Experimental Design
Our sample frame for the study is drawn from the Community Action Project’s (CAP) free Volunteer Income Tax Assistance (VITA) tax preparation service for low and moderate income (LMI) individuals. CAP is based in Tulsa, Oklahoma and provides a range of social services including early childhood education and first-time homebuyer’s assistance. It runs one of the largest per-capita free tax preparation programs in the country, and delivered this service at eight different locations throughout Tulsa County during the 2010 tax season. CAP gave the research team access to the three busiest sites, and the research team approached potential subjects at these sites from January 26 through April 15.
High volume at the tax preparation sites resulted in long waiting times for many individuals visiting the CAP centers, and we used this downtime to enroll people into the study. The CAP receptionist or a surveyor from the research team would periodically ask people in the waiting room if they were interested in completing a survey about their financial well-being. The tax-preparation staff also encouraged people they were assisting to complete the survey after their tax-preparation session. Individuals were offered a $5 gift card redeemable at a local gas station and convenience store as an incentive to participate in the survey. Interested individuals were directed to a member of the research team. People granting consent to participate in the study (by complete the survey and granting permission for the research team to do “soft pulls” of the subject’s credit report that day, and going forward) were then interviewed. Survey interviews took about fifteen minutes and covered basic demographics, financial product use, respondent assessments of their household’s financial condition, and some basic questions designed to measure behavioral factors (time-inconsistency, exponential growth bias, and limited attention). 1432 interviews were completed over the 2.5 months of surveyor/marketer operations. Upon completing the survey, the interviewer used a random number generator in Excel to randomly assign the interviewee (the subject) to either receive a BoLT offer (details below in II-D), following the survey, or not. This process and assignment was not visible to the interviewee. Our sample frame consists of 465 individuals that qualified for BoLT by having an auto loan or a credit card with a positive balance at baseline. Of these, 238 were assigned to be offered BoLT (treatment) and 227 assigned to not receive BoLT (control).
We designed BoLT with an eye towards offering a simple debt reduction product that counters behavioral biases toward indebtedness and ultimately could be offered to a mass market of consumers interested in accelerating debt repayment (and/or controlling new borrowing). The version of BoLT piloted here has three features: a planning/goal setting tool, a commitment option, and reminders. Everyone completing a survey was randomly assigned to receive a BoLT offer or not, immediately upon finishing the survey (the randomization was conducted in real-time and unbeknownst to the subject, on the market/surveyor’s laptop). Subjects assigned to the control group received their survey compensation (a gift card), and thanks for their time. Subjects assigned to receive a BoLT offer also got a marketing pitch with a brief overview of BoLT’s features. Subjects expressing interest in BoLT were then taken through a more thorough, but still brief (typically twenty minutes from start-to-finish for those who signed up), planning and set-up process.
Step 1 of the BoLT marketing and intake process is identifying a suitable debt. In the pilot, this involved the marketer inviting the prospective client to discuss the client’s credit report, and using this information and the client’s recall to identify an auto loan or credit card account with a nontrivial balance and a high APR (credit reports do not include pricing information). Due to operational constraints in the pilot (namely making BoLT marketing and intake as simple and quick as possible in order to capture as large as a sample as possible while they were waiting to get their taxes done), we offered BoLT for only a single loan per client.
Step 2 of the intake process is using a decision aid to help make a concrete plan for paying down debt more quickly. The idea is to present someone with a simple but effective planning/goal setting tool that counters the many potential cognitive obstacles to motivating, setting, and implementing a realistic debt reduction goal, including limited numeracy or literacy, exponential growth bias re: interest expense, information overload, and planning fallacies. In the pilot studied here, the tool was a simple repayment schedule calculator, used by the on-site surveyor/marketer to help interested clients craft a realistic goal and schedule for accelerating the repayment of a single credit card or auto loan debt. Marketers typically reviewed a number of hypothetical payment schedules with a client to demonstrate the potentially dramatic reductions that small increases in monthly payments can have on repayment time and total interest paid (demonstrating this is a goal of one the new credit card disclosures required by the CARD Act). Once someone drafted a repayment schedule featuring the usual equal monthly payments, marketers presented the option of a payment schedule that escalated $10 each month.
Step 3 of the process is the offer of a commitment device, in this case the option of signing up one or more “peer supporters”, who would be notified and asked to provide encouragement (but not financial support), in the event that the BoLT client fell behind on her repayment schedule. The approach is to use the ex-ante prospect of peer notification, and any ex-post peer encouragement, as additional incentives for sticking to the debt reduction plan. Individuals who made a repayment plan in Step 2 were offered the Peer Support option, with a few exceptions. A client selecting the option simply gave the marketer contact information for one or more friends, family members, and/or co-workers.
The fourth step of the BoLT intake process had the client choose whether she wished to receive monthly reminders by phone or email. The idea here is to counter limited attention by keeping the client’s debt reduction goal and plan at “top of mind” (see, e.g., Karlan et al. 2011 and cites therein). Stand-alone reminders may be particularly important in settings, like our pilot, where the institutional or contractual environment does not tend to produce regular follow-up contacts (monthly statements, advertising). Reminders were administered by research team staff.
Besides administering reminders, the other follow-up administration involved in this implementation of BoLT is monitoring repayment progress so that peer supporters can be notified if the supported client falls behind their scheduled repayment plan. The research team does this monitoring monthly by comparing information from the client’s latest credit report (obtained with a soft pull) to her BoLT repayment plan.