A budget analysis of a centralized loan management platform
Executive Summary
The federal government holds more than $5 trillion in direct loans and loan guarantees across more than 100 credit programs at some twenty agencies. Each agency procures, operates, and modernizes its own loan systems, and no common data standard connects them. The government cannot readily assess its own overall credit exposure, compare performance across programs, or determine whether an applicant denied at one agency has applied at another.
The Federal Loan Systems Modernization Act of 2026 would authorize a shared lending platform built on commercially available loan-management technology, with GSA responsible for acquisition and OMB for coordination. Program authorities stay where Congress placed them. What consolidates is the back office, meaning intake, underwriting workflow, servicing, document management, and fraud detection. Operations are funded by a cost-recovery fee agencies pay from the accounts they already use to run their existing systems.
The bill should score at close to zero. Enactment requires a small discretionary appropriation, plausibly in the low tens of millions over five years, and the fee moves money between federal accounts, so it nets out. That, though, significantly undercounts what the bill would save. Consolidating a civilian loan-system IT base of roughly $0.5 billion to $1.5 billion a year at a 30 percent rate produces about $250 million to $300 million annually once the platform matures, and $1.0 billion to $1.5 billion over ten years. Past efforts at federal data center and payroll consolidations support both the rate and the method. Agencies are already funding this work separately, at FHA, VA, and USDA.
Reduced fraud increases actual savings considerably. In federal lending, fraud occurs mostly at origination, where the pandemic lending programs demonstrated the importance of intake controls. Fraud prevention could impact a future score in a number of ways, by lowering default assumptions on new cohorts, for instance, or by reducing upward subsidy reestimates on loans already made. Both require program-level evidence that no credit program can produce today. Common data across programs would produce it across the government’s vast loan portfolio.
1. Introduction
The federal government is one of the largest creditors in the world. Its direct loans and loan guarantees exceed 5 trillion dollars, a portfolio larger than the assets of the largest bank in the United States. That portfolio finances farms, homes, small businesses, students, exports, and energy projects, and for millions of Americans it is the only source of credit available on affordable terms. The Department of War has entered this space as well, through the Office of Strategic Capital, which makes direct loans and guarantees supporting critical technologies and the defense industrial base.
The portfolio comprises more than 100 credit programs at some twenty agencies, administered on back-office systems that are aging and fragmented, several of them decades old. Each agency develops, procures, and maintains its own loan systems. There are no common data standards across them. As a result, the government cannot easily produce a reliable picture of its own credit exposure, compare performance across programs, or determine whether an applicant at one agency has been denied, flagged, or defaulted at another.
| Five federal credit programs, by size | |
| FHA Mutual Mortgage Insurance Fund | $1.6 trillion in mortgages outstanding, end of 2025 |
| Education Direct Loan program | $1.44 trillion in principal outstanding, 2025 |
| FHA General and Special Risk Insurance | $153 billion in guarantees outstanding, end of 2025 |
| VA housing loan guarantees | $206 billion guaranteed on 528,340 loans in FY2025 |
| USDA Farm Service Agency loan portfolio | $36 billion outstanding to more than 123,000 farmers and ranchers |
| Source: OMB, Analytical Perspectives, Fiscal Year 2027, Chapter 4, Credit and Insurance. | |
Senators Marsha Blackburn (R-TN) and Maggie Hassan (D-NH), and Representatives Brad Finstad (R-MN) and Raja Krishnamoorthi (D-IL), introduced the Federal Loan Systems Modernization Act of 2026 in March 2026 as S. 3980 and H.R. 7789, respectively. The bill would address that condition by authorizing a shared federal lending platform based on commercially available loan-management technology. Members of both parties have taken an interest in the bill because the problem is not political. No one defends the current arrangement. The obstacle is that the congressional budget process does not capture what the arrangement actually costs.
