Abstract
In India, self-help groups aim to eradicate poverty. Groups access microcredit via banks, government, or nongovernmental organizations. A vast but inconclusive literature exists on the impacts of heterogeneity and freedom of participation on group functioning. We used survey data and curve estimation to study the effect of these variables on collective action. Survey results were triangulated with in-depth interviews. The applied mixed methods design is useful for empirical studies where the functional form of one variable on the other variable is contested and no a priori model exists for theory-driven deductive empirical testing. We found that both variables are mediated by their institutional context. Heterogeneity promotes the emergence of leadership in bank groups, but it reduces collective action in government-initiated groups.
The success of India lies in the prosperity of its villages. Agricultural productivity continues to increase, and in spite of the rapidly growing service sector, agriculture still accounts for large shares of gross domestic product and employment. Yet rural development and poverty alleviation continue to be major challenges for the country (Bardhan, 2010). Although the relative poverty ratio has decreased since 1947, the absolute number of poor people is still very high, with approximately 269 million Indians living below the poverty line in 2012. The problem is particularly severe in rural areas where 25.7% of the population is rated as poor. In comparison, only 13.7% of the population is rated poor in cities (Government of India, Planning Commission, 2013).
One of the major causes of poverty in rural India is the lack of access to productive assets and financial services for both individuals and communities (Puhazhendhi, 2013). Microcredit addresses these challenges and has become a dominant instrument to provide financial services to the rural poor in India and elsewhere (Armendáriz & Morduch, 2007). In the absence of collateral, repayment rates can be increased and information asymmetries reduced by using peer pressure, group-lending mechanisms, and local information (Besley & Coate, 1995; Bonus, 1986; Stiglitz, 1990). In the southern states of India, microcredit is most prominently provided via self-help groups (SHGs)—democratically controlled organizations of mutual economic help—with an increasing role for commercial banks in lending (Puhazhendhi, 2013). Recently, microcredit has been criticized because of high interest rates, the use of violence to enforce loan repayment, or rollover debt in household lending (Taylor, 2011). On the other hand, the positive effects of SHGs on women’s participation in nonagricultural labor, women’s savings, participation in intrahousehold decision making, and civic engagement are well documented (Desai & Joshi, 2013).
In this regard, it is important to note that Indian SHGs are institutionally diverse. Ideally, they can be categorized into three different models of service provision: (a) bank promoted, (b) government promoted, and (c) nongovernmental organization (NGO) promoted. Each of these models operates under a different legal framework. In addition, interest rates, repayment, and the amount of government subsidies received differ substantially. Relatively little information is known about how these institutional differences among the promoting organizations affect the respective collective action processes and how they are perceived by SHG members.
In this article, we used a sequential mixed methods approach to understand the effect of a number of community attributes on the level of collective action in these SHGs. Derived from literature, we studied two particularly important factors that may affect the level of collective action in microfinance SHGs: the degree of heterogeneity among participants and the degree of freedom of participation. In the first step, in-depth interviews with key informants were used to understand the organizational differences between the three credit delivery models. In addition, interview information guided the sampling for the second, quantitative stage of the research design. In this step, a survey with members from nine different SHGs was conducted. Curve estimation was used to explore the functional form of the two variables of interest on the level of collective action. These results were then interpreted against the background knowledge gained from the interview data.
Our article contributes to an ongoing academic debate, namely, the discussion on the role of institutions in mediating heterogeneity problems in collective action (Poteete & Ostrom, 2004; Varughese & Ostrom, 2001). Empirically, we aim to provide policy-relevant knowledge regarding the functioning of the Indian SHG system. Methodically, we demonstrate how curve estimation embedded in a mixed methods design can be used to explore and explain the functional form of community attributes on collective action when this form is theoretically and empirically contested.
