRAC and the Poultry Workers charged a vaccination fee for providing the service.
  
  
• Feed Seller-The feed sellers were trained to mix feed or sell pre-mixed feed as supplementary feed to the poultry keepers. They prepared balanced chicken rations from locally available feed materials supplemented by purchased nutrients.
  
  
• Egg Collectors-Table eggs were collected from the Key Rearers by Egg Collectors to be supplied to a community sale center or to the wholesaler at the nearby market.

Apart from the above technical components, there was an organizational support system component including training in various aspects of poultry rearing, organizing target participants into groups, provision of credit, input, extension and health services. Several of these inputs and services required access to the DLS, and these were obtained through informal collaboration with the DLS staff and offices in Manikgonj district, where the action research sites were located. The DLS headquarters in Dhaka later extensively examined the model at work in Manikgonj and considered it viable and replicable.

Based on the experiences of the pilot tests in Manikgonj, BRAC and DLS together replicated the model during 1985-87 on a pilot scale in 32 thanas mainly in the northwestern part of the country through 54 Area Offices of BRAC with assistance from FAO/UNDP. The credit component of the model was adapted from the Grameen Bank's group collective debt responsibility approach. In this approach a group was formed with five credit beneficiaries and a Village Organization was formed with eight groups. The point of contact in terms of technical poultry extension and debt collection is the responsibility of a BRAC Programme Assistant who would regularly contact 10-15 Village Organisations representing 400 - 600 farmers (usually women) on a weekly basis (Saleque and Mustafa 1997; Fattah 2000).

The outcomes of the extended pilot project appeared to be positive: bird population increased due to reduced mortality among participating flocks and participating women had increased their income, and through this effort BRAC and DLS, an NGO and a government department, learned to work closely together and they also came closer to people who needed their services. Lessons were also learnt about the advantages and problems of functional relationships among the various enterprises in the model and the optimal size of each enterprise, which helped to modify the model components later (Saleque 2000).

Replication of the model and changes introduced

Based on these experiences, the model was further modified and replicated through three large projects during 1992-2003 with assistance from DANIDA, IFAD, ADB and the Government of Bangladesh (Table 1). These projects targeted 873,000 poor women in 197 of the 464 thanas, mainly in the north, west and south of the country. The main objectives of these projects were to increase per capita income and animal protein consumption among rural poor through participation in the poultry production model. The DLS was the implementing agency in all three projects. Its role was to coordinate, monitor, control and to provide technical support. Specifically, the DLS was responsible for : 1) activities and facilities for implementation of a breeding programme, 2) activities and facilities for establishing a Management Information System, 3) activities and facilities for establishing an international training institution, 4) budgets for applied research activities and a comprehensive human resource development programme.

Apart from BRAC, two other large NGOs- Proshika and Swanirvar Bangladesh- were involved in the implementation SLDP I. Each NGO was responsible in its mandate area for : 1) establishment of an Area Office for each 3,000 to 6,000 women members, 2) selection of potential beneficiaries, 3) organization of village groups, 4) commence a saving programme, 5) training of beneficiaries, 6) creation of an enabling environment by establishing income generating activities such as input suppliers, veterinary service activities, and marketing, 7) provision of loans and assist each of the beneficiary in establishing an income generating activity, 8) technical support for operation of the different activities.

Under one Area Office, there would be on average 3853 Key Rearers, 40 Chick Rearers, 6 Mini Hatcheries, 24 Model Breeders, 10 Feed Sellers, 100 Poultry Workers and 10 Egg Sellers. Thus the model centred around the Key Rearer with some 10 improved hens supported by a number of small entrepreneurs, all available in the village, to provide the inputs and the services needed to maintain the flock. The Key Rearers were wrapped together by community groups, awareness programmes, training and access to micro-credit. The model was designed to create an enabling environment in which all inputs and services needed were available in the village to minimize the risks of investment in a smallholder activity. Even though different enterprises were established as an integrated production chain, each unit would operate on free market principle and was free to sell to customers outside the chain. Furthermore, no subsidies were provided at the user level. A poultry activity was compulsory for the first loan, but after repayment of the first loan the beneficiary was entitled to a new loan for an activity of her own choice.

