Aga Khan Foundation, Sandra Calligaro

Download this publication

How to cite this publication:

Magnus Hatlebakk (2022). Community driven development in fragile contexts. Bergen: Chr. Michelsen Institute (CMI Report R 2022:2)

This review of community driven development discusses the incentive problems that arise when an external agency attempts to involve the local community in the planning and implementation of development projects. 


This report is part of a review for the Ministry of Foreign Affairs, Norway, of Norwegian support for Community Driven Development in Afghanistan. Afghanistan specific documents and data collection are reported in the main report, while we here report on a desk review of the literature on community driven development, with a focus on experiences from countries and areas of fragility. Our general findings are included in the main report, while we here add empirical evidence from other countries than Afghanistan.

Magnus Hatlebakk
CMI, Bergen, May 2022

This report is commissioned by the Ministry of Foreign Affairs under a framework agreement between the Ministry and Chr. Michelsen Institute (CMI).



This review of community driven development discusses the incentive problems that arise when an external agency attempts to involve the local community in the planning and implementation of development projects. There is a risk of local elite capture, both of democratic processes and the benefits of the programme. This is a complex strategic situation where the sequence of decisions may affect the final outcome. It may matter whether the donor, local leaders, or the local population moves first and state their preferences for what project to fund and how to implement it, and what steps that follow in the decision process and ultimately implementation of the programme. The review discusses some of the game-theoretic models that have been applied to analyze this strategic game. Then the review goes on to discuss empirical evidence, in most cases from randomized interventions, with a focus on fragile contexts. Most of the projects come in form of World Bank supported village level infrastructure projects. As we may expect, these improve the local access to infrastructure, while the impacts on democratic processes are more mixed. Also the degree of elite capture vary between contexts. There is some evidence that one can counteract elite capture by involving alternative local experts in the implementation of the programme, preferably people identified with help of the local population. 

1. Introduction

Community driven development (CDD) is not a well defined concept, and tends, strangely, to describe development projects that are externally funded and monitored against goals set by external agencies. The community element includes attempts at involving the local population in identifying target groups, sub-goals, prioritizing between sub-projects and spending of funds, and/or involvement in implementing the projects in terms of labor, co-financing, or management. The funding agency will normally have goals that can reasonably be linked to community development, such as local governance, democracy and an active civil society, and quite often the projects focus on sectors that are likely to have direct local impacts such as education, primary health, microfinance, agriculture, microbusiness, and local infrastructure. The CDD term thus covers a wide range of development projects. One may argue that the main success criteria, which will separate it from any other development project, should be the degree of involvement of the local population. While general impact evaluations that focuses on other goals, such as poverty reduction, income generation etc, should be secondary, unless community involvement is considered as a means to reach these other goals, rather than being a goal in itself.

Below we will review both issues, that is, CDD as community involvement, and as a means to reach other targets. In both cases we will focus on externally funded projects, and thus not review variations on, and preconditions for, the potentially more effective community driven development that happens independently of national governments or aid agencies. The narrow focus is motivated by a need to judge the efficiency of aid funding. We will still bring independent community development into the picture, as this will be an alternative to any externally funded development project, and thus constitute the control where the community manage without external aid. One may well imagine that external projects have even a negative impact on local development, if they crowd out local initiatives. The review will start with a general discussion, and then switch towards a review of empirical contributions, with a particular focus on fragile contexts. The review is motivated by Norway’s extensive support to Afghanistan, and the review of empirical evidence thus starts with Afghanistan, and then goes on to comparable countries in conflict, such as Myanmar and Sierra Leone, for which we have found relevant research.

2. CDD as community involvement

A core aim of CDD projects is to involve the local community in the development process. Ideally this includes involvement at all stages of the project, from setting the goals, via identifying target groups, prioritizing sub-projects and budgets, to monitoring implementation and evaluation of the results. But since CDD projects, of the type we discuss here, comes with external funding, there will immediately be competition for those funds, and local strongmen will automatically be better positioned to take their share. Such local capture may even be appreciated by the local population as they in fact understand the political economy of aid, and thus that the externally funded projects may not have appeared in the first place without the efforts of the same village leaders. What will then be the best strategy for an aid agency that also knows the political economy of aid? That is, how can one ensure real local involvement beyond local leaders? This is covered by the literature, normally using principal-agent theory, which is a sub-field of game theory where the principal (the aid agency, or even the local elite) moves first taking into account the optimal response from the agents, potentially with the local leaders moving before the general local population in a three-step model, or even the local elite moving before the funding agency.

