The looming impact of student loan repayment on the FHA portfolio?

Much much larger than I previously understood

Note: This post is for paid subscribers only for a few days. I will release it publicly on Tuesday morning, but wanted those who pay to support me to have first access particularly because I think this impact is widely unknown/unappreciated.

By my calculations roughly 2.5 million FHA borrowers will be facing a payment shock of on average $625 dollars a month starting sometime between September 2025 and April 2026. The cause will be the forced shift of these borrowers from a SAVE student loan repayment plan (which are still in forbearance and haven’t required payment in 5 years) to the new Repayment Assistance Plan (“RAP”) in the proposed student loan overhaul legislation. That legislation advanced out of the House Education Committee recently and is part of the broader budget reconciliation “big beautiful bill” so while it could face further amendment, it seems likely to become law largely as is sometime this summer.

The likely timeline is:

1. The “Big Beautiful Bill” which includes the student loan overhaul is enacted late July (cant be too much later with X-date in mid/late August)

2. With SAVE officially deprecated, servicers keep borrowers in administrative forbearance until they can implement the new Repayment Assistance Plan

3. Maybe as early as September but more likely near the turn of the year Servicers complete implementations and switch borrowers over. This may happen on different timeframes for different servicers

4. The proposed legislation requires switchover within 9 months so all borrowers switched over to RAP by April 2026.

These 2.5m borrowers have considerably higher incomes on average (101k/yr) vs the FHA borrowers who do not (64k/yr) so they are better positioned to deal with the payment shock but even so it seems likely to have a striking impact, particularly since no student loan payment has been required from these folks for 5 years.. To put the amounts in perspective, the average Principal and Interest payment on the FHA loan for these borrowers is $1361.

No way Out

Can the borrower elect to just not pay the student loan? Yes, but as I pointed out last weekend, if they become delinquent on the student loan they will be shut off from FHA home retention loss mitigation (no partial claims or modifications). Furthermore, they can only avoid payment for a little while, because when the loan goes into default at 270-360 days, wage garnishment will start. With wage garnishment at 15% of discretionary income, the federal government is going to get as least as much as the borrower would have been required to pay in the new RAP plan so it will make no financial sense to default if you have a job where income can be garnished.

FHA was best option for high student loan debt borrowers to get a mortgage

Initially, it didnt jump out at me me that FHA loans would be a magnet for student loan debt, in particular for higher income borrowers with high student loan balances, but when you think about it, if those borrowers wanted wanted to buy a home, and either didnt have or didnt want to put alot down on the home, and needed a program that allows high DTIs to accommodate their high student loan debt. Where else were they going to go, even if their credit and incomes were relatively high. The FHA lowering the imputed debt to 0.5% in August 2021 made it all the more attractive.

The rest of the piece explains in detail how Ive reached this conclusion including all the data Ive used and assumptions I have made. I believe the analysis is solid but as always hold myself out for critique, if you think any of my assumptions are off based please do let me know and if warranted I will adjust the model and update the conclusions.

2.5 million FHA borrowers currently have a student loan enrolled in the SAVE plan

In August 2021, the FHA changed how they counted student loan debt in a borrowers debt to income ratio (“DTI”) lowering the imputed monthly payment for student loans that were paying $0 a month (loans that were in forbearance or deferment) from one percent to one half of one percent of the loan balance. So if a borrower had a $50,000 loan balance but the loan was in forbearance (as they all were in 2021) only $250 would count to the DTI instead of $500. To support that change HUD issued a press release which identified that 45% of FHA borrowers who are first time homebuyers (80% of the FHA borrowers) carried student loan debt. Read it here: https://archives.hud.gov/news/2021/pr21-103.cfm

Granted the % of student loan borrowers in the 20% of FHA borrowers who are not 1st time homebuyers is likely less than 45% but it is also similarly likely that homebuyers with large student loan debts were attracted to FHA loans post summer 2021 both because of higher DTI ranges for FHAs and student loans only counting half as much towards DTI. Accordingly a 40-45% range of FHA borrowers also having student loans looks reasonable.

