UNIVERSITEIT VAN AMSTERDAM
Community Driven
Development, Social
Capital & Inequality
Assessing the Impact of the Indonesian Urban
Poverty Project on Social Capital
Supervisor: prof. dr. M.P. Pradhan Rik Vegter (5872421)
8/14/2016
Using a household panel dataset on 158 treatment- and 97 control villages, the effect of an Indonesian community driven development project on social capital is estimated. Additionally, the role of inequality in the successfulness of this project is investigated. Difference in differences regressions showed no impact of the project on social capital. However, a differential treatment effect for inequality was observed. In villages with low inequality the project increased participation in governmental meetings and activities while in villages with high inequality participation decreased.
Contents
1. Introduction ... 2
2. Community driven development and social capital ... 3
Social capital ... 4
How does social capital affect welfare? ... 7
How does community driven development affect social capital? ... 8
Inequality, elite capture and CDD ... 9
UPP Indonesia ... 9
3. Data & methods ...10
Difference-in-differences ...12
Baseline social capital ...13
Baseline summary statistics ...13
DID estimator ...16
Inequality interaction effect...16
4. Results ...16
Social capital outcomes ...16
5. Conclusion & Discussion ...19
Literature ...19
Appendix ...22
Verklaring eigen werk
Hierbij verklaar ik, Rik Vegter, dat ik deze scriptie zelf geschreven heb en dat ik de volledige verantwoordelijkheid op me neem voor de inhoud ervan. Ik bevestig dat de tekst en het werk dat in deze scriptie gepresenteerd wordt origineel is en dat ik geen gebruik heb gemaakt van andere bronnen dan die welke in de tekst en in de referenties worden genoemd.
De Faculteit Economie en Bedrijfskunde is alleen verantwoordelijk voor de begeleiding tot het inleveren van de scriptie, niet voor de inhoud.
1. Introduction
Community driven development (CDD) is a form of development aid that is increasingly used worldwide. Besides claims of CDD being better at targeting the poor and increasing the livelihoods of people than traditional forms of development aid, one of the promises is that it increases social capital and strengthens governance. While extensive reviews are written on community driven development (Mansuri & Rao, 2004; Dasgupta & Beard, 2007; Tanaka, Singh & Songco, 2006), literature that examines the relationship between community driven development and social capital is not that well established. This thesis tries to add to that by examining the effect of the Indonesian Urban Poverty Project (UPP) on social capital. The work of Darmawan & Klasen (2013) -who found that elite capture was present in UPP Indonesia-, inspired me to also investigate the relationship between inequality and its effect on the successfulness of this project. Consequently, the research question of this thesis is: What is the effect of the Urban Poverty Project Indonesia on social capital and how is this effect influenced by inequality? The answers to these questions may provide some guidance in finding out under which conditions CDD projects thrive. Under UPP Indonesia, groups of individuals could submit proposals for small scale loans. These loans were aimed at individuals with small businesses who would benefit from an increase in liquidity after the economic downturn of 1997 in East-Asia. In each of these villages an elected group of people would decide which proposals were accepted and which were not. Besides providing resources to the poor, the project was designed in a way that it would increase community participation. In order to determine the project’s effect on social capital, indicators that reflect social capital are used for the analysis. These are measures of participation in different governmental and community organizations as well as the participation in community activities. Household-survey panel data is used in the analysis, which were collected before the implementation of the project and after the project had been finished. A total of 255 villages were used for the analysis, consisting of 12780 individuals. The difference in differences estimates show that UPP did not increase social capital in treatment villages. However, when inequality is accounted for, significant results for some of the social capital indicators are found. This indicates that inequality may play a role in community driven development projects and their effect on social capital.
This thesis is organized as follows. Section 2 discusses existing literature on CDD, social capital and inequality as well as the UPP. Section 3 describes the data & methods that were used for this research. Section 4 discusses the results, while section 5 concludes and brings forward some points for discussion and further research.
2. Community Driven Development and Social Capital
Community driven development (CDD) is one of the fastest growing mechanisms for channelling development assistance. According to Mansuri & Rao’s (2004) conservative estimates, the World Bank’s lending for CDD projects has increased from US$ 325 million in 1996 to $2 billion in 2003. Between 2000 and 2005 the World Bank supported 190 lending projects for a total of $9.3 billion (Tanaka et al., 2006). In the past decade the World Bank has approved over 500 CDD projects with a total value of over $28 billion (World Bank, 2016). The increasing interest in community driven development can be explained by the criticism on traditional (centralized) distribution of development aid, which is sometimes claimed to be ineffective because of its top-down approach. The World Bank defines CDD as an approach that gives control over planning decisions and investment resources for local development projects to community groups. The motive of this approach is that communities that receive aid are the best judges of how their lives and livelihoods can be improved. With some support in the form of resources and access to
information they can organize themselves to provide for their immediate needs. Tanaka et al. (2006) present five defining elements of CDD projects.
