Evaluating the Impact of Commitment Savings Accounts Linked to Mobile Money

The study will be conducted in two phases: (i) a pilot and (ii) a large-scale study. During phase 1, we will inform people of the various mobile savings products, and elicit their interest in each of these. Then we will randomly assign each of these people to one of the mobile savings products described below, and collect data on product takeup and use. Phase 1 will involve four different products. The phase 2 randomized controlled trial may involve a sub-set of these, conditional on the phase 1 results, as well as power calculations, sample size, and budgetary considerations.

The four products are:
T1: A basic m-savings account, which is fully liquid and accessible via the mobile money ecosystem.
T2: An enhanced version of the basic m-savings account where the user can specify automatic transfers from mobile money float to the m-savings account.
T3: An account that also adds soft commitment, in the form of changeable liquidity restrictions.
T4: An account that adds hard commitment, in the form of an irreversible liquidity restriction.

In phase 2, we will also randomly vary whether a study participants spouse receives an information treatment, in which he or she is informed of the terms and conditions of the account that his or her spouse has just taken up.

2017-07-01

2019-12-31

Savings behavior; commitment account settings and targets; goal attainment; investment; expenditures; transfers of money; income; intra-household decision-making power over savings; usage of financial products and services

Savings behavior: deposits and withdrawals into mobile money and mobile-money based savings accounts, other cash savings, and non-cash savings, and how savings allocation across different places shifts in response to treatment.

Commitment account settings and targets: deposit targets setting (in terms of amount and frequency), withdrawal restrictions setting, the extent to which the soft commitment on the deposit side is broken due to lack of funds, and the extent to which the soft commitment on the withdrawal side is broken by the participant. Interactions between targets set and subsequent savings behavior. Evolution of people’s target setting, and the extent to which their behavior with respect to violating targets and breaking soft commitments evolves over time.

Goal attainment: here we will link baseline and endline data to measure the extent to which a respondent achieves his or her savings goals (e.g. investment, purchase of asset, paying school fees, etc.).

Investment: purchases of productive and non-productive assets, as well as investment in human capital (e.g. children’s education).

Expenditures: here we will measure outlays by study participants on themselves, spouses, other family members, and non-family members, both within and outside the household.

Transfers of money from study participants to spouses, other family members, and non-family members. We will also measure transfers received by study participants from others.

Income by source, including wage labor, self-employment income, and agricultural income.

Intra-household decision-making power over savings.

Usage of financial products and services: e.g. debt, bank loans, and other financial products for which bank account ownership could serve as a gateway.

Our key research questions are:
- How does setting a default amount to save regularly, with the assistance of auto-deposit, impact key outcomes?
- What is the impact of allowing individuals to specify self-imposed, non-binding withdrawal restrictions on their mobile-money savings?
- What is the impact of requiring individuals to specify self-imposed binding withdrawal restrictions on their mobile-money savings?
- How does disclosing the degree of mobile-money savings account liquidity to an account-holders spouse impact key outcomes?

The study will be conducted in two phases: (i) a pilot and (ii) a large-scale study. During phase 1, we will inform people of the various mobile savings products, and elicit their interest in each of these. Then we will randomly assign each of these people to one of the mobile savings products described below, and collect data on product takeup and use. Phase 1 will involve four different products. The phase 2 randomized controlled trial may involve a sub-set of these, conditional on the phase 1 results.

THEORY OF CHANGE

The first central assumption in our theory of change is that individuals in our study area have an unmet need for secure mobile-based savings services, e.g. because their current methods of saving are costly, inconvenient, or insecure. This assumption is supported by existing literature and our own scoping work: nearly half of individuals we interviewed during our scoping work stated that they would be willing to pay for a mobile-linked savings account with NVRB. Moreover 63 percent of NVRB customers participating in a deposit-collection pilot requested mobile banking when asked what services they would like NVRB to offer going forward. A second critical assumption is that individuals in our catchment areas have access to mobile phones, that the mobile money network is reliable, that mobile money services are appropriately priced, and that individuals are able to navigate the mobile money system. We believe that these are such that the basic requirements for the study are met.