I confirmed the reasonableness of this number by doing a loan level analysis. Longtime readers may recall my post late last year where I described linking the MBS performance data with the HMDA origination data (Read it here:

). That combined dataset provides me both precise DTI (from the MBS data) and qualifying income (from the HMDA data) on about 3 million of the 6.8 million loans active in GNMA II MBS pools in April. Not all of them but enough to be broadly representative.

For each of those loans I then apply the DTI to the qualifying income at origination and subtract out the housing portion of the monthly debt to get the Total Nonhousing monthly debt contribution to the DTI. The non housing portion of DTI is basically auto loans, credit card minimum payments, other installment loans, child support/alimony and the wild card of student loans. For each loan I estimated a $400 - $450 (depending on the origination year) car loan per borrower on the loan and then a blanket $350 a month per borrower for everything else (CC, installment etc.) to isolate out the student loan contribution to the nonhousing portion of the DTI for each loan. Using those numbers, which I think are reasonable in an aggregate sense, leaves about 43% of the 3 million loans I can calculate on with a student loan. In line with the FHAs reported numbers per that press release.

Applying this percent to the full FHA portfolio yields 43% * 7.8 million loans = 3,354,000 loans with student loans. But critically, how many of those loans are likely to be SAVE plan loans. This matters because non-SAVE plan loans have already faced their payment shock and per my calculations FHA borrowers in non-SAVE plan loans have dramatically lower balances of student loan debt and thus payments (average of ~$100 a month) . Surely there are some FHA borrowers with non-SAVE plan loans who are delinquent/defaulted and it should echo in the MBS performance data soon as fewer loans are partial claimed or modified (delinquent student loan debt means no FHA loss mitigation) and eventually the stress of wage garnishment reflects in more delinquency. But I dont expect to see much from this cohort, There isnt nearly as many of these loans (800k or so) and the avg payment shock ~$100 / month will be much easier to absorb.

How do I determine that nearly 2.5 million of those 3,354,000 FHA borrower student loans are in the SAVE plan? Because I calculate what the monthly payments would be for each borrower under the SAVE plan in vs. the other major student loan plans in 2024 and make the assumption that borrowers likely chose the plan that offered them the lowest payment. SAVE was also generous with forgiveness making it even more attractive.

The following is my detailed methodology.

Determining Loan Balance, Adjusted Gross Income and family size

1. I estimate the student loan balance as 100x (1%) the monthly student loan contribution to the DTI for loans originated prior to August 2021 and 200x (0.5%) the monthly loan amount for loans originated after August 2021. Only loans that originated between 2018 and 2023 are in my dataset since HMDA data is only available for those years currently. To be sure this will tend to overestimate some student loan balances but for the repayment plans where the loan balance factors in to the payment (standard repayment) I cap the monthly payment at the monthly student loan contribution to the non housing portion of the DTI

2. For all borrowers I determine their Adjusted Gross Income (“AGI”) by taking their qualifying year income, increasing it to 2024 levels by growing it per the Social Security Average Wage Index, and the multiplying that 2024 income level by 85% to account for deductions etc. that lower headline income to AGI.

3. I determine the number of children under 17 by assuming borrowers under 25 have 0, borrowers 25-34 have 1, borrowers 35-54 have 2, borrowers 55-64 have 1 and 0 above that. (HMDA data includes age data on the borrowers)

4. I determine family size to be the number of borrowers on the loan + the number of children under 17.

Calculating each plan payment

SAVE - I determine the borrowers discretionary income as their AGI - 225% of the federal poverty line for the family size. I then calculate the monthly payment as 7% (as a blended rate that roughly reflects the split between undergraduate (5%) and graduate (10%) loans out there) of that discretionary income /12 months.

Statutory Income Based Repayment (IBR) - I determine the borrowers discretionary income as their AGI - 150% of the federal poverty line for the family size. I then calculate the monthly payment as 10% (which is conservative since borrowers with any fed student loan balance prior to 2014 would have 15%) of discretionary income / 12 months.

Standard Repayment - I provide the student loan balance, determine the likely term of the loan assuming it is a consolidated loan, apply a presumed interest rate of 6.3% (the average student loan interest rate) and amortize it over the period.