(1) They are community focussed because the target beneficiary is some form of a community-based organization (CBO).
(2) They involve participatory planning and design.
(3) The community controls the resources, which ensures that there should be at least some form of resource transfer to the community/CBO.
(4) The community is involved in implementation through direct supply of inputs, labor, or funds, or indirectly through management and supervision of contractors or operation and maintenance functions.
(5) CDD projects employ community-based participatory monitoring and evaluation to ensure downward accountability to the community.
Donger et al. (2003) bring forward different reasons why CDD could be an effective way of delivering development aid. Among other things it can: enhance sustainability, improve efficiency and effectiveness, make development more inclusive, empower poor people, build social capital and strengthen governance. This is achieved by reducing information problems that face both the social planner and beneficiaries and by allowing communities to identify projects as well as eligible recipients. Furthermore, it strengthens the civic capacities of communities by enabling them to acquire skills and organizational abilities that
beneficiaries in the decision making and planning process funds will be better distributed and communities are empowered, or as Donger et al. (2003) put it:
“It is expected that this (CDD) will result in the allocation of development funds in a manner that is more responsive to the needs of the poor, better targeting of poverty programs, more responsive government and better delivery of public goods and services, better maintained community assets, and a more informed and involved citizenry that is capable of undertaking self-initiated development activity.”
While CDD can achieve meaningful results, the design of it also has some limitations. Tanaka (2006) brings forward three of these limitations. Firstly, heterogeneity of communities raises concerns about elite capture. Poor communities are often unequal and stratified, therefore prone to elite capture. Furthermore, CDD bypasses local existing institutions and creates structures that are not sustainable after the close of the project. It may crowd out other initiatives. Finally, the cost of participation for poorest and most vulnerable is another concern. For poor individuals it may be very costly to participate in community driven development projects because of the valuable time and resources that need to be invested.
Social Capital
Social capital is a concept that has received growing attention in development literature. Portes & Landolt (2000) ascertain that the current increasing interest in the concept of social capital comes from the limitations of a purely economic approach towards the prevailing development goals of sustained growth, equity and democracy. They observe that orthodox economic theory is unable to explain the contradictory outcomes that economic policies implemented by the IMF and US treasury have had on countries at varying levels of development. Additionally, they remark that the removal of government protections has paved the way for the forces of the free market, which has created an ‘every man for himself’
environment. The concept of social capital tries to stress the importance of social interaction, also from an economic point of view. Grootaert states (1999) “The recognition that social capital is an input in a household’s or a nation’s production function suggests that the acquisition of human capital and the establishment of a physical infrastructure needs to be complemented by institutional development in order to reap the full benefits of development investments.”
Why is social capital an important concept in the context of community driven development? To answer that question it is first important to establish a proper definition of social capital. Various sociologists as well as economists have contributed to the concept of social capital. The first being the French sociologist Pierre Bourdieu, who in 1985 defined social capital as “the aggregate of the actual or potential resources which are linked to possession of a durable network of more or less institutionalized relationships of mutual acquaintance or recognition.” The current prevailing definition however can be attributed to Portes (1998). He states that: “social capital stands for the ability of actors to secure benefits by virtue of membership in social networks or other social structures.” Both definitions emphasize that engaging in relationships with others and being part of social networks can result in certain benefits. What could these
benefits be then? Grootaert (2004) brings forward two important examples. Firstly, participation in social networks increases the availability and access to information and lowers its costs. For example, if this information is connected to crop prices, credit sources or new technologies it may positively affect returns from agriculture and trading. Furthermore, participation in local networks and increasing mutual trust enables communities to reach collective decisions and implement collective projects. Community driven development tries to increase social capital by giving local community groups control over planning decisions and investment resources for development projects. These social networks that are created strengthen community ties and may form the basis for future partnerships. While most literature focuses on the positive effects of social capital it must also be noted that there can also be negative effects that can result from the increase in social interactions and networks. Portes & Landol (2000) make an important remark on what is probably the most important issue with social capital. Namely, that the same strong ties that enable group members to obtain privileged access to resources, bar others from securing the same assets. Exclusion of outsiders is not the only negative consequence that social capital can have. Excess claims on group members and restrictions on individual freedoms are among other adverse effects that social capital can have (Portes & Sensenbrenner 1993). If we want to determine how a certain project affected social capital then we have to find a way to measure social capital. In their book ‘Understanding and Measuring Social Capital’, Grootaert & Van Bastelaer (2004) provide quantitative as well as qualitative techniques to measure social capital. An important remark they make is that social capital can differ in form and scope.
