Student Loan Forgiveness

Student debt forgiveness has a much larger MPC than mortgage balance reductions or stock appreciation.

“Compared to mortgage balance reductions (Agarwal et al., 2017) or stock appreciation (Chodorow-Reich et al., 2021; Di Maggio et al., 2020), student debt forgiveness has a much larger MPC.”

• cites Agarwal et al., 2017Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Chodorow-Reich et al., 2021 sourceSupported
• cites Di Maggio et al., 2020Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Consumption and spending can differ dramatically when households respond by making expensive durable purchases.

“When considering consumption responses, an important distinction is the difference between consumption and spending, which can differ dramatically when households respond by making expensive durable purchases.”

• cites Laibson et al., 2022 sourceSupported

Student loan forgiveness leads to decreased earnings.

“For labor market outcomes, we estimate that forgiveness leads to decreased earnings and increased switching across industries, with effects increasing over time.”

Biden's proposed forgiveness would have produced more progressive forgiveness than the actual policy.

“We estimate that Biden's proposal would have produced more progressive forgiveness, with the average forgiven dollar accruing to a borrower with $4,868 in predicted monthly income compared to the $5,562 of the actual policy.”

The marginal propensity for consumption from student loan forgiveness is 0.27.

“Adding up the effects, we estimate a marginal propensity for consumption of 0.27 and a marginal propensity for earnings of -0.49.”

Forgiveness does not increase the rate of job switching.

“While we do not see an increase in the rate of job switching, we see that the nature of job switching changes: more switching across industries and out of public service.”

The marginal propensity for earnings from student loan forgiveness is -0.49.

“Adding up the effects, we estimate a marginal propensity for consumption of 0.27 and a marginal propensity for earnings of -0.49.”

Among hourly workers, forgiveness causes a drop in hours worked.

“Finally, among hourly workers, we estimate a drop in hours worked, where the total earnings drop comes half from an hours reduction and half from a wage reduction.”

The increase in other debt after student loan forgiveness is primarily driven by mortgage debt.

“This increase is primarily driven by mortgage debt.”

The MPX is 27 times the monthly increase in permanent income.

“The MPX is 27 times the monthly increase in permanent income, which mostly reflects the up-front increase in durables balances.”

Six percent of borrowers in the sample received forgiveness since March 2021.

“We estimate that 6% of borrowers in our sample have received forgiveness since March 2021, with an average of $32,000 discharged.”

Borrowers transfer 0.64 of the annuitized shock into the current period.

“We combine our MPC and MPE estimates, along with adjustments for down payments, labor income taxes, and novel data on the change in borrowers' expectations of forgiveness, and estimate that borrowers transfer 0.64 of the annuitized shock into the current period.”

Forgiveness targets borrowers with $115 higher predicted monthly earnings than non-forgiven borrowers, after controlling for age and when the borrower took out loans.

“Much of this difference reflects that forgiven borrowers are older and at the higher-earning part of their lifecycle, but when we control for age and when the borrower took out loans, a difference of $115 remains.”

Forgiven borrowers have $703 higher average monthly predicted income than student loan borrowers who did not receive forgiveness.

“Relative to student loan borrowers who did not receive forgiveness, forgiven borrowers have $703 higher average monthly predicted income.”

The MPX mostly reflects the up-front increase in durables balances.

“The MPX is 27 times the monthly increase in permanent income, which mostly reflects the up-front increase in durables balances.”

Among borrowers subject to a payment pause, borrowers transfer 0.82 of the annuitized shock into the current period.

“When we focus, however, on the subset of borrowers who were subject to a payment pause and thus forgiveness is primarily a wealth shock, we estimate that borrowers transfer 0.82 of the annuitized shock into the current period.”

Under the actual policy, the average forgiven dollar accrued to a borrower with $5,562 in predicted monthly income.

“We estimate that Biden's proposal would have produced more progressive forgiveness, with the average forgiven dollar accruing to a borrower with $4,868 in predicted monthly income compared to the $5,562 of the actual policy.”

Student loan forgiveness has little effect on non-student loan delinquencies.

“We find little effect on loan delinquencies, other than for student loans.”

The extensive margin auto loan effect is counteracted by reduced balances on pre-existing debt.

“We also see an extensive margin effect on whether borrowers have any auto loan, which is counteracted by reduced balances on pre-existing debt.”

The gap in predicted earnings between forgiven borrowers and individuals who did not attend college is even larger than the gap with the general credit-history population.

“Unsurprisingly, the gap is even larger when comparing forgiven borrowers to individuals who did not attend college.”

Among previously defaulted borrowers, forgiveness leads to earnings increases.

“When we focus on previously defaulted borrowers, we find earnings increases, consistent with Di Maggio et al. (2019).”

• cites Di Maggio et al. (2019)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The decrease in monthly earnings exceeds $75 in the sixth month post-forgiveness.

“We estimate drops in monthly earnings of $44 (or 2.3%) pooled over the first six months post-forgiveness, with the decrease exceeding $75 in the sixth month.”

Borrowers experiencing forgiveness increase credit card borrowing by $220 over the six months following forgiveness.

