Incentivizing Savings

“Incentivizing Savings” is a randomized impact evaluation that is currently taking place in the Metropolitan Region, Chile, executed by J-PAL LAC, institution based in the Pontificia Universidad Católica de Chile, in partnership with BancoEstado, a state-owned, autonomous financial institution. The investigation aims to determine and compare the effect of different saving incentives, as saving reminders through cell phone text messages (SMS) and automatic savings plans, in variables like financial inclusion and general well-being of the individuals. A more specific description of the treatments is presented below:
• Treatment 1: Automatic savings plan. This treatment consists in the offering of an automatic savings program.
• Treatment 2: Reminder text messages. This group receives one text message per month, recalling their savings target mentioned in the baseline survey.
• Treatment 3: Short Training. A four-part treatment, it begins with a 3 minutes animated video with strategies that could eventually help people to lower their consumption in Temptation goods. Second, a pack is given to those assigned to this treatment with a magnet and a card holder with the logo of the project as "reminders". Third, a book that insists in the strategies is given to the participants assigned to this treatment. To increase the retention of the books, they are made as calendars. And fourth, SMSs is sent every month with the strategies mentioned in the video.
• Control group: People in this group just open a savings account.
The sample was arranged with BancoEstado, an institution with a strong presence in the whole territory, and consists of BancoEstado clients (over 18 years old) recruited in the branches when the Bank executives inform them about the investigation and then offer them to participate. The branches have been chosen considering their vulnerability; the poorer municipalities of the Metropolitan Region. The number of people being considered is 6,000; 1,800 in the Control group, 1,800 in Treatment 1, 1,200 in Treatment 2 and 1,200 in Treatment 3.
As mentioned, the participants are recruited in the branches once they approach the Bank executives with the intention of opening a saving account. Once the clients mention their intention to open a savings account, the executive offers them to participate. If they agree, the executives must handle them the written consent, wait for them to sign it and then handle the baseline survey, which is performed in a tablet device provided by the research team that includes also a presentation video. The randomization and subsequent treatment assignment is made by the same survey software using the last digit of the RUT number (the National Identification Number that is correlated only with the order of birth registration in a short period of time and not related with any socioeconomic or personal characteristic) that was filled as the first question of the survey.

2015-08-30

2016-11-30

To measure those outcomes this evaluation will use two main sources of data: (1) administrative data from the bank and (2) surveys coordinated and executed by the research team or an associated organization.
1. Administrative data: During the intervention BancoEstado will weekly send the research team checking and savings accounts information (balances) of all participants, together with banking (number of accounts, date of opening, state and characteristics the accounts) and socioeconomic information that the bank independently collects (this information will be cross-checked with the baseline survey information). Starting in the intervention onwards, every three months, the Bank will also send complete information about the movements of all of the participant’s accounts. This will allow to stufy the effects of the treatments in:
• Saving and checking account balances
• Financial behaviour
• Financial products adquired
2. Surveys: A baseline survey will be conducted in a tablet device at the moment of the offering. A follow up, household survey will be eventually conducted after one year (funds are still being applied for its execution). A survey firm will be hired to collect that data. The questionnaire will include the following modules:
• Household member characteristics (gender, age, education)
• Employment and income
• Entrepreneurship history
• Measures of cognitive abilities (such as a digit-span test or a version of a Raven test)
• Savings and credit behavior
• Household financial behavior
• Children’s behaviors (such as labor or school attendance)
• Female empowerment and self-esteem

The female empowerment module in the questionnaire addresses both psychological and financial empowerment. Questions regarding the following variables will be included:
Individual characteristics:
• Autonomy: Includes initiative to do things, decision making, proactive, how people face new situations.
• Auto-efficacy: Personal skills related with self-confidence, how to solve problems, make plans and perseverance.
• Locus of control: How people control their own life, for instance the reasons behind success in reaching goals: luck, chances, effort
Financial competences:
• Financial system´s knowledge: People familiar with banks, how banks or financial institutions work, their functions, financial products, etc.
• Financial literacy: Knowledge of interest rate, inflation, fees, and product´s conditions, etc.
• Financial management: How people feel when go to a bank, Capable? Comfortable? Confident?
Other variables:
• Social support: Relationships, confidence on friends, family, people; variables that are related with asking for help.
• Family ambient: Information regarding how family members interact. Discussions, fights, etc.