The distinction is made between structural and cognitive social capital. Structural social capital refers to objective and externally observable social structures; these can be all forms of networks and associations. Cognitive social capital refers to the subjective and intangible elements such as norms of behavior, generally accepted attitudes, shared values and trust. These two forms of social capital are mutually reinforcing, however each form can also exist on its own. Another way of isolating the different elements of social capital is based on its scope. On the micro level social capital can be observed in the form of horizontal networks of individuals and households and the associated norms and values that underlie these networks. The meso level is situated between individuals and society as a whole; it captures horizontal and vertical relations among groups. Finally, social capital can be observed at the macro level, in the form of the institutional and political environment that serves as a backdrop for all economic and social activity, and the quality of governance arrangements. Grootaert & Van Bastelaer (2004) presented the forms and scope of social capital on a continuum as can be seen in figure 1.
Figuur 1, the forms and scope of social capital (Grootaert & van Bastelaer, 2004)
The continuum in figure 1 can serve as a starting point for measuring social capital in different settings. With the different forms and scope of social capital in mind Grootaert & Van Bastelaer developed an assessment tool for measuring social capital (SOCAT) that must fulfil four criteria.
- It must recognize cultural variation while providing a unifying conceptual framework - It must take into account structural and cognitive dimensions of social capital
- It must build primarily on activities local people consider appropriate for collective execution - It should be constructed using qualitative and quantitative methods.
The SOCAT will be used as guidance for the quantitative analysis. Like the continuum presented in figure 1 the SOCAT makes the distinction between structural social capital and cognitive social capital. When measuring social capital both need to be taken into account.
Structural social capital is measured along three dimensions; density of membership, diversity of membership and participation in decision making.
- The density of membership at community level is measured as the number of existing organizations in a community. At the household level, it is measured as the average number of membership of each household
- The diversity of membership measures heterogeneity within organizations. The diversity can be rated using the following criteria: kinship, religion, gender, age, political affliction, occupation and education.
- Participation in decision-making to what extent organizations follow a democratic pattern of decision-making. It also measures the overall effectiveness of the organization leader.
Cognitive social capital is also measured along three dimensions; solidarity, trust and cooperation and conflict & conflict resolution
- Solidarity is measured as the degree of solidarity and mutual support within a community. It is measured by asking people whether the village would get together to deal with a crisis that affects the whole community. Or the degree of solidarity that would be showed if an unfortunate event happened to an individual of the village.
- Trust and cooperation is difficult to measure because individuals may have different perceptions of these concepts. Therefore generalized trust (the extent to which one trusts people overall) is measured. A statement that is used to measure this is for example “most people in this village or neighbourhood are basically honest and can be trusted “
- Conflict & conflict resolution is measured by asking whether the village or neighbourhood is peaceful or in conflict, whether people make any contributions to common goals, and whether relations are harmonious or disagreeable.
How does social capital affect welfare?
The Local Level Institutions Study (LLIS) collected large-scale household survey data in Bolivia, Indonesia and Burkina Faso to investigate the role of local institutions in providing service delivery and in affecting welfare and poverty outcomes. It contains detailed information about social relationships and structures as well as traditional economic variables. Grootaert, Oh and Swany (2002) examined the impact of social capital on two outcomes; (1) the impact of social capital on per capita consumption expenditure of the household and (2) the role of social capital in facilitating access to credit. Using instrumental variable estimation to control for possible endogeneity of the social capital variable, they found that a 5% increase in social capital endowment of a household increases household expenditure per capita by 2.7%. An effect which is 5-11 times larger than the effect found for human capital (schooling). Also, the positive role of social capital in gaining access to credit was confirmed. Using a specially designed large-scale survey, Narayan & Pritchett (1999) measure the degree of associational activity as a proxy for social capital and trust among households in rural Tanzania. They find that a one standard-deviation increase in the village social capital index, which corresponds to half of the population joining one additional association, would lead to an increase of at least 20% in expenditures per person in each household in the village. Thus it seems that investing in social capital can be an attractive way of increasing household welfare. Another interesting contribution to the social capital literature is made by Feigenberg, Field and Pande (2013). They conduct an experiment on micro financing using 100 women that are first-time borrowers. These women were divided in groups of 10 that would have meetings on varying intervals. By varying these intervals, the impact of social interaction on economic cooperation could be measured. Their findings suggest that a development program that encourages repeat interactions can increase long-run social ties and enhance
social capital among members of a localized community in a very short amount of time. With only the meetings as an outside stimulus, a feeling of solidarity with close neighbours was created that enhanced economic interaction and cooperation. This provided a buffer against economic shocks that could lead to default.