The remainder of the article is structured as follows. We first review the literature on collective action problems caused by group heterogeneity and (the lack of) freedom of participation. We then describe the methods and tools used in this study and provide background information on the study area and sample. Finally, we discuss the empirical results and draw a conclusion.
Collective Action and Heterogeneity
A collective action situation exists
when a number of individuals have a common or collective interest—when they share a single purpose or objective—[and when] individual, unorganized action [ . . . ] will either not be able to advance that common interest at all, or will not be able to advance that interest adequately. (Olson, 1965, p. 7)
Mancur Olson’s (1965) theory of collective action on the voluntary provision of public goods states that people will typically free-ride on the contribution of others. In his view, public goods will be generally underproduced unless selective incentives’—private goods attached to the public good—are implemented. From a similar perspective, Hardin (1968) argues that open access common pool resources are overexploited unless privatized or centrally governed by a leviathan (i.e., the state).
This view was challenged by the work of Ostrom (1990) who has repeatedly shown that common pool resources can be managed by local communities if certain institutions—in the broadest sense and throughout this article understood as “sets of rules”—are put into practice. Collective action is likely to be successful if members can communicate frequently and at low costs, if group rules are implemented and monitored, and if violations of rules are gradually sanctioned. Another important prerequisite is that group activities are supported and respected by higher level authorities. In later contributions, Ostrom (2005, 2009) also developed a broader list of additional contextual variables that affect the likeliness of successful collective action in social dilemma situations. These include institutional variables such as well-defined group boundaries, the general freedom for group members to participate in rule-making and rule-changing, and the degree of homogeneity among them with respect to preferences and individual interest.
There are many potential dimensions of heterogeneity (Baland, Bowles, & Bardhan, 2006), and the impact of heterogeneity on collective action is highly contested (Poteete & Ostrom, 2004). For instance, Olson (1965) argues that heterogeneous interests in producing a public good will foster its voluntary provision. This is especially true if thresholds in production—a critical mass—must be passed (Marwell & Oliver, 1993). In other words, group heterogeneity and the shape of the production function for a collective good or service may interact in social dilemma situations (Kollock, 1998). Because it is sometimes difficult to identify and observe benefits from a public good at the individual or household levels, most empirical work focuses on the effect of heterogeneity in income, wealth, technology, gender, or ethnicity on collective action (Baland et al., 2006; Habyarimana, Humphreys, Posner, & Weinstein, 2009; La Ferrara, 2002). A frequent argument is that people who share common traits are more likely to share social networks, affecting subsequent interaction and cooperation (McPherson, Smith-Lovin, & Cook, 2001). People who are alike may have preferences for working with each other; they may also have better knowledge on how to sanction and reward each other, which is important for establishing cooperation (Habyarimana et al., 2009). On the other hand, socioeconomic heterogeneity is often correlated with diversity in approaching problems, which is important for problem solving (Page, 2008). Stated differently, there are costs and benefits of heterogeneity, suggesting the existence of an optimal degree of heterogeneity for solving a particular collective action problem. This point is also stressed by Baland and Platteau (1999):
[W]ealthier users, because they usually have more incentives to “cooperate”, tend to contribute more to collective action. On the other hand, when inequality is large, “small” users internalize such a tiny share of the benefits that they are not prompted to participate in the collective effort. Increasing inequality thus enhances the incentive of the big users to voluntarily contribute and simultaneously encourages the small users to free ride on the former’s contributions. Consequently, the net impact of inequality on collective action will hinge upon the respective strengths of these two opposite effects. (p. 777)
Another important point is raised by Varughese and Ostrom (2001), who, in a study on forest user groups in Nepal, focused on the role of institutions in mediating heterogeneity. They concluded that heterogeneity does not have a determinant impact on the success or failure of collective action. Rather, it is important how collective action is organized and how heterogeneities are addressed in a group’s institutions. Often groups are not completely free to craft their own rules and operate within the limits of higher level authorities. It is important to pay due respect to the institutional environment of a collective action situation.