The analysis that follows reaches three conclusions. The bill would score at close to nothing, because the fee that funds the platform moves money between federal accounts. The operating savings are real and run to hundreds of millions of dollars a year, though current scoring conventions do not record them. And preventing fraud at loan origination reduces mandatory spending through the annual subsidy reestimate, which is the one route by which any of this could appear in a future score.
Two questions follow. How would the Federal Loan Systems Modernization Act score under current budget rules, and what does that score leave out?
2. What the bill does
The Act authorizes a shared platform based on commercially available loan-management technology, with the General Services Administration responsible for planning and acquisition and the Office of Management and Budget responsible for coordination across agencies. The design follows the pattern of federal payroll consolidation rather than a single agency-wide system procurement. Provider agencies operate the service on a reimbursable basis, and the platform matures over time into a shared-services marketplace.
The bill directs a six-month migration sequencing plan and applies a migrate-or-justify standard to agencies that administer loans. An agency that concludes the platform does not fit its program must explain why. The standard preserves agency judgment while ending the default assumption that every credit program requires its own technology stack.
The Act preserves each agency’s program authorities in full. Eligibility rules, underwriting criteria, interest rates, terms, and servicing policy remain where Congress placed them. What the bill modernizes is the back office, meaning application intake, underwriting workflow, servicing, document management, and fraud detection. An agency that migrates continues to run its own program under its own statute, using shared infrastructure to do it.
Ongoing operations are funded through a discretionary remittance fee applied to each federal loan serviced through the platform. Agencies pay the fee from the accounts they already use to operate their loan systems. The fee is set to recover the cost of the service and no more.
3. How a bill like this scores
In plain terms, the bill asks Congress for a modest appropriation to plan the platform, set common data standards, and oversee the migration. After that, agencies pay a fee to use the platform instead of paying to operate their own loan systems. The fee moves money from one federal account to another, so the total the government spends does not go up, and we anticipate the Congressional Budget Office would record a cost close to zero. But, that is a statement about what the budget rules count, not what the bill would actually save.
CBO sorts the effects of a bill into three categories.
- Mandatory spending. Spending that occurs automatically once a law is enacted, without any further action by Congress.
- Revenues. Taxes, duties, and similar collections.
- Discretionary spending. Spending that occurs only if Congress provides money for it in a later appropriations bill.
A bill that authorizes an activity without appropriating money for it produces discretionary costs, and CBO estimates those costs subject to appropriation, which is a way of saying the money will be spent only if Congress decides to spend it.
Federal credit programs occupy a particular place within that structure. Under the Federal Credit Reform Act of 1990 (FCRA), the budget records the estimated lifetime subsidy cost of a loan or guarantee at the time it is made, rather than the cash moving in and out each year. For most credit programs that subsidy cost is appropriated, which means it competes for dollars within the 302(b) allocations that govern discretionary spending. Only a small number of credit programs are mandatory. Many guarantee programs carry a subsidy cost of zero or less, because the fees borrowers pay are set to cover expected losses, and those programs require no subsidy appropriation at all.
What happens after a loan is made follows a different path. FCRA requires agencies to reestimate the subsidy cost of each annual cohort of loans, every year, for as long as loans in that cohort remain outstanding, to reflect how the loans are actually performing. When actual losses exceed the original estimate, the resulting upward reestimate is paid from permanent indefinite budget authority provided by section 504(f) of the Act. Agencies do not request an appropriation to cover it, and that is true of discretionary and mandatory credit programs alike. Spending that occurs automatically, without further action by Congress, is mandatory spending. The budgetary consequences of loan performance therefore impact the budget on the mandatory side, whatever the classification of the program that made the loan.
The predicted score follows from that structure. Implementing the Act would cost a small amount of discretionary money, plausibly in the low tens of millions of dollars over five years, covering GSA planning and acquisition support, governance, development of data standards, and oversight. The Act would have no net effect on mandatory spending or revenues, provided the fee the legislation authorizes is paid by agencies and set at cost recovery.