“Borrowers experiencing forgiveness increase mortgage borrowing by $2,300, auto loan borrowing by $230, and credit card borrowing by $220 over the six months following forgiveness.”

Student loan forgiveness leads to increased switching across industries.

“For labor market outcomes, we estimate that forgiveness leads to decreased earnings and increased switching across industries, with effects increasing over time.”

Forgiveness increases switching out of public service.

“While we do not see an increase in the rate of job switching, we see that the nature of job switching changes: more switching across industries and out of public service.”

The average amount discharged among forgiven borrowers in the sample is $32,000.

“We estimate that 6% of borrowers in our sample have received forgiveness since March 2021, with an average of $32,000 discharged.”

The large mortgage effect is driven primarily by increases in the probability borrowers had any mortgage rather than balance changes for pre-existing mortgages.

“This large mortgage effect is driven primarily by increases in the probability borrowers had any mortgage rather than balance changes for pre-existing mortgages.”

Forgiven borrowers have $504 higher average monthly predicted earnings than the general population of individuals with a credit history.

“Forgiven borrowers have $504 higher monthly predicted earnings on average, which drops to $193 when adding lifecycle controls.”

Forgiveness causes a small drop in whether borrowers are in the employment data.

“We estimate a small drop in whether the borrowers are in the employment data, though when we condition on a balanced sample of individuals always in the data, our intensive margin estimates are very similar to the unconditional estimates.”

The expenditure associated with the consumption response is concentrated right after forgiveness.

“On the other hand, the expenditure associated with the consumption is concentrated right after forgiveness.”

Borrowers experiencing forgiveness increase auto loan borrowing by $230 over the six months following forgiveness.

“Borrowers experiencing forgiveness increase mortgage borrowing by $2,300, auto loan borrowing by $230, and credit card borrowing by $220 over the six months following forgiveness.”

Forgiven borrowers are no more likely to move to higher earnings metropolitan areas.

“Forgiven borrowers are also no more likely to move to higher earnings metropolitan areas and credit scores hardly move.”

For each dollar of student loan forgiveness, other debt increases by 8.8 cents.

“For each dollar of student loan forgiveness, other debt increases by 8.8 cents.”

Forgiveness reduces monthly earnings by $44, or 2.3%, pooled over the first six months post-forgiveness.

“We estimate drops in monthly earnings of $44 (or 2.3%) pooled over the first six months post-forgiveness, with the decrease exceeding $75 in the sixth month.”

Under Biden's proposal, the average forgiven dollar would have accrued to a borrower with $4,868 in predicted monthly income.

“We estimate that Biden's proposal would have produced more progressive forgiveness, with the average forgiven dollar accruing to a borrower with $4,868 in predicted monthly income compared to the $5,562 of the actual policy.”

Forgiveness has an extensive margin effect on whether borrowers have any auto loan.

“We also see an extensive margin effect on whether borrowers have any auto loan, which is counteracted by reduced balances on pre-existing debt.”

Forgiveness increases credit card debt on both the extensive and intensive margins.

“For credit card debt, we see increases on both the extensive and intensive margins.”

Among hourly workers, the total earnings drop comes half from an hours reduction and half from a wage reduction.

“Finally, among hourly workers, we estimate a drop in hours worked, where the total earnings drop comes half from an hours reduction and half from a wage reduction.”

The labor market effects of student loan forgiveness increase over time.

“For labor market outcomes, we estimate that forgiveness leads to decreased earnings and increased switching across industries, with effects increasing over time.”

Forgiveness targets borrowers with $193 higher predicted monthly earnings than the general population of individuals with a credit history, after adding lifecycle controls.

“Forgiven borrowers have $504 higher monthly predicted earnings on average, which drops to $193 when adding lifecycle controls.”

Earnings drops are larger for individuals initially employed in public service jobs.

“and indeed we estimate larger earnings drops for individuals initially employed in public service jobs.”

Borrowers experiencing forgiveness increase mortgage borrowing by $2,300 over the six months following forgiveness.

“Borrowers experiencing forgiveness increase mortgage borrowing by $2,300, auto loan borrowing by $230, and credit card borrowing by $220 over the six months following forgiveness.”

Consumption effects of forgiveness are larger for young workers.

“Consumption effects are larger for young workers.”

Credit scores hardly move after forgiveness.

“Forgiven borrowers are also no more likely to move to higher earnings metropolitan areas and credit scores hardly move.”

These borrowers earn an average of $900 (19%) more per month than the average student loan borrower.

“These borrowers earn an average of $900 (19%) more per month than the average student loan borrower.”

The mean dollar of forgiveness under the approximated Biden proposal accrues to a borrower with $3,874 in predicted monthly earnings, for a difference of $2,958.

“The mean dollar of forgiveness under the approximated Biden proposal accrues to a borrower with $3,874 in predicted monthly earnings, for a difference of $2,958.”

The actual earnings gaps tend to be larger than the predicted earnings gaps.

“The actual earnings gaps tend to be larger than the predicted earnings gaps.”

The mean dollar of forgiveness under the implemented forgiveness policy accrues to a borrower with $6,832 in actual monthly earnings.

“The mean dollar of forgiveness under the implemented forgiveness policy accrues to a borrower with $6,832 in actual monthly earnings.”

