Corresponding author: Michael Alexeev ( malexeev@indiana.edu ) © 2021 Non-profit partnership "Voprosy Ekonomiki".
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits to copy and distribute the article for non-commercial purposes, provided that the article is not altered or modified and the original author and source are credited.
Citation:
Alexeev M, Nurmakhanova M, Polishchuk LI (2021) Institutions and social capital in group lending. Russian Journal of Economics 7(4): 269-296. https://doi.org/10.32609/j.ruje.7.76647
|
Formal institutions and social capital interact with each other in multiple ways. We argue and show empirically at the cross-country level that in the case of group lending, contract enforcement complements bonding social capital and substitutes for bridging one. It means that payoff to social capital in group lending depends on social capital type and is contingent on the quality of contract enforcement which serves as a sorting factor, working in the opposite directions for different stripes of social capital. These results are robust to various estimations, sets of controls, and social capital measures.
social capital, contract enforcement, microfinance.
Over the last few decades, a vast empirical literature has firmly established the high significance of institutions in determining various economic outcomes (see, e.g.,
On the one hand, institutions and culture serve similar purposes, both being coordination devices, supplying “rules of the game” for economy and society. This makes one expect that institutions and culture are substitutes. Indeed, low trust societies and communities, which lack confidence in their grassroots ability for cooperation and coordination, yearn for greater government regulation (
On the other hand, the performance of formal institutions depends on cultural attitudes and norms in society. For example, regulation could be ineffective due to cultural reasons, such as tolerance for corruption (
The interplay between formal institutions and culture is important for the assessment of economic payoffs to either of these factors. This includes measurement of payoff to social capital, commonly defined as a nexus of cultural traits required for cooperation and collective action (
The complexity of the interaction between social capital and institutions calls for further studies, including those of specific sectors and types of collective action, where the linkages of interest are more tangible, and hence easier to identify and conceptualize, than at the macro level. In this paper, we follow such a path by exploring the joint role of institutions and social capital in determining the patterns of lending in microfinance, and explaining the reliance on group lending.
Microfinance institutions (MFI) offer group lending as an alternative to the conventional individual loans, especially when the latter are unavailable to poorer households and small businesses that are unable to pledge collateral against a loan. Secured lending is a well-known means to deal with moral hazard and adverse selection, which would otherwise make lending problematic. To unlock access to finance in the absence of secured lending, microfinance institutions, many of them inspired by the Grameen Bank in Bangladesh, resort to group lending. This financial instrument relies on “social collaterals” (
In group lending, borrowers form self-selected teams in which they are jointly liable for each other’s personal loans; such liability could include joint responsibility to repay all loans and/or shared reputational consequences if some of the loans are not repaid. More generally, under joint liability, all group members are considered in default, if any of the members fails to repay his/her loan (
Group lending is well suited as a testing ground for theories and hypotheses concerning the interplay of formal institutions and cultural traits, especially those related to social capital. First, obtaining and repaying a group loan is a quintessential collective action, and as such, it should be rooted in social capital, which embodies social collateral (similarly to physical capital pledged as a conventional tangible collateral in individual lending). Indeed, mutual trust, shared values, social networks, peer monitoring and other ingredients of social capital are all prominent features and prerequisites of collective lending (
We argue that joint presence, relevance, and interaction of social capital and institutions could be behind the patchwork of findings reported in the earlier studies of the impact of social capital on group lending, and that heretofore loosely associated and sometimes even contradictory observations from those studies can be integrated into a coherent picture once institutions are factored in. In doing so, we explicitly incorporate in our reasoning the asymmetric information problem essential for lending and borrowing. We retain for our analysis a general measure of contract enforcement, representing institutions, and distinguish between bonding and bridging types of social capital (
We outline a conceptual framework, in which the impacts of bonding and bridging social capitals on group lending are conditional on the strength of formal institutions. Our logic is centered on the informational asymmetry in group lending, where both “good” and “bad” collective loan applications could be expected, and lenders cannot tell one from the other, creating a well-known “lemons” problem (
We take these predictions to empirical testing using Microbank Bulletin dataset comprising 450 MFIs from 44 countries for the period of 2006–2010. Our social capital measures are conventionally drawn from the World Values Survey, and the rule of law index — from the Worldwide Governance Indicators (WGI) for the same period. Estimations of regression models combining data from the above sources fully and robustly support our hypotheses, indicating that the impact of social capital on group lending remains undetermined unless different stripes of social capital are considered separately, and until the strength of contract enforcement, serving as a “sorting factor,” is factored in. Furthermore, this sorting factor works in opposite directions for bonding and bridging stripes of social capital.
