Experimental Design
The partnering bank invites by e-mail all customers with an outstanding mortgage who can be contacted electronically and have consented to such contact (around 70,000 customers are eligible; the bank sends invitations 2026-09-03 and sends a reminder on 2026-09-15). Participation is voluntary and compensated with a voucher.
Participants first answer questions about their property and mortgage. Participants who refinanced before mid-2022, that is, in the low-rate period, report the terms of and reasoning behind that refinancing and do not take part in the information experiment. All other participants work through a hypothetical refinancing scenario: the rate-fixation period of a €110,000 mortgage ends, the rate falls from 3% to 1%, and the bank’s standard offer lowers the monthly payment by €100 while leaving the balance and the date of full repayment unchanged.
Before making their choice, we randomly assign participants to one of two arms with equal probability. The control group receives neutral definitions of loan terms: the monthly payment comprises an interest and a repayment component, the interest rate in the scenario is 1% per year, and the remaining term is the time until the loan is fully repaid. The treatment group instead receives information that additional repayment (a higher monthly payment or a one-time repayment) earns a fixed return equal to the loan rate of 1% per year, that money used for repayment forgoes any potentially higher return available elsewhere, and that it is tied up in housing wealth rather than available for expenses or emergencies. Both groups answer a comprehension question on their respective text and receive the correct answer as feedback. Participants then choose whether to accept the bank’s offer, change the monthly payment (from €100 below to €200 above the offer), and/or make a one-time repayment out of their savings (up to €30,000), with the implied payment and repayment date displayed live. They subsequently state the reasons for their choices in an open-ended question, indicate whether the level of the new rate, access to the money, and alternative uses of the money spoke for or against additional payments, rank the most important reasons, and report how they would adjust consumption, savings, investment, and hours worked.
If faster repayment reflects a mistake, the treatment should reduce additional payments, shift stated reasoning toward opportunity costs and liquidity, and shift intended uses of the freed-up money toward savings and investment. If faster repayment reflects a preference, choices should not respond to the information.
In a second, smaller experiment shown to all participants, respondents choose between using a freed-up €100 per month from a refinancing at a lower rate for faster repayment or for a savings product. We randomly assign each respondent to one of three products with equal probability: a liquid deposit at the loan rate (1%), a liquid deposit at the loan rate plus one percentage point (2%), or a diversified equity fund savings plan with a higher but uncertain expected return. The 2% deposit strictly dominates faster repayment. The share preferring repayment across the three arms measures how sensitive the repayment preference is to the return and liquidity of the alternative.
The survey closes with a mirror scenario in which the rate rises, questions on financial traits and interest-rate perceptions, a hypothetical income-windfall allocation, and background questions on income and the balance sheet. Bank data allow us to relate stated choices and reasoning to actual refinancing behavior and to study heterogeneity in treatment effects.