x

We are happy to announce that all trial registrations will now be issued DOIs (digital object identifiers). For more information, see here.
Testing the Effectiveness of Consumer Financial Disclosure: Experimental Evidence from Savings Accounts
Last registered on March 31, 2019

Pre-Trial

Trial Information
General Information
Title
Testing the Effectiveness of Consumer Financial Disclosure: Experimental Evidence from Savings Accounts
RCT ID
AEARCTR-0004053
Initial registration date
March 25, 2019
Last updated
March 31, 2019 11:00 PM EDT
Location(s)
Primary Investigator
Affiliation
MIT
Other Primary Investigator(s)
Additional Trial Information
Status
Completed
Start date
2015-01-01
End date
2016-10-01
Secondary IDs
Abstract
Disclosure—the practice of providing information to support decision making—has been widely mandated in public policy but is routinely ignored by consumers and subject to obfuscation by firms. Yet most evidence on the effectiveness of consumer financial disclosure stems from lab experiments where subjects do not have competing demands on their attention or from analysis of borrowing decisions where optimality is hard to characterize. In this paper, we provide field evidence from randomized-controlled trials with 124,000 savings account holders at five UK depositories. Treated consumers received varying degrees of salient information about alternative products, including one with their current provider that strictly dominated their current savings product. Motivated by work on search frictions, switching costs, and inattention, our experimental variation is designed to allow us to examine the importance of each in inhibiting effective disclosure. Despite the switching process taking 15 minutes on average and the moderate size of average potential gains (£123 in the first year), attention to disclosure is low, significantly limiting its potential effectiveness, motivating explicit disclosure-design rules, and demonstrating the nature of deposit stickiness.
External Link(s)
Registration Citation
Citation
Palmer, Christopher. 2019. "Testing the Effectiveness of Consumer Financial Disclosure: Experimental Evidence from Savings Accounts." AEA RCT Registry. March 31. https://doi.org/10.1257/rct.4053-1.0.
Former Citation
Palmer, Christopher. 2019. "Testing the Effectiveness of Consumer Financial Disclosure: Experimental Evidence from Savings Accounts." AEA RCT Registry. March 31. https://www.socialscienceregistry.org/trials/4053/history/44369.
Experimental Details
Interventions
Intervention(s)
To inform policy remedies aimed at increasing competition in the savings account market, the UK Financial Conduct Authority partnered with five retail financial institutions to conduct multiple RCTs testing the extent to which disclosures about comparable product interest rates would be useful to consumers. Over 124,000 savings account holders were randomly assigned into treatment and control groups, and we collect rich administrative data on account balances, demographics, and switching behavior from each of the five financial institutions. Consumers in the treatment groups received various forms of information about their account. For a subgroup of consumers, we conduct follow-up surveys to supplement our analysis of behavioral responses to trial interventions with direct explanations for consumer behavior.
Intervention Start Date
2015-01-01
Intervention End Date
2016-10-01
Primary Outcomes
Primary Outcomes (end points)
Our outcome of interest is whether a consumer switched her savings account product.
Primary Outcomes (explanation)
We distinguish between internal and external switching. When customers convert their account to another instant-access savings product or open a new instant-access savings account with the same firm, move some money into it, and empty their old account, we define this action as internal switching. We define all remaining switching that does not fall into internal switching as external switching, including cases where, for example, the depositor converted their savings account to an investment account with the same firm. Internal and external switching are mutually exclusive and always sum up to any switching. Where our measures overlap, we record the latest action as the final action. For example, if a customer first converts an account and then closes the converted account, we classify that as external switching.
Secondary Outcomes
Secondary Outcomes (end points)
Our secondary outcomes include information elicited from our follow-up survey with a subset of consumers, including beliefs, preferences, attentiveness, and awareness.
Secondary Outcomes (explanation)
We query consumers about their beliefs on the difficulty of searching for alternative savings products, the difficulty of switching accounts, and prevailing market interest rates. We ask them about their preferences, including what factors are most important to them when choosing a savings product and what would make them most likely to switch. We measure their attentiveness by asking them if they recall recent disclosures from their account provider and are aware of their current interest rate.
Experimental Design
Experimental Design
Partnering with five UK depositories, we tested multiple disclosure designs in a range of field trials. Each financial institution helped to complete one trial for a total of five trials. All trials were conducted with customers who held an easy-access savings account with one of the partnering UK financial institutions at the time of random assignment. Customers were experiencing a rate decrease in three trials (Trials 2, 4 and 5) and were already on a relatively low rate in two trials (Trials 1 and 3). Trials 1, 2 and 3 provided customers with forward looking information about interest rates currently available to them. The trials varied in terms of the situation in which the customer received the disclosures (whether at the point of an interest rate decrease) and the specific disclosure designs tested within each trial.