While the results from the aforementioned studies offer promising results for future projects that aim at improving social capital, they do not comment on the extent to which these results can be extrapolated. It is plausible to assume that the relationship between social capital and economic welfare is not a linear one. There can be thought of mechanisms that cause social capital to have decreasing marginal returns to welfare.
How does community driven development affect social capital?
There is no extensive amount of research that has focussed on this relationship. Nevertheless, two recent studies deserve to be mentioned. Chase & Christensen (2004) looked at the Social Investment Fund (SIF), a community driven development project initiated in 1998 by the Thai government. The SIF was
established to provide resources for local and grassroots organizations to implement their development projects. Communities that wanted to receive grants had to follow sub-project procedures for proposals, management and monitoring. Chase & Christensen were interested in how this project had affected social capital. One important finding is that they observe a selection effect. Villages with more social capital were more likely to participate in the project. However, when the selection effect is isolated they still find that the project enhanced information sharing, leadership and empowerment. It helped build local leadership by encouraging leaders to get things done outside of the formal government system. Another interesting finding is that in villages where more people work in agriculture and do not own their farms, higher levels of social capital are found. Furthermore, more education was associated with less social capital. These findings might show that working together on farmland ‘creates’ social capital and that these norms and networks are maintained better by the less educated. Labonne & Chase (2011) used a panel data set of 2100 households to explore the social capital impacts of a community driven development project in the Philippines. A total of 66 treatment and 69 control communities were compared in this project where communities could compete for block grants for infrastructure investment. Labonne & Chase used the following definition of social capital: ‘the ease with which community members act collectively’. This definition coincides with the view of social capital being a community-level aggregate instead of a household-specific characteristic. They find that the CDD project increased community participation in village assemblies and that it also increased the frequency with which local officials meet with residents. However, the project decreased the time spent on collective action activities such as construction and maintenance of community infrastructure (Labonne & Chase, 2011).
Inequality, elite capture and CDD
The success of a community driven development project depends largely on the input from the
community itself. In a CDD, different households/individuals have to work and make decisions together in order to reach common goals. Who is included in the project, the way these individuals work together and the way decisions are reached upon, affect the way resources are divided and the eventual outcome of the project. Inequality may play an important role in this process, which can affect project outcome in various ways. One way it can affect the successfulness of CDD projects is through elite capture. Different studies find that participants of poverty government programs tend to be wealthier, more educated, and higher social network and male (Darmawan & Klasen, 2013, Platteau, 2004). These findings suggest that in some cases, participatory programs actually exclude the individuals they are targeting. The result is that the allocation process mainly reflects the preferences of the elite group and the marginalized poor benefit less. When communities are more unequal, the probability that decision making in is the hands of a powerful elite group is larger. Darmawan & Klasen (2013) indeed find that in the Urban Poverty Project inequality increased elite capture. Another interesting finding is that when the individuals that decide on projects closely share characteristics with the poor, altruistic behaviour exists and the decisions that are made favour the poor. Inequality may affect project outcome in another way than through elite capture. Because social ties tend to be weaker in more unequal communities, inequality may hamper individuals in working together and collectively making decisions.
UPP Indonesia
The Urban Poverty Project (UPP) was a World Bank project designed in 1997 to alleviate some of the quickly rising poverty in Indonesia after the emerging 1997 Asian Economic Crisis. When a year later the (corrupt) Suharto administration collapsed, the government of Indonesia sought ways to rapidly and transparently distribute financial resources to the urban poor. The World Bank assisted by creating the UPP. The objective of the program was stated in the project implementation report:
“Through a bottom-up and transparent approach, the project seeks to improve basic
infrastructure in poor urban neighbourhoods and to promote sustainable income generation for its poor urban residents who are mostly long-term poor, have incomes eroded by high inflation, or lost sources of income in the economic downturn. Also, the project seeks to strengthen the capability of local agencies to assist poor communities.” (World Bank, 2005)
The project targeted around 2800 of the poorest kelurahans (a collection of small villages) in the part of Northern Java that was most affected by the economic crisis. In total, these kelurahans accounted about 7.8 million households or 31.2 million people. The project consisted of four programs, with a total of 129.7 million dollars divided over these programs: (1) Grants for sub-projects, (2) Technical assistance and services for management and implementation, (3) Technical assistance for monitoring and evaluation, and (4) Increased government administration.
The first phase of the project was implemented successfully, which led to a request from the Indonesian government to expand the UPP to other provinces. In June 2002, the second phase of UPP (UPP2) was approved and eventually also a third phase (UPP3) was implemented in 2005.