Note that a core problem will be that each local community will support their own leaders in the competition between villages for external funds, and in the process probably accept that the leaders take a share of the allocated funds. Such rent-seeking activities can in principle be minimized by use of objective targeting criteria and uniform programmes across villages. This will, however, be in conflict with the ideals of community involvement, where each village may come up with its own ideas for local development. The solution may be to allocate funds according to a pre-defined rule to avoid rent-seeking activities, but leave the villages to decide how to spend the funds.

But even if competition for external funds is minimized, the within village competition for the distribution of funds will still exist. The donors may attempt to side-track village leaders by working with NGOs or other community based organizations (CBOs). But this may only move the political game to these organizations, with local leaders maneuvering to control the local NGOs and CBOs. Thus tilting the power dynamics within a village is far from straight forward, and may not even be preferable, as it can be seen as an attempt to overrule local democracy. Thus community driven development may in reality imply to work with local leaders, and thus implicitly accept that they receive their share of the funds. Some donors will, however, attempt to tilt the local balance of power by working with alternative local leaders, such as women’s groups, trade-unions, human rights groups, and other community organizations. Before we go on to the empirical literature, what can we learn from the formal modelling of these complex principal-agent relations? 

3. Strategic aspects

It is useful to discuss in some detail the strategic relations between a donor agency, village leaders and the local population, where the latter may be split into a target and non-target group. The donor agency comes with a budget and normally some goals. In the simplest case the goal is to let the population decide how to spend the budget. In general, however, the donor will have more specific goals, such as poverty reduction. The latter requires that the target group of poor people is identified. The local population want to benefit from the budget, and realize that this may depend on coordination among themselves and the behavior of the donor and potentially the village leaders. The village leaders may care for the population as well as their own welfare. This all sums up to a very complicated game, with different interests within each group, potentially also within the donor agency. The outcome may depend on the sequence of decisions made by different actors, as indicated by the game-theoretic literature mentioned above.

The village leaders may first suggest a use of the budget, which within a CDD framework may then be discussed with the villagers before it is submitted to the donor agency. Or the donor may suggest a use of the budget that they discuss directly with the village population, which will then include the local leaders. The first will be a sequential game with three steps, the second a sequential game with two steps, where the village leaders and population need to coordinate in the second step. In both cases the first movers may take into account what they expect to happen in the following step(s). In theory one may also imagine that the villagers suggest a use of the budget that is sent either directly to the donor, or via the local leaders.

In sequential games the last mover will normally loose, in particular if the first mover knows what the last mover will accept. In that case the first mover can offer as little as possible. If the first mover, such as an aid agency, care about the last mover this may no longer be the case. This, in turn, implies that the donor may prefer to move first and suggest a use of the budget, which in turn is taken to the villagers, while taking into account that the local leaders may have different interests than the population. While in a proper CDD framework one may imagine that villagers are allowed to move first, and suggest a use of the budget. If village leaders are able to get involved in that decision they may take a larger share than in the case where the donor agency suggests a budget.

Now this game has only described one village, but there may be competition between villages, so that the game is not only about the use of a fixed budget, but also a competition for the size of the budget. In this case the villagers may prefer to involve local leaders as they may be better placed to negotiate a larger budget. Thus for each village a CDD approach, where the villagers coordinate with the village leaders, is preferred to a standard model where the donor moves first. For the country at large, however, this may lead to rent seeking activities where leaders from all villages compete for the total donor budget. This problem can be solved by the donor by implementing simple allocation rules that does not allow for rent-seeking activities.

The conclusion is that a donor agency should not open up for costly negotiations about the distribution of funds within a country, but rather go for some simple allocation rules when it comes to the budget. This may still allow for local discussion of the within village allocation of budgets. If villagers understand that the total budget does not depend on the efforts of the local leaders, then they may end up with more say in the final use of the money.

The final outcome may, however, depend on the degree of local control in the hands of local strongmen. The game we have described is part of a larger power game, and any surplus villagers gain from a programme can potentially be confiscated by other means. So a donor should take into account these broader power structures, and attempt to find ways of strengthening the bargaining position of the poor and other marginalized groups. Again strategic aspects are important, the donor should attempt to strengthen the outside option of villagers to make them less dependent on local strongmen. This may involve transfer of assets, education, or credit and labor programmes that may provide villagers with an alternative to interactions with local employers and moneylenders. In doing so the donor may, however, have to take into account the local strongmen’s ability to hinder implementation of the development programme.