The establishment cost and the first 3 years operational costs of an Area Office was covered by the project (Donor). After that, the profit margin from loans, sales of inputs and service fees from the 3,000 to 6,000 members was assumed to be enough to cover the NGO's cost of maintaining and operating the office.

Under PLDP and SLDP II, 10 NGOs including the three large ones involved in SLDP I, were involved in implementation along with DLS. The credit funds under PLDP and SLDP II were channelled from ADB through a semi-autonomous apex micro financing organization "Palli Karma-Sahayak Foundation" (PKSF) to the NGOs who were registered as partners of PKSF.

A new phase of the project has been recently agreed between IFAD and the Government of Bangladesh through the Ministry of Finance, with PKSF as the executing agency mainly to support capacity building efforts to make the poultry model more widely used for the benefit of the poor. And this project is first and foremost a micro-credit project but with support for capacity development on livestock to be derived from the DLS on a consulting basis, as most of the micro-finance investment is made on livestock.

An important aspect of the evolution of the model was to accumulate experiences from previous and ongoing projects and ensure that these experiences were reflected in formulation of new projects. Any sustainable model has to be dynamic. A conceptual framework provide elements of the model but these elements- their nature and dimension, may change over time under changing conditions. In the evolution of the model, feedback and learning have played an important role (Dolberg 2001; Saleque 2000). Among the NGOs, BRAC operationalised some informal and formal feedback systems both upward and downward. Feedback took place through the numerous meetings and dialogue that was held regularly at all levels (i.e. village, area, regional and head office level). Feedback from and to villagers (various enterprises) provided a foundation for learning. The Programme Organisers met regularly with village groups, discussing issues and problems. Regional Managers and head office Staff visited village meetings or visit with individual enterprise operator when they were in the field. These meetings, together with informal discussions, formed the basis for feedback of the Programme from and to the village groups. In short, there were strong elements reminiscent of the Kolb (1984) learning circle, which underlines learning on the basis of experience.

Experiences gathered from implementation of SLDP-I were used later in redesigning the model in PLDP and SLDP-II (Dolberg et al 2002). Some of the important changes include the following:

• The use of Sonali crossbreed (Rhode Island Red cock x Fayoumi hen) instead of Fayoumi. On-farm research in SLDP I showed Sonali to be the most appropriate breed for semi-scavenging system (Rahman et al 1997).
  
  
• The introduction of feed choice system in PLDP to allow hens of Key Rearers to choose own feed supplement. It used a simple bamboo pole split in half as a trough with three compartments for the protein (soybean), carbohydrate (maize) and the shell (oyster) supplements.
  
  
• Introduction of hatching and rearing of Sonali chicks by the Key Rearers in PLDP using local hens. This means the Key Rearer could produce their own replacement pullet, and eliminate the need for both Mini-Hatcheries and government farms as sources of day-old chicks.
  
  
• Making mini-hatchery an optional enterprise in the model as this faced difficulty to be competitive in the market due to subsidized day old chick supplies by the DLS. Moreover, it was the most management and labour demanding enterprise as eggs had to be turned every four hours including at night, which poor women had difficulty to cope with. Mini-hatcheries appeared to work better for duck eggs, which have thicker shells and are less likely to break than chicken eggs. The hatchability of duck eggs was reported by PLDP to be 80% to 90% as compared with 50% to 70% for chicken eggs.
  
  
• Number of beneficiaries per area office was increased by about 50% requiring support staff to handle more households per head, the loan size was increased considerably, for key rearers up to 100%. However, the poverty criteria used in SLDPI seem not to have been used in the later projects, thus allowing field staff of most of the new implementing NGOs to select relatively well-off households as participants. Moreover, in SLDPI, the three large NGOs involved had an exit strategy. Once the project was finished, the participants were to be transferred to regular poverty alleviation and rural development programmes of these NGOs, so participants would continue to have access to credit and other services for poultry or any other economic activity they wanted to pursue. The smaller NGOs participating in PLDP and SLDPII apparently did not have any exit strategy, nor did they have infrastructure or experience in livestock related programmes to implement an exit strategy. This aspect might have contributed to larger drop out rates and poor performance of the later projects, as will be shown later.

