Experimental Design
The three experiments were implemented by three different banks in three different country “sites”: Bolivia, Peru, and the Philippines. The sample for each experiment is comprised of new commitment savings account holders. What is meant by “commitment savings account” differs in each of the three sites. After opening the account, the bank randomly assigned reminder treatments: reminder or not, and then content and timing within the reminder group. Banks did not mention or advertise reminders prior to random assignment, nor did they create random assignments for non-takers, and hence we conduct our analysis on account-openers only. In two of three sites – Bolivia and Philippines - the reminders were not even announced at account-opening; the bank just started sending them at no charge to clients. Each bank also had its personnel collect some “baseline” data prior to making the product offer. Products, marketing, and some reminder features varied across sites.
First Valley Bank (FVB), a for-profit bank operating in Western Mindanao, Philippines, worked with us to randomize reminders as part of the rollout of its new Gihandom (Dream) Savings product. Between April and August 2007, bank marketing employees conducted door-to-door marketing visits in rural and small urban areas and offered 10,056 individuals the opportunity to open a Gihandom account. As part of this marketing visit, the bank employee also conducted a brief five to ten minute survey. Bank staff used personal digital accessories (PDAs) for the baseline survey and random assignment to treatments. Of the 10,056 offers, 2,314 (23%) opened an account. Gihandom allows a client to set her own commitment amount (US$50 or above) and commitment end-date (from three months to two years after opening). Except in hardship cases, clients can then access funds only once both commitments - amount and date - have been met. Once the client opens the account with a minimum deposit of US$2.50, there is no fixed deposit schedule to fulfill. The client receives a savings lockbox and is encouraged at sign-up to make small deposits on a daily basis. When the client desires, the client goes to the bank to deposit the money in the lockbox (e.g., when it is full). Clients also can, and do, go to the bank and make normal deposit, without the lockbox. We do not have data on whether they client had the lockbox with them for the deposit. Among clients with a cell phone (66% of those who opened accounts), the bank randomly and independently assigned some clients to receive “regular” and/or “late” text message reminders to come to the bank to make a deposit each month. The bank sent the late reminder only if the client did not make any deposit in a given month. Reminders were further randomized to gain or loss frame language with respect to “making your dream come true”. A client assigned to receive both regular and late reminders got the same frame on all messages.
In Peru, the Government-owned bank Caja de Ica worked with us to randomize reminders as part of the rollout of its new product Plan Ahorro (“Saving Plan”). The bank marketed the product on television and radio in the Ica metropolitan area (urban and rural), and clients signed up over the course of several months. When opening an account, Plan Ahorro clients selected a commitment end-date (between 6 months and 12 months post-opening), a minimum commitment amount to deposit each month, and a goal (specific expenditure) label from 14 pre-established categories. Clients were required to make each planned deposit within ten days of each monthly due date in order to meet their commitment. Commitment compliance was rewarded with an annualized interest rate of 8% per annum rather than the normal 4% per annum. As at our other sites, the bank randomly assigned reminders to clients after they signed up for the product. The bank sent letters because low cell phone prevalence made text messages impractical. As in the Philippines, the bank did independent randomizations for regular and/or late reminders that were assigned to the same gain or loss frame. The bank sent regular reminders with a target client-receipt date seven days before the due date for that month's scheduled deposit. The bank also randomly assigned regular-reminder clients to have their reminders signed by either the bank, or the client herself, with her signature recorded at account-opening. As in the Philippines, the bank sent a late reminder only if the client was late (i.e., if they had not made a deposit three days after their scheduled deposit date). All late-reminder letters were signed by the bank. The bank implemented two additional treatments designed to increase the salience of the client's specific expenditure goal. One treatment randomly assigned some in the reminder group to get a letter that focused on their particular goal. Another treatment independently and randomly assigned the gift clients received upon opening the account: a jigsaw puzzle of their goal, a photo of their goal, or a pen. Those in the jigsaw puzzle group received a piece of the puzzle after each deposit.
Ecofuturo, a for-profit bank in Bolivia, worked with us to implement a text message reminder program for its established product Ecoaguinaldo. “Aguinaldo” is the year-end bonus, equal to one month's pay, that employers are required to pay salaried employees in Bolivia. Ecofuturo markets Ecoaguinaldo as a product designed to help its clients, many of whom are self-employed, save up all year for their own year-end payout. The product is marketed for three months between January and March on television and radio in urban areas of Bolivia close to Ecofuturo's branches. Clients who opened an account during January-March 2008 were brought into the study and eligible to make savings deposits until the December 2008 commitment end-date. At sign-up, clients chose a monthly minimum deposit amount (with a floor of US$1.41). Clients making all of their committed monthly deposits received a bonus interest rate of 6% for their first ten months following enrollment in the program (compared to a regular interest rate of 3%) as well as free life and accident insurance.10 Clients missing one deposit, or withdrawing money before the payout date, forfeited the higher interest rate and had their insurance policies canceled. As in the Philippines, clients with a cell phone were randomly assigned to receive text message reminders or not. In Bolivia the experimental design called for everyone assigned to get a reminder to get one every month, in advance of their scheduled deposit; i.e., all reminders were supposed to be “regular,” and unlike the other two sites the bank did not randomly assign variation in regular versus late reminders. But the bank deviated from this design in two ways. First, it did not start sending reminders until several months after account opening (which took place January-March 2008), beginning in May. Second, the bank switched from a “regular” to “late” reminder rule starting in August: from August-November, clients were only sent a reminder if they had not yet made a deposit that month. As a consequence most clients received fewer reminders than prescribed by the design. (We use the random assignments in our analysis, rather than actual treatment status, to avoid bias from the endogeneity of not yet having made a deposit.) Interestingly, we see that lower-frequency messaging does not seem to reduce the potency of reminders. This suggests that it is the initial reminders that matter (consistent with “tuning out” or habituation over time), although we lack the random variation required to test that hypothesis. Every reminder mentioned the savings goal, broadly speaking: “Aguinaldo,” the year-end bonus (note that this goal is somewhat more generic than at the Peru site, where many clients indicated a more specific future goal like school fees). Besides the boilerplate script content was randomized 2x2: (gain or loss frame) x (mention insurance incentive or not). The latter treatment parallels the Peru design in the sense that it produces some reminders that mention both a financial incentive and a savings goal (in this case, the “Aguinaldo” or year-end bonus).