4. Empirical evidence

The strategic aspects discussed in the previous section imply that the outcome of any CDD programme will depend on the local context, in particular the internal power relations both at the village and national level, as well as the design of the programme. This in turn means that any impact evaluation of a particular programme has limited external validity. We will still discuss some findings, and with an attempt to keep the discussion of strategic elements in mind. The focus will be on fragile contexts. We start with Afghanistan, then gradually move further afield. There are already other reviews, and we rely to some extent on these.

4.1 Afghanistan

A major CDD programme is the National Solidarity Program in Afghanistan. A large Randomized Controlled Trial (RCT) was conducted to measure the impacts of the programme, which in turn led to a series of articles in good journals. These covered a number of outcome variables. The programme allocated block grants according to a pre-defined rule, which will thus leave for the local community to decide only on the allocation of those funds between projects. The RCT was design to investigate precisely what we have discussed above, that is, allocation by way of local discussions (where local leaders may affect the outcome) versus direct decisions by the population (via a referendum). As expected it was found that community discussions led to elite capture, while project implementation and welfare measures were not affected by the allocation method. That is, welfare improved equally much, and compared to a (third) control group of villages there was progress on access to drinking water and electricity, acceptance of democratic processes, perceptions of economic well-being and attitudes towards women.

A similar programme implemented in Afghanistan was also studied with help of a large RCT. That is, the Targeting the Ultra Poor programme. This builds upon a model implemented in a number of countries, which is found to lift people out of poverty, also in the longer run. The idea is that poor people meet multiple constraints, and if all constraints are targeted at the same time, one can lift people permanently out of poverty. It is also found that if one drops some of the programme components, then the programme may not work. The underlying constraints that keep people in a poverty trap are a combination of being initially poor (lack of assets, human capital, and collateral to raise loans) and market failures that result from a combination of people being initially poor and other underlying constraints (asymmetric information, large fixed costs) that limit competition and may leave people even without market access. Multifaceted programmes that target all these constraints at the same time will potentially help people out of poverty and may thus only need to run for a limited period of time. Similar to findings from other countries, in Afghanistan it was found “significant and large impacts across all the primary pre-specified outcomes: consumption, assets, psychological well-being, total time spent working, financial inclusion, and women’s empowerment”.

The ultra poor programme is not a full scale CDD programme, since the local community is not involved in the design of the programme. But it has a CDD component in terms of a Participatory Rural Appraisal (PRA) to identify the poor. And with the focus on village level poverty traps, it is very much a community development programme, and thus an alternative against which full-scale CDD programmes can be judged. Another similarity is that both programmes collaborated with the Community Development Council in each village.

The findings from Afghanistan seems to confirm findings from broader reviews of the literature. CDD programmes, or programmes having CDD components, appear to be effective in providing services for the poor, while the impacts on community involvement are more uncertain, and may depend on the design of the programme. One review concludes: “I … find that CDD effectively delivers public goods and modest economic returns … There is little evidence, however, that CDD transforms local decision-making or empowers the poor in any enduring way.” Another review agrees on the first point, and is slightly more positive on the second: “The empirical evidence … confirms that community-driven development programmes provide much needed productive economic infrastructure and services … They also provide villagers … with a voice in how development funds are used to improve their welfare.”

4.2 Pakistan

Afghanistan’s neighbor Pakistan has had a similar World Bank supported CDD project. We have not been able to find a proper impact evaluation of the scheme. There is one project document that is named as an impact assessment, where there is a control group. The focus in on an irrigation scheme, and it is considered as positive that all in the treatment group cultivates land, compared to only 64% in the control group, thus clearly not a balanced sample. The control group seems to be selected from the same areas, but we find no further information on how the control group is selected. The control group is also not really utilized, and there seems to be a very low response rate in the treatment group in many of the tables.