The paper contributes to the literature by offering a conceptual framework for analysis of the role of different types of social capital in group lending, conditional on the strength of formal institutions. This framework provides a unified explanation of diverse results of earlier empirical studies of group lending, which reflect various combinations of institutions and social capital. We demonstrate qualitatively different roles of bridging and bonding stripes of social capital in the same economic activity, and show that both types of social capital could be either assets or liabilities depending on the quality of institutions. Therefore, the paper generates new insights into the interplay between institutions and culture in the presence of informational asymmetry.
The rest of the paper proceeds as follows. Section 2 provides a brief overview of existing literature, Section 3 outlines the theoretical framework, Section 4 describes the methodology and data used in the analysis, Section 5 presents and discusses the results, and Section 6 concludes.
The type of lending provided by MFIs depends on the relative advantages and disadvantages of individual vs. group loans. Other things being equal, small-scale entrepreneurs in developing countries normally prefer to borrow as individuals, just as observed in developed countries, since entrepreneurs typically do not want to be monitored and directed by others and/or be responsible for someone else’s performance. However, in the absence of tangible collaterals, or in an environment where collecting on a collateral is prohibitively costly, individual lending becomes problematic, and group lending emerges as a practical alternative. Group lending enables previously excluded borrowers to jointly create “social collateral” and use it to secure otherwise unavailable credit (
Group lending received significant attention in the literature, both theoretically and empirically. Theoretical studies highlight comparative advantages and disadvantages of group vs. individual lending, and in particular focus on how “social collaterals” can alleviate the adverse selection and moral hazard problems through team formation and monitoring.
A number of influential papers address within-group monitoring and enforcement issues and other relational advantages of group lending.
Overall, comparative advantages of group lending over individual lending summarize as follows:
However, group lending also carries significant disadvantages and additional risks in comparison to individual lending, which are as follows.
Reflecting these benefits and costs, numerous empirical studies present a mixed picture of successes and limitations of group lending. Thus, in Mongolia, access to group lending increases consumption and boosts entrepreneurship, whereas access to individual lending has no such effects (
Empirical research is also ambiguous on the profitability of the two types of lending to MFIs. While
Some empirical studies confirm the importance of monitoring and within group enforcement, indicating that social capital stock plays an important role.
The above evidence highlights the relevance of institutional and socio-cultural factors for the performance and effectiveness of group lending. In the next section, we outline a conceptual framework, which identifies such factors and generates testable hypotheses about their joint impact for group lending.
As argued in the Introduction, since group lending is a collective undertaking, the capacity for collective action, known as social capital, should be relevant for the success of this MFI instrument. Indeed, the concept of social capital, variously defined and interpreted, features prominently in the microfinance and group lending literature (see, e.g., surveys in
The concept of social capital allows multiple interpretations, which are extensively debated in the literature (see, e.g.,
Obviously, all of the above factors facilitate the formation of borrowing teams that are prospective recipients of group loans and increase teams’ confidence in accepting collective liability for loan repayment and in the ability to successfully implement projects of team members. This, in its turn, gives lenders additional assurance in loan repayment, true to the spirit of “social collateral.”
The literature draws a distinction between different types of social capital as contributing factors to group lending. One important distinction is between “bonding” social capital confined to smaller close-knit communities and ultimately to borrowing groups, and “bridging” one, encompassing large segments of the society (see, e.g.,
One could expect that bonding social capital should have more immediate bearing on group lending, which requires collective action within small teams, and this is indeed what
Furthermore, the payoff to bonding social capital in group lending is uneven across countries and societies. According to
A possible explanation of such discrepancy could be in the strength of formal institutions, which affect the payoff to social capital in group lending. According to
The assortment of “fascinating, but somewhat contradictory results” (
Our logic, briefly outlined in the Introduction, elaborates as follows. First, bonding social capital helps in the formation of smaller borrowing groups and in sustaining cooperation inside such groups. This effect makes group members more willing and better able to be engaged in collective undertakings. However, group lending unlocks two types of group applications, first of which are bona fide low-risk (for short — “good”) borrowing projects, and the second — high-risk or fraudulent (“bad”) projects, comprising “willful defaulters” or excessively risky projects. Bonding social capital is a hatchery for both types of group projects, as it enables collective action and collaboration within smaller groups irrespective of the goal of the collective action. The goals could be either to collectively repay a loan required for a good project (in which case group members monitor and/or assist each other), or just to obtain a loan for a group with a bad project, when group members are aware that the project is too risky or even fraudulent, and collude with each other against the lender. Bonding social capital in and of itself puts no filters to bad projects due to limited morality and low radius of trust within small project groups, so it is possible that group members have no scruples about offloading external costs of collective opportunistic behavior upon outsiders — in the case of collective loans, upon lenders.
