Savings and consumption responses to student loan forbearance

See analysis plan.

Participants read a scenario where the US government gives an unexpected $300 cash payment to a specific population. They then read a description of their balance sheets when they receive the cash payment, which includes income, savings account balances, credit card debt, and student loan debt. I quote the interest rate on the savings account, credit card debt, and student loan debt as an annual percentage rate. I then ask participants to indicate how much of their cash payment they would use to increase spending / savings in the following categories 12 months after receiving the payment:
• Durable goods spending (with description)
• Non-durable goods spending (with description)
• Other spending.
• Increased savings account balances.
• Increased credit card debt payments.
• Amount left unspent in checking account after 12 months.

For each participant, the key dependent variables are the amount of total spending allocated towards each category.

The experimental variation involves showing participants four different descriptions which vary (i) the population that the government targets with the payments; and (ii) the interest rate on student loans. For (i), the target population is described as either “all US households” or “households with student debt.” For (ii), the APR on student loans is quoted as either 0% or 7%.

For heterogeneity analysis, I will also collect the following information:
• Whether the person made payments on federal student debt after the federal student loan pause began in March 2020.
• Answers to a financial sophistication instrument. I describe a scenario where participants have two credit cards with interest-bearing balances. One card has a lower balance but higher APR, and the other card has a higher balance but a lower APR. I then ask participants how they would allocate a fixed budget towards repaying each card. I classify borrowers as “making a financial mistake” if they put any of their budget towards the low APR card.

Finally, I will collect demographic information on age, household income, and amount of debt in various categories.

2023-03-07

2023-03-22

For each participant, the key dependent variables are the amount of total spending allocated towards each category in the intervention described above.

For heterogeneity in the experiment, I will also collect the following information:
• Whether the person made payments on federal student debt after the federal student loan pause began in March 2020.
• Answers to a financial sophistication instrument. I describe a scenario where participants have two credit cards with interest-bearing balances. One card has a lower balance but higher APR, and the other card has a higher balance but a lower APR. I then ask participants how they would allocate a fixed budget towards repaying each card. I classify borrowers as “making a financial mistake” if they put any of their budget towards the low APR card.

Finally, I will collect demographic information on age, household income, and amount of debt in various categories.

In addition to running the experiment described above, I will also analyze data on whether and why borrowers continued to make payments on their federal student debt after the start of the federal student loan forbearance program. I also include questions to evaluate borrower familiarity with program terms, and subjective explanations for behavior that is not optimal from the perspective of standard lifecycle savings and consumption models.

I will conduct two types of analysis with this additional data. First, I will examine program familiarity, how stimulus checks were spent, and performance on the financial sophistication instrument, splitting the sample based on whether or not borrowers with federal student debt continued to make payments once forbearance began. Second, I will report subjective justifications for continuing to make federal student loan payments after forbearance, as well as subjective justifications for non-optimal behavior, for borrowers who continued to make federal student loan payments after forbearance began.

This analysis is descriptive, so I will not pre-register any specific hypotheses.

See analysis plan.

There are four treatment groups: (all US households, households with student debt) x (0% interest rate, 7% interest rate). Participants will be assigned to one of these four experimental conditions.

I will rely on randomization software in the Qualtrics survey platform. Randomization will aim for an equal number of people in each treatment group.

Randomization is at the person level. Treatment is not clustered.

No

800 people.

800 observations.

200 people in each treatment condition.

I conducted power calculations for versions 1-3 for two outcomes without controls: (i) fraction spent on student loans; and (ii) fraction spent on total spending. Standard deviations are based on pilot results and results in Coibion, Gorodnichenko and Weber (2020). Versions 1 and 2: • Fraction spent on student loan differences: o With control group mean of 0.08 and standard deviation of 0.28, >80% power to detect effects of >6pp. • Fraction spent on total spending: More uncertainty over standard deviations, so conduct sensitivity. o Optimistic case: Control group mean of 0.24, standard deviation of 0.28, >80% power to detect effects of >6pp. o Pessimistic case: Control group mean of 0.24, standard deviation of 0.4, >80% power to detect effects of >8pp. Version 3: • Fraction spent on student loan differences: o With control group mean of 0.08 and standard deviation of 0.28, >80% power to detect effects of >8pp. • Fraction spent on total spending: o Optimistic case: Control group mean of 0.24, standard deviation of 0.28, >80% power to detect effects of >8pp. o Pessimistic case: Control group mean of 0.24, standard deviation of 0.4, >80% power to detect effects of >12pp.

Analysis plan - Forbearance survey