3. Data & Methods
The goal of this research is to determine if the urban poverty project (UPP) has affected social capital outcomes in the beneficiary villages. It attempts to answer the question of how much the social capital outcome indicators have changed purely as a result of the project. Furthermore, it will attempt to ascertain the influence of inequality on the project effect. To isolate the effect of the project, the changes in the social capital outcomes that would have occurred in absence of the project need to be estimated. Since those changes are unobservable for the treatment group, these changes will need to be estimated by using a control group. Ideally the villages in this control group are very similar in characteristics compared to the treatment group. So that if we compare the changes in social capital before and after treatment, the true impact of the project can be estimated.
This research follows the methods of Pradhan, Rao and Rosemberg (2010) in finding control villages to answer the research question. The UPP is designed in such a way that the ‘poverty score’ of a sub-district (kecamatan) determines if it is eligible for the project. In every district the richest 20% of sub-districts were excluded from the project. From each the sub-districts that were eligible about half of the villages (kelurahans) were randomly selected for the project. The number of sub-districts that could be included was set by project management. Therefore, the threshold varied by district. The poverty score threshold was used in two ways to select sub-districts for the control group. Firstly, they matched within-districts; sub-districts that were just below the poverty score threshold (just too rich) were used as control group and matched with sub-districts that were just above the threshold(just too poor). As a result, the poverty scores of the control and treatment groups are close, which is beneficial for the impact evaluation. This within-district matching however does not sample for the poorest sub-districts. Therefore they also matched across districts (using the fact that poverty thresholds were different in every district); randomly sampled treatment sub-districts were matched with sub-districts with a similar poverty score who were below the threshold in that district. In this way, also poor sub-districts were included in the evaluation. The two matching techniques can be seen in figure 3.
Figure 2, (Pradhan, Rao and Rosemberg, 2010)
Using this method, a total of 255 villages have been selected for the analysis. These villages received surveys both before (baseline) and after (final) the project was implemented. The baseline survey was conducted in the beginning of 2004, a year before the first village received grants, whereas the final survey was done in the end of 2007, about two years after the first disbursement. For both the baseline as well as the final round, data on three different levels were collected. Data on the village level were collected through interviews with the head of the village (lurah). This consisted of information on population, land & housing, village finance, and health & education. Data on household level were collected through interviews with the head of the household, which was done with 32 households per village. These interviews included information on household members’ basic characteristics, housing, economic status, household consumption, loans & savings and the use of health & social safety net facilities. In each of these households 1 male and 1 female were interviewed to obtain data on an individual level. These data included questions on ethnicity, social network and the participation in governmental & community organizations and community activities. For the analysis, data on the village level was used. Since this research is interested in the impact of the project on village level, the household level data and individual level data were averaged on village level.
Difference-in-differences
Difference-in-differences (DID) is the technique that is used to estimate the impact of the project on our social capital outcomes of interest. The DID estimator compares the changes in social capital outcomes over time between treatment and control villages. The DID estimator is the average change in our
outcome variable for the treatment group minus the average change in our outcome variable in the control group. This gives us:
= , , , , )
Where Y is the social capital variable of interest and is the DID estimator (Stock & Watson, 2003). If the treatment is randomly assigned, then is an unbiased and consistent estimator of the causal effect. In this project the treatment was not completely random assigned, but the method of determining treatment and control villages as described in the previous section assures an as if randomly assigned approach. Another important aspect of the DID estimator is the parallel trend assumption. This assumes that both control and treatment villages would have followed the same trend (∆ ∆ ) in absence of the program. This parallel assumption is illustrated in figure 4.
Baseline social capital
To determine the impact of the project on social capital, several outcome variables that reflect social capital are chosen for the analysis. These outcome variables are based on Portes (1998) definition of social capital; the ability of actors to secure benefits by virtue of membership in social networks or other social structures. These outcome variables include participation in governmental and community organizations, participation in community activities as well as the density of membership and the amount of time, money and material invested in these organizations/activities. Governmental organizations are different kinds of organizations that are related to governance at the village level, such as the village council. Community organizations include different types of organizations that provide services to its members. These services can range from services that provide basic needs (education, health and security) to financial services and services related to religion. Community activities encompass various activities that are related to social interaction. Examples of community activities are sports and music, death rituals and voluntary collective work. Table 1 shows the average baseline values of these variables for both control and treatment villages (variables marked with an asterix refer to 3 months prior to the survey). It can be seen that there are some differences in the participation, time spent and money & material contribution to these organizations and activities. The average individual participated 0.88 times in governmental activities in the last 3 months and contributed a total of 7,020 Indonesian Rupiah (IDR) to these organizations. Time spent on these organizations amounted to an average of 4.93 hours. Number of times participated in community organizations was 2.13, with total contributions at 5,022 and an average of 4.07 hours of time spent. Participation in community activities is remarkably higher with 7.81 times in the last three months, which is also reflected in the amount of money and material contributed (89,550) and the time spent (17.66 hours) on these activities.