There is, however, an earlier study of community involvement in infrastructure projects. This study focused on community maintenance of projects that were initially externally initiated (by the government or the NGO Aga Kahn), and what was termed as comparable projects. 33 villages had at least two such projects, allowing for within village comparison, while another 66 villages had only one project, thus in total 132 projects. These were a subset selected among 651 projects conducted in 99 villages in the north-east Himalayan corner of Pakistan. Most of the projects were related to irrigation and regulation of rivers. The current state of the project was the outcome measure, and the research team investigated whether characteristics of the project, or the local community, affected the level of maintenance. They find that community participation in the project matters for maintenance, but in two different ways, with one effect being negative. Community participation in non-technical decisions (selecting project type, usage rules, etc.) positively affected maintenance, while community participation in Technical Decisions (deciding project design, scale, etc.) had a negative impact.

4.3 Myanmar

Myanmar has implemented a flagship CDD programme with support from the World Bank and other international donors, which is thus similar to the National Solidarity Project in Afghanistan. Any evaluation of the programme has to rely on ex-post data, as concluded by a pilot evaluation: “The design of an impact evaluation of the NCDDP is not straightforward. Lack of a baseline survey and non-random assignment of townships restrict the available options”. The pilot evaluation ended up recommending a regression discontinuity design (RDD), which is a standard approach when RCT is not available. The idea is to use neighboring villages as the control group. In the pilot they had only 19 villages and no significant findings to report. We have not been able to find a follow-up full-scale version of this evaluation, or any other attempts at establishing a counterfactual for an impact evaluation of the NCDDP.

The closest we get is a descriptive study that compares the NCDDP to a similar programme supported by Korea, where the latter to a larger extent collaborated with the government both at the central and local level, and thus used government extension workers as change agents in the local community. The World Bank supported project, in contrast, intentionally attempted to avoid government agencies and worked with NGOs and other private actors. The authors use the two programmes as examples of respectively a developmental state and a neoliberal CDD modality. There is no attempt at providing evidence in support of either model, but the reader gets the impression that they do not favor the neoliberal modality, although there are no strong arguments raised in favor of collaboration with the Myanmar government. It is not clear to us, though, why the use of NGOs in the NCDDP defines it as a neoliberal programme. To us it appears that NCDDP is a standard CDD programme, while the developmental state modality seems to fit for the comparison programme.

4.4 Indonesia

In Indonesia different modalities of community involvement in identifying target groups have been implemented. One of these is the governmental PKH cash-transfer programme. A recent study used the expansion of the programme in 2010-11 to collect data prior to an intervention where the involvement of elites was randomized across community meetings that were used to identify target groups. In this programme they found no elite-capture. They also measured elite capture in some other ongoing programmes, where it was found to be relatively low. Based on this study, the authors argue that elite capture of community driven development is maybe a smaller problem than one may expect, and that “improving the administrative and management skills of local leaders may contribute more to welfare than cutting them out of the whole process to avoid capture, even if this means that the elites sometimes pocket some of the resources”.

4.5 Philippines

The Kalahi-CIDSS CDD programme is implemented by the government with support from the World Bank and the US via the MCC programme. The CDD programme has a large number of sub-projects (about 40 000 at the time or writing). These projects are implemented at the village level. Government facilitators trains community members and local governments in choosing, designing, and implementing the projects. The selection is done at a higher (municipality) level than the village, but based on locally selected criteria to judge proposals from different villages. Most subprojects have been implemented within six months. Projects include flood control, roads, schools, from what we can see mostly infrastructure. In one RCT municipalities were first matched on observable characteristics, and then for each of 99 pairs of municipalities one was randomly selected to receive the programme. Baseline data were collected from 6000 households in 2011 and endline data in 2015. In an earlier RCT there was eight matched pairs of municipalities with 66 treated villages and 169 control villages. In this early study there was 2400 households in the baseline from 2003, a second round in 2006, and a final round in 2010.

From 2003 to 2006 Labonne and Chase found that “the project increased participation in village assemblies and the frequency with which local officials meet with residents.” Saguin studied the change from 2003 to 2010, using the same data, and found that the increase in participation in village assemblies happened only among non-poor households. Labonne, in a longer report, investigated a number of indicators, and found that income increased with about 12%, more households (9%-points) got year-round road access, and more households were willing to contribute to local projects. This report, however, tested many hypotheses, so that some findings may turn up as significant just by coincidence. There are 40 tables with impact estimates, but many of them measure basically the same, so that there are fewer dimensions, which in turn means that we may trust those that are in fact significant.

From 2011 to 2015 Beatty et al. found that the programme increased participation in community organizations and awareness of local governance. And there were immediate impacts of roads on transportation costs, and schools on enrollment and teacher/student ratios. Also here multiple hypotheses were tested within a smaller number of domains. The effect size varied within each domain, and the mentioned findings are those with the largest impacts (as measured in standard deviation units). The general conclusion is that the programme had some, we would say modest, effects on community participation and development outcomes.