Higher stock of bonding social capital increases the aggregate demand for group loans, potentially adding more of both good and bad project applications. However, the mix of such increase depends on contract enforcement institutions. If the latter are strong, “lemon” projects would be brought to effective collection, or ruin the credit ratings of applicants, and hence such projects would be less appealing and perhaps altogether avoided, staying “outside the equilibrium path.” Therefore, all or nearly all of the demand for group loans increases, fueled by bonding social capital, and would comprise good projects, which lenders would willingly fund. This leads to a conclusion that under strong contract enforcement the incremental contribution of bonding social capital to group lending is positive, and this is exactly what was observed by
Weaker rule of law makes the marginal effect of bonding social capital for group lending weaker, too. Indeed, laxer contract enforcement increases the appeal of bad projects, since opportunistic borrowing teams stand better chances to get away with those. It means that “lemons” comprise an increasing share of aggregate demand. Lenders respond to greater risks of borrowers’ default by raising interest rates to include higher risk premium.
Monotonic increase of the marginal effect of bonding social capital in the strength of contract enforcement means that these two factors complement each other in group lending. Whether this complementarity is pronounced enough to pull the marginal effect of bonding social capital in the negative territory when contact enforcement is poor, is a priori unclear, but evidence from some developing countries, including the aforementioned study of Thailand (
We now turn to bridging social capital, which is based on broadly applicable ethical norms, trust, and reciprocity, extending beyond smaller groups, and which reduces the likelihood of opportunistic behavior even against socially distant victims, including lenders. Such cultural traits give lenders extra assurance that borrowers would be implementing their projects in good faith and do not intend to defraud lenders and/or take advantage of zero (tangible) liability and select excessively risky projects. Thus, bridging social capital affects the supply side of unsecured lending, where its role is similar to formal contract enforcements, making these two factors substitutes.
This means that, in agreement with the literature reviewed above, social capital could be, depending on its type, either the substitute of, or complement to, formal institutions. In both cases payoff to social capital monotonically depends on the strength of contract enforcement, making the latter a sorting factor in relation between social capital and group borrowing, but such a factor works in the opposite directions for bonding and bridging types of social capital.
To gain better insight into the impact of bridging social capital, conditional on the strength of institutions, consider first the case of weak contract enforcement. In such an environment, general morale and trust of bridging social capital are the only assurance devices against the “lemons,” and as such facilitate group lending. A high stock of bridging social capital greatly reduces the lending risks, making good projects predominant in the loan portfolio, and suppresses the adverse selection effect that would otherwise put limits to group lending. Low-risk portfolio pulls down interest rates, which further increases the number of good applications with little if any rise of bad ones, and the group loan portfolio expands. Therefore, against the backdrop of weak contract enforcement, bridging social capital makes a positive contribution to group lending (which mirrors its role at the macro level, as established in
Substitution between the rule of law and bridging social capital makes one expect that as contract enforcement grows stronger, the significance of bridging social capital should be declining. Again, it is an empirical question whether the payoff to bridging social capital turns negative over the higher range of contract enforcement — this could be expected if bridging social capital helps eliminate the few remaining “lemons,” which lenders were willing to tolerate in predominantly low-risk loan portfolios.
The above discussion summarizes to the following testable hypotheses.
Hypothesis BO1. Bonding social capital and contract enforcement are complements is group lending.
Hypothesis BO2. When contract enforcement is strong, bonding social capital has a positive marginal impact for group lending. When contract enforcement is weak, the sign of the bonding social capital’s impact is theoretically uncertain, but is expected to be negative.
Hypothesis BR1. Bridging social capital and contract enforcement are substitutes in group lending.
Hypothesis BR2. When contract enforcement is weak, bridging social capital has a positive marginal impact for group lending. When contract enforcement is strong, the sign of the bridging social capital’s impact is ambiguous in theory but is expected to be negative.
It is noteworthy that individual lending should be much less sensitive to social capital stocks than a group lending. This is particularly clear in the case of bonding social capital, which by definition facilitates collective action in small groups, and as such is idled in individual lending. Bridging social capital, in its turn, affects social collateral, essential for group lending, but has no bearing on physical collateral, which underpins individual lending, and therefore the latter is less affected by bridging social capital as well. This means that the above hypotheses about the impact of social capital of both types on group lending should also hold for shares of group lending in MFI portfolios (the rest of portfolios comprise individual loans). We make use of such extension in the empirical analysis that follows.
In contrast, formal institutions should be expected to be relevant both for group and individual lending. One channel of such influence is the ease of obtaining property titles on assets owned de facto, that would allow for the pledging of such assets as tangible collaterals for individual loans. It is well known since
While some of the above hypotheses find occasional support in the cases presented in the earlier literature, our paper brings them together in a systematic and coherent way. We now turn to testing these hypotheses directly at the cross-country level. Such sample selection is determined by the nature of our research question, i.e., to study the joint impact on group lending of institutions and social capital. While the former can exhibit significant variations between various communities and localities, the latter varies between larger jurisdictional units, in particular across countries, which in the next section are our observations.