The treatment of the remittance fee accounts for the second half of that conclusion. Agencies would pay the fee into a revolving fund that charges only what the service costs. The fee is an outlay for the agency paying it and a collection for the fund receiving it. Because both sides of the transaction occur inside the federal budget, they cancel each other, and the net effect on the deficit is approximately zero. CBO refers to a collection of this kind as an offsetting collection, meaning a receipt credited against spending rather than counted as revenue. Its explanation of the term appears in CBO publication 57660.
The reason the score misses the savings is a feature of conventions rather than a flaw in the analysis. The bill does not lower any appropriation. CBO does not credit efficiency savings that cannot be tied to a specific reduction in spending, because an agency that spends less on one activity may spend the difference on another. An agency that reduces the cost of operating its loan systems keeps that difference within its own budget unless Congress reduces its appropriation. The savings described in the next section are real, though, even if they fall outside what CBO conventions record.
That structure creates the one exception to the conclusion above. Improved loan performance does produce savings the budget records, through two routes. A lower default assumption lowers the subsidy rate on new cohorts, which reduces the appropriation an agency needs and frees room within its discretionary allocation. Better performance on loans already made produces smaller upward reestimates, or downward ones, which reduces mandatory outlays directly. CBO would need verifiable program-level evidence to attribute either effect to a specific control, and evidence of that nature is sparse today. Section 5 addresses why, and what the Act would do about it.
4. The savings the bill would achieve
The savings from consolidating loan systems can be estimated with reasonable discipline, using a documented base, a rate drawn from federal shared-services experience, and precedents from two comparable consolidations. The estimate has three steps.
- The base. Total federal loan-system information technology, covering operations and modernization across the whole credit portfolio, costs about $2 billion to $3 billion dollars a year. The figure is corroborated by loan and credit systems accounting for roughly 3 to 4 percent of civilian federal IT spending. The base covers civilian agencies. Defense lending is excluded from it.
- The carve-out. Excluding Education and Federal Student Aid loan servicing, which operates as its own program and was recently re-contracted, the consolidatable base is about 0.5 to 1.5 billion dollars a year. It includes the Department of Veterans Affairs, the Federal Housing Administration, the Department of Energy’s Loan Programs Office, the Export-Import Bank, the Small Business Administration, the Department of Agriculture, and Ginnie Mae.
- The rate. The estimate assumes savings of 30 percent of the consolidatable base, the top of OMB’s documented shared-services adoption range. The top of the range is the appropriate assumption here because the duplication is unusually complete. Every agency in the carve-out separately procures, operates, and modernizes its own stack, none of them share a data standard, and each is independently replacing systems that are degrading while policymakers rely on them more heavily to move money to constituencies. The duplicated spending underway right now appears in Section 6.
Two comparable consolidations give the approach a defensible basis. Federal data center consolidation delivered $6.24 billion in verified savings and cost avoidance between 2012 and 2020, according to GAO. Federal payroll consolidation, the model the Act follows most closely, projected savings of about 1 billion dollars over ten years, and it consolidated a function considerably simpler than federal lending.
At 30 percent, savings amount to about $250 million to $300 million a year once the platform matures, and roughly $1 billion to $1.5 billion over ten years. Those figures are net of implementation and transition costs, and they are back-loaded behind a three-year migration. It’s notable that a bill CBO would score at close to zero produces savings in the hundreds of millions of dollars a year.
| The estimate at a glance | |
| Total federal loan-system IT | About $2 billion to $3 billion a year |
| Consolidatable base, civilian agencies | About $0.5 billion to $1.5 billion a year |
| Assumed savings rate | 30 percent, the top of the documented shared-services range |
| Annual savings once mature | About $250 million to $300 million |
| Ten-year savings | About $1.0 billion to $1.5 billion, net of transition costs |
| Excludes Education and Federal Student Aid loan servicing, and excludes defense lending. | |
5. Fraud prevention
The estimate in Section 4 covers the cost of operating loan systems. It does not include the money the government loses to fraud in the programs those systems support. Those losses are an even bigger problem, but on that modern loan technology is designed to prevent. The figures in this section are not added to the 250 to 300 million dollars a year estimated above. They describe a benefit the government would capture whether or not the budget process records it.