Baseline summary statistics
It is not possible to see if the control and treatment group have followed the same trend in outcome variables before the program was implemented because there is no data available to investigate this. What can be done however is comparing the baseline characteristics of control and treatment villages. If there are no significant differences between these characteristics, it is fair to assume that in absence of the project the treatment and control villages would have followed the same trend. And therefore, the estimation of is a reflection of the true project effect. Table 1 describes baseline summary statistics of various variables for both treatment and control villages. As can be seen, the means of most variables are within close proximity of each other, which suggests that the treatment and control villages do not differ much in terms of baseline characteristics. This is confirmed when t-tests on all of the variables that are reported in table 1 are performed. Only one of the variables (Money contributed to governmental organizations) did not pass the two sample t-test for equal means. The t value of 2.83 indicates that control villages contributed significantly more money to governmental organizations in the baseline compared to treatment villages. Since this variable is one of our indicators of social capital it is likely to
affect the estimation of the project effect on this specific variable. Moreover, it may also affect the estimates of the other social capital indicators. The indicators of social capital that are used in the regressions are displayed in bold in Table 1. The amount of money contributed to governmental
organizations in control villages at baseline relatively high (at least higher than in treatment villages), which is expected to have a negative effect on the trend for control villages (already high levels of money
contribution may leave less room for growth). If this is indeed the case, the common trend assumption is violated and the effect of the project on money contributions to governmental organizations is
overestimated. Thus, the high average amount of money contributed to governmental organizations in control villages leads to an upward bias of the project effect on social capital. While the difference in money contributed to governmental organizations is significant, it is still relatively small. If we compare the average amount of money contributed to governmental organizations in control villages (7,239) to the average amount of money contributed to community activities in control villages (82,459) we can see that the former is less than 10% of the latter. Therefore, the results of the comparison of baseline summary statistics gives no reason to abandon the common trend assumption of the DID regression.
Table 1 – Baseline summary statistics
Baseline summary statistics Control Treatment
Variables N Mean N Mean Difference t-value Average number of households 90 1,224 153 1,242 18.00 0.14 Average total population 96 5,934 156 5,644 -290.00 -0.34 Average village size 97 353.80 158 317.70 -36.10 -0.35 Average size of farm area 97 144.40 158 121.6 -22.80 -0.56 Average size of non-farm area 97 164.6 158 118.3 -46.30 -0.60 Number of mosques 97 6.54 158 6.56 0.03 0.03 Village has orphanage 79 0.139 131 0.221 0.08 1.54 Number of doctor practices 97 2.13 158 1.67 -0.47 -1.22 Number of midwives 97 0.887 158 1.04 0.15 0.86 Number of primary schools 96 3.24 156 3.63 0.39 1.25 Number of junior schools 49 1.74 91 1.64 -0.10 -0.49 Number of senior schools 50 1.94 68 1.99 0.05 0.13 Gini coëfficient 97 0.33 158 0.31 -0.01 -1.32 Years of education (head of household) 97 8.90 158 8.94 0.04 0.18 Famliy size 97 4.65 158 4.61 -0.04 -0.55 Consumption per capita 97 231,432 158 220,083 -11,349 -0.94 Participation in governmental activities 97 0.900 158 0.867 -0.03 -0.51 Percentage board member governmental
organization 97 0.206 158 0.232 0.03 1.01 Attendance of governmental organization
meetings* 97 2.63 158 2.51 -0.12 -0.58 Money contribution to governmental
organizations* 97 7,239 158 4536.00 -2703.00 -2.65** Material contribution to governmental
organizations* 97 1,804 158 1242.00 -562.00 -1.26 Time spent on governmental activities (hours)
* 97 5.21 158 4.76 -0.45 -1.03 Participation in decision making process of
governmental organizations* 97 0.404 158 0.430 0.03 0.61 Percentage member of community
organization 97 0.264 158 0.305 0.04 1.23 Participation in community organization
activities* 97 2.03 158 2.19 0.16 0.44 Money contribution to community
organizations* 97 3,375 158 5,104 1,729 1.65 Material contribution to community
organizations* 97 670.5 158 517.3 -153.20 -0.68 Time spent on community organizations
(hours) * 97 3.85 158 4.21 0.36 0.54 Participation in decision making process of
community organizations* 97 0.131 158 0.152 0.02 1.04 Number of community activities joined 97 1.89 158 1.78 -0.11 -1.39 Number of times participated in community
activities* 97 8.06 158 7.65 -0.41 -0.68 Money contributed to community activities* 97 82,459 158 75,464 -6,995 -0.71 Material contributed to community activities* 97 11,676 158 11,272 -404 -0.19 Time spent on community activities (hours)* 97 17.95 158 17.49 -0.46 -0.35
DID estimator
To estimate the impact of the project (treatment) on the social capital outcomes we write the OLS estimator of as
∆ = + + + (Model 1)
Where ∆ is the change over time in our social capital variable of interest for village i, is a constant, is our treatment effect of interest, is a dummy variable which equals 1 if village i was a treatment village and 0 if village i was a control village, is a set of baseline control variables and is the error term. By focussing on the change in Y over the course of the experiment, the DID removes the influence of initial values of Y that vary between the treatment and the control groups. Nevertheless, we would like to have treatment and control groups that are very similar in characteristics at the baseline. This validates the parallel trend assumption that underlies the difference-in-differences estimation.