Customers in the reverse-page switching box trial (2) and in the reminder trials (4 and 5) faced an interest-rate decrease to a level that was significantly below the average of what new customers could obtain. In these trials, the firms sent letters to customers informing them of the old and new interest rates and some general contact details for further information no later than 60 days before the interest rate decrease, in accordance with EU regulatory requirements. In the reverse-page switching box (2) and the SMS reminder trials (5), the rate decrease applied to all customers holding the particular type of the account and occurred on the same date for everyone. In the digital reminder trial (4), the rate decrease occurred a fixed period of time after the individual account opening date and was part of the account terms and conditions. Customers in the front-page switching box (1) and the switching-form trials (3) faced no interest rate decrease but were already receiving an interest rate that was significantly below market average of what new customers could obtain. These customers received no other specific communication in advance of the information sent out during the trial.

In Trial 1, consumers were already on a low rate and were due to receive their annual statement in autumn 2015. Customers were randomly selected into five equally sized groups. The control group received an annual statement with no additional information on the front page. For treatment groups, different information was added to the front page of the annual statement depending on the treatment arm. This included a simple encouragement to shop around for another account; a comparison of the currently applicable rate with the highest rate available on a comparable account with the current provider (best internal rate); the best internal rate with the same provider and the average of three highest rates on comparable accounts with competitors (best competitor rates); and a final variant which added a graphical illustration of gains from switching. The monetary gains used in the illustration were based on an illustrative balance of £100, £1,000 or £10,000, depending on which was the next lowest to the customer’s actual savings balance at the time (for example, for an actual balance of £250 the illustration was for £100). We excluded customers with balances lower than £100.

In Trial 2, customers at the time of disclosure had rates close to market rates and were being notified of an impending rate decrease. Customers who had opted out of marketing communications were excluded from this trial. All letters were sent more than two months ahead of the rate decrease. Customers were randomly assigned to one of five equally sized groups: a control group and four treatment groups. The control group received a letter which notified the customer of the rate decrease on all affected instant-access accounts early summer of 2015. The control letter included no additional information about internal or external rates. The treatment groups received additional information on the back page of the letter, formatted into a call-out box (referred to as the “switching box”) with a graphical comparison of interest rates. Each treatment group received information about the best available interest rate with the current provider. The four treatment groups differed by whether the disclosure was personalized and whether it included information on the best competitor rates (the average of the three highest rates on comparable accounts), for a total of four possible combinations. The two non-personalized treatments (one displaying external rates and one not) had an illustration of monetary gains from switching based on an assumed balance of £5,000, and the personalized disclosures used each individual's balance at the time of mailing to illustrate the gains.

Trial 3 featured a disclosure bundled with a switching-process improvement designed to lower both the expected time cost of switching and its uncertainty. In addition to providing information as in Trials 1 and 2, we test the effect of providing a form that can be completed and returned to the firm in order for the customer to be switched to an identical, “front book” product paying a better rate at the same provider. In August 2015, the provider sent a one-off mailing to encourage long-standing customers to switch to an equivalent internal account with a significantly higher rate. The customers were selected randomly into two equally sized groups. The control group received a letter with a switching box that included the best internal rate and the best competitor rate as noted above, as well as potential gains from switching based on a non-personalized balance example (£5,000). The treatment group received the same letter, but with a tear-off return switching form pre-filled for a switch to the best internal rate and a prepaid, addressed envelope.

