Experimental Design
1650 slum dwellers from the city of Pune, India, will be enrolled in the trial. 825 participants will be randomly allocated to the treatment group and 825 participants to the control group. Randomization will be stratified by participant sex, income, and present bias. Eligibility criteria are a) being 18 years and older, b) having some income (employment or other) at least once per week or on a monthly basis, and c) holding at least some decision making power over how money is spent. The study aims for a roughly equal share of female and male participants.
***
Additional study component: Trust game
Design and procedures:
Spousal trust and trustworthiness were quantified by a simplified binary trust game, in which spouses were randomly selected to act as the first and second mover, respectively (see Figure 1 below). The first mover was asked to decide between two payout options, namely (A) relying on the spouse’s decision to determine the final payout and accepting a potential loss in case the spouse will choose to default (proxy for trust), or (B) defining the final payout without relying on the spouse’s decision and thus a priori accepting a lower final payout (proxy for mistrust). If the first mover opted for decision (A), the second mover determined the final payout by choosing (A) whether to cooperate (proxy for trustworthiness) and receive equal payouts or (B) whether to default and thereby maximize the individual payout (proxy for untrustworthiness).
Since the trust game was a sequential game and was only played once, each spouse only made one choice (single-role game). Thus, for every spouse, the game yields one decision, either in terms of trust or trustworthiness, depending on whether the spouse was randomly selected to be the first or second mover, respectively. Choices were kept private such that both spouses made their decisions without knowledge about the other movers’ choice.
Payouts were selected to reflect gender-specific preferences, namely embroidered handkerchiefs for female subjects and plain handkerchiefs for male subjects. The payout amounts varied in the number of handkerchiefs to be received in the different scenarios determined through spouses’ choices and were chosen as such to maintain an incentive to maximize the own payout, i.e. by ensuring a sufficient gain in utility of receiving an additional handkerchief. The selection of gender-specific payouts was crucial in order to adequately measure individual trust and trustworthiness between spouses. Interchangeable payouts would be valued equally by both spouses and might thus motivate individuals to maximize payouts differently, e.g., according to a joint household payout. Moreover, gender-specific (as opposed to interchangeable) payoffs may reduce potential reciprocal or reputational effects of conducting a trust game between non-anonymous, cohabiting players.
The team of enumerators that administered this study component was deliberately different from the one that conducted the main survey data collection. To internalize the rules and procedures of the trust game and to discuss potential difficulties during its conduction in the field, the team underwent an extensive three-day training. For the implementation of the trust game during home visits, we assigned two enumerators per household instead of one in order to ensure a clean implementation. This assured the isolation of spouses from each other during their decision-making processes and choice determination in the trust game as well as a detailed presentation of the choices and payouts.
Constructed variables:
The data from the trust game was used to create two types of variables, which will be used in the empirical analysis to detect heterogeneity in treatment effects:
1. Individual choices in trust game:
This refers to the wife’s and the husband’s individual decisions to cooperate/deflect in the trust game (regardless of what the respective other mover chose). The corresponding variables indicate spouses’ trust and trustworthiness (each coded as 1 if a person cooperated, and as 0 if a person deflected).
2. Alignment of choices in trust game:
We also account for whether the decision of one mover in the trust game was aligned with the other mover’s decision. We created two variables to quantify the alignment of choices in the trust game: First, a binary variable, coded as 1 if choices were aligned (i.e., both cooperated or both did not cooperate) and as 0 if they were misaligned. Second, a categorical variable that also specifies the type of alignment, i.e., that distinguishes between positive alignment (both cooperated), negative alignment (both did not cooperate) and no alignment.