Finding the best breed under scavenging or semi-scavenging conditions remained a problem. In order to support adaptive research to determine the "best" breed or breed combination under scavenging/semi-scavenging systems, an allocation of US$ 800 000 was made for field research within SLDP II.

Impact assessment on the BRAC-DLS model

Both SLDP and PLDP project designs required establishment of systematic monitoring and evaluation procedures by the implementing agencies (DLS and the concerned NGOs) to monitor the progress in implementation and performance of the projects as well as assess impact of the projects. However, DLS was not successful in establishing an effective monitoring and impact assessment system in either SLDP or PLDP. Most of the results reported on the progress of implementation of the model came from participating NGOs as a part of their routine programme report without detail analysis. Donor evaluation missions consisting of consultant teams produced some field observation based reports without using any systematic impact assessment design (Dolberg et al 2002).

In the absence of systematic monitoring and evaluation within the projects, several small and large studies have been conducted at various stages of the projects specifically to assess their impacts. An inventory of these studies is shown in Table 2.

The BRAC-DLS poultry model was replicated through three large projects, each implemented in different geographical areas, in different periods of time with some overlap, by different combination of NGOs with varying capacity and experience. Although the basic poultry model structure remained fairly similar, several changes were introduced at different stages. All of these factors might have implications for how the model performed under different projects, how the various technical, institutional and organizational components of the model performed under different projects and situations, what impacts the projects and their enterprises made through which pathways. If beneficiaries succeeded to get out of poverty, how did they do so and what did they do once they passed the threshold of poverty? If some failed to successfully use the enterprise to help them get out of poverty why did they not succeed? These and other related questions should have been addressed through appropriate sampling frame and analytical approaches to assess the impact of the poultry projects under discussion. But most of the studies listed above suffer from important sampling and analytical limitations, which need to be kept in mind while interpreting their results and their implications. So these limitations are summarised below before discussing the results.

Methodological limitations of the impact studies

Sample size

Given that each project had a large number of target beneficiaries, a representative and adequate sample size would be needed to draw statistically valid conclusions. Due to various factors that might influence performance and impact of the model as outlined above, application of appropriate stratification criteria would be needed to draw representative samples. Alam (1997), Nielsen (1998), DARUDEC (2003) and Raha (2003) used fairly large samples but they did not consider all the important stratification requirements in sampling design. Raha selected three upazilas purposively in accordance with the advice of the NGO, which had operational mandate in those upazilas, and the upazilas with reasonably high success records were recommended and selected. Moreover, only households that were operating at the time of the survey were selected, consequently the results are biased upwards and can't be generalized. Seeberg (2002) and DARUDEC (2002) used very small, unrepresentative samples drawn from observed high performing areas often at the suggestion of the implementing NGOs, so they did not allow rigorous statistical analysis and inference. Rahman et al (1997) also had a fairly small sample though it was subjected to statistical analysis. Nielsen (2003) used a case study approach with a fairly small purposively selected sample, which were interviewed at three different points in time. Case study approach often allows in depth analysis of representative cases for drawing general conclusions, hence Nielsen's results could be considered valuable but by her own account, the samples were selected by the NGOs operating in the selected project areas, so there perhaps was a bias towards selecting more well-off and more successful cases, thereby limiting the degree of generalization that can be derived from the study.

Basis of comparison

The projects operated under a dynamic socio-economic environment where poor peoples' conditions might have changed to some degree without the poultry projects. Therefore, to assess the net effect of the projects at household or higher levels, both before-after (for participants) and with-without (including both participants and non-participants or control group) comparisons would be needed (Table 3).

