Experimental Design Details
We use a microfinance repayment game in a lab-in-the-field setting to analyze the enforcement problem (ex-post moral hazard) under risk. We exogenously vary the liability structure and the availability of repayment flexibility in a 2 × 2 design: individual vs joint liability and flexibility vs no flexibility.
The standard game (individual liability and no flexibility, IL) models a simple credit repayment choice under risk over three periods. We abstract from the investment choice and the presence of any savings technology and assume the investment yields a steady income y=2 per period. In each period, the bank requires a repayment R=1, and clients face the choice of repaying or using the entire income for consumption. Individual idiosyncratic shocks occur with probability 0.25 and destroy the entire income of a period, thus leading to non-repayment. Clients make their repayment choice conditional on being able to make the repayment installment. When they suffer an income shock, they can neither repay nor consume in that period. Following the first non-repayment, whether by choice or by bad luck, clients are in default for the rest of the game. If the client repays in all three periods, she receives the continuation value V, a monetary reward which symbolizes the value of the future relationship with the bank. As this value only materializes in the future, V is paid one month after the experiment. In contrast, all experimental income not spent on repayment or lost to the shock can be spent right after the experiment on a vast selection of consumption items – resembling the temptation of immediate consumption. We artificially induce temporal discounting by reducing the consumption value of income tokens from round to round.
We model joint liability (JL) as a two-person borrowing group that is jointly responsible for repaying 2R in each period. Joint liability is automatically enforced in case of non-repayment of any member of the borrowing group. While the reduction of the usual five-person group to two persons is a simplification of reality, automatic enforcement is a realistic representation of how microfinance institutions put joint liability into practice. The repayment choice then becomes a coordination game: Clients simultaneously choose whether to repay or not. If they choose to repay, but their partner does not, they automatically repay for their partner as well. To keep the experiment simple, we only elicit unconditional repayment choices and the belief about the partner’s repayment choice. The bank does not distinguish between the source of repayment: as long as R=2 in each period, both clients will receive V. A measure of peer pressure and punishment is introduced via the possibility to send “dislike” messages to one’s partner, conditional on their behavior. Punishment has a cost (very small in magnitude relative to the other incentives) to the sender, and a 1:3 cost to the receiver. Shock outcomes and repayment decisions are perfectly observable when making punishment choices.
In the first set of experiments, conducted in March/April 2016, repayment flexibility is the option to defer one repayment to the next period. This sets the repayment obligation for the current period to 0, but requires a double repayment in the subsequent period. In this way, default in the current period is prevented for sure. Flexibility is only available in periods 1 and 2, and can only be exercised once. This is implemented using one flexibility token that can be used in either period 1 or 2, resulting in three possible realizations of flexibility: First, the borrower does not use flexibility at all . Second, the borrower uses flexibility in the first period and is then required to make up her missed repayment by a double repayment in the second period. Third, the borrower uses flexibility in the second period and is then required to make up her missed repayment by a double repayment in the third period. Failure to make double repayments results in the loss of V, as do shocks once the the flexibility token has been used up. While the provision of flexibility is intended to allow the borrower to self-insure her repayment against the idiosyncratic shock, it can also be used to increase early consumption (i.e. absent shocks). Early consumption is attractive because future income is discounted. We will refer to this as a misuse of flexibility, since the insurance value of flexibility is traded for early consumption.
We examine the combination of joint-liability and flexibility in JL-flex. In a two-person borrowing group, both partners have one flexibility token and can defer one repayment obligation to the next period in any of the first two periods. In our setting, this has two important implications. First, when a borrower uses the flexibility token, she is unable to insure her partner’s repayment obligation in the same and in the next period. By using the flexibility token, the repayment obligation of the borrower is reduced to zero and no repayment can be made by this borrower in this period, not even to cover her partner’s repayment. In the next period, the borrower is required to make a double repayment, and hence use her entire income for her own loan repayment, which again leaves no scope for insuring her partner. In addition, if she faced a shock when the double repayment is due, her partner would be unable to insure her, since group repayment obligation (R=3) exceeds group income (y=2) in this case.. The use of flexibility in our setting does not allow for mutual insurance in the following period, which results in a substitution of mutual group insurance by individual self-insurance. These very strong assumptions are necessary to simplify the design and make it understandable and practicable. As before in the JL treatment, the possibility to punish repayment choices as well as the use of flexibility is known. All punishment decisions are made under full information, i.e. clients know whether her partner defaulted voluntarily.
Subject to receiving funding, we tentatively plan to conduct a second set of experiments, with a new type of flexibility: A “repayment pause”, which works without double repayments. A pre-specified number of repayment breaks are calculated into the installments. A real-life analogue would be a repayment schedule of 48 installments over 50 weeks, where the borrower can spontaneously decide in which two weeks to take a break. This allows the borrower to respond to shocks, while at the same time avoiding the burden of double installments. A downside is the pre-commitment to a fixed number of breaks, which is necessary to calculate installments. The planned second set of experiments allows us to test the key hypotheses of our study absent the discussed crowd-out between mutual insurance and flexibility. We will implement this in the same lab-in-the-field setting as the first set of experiments. To remove the crowd-out mechanism, subjects will receive y=3 per period, with three repayments of R=1 in the no-flex conditions, and two repayments of R=1.5 in the flex conditions. The key difference to studies varying the frequency of repayment is that borrowers choose themselves in which period to take a break. This allows them to respond to shocks, while maintaining a high-frequency schedule in times of stable income flows.
We use a mixture of within- and between-subjects design. All decisions are elicited by pen and paper using the strategy method throughout the experiment. We randomize treatment allocation on session level, and conduct sessions consisting of IL and IL-flex, as well as sessions consisting of IL, JL and JL-flex. Due to the natural order of these treatments and to keep the experiment as simple as possible, the order of treatments stayed the same in all sessions. At the end, we randomly select one treatment to be paid out, and realize the shocks (and where applicable the group matching, without revealing partners’ identities). To provide a behavioural measure of time-inconsistent preferences, participants are allowed to revise their choices after the treatment selection, but before the shock realizations.