Borrowers who receive forgiveness tend to be slightly older than the median borrower.

“Borrowers who receive forgiveness tend to be slightly older than the median borrower, consistent with the details of forgiveness policies.”

Forgiveness leads to a substantial effect on student loan balances, with the average balance dropping by $31,800 following forgiveness.

“We find that forgiveness leads to substantial effects on student loan balances, with the average balance dropping by $31,800 following forgiveness.”

Relative to all individuals with a credit history, forgiven borrowers are even more positively selected on income: a mean monthly predicted earnings difference of $193 with lifecycle controls and $504 without.

“We find that forgiven borrowers are even more positively selected on income relative to this comparison group: a mean monthly predicted earnings difference of $193 and $504, with and without lifecycle controls, respectively.”

The mean dollar of forgiveness under the approximated Biden proposal accrues to a borrower with $4,868 in predicted monthly earnings, for a difference of $694.

“The mean dollar of forgiveness under the approximated Biden proposal accrues to a borrower with $4,868 in predicted monthly earnings, for a difference of $694.”

Borrowers who receive forgiveness have higher earnings (both unconditional and conditional on positive earnings) on average than individuals in the other two groups.

“As we saw in Table 2, borrowers who receive forgiveness have higher earnings (both unconditional and conditional on positive earnings) on average than individuals in the other two groups.”

All the estimates on student loan outcomes are statistically significant at the 1% level.

“All the estimates are statistically significant at the 1% level.”

The mean dollar of forgiveness under the implemented forgiveness policy accrues to a borrower with $5,562 in predicted monthly earnings.

“The mean dollar of forgiveness under the implemented forgiveness policy accrues to a borrower with $5,562 in predicted monthly earnings.”

The paper identifies 58,538 borrowers (about 6% of borrowers in the sample) who receive student loan forgiveness in the TransUnion data.

“With these filters, we identify 58,538 borrowers (about 6% of borrowers in our sample) who receive student loan forgiveness in the TransUnion data.”

For the six months following forgiveness, mortgage, auto, and credit card balances increase by approximately $2,800 in total.

“For the six months following forgiveness, we estimate increases in mortgage, auto, and credit card balances totaling approximately $2,800.”

The mean difference in predicted monthly earnings (in December 2019) between forgiven borrowers and non-forgiven borrowers is $703.

“The mean difference in predicted monthly earnings (in December 2019) between forgiven borrowers and non-forgiven borrowers is positive and substantial: $703 (column 1).”

The gap in predicted monthly earnings is slightly smaller among workers in public service oriented industries.

“In the remaining columns we show that the gap is slightly smaller among workers in public service oriented industries.”

Forgiven borrowers are more likely to be employed in public service jobs.

“Forgiven borrowers are also more likely to be employed in public service jobs, which is consistent with the PSLF program contributing to many of the new forgiveness events.”

Payments due drop by approximately $64 following forgiveness.

“Payments due drop by approximately $64, a relatively small number due to many borrowers being subject to the payment pause in much of our estimation window.”

Except around the announcement of Biden's proposed policy, baseline forgiveness expectations are around 25%.

“Except around the announcement of Biden's proposed policy, baseline forgiveness expectations are around 25%.”

Delinquencies drop sharply by 2.4 percentage points following forgiveness.

“We also see a sharp drop in delinquencies, by 2.4 percentage points.”

The estimate of the MPX using the savings from student loan payments is 57.5.

“using the savings from student loan payments is 57.5.”

• AssumptionThe authors state this as an assumption: The calculations so far assume that the changes are unexpected.

For the full sample of workers, forgiveness causes a 0.4 percentage point decrease in the probability of having any earnings reported or being employed.

“For these workers, we estimate a 0.4 percentage point decrease in the probability of having any earnings reported or being employed.”

For the payment pause subsample, the estimates on mortgage, auto, and credit card debt are quite similar to the baseline estimates.

“The estimates are quite similar to our baseline estimates”

Forgiveness causes a drop in monthly earnings of $44, or 2.3% when using log earnings.

“We estimate a drop in monthly earnings of $44, or 2.3% when we use log earnings.”

The MPX is around 100 times the MPC estimates.

“Note that the MPX is around 100 times our MPC estimates.”

The estimate of the monthly MPC using the annuity method is 0.27.

“Our estimate of the monthly MPC using the annuity method is 0.27”

• AssumptionThe authors state this as an assumption: The calculations so far assume that the changes are unexpected.

The main results are driven entirely by borrowers who never defaulted.

“Our main results are driven entirely by the set of borrowers who never defaulted.”

For other forms of credit, there is a statistically significant increase in the number of lines in each credit type.

“For the other forms of credit, we see a statistically significant increase in the number of lines in each credit type.”

Forgiveness causes a decrease in the number of student loan lines.

“As expected, we estimate a decrease in the number of student loans lines.”

Spreading the wealth shock equally every month over 41 years is a savings of around $104 per month.

“Spreading the wealth shock equally every month over 41 years is a savings of around $104 per month.”

Earnings effects are larger for workers in public service.

“We find that earnings effects are larger for workers in public service.”