In rural finance, the classic example of collective action problems are credit cooperatives or Rotating Savings and Credit Associations, which can help in overcoming some of the problems with information asymmetries as well as enforcement and monitoring typically found in rural credit markets (Bonus, 1986). Some studies focus on the effect of heterogeneity or institutions on trust and collective action in such groups (Fadiga & Fadiga-Stewart, 2004; van Bastelaer, 2000). In India, however, such studies are rare. Here, most empirical work is on collective action for the provision of irrigation infrastructure (Bardhan, 2000; Wade, 1987). Nonlinear relationships between economic inequality and collective action are found (Balasubramanian & Selvaraj, 2003).
Organizations of mutual self-help have a long history in rural areas of the developing world (Kwapong & Hanisch, 2013). Formally, most of these organizations were democratically controlled and followed an open membership policy. In practice, however, such is not the case. For example, agricultural marketing cooperatives were often misused to monopolize international trade with cash crops via state marketing boards, thereby taxing the rural population in favor of urban élites (Bates, 1981). Likewise, in India, rural self-help organizations in finance were abused as an effective means for achieving political aims, especially before elections, and it has been claimed that depoliticization is required for the system to actually serve the poor (Dev, 2006). Similarly, Ostrom’s (1990) design principles highlight the absence of interference from higher level authorities for successful collective action. The freedom to actively engage in group decision making becomes a critical success factor as indicated by the literature (Bates, 1981; Kwapong & Hanisch, 2013).
A key lesson of the literature review is that the variables affecting collective action are interlinked, and their relationships are complex. At the same time, people act in a specific institutional context which mediates these variables. In such a situation, it is important to analytically focus on a few important factors (Agarwal, 2001). Our focus is on the level of heterogeneity and the degree of freedom of participation as particularly important variables. The literature on the effect of the two variables on collective action still lacks consensus. Also, empirical work has found diverse functional relationships between heterogeneity, freedom of participation, and collective action. One reason for the lack of causal clarity is that groups can actively adapt to problems caused by heterogeneity by crafting particular rules. The above description of analytical problems fully applies to the study of the different models of microcredit provision via SHGs in India. To address the question of how heterogeneity affects collective action under different credit models, we propose a mixed methods approach which triangulates results of curve estimation with results of interviews.
Mixed Methods Research Design
In the most general sense, mixed methods research involves the combination of two or more methods within a single study. Historically, this research design especially refers to the combination of qualitative and quantitative methods (Johnson, Onwuegbuzie, & Turner, 2007; Teddlie & Tashakkori, 2009). Under the term q-squared, this definition has reached some popularity also in development research (Kanbur, 2004; Shaffer, 2013a, 2013b). Bryman (2006) highlights several dimensions for which a typology of mixed methods studies is suitable. Specifically, one may analyze (a) the succession of data collection which can take place either sequentially or in parallel, (b) the priority given either to the qualitative or the quantitative part, (c) the function fulfilled by combining methods (e.g., triangulation or exploration), (d) the stage at which integration is occurring (e.g., in data collection or in data analysis), and finally (e) the number of data strands.
A different stance is taken by Yin (2006), who highlights that the combination of exclusively qualitative methods (e.g., history and case study analysis) or exclusively quantitative methods (e.g., experiments and surveys) still qualifies as a mixed methods study. Instead, emphasis should be placed on the degree of integration between methods. Yin (2006) develops a framework of questions that could serve as a research heuristic to evaluate the degree of genuine integration between the methods combined within one particular empirical study. More specifically, both methods should (a) address the same research questions, (b) focus on the same level of analysis (although not necessarily involving the same level of data analysis), (c) have some overlap between samples, (d) use comparable instruments in data collection, and (e) be evaluated with respect to the similarity of their analytic strategies. The more criteria are integrated among combined methods in a particular study, the more it will follow a genuine mixed methods approach. Otherwise, a mere parallel of two different studies is likely to occur (Yin, 2006).