Either before-after or with-without comparison on its own may generate biased estimates of the effects of the project if there were initial differences between participants and non-participants and/or there were changes in peoples' conditions (positive or negative) without the project. The before-after comparison also needs to address the time path of adoption as not all participants joined the project in the beginning or at the same time, and problem of using constant or current prices in valuing products and inputs. If the number of participants joining the project has an unequal (skewed) distribution over the life of the project, the sample should also reflect that distribution in order to obtain an unbiased estimate of the impact of the project.

Another aspect of sampling for impact assessment is how to treat active vs dropout participants. Empirical studies on agricultural technology adoption generally divide a population into adopters and non-adopters, and analyse the reasons for adoption or non-adoption at a point in time and then the impact of adoption may also be measured. In reality, technology adoption is not a one-off static decision rather it involves a dynamic process in which information gathering, learning and experience play pivotal roles particularly in the early stages of adoption. The adoption pathway may involve a process in which farmers move from learning to adoption to continuous or discontinuous use over time (Jabbar et al 2003). Inclusion of drop outs in the sample and analysis of their profiles are very important where high drop out rates, permanent or temporary, are observed, as apparently the situation in all three poultry projects under discussion.

All the impact studies under review except that of Nielsen (1998) and Nielsen (2003) used only simple before-after comparison, some included dropouts in the sample but in most cases the depth of analysis remain low. Only Nielsen (1998) used a before-after as well as with-without sampling frame including dropouts but here also the depth of statistical analysis to draw robust inferences remain poor. Nielsen (2003) used a panel collecting data for three points in time, including dropouts so could provide a more robust basis to assess the impacts if the sample size was not too small.

Impact on the beneficiaries

Types of beneficiaries

Alam's (1997) sample comprised 75.2% key rearers, 6% chick rearers, 4% model rearers, 4% feed sellers, 0.8% mini hatcheries and 10% poultry workers and no egg sellers. Male and female headed households in the sample were 76.5% and 23.5% respectively. The study mentioned that the number of samples in each beneficiary group was more or less proportional to the total number of beneficiaries in each group (Alam 1997, p2). In theory, all the enterprises in the model are targeted to poor women and about 95% of the beneficiaries are supposed to be Key Rearers around whom other enterprises are built. Therefore, there seems to be a major divergence between the expected theoretical composition of the model and its actual composition in the field. Most likely the imbalance occurred because egg sellers were not yet a part of the model when this survey was done and Alam did not include dropouts (majority of which would be key rearers, the largest beneficiary group) in the sample rather took a reasonable number from each of the other categories to allow meaningful analysis.

Nielsen (1998) did not give a breakdown of its 1085 samples according to active beneficiaries and dropouts by type of enterprises, and control group. However, an aggregate gender breakdown was given showing that among the sample active beneficiaries 15.4% were female headed compared to 8.6% for dropouts and 12.2% for control group. Others were male-headed households, which indicate that either the sample was biased toward male headed households or that the project did not really reach its targets who should be principally women. Among the active beneficiaries, 44% could read and write, 52% could sign and 4% were illiterate; the corresponding figures for the control and dropout groups combined were 25, 40 and 36% respectively.

Similar problems in the sampling composition to identify beneficiary types remain in most other studies. For example, Swan (2000) identified three categories of beneficiaries as follows without giving any quantitative figure about the distribution of these beneficiaries: (1) women from poor and landless farmer households and female-headed households; (2) poor landless farmers who operate less than 0.5 acres (0.2 ha) of land and depend on the sale of more than 10 days per month of their manual labour as the main source of income; and (3) poor and marginal farmers with between 0.5-1.0 acres (0.2-0.4 ha) of land and an average daily income of less than Tk 17/day (about US$ 0.35/day or $ 128/year). Seeberg (2002) reported that 74% of the beneficiaries had less than 0.50 acre of land and 12% had 0.51-1.00 acre and 4.3% had 1.01-1.50 acres. This is also an indication that a good proportion of the participants were well-off households (with over 0.5 acre land and perhaps some livestock, as land and livestock ownership has good correlation), who were not the basic target of these projects.