4.6 Sierra Leone

Sierra Leone implemented a CDD programme with support from the World Bank, with the peak project period being 2005-2009, and with a less intensive phase that seems to have faded out in 2012. The first phase is reasonably well documented. The programme was randomly assigned to half of 236 villages, and followed for 11 years. The research team has published findings in good journals along a number of dimensions. We will here focus on an article that go more in detail on alternative modalities for community involvement. But let us first report on the broader long-run impacts from the first paper referred above. A core element of the programme was to set up a Village Development Committee. Many control villages also had these, but many more programme villages had one even a decade after the programme started. The programme also supported local public goods and development of local markets, with a long list of outcomes being measured. Also when they control for the fact that some will by coincidence be significant, they report “large persistent gains in local public goods and market activity”.

We now turn to the issue of how the community can be involved. Towards the end of the programme period, the local governments implemented a competition between local communities for smaller infrastructure grants. Normally the local chief would lead the work with the application. This opened up for an additional low-cost treatment. In all villages the community was asked to identify five high skilled people who may contribute, and these, as well as the chief, conducted a written test. Then a lottery was done between the high scorer and the village chief, with the outcome only nudging the village to select the high scorer to lead the application process. In about half of the villages the skilled people also got some training as an additional treatment. Ultimately the chief could anyhow take over, despite the random nudge to select a technocrat to lead the process. The way we read the paper (it is not really clear on this), those who get the random nudge that the chief should lead the application process are considered as the status quo treatment, even though they also had to identify high skilled people. In total there are thus six different combinations of treatment, CDD or not times status quo, technocratic solution and technocratic plus training.

It is reported that the nudge towards selecting a more skilled person to lead the application process in fact leads to the community selecting a skilled person, a better application and a higher likelihood of winning the infrastructure grant. When we look into the detailed findings there are some interesting interaction effects (Figure 3 in the article), which are our interpretation of the findings: The marginal effect of selecting skilled people matters less in villages that have been in the CDD programme, and note they had been in the programme for many years. The extra intervention of training these skilled people, however, matters most in the CDD villages. Thus selecting skilled people seems to be a short cut to success, independently of the CDD programme. But, on the other hand, training of these people is most effective in CDD villages. Thus this long after the CDD started the programme did not contribute much in itself when the village had to solve a new task, while the use of skilled local people helped independently of the CDD programme. Finally, training of those skilled people was more effective in CDD villages, suggesting some deeper long-term benefits of the programme.

4.7 Sudan

In Sudan the World Bank supported the Community Development Fund programme that was implemented during 2008 to 2011 in 616 villages, in which 915 local infrastructure projects were conducted. Research was conducted in some of the villages, in North Kordofan and Kassala states, with 16 villages randomly getting the programme and eight villages being assigned as control villages. It is not clear to us, from the published paper, how those 16 villages were selected among the 616 covered by the programme. In a survey it was found that people in programme villages to a larger extent reported that community governance was participatory, and they themselves participated more in local governance. While there were no significant difference in social networks and norms between programme and control villages.

4.8 Liberia

DFID supported a community driven reconstruction (CDR) project in Liberia. The project was implemented by the International Rescue Committee (IRC), and the project “attempted to build democratic, community-level institutions for making and implementing decisions about local public goods”. The programme ran from the fall 2006 till early 2008 and covered 42 villages in Northern Liberia. A core component was to create community development committees through local elections. The committee managed a community-wide process to select and implement a two-step development project, with total value of about USD 20 000. The 42 villages were selected at random together with another group of 41 control villages. The research team then measured social cohesion in both programme and control villages after the programme ended using a standard World Bank survey module on social capital.

In addition they ran a so-called public goods experiment where people received some money that they could contribute to a new community project, knowing that the amount would be supplemented with more funds by the research team. In both programme and control villages people contributed most of the money received, respectively 75% and 81%, which, we believe, is at the upper end of contributions to public goods games. The treatment effect though is, in our mind, not very large (the difference between 81% and 75%). The research team is not in agreement, they conclude that the findings are striking, suggesting “that changes in community cohesion can take place over a short period of time”.

