The Semblance of Success in Nudging Consumers to Pay Down Credit Card Debt

The trial applied to consumers who are in the process of applying for a new credit card with that lender.

When a consumer is applying for a credit card online and had been provisionally accepted they reach a screen providing them with the option to setup a direct debit. At this point the individual is randomly allocated to either treatment or control group via a random number generator. This trial remained live until the target sample size had been reached.

The control group were presented with the standard automatic payment (Autopay / Direct Debit) options: automatic full payment, automatic fixed payment, automatic minimum payment).

The treatment group were not presented with an explicit automatic minimum payment option. They only saw automatic full payment and automatic fixed payment as options.

Once allocated to control or treatment they would view these same choices if they returned to the pages to set-up or change their automatic payment choices within 30 days of applying for the card. One reason for doing so may be that they have wanted to sign-up for automatic payments but did not have bank details to hand so returned later on. If a consumer phoned the lender’s call centre they had the normal automatic payment options available to them.

2016-11-01

2017-12-31

The 10 primary outcome variables we will focus on are as follows. Outcomes 1-6 are measured on target credit card from lender data, outcomes 7-10 on credit card portfolio from credit file data:

1. Minimum payment: Credit card repayment of only contractual minimum amount (binary).

2. Full payment: Credit card repayment of full balance.

3. Insufficient payment: Credit card repayment below contractual minimum amount or no payment made when due (binary)

4. Outstanding debt: (Credit card statement balance – repayments) / (Credit card statement balance)

5. Cost: (Credit card interest + credit card fees) / (Credit card statement balance)

6. Consumption: (Value of credit card transactions) / (Credit card statement balance)

7. Portfolio minimum payment: Proportion of credit cards where repayment of only contractual minimum amount

8. Portfolio full payment: Proportion of credit cards where repayment of full balance

9. Portfolio insufficient payment: Proportion of credit cards where repayment below contractual minimum amount or no repayment made when due

10. Portfolio outstanding debt: (Sum of credit card statement balances – repayments) / (Sum of credit card statement balance)

These primary outcomes were selected on the basis of being outcomes of economic importance and those where we can robustly estimate the treatment effects as the variables are bounded. These primary outcomes are intended to examine the ‘target card’ (subject to the trial) as well as the broader ‘credit card portfolio’ of that cardholder.

We focus on credit card outcomes as primary outcomes of importance as our trials apply to credit cards. Credit cards account for a considerable amount of consumer credit debt and credit cards put the onus on consumers to choose their repayment amount (whereas many other products which have a fixed amortisation schedule). We look at the broader portfolio of credit products held in secondary outcomes.

Three variables are straightforward binary variables (or dummy variables) for whether the consumer’s payment is repays their balance in full, repays only their contractual minimum amount or repays nothing (or less than) the contractual minimum payment. When looking at the credit card portfolio the analogous variables are the proportion of credit cards with those repayment characteristics (full, contractual minimum or below contractual minimum).

Choosing other primary outcomes to assess outstanding debt and repayments was less straightforward. The distribution of credit card balances (and variables derived from them such as repayments and interest costs) have what is known as a ‘fat tails’ problem. This means that there is a long tail to the distribution of such variables which can mean that estimates of averages are heavily influenced by the extreme values in the tail of the distribution. Such extreme values (e.g. those with very large credit card debts) can be the observations potentially of most interest so we do not want to drop them in our primary analysis. We examined ways to normalise these variables by income estimates using a representative sample of CRA data but the ‘fat tail’ problem remained and doing this also made the estimates less easily interpreted.

Instead we therefore construct primary outcome variables for assessing debt, cost and consumption as ratios of statement balances. This is an imperfect solution as it means that both the numerator and denominator are endogenous in the construction of such ratios (For the cost variable the interest component is proportionate to the balance). However, doing this enables the outcomes to be bounded in order to be able to be precisely estimated while retaining the ability to be easily interpreted. We address the issue of the numerator and denominator both being affected by decomposing it into its component parts (with ‘fat tails’) to assess the relative contributions.

In cases where the total value of repayments within a statement cycle is greater than the statement balance we plan to limit repayments to the statement balance amount. The same approach is planned to be used for the value of transactions. In cases where repayments, transactions or statement balances are negative we will limit these to zero. In cases where a zero balance is due this is recorded as full payment as no debt is outstanding.

Secondary analysis considers a broader set of outcomes and empirical approaches to check the robustness of the primary results and understand the
mechanisms driving the results in greater detail. Conducting secondary analysis depended on the results from the primary outcomes.