Fraud in federal lending occurs predominantly at origination, meaning the period running from application through underwriting to disbursement. An applicant misstates identity, eligibility, income, collateral, or the existence of the enterprise being financed, and the loan is approved on those representations. Misrepresentation during underwriting, such as fabricated documentation of a disaster loss, falls in that same period. Some fraud does occur later, mostly identity takeover and false claims for discharge or forbearance, but the great majority of the exposure is created before the money is loaned. The controls that prevent fraud are employed primarily at application, which is the function this legislation modernizes most directly, and a fraudulent loan stopped at intake is a loss the government avoids for the entire life of a loan it never made.
The structural problem is the same fragmentation described in the introduction, viewed from the perspective of an applicant with bad intent. More than 100 credit programs operate on separate systems with no shared data standard. No agency can determine whether an applicant has been denied, flagged, or defaulted somewhere else in the federal government. An applicant turned away at one agency can apply successfully at another the same week, and nothing in the current architecture would detect it. Commercial loan platforms handle this as a routine function. Identity verification, entity resolution, and duplicate detection occur at application intake, before money moves, rather than through investigation and recovery afterward.
The cost of weak intake controls has been measured. The Small Business Administration’s Inspector General estimated that more than 200 billion dollars in COVID-19 Economic Injury Disaster Loan and Paycheck Protection Program funds, at least 17 percent of the total disbursed, went to potentially fraudulent actors after the agency relaxed controls to move money quickly.
The effect of strong intake controls has also been measured, in the same programs. When SBA moved screening to the front of the lending process and onto a commercial loan platform, the results were observable within months. The commercial provider SBA engaged to modernize its loan systems during the pandemic reports preventing millions of dollars in attempted fraud each month, with higher volumes when a declared disaster produced a surge in applications. SBA’s own reporting is consistent with that account. The agency blocked 21.3 million duplicate, ineligible, and fraudulent applications across its four largest pandemic programs, retaining $511 billion in funding, and the two relief programs designed after those controls were in place, the Shuttered Venue Operators Grant and the Restaurant Revitalization Fund, held estimated fraud rates below 1 percent. The difference between a program with pre-approval screening and one without it, in the same agency during the same emergency, is the clearest available evidence of what the technology does.
Some of the necessary tools already exist and cost agencies nothing. Treasury’s Do Not Pay service screens payees against death records, debarment lists, and other authoritative sources before money goes out. Treasury reports that Do Not Pay helped agencies prevent, detect, and recover $11.7 billion in FY2025. Agencies connect to it one at a time, through whatever their legacy systems permit, which is why adoption remains uneven across credit programs. A shared platform connects once, on behalf of every program that migrates to it.
The scale of the underlying exposure is substantial. GAO estimates that the federal government loses between $233 billion and $521 billion a year to fraud across all programs and operations. Fifteen agencies reported $185.8 billion in improper payments across 64 risk-susceptible programs in fiscal year 2025, a related but differently constructed measure. Federal credit programs are a subset of that exposure, and no credit-specific estimate has been published.
| Credit programs in the FY2025 improper payment data | |
| SBA, PPP Loan Forgiveness | 19.2 percent estimated improper payment rate |
| SBA, PPP Loan Guaranty Purchases | 13.3 percent, down $1.5 billion from FY2024 |
| FHA, VA, USDA guaranteed housing, Direct Loan | Do not appear among the programs reporting rates above 10 percent or amounts above $5 billion |
| Source: GAO analysis of PaymentAccuracy.gov data, GAO-26-108694. Fifteen agencies reported $185.8 billion across 64 programs. | |
Fraud prevention benefits taxpayers in three ways. A fraudulently originated loan defaults at a rate far above the assumption used to price its cohort, and that cost surfaces in later years as an upward reestimate paid from permanent indefinite budget authority, which is mandatory spending. Preventing the origination prevents the outlay. Sustained evidence of lower fraud losses also lowers the default assumptions used to set subsidy rates on new cohorts, which reduces the subsidy appropriation an agency requires and frees room within the discretionary allocation, where credit programs compete with everything else in the bill. And in disaster and emergency lending, money disbursed to a fraudulent applicant is rarely recovered in full, while the pursuit itself consumes inspector general resources, Department of Justice prosecutions, and Treasury collection efforts.