The complexity and capital and skill requirements are not the same for the different enterprises, and the enterprises need to be adopted as independent individual choices on a voluntary basis but function as an integrated chain. Therefore, finding the ratio of different entrepreneurs and their profiles (asset and income base, skill level, demographic characteristics, market orientation and access etc) would be useful in explaining which type of poor women adopted which enterprise and why, what participant characteristics contributed to non-participation, participation, success or failure and why there was an imbalance in the model composition, if any. Such detailed analysis was missing in all the studies.

Some studies mentioned that benefits from the poultry projects have spilled outside the model participants. Rather than using the loan for poultry, some members used part or entire loan for other enterprises, some initially took up poultry but gradually moved into other things such as rickshaw pulling, petty trading. But these results could also be interpreted differently. Rather than spillover benefits, they might indicate and explain the high drop out rates (see below). Perhaps these participants did not have any interest in poultry in the first place but joined the project to access credit and soon moved out of poultry to engage in what they wanted to do. This implies that given freedom to choose a business with micro-credit, not everybody would take up poultry as the initial or first choice as required by the poultry projects. Mini-hatcheries and model breeders were supposed to sell their improved chicks to other enterprises in the model but some sold to local farmers outside the model either because there was inadequate number of key rearers in the model or because they produced a surplus. Other secondary beneficiaries include farmers in the locality who bought vaccination services from the poultry workers, and feeds from feed sellers. The extent of such spillover beneficiaries of the project has not been analysed in any study.

A more fundamental question is that the concept has been developed to target poor women, especially the ultra poor, but it is unclear from the various impact studies if the poorest women were targeted and reached in the various projects. An objective characterisation of the actual beneficiaries is essential because, while discussing the principles and problems of adaptation of the Bangladesh model elsewhere, Jensen (2002) mentioned that " it is surprising so many organizations and individuals have a policy to target the poorest, but either on purpose or in reality exclude the poorest segment of the population from their activities. Common phrases are : the poorest do not have the capability to learn, the poorest are lazy, the poorest have chosen to live in poverty, or it is better to start with the better off and then the poor will benefit through a trickle-down effect -an approach which has been rejected long ago".

Drop out

None of the formal impact studies tried to estimate drop out rates though some studies included drop outs in the sample. Nielsen (1998) reported drop out rates of 2-3% based on estimates of the implementing NGOs. Quoting PLDP monitoring system data, Nielsen (2003) reported that between 2001 and 2002, 59% of the Key Rearers in the PLDP project areas, who constitute 95% of the target beneficiaries, have dropped out. In her own case studies, 26% dropped out completely between 1999 and 2003 and another 30% stopped PLDP activities but have maintained contact with the NGO for savings and participation in group meetings. This latter group perhaps continues to maintain contact with the NGO to access credit to undertake activities of their own choice in the future. An IFAD review and evaluation mission reported that 50% of SLDP participants appeared to have dropped out after the project ended and 35-40% of PLDP beneficiaries dropped out even before the project was completed (Anonymous 2002).

Nielsen (1998 and 2003) mentioned losses suffered due to high chick mortality, lack of technical services and low productivity, problems in repaying loan, family problems e.g. advice of husband to stay away to avoid being indebted, as reasons for drop out. About 33% of the dropouts showed interest to join the project again. There are indications that some people dropped out because they joined the model only to obtain credit and initially adopted a poultry enterprise but soon moved into their choice of business leaving poultry partly or fully. The IFAD mission mentioned several reasons for drop out : (a) small loan size and very small poultry enterprise do not provide a pathway out of poverty, (b) some poultry enterprises, especially mini hatcheries and model rearers, were not profitable, partly due to low market price of eggs and other market distortions, (c) the design being fixed with type, scale and relative number of enterprises, adopters did not always had the freedom to choose the enterprise they liked best, (d) the ratio of participant to NGO field staff was much higher than normal NGO micro-credit programmes, especially under PLDP and SLDP-II, hence supervision, training, input supply and services and regular contacts with participants could not be maintained at optimal level.