In a second paper the team focused on a second cross-treatment, with only women participating in the public goods game. In those villages there were no impact of the CDR programme, women contributed 84% irrespectively of the CDR programme. In villages where men participated in the game, there is thus a larger effect of the CDR programme, with 82% in CDR villages, and 67% in the other villages. Thus women contribute to public goods anyhow, while the community development programme brought men up to the same level of community cooperation as women. While we considered the average effect above as small, this effect on men of the programme is large.

4.9 Gambia

The World Bank supported a CDD programme in Gambia that from 2008 to 2010 funded development projects in about a third of the villages of the country. Among 900 eligible villages 500 were selected at random to receive support. The programme involved the local community in different stages, from identification of potential sub-projects to be funded, all the way to maintenance. Village development committees identified priorities and development plans, and managed the funds and implementation of the sub-projects. The funds were relatively large (about USD 10 000 per village, with on average 50 households per village). The most common sub-projects were related to agriculture, which potentially benefitted larger land owners, according to the published article: “farm implements and inputs (tractors, other machinery, seeds and fertilizer), milling machines, water pumps, seed stores and cereal banking, and draft animals”.

The research is based on a survey that was conducted in 2014 in 28 programme and 28 control villages. All households in each village were interviewed to be able to identify all transactions between households within each village. It was found that in programme villages the number of transactions were lower (one in six transactions) than in the control villages. This includes fewer economic transactions, fewer social links, and less participation in community-based organizations and village meetings. The researchers report that the reduction appears to be largest for marginalized groups. They find only modest effects on economic welfare indicators, although again they report heterogenous effects that suggest elite capture of the development projects. They suggest that the development projects led to a more formalized local economy with external links, and thus less need for interactions within the village.

4.10 Democratic Republic of Congo

In the Democratic Republic of Congo (DRC) a large CDR programme was implemented by two NGOs (International Rescue Committee and CARE) in villages in the eastern parts of the country, starting in 2007. The local population was trained in democratic practices and accountability, and thereafter selected a development project for the village and elected a committee to manage the project funds received from the programme. Two years after the programme started researchers implemented a new intervention, in form of an unconditional cash transfer of USD 1000 to 457 villages, with half of the villages being in the CDR programme. A later study by the same team that focused on longer term impacts say, however, that 560 clusters with 816 villages were included in the first round. The short-run impacts were measured two years after the CDR project started. The findings are negative, there is no difference between CDR and other programmes in the use of the funds, including no difference in elite capture. And there is no difference in democratic practices. Note that the intervention is not the programme in itself, but a new intervention two years into the programme to see whether the programme led to changes in democratic practices.

The second study reported on longer-term impacts eight years after the project started. We cannot see that a new intervention was implemented at that time, and as in the first round the difference between the two random samples is participation in CDR, or not. Also in this study there are negative findings when it comes to outcomes beyond the immediate impacts of use of the funds. The infrastructure (primary schools and hospitals) in CDR villages have better quality, but they find no difference between CDR villages and the controls in other outcomes, including health, education, economic welfare, women’s empowerment, governance, and social cohesion.

Another CDD programme was implemented by the governmental Social Fund with support from the World Bank. It was part of the Eastern DRC Recovery Project (STEP in French), and ran from 2016 to 2020. The programme was randomly assigned to about two thirds of 400 communities with eligible project proposals, with 127 communities in the control group with no intervention. The programme is large, with USD 100 000 to an infrastructure project in each community. Local NGOs provided training to existing local development committees in participatory management of the project. These committees were responsible for project management and maintenance. A random half of the communities, represented by formal and informal leaders, also received training from the same NGOs in conflict mitigation. Interviews were conducted with 10 random households in each community before and after the programme to measure impacts.

The researchers found no impact on access to infrastructure. They worry that the questionnaire used was not able to pick up impacts from the wide variety of infrastructure projects supported, which ranged from rehabilitation of schools to building of markets. They find a reduction in community organization, while they find an increase in violence between different social groups. There is not much discussion of this negative impact in the paper. But with such a large allocation of funds to a community, one can easily imagine that social cleavage increases. In a World Bank appraisal of additional funding of STEP we cannot find any reference to this evaluation, instead they refer to the slightly more positive evaluation of the previous programme implemented in Eastern Congo that we discussed above.