Offering individuals a basic NVRB mobile savings account (T1) may help ease savings constraints via a mental accounting/labeling effect: the basic m-savings account does not offer any amenities not found on an MTN mobile money account. However, individuals may find it easier to keep savings in a separate account earmarked specifically for their savings goal. The mobile savings plus auto deposit account (T2) may further increase savings rates if automatic withholding increases the flow of funds into the m-savings account and if individuals do not consider resources in their m-savings and mobile money accounts to be fully fungible. Adding a soft withdrawal restriction to the m-savings account (T3) could help increase savings balances if (a) making the commitment increases the perceived non-fungibility between the m-savings and mobile money accounts or (b) being able to claim that funds are locked away helps deter requests from members of an individual’s social network. Making the withdrawal restriction binding (T4) will have an ambiguous effect on deposits: individuals who understand that they are present biased may prefer to save more in such an account to force their future selves to spend money responsibly. Similarly, the binding withdrawal restriction could be useful for spouse or other-control purposes. Conversely, the hard withdrawal restriction could make the account less attractive to individuals who wish to access savings in the event of a negative economic shock.
We hypothesize that the spousal information treatment will have different impacts depending on the banking treatment arm: the information arm should make individuals with fully liquid accounts (T1-T3) more vulnerable to transfer requests from the spouse, while it should either have no effect or make it easier for individuals with fully committed accounts (T4) to refuse transfer requests from their spouse.


IDENTIFICATION STRATEGY
We will begin by conducting public outreach sessions in communities surrounding five NVRB branches. The outreach sessions will address in general terms the features of the mobile-savings accounts, without providing specific details on each of the four different product variants, and will inform individuals that if they express interest, they will be entered in a lottery to be offered one of the mobile savings products. Only individuals who meet the eligibility criteria for inclusion in the study that are described in section 2.1 will be included in the study. We will enroll 3,000 eligible individuals in the study sample. Following a baseline interview, individuals in the study sample will be randomly assigned, at the individual level, to be offered one of the mobile savings products, and he or she will either take-up that product (by completing any necessary procedures to open the mobile-savings account), or not.
Individuals will be randomly assigned to one of T1-T4 with equal probability, and to either I0 or I1 with equal probability. The randomization will be stratified on gender, bank branch, whether or not the individual is a pre-existing MTN mobile money customer, and whether or not the individual has a pre-existing account with a formal bank (these variables will be collected on the expression of interest form). Since we have a total of 8 experimental cells, each cell will include 375 individuals.


EXPERIMENTAL DESIGN
For each treatment condition, we have listed the behavioral savings channels that we hypothesize may be at play - e.g. for private information plus basic M-savings, we hypothesize that the accounts may change savings behavior because they offer individuals goal setting, planning, other control, mental accounting, and spouse control benefits (note that goal setting, planning, and other control are listed as “common” assets to all accounts).

in the below,
T1 = basic m-savings
T2 = basic m-savings + auto-deposit
T3 = auto-deposit + soft commitment
T4 = auto-deposit + hard commitment
I0 = private information
I1 = public information

Behavioral savings channels that we hypothesize may be at play inside each cell. Note, (+)/(++) indicates a strengthening of an already-existing benefit. "M.A." stands for mental accounting.
T1 + I0: Spouse control, M.A.
T2 + I0: Spouse control, M.A., M.A. x self-control
T3 + I0: Other control (+), spouse control (+), M.A., M.A. x self-control
T4 + I0: Other control (++), spouse control (++), M.A., M.A. x self-control, self control, illiquidity
T1 + I1: M.A.
T2 + I1: M.A., M.A. x self-control
T3 + I1: Other control (+), M.A., M.A. x self-control
T4 + I1: Other control (++), spouse control (++), M.A., M.A. x self-control, self control, illiquidity

Randomization done in office by a computer.

Individual person level randomization.

No

N/A

The phase 1 study sample is planned to be approximately 300 individuals. The phase 2 study sample is tentatively planned to include 3,000 individuals, although we expect to adjust this based on phase 1 findings.

For phase 2, we anticipate a total of 8 experimental cells. We plan to test four different savings products (or product variants), each described above under "intervention". Each variant will be offered to 750 people. This will create 4 treatment cells. And within each treatment cell, we will also randomly vary whether a study participants spouse receives an information treatment (with half receiving, and half not). This will thus result in a total of 8 treatment cells, each with 375 individuals.

Note: the final sample size and study design will make use of updated power calculations based on phase 1 results. Tentatively, our power calculations are as follows: For cell-to-cell comparisons between our 8 treatment cells, we have targeted a standard minimum detectable effect size of 0.25, assuming 85 percent take-up. For some specifications we will be able to pool across the two information treatment arms, thus reducing the total number of cells to 4, and our power for these is 0.18 SMDE, again assuming 85 percent take-up. To put these numbers in context, suppose 35 percent of the control group meets their savings goal. A 0.18 SMDE translates to an 8.6 percentage point increase in goal attainment and a 0.25 SMDE translates into a 0.12 percentage point increase in goal attainment.We believe our take-up assumption of 85 percent is justified because it is conditional on someone attending an information session, receiving general information regarding NVRBs mobile-savings accounts, and then expressing interest in opening one of the accounts. Past experience gives us reason to believe that take-up conditional on a formal expression of interest will be high. However, we will validate our take-up assumptions during phase 1, and will adjust the phase 2 sample size and study design accordingly.