These features led the IFAD mission to express concern about the sustainability of the poultry production system. A detailed analysis of the drop out rates and profiles of drop outs vs those who stayed in the business might be very helpful in increasing the sustainability of the model and the projects as targeting potential adopters and their training and other needs may be more objectively and accurately done in future projects.

Profitability of poultry enterprises

A number of studies assessed profitability of different enterprises and found them positive. Some of the important results are summarised in Table 4. The absolute values of profit for some enterprises appear to be very low while for others they are reasonably good. However, there is no detailed analysis of profit variability and related causes or factors. Cost of day old chicks, supplementary feeds, if any, health costs and output prices may have influence on profitability. Also there may be differences between enterprises, locations and operating areas of NGOs. For example, the IFAD review and evaluation mission observed that mini hatcheries and model breeders found it difficult to run a viable business producing hatching eggs and chicks because DLS farms supplied chicks at below commercial prices. Mini hatcheries also appeared to be technically difficult to operate because of high labour and management demand. The mission also observed egg prices lower than usual in several locations, especially near the Indian border as eggs come across the border and drive down local prices (Anonymous 2002).

Market access for products may be a major factor in profitability. The implicit assumption is that poultry producers in the model should have no problem selling their small quantities of outputs locally. Egg sellers in the model buy eggs and sell usually in local markets. But if a large proportion of the poor families take up poultry and expand the size of the enterprise under semi-scavenging conditions, aggregate increased production may glut local markets unless traders can access larger urban markets. Moreover, increasingly these bigger markets and towns may be supplied by the rapidly expanding private commercial sector. Commercialisation may lead to falling poultry prices relative to feed cost but commercial farms may remain competitive due to economies of scale and they may push scavenging and other small-holder poultry units out of business. It is therefore necessary to analyse the competitiveness of scavenging/semi-scavenging poultry and identify production and market niches where they may have potential for survival.

Income of the beneficiaries

Several studies assessed income from poultry and its contribution to total household income. However, it is unclear how poultry income and total household income have been actually defined and if all the definitions are the same or comparable. It is also unclear if current or inflation adjusted prices have been used because inflation and currency devaluation would make comparison of values generated at different times difficult. Results derived from Alam (1997) are summarised in Table 5. The average weekly household income of the beneficiaries increased from taka 268 to 398, i.e by taka 130 or 48%. This was more or less equivalent to the income derived from the poultry enterprise at the time of the survey though in a few cases, poultry income was higher than the overall change in household income. The survey also showed that 28% of the households had increased their income above the poverty line (Nielsen 1997 based on Alam 1997) although no details was provided on whether they could move up any further after crossing the poverty line and what pathway they took to move up the ladder. Therefore, it appears as though the entire change in average household income came from participation in the poultry project. Poultry income accounted for about 35% of household income at the time of the survey, the figure in the beginning of the survey was not available. There was considerable variation in household and poultry income and its share between the different enterprises and also between different districts.

Neilsen (1998) found that net poultry income and its share in total household income were higher for project beneficiaries before the project and increased five times among project beneficiaries but only marginally among the control group(it is not clear if control group include the dropouts) (Table 6). Project members' own contribution to household income increased from 16 to 30% among the beneficiaries but only from 9 to 10% among the control group. Breakdown by type of enterprises was not available nor was any indication of the extent of variation across districts and NGOs.

Nielsen (2003) found average monthly beneficiary income of key rearers from PLDP poultry project activities was Tk 1470 in 2003, which was higher than the average for all PLDP activities though this comparison was perhaps not meaningful as the sample of different enterprises was not proportional. About 75% of the beneficiaries had more than one source of income in 1999 as well as in 2003. Average monthly income from non-PLDP activities increased from Tk 255 in 1999 to Tk 775 by the end of 2002.

DARUDEC (2002) showed that average monthly income of different beneficiary groups increased by 14-163% (Table 7). However, the small sample size and its lack of representativeness may have highly biased some of the estimates.