Repayment Assistance Plan (RAP) - I determine the borrowers base payment per the proposed statute which for those unfamiliar with it is 120 dollars/year for borrowers with AGIs less than $10,000, 1% of AGI for AGIs between 10-20k, 2% for AGIS between 20-30k, one percent more for each additional 10k until capping at 10% for AGIs > 100k. I then subtract $50 a month for each of the imputed children and enforce the $10 minimum payment (not really relevant honestly)

Example Loan

DisclosureSequenceNumber: 1517127443 originated in December 2020 with an original principal balance of $223,000, qualifying income of $68000 for 2 borrowers with 2 imputed children with a DTI of 53. PITI on this loan is $1161

Monthly Debt at origination = .(53) * (68000/12) = $3003/month.

Non Housing Monthly Debt = 3003 - 1161 = $1841/month.

I assume 2 cars at 400/month + 700/month (350/borrower for other debt not student loan like credit card minimums, installments etc.) leaving 1841 - 1500 = 341/month as the likely student loan debt amount.

Since this loan was originated prior to August 2021, I assume a student loan balance of $34100

In late 2023/2024 the standard repayment for the loan 34100 amortized over a 20 year term at 6.3% = $250

For IBR I assume the current income is 84695 after increasing it by Social Security’s AWI and calculate an AGI of ~72,000. Poverty level for a family of 4 is ~31,000 * 150% = ~46,500. So discretionary incomes is 72,000 - 46,500 = $25,500. 10% of that is ~2,550/12 = ~$210

For SAVE. ~31,000 * 225% = 69,750. So discretionary income of ~ 3250. 7% of that is 227.5/12 or ~ $19 a month.

Given the choice between $250 a month, $210 a month, and $19 a month I assume the borrowers chose SAVE.

Under RAP however, Since AGI is ~72,000 this borrower will have a payment of 7% of that 72,000*.07 = 5040/12 = 420 and less 100 (50 dollars for each kid) = $320

This borrower should probably switch to the existing IBR plan now while they still can before the legislation gets enacted and they are forced into RAP and trapped in it.

And what if the borrower doesn’t pay. Once wage garnishment kicks in, the borrower will face an even larger hole in their monthly budget until they catch up on the student loan arrears. For that loan $84,695 * .15 = 12704.25 *.78 (assuming 22% for deductions like taxes) = 9909 / 12 = a whopping $825 a month.

Getting to 2.5m FHA borrowers in SAVE

Every borrowers situation is different of course and changes to AGI, family size and loan balance changes the calculus on each calculation but generally, if the loan balance was very low, it likely made sense for borrowers to stay in the standard plan, otherwise it vastly favored the SAVE plan.

After calculating likely student debt load and payments under the various plans. 994,525 of the 2,934,594 HMDA linked loans in my dataset are likely to be in SAVE. Scale that number to the full 7.8m FHA portfolio and you get ~2.5m loans.

Key Assumptions

This analysis rests on some key assumptions, if any of them are materially wrong, then the analysis will likewise be materially off.

1. FHA portfolio has 40-45% of loans with student debt - Evidence from the 2021 FHA press release and the loan level data analysis seems to support

2. $400-450 per car and $350 /month (per borrower) for all other non housing debt (credit card min payments, installment loans, child support) is a reasonable estimation. If its too low vs. reality it will inflate the imputed student loan debt and mean I have over identified FHA loans in SAVE

3. That the student loan overhaul does in fact become law this summer without being materially amended/changed again later in the reconciliation process.

Since the RAP payments are formulaically tied to income, my calculation of the avg payment shock should hold firm. The income assumptions (increasing it from origination by the Social Security Average Wage Index)) should not be far off. So its mostly a question of did I get the number of FHA loans likely to have student loans in SAVE repayment plan correct.

Conclusion

If those assumptions and my analysis are correct, then the student loan default data thus far has largely been irrelevant towards the FHA portfolio. But with 2.5m borrowers facing an average $625 a month payment shock sometime in the next 9-10 months, that reckoning is coming.

Thanks for reading and your continued support,

John