Our main variable of interest is a measure of group lending by MFIs. There are three basic models of lending employed by such institutions: solidarity group, village banking and individual loans. We lump solidarity group and village banking in the group lending category, whereas the balance is individual lending. Data on lending methodology employed by MFIs come from Microbank Bulletin (MBB). Information is provided for 450 MFIs from 71 countries for the period of 2006-2010. Unfortunately, our data source does not provide information on the volume of group lending, and we focus instead on the proportion of MFIs in a given country that deal in group lending. This variable (hereafter Group) in fact reflects the relative scale of group lending vs. individual loans, but as explained at the end of the previous section, it is suitable for testing our hypotheses on the role of social capital, conditional on formal institutions, in group lending. Our data is insufficient for a well-balanced panel, and we instead resort to cross-sectional analysis at the country level. Limitations of the data on independent variables introduced below leave a sample of 44 countries, listed in the Appendix A.
We conventionally use World Values Survey (WVS) data to evaluate social capital in different countries. This global research project explores people’s values and beliefs and provides information on trust, integrity, tolerance, solidarity, attitudes to community life and government institutions, etc. Theoretical descriptions of bonding and bridging social capital by
To gauge bonding social capital, we use three questions that are related to trust within various close-knit groups, including trust in people respondents know personally and in people within their family and neighborhood (1 — trust completely, 2 — somewhat, 3 — not very much, 4 — no trust at all).
To measure bridging social capital, we select ten questions related to general social trust, reciprocity and civic responsibility. Some questions ask respondents to evaluate their trust in people that they met for the first time, people of another nationality and people of another religion. Responses to these questions use a 4-point scale (from 1 — trust completely to 4 — not trust at all). Another question, “Generally speaking, would you say that most people can be trusted or that you need to be very careful in dealing with people?,” particularly popular in empirical social capital studies (
To measure adherence to moral norms, we select four questions related to the acceptance of some forms of unethical or unlawful behavior, such as: cheating on taxes if you have a chance; dodging a fare on public transport; claiming government benefits to which you are not entitled; and accepting a bribe in the course of someone’s duties.
We aggregated the answers to the above questions into three variables — Narrow Trust (NT), General Trust (GT), and Integrity (IN) — using conventional factor analysis described in Appendix B. Narrow trust serves as a proxy for bonding capital while general trust and integrity reflect different aspects of bridging social capital.
Following earlier literature (e.g.,
To test the hypotheses listed in the previous section, we use the above measures of social capital and contract enforcement, and interactions thereof, as independent variables in regression models with the share of group lending as the dependent variable. We include in our regressions different sets of control variables with potential relevance for group lending. Thus, MFI portfolios could be affected by the overall performance of the country’s financial sector (
Another control variable, the Gini coefficient, measures income inequality in the country, and thus reflects the degree of poverty, which is relevant to the availability of collateral and, therefore, might play a role in the choice of lending methodology. We employ World Bank estimates of the Gini coefficients expressed in percent. In addition, we use the World Bank data on real per capita GDP, which also reflects the availability of collateral, and more generally, accounts for differences in the degree of development across countries.
Finally, we control for Legal Origin (1 for common law, 0 otherwise) to account for different approaches to property and contracts, which is a convention in comparative institutional analyses.
Since the data on lending methodology employed by MFIs is provided for the period of 2006–2010, the data on financial development, Gini coefficient and per capita GDP are also averaged for the same period. Per capita GDP is calculated based on purchasing power parity and 2011 dollars were used to convert current dollars. For social capital data, we calculate measures of the bonding and bridging social capital by averaging the relevant data from several waves of the World Values Survey (1990–2014), which is justifiable by slow changes in cultural indexes (
Full descriptions of main variables and their sources, and the descriptive statistics are presented in Appendix C.
In order to test the hypotheses formulated in Section 3, we need to model the non-linear relationship between group lending and our main variables. We do this by introducing pairwise interactions between Contract Enforcement and different components of social capital: Integrity, General Trust and Narrow Trust. Specifically, we estimate the following regressions:
Group = β 0 + β 1 CE + β 2 GT + β 3 IN + β 4 NT + β 4 CE × GT +
+ β5 CE × IN + β6 CE × NT + γX + ε, (1)
where Group denotes the share of group lending in the country; CE, GT, IN, and NT stand for Contract Enforcement, General Trust, Integrity, and Narrow Trust, respectively; and X is a vector of control variables described in the previous section.
The use of interactions between institutions and social capital is essential to properly estimate their effects. This is because according to our hypotheses, the effects of social capital have opposite signs, depending on the strength of formal contract enforcement. Therefore, in a simple linear regression the coefficients might be insignificant because the effects of social capital at different levels of formal rule of law would be canceling each other and their signs would be theoretically indeterminate. This is precisely what we observe in a model without interaction terms where the coefficients of our main variables of interest are statistically insignificant (see Appendix E).
