Experimental Design
The study is carried out in collaboration with the Indian NGO Basix Sub-k. It builds on the recent expansion of the business correspondent model (BCM) in India. In collaboration with formal banks, Basix Sub-k is opening bank accounts in different States over the country. In rural areas, the NGO selects one villager to become the local banker (the business correspondent sub agent or BCSA). He receives training, a receipt machine, a finger print recognition device and a mobile phone connected through the mobile network to the partner bank. The BCSA can then use those devices to perform standard transactions on their accounts: deposits, withdrawals and transfers.
In villages where Basix Sub-k has recently opened BCSA accounts, we do weekly interviews during 7 to 13 weeks, to gather detailed information about the evolution of their household composition and the various earnings and expenditures of their household members. Because those surveys are extremely demanding, each participant is offered Rs150 after each interview.
Phase 1
To test the above hypotheses we designed the study as follows. We operate in an area with recent BCSAs. We select a random sample of people who opened an account (group A) and a random sample of people who did not. That second sample (of people without account) is subject to the first treatment: we give an account to two third of them. After this first treatment, we are left with three groups: A – already had an account, B – received a new account, and C – did not have an account and is not offered one.
The second treatment in the experiment is a randomization of the way the weekly compensation is paid. Half of the respondents with a BCSA account receive Rs150 directly on their account (treated), while the others receive it in cash (control).
We now have 5 different groups: Ac had an account and is paid cash, Aa had an account and is paid on it, Bc just received an account and is paid cash, Ba just received an account and is paid on it, C does not have an account and is paid cash. The comparison between these five groups will provide clear information on the impacts of the different steps of the financial inclusion process.
The first randomization, providing bank accounts to people who initially did not self-select into the new banking system, has two main goals:
- Measuring the first step of the process: having a bank account versus not having one
- Evaluating the importance of the initial self-selection. Existing studies suffer from the caveat that they only observe impacts on the pool of self-selected individuals (corresponding to our group A). Those impacts may not be representative of what a large scale financial inclusion plan (with accounts opened for everyone) would achieve.
The second randomization (cash versus account) aims at measuring the second step of the financial inclusion process: going from having a bank account but receiving cash payments, to having a bank account and being paid on that account. Importantly, given that (i) all respondents have an account (in groups A and B); (ii) there is no direct cost in depositing or withdrawing; (iii) all respondents in a village receive their compensation at the same location, close to the BCSA; we can attribute the treatment effect to the ‘default option’ described above.
Phase 2
One month after the last week of interviews of phase 1, we will re-start the weekly surveys for an additional period of four weeks. During these weeks, we will compensate everyone in cash as to test whether the initial treatments have impacts that last in the medium term.