Bankruptcy Information Survey Experiment

Last registered on October 16, 2026
View Trial History

Pre-Trial

Trial Information

General Information

Title
Bankruptcy Information Survey Experiment
RCT ID
AEARCTR-0019916
Initial registration date
October 08, 2026

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
October 16, 2026, 2:06 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.

Locations

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information

Primary Investigator

Name
Samuel Antill
Affiliation
Harvard Business School
Contact Primary Investigator

Other Primary Investigator(s)

PI Name
Raymond Kluender
PI Affiliation
PI Name
Eleanor Jenke
PI Affiliation

Additional Trial Information

Status
In development
Start date
2026-10-09
End date
2026-12-31
Keywords
Behavior, Finance & Microfinance, Welfare
Additional Keywords
JEL code(s)
D14, G4, G5, H31, K35
Secondary IDs
Prior work
This trial does not extend or rely on any prior RCTs.
Abstract
Many financially distressed households who could benefit from Chapter 7 bankruptcy never file. Common misperceptions about bankruptcy – such as overestimating the negative impacts on one’s net worth and access to credit – may discourage filing. This randomized experiment will test the following question: does providing information about Chapter 7 filing outcomes increase interest in filing for bankruptcy? We test one hypothesis for why individuals do not file: individuals underestimate the probability that bankruptcy will lead to a debt discharge and overestimate the probability that they will have to surrender property. We partner with Upsolve, a nonprofit that helps individuals file for Chapter 7 bankruptcy. Upsolve will survey individuals who created Upsolve accounts but did not file. The survey will elicit perceptions about bankruptcy outcomes and provide information to correct these misperceptions. By exogenously shifting beliefs about the effects of bankruptcy, we will estimate the effect of the information provided on (i) engagement with the filing process through Upsolve and (ii) stated propensity to file.
External Link(s)

Registration Citation

Citation
Antill, Samuel, Eleanor Jenke and Raymond Kluender. 2026. "Bankruptcy Information Survey Experiment." AEA RCT Registry. October 16. https://doi.org/10.1257/rct.19916-1.0
Sponsors & Partners

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information
Experimental Details

Interventions

Intervention(s)
Upsolve will send survey invitations via email to individuals who (i) completed the bankruptcy screener on Upsolve’s website, (ii) were deemed qualified to file, (iii) created an Upsolve account, and (iv) did not file. The survey will first ask respondents about their beliefs about the effects of bankruptcy, then respondents will be randomly assigned to the treatment or control group with equal probability. The “net worth” treatment group learns that (i) 3% of Upsolve users who filed for Chapter 7 bankruptcy surrendered assets and (ii) 97% received a full discharge of eligible debt. Meanwhile, the control group learns that the labor force participation rate was 62% in June 2026.

The labor force participation statistic is from the Federal Reserve Bank of St. Louis. The discharge and surrender statistics are calculated using Upsolve filers successfully linked to the Federal Judicial Center Integrated Database, conditional on attendance at the required 341 meeting of creditors, completion of the required debtor education course, and closure of the bankruptcy case.

Intervention Start Date
2026-10-09
Intervention End Date
2026-12-31

Primary Outcomes

Primary Outcomes (end points)
1. Requests link to Upsolve
a. Definition: A binary variable equal to one if the respondent indicates they would like to be given the link to Upsolve at the end of the survey.
Primary Outcomes (explanation)

Secondary Outcomes

Secondary Outcomes (end points)
1. Interest in bankruptcy
a. Definition: A binary variable equal to one if the visitor reports that they would be “Somewhat likely”, “Likely”, or “Extremely likely” to file for Chapter 7 bankruptcy if they were to have trouble paying their debts.
2. Bankruptcy questionnaire completion
a. Definition: A binary variable equal to one if the respondent completes the Upsolve bankruptcy questionnaire.
3. Bankruptcy filing
a. Definition: A binary variable equal to one if the respondent files for Chapter 7 bankruptcy through Upsolve.
Secondary Outcomes (explanation)

Experimental Design

Experimental Design
Upsolve will send survey invitations via email to individuals who (i) completed the bankruptcy screener on Upsolve’s website, (ii) were deemed qualified to file, (iii) created an Upsolve account, and (iv) did not file. The survey will first ask respondents about their beliefs about the effects of bankruptcy, then respondents will be randomly assigned to the treatment or control group with equal probability. The “net worth” treatment group learns that (i) 3% of Upsolve users who filed for Chapter 7 bankruptcy surrendered assets and (ii) 97% received a full discharge of eligible debt. Meanwhile, the control group learns that the labor force participation rate was 62% in June 2026.

The labor force participation statistic is from the Federal Reserve Bank of St. Louis. The discharge and surrender statistics are calculated using Upsolve filers successfully linked to the Federal Judicial Center Integrated Database, conditional on attendance at the required 341 meeting of creditors, completion of the required debtor education course, and closure of the bankruptcy case.

Experimental Design Details
Not available
Randomization Method
Respondents will be randomly assigned to treatment with 2/3 probability and to control with 1/3 probability. Treatment assignment will be determined automatically when the respondent reaches the information provision section of the survey.

Two considerations motivate the unequal split. First, our primary outcome is binary, so its within-group variance is p(1−p). We expect a higher link-request rate under treatment, and p(1−p) rises steeply with p when the base rate is low, so outcome variance should be greater among treated respondents than among controls. Neyman allocation — which sets the ratio of group sizes equal to the ratio of their outcome standard deviations — therefore favors assigning more respondents to treatment. Under our assumed control mean of 10% and treatment mean of 20%, the variance-minimizing treatment share is 57%. Second, the treatment provides accurate information about Chapter 7 filing outcomes to respondents already assessed as qualified to file, and Upsolve prefers that a larger share of the sample receives this information. We therefore allocate somewhat more to treatment than variance minimization alone would imply, at a small cost in precision: at an assumed control mean of 10%, the minimum detectable effect is 7.0 percentage points under a 2/3–1/3 split, compared with 6.8 percentage points under a 57%–43% split.
Randomization Unit
Randomization will be at the respondent level.
Was the treatment clustered?
No

Experiment Characteristics

Sample size: planned number of clusters
Upsolve will send survey invitations via email to approximately 275,000 eligible users and expects a response rate of 0.25%, resulting in an expected sample size of approximately 688 respondents.
Sample size: planned number of observations
Upsolve will send survey invitations via email to approximately 275,000 eligible users and expects a response rate of 0.25%, resulting in an expected sample size of approximately 688 respondents.
Sample size (or number of clusters) by treatment arms
We anticipate approximately 459 respondents in the treatment group and 229 in the control group, although this will depend on the response rate.
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Assume a two-sided test with a 5% significance level and 80% power. With a sample size of 688 (0.25% of 275,000), we are powered for a minimum detectable effect of 7.4 percentage points on the rate of requesting a link to Upsolve. This represents 74% of our assumed control mean of 10%. The minimum detectable effect lowers to 5.7 percentage points (114%) if we halve the control mean assumption to 5% and increases to 9.5 percentage points (48%) if we double the assumption to 20%.
Supporting Documents and Materials
IRB

Institutional Review Boards (IRBs)

IRB Name
Harvard University IRB
IRB Approval Date
2025-06-13
IRB Approval Number
IRB25-0648
Analysis Plan

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information