Borrowers' beliefs that they would receive forgiveness, prior to actually receiving forgiveness, averaged 25%.

“borrowers' beliefs that they would receive forgiveness, prior to actually receiving forgiveness, averaged 25%.”

The age of the typical borrower in the sample is around 39.

“the age of the typical borrower in our sample is around 39”

The paper finds opposite signed results for borrowers who had previously defaulted.

“Interestingly, we find opposite signed results for borrowers who had previously defaulted.”

The estimate of the MPX using the annuity method is 27.4.

“Our estimate of the MPX using the annuity method is 27.4”

• AssumptionThe authors state this as an assumption: The calculations so far assume that the changes are unexpected.

There is no change in the rate at which borrowers have new employment post-forgiveness.

“We do not see any change in the rate at which borrowers have new employment postforgiveness”

The reduction in student loan balances is approximately $32,000 on average.

“the reduction in student loan balances, which is approximately $32,000 on average.”

The estimated MPX, based on the combined increase in mortgage, auto, and credit card balances, and normalized by the reduction in average student loan balances, is around 0.12.

“Our estimated MPX, based on the combined increase in mortgage, auto, and credit card balances, and normalized by the reduction in average student loan balances, is around 0.12.”

• AssumptionThe authors state this as an assumption: The calculations so far assume that the changes are unexpected.

The MPE is -0.88 to -0.97 when considering only the reduction in payment due.

“responses are higher, -0.88 to -0.97, when considering only the reduction in payment due.”

• AssumptionThe authors state this as an assumption: The calculations so far assume that the changes are unexpected.

The effects on labor market outcomes are increasing in magnitude over time.

“we see that the effects on the labor market outcomes are increasing in magnitude over time.”

The effect of forgiveness on earnings is larger for hourly workers than for salaried workers.

“We find larger effect for hourly workers, who are likely more elastic and are more able to request more hours of overtime, or cut hours in response to changed incentives.”

Forgiveness causes a small (1-2 point) decrease in credit scores.

“We estimate a small (1-2 point) decrease in credit scores.”

The dynamic difference-in-difference and two-way fixed effect estimators deliver very similar results for labor market outcomes.

“As with the credit outcomes, the dynamic difference-in-difference and two-way fixed effect estimators deliver very similar results.”

The effects on credit outcomes appear quite stable out to 12 months post-forgiveness.

“In Appendix Table C.4, we show that the effects appear quite stable out to 12 months post-forgiveness”

The average reductions in payment due in Table 4 are smaller, rather than larger, than the evenly-spread wealth shock.

“the average reductions in payment due in Table 4 happens to be smaller, rather than larger, than the evenly-spread wealth shock”

The estimate of the monthly MPC using the flow values of payments is 0.57.

“using the flow values of payments is 0.57.”

• AssumptionThe authors state this as an assumption: The calculations so far assume that the changes are unexpected.

The MPE is around -0.5 when annuitizing the reduction in student loan balances.

“we find an MPE of around -0.5 when annuitizing the reduction in student loan balances”

• AssumptionThe authors state this as an assumption: The calculations so far assume that the changes are unexpected.

The estimate of α for the full sample is 0.64, a bit below 1.

“We estimate α = 0.64, a bit below 1.”

• AssumptionThe authors state this as an assumption: The paper ignores precautionary savings considerations.

The IV estimates of the effect of loan forgiveness on outcomes are generally similar in magnitude to the difference-in-differences estimates.

“Generally, the magnitudes are quite similar to those in Tables 4 and 5”

• AssumptionThe authors state this as an assumption: The exclusion restriction for the IV analysis holds: servicers only affect forgiveness probabilities and not other factors related to borrower outcomes.

When decomposing the total earnings effect for hourly workers into hours and wages, half of the decreased earnings is attributed to reduced hours and half to reduced wages.

“When we decompose the total earnings effect for hourly workers into hours and wages, we attribute half of the decreased earnings to reduced hours and half to reduced wages.”

For hourly workers, there is a drop in the hourly wage.

“For hourly workers, we also see a drop in the hourly wage in column (7) of Table 7.”

Not incorporating baseline forgiveness beliefs lowers α to 0.48.

“The fourth row shows that not incorporating baseline forgiveness beliefs lowers α to 0.48.”

Ever being forgiven reduces student loan delinquency by 0.035 (SE 0.011).

“Ever Forgiven | -35612.402 * * * | -108.030 * * * | -0.035 * * *”

In the dynamic DID specification, forgiveness reduces earnings by $44.28 (SE 11.88) over a 6-month horizon post-forgiveness.

“Dynamic DID | -44.28 * * * | -0.023 * * * | -0.004 * * * | 0.0005 | 0.0005 * * | -0.002 * * | -1.02 * * *”

Respondents who ever receive forgiveness have no statistically significant difference in their beliefs prior to forgiveness in either the short or long-run.

“We find that respondents who ever receive forgiveness have no statistically significant difference in their beliefs prior to forgiveness in either the short or long-run.”

Ever being forgiven reduces student loan payments due by $108.030 (SE 24.614).

“Ever Forgiven | -35612.402 * * * | -108.030 * * * | -0.035 * * *”

Student loan forgiveness increases forgiven borrowers' credit card balances.

