Selective Attention in Consumer Finance: Evidence from a Randomized Intervention in the Credit Card Market
How to cite
Medina, Paolina. 2018. "Selective Attention in Consumer Finance: Evidence from a Randomized Intervention in the Credit Card Market." AEA RCT Registry. AEARCTR-0003034. https://doi.org/10.1257/rct.3034-1.0
Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.
* Primary investigator
Abstract
In partnership with a Personal Finance Platform in Brazil, I implement a randomized intervention to measure the effect of reminders for timely payment of credit card bills. While I find a 13% reduction in the cost of late fees paid, 31% of the users that avoid credit card late payments, incur instead overdraft fees on their checking accounts. This leads to heterogeneous gains from the intervention with some users saving 15% in total fees paid, and others incurring increased fees of 5%. I interpret these results using theories of salience-driven selective attention, and argue that informational nudges may inflate the perceived importance of the target task. The results of this experiment suggest that when designing policy interventions, one size may not fit all, and targeting nudges to those who are more likely to benefit from them has the potential to significantly increase the gains from the intervention.
Intervention
For this intervention, five messages were defined to be sent at different moments of the credit card billing cycle. Three of these messages were designed to be sent early in the billing cycle, to encourage planning. In an attempt to strengthen individuals' mental accounts for credit card payments, the messages encourage users to set some money aside for their next credit card payment. These messages are referred to as "Planning" messages. Two more messages were designed to be sent close to the credit card due date, to invite consumers to make their credit card payment on time. This last set of messages is referred to as "DD Alert" messages.
Treatment Group 1 received both Planning and DD Alert messages and is referred from now on as the Full-Treatment Group. Treatment Group 2 received only DD Alert messages and is referred to as the DD Alert Group from now on. Finally, the Control Group received no messages. All users in the experimental pool were removed from regular message lists sent by the Partner, and therefore received no additional messages.
In addition to the main experimental design, I set aside six thousand randomly selected users as part of a Robustness Treatment Group. Users in this group received different placebos in each of the treated billing cycles. In the first treated billing cycle, they received only the first planning message (24-27 days before the
DD). In the second treated billing cycle, they received five messages from the pool of messages regularly sent by the Partner. These messages were chosen for having the same login rates as the main treatment
messages, but contain no information about late payment fees or credit card due dates.
Experimental design
To run the experiment, the Partner company agreed to provide access to the information of 26,069 users, to send them push notifications containing information on credit card due dates and late payment fees over two billing cycles (September 25, 2015 to December 13, 2015), and to send each participant a survey ten days after receiving the last push notification.
The experimental pool corresponds to the universe of users that were charged a late payment fee between August 2014 and April 2015, and whose last login was on or after May 2015. Access is granted for available transactions taking place between August 1, 2015 and January 31, 2016.
The main design consists of randomly assigning 26,069 users into four groups: one Control Group, two main Treatment Groups, and one additional Robustness or alternative-control Treatment Group. This last group is used to investigate explanations alternative to the main limited attention channel proposed in this paper.
The sample is stratified according to the following variables: a dummy variable indicating whether the user had enough balance in his checking accounts at the time when he was charged a late payment fee, a dummy variable indicating whether or not the user receives periodic paychecks, a discrete variable indicating the due date (DD from now on) on the main credit card of every user, and another dummy variable indicating whether or not the gap between DD and paycheck date is between 0 and 15.
Design details
Cited by
No citation data available for this registration yet.