Our research design qualifies as being sequential. In the first step, semistructured interviews were conducted with key informants. Representatives from all three models of microcredit delivery (bank promoted, government promoted, and NGO promoted) were interviewed. Figure 1 displays their organizational structure.
Description of microcredit delivery models.
Interviewees provided information on the performance of groups. It was decided beforehand that three key informants should be interviewed per model at different levels of the respective organizations. This approach resulted in nine interviews with representatives from the microfinance provision organizations. These interviews focused on heterogeneity, monitoring, and group organization. A common set of guiding questions was used (see Bharamappanavara, 2010), and the interviewer followed up on interesting points. In addition to these interviews, two government officials, two directors of district NGO foundations, and a representative from the refinancing bank were interviewed using the same interview guidelines and complemented by additional questions. All interviews were conducted by one person, thereby reducing interviewer bias. The interviews were documented through field notes and video recordings. Data were analyzed following an analytical strategy focusing on frequencies of points mentioned and evaluations of informants (Yin, 2008). An overview of the interviews and the topics discussed are summarized in Table 1.
Semistructured Interviews of Qualitative Research Phase.
Note. SHG = self-help group; NGO = nongovernmental organization.
In the quantitative part, a survey of members from the nine SHGs was conducted. The survey was stratified by taluk (subdistrict) and microfinance delivery mode. The three taluks were selected to represent the variation in agri-ecological conditions: Jagalur, a rather poor dry-land area; Harapanahalli, a rather average area with some irrigated land; and Davanagere, a relatively wealthy area with mostly irrigated fields. Sampled SHGs were selected randomly using simple random sampling from a register of all groups in the area.
From a structured questionnaire with closed questions, we gathered information on the socioeconomic background of members, their loans, perceptions of group functioning, and group composition. Groups usually have between 10 and 20 members. To avoid selection bias, we aimed at a larger number of members and chose the minimum group size (i.e., 10 members, resulting in a total of 90 filled questionnaires). The questions that are of main interest in this article refer to the level of collective action (which is also a key performance indicator), the perceived degree of group heterogeneity, and the perceived freedom of participation. Ranking scales were used for all variables which are adapted from the study of the National Bank for Agriculture and Rural Development (also see Bharamappanavara, 2010). The dependent variable was analyzed on a 4-point scale as summarized in Table 2.
Ranking Criteria for the Level of Collective Action.
Source. Adapted from Bharamappanavara (2010).
The degree of heterogeneity was measured as the perception of members on a 5-point scale. After a short introduction, people were asked to state how they perceived the group from (1) “all members are from the same background” to (5) “all members are from very different backgrounds.” In a similar way, data on the perceived freedom to participate in the affairs of the group were gathered. People were asked to rank their influence on a 4-point scale from (1) “I am very free to participate in group affairs” to (4) “I am not free at all to participate in group affairs.”
To explore the functional relationship between the perceived level of collective action and the two independent variables heterogeneity and freedom of participation, a curve estimation tool was used. Curve estimation is a simple explorative procedure to detect the best fit of a nonlinear functional relationship between an independent and a dependent variable. It is implemented in various statistical software packages, including SPSS (Norušis, 2010) or Stata (Wei, 2013). Typically, these packages will estimate various models—linear, logarithmic, inverse, quadratic, cubic, power, compound, S-curve, logistic, growth, and exponential—and, based on goodness-of-fit measures, these packages will select the best performing model and provide respective graphical plots. For our analysis, we used SPSS 16.
Curve estimation is explorative and inductive in its approach, and it is typically used to develop explanation and theory based on the functional relationships observed. Therefore, a good understanding of the case at hand is critical. Typically, only one independent variable is regressed on one dependent variable, and other factors are neglected. For interpretation and triangulation of the observed functional forms, in-depth knowledge is essential. This makes curve estimation particularly prone to systematic complementation by qualitative methods resulting in a mixed methods design. The research design is summarized in Figure 2.