“| Dynamic DID | 222.048 * * * | 5.337 * * * | 0.004 * * |”

• Also stated in: § B.2 Data Validation · plain

  “Dynamic DID 0.042 * * * 0.005 * * * 98.254 * * * 1.995 * 0.001”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

In the dynamic DID specification, forgiveness reduces earnings for never defaulters by $55.75 (SE 13.88).

“Dynamic DID | -73.77 * * | -20.61 * * | -51.14 * * * | -31.90 * * | -55.75 * * * | 63.65 * *”

In the dynamic DID specification, forgiveness reduces public service employment by 0.002 (SE 0.001).

“Dynamic DID | -44.28 * * * | -0.023 * * * | -0.004 * * * | 0.0005 | 0.0005 * * | -0.002 * * | -1.02 * * *”

Student loan balances decrease after forgiveness for each of the four most common forgiveness policies (disability, borrower defense, PSLF, and IDR).

“Balances -12144.399* * * -27553.796* * * -41065.647* * * -37550.854* * *”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

In the dynamic DID specification, forgiveness reduces earnings for public service workers by $51.14 (SE 16.85).

“Dynamic DID | -73.77 * * | -20.61 * * | -51.14 * * * | -31.90 * * | -55.75 * * * | 63.65 * *”

In the dynamic DID specification, forgiveness reduces earnings for non-public service workers by $31.90 (SE 14.56).

“Dynamic DID | -73.77 * * | -20.61 * * | -51.14 * * * | -31.90 * * | -55.75 * * * | 63.65 * *”

In the dynamic DID specification, forgiveness reduces hours worked by 1.02 (SE 0.466).

“Dynamic DID | -44.28 * * * | -0.023 * * * | -0.004 * * * | 0.0005 | 0.0005 * * | -0.002 * * | -1.02 * * *”

In the first stage, servicer leniency is positively associated with ever receiving forgiveness, with coefficients of 0.993 (SE 0.007), 0.974 (SE 0.009), and 0.964 (SE 0.010) across columns.

“Servicer Leniency | 0.993 * * * | 0.974 * * * | 0.964 * * *”

In the dynamic DID specification, forgiveness reduces log earnings by 0.023 (SE 0.006).

“Dynamic DID | -44.28 * * * | -0.023 * * * | -0.004 * * * | 0.0005 | 0.0005 * * | -0.002 * * | -1.02 * * *”

Student loan forgiveness increases forgiven borrowers' mortgage payments due.

“| Dynamic DID | 2318.755 * * * | 17.468 * * * | -0.000 |”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

Student loan forgiveness reduces forgiven borrowers' student loan balances.

“| Dynamic DID | -31843.308 * * * | -64.293 * * * | -0.024 * * * |”

• Also stated in: § B.2 Data Validation · plain

  “Dynamic DID -2.046 * * * -0.883 * * * -31778.402* * * -63.134 * * * -0.024 * * *”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

Student loan forgiveness increases forgiven borrowers' auto loan balances.

“| Dynamic DID | 232.043 * * * | 4.804 * * * | 0.000 |”

• Also stated in: § B.2 Data Validation · plain

  “Dynamic DID 0.009 * * * 0.007 * * * -637.807* * * -11.736 * * * -0.000”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

Student loan forgiveness has no statistically significant effect on forgiven borrowers' credit scores.

“Dynamic DID -1.062 -0.001 (0.818) (0.001)”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

In the dynamic DID specification, forgiveness reduces any employment by 0.004 (SE 0.001).

“Dynamic DID | -44.28 * * * | -0.023 * * * | -0.004 * * * | 0.0005 | 0.0005 * * | -0.002 * * | -1.02 * * *”

After forgiveness, forgiven borrowers' hours worked decrease.

“Post Forgiveness -68.57* * * -0.036 * * * -0.010 * * * 0.0006 * * 0.0007 * * * -0.008 * * -1.545 * * *”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

Student loan forgiveness reduces delinquency on forgiven borrowers' student loans.

“| Dynamic DID | -31843.308 * * * | -64.293 * * * | -0.024 * * * |”

• Also stated in: § B.2 Data Validation · plain

  “Dynamic DID -2.046 * * * -0.883 * * * -31778.402* * * -63.134 * * * -0.024 * * *”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

Ever being forgiven reduces student loan balances by $35,612.402 (SE 4247.419).

“Ever Forgiven | -35612.402 * * * | -108.030 * * * | -0.035 * * *”

After forgiveness, forgiven borrowers' probability of being employed decreases.

“Post Forgiveness -68.57* * * -0.036 * * * -0.010 * * * 0.0006 * * 0.0007 * * * -0.008 * * -1.545 * * *”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

Over a 12-month horizon after forgiveness, forgiven borrowers' auto loan balances and payments due increase.

“Dynamic DID 218.211 * * * 4.704 * * * -0.001”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

Student loan forgiveness has no statistically significant effect on forgiven borrowers' probability of relocating.