Employing the interaction terms between the measures of social capital and the rule of law naturally allows for the empirical tests of our hypotheses. For example, in order to test hypothesis BO1, we need to evaluate the coefficient of the product of the bonding capital measure (NT) and the measure of formal contract enforcement (CE). Since we hypothesize that bonding capital and the rule of law are complements, we expect the coefficient of their product to be positive, implying that stronger formal contract enforcement increases the positive effect of Narrow Trust on the propensity of MFIs to engage in group lending.
Our second hypothesis, BO2, states that the impact of Narrow Trust on group lending depends on the strength of contract enforcement. When the latter is strong, Narrow Trust positively affects the propensity for group lending. Conversely, when the rule of law is weak, the effect of Narrow Trust becomes theoretically uncertain, but, as we argued earlier, it is most likely negative. Therefore, to test this hypothesis we need to use an empirical specification that allows for the dependence of the marginal effect of Narrow Trust, , on the strength of formal contract enforcement. This is accomplished by the formulation in (1) where
For the marginal effect in (2) to be positive for high values of CE and negative for low ones, requires that β4 < 0 and β7 to be sufficiently highly positive (specifically, β7 CE^ ≫ β4, where CE^ denotes maximum possible value of contract enforcement). Similar considerations apply to our expectations of the signs of other coefficients implied by hypotheses BR1 and BR2. Namely, we expect β2 and β3 to be positive while β5 and β6 to be sufficiently strongly negative, implying that bridging capital and formal contract enforcement are substitutes and making marginal effects of bridging capital change signs from positive to negative as formal contract enforcement increases. The estimates presented below strongly support all of our hypotheses.
We estimate equation (1) using OLS and Tobit. The latter is a more appropriate specification because MFIs in five of our 44 countries do not use group lending, implying that we have a significant share of observations of the dependent variable bunched at one of the extremes.
Estimation results of the above equation are reported in Table
Variables | OLS | Tobit | ||||||
(1) | (2) | (3) | (4) | (5) | (6) | |||
Contract enforcement | 1.201 (0.162) | 1.564* (0.088) | 0.859 (0.205) | 1.978** (0.029) | 2.809*** (0.001) | 2.076*** (0.000) | ||
General trust | 0.915** (0.035) | 0.705 (0.157) | 0.649 (0.129) | 1.559*** (0.003) | 1.707*** (0.001) | 1.598*** (0.000) | ||
Contract enforcement × General trust | –3.091** (0.010) | –2.821** (0.045) | –2.610** (0.028) | –5.315*** (0.001) | –6.325*** (0.000) | –5.921*** (0.000) | ||
Integrity | 0.804* (0.082) | 0.903** (0.030) | 0.559 (0.104) | 1.351** (0.023) | 1.727*** (0.000) | 1.328*** (0.000) | ||
Contract enforcement × Integrity | –2.422* (0.057) | –2.774** (0.028) | –1.703* (0.096) | –4.158** (0.015) | –5.200*** (0.000) | –4.000*** (0.000) | ||
Narrow trust | –0.830 (0.113) | –0.843 (0.109) | –0.597 (0.169) | –1.491** (0.022) | –1.726*** (0.001) | –1.424*** (0.000) | ||
Contract enforcement × Narrow trust | 3.300** (0.031) | 3.180** (0.031) | 2.750** (0.020) | 5.637*** (0.010) | 6.440*** (0.000) | 5.805*** (0.000) | ||
Logarithm of per capita GDP | –0.149*** (0.004) | –0.097** (0.036) | –0.214*** (0.000) | –0.155*** (0.001) | ||||
Gini coefficient | –0.002 (0.728) | –0.003 (0.492) | 0.002 (0.761) | 0.000 (0.978) | ||||
Financial development | 0.209 (0.495) | 0.047 (0.855) | 0.456 (0.125) | 0.259 (0.241) | ||||
UK legal origin | 0.296*** (0.000) | 0.286*** (0.000) | ||||||
Constant | –0.242 (0.221) | 1.080** (0.049) | 0.813* (0.067) | –0.521** (0.024) | 0.945* (0.065) | 0.665* (0.082) | ||
R-squared | 0.264 | 0.422 | 0.582 |
The results for marginal effects of all three measures of social capital are presented in Figs
Again, we find strong confirmation of our hypotheses. Indeed, according to Fig.
We observe mirror reflection pictures for both measures of bridging social capital (Figs
The coefficients of our main variables of interest and their marginal effects are robust to the inclusion of several other control variables. As another robustness checks, we replace the principal components used as social capital measures by answers to specific questions from the World Values Survey, that appear to be most representative for the broad categories of narrow trust, general trust, and integrity. To this end, we retain the following questions for the above categories: “Trust people known personally” for narrow trust; “Most people can be trusted” for general trust; and “Not justified to claim government benefits to which you are not entitled” for integrity. The results shown in Appendix D are qualitatively indistinguishable from those based on the principal components. In particular, they also strongly support our hypotheses.