4.11 Ethiopia

One can also consider community forest user groups as CDD projects, and an RCT was implemented in the Oromia region of Ethiopia to compare internal and external monitoring. Three different modalities were randomly assigned to different villages: top-down monitoring by government officials with possible punishment for under-performing leaders, top-down monitoring by government officials with rewards for the best leaders (instead of punishments), and internal monitoring by group members who can punish under-performing leaders. All 132 user groups covering 3600 members in one area (Adaba and Dodola districts) were involved, and randomly assigned to the three treatments plus a control.

For the full sample it is found that external monitoring with punishment works best. If the sample is split, though, it is found that internal monitoring improves economic outcomes for villages where forest resources are economically important. In villages where the forests is not so important, external monitoring works better. These findings make sense, people will only invest in monitoring if they have something to gain from it.

4.12 Summary of the empirical evidence

In this report we have focused on evidence from randomized interventions that in most cases have been published in good international journals. For Asian countries in conflict, which may be considered similar to Afghanistan, we have included other studies when no randomized intervention could be found. Randomization ensures that the control group is similar to the programme group on all other aspects than the CDD programme itself. As it turned out, most of the programmes have been supported by the World Bank. The standard modality is to support small infrastructure projects, where the local community selects, builds, and maintains the infrastructure. The programmes intend both to improve economic and social welfare via the investment in infrastructure, and support community organization, while avoiding elite capture of the benefits of the programme. As a result the impact evaluations focus on these three outcomes. Some of the studies also compared different CDD modalities, but in most of them the programme was simply compared to the alternative of having no CDD programme.

For Afghanistan it was found that direct decisions by community members (via a referendum) could reduce elite capture, as compared to a model where the community had to discuss what project to prioritize, which more easily could be controlled by local leaders. For both modalities the CDD programme had good impacts both on access to infrastructure and uptake of democratic practices. The finding that infrastructure projects improve the access to infrastructure is not very surprising, and seems to be a solid finding across many CDD programmes. While the impacts on democratic processes and community organization is found to be more mixed, as is also summarized in the reviews mentioned above.

There are important heterogenous findings, with some CDD programmes having impacts only for sub-groups. One study from Liberia indicates that men can benefit more than women from a CDD programme. While women had cooperative behavior independently of the CDD programme researchers found that men became more cooperative as a result of the programme. In the Philippines the CDD programme was found to improve community organization and communication between leaders and the population, but when researchers split the sample, they found that this happened only for the non-poor, indicating some degree of elite capture.

The findings on elite capture is, however, mixed. While there was some evidence of elite capture in Afghanistan, it is not found in the study from Indonesia, where it is concluded that the benefits of involving local leaders may be more important than the risk of elite capture. In Gambia it was found that the CDD infrastructure projects led to a more formalized economy, and thus less need for social and economic interactions within the village, with fewer interaction taking place primarily for marginalized groups that in turn would benefit less from the more formalized economy. The reduction was, however, modest. A study from DRC suggests that the size of the projects matter, the larger is the gain from the CDD project, the larger is the risk of increased social conflict, assumedly due to more conflict related to competition for the gains from external funds.

Involvement of local people in the implementation of projects can be beneficial in different ways. A study from Sierra-Leone found good effects of asking the local community to select particularly skilled people to lead the CDD project, and then train these people. A study from Pakistan found that local involvement in non-technical issues was beneficial, while the opposite was the case for technical issues. Another study from Ethiopia also indicates that involvement of local people in monitoring seems to work, at least for the case of forest user groups, but only in villages where forest resources are important for local livelihoods. So in this case the economic importance of the project had a beneficial effect, apparently in contrast to the mentioned study from the DRC.

5. Conclusions

Community driven development is, surprisingly, used as a term for externally initiated and funded projects, where the donors want some local involvement. There is some, although mixed, evidence that involvement of local people means involvement of local leaders, and thus the risk of elite capture. Donors can to some extent avoid this by having direct contact with the local population, e.g. through a referendum, or by asking the local population to identify local experts that can supplement, or replace, the local leaders in selecting and implementing the programmes. Further training of these experts is likely to improve the outcome. Still, involvement of local leaders seems unavoidable, and as a result some level of elite capture can be expected. One should also be aware that benefits for the upper rungs of society can come in many different forms, from involvement in local decision making, via direct control of programme components, such as agricultural investments or investments in public services such as schools and health services, to disproportional benefits from improved market access due to investments in roads, transportation or other forms of market facilitation.

Magnus Hatlebakk

Senior Researcher; Coordinator: Poverty Dynamics