“Dynamic DID -1.062 -0.001 (0.818) (0.001)”

• Also stated in: § B.2 Data Validation · boosted

  “We estimate the effect of forgiveness on relocation and present the estimates in TableC.3(event study graphs in Figure C.7).We see a precise null effect.25”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

The reduced-form MPX (mortgage+auto+CC balances) estimate is 2772.847 (SE 474.467).

“MPX | 2772.847 * * * | 0.116 * * * | 27.352 * * * | 57.505 * * *”

Most of the additional mortgage properties acquired by forgiven borrowers are in the borrower's original zip code.

“While the additional mortgage line likely reflects a new property purchase, most of these properties are in the borrower's original zip code.”

Student loan forgiveness reduces the payments due on forgiven borrowers' student loans.

“| Dynamic DID | -31843.308 * * * | -64.293 * * * | -0.024 * * * |”

• Also stated in: § B.2 Data Validation · plain

  “Dynamic DID -2.046 * * * -0.883 * * * -31778.402* * * -63.134 * * * -0.024 * * *”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

In the dynamic DID specification, forgiveness reduces earnings for hourly workers by $73.77 (SE 27.04).

“Dynamic DID | -73.77 * * | -20.61 * * | -51.14 * * * | -31.90 * * | -55.75 * * * | 63.65 * *”

Student loan forgiveness increases forgiven borrowers' mortgage balances.

“| Dynamic DID | 2318.755 * * * | 17.468 * * * | -0.000 |”

• Also stated in: § B.2 Data Validation · plain

  “Dynamic DID 0.008 * * * 0.007 * * * -216.917-0.095 -0.000”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

Over a 12-month horizon after forgiveness, forgiven borrowers' credit card balances and payments due increase.

“Dynamic DID 251.042 * * * 6.135 * * * 0.004”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

In the full sample, the MPE is -0.489 (SE 0.194).

“Full sample | 0.272 * * * | -0.489 * * | -0.117 * * | 0.762 * * * | 0.644 * * *”

In the dynamic DID specification, forgiveness increases earnings for ever defaulters by $63.65 (SE 27.63).

“Dynamic DID | -73.77 * * | -20.61 * * | -51.14 * * * | -31.90 * * | -55.75 * * * | 63.65 * *”

In the dynamic DID specification, forgiveness reduces earnings for salary workers by $20.61 (SE 11.05).

“Dynamic DID | -73.77 * * | -20.61 * * | -51.14 * * * | -31.90 * * | -55.75 * * * | 63.65 * *”

Among forgiven borrowers in the no-payment group, student loan balances decrease after forgiveness.

“Dynamic DID -31785.419* * * 3.331 0.003”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

After forgiveness, forgiven borrowers' log earnings decrease.

“Post Forgiveness -68.57* * * -0.036 * * * -0.010 * * * 0.0006 * * 0.0007 * * * -0.008 * * -1.545 * * *”

• AssumptionRests on an assumption: Absent forgiveness, forgiven borrowers and the comparison borrowers (those not yet forgiven or never forgiven in sample) would have followed parallel trends in the outcomes studied.

In the full sample, the MPT is -0.117 (SE 0.047).

“Full sample | 0.272 * * * | -0.489 * * | -0.117 * * | 0.762 * * * | 0.644 * * *”

In the full sample, MPC-MPE is 0.762 (SE 0.204).

“Full sample | 0.272 * * * | -0.489 * * | -0.117 * * | 0.762 * * * | 0.644 * * *”

In the full sample, MPC-MPE+MPT is 0.644 (SE 0.160).

“Full sample | 0.272 * * * | -0.489 * * | -0.117 * * | 0.762 * * * | 0.644 * * *”

Substantial policy uncertainty would have made it difficult for many borrowers to anticipate forgiveness.

“We expect that for many borrowers, the substantial policy uncertainty would have made it difficult to anticipate forgiveness.”

Student loan forgiveness is primarily an unexpected, moderately-sized wealth shock.

“We characterize forgiveness as primarily an unexpected moderately-sized wealth shock.”

Different loan servicers were more or less likely to process loan forgiveness due to administrative procedures or information campaigns.

“Different loan servicers were more or less likely to process loan forgiveness, due to various administrative procedures or information campaigns.”

The consumption response to student loan forgiveness is small.

“On the one hand, the consumption response is small and the reduction in labor earnings implies less economic activity.”

Student debt potentially locks borrowers into jobs based on repayment benefits.

“We especially highlight that student debt has large income effects on labor supply and potentially locks borrowers into jobs based on repayment benefits.”

Student loan forgiveness disproportionately went to higher-income borrowers.

“Importantly, our results suggest that forgiveness disproportionately went to higher-income borrowers.”

The reduction in labor earnings from student loan forgiveness implies less economic activity.

“On the one hand, the consumption response is small and the reduction in labor earnings implies less economic activity.”

A 2016 Government Accountability Office report found that 98 percent of reinstated disability discharges occurred because borrowers did not submit requested documentation, not because earnings were too high.

“they cited a 2016 report by the Government Accountability Office which found that 98 percent of reinstated disability discharges occurred not because earnings were too high, but because borrowers simply did not submit the requested documentation (U.S.Department of Education, 2023; U.S. Government Accountability Office, 2016).”