The above empirical analysis demonstrates high degree of robustness of our findings and conclusions to variations in controls, estimation techniques and social capital measurement.
Group lending has been a significant institutional innovation that allowed MFIs to service clients without tangible collateral, thus sidestepping a major obstacle to secure finance and ultimately to economic development. Group lending relies on social collaterals as a palliative means to deal with moral hazard and adverse selection, inherent in credit markets. As such, it is based both on cultural traits, especially those relevant for collective action, and institutions inherent to lending and borrowing. Hence, group lending presents a natural case for analysis of the interplay between institutions and social capital. Furthermore, it allows to isolate the impacts of bonding and bridging stripes of social capital and show that they are contingent on the quality of contract enforcement.
Bonding and bridging social capitals play different roles in group lending — the former facilitates the formation of borrowing groups and cooperation therein but gives no assurance against moral hazard that concerns the lender, whereas the latter provides such assurance by upholding integrity and universal morality that straddle over group boundaries and make moral hazard less likely. These effects operate on different sides of financial markets, affecting, respectively, demand for and supply of group loans. Contract enforcement is an alternative third party means to prevent opportunistic behavior and keep moral hazard in check, and as such, it substitutes for bridging social capital, which is embedded in norms and values and serves a similar purpose in group lending. Therefore, contract enforcement and bridging social capital substitute for each other, which can be clearly seen in the data. On the other hand, contract enforcement reinforces the payoff to bonding social capital in group lending, by allowing to make use of the capacity for collective action within small borrowing teams, while putting roadblocks to “lemon” projects, to which bonding social capital alone is not a filter.
These arguments are rather general and can be applied to other markets and types of collective action, where teams are engaged in various transactions with outside counterparts. One specific example would be a team of workers contracting with a customer for a small construction project such as replacing a roof or building a deck, where the customer pre-pays for a substantial portion of the work. Another example is coproduction of urban infrastructures in Brazil described in
Leonid Polishchuk gratefully acknowledges support of the Basic Research Program of HSE University, Moscow, Russia.
Argentina | Ethiopia | Mexico | Russia |
Armenia | Georgia | Moldova | Rwanda |
Azerbaijan | Ghana | Montenegro | Serbia |
Brazil | Haiti | Morocco | South Africa |
Bulgaria | India | Nigeria | Tunisia |
Burkina Faso | Indonesia | Pakistan | Turkey |
Chile | Jordan | Palestine | Ukraine |
China | Kazakhstan | Peru | Uruguay |
Colombia | Kyrgyzstan | Philippines | Uzbekistan |
Ecuador | Malaysia | Poland | Vietnam |
Egypt | Mali | Romania | Zambia |
We use the following WVS questions to construct our main variables.
Bonding social capital
Bridging social capital
To aggregate the relevant questions from WVS into composite indexes, we begin with data compression by exploratory factor analysis (EFA) using principal component extraction with promax rotation. This is a standard approach to identify latent constructs from a larger set of correlated variables (Finch and West, 1997; Fabrigar and Petty, 2001; Norris and Lecavalier, 2010).
Next, we employ confirmatory factor analysis (CFA) to test the consistency of the factors with the nature of the proposed construct. The aim of conducting the CFA is to test the fit of the data and hypothesized research model (
Next, we turn to exploratory principal factor analysis, using scree plot of eigenvalues (Fig.
Initially, there are four factors above the eigenvalue 1 threshold.