We measure outcomes using microdata collected from lenders and credit file data collected from a credit reference agency.

Our experiment is an RCT. We use 0.5% as the threshold for statistical significance. 5% where results are suggestively significant and also highlight where results are significant at the 1% level. In addition, for the primary outcomes, we will distinguish between statistically insignificant results where we confident in there being no effect (i.e. precisely estimated null results) from results where we cannot rule out there being a treatment effect with confidence (i.e. low powered). This will be done through reporting the minimum detectable effect size (MDE) which shows the effect that would have been detectable with 80% power at the 5%, 1% and 0.5% statistical significance level.

We will construct an unbalanced panel with one observation per credit card, per month observed. The panel is unbalanced as some cards will have been part of a trial for a longer period of time than other cards. We expect to observe each individual to only have one of their credit cards subject to the trial.

We plan to estimate an OLS regression with standard errors clustered at the individual card-level. We expect to observe multiple credit card statement cycles from the start of the trial. Dummy variables for these will be included to allow for the treatment effect to be measured for each cycle from the start of the trial. We hypothesise that treatment effects will vary over time (for example due to consumer learning to optimise their payments as their balances change) but we do not impose a functional form on these as it is unclear what the appropriate functional form would be. The target card is the credit card subject to the trial.

We plan to estimate an OLS regression with standard errors clustered at the individual card-level. We expect to observe multiple credit card statement cycles from the start of the trial. Dummy variables for these will be included to allow for the treatment effect to be measured for each cycle from the start of the trial. We hypothesise that treatment effects will vary over time (for example due to consumer learning to optimise their payments as their balances change) but we do not impose a functional form on these as it is unclear what the appropriate functional form would be. The target card is the credit card subject to the trial.

We plan to include a constant, a series of time-invariant control (CONTROLS) variables (constructed using information on the target credit card and card-holder from before the start of the trial and dummies for the month and year (MONTH) the outcome is observed, dummies for statement cycle number (CYCLE), and an interaction between TREATMENT and CYCLE. The controls are designed to soak up variation not attributable to the trial in order to make our estimates of treatment effects more precise. We are not examining the coefficients on these in the primary analysis. This is because with the lags of the outcome included as controls, the coefficients on other control variables do not measure the effects of these, they measure the effect of controls on the change in outcome.
In this specification the coefficients on the interaction between CYCLE and TREATMENT shows the treatment effect cycles since the start of a trial.

CONTROLS were Gender, Age, Age squared, Log Estimated Income, Credit Score, Unsecured Debt-to-Income (DTI) Ratio, Any Mortgage Debt, Log Credit Card Credit Limit, Credit Card Purchases Rate, Subprime Credit Card, Any Credit Card Promotional Rate, Any Credit Card Balance Transfer, Credit Card Open Date, Credit Card Statement Day, Any Credit Card Secondary Cardholder.

These were all from the time of card origination (or month preceding card origination where consumer rather than Credit Card specific variables). For outcomes constructed from credit reference agency (CRA) data up to eleven dummies for lags of outcomes were included for months preceding the start of the trial.

CYCLE and MONTH are both included because statement cycles do not perfectly align with calendar months and trials went into the field at different points-in-time.

As the consumers are not known in advance, the randomisation process is carried out ‘live’ without exclusion restrictions. Randomization was carried out through a random number generator JAVA script created by the lenders.

Randomized at the consumer level.

No

At least 40,000 consumers for lender 1. At least 10,000 consumers for lender 2.

At least 40,000 consumers for lender 1. At least 10,000 consumers for lender 2.

At least 20,000 consumers for each of control and treatment for lender 1. At least 5,000 consumers for each of control and treatment for lender 2.

We run a field experiment and a survey experiment to study an active choice nudge. Our nudge is designed to reduce the anchoring of credit card payments to the minimum payment. In our field experiment, the nudge reduces enrollment in Autopaying the minimum from 36.9% to 9.6%. However, the nudge does not reduce credit card debt after seven payment cycles. Nudged cardholders tend to choose Autopay amounts that are only slightly higher than the minimum payment. The nudge lowers Autopay enrollment resulting in increasing missed payments. Finally, the nudge reduces manual payments by cardholders enrolled in Autopay.

The Semblance of Success in Nudging Consumers to Pay Down Credit Card Debt (2023) Guttman-Kenney, Benedict & Adams, Paul & Hunt, Stefan & Laibson, David & Stewart, Neil & Leary, Jesse. NBER Working Paper No. 31926.

https://www.nber.org/papers/w31926