Mixed methods research design.
As seen from the figure, we used a sequential approach to address the question of how heterogeneity affects collective action in this study. Literature revealed a set of diverse theories and empirical results on this topic. Furthermore, it is stressed that the institutional context matters for the mediation of heterogeneity. Background information on this context was collected in the qualitative part of the study. In-depth interviews with representatives from all three microcredit delivery models provided valuable insights into governance aspects and group performance. In a survey, information about members’ assessment was used to produce curve estimation plots and estimate equations. Results were then discussed against the background information extracted through our own interviews and a literature review.
Study Area and Sampling
The study was conducted in three taluks (subdistricts) of Davanagere district in Karnataka, India. A map of the study area is provided in Figure 3.
Map of the study area.
Davanagere is located in the central part of Karnataka encompassing six taluks, which cover an area of 0.597 million hectares. The district is predominantly agrarian with no major- or medium-sized industries. The total population of the district is 1,790,952 with a little more than two thirds of the population living in rural areas. Literacy rates are 76.4% for males and 58% for females, which is comparable with the country average (Government of Karnataka, 2008). Microfinance through SHGs has gained popularity over the last few years, and more than half of the rural loans have been provided through the SHG system. The remaining loans are met informally through family, friends, and moneylenders. From the 27 districts in the state, Davanagere was purposively selected based on data on the proportion of the three microcredit modes under study. Taluks were selected to represent the variation in wealth within the district.
In Davanagere, 19 NGOs operate which organize the rural population into SHGs and establish the links to banks. The Women and Child Development Department (WCDD) is playing an active role in the formation of Sthree Shakti groups, and there are nine banks involved in promoting SHGs in the district. In total, 15,343 SHGs exist in the district, out of which 12,455 provide microcredit within one of the three models.
From a full list of all SHGs, one group per chosen taluk and delivery model was selected using simple random sampling. In each group, 10 members were interviewed, resulting in a total of 90 observations. An overview on the sampled groups is provided in Table 3.
Description of Sampled Self-Help Groups (SHGs).
Note. Tq = taluk; SHG = self-help group; NGO = nongovernmental organization.
The table shows how groups are relatively homogenous with respect to size. Some variation exists with respect to the year of group foundation.
Results
Information From Interviews and Descriptive Evidence
Tables 4 and 5 summarize the findings and provide some sample statements from the in-depth interviews by delivery models with respect to the key performance measure of the group, initial hurdles for group formation, importance of leadership, documentation and bookkeeping of the group, the level of political interference, and the role of trainings on income-generating activities.
Findings From In-Depth Interviews.
Note. NGO = nongovernmental organization.
Sample Statements From In-Depth Interviews by Microcredit Delivery Models.
Note. SHG = self-help group; NGO = nongovernmental organization.
It can be seen that members in bank-promoted SHGs have to carry a high burden for initiating a group. In the bank’s view, the success of a respective SHG mainly depends on its repayment performance. Little assistance is provided for bookkeeping and documentation. Leaders play an important role. Government- and NGO-promoted models are mostly valued by their usefulness in achieving political goals of poverty alleviation. In NGO-promoted groups, a focus is placed on providing training on income-generating activities and the productive use of investments. Political interference is particularly high in the government-promoted model.
Table 6 presents relative and absolute frequencies of key socioeconomic characteristics such as age, education, marital status, type of family, family size, occupation, religion, caste, and annual income.
Frequencies of Key Characteristics by Delivery Models.
Note. SHG = self-help group; NGO = nongovernmental organization.