Inequality interaction effect
To ascertain the influence of inequality on the project impact, model 1 is extended and an interaction term is included ( ∗ ∗ ), where is the Gini coefficient for village i.
∆ = + ∗ + ∗ ∗ + + (Model 2)
Now
is the differential treatment effect in which we are interested. By estimating this we can check if inequality influences the project’s effect on social capital. This means we can see if the project works better in areas with low or high inequality. As in model 1, is the treatment effect.4. Results
This section will discuss the project impacts on village-level social capital. The change in the social capital indicators between treatment and control villages over the course of the project is compared. Results for which significant coefficients are found are presented. Results for some of the other outcome variables can be found in the appendix.
Social capital outcomes
Tables 2, 3 and 4 show the project impact on participation in governmental activities, attendance in governmental meetings and time spent on governmental activities. All of the variables are aggregated at the village level. The social capital variables as well as ‘consumption per capita’ and ‘years of education’ are in logarithms. The first model (difference-in-differences estimator with consumption, education and inequality as control variables) finds no significant impact of the project on our social capital outcomes.
However, when the model is extended and an interaction effect for treatment and inequality (Gini coefficient) is included, significant impacts are observed. Table 2 shows the effect of the project on participation in governmental activities, measured by the average amount of different governmental organisations an individual participated in, the last 3 months before the survey. The project on average increased participation in governmental activities by 78% in treatment villages. The effect of the project on attendance in governmental meetings is of the same magnitude (75%) as can be seen in table 3. Looking at table 4 we can see that the project increased time spent on governmental activities by 130% and this effect is significant at the 1% confidence interval.
Table 2 – DID regression for participation in governmental activities
Table 4 – DID regression for time spent on governmental activities
The previous results suggest that the project has helped increasing at least some aspects of social capital. In this case; participation in governmental activities, attendance in governmental meetings and time spent on governmental activities. These findings are in line with the research of Labonne & Chase (2011), who find that a community driven development program in the Philippines increased the percentage of households participating in village assemblies by about 21 percentage points. Additionally, they find a positive impact on the number of times a year the village elected leader meets with the villagers. In both this study and Labonne & Chase’s an increase in participation in local governance as a result of the community driven development project is observed. However, where Labonne & Chase find a negative effect of the Philippine project on participation in group and collective action activities, the difference-in-differences estimations in this study show no significant impact of the project on participation, time, money and material invested in community organizations and community activities. These results hold if the interaction effect (Treatment*Gini) is included. Furthermore, what the estimations in tables 2, 3 and 4 show is that inequality plays an important role in the effect of the project on our social capital outcomes; the interaction effect is significant and has the expected (negative) sign. If, for instance, the project impact on participation in governmental activities (table 4) for a village with the average Gini coefficient of our sample (0.32) is compared with the impact in the most unequal village in our sample (Gini coefficient of 0.78) substantial differences are observed. The total effect of this difference in inequality on the treatment
effect is equal to ((0.97*0.32) - (2.25*0.32)) – ((0.97*0.78) – (2.25*0.78)) = 0.5888. So, the project leads to 59% more participation in governmental activities in a village with average inequality compared to a village with very high inequality.
5. Conclusion & Discussion
Using a difference in differences estimation approach, this thesis estimated the effect of the Urban Poverty Project in Indonesia on village-level social capital. Additionally, the effect of within village inequality on the project’s outcome was researched. This was done using a household panel dataset which contains data from before the project started and two years after the project was finished. The dataset includes data on 12780 individuals in 97 control- and 158 treatment villages. Overall, the results indicate that the UPP did not increase village-level social capital in treatment villages. The estimates show no significant impact of the project on participation in governmental organizations, community organizations and community activities neither on the time and money invested in these organizations. However, when accounted for differences in income inequality, significant effects of the project on some of the social capital indicators are found. In villages with low inequality, the project increased participation in governmental organizations, attendance in governmental meetings and time spent on governmental activities. In villages with high inequality, the effect of the project on these social capital outcomes was negative. Comparison of a village with average inequality with the village with highest inequality shows a 59% difference in participation in governmental activities as a result of the project. These results indicate that inequality may be a factor that needs to be taken into account when implementing community driven development projects that are aimed at increasing social capital. The mechanism behind the differential treatment effect of inequality was beyond the scope of this thesis. However, it would be an interesting topic for further research.