Regarding the identification strategy, it relies on the random assignment of each eligible individual to a treatment or control group, which guarantees that individuals in each group are, on average, the same. We will use administrative and survey data to test the balance among groups and the effectiveness of the randomization.

One of the most important stages of the evaluation is the offering process, which will take 3 months in around 25 BancoEstado branches, because it is the only instance (before the follow up survey) when a personal contact with the participants takes place. The participants are offered to participate when they come to the branch to open a savings account, by the same executive in charge of opening it. If the client agrees to participate in the investigation, the executive must provide the consent and the baseline survey (in the tablet device). Once these documents are completed, the executive, with a hidden button in the tablet, must check whether the client should be assigned to one of the "in situ" treatments, offering an automatic savings program or showing the educational video. If they are not assigned to any of these treatments, the executive should proceed to open the account desired by the customer.

The equations for this stage is:
Second Stage
Yij = b0 + b1*T1 + b2*T2 + b3*T3 + a1*Xij + uit
Where Yij is an outcome variable (e.g. amount saved), T1 to T3 are indicators of the treatment status, and Xij is a set of characteristics and dummy variables used in the stratification process. The coefficients b1, b2 and b3 will provide the program effect (as it is currently implemented and with its extension) on the outcome of interest and provide information on the intention to treat. Coefficients significantly different from zero will provide evidence of an effect of the program´s intervention. In particular, we will obtain the effect of each treatment through the following comparisons:
• b1 = Possessing a savings account and an automatic savings program compared to only having a savings account (control group).
• b2 =Possessing a savings account and monthly receiving SMS reminders compared to people only having a savings account (control group).
• b3 = Possessing a savings account, watching a short educational video and receiving a notebook with financial rules of thumb compared to only having a savings account (control group).

Regarding the sampling strategies, the research team will chose around 25 BancoEstado branches to execute the intervention. They will be chosen following a socio-economic evaluation, focusing on those concentrating more vulnerable population. We will stratify the sample to randomly assign participants to treatment and control groups. The stratification variables are (1) saving goal and (2) reception of subsidies.

Given the experimental nature of the evaluation and implementation of the method, it’s impossible to randomize ex post the offer, at least for the treatments that include the automatic savings and the educational video (treatments 1 and 3). This is because the acceptance of an automatic savings mechanism requires the signing of consents and legal documents that can only be filled on the branch by a trained executive. Thus, the randomization will be performed at the branch by the device used to execute the survey, a tablet. This tablet will be specially programmed to carry out this work and stratify by type of saving goal and the receipt of state subsidies.

As randomization instrument, the last digit of the National Identification Number will be used. This national identification number, assigned at birth, depends on the order of inscription and the National Register, and therefore is not related to any personal or socioeconomic variable.

No

There won't be clusters in this RCT, individuals subjects are randomly assinged to the different treatments.

Regarding the size of the sample, for this stage the number of people being considered is 6,000.

1,800 in the Control group, 1,800 in Treatment 1, 1,200 in Treatment 2 and 1,200 in Treatment 3. The differences in the number of observations by treatment arm are due to the differences in their take-up rates (see below)

Considering the following take-up: Control ->100% Treatment -> 40% Treatment 2 -> 80% Treatment 3 -> 95% Then the MDEs when comparing to the control groups are: Treatment 1 --> 0,2524 SD Treatment 2 --> 0,1305 SD Treatment 3 --> 0,1099 SD The MDs of the Treatment 2 (SMS) compared to the other treatments: Treatment 1 --> 0,2822 SD Treatment 3 --> 0,1430 The MDs of Treatment 3 compared to the Treatment 1 --> 0,2822 SD