The data are presented by delivery models, and it can be seen that on average respondents are relatively similar across models. Some differences exist with respect to caste and occupation. Respondents in the NGO-promoted model often have a scheduled caste background and are more likely to work as agricultural laborers. However, this fact does not have strong effects on their income. The last row displays the average number of members per group, indicating also that no substantial differences exist. We can conclude that the groups are socioeconomically similar. This is important for the curve estimation, and we do not control for heterogeneity among individuals or the groups. In Table 7, summary statistics for the variables used in the curve estimation are presented.
Summary Statistics of Model Variables.
The summary statistics show that, at least on the group level, there is no strong confounding of the observed variables with the delivery models. The level of collective action is somewhat lower in bank-promoted groups. Measures of heterogeneity and freedom to participate are fairly close across models. As shown in the summary statistics, no statistically significant differences of the means exist between the three different delivery models for a normal distribution. Also, nonparametric tests accept the null hypothesis of equal samples across models for both the heterogeneity (Kruskal–Wallis test; df = 2; χ2 = 1.338; p = .5122) and the freedom of participation (Kruskal–Wallis test; df = 2; χ2 = 1.388; p = .4996) variables.
Curve Estimation Results
Figure 4 shows the results from the curve estimation by delivery model and study variable.
Curve estimation plots by delivery models and variables.
The y-axis shows the level of collective action as described above. The x-axis of the first row indicates the perceived degree of heterogeneity; in the second row, the x-axis displays the perceived freedom of participation. The estimated equation is displayed below each graph. At a low level of heterogeneity (X = 1), the level of collective action is similar across delivery models with values slightly below three. However, with increasing heterogeneity, the level of collective action changes quite differently depending on the particular model. In the bank model, an increase in heterogeneity leads to a decrease in collective action in the first quarter of the graph; then, the curve moves up. In the government model, heterogeneity has a negative effect over the scale of the variables used. The curve approaches one, and then, predicted values even leave the scale as they decrease below a value of one. Also, in the NGO-promoted model, heterogeneity seems to reduce the level of collective action. The effect is much smaller though, and even at the highest level of heterogeneity, collective action is still around a value of two.
The curves on freedom of participation also show substantial differences between delivery models. Here, the departure points show more variation with roughly 2.5 in the bank model, about two in the government model, and a value of about 2.7 in the NGO-promoted model. The effect of freedom of participation on collective action is very small in the bank model as seen by the constant line, which is almost parallel to the x-axis. In the government model, the curve first grows, reaches a maximum at about 1.5, and then quickly decreases below 1 and ultimately leaving the defined interval at approximately 3.5. In the NGO-promoted model, the curve starts at a relatively high level and then increases exponentially, reaching a level of 3.5 at the maximum of freedom of participation variable. Recall from the scaling that high values actually represent a low degree of participation in the group.
All models were tested for first-, second-, and third-degree polynomial functions, logarithmic, logistic, and s-shaped relationships. The presented models were selected based on the best fit. R 2 values and p values of F statistics are available from the authors on request. Cross-validations are also provided in Tables 8 and 9.
Cross-Validation Heterogeneity Models.
Note. Figures in parentheses indicate standard errors. N = 30, n = 15.
Cross-Validation Freedom of Participation Models.
Note. Figures in parentheses indicate standard errors. N = 30, n = 15.
In these cross-validations, half of a sample of a particular delivery model was selected randomly. The model was then run with this reduced set of observations. Next, coefficients were compared. The results show that coefficients are relatively unstable when reestimated. With a smaller sample, the shape of some curves may change quite substantially. One problem here is the relatively small sample size. Repeated cross-validations yield different results. A larger sample with more variation in the data would be necessary for further investigation.