Item | Narrow trust | General trust | Integrity |
Trust people in family | 0.875 | ||
Trust people in neighborhood | 0.657 | ||
Trust people known personally | 0.513 | ||
Most people can be trusted | 0.648 | ||
Trust people met for the first time | 0.792 | ||
Trust people of another nationality | 0.752 | ||
Not justified to claim government benefits to which you are not entitled | 0.748 | ||
Not justified to avoid a fare on public transport | 0.786 | ||
Not justified to cheat on taxes if you have the chance | 0.825 | ||
Not justified to accept a bribe | 0.804 |
Factor | Scale reliability Cronbach alpha coefficient |
Narrow Trust | 0.69 |
General Trust | 0.74 |
Integrity | 0.79 |
Next, we perform factor analysis on the remaining ten items with consequent oblique rotation of factors assuming that constructs would be correlated. As a result, we retain three factors, interpreted as Narrow Trust, General Trust, and Integrity. Narrow Trust social capital includes various measures of trust within a smaller group (such as family, neighbors, friends, etc.), while General Trust reflects trust in people outside of such groups. Integrity reflects rejection of unlawful/unethical behavior even against socially distant parties such as the state. Rotated factor loadings are as follows (Table
Next, we calculate Cronbach’s alpha coefficients for each factor to measure the reliability of factors and their internal consistency, that is, how closely the items within a factor are related to each other (Table
Variables | Definition |
Group lending | Proportion of MFIs in the country that employ group lending. Source: The MIX platform (https://www.themix.org/) and authors’ calculations. |
Contract enforcement | Rule of Law index. Source: World Governance Indicators: https://databank.worldbank.org/source/worldwide-governance-indicators |
Integrity | Social capital aspects related to attitude toward civic duty and responsibility. Source: World Values Survey (http://www.worldvaluessurvey.org/WVSDocumentationWVL.jsp) and authors’ calculations. |
Narrow trust | Social capital aspects that reflect aspects of trust within a group (might include a family, neighbors, friends, etc.). Source: World Values Survey (http://www.worldvaluessurvey.org/WVSDocumentationWVL.jsp) and authors’ calculations. |
General trust | Social capital aspects reflecting trust for people outside of own group. Source: World Values Survey (http://www.worldvaluessurvey.org/WVSDocumentationWVL.jsp) and authors’ calculations. |
Financial development | Composite index of financial development. Source: World Bank Financial Development Database (https://databank.worldbank.org/source/global-financial-development). |
Gini coefficient | Gini coefficient measures income inequality in the country. Source: World Development Indicators (https://datacatalog.worldbank.org/dataset/world-development-indicators). |
Per capita GDP | Real per capita GDP in PPP terms in 2011 U.S. dollars. Source: World Development Indicators (https://datacatalog.worldbank.org/dataset/world-development-indicators). |
UK legal origin | A dummy variable that takes on a value of 1 for common law countries (British legal origin), and 0 otherwise. Source: La Poprta et al. (1999). |
Variable | Descriptive statistics | |||
---|---|---|---|---|
Mean | Std. dev. | Min | Max | |
Group lending | 0.566 | 0.314 | 0 | 1 |
Descriptive statistics for individual survey questions | ||||
Most people can be trusted | 1.264 | 0.441 | 1 | 2 |
Not justified to claim government benefits to which you are not entitled | 8.427 | 2.481 | 1 | 10 |
Not justified to avoid a fare on public transport | 8.437 | 2.426 | 1 | 10 |
Not justified to cheat on taxes if you have the chance | 8.728 | 2.237 | 1 | 10 |
Not justified to accept a bribe | 9.180 | 1.830 | 1 | 10 |
Trust people in family | 3.801 | 0.491 | 1 | 4 |
Trust people in neighborhood | 2.883 | 0.815 | 1 | 4 |
Trust people known personally | 2.983 | 0.783 | 1 | 4 |
Trust people met for the first time | 1.961 | 0.789 | 1 | 4 |
Trust people of another nationality | 2.210 | 0.859 | 1 | 4 |
Descriptive statistics for factors generated by factor analysis | ||||
Integrity | –0.070 | 0.487 | –1.725 | 0.523 |
General trust | –0.168 | 0.307 | –1.006 | 0.414 |
Narrow trust | –0.050 | 0.422 | –1.690 | 0.692 |
Integrity (standardized) | 0.736 | 0.217 | 0 | 1 |
General trust (standardized) | 0.590 | 0.216 | 0 | 1 |
Narrow trust (standardized) | 0.688 | 0.177 | 0 | 1 |
Descriptive statistics for other variables | ||||
Contract enforcement | –0.403 | 0.550 | –1.343 | 1.301 |
Contract enforcement (standardized) | 0.356 | 0.208 | 0 | 1 |
Financial development | 0.357 | 0.157 | 0.061 | 0.641 |
Gini | 43.653 | 10.297 | 26.700 | 63.200 |
Ln(GDP) | 8.151 | 0.905 | 5.693 | 9.366 |
UK legal origin | 0.159 | 0.370 | 0 | 1 |
In this Appendix, rather than using principal components, we present the estimates based on answers to single questions in the WVS. We present both OLS and Tobit regressions (Table
Results for group lending with interaction terms (based on single questions).