In Trials 4 and 5, we test the effect of timely repetition of informational disclosures through reminders sent via email or SMS. In Trial 4, we test email or SMS reminders sent to consumers who held accounts that experienced scheduled rate decreases during June-September 2015. The trial sample consisted only of customers who all had an email address and a mobile phone number on record. Over 90% of customers in the sample had both email and phone number on record. The partnering institution reported that around 2% of email reminders and around 10% of SMS reminders could not be delivered due to invalid records. We do not adjust trial results for failed deliveries. Customers were randomly selected into three equally sized groups. The control group received only an initial letter sent at least 60 days before the rate decrease, as mandated by current regulation. Two treatment groups were then issued either an email reminder or an SMS reminder in addition to receiving the same mailing as the control group. The email reminder was similar in its content to the letter sent to all groups. It included information about the previous and new interest rates, and in addition to the initial letter it included the best interest rate available on a comparable account with the firm. The SMS reminder was shorter and included no information on interest rates. Due to logistical constraints we sent the reminders on one actual date to all customers, as it was not possible to randomly allocate reminders to be sent at different points of time. Each customer account had an interest rate decrease date which was within eight weeks before and seven weeks after the date of sending the reminders.

In Trial 5, we test the effect of an SMS reminder around the time of a rate decrease in early summer 2015. All customers in the trial had a mobile phone number on record. The partnering institution evaluated that around 8% of reminders were not delivered to customers in the treatment groups due to invalid phone number records. We do not adjust trial results for failed deliveries. Customers were randomly selected into one of five groups. The control group received no further communication following the initial letter sent 60 days or more before the rate decrease. Customers in the four treatment groups received an SMS reminding them of the rate change, one week before, one week after, or on the day of the rate decrease. For those receiving the SMS on the day of the rate change, the SMS either encouraged switching or said that there was no higher rate on a comparable product available. Each treatment group included 16% of the trial sample and the control group included the remaining 35% of the sample. Customers who switched between assignment and the due date of the reminder still received the reminders and were retained in the sample to ensure that the comparison of effects of timing is consistent across all treatment groups.
Experimental Design Details
Randomization Method
Randomization done in office by a computer
Randomization Unit
individual account
Was the treatment clustered?
No
Experiment Characteristics
Sample size: planned number of clusters
Our experiment consists of five trials, each with a different anonymous partnering financial institution. Each trial had a different number of treatment arms. Trials 1, 2, and 5 had four treatment arms. Trial 3 had one treatment arm. Trial 4 had two treatment arms.
Sample size: planned number of observations
124,832 consumers
Sample size (or number of clusters) by treatment arms
Trial 1: 12,723 control; 49,156 treatment (12,441 in T1; 12,628 in T2; 12,320 in T3; 11,767 in T4)
Trial 2: 2,659 control; 10,602 treatment (2,637 in T1; 2,652 in T2, 2,632 in T3; 2,681 in T4)
Trial 3: 1,999 control; 2,004 treatment
Trial 4: 5,180 control; 10,307 treatment (5,115 in T1; 5,192 in T2)
Trial 5: 10,200 control; 20,002 treatment (4,998 in T1; 4,999 in T2; 4,998 in T3; 5,007 in T4)
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
MDE = k*sqrt(4*sigma^2/N) = 0.44 percentage points when k=2.8 and N=124,832 and sigma2 = mu*(1-mu) for our binary dependent variable other switching that has mu = 8.7%.
IRB
INSTITUTIONAL REVIEW BOARDS (IRBs)
IRB Name
Financial Conduct Authority
IRB Approval Date
2015-01-01
IRB Approval Number
N/A
Post-Trial
Post Trial Information
Study Withdrawal
Intervention
Is the intervention completed?
No
Is data collection complete?
Data Publication
Data Publication
Is public data available?
No
Program Files
Program Files
Reports and Papers
Preliminary Reports
Relevant Papers