Interpretation and General Discussion
The results from the curve estimations show that the functional forms vary between delivery models. In the bank model, the degree of heterogeneity has a positive effect on the level of collective action for large parts of the curve. In applying for bank loans, the largest bureaucratic hurdles must be taken by the group. Much of this burden is left to the group, and relatively little assistance is granted to members in this process in comparison with the other delivery models. If all group members are from similar backgrounds and similarly poor, it would be difficult for them to collect the necessary resources and to reach consensus on who should take these burdens. As stressed by Olson (1965) or Marwell and Oliver (1993), some heterogeneity may be helpful for passing threshold levels in producing a good for the group. Leadership may more easily emerge from the group. Of note, the slope is negative in the first part of the curve, indicating that this effect only appears with a somewhat higher degree of heterogeneity. In other words, in our sample, bank-promoted SHGs function better if they are either rather homogeneous or heterogeneous. One interpretation would be that in the first part, heterogeneity causes some conflict on lower levels, but the benefits from heterogeneity quickly crop up with respect to the many hurdles which must be taken.
In the government-promoted model, heterogeneity has a negative effect on the level of collective action. Participation in a government-promoted group is much less demanding, and one can easily become a rather passive member. Even without regular meetings and without implementing all rules, groups will not dismantle. Once formed, a government-promoted SHG will typically continue to exist even if it performs poorly. In this situation, heterogeneity could prevent group cohesion and trigger conflicts as sometimes argued in the development economics literature (La Ferrara, 2002).
Similar arguments apply to the NGO-promoted model. Here, group members can also rely on help and assistance in many processes, especially in comparison to the bank-promoted model. On the other hand, groups can be dismantled, and some collective action is also necessary for group functioning. This observation could be an explanation for the somewhat smaller effect as displayed in the figure above.
In the bank model, freedom of participation hardly has an effect on collective action. Whether single members are very active or hardly participate in group decisions is insignificant for group functioning. Our interpretation is that the role of internal leaders is most important in bank-promoted models. Without strong leadership of one or two members, a bank-promoted SHG is more likely to collapse, relatively independent of how actively or passively individuals participate in decision making. Leaders will be responsible for ensuring a certain level of collective action. Members will not participate in group efforts, not because of their perceived freedom or lack of freedom to do so but rather because leaders take the decision in these aspects.
In the government-promoted model, active participation of members matters for the group’s level of collective action. The average level is relatively low, and when members’ participation decreases, it breaks down completely as seen in the curve. Members are responsible for ensuring that rules are properly implemented and regular meetings take place. As the interview data have shown, monitoring and implementation in this regard is rather poor. Groups continue to exist even under poor performance. For the level of collective action and in our sample, the importance of freedom to participate is most important in the government-promoted SHG model.
In the NGO-promoted model, the relationship differs from the government model. Here, with a decrease in participation, the level of collective action increases. Our interpretation is that if group members control group affairs, collective action reaches fairly high levels. However, most NGO staff is quite experienced and motivated to ensure attendance of members in meetings, spread information, and monitor and enforce rules. They will let members assume these tasks if they have the feeling that the members are free to take part in group activities. If this feeling is not the case, they will drop in and at times do a better job than actual members.
A drawback of the presented quantitative data and results is the relatively small sample size and the ordinal scale of the variables. There are too few observations to perform reliable cross-validations from the data. Also, curve estimations run on continuous data would certainly be more robust and more useful to show greater variation in the functional forms. These limitations in degrees of freedom also do not allow us to control for observed heterogeneity or to incorporate interactions of variables. As discussed in the literature section, many of the variables that are of interest in the study of collective action do not naturally lend themselves to measurement and are difficult to quantify on a continuous scale. Indices constructed from extensive lists of categorical variables would be a possible way to address this drawback. It could also be interesting to reverse the order of the quantitative and the qualitative parts of the mixed methods design. Key experts in the field could be confronted with curve estimation graphs, suppressing information which graph represents which model. Their judgments would then provide an additional source of triangulation.