Variables | OLS | Tobit | ||||||
(1) | (2) | (3) | (4) | (5) | (6) | |||
Contract enforcement | 0.680 (0.200) | 0.835 (0.183) | 0.540 (0.238) | 0.800 (0.119) | 0.924* (0.081) | 0.637* (0.092) | ||
General trust | 0.989*** (0.004) | 0.999*** (0.005) | 0.847*** (0.005) | 1.194*** (0.001) | 1.195*** (0.000) | 1.035*** (0.000) | ||
Contract enforcement × General trust | –3.960*** (0.001) | –3.907*** (0.003) | –3.265*** (0.002) | –4.881*** (0.000) | –4.872*** (0.000) | –4.178*** (0.000) | ||
Integrity | 0.557* (0.059) | 0.638** (0.026) | 0.328 (0.194) | 0.821*** (0.008) | 0.955*** (0.001) | 0.601** (0.021) | ||
Contract enforcement × Integrity | –1.988*** (0.009) | –2.020** (0.013) | –1.443** (0.023) | –2.997*** (0.001) | –3.091*** (0.000) | –2.385*** (0.000) | ||
Narrow trust | –0.410* (0.076) | –0.577** (0.015) | –0.418* (0.078) | –0.653*** (0.007) | –0.843*** (0.000) | –0.653*** (0.003) | ||
Contract enforcement × Narrow trust | 2.659*** (0.000) | 2.549*** (0.001) | 2.092*** (0.002) | 3.803*** (0.000) | 3.686*** (0.000) | 3.108*** (0.000) | ||
Logarithm of per capita GDP | –0.117** (0.015) | –0.075* (0.092) | –0.153*** (0.002) | –0.105** (0.020) | ||||
Gini coefficient | –0.002 (0.602) | –0.005 (0.263) | –0.002 (0.694) | –0.005 (0.274) | ||||
Financial development | 0.315 (0.311) | 0.141 (0.577) | 0.592* (0.081) | 0.383 (0.152) | ||||
UK legal origin | 0.271*** (0.002) | 0.267*** (0.001) | ||||||
Constant | –0.086 (0.573) | 0.952** (0.049) | 0.863* (0.053) | –0.147 (0.316) | 1.112** (0.019) | 0.987** (0.022) |
In this Appendix, we present the results of regressions without interaction terms using both the principal components (Table
Group = β 0 + β1 CE + β2 GT + β3 IN + β4 NT + γX + ε. (E1)
As noted in the main text, the coefficients of the institutional and social capital variables are not statistically significant at 5% confidence level.
Estimation results for the principal components data (no interaction terms).
Variables | OLS | Tobit | ||||||
(1) | (2) | (3) | (4) | (5) | (6) | |||
Contract enforcement | –0.123 (0.558) | 0.002 (0.994) | –0.063 (0.698) | –0.184 (0.419) | –0.112 (0.582) | –0.164 (0.337) | ||
General trust | –0.248 (0.248) | –0.362 (0.103) | –0.328 (0.107) | –0.317 (0.196) | –0.439* (0.072) | –0.397* (0.063) | ||
Integrity | –0.086 (0.701) | –0.071 (0.711) | –0.078 (0.649) | –0.097 (0.723) | –0.020 (0.930) | –0.044 (0.824) | ||
Narrow trust | 0.440 (0.142) | 0.357 (0.232) | 0.359 (0.180) | 0.519 (0.146) | 0.431 (0.203) | 0.427 (0.145) | ||
Logarithm of per capita GDP | –0.168*** (0.001) | –0.115*** (0.010) | –0.207*** (0.000) | –0.146*** (0.004) | ||||
Gini coefficient | –0.002 (0.628) | –0.005 (0.302) | –0.001 (0.902) | –0.004 (0.461) | ||||
Financial development | 0.473 (0.172) | 0.250 (0.323) | 0.711* (0.084) | 0.445 (0.125) | ||||
UK legal origin | 0.327*** (0.002) | 0.341*** (0.001) | ||||||
Constant | 0.217* (0.084) | 1.717*** (0.000) | 1.379*** (0.001) | 0.218 (0.107) | 1.897*** (0.000) | 1.529*** (0.000) | ||
R-squared | 0.066 | 0.241 | 0.461 |
Variables | OLS | Tobit | ||||||
(1) | (2) | (3) | (4) | (5) | (6) | |||
Contract enforcement | –0.118 (0.589) | –0.072 (0.728) | –0.089 (0.595) | –0.175 (0.455) | –0.186 (0.397) | –0.188 (0.286) | ||
General trust | –0.212 (0.278) | –0.201 (0.388) | –0.136 (0.481) | –0.263 (0.211) | –0.266 (0.272) | –0.191 (0.323) | ||
Integrity | –0.228 (0.300) | –0.137 (0.546) | –0.261 (0.141) | –0.296 (0.228) | –0.171 (0.466) | –0.303 (0.104) | ||
Narrow trust | 0.466** (0.027) | 0.268 (0.295) | 0.262 (0.230) | 0.552** (0.023) | 0.336 (0.233) | 0.324 (0.164) | ||
Logarithm of per capita GDP | –0.126** (0.015) | –0.071 (0.169) | –0.154*** (0.008) | –0.092* (0.092) | ||||
Gini coefficient | –0.002 (0.694) | –0.005 (0.302) | –0.001 (0.846) | –0.005 (0.359) | ||||
Financial development | 0.458 (0.222) | 0.188 (0.477) | 0.688 (0.121) | 0.374 (0.217) | ||||
UK legal origin | 0.347*** (0.001) | 0.363*** (0.001) | ||||||
Constant | 0.251** (0.020) | 1.331*** (0.007) | 1.095** (0.038) | 0.266** (0.020) | 1.499*** (0.006) | 1.230** (0.027) | ||
R-squared | 0.131 | 0.213 | 0.449 |