• cites U.S.Department of Education, 2023Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites U.S. Government Accountability Office, 2016Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The increase in whether borrowers have a loan of any type suggests a fairly high change in durable ownership.

“The increase in whether borrowers have a loan of any type suggests a fairly high change in durable ownership, with a large change in debt and somewhat smaller change in current consumption.”

Loan servicers are quasi-randomly assigned.

“We use the fact that loan servicers are quasi-randomly assigned (Cornaggia and Xia, 2024).”

• cites Cornaggia and Xia, 2024Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Forgiven borrowers are positively selected on income.

“Thus, the forgiven borrowers are positively selected on income.”

The paper validates its inference of forgiveness by comparing it to public statistics released by the Department of Education.

“We validate our inference of forgiveness by comparing it to public statistics released by the Department of Education.”

The unearned income from winning the lottery generates a very similar MPE as student debt forgiveness.

“The most similar shock is perhaps the unearned income from winning the lottery (Imbens et al., 2001), which generates a very similar MPE as student debt forgiveness.”

• cites Imbens et al., 2001 sourceNot checked — no full text of the source

The two empirical strategies yield similar results.

“Reassuringly, our two empirical strategies yield similar results.”

When conditioning on a balanced sample of individuals always in the data, intensive margin estimates are very similar to the unconditional estimates.

“We estimate a small drop in whether the borrowers are in the employment data, though when we condition on a balanced sample of individuals always in the data, our intensive margin estimates are very similar to the unconditional estimates.”

The Biden administration's changes to existing student loan policies led to forgiveness for around 3 million borrowers.

“These changes have since led to forgiveness for around 3 million borrowers (U.S.Department of Education, 2024a).”

• cites U.S.Department of Education, 2024aNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Student debt has large income effects on labor supply.

“We especially highlight that student debt has large income effects on labor supply and potentially locks borrowers into jobs based on repayment benefits.”

The dynamic difference-in-differences design absorbs all time-invariant individual-specific factors that could be correlated with who received forgiveness.

“We thus absorb all time-invariant individual specific factors that could be correlated with who received forgiveness.”

• AssumptionRests on an assumption: The parallel trends assumption holds for the dynamic difference-in-differences design: absent forgiveness, outcomes of forgiven and not-yet-forgiven borrowers would have evolved similarly.

The IV estimates identify the local average treatment effect for compliers-borrowers whose forgiveness outcome depends on their servicer.

“Our estimates will identify the local average treatment effect for compliers -i.e., borrowers whose forgiveness outcome depends on their servicer.”

• AssumptionThe authors state this as an assumption: The exclusion restriction for the IV analysis holds: servicers only affect forgiveness probabilities and not other factors related to borrower outcomes.

The debt forgiveness studied can be thought of as a moderately-sized wealth shock.

“The debt forgiveness we study can be thought of as a moderately-sized wealth shock.”

The Biden proposal was considerably more progressive than the forgiveness that has been implemented.

“The Biden proposal was therefore considerably more progressive than the forgiveness that has been implemented.”

Removing flagged student loan closures for which 75% or more of the initial balance returns removes approximately 46% of initially flagged borrowers.

“This removes approximately 46% of initially flagged borrowers.”

The paper's estimated forgiveness rate lines up closely with the Department of Education's statement that 7% of borrowers have received forgiveness.

“This lines up closely with the Department of Education's statement that 7% of borrowers have received forgiveness.”

Student loan balances in the employment records data are very similar to TransUnion balances.

“Comparing student loan balances to TransUnion, the balances are reassuringly very similar.”

The change in consumption from a reduction in committed payments of one dollar is equivalent to an increase in unearned income of one dollar in the period.

“the change in consumption from a reduction in committed payments of one dollar is equivalent to an increase in unearned income of one dollar in the period”

Removing flagged student loan closures that coincide with unexplained drops in mortgage, auto loan, or credit card debt removes approximately 6% of the initially flagged borrowers.

“This removes approximately 6% of the initially flagged borrowers.”

The paper's aggregate and per capita measured state-level forgiveness closely matches data from the DOE.

“our aggregate and per capita measured state-level forgiveness closely matches data from the DOE.”

Di Maggio et al. (2019) find evidence for debt overhang reducing labor supply for defaulted private student loan borrowers.

“Di Maggio et al. (2019) find evidence for debt overhang reducing labor supply for defaulted private student loan borrowers.”

• cites Di Maggio et al. (2019)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Most forgiveness during the sample period is likely driven by a set of opaque policy changes rather than individuals' actions.

“most forgiveness during our sample period is likely driven by a set of opaque policy changes rather than individuals' actions.”

The opacity of policy changes limits the ability of individuals to sort endogenously into treatment.

“it also limits the ability of individuals to sort endogenously into treatment.”

Most of the difference in progressivity between the Biden proposal and implemented forgiveness is driven by the earnings cutoff of $125,000 rather than variation in balances below $10,000.

“Most of this difference is driven by the earnings cutoff of $125,000 rather than variation in balances below $10,000.”

When borrowers take out their student loans, they are randomly assigned to a servicer.

“We use the institutional feature that when borrowers take out their students loans, they are randomly assigned to a servicer”

• cites Cornaggia and Xia, 2024Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

This extensive margin response is consistent with borrowers using credit to finance consumption of new durables.