Finally, it is important to evaluate the mixed methods approach. According to Yin (2006), this especially concerns the degree of integration between methods. An evaluation can be conducted by reviewing research questions, the level of analysis, overlap between samples, instruments and data collection, and analytic strategies. In our study, both methods inform each other, and both methods concern the effect of heterogeneity and freedom of participation on collective action in SHGs in rural India. While the qualitative part also investigates institutions and structural aspects at the level of the organizational model, the quantitative part focuses on the individual level. The same level of analysis (i.e., the group level) is used, although data are analyzed on different levels: the group level for the in-depth interviews and the individual level for the curve estimation. Samples target different groups of interest and do not overlap. The qualitative part was based on information provided by key informants; the quantitative part focused on members. Also, instruments used in the analysis are not strongly integrated, and different instruments have been used for the two samples. Yet a similar analytic strategy has been applied to both cases. This strategy especially concerns the way in which dependent and independent variables and their (contested) relationship has been constructed. In both methods, we have focused on understanding the functional relationships of heterogeneity and freedom of participation on the level of collective action. Structural aspects and their interaction with collective action are covered in particular by the qualitative part.
Taken together, the degree of integration between methods in our study can be described as moderate. To achieve greater integration in future work, instruments and sampling could especially be put forward. For instance, qualitative interviews could also be conducted with members, and a survey could be distributed among officials and representatives. Along the same lines, members could be questioned about their perceptions of structural aspects, and key officials could provide more information on group performance.
Summary and Conclusion
Collective action is a complex phenomenon, which is not easy to analyze and compare across different contexts. We have studied the effect of members’ homogeneity and freedom of participation on the level of collective action while using a sequential mixed methods research design. The comparison has been made between three different microcredit delivery models in Davanagere district, Karnataka, India. Our results suggest that the functional forms of these variables are mediated by the institutional context. Heterogeneity is rather beneficial for groups promoted by banks, has relatively large negative effects on government-promoted groups, and has somewhat smaller negative effects on NGO-promoted groups. Freedom of participation is not decisive for collective action in bank-promoted SHGs. Low freedom of participation hinders collective action in government groups and has a positive effect on NGO-promoted groups. These findings have been compared with in-depth interviews, allowing possible explanations for the observed relationships.
Our results show that collective action is developed best in the NGO-promoted model and worst in the government-promoted model. This finding is consistent with recent literature on the positive effect of NGOs for development. One way to improve the level of collective action of government-promoted SHGs may be to adopt monitoring patterns of NGO-promoted groups. This is particularly important as the level of collective action affects the performance of the group. Our data support the notion that the success of the NGO-promoted model can to a large extent be attributed to its well-trained staff. In principle, proponents of the other models could focus on training their staff in a similar manner to improve performance. However, this eventually bears the risk that SHGs become fully dominated by their promoting organizations. Group independence and a minimum degree of members’ participation should be ensured even at the cost of poorly functioning groups. Otherwise, the independent and voluntary character of these groups may vanish. This can already be observed for some of the government-promoted SHGs. Tools of poverty alleviation then become mere tools of rural politics through which politicians distribute money for votes before elections are held. Methodically, our approach has demonstrated how curve estimation can provide a quantitative basis for empirical work when diverse theoretical predictions exist. However, in-depth knowledge of the case is a prerequisite.
In summary, collective behavior is multifaceted, and the level of collective action is different in all three microcredit models with respect to heterogeneity and freedom of participation. Eventually, success and failure of SHGs will to a great extent depend on these variables, which are mediated by the delivery model’s institutions.
Footnotes
Acknowledgements
We thank the editor and two anonymous referees for their helpful and encouraging comments. Additionally, Saikumar Bharamappanavara also thanks the German Academic Exchange Service (DAAD) for a PhD scholarship received under the program “Future Megacities – Energy- and Climate-Efficient Structures in Urban Growth Centres”.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: We are grateful for financial support from the German Federal Ministry of Education and Research (BMBF) as part of the “Sustainable Hyderabad” project (Grant Number: FKZ 01LG0506A1). Saikumar C. Bharamappanavara is also grateful for financial support received under the European Commission’s Erasmus Mundus program for his studies in the International Master of Rural Development.