A little skin in the game: reducing moral hazard in joint liability lending through a mandatory collateral requirement∗ Jon Einar Flatnes,y Michael R. Carterz 2/5/2018 Abstract Both heavily collateralized individual loan contracts and joint liability group lending con- tracts have received much attention in the microfinance literature, yet each has been shown to have important limitations. In this paper, we ask whether a contract that combines joint liabi- lity with a collateral requirement could be optimal relative to an uncollaterlized joint liability contract. Specifically, we show theoretically that adding a collateral requirement to joint liabi- lity reduces moral hazard and has an ambiguous effect on credit market participation. Using a framed field experiment among active credit group members in Tanzania, we demonstrate that a modest (20%) collateral requirement reduces moral hazard (effort diversion), and that this behavioral change reduces defaults by 15-20% depending on the level of social connectedness. Moreover, while the fraction of the population willing to borrow decreases by 7% overall, we find that it increases for group members who are highly socially connected. Keywords: Credit Rationing, Moral Hazard, Collateral, Field Experiments, Joint Liability JEL Classification Codes: D82, G21, O16, Q14 ∗We would like to thank Travis Lybbert, Stephen Boucher, Ghada Elabed, Jean Paul Petraud, Isabel Call, Wenbo Zou, Emilia Tjernstrom, Thomas Barre, Eliana Zeballos, Patrick McLaughlin, Jacob Humber, Asare Twum Barima, Abbie Turianski, and an anonymous referee for their insightful comments and feedback. We would also like to thank VisionFund Tanzania, the enumerators, community leaders and participants in the experiment for their time and service. This research was made possible, in part, through funding provided by the U.S. Agency for International Development through the BASIS Assets and Market Access Collaborative Research Support Program. yAssistant Professor, Department of Agricultural Environmental, and Development Economics, The Ohio State University, Columbus, OH 43210; email: fl[email protected] zProfessor, Department of Agricultural and Resource Economics, University of California, Davis, One Shields Avenue, Davis, CA 95616; email: [email protected] 1 1 Introduction Over the past three decades, microfinance has eased liquidity constraints for millions of low-wealth people around the developing world.1 In making loans to low-wealth borrowers, microfinance lenders face not inconsequential problems of moral hazard and adverse selection given that small loan sizes and fixed information costs make it expensive to select and monitor borrowers. To combat these challenges, lenders typically rely on either conventional collateral mechanisms or on joint liability loan structures, with the latter having driven the so-called microfinance revolution. Both devices have their limitations. Collateral requirements ration out borrowers who lack the assets required to qualify for loans. Even individuals with sufficient assets may choose not to put their collateral at risk and hence fail to borrow (a phenomenon Boucher et al. (2008) label risk rationing). In contrast, joint liability loan contracts are designed to harness the social collateral among community members, thus managing default risk for lenders typically without requiring conventional collateral. At the same time, joint liability lending exposes borrowers to a new source of risk, namely the morally hazardous behavior of others in their borrowing group for whom they are financially liable. While much of the theoretical economics literature has considered individual collateral and joint liability in isolation from each other, there is a continuum of contracts that offer different mixes of individual collateral and joint liability. As such, individual liability with collateral or full joint- liability with no collateral can be viewed as corner-solution contracts, and it is not obvious that these contracts are optimal relative to contracts that combine collateral with joint liability. In an effort to more fully explore the contract continuum, in this paper, we develop a theoretical model to explore the interactions between joint liability and individual collateral, focusing on how different interior contracts along that continuum perform with respect to moral hazard, loan repayment and credit rationing. Intuitively, while collateral has an obvious incentive effect under individual liability by increasing the cost of failure, it also has an additional effect under joint liability in that it makes the successful state more lucrative, since the collateral serves as partial insurance against having to cover other group members' losses. As such, collateral may act both as a stick and as a carrot under joint liability. Moreover, while collateral is unambiguously a burden for borrowers under individual liability, adding collateral to a joint liability contact may actually increase the utility of safe borrowers by reducing the expected amount that the group will have to cover for defaulting members.2 In particular, our model shows that collateral has an ambiguous impact on credit market participation under joint liability, increasing it for socially connected people, and decreasing it for risk-averse people with low social collateral. To test the predictions of the theoretical model, we conducted a framed field experiment, in which a sample of Tanzanian smallholder sunflower 1The impacts of microfinance are subject to debate. For example, see the discussion in Banerjee et al. (2015). 2This assumes that interest rates do not adjust in response to the addition of collateral. If interest rates are endogenous, it is possible that certain borrowers under individual liability prefer a collateralized contract over a non-collateralized one. In our model, we explore both the case of exogenous and endogenous interest rates. 2 farmers already enrolled in joint liability groups played a series of microfinance games, similar to those conducted in Gin´eet al (2009). The results of our experiments show that adding a collateral requirement to a joint liability loan reduces moral hazard (effort diversion), and that this behavioral change by itself reduces defaults by almost 20% when social collateral is modest. Also consistent with the theory, the experimental findings show an ambiguous impact of the collateral requirement on credit market participation, increasing it for highly socially connected people, but on average, reducing the fraction of the population willing to borrow by 7%. To the best of our knowledge, this is the first paper to explicitly model the effect of combining joint liability with a collateral requirement. It is also the first paper to experimentally test the effect of adding collateral to an existing joint liability contract. Rooted in the Grameen Bank model developed by Nobel Peace Prize laureate Muhammad Yunus, joint liability group lending has received much attention in the economics literature. Under a typical group liability contract, borrowers form groups consisting of five to twenty members, who receive individual loans but are all mutually responsible for the total repayment to the lender. Proponents of group liability contracts argue that this feature can help overcome problems of moral hazard and adverse selection in small-scale lending where lenders lack credit information about potential borrowers and do not have the resources to monitor individuals. In particular, positive assortative matching in group formation can reduce informational asymmetries (Ghatak, 1999; van Tassel, 1999), and strong social collateral within existing groups might help reduce voluntary default and other morally hazardous behavior (Besley and Coate, 1995; Anderson and Nina, 1998). Despite these potential benefits of group liability, there is mixed evidence, both theoretical and empirical, on the relative merits of group versus individual liability (Morduch, 1999; Huppi and Feder, 1990; Gin´eand Karlan, 2014). Critics of group liability argue that such contracts are prone to free-riding and collusion (Gin´eet al., 2009; Laffont and Rey, 2003), that they deter relatively safe borrowers who would not default under an individual contract but would do so under a group lending contract (Besley and Coate, 1995), and that they may induce excessively prudent behavior (Banerjee et al., 1992). Moreover, positive assortative matching may only work partially, especially if group sizes are large. Finally, social hierarchies within groups or strong social ties among group members may distort or weaken the value of the social collateral, as successful borrowers might be reluctant to pressuring less fortunate, less abled, or more powerful group members to repay (Wydick, 1999; Gin´eand Karlan, 2014). Limitations to conventional individual loan contracts are also well-studied. Stiglitz and Weiss (1981) demonstrate in their seminal paper, that a lender's inability to distinguish between different borrower types and to enforce and monitor borrower behavior results in adverse selection and moral hazard, respectively. To combat these issues, a profit-maximizing lender will set an interest rate lower than the market-clearing rate, thus leading to credit rationing. In response to these findings, Bester (1985a) extends the Stiglitz and Weiss model and shows that when banks can 3 set the interest rate and the collateral requirement simultaneously, perfect sorting of borrowers is possible, thus eliminating adverse selection and thereby credit