Barriers to Bankruptcy Relief

2025-02-25

2025-12-31

Our primary outcome is an indicator for whether the respondent files for bankruptcy after the baseline survey, as measured in credit report data from a credit reporting bureau.

We will additionally test the following secondary outcomes:
• Reported likelihood of filing for bankruptcy (baseline survey)
• Willingness to pay for bankruptcy information and assistance filing (baseline survey)
• Clicking through to a webpage providing bankruptcy information and assistance filing (baseline survey)
• Whether the respondent has contacted a bankruptcy attorney (follow-up survey)
• Whether the respondent plans to file for bankruptcy in the next year (follow-up survey)
• Whether the respondent started the bankruptcy screener tool at Upsolve (follow-up survey)
• Credit card balances (credit reports, measured one and two years after the survey)
• Credit score (conditional on having one) (credit reports, measured one and two years after the survey)
• Total amount of debt past due (credit reports, measured one and two years after the survey)
• Total amount of debt in default (credit reports, measured one and two years after the survey)
• Total amount of debt in collections (credit reports, measured one and two years after the survey)
• Income (credit reports, measured one and two years after the survey)

We have three sets of secondary outcomes: (1) outcomes from the baseline survey, (2) outcomes from the follow-up survey, and (3) other outcomes from the credit report data. These are described below.

Baseline survey outcomes: We measure three outcomes in the baseline survey. First, we ask participants to estimate the likelihood that they would file for bankruptcy in the next year if they had trouble paying debts. We form an indicator for whether the likelihood is “Somewhat Likely,” “Likely,” or “Extremely Likely.” Second, we ask participants to enter a lottery to win their choice of either (i) a cash prize of $30 or (ii) access to a service that provides information about bankruptcy and assistance with filing. One out of every 1,000 participants will be randomly selected as a lottery winner and receive whichever option they selected ($30 cash or links to Upsolve, a nonprofit providing bankruptcy filing information and assistance). We form an indicator for whether the participant chooses information and assistance over the $30 cash prize. Third, all treatment group respondents, regardless of willingness to pay or lottery outcome, will be given the link to Upsolve at the end of the survey. We form an indicator for whether respondents click this link. This indicator is set to zero for the control group.

Follow-up survey outcomes: Between two and four months after the baseline survey, we will email a link to a follow-up survey to participants who completed the baseline survey, passed attention checks, and do not own real estate. We will measure and form indicators for: (i) whether the participant contacted a bankruptcy attorney, (ii) whether the participant plans to file for bankruptcy in the next year, and (iii) whether the participant started the bankruptcy screening survey on Upsolve.

Credit report outcomes: We will obtain annual credit reports from one year before (t = -1) to two years after (t = 2) the initial survey. Using this data, we will measure borrowing behavior (sum of credit card balances), access to credit (credit score conditional on having one), financial distress (sum of balances for tradelines with payments 30+ days past due, sum of balances for tradelines with payments 90+ days past due (i.e., in default), sum of debt in collections), and income. This will enable us to estimate both the effects on behavior of the informational intervention and the effects of filing for the subset of individuals who are nudged into filing.

A credit reporting bureau will provide a list of one million email addresses corresponding to high-debt adults in the US for whom knowledge of the bankruptcy system could be useful for their financial decision-making. To arrive at this sample, the credit reporting bureau applied a Chapter 7 bankruptcy prediction model to the entire US population that estimated the likelihood of filing based on six months of bankruptcy data. The model incorporated key credit report attributes including income, credit score, and credit utilization. Focusing on adults that have never filed for bankruptcy and that have an associated email address, the credit reporting bureau ranked adults by this propensity score. The top (i.e., highest propensity) one million never-bankruptcy adults with associated email addresses were selected.

We will invite these individuals to take our online survey until we have emailed all 1 million email addresses or until we have 30,000 respondents who consent to the study, complete the baseline survey, confirm in the survey that they do not own real estate, pass our two initial attention checks, and indicate in the survey that they do not expect to fully repay their debt. Our final sample includes only the individuals that meet these criteria. For outcomes only measured in the follow-up survey, we will additionally restrict the sample to those who complete the follow-up survey.

Our baseline survey is structured as follows. First, we ask respondents about their household finances, which helps us estimate the respondent’s vehicle equity and the amount of debt they might be able to discharge in Chapter 7 bankruptcy. Next, we ask participants to estimate the likelihood that they would file for bankruptcy if they had trouble paying debts. We then measure knowledge of bankruptcy and concerns with bankruptcy. Next, we randomize participants and provide the information intervention. Then, we measure the baseline survey outcomes. We conclude by measuring participant demographics, notifying lottery winners of their prize, and providing the Upsolve link to treatment group respondents.

Not available

Randomization done within the Qualtrics survey.

Randomization will be at the user level.

No

n/a

Up to 30,000 respondents. Our sample size will be smaller if we exhaust our list of email addresses without reaching 30,000 responses that meet the criteria described above.

Up to 7,500 respondents per treatment arm. Our sample size will be smaller if we exhaust our list of email addresses without reaching 30,000 responses that meet the criteria described above.

With the maximum sample size of 30,000, we are powered for a minimum detectable effect of 1.1 percentage points on our primary outcome of bankruptcy filing. This represents approximately 20% of our assumed sample mean based on our estimate that 5.7% of individuals with more than $40,000 in non-student, non-mortgage debt file for bankruptcy annually based on the Survey of Consumer Finances.