Literature
Avdeenko, A., & Gilligan, M. J. (2015). International interventions to build social capital: evidence from a field experiment in Sudan. American Political Science Review, 109(03), 427-449.
Beard, V. A., & Dasgupta, A. (2006). Collective action and community-driven development in rural and urban Indonesia. Urban Studies, 43(9), 1451-1468.
Bebbington, A. (2006). The search for empowerment: social capital as idea and practice at the world bank. Kumarian Press. Chase, R. S., & Christensen, R. N. (2014). Picking Winners or Making Them? Evaluating the Social Capital Impact of Community Driven Development (CDD) in Thailand. Case Studies In Business, Industry And Government Statistics,3(2), 95-107
Darmawan, R., & Klasen, S. (2013). Elite Capture in Urban Development: Evidence from Indonesia (No. 145). Courant Research Centre: Poverty, Equity and Growth-Discussion Papers.
Dasgupta, A., & Beard, V. A. (2007). Community driven development, collective action and elite capture in Indonesia. Development and change,38(2), 229-249.
Gupta, M. D., Grandvoinnet, H., & Romani, M. (2004). State–Community Synergies in Community-Driven Development. Journal of Development Studies,40(3), 27-58.
Dongier, P., Van Domelen, J., Ostrom, E., Ryan, A., Wakeman, W., Bebbington, A., ... & Polski, M. (2003). Community driven development. World Bank Poverty Reduction Strategy Paper.
Fritzen, S. A. (2007). Can the design of community-driven development reduce the risk of elite capture? Evidence from Indonesia. World Development, 35(8), 1359-1375.
Galasso, E., & Ravallion, M. (2005). Decentralized targeting of an antipoverty program. Journal of Public economics, 89(4), 705-727.
Grootaert, C. (1999). Social capital, household welfare, and poverty in Indonesia. World bank policy research working paper, (2148).
Grootaert, C. (2003). On the relationship between empowerment, social capital and community driven development. World Bank working paper, (33074).
Grootaert, C., & Van Bastelaer, T. (Eds.). (2002). Understanding and measuring social capital: A multidisciplinary tool for practitioners (Vol. 1). World Bank Publications.
Khwaja, A. I. (2005). Measuring empowerment at the community level: An economist’s perspective. Measuring Empowerment: Cross-Disciplinary Perspectives (W ashington DC, The W orld Bank), 267-284.
Labonne, J., & Chase, R. S. (2011). Do community-driven development projects enhance social capital? Evidence from the Philippines. Journal of Development Economics, 96(2), 348-358.
Mansuri, G., & Rao, V. (2004). Community-based and-driven development: A critical review. The World Bank Research Observer, 19(1), 1-39.
Mansuri, G., & Rao, V. (2012). Can participation be induced? Some evidence from developing countries. Some Evidence from Developing Countries (July 1, 2012). World Bank Policy Research Working Paper, (6139).
Narayan-Parker, D. (Ed.). (2002). Empowerment and poverty reduction: A sourcebook. World Bank Publications.
Onyx, J., & Bullen, P. (2000). Measuring social capital in five communities.The journal of applied behavioral science, 36(1), 23-42.
Platteau, J. P. (2004). Monitoring elite capture in Community‐Driven development. Development and Change, 35(2), 223-246.
Portes, A., & Landolt, P. (2000). Social capital: promise and pitfalls of its role in development. Journal of Latin American Studies, 32(02), 529-547.
Portes, A., & Sensenbrenner, J. (1993). Embeddedness and immigration: Notes on the social determinants of economic action. American journal of sociology, 1320-1350.
Pradhan, M., Rao, V. & Rosemberg, C. (2010) The Impact of the Community level activities of the Second Urban Poverty Project (UPP). World Bank
Rao, V. (2005). Symbolic public goods and the coordination of collective action: A comparison of local development in india and indonesia. World Bank Policy Research Working Paper, (3685).
Stone, W. (2001). Measuring social capital. Australian Institute of Family Studies, Research Paper, 24. Stock, J. H., & Watson, M. W. (2003). Introduction to econometrics (Vol. 104). Boston: Addison Wesley. Tanaka, S., Singh, J., & Songco, D. (2006). A review of community-driven development and its application to the Asian development bank. Asian development bank (ADB).
World Bank (2016) Community Driven Development strategy overview (2015, September 29) Retrieved from http://www.worldbank.org/en/topic/communitydrivendevelopment/overview#2