“This extensive margin response is consistent with borrowers using credit to finance consumption of new durables.”

The full sample does not provide the cleanest test of the permanent income hypothesis for two reasons.

“Our full sample, though, does not provide the cleanest test of the permanent income hypothesis for two reasons.”

The forgiveness event does not affect credit supply, at least the extent of supply that varies with credit score.

“Thus, the forgiveness event does not affect credit supply, at least the extent of supply that varies with credit score.”

The MPE estimates imply this forgiveness reduced earnings by around $1.4 billion per year.

“our MPE estimates imply this forgiveness reduced earnings by around $1.4 billion per year.”

There is a minimal relationship between the servicer's forgiveness rate and characteristics of the state where the borrower lives.

“Similarly, we show a minimal relationship between this forgiveness rate and characteristics of the state where the borrower lives”

The similarity in IV and DiD estimates suggests that both strategies capture similar treatment effects of loan forgiveness on credit outcomes.

“suggesting that both strategies capture similar treatment effects of loan forgiveness on credit outcomes.”

• AssumptionThe authors state this as an assumption: The exclusion restriction for the IV analysis holds: servicers only affect forgiveness probabilities and not other factors related to borrower outcomes.

The similarity of estimates in the payment pause subsample suggests that most of the policy effects in the full sample are driven by wealth effects rather than liquidity effects.

“which suggests that most of the policy effects in the full sample are driven by wealth effects rather than liquidity effects.”

Since this represents around 7% of outstanding debt, forgiving all student loan debt would increase these aggregate effects by over 10 times.

“Since this represents around 7% of outstanding debt, forgiving all student loan debt would increase these aggregate effects by over 10 times.”

The estimates of the MPE using the annuity method are comparable with Golosov et al. (2023), who also find an MPE around -0.5 among U.S. lottery winners.

“Our estimates of the MPE, using the annuity method, are comparable with Golosov et al. (2023), who also find an MPE around -0.5 among U.S. lottery winners.”

• cites Golosov et al. (2023)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

A borrower's age and credit score are not strongly related to her servicer's (leave-out) forgiveness rate.

“Table C.5 shows that a borrower's age and credit score are not strongly related to her servicer's (leave-out) forgiveness rate.”

The debt overhang effect is much more likely to be relevant for borrowers in default who anticipate some form of future discharge.

“This effect is much more likely to be much more relevant for borrowers in default who anticipate some form of future discharge.”

The smaller average reductions in payment due are likely due to factors like the payment pause, forbearance, and especially IDR.

“likely due to factors like the payment pause, forbearance, and especially IDR.”

The MPX estimate suggests that forgiveness so far has led to an increase in upfront spending of $11.5 billion.

“Our MPX estimate suggests that forgiveness so far has led to an increase in upfront spending of $11.5 billion.”

If households were textbook PIH consumers, this would make the paper's MPCs a lower bound.

“If households were textbook PIH consumers, this would make our MPCs a lower bound.”

The event-study estimates show minimal evidence of pre-trends.

“Figure C.6 shows minimal evidence of pre-trends.”

The TransUnion-based estimates of student loan forgiveness closely match the Department of Education's reported trends and magnitudes of forgiveness, including at the state level.

“We see very similar trends and magnitudes of forgiveness between the DOE reporting and our estimates, with TransUnion numbers lagged those of the DOE.”

• Also stated in: § B.2 Data Validation · plain

  “We again find that our estimates closely match those of the DOE reporting, even at the state level.”

Servicer leniency is not significantly associated with borrower demographics (age, credit score, log household income, political affiliation, sex, race), supporting its use as an instrument.

“Servicer Leniency 9.859 * 11.903 0.040 -0.003”

• Also stated in: § B.2 Data Validation · plain

  “Share Female Share Black Share Asian Share Hispanic Servicer Leniency 0.001 0.004 0.005 0.007”

Had the paper not considered the earnings responses, it would have decisively rejected the permanent income hypothesis.

“Had we not considered the earnings responses, we would have decisively rejected the permanent income hypothesis.”

Koustas et al. (2024) find that households change their repayment behavior and nondurable spending in response to large changes in expectations.

“Indeed, Koustas et al. (2024) do find that households change their repayment behavior and nondurable spending in response to large changes in expectations.”

• cites Koustas et al. (2024)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Because balances on durables exclude down payments, using the increase in balances underestimates spending on durables.

“Balances on durables exclude down payments, which will cause us to underestimate spending using the increase in balances.”

The same process used to identify forgiveness in the employment records data yields nearly identical rates of student loan forgiveness.

“We employ the same process to identify forgiveness in the employment records data which yields nearly identical rates of student loan forgiveness.”

Because the panel is balanced and nearly all groups receive treatment well before the end of the panel, the group weights hardly change across different values of l.

“Because we have a balanced panel and nearly all groups receive treatment well before the end of the panel, the group weights hardly change across different values of l.”

The employment records provide national coverage of employment history and monthly earnings from approximately 150 million unique individuals and 3 million employers.

“The employment records provide national coverage of employment history and monthly earnings from approximately 150 million unique individuals and 3 million employers.”