| What earlier intervention is worth, in one program | |
| Defaulted VA-guaranteed loans resolved | More than 99 percent, FY2025 |
| Veterans assisted in retaining a home or avoiding foreclosure | 173,478 |
| Guaranteed claim payments avoided | More than $9.9 billion |
| Source: OMB, Analytical Perspectives, Fiscal Year 2027, Chapter 4. These are servicing interventions rather than fraud controls, and VA measures them within its own program. | |
None of these savings appears in a CBO score, because no credit program can produce reliable evidence tying a specific control to a specific decline in losses. Upward reestimates draw on permanent indefinite authority, so an agency faces no appropriation consequence when a cohort performs badly, and nothing in the current architecture measures what prevention at intake is worth. Common data across credit programs would produce that evidence for the first time. The Act therefore does more than prevent fraud. It creates the measurement the budget process requires before it will recognize that fraud has been prevented.
6. The cost the government is already paying
The alternative to consolidation is not the status quo. It is continued modernization agency by agency, which is happening now and which Congress is already funding one appropriation at a time.
Ginnie Mae holds a modernization contract valued near $127 million. FHA is spending about $92 million on Catalyst. The Department of Veterans Affairs is re-competing a home-loan guaranty technology stack that cost $328 million over four years. USDA announced its own loan modernization effort in June 2026. Each of these is a single agency replacing its own deteriorating systems alone, to its own specifications, with no common data standard and no capacity to exchange information with any other credit program.
The money is being spent either way. The question before Congress is whether it produces twenty incompatible systems or one platform that every program can use. A shared platform consolidates the spending already underway and prevents the next round of incompatible replacements, which agencies will otherwise undertake on the same schedule a decade from now.
7. The asymmetry
The budget process would record a bill that costs almost nothing to enact and, once the fee funds operations, almost nothing to run. That is an accurate description of the score and an incomplete description of the legislation.
The operational benefit, though, is enormous and would impact lenders and borrowers quickly. Eligible farmers waiting on operating loans before planting, first-generation homebuyers, and small businesses without access to bank credit all get badly needed money faster and with fewer opportunities for fraud along the way. Agencies and lenders gain one modern platform in place of twenty decaying ones. Program staff spend their time on eligibility and service rather than on reconciling systems that cannot exchange data.
The institutional benefit is equally concrete. Enterprise visibility into a $5 trillion portfolio. Common data standards that make subsidy-cost estimates more auditable, which improves the quality of the budget numbers Congress relies on for credit programs. A smaller cybersecurity attack surface across the credit portfolio. Earlier detection of risk, both credit risk and fraud risk, at the point of application rather than after disbursement.
A bill the budget process records as near-costless, and that borrowers, lenders, and agencies experience as faster and more reliable, is worth enacting on those grounds alone. The operating savings are significant. The fraud losses avoided are larger still, and the Act would make them measurable for the first time.
8. A no brainer
Congress should enact the Federal Loan Systems Modernization Act. It costs almost nothing to pass, it pays for its own operations through a cost-recovery fee, and it consolidates spending that agencies are undertaking separately today.
OMB could require agencies to show operational and avoided-modernization savings in their budget submissions, so that the value of consolidation becomes visible to the budget process over time rather than remaining outside it.
OMB and Treasury could require every credit program migrating to the platform to screen applications against Do Not Pay or a comparable service, and to report losses prevented in a form that can be tied to cohort performance. That reporting would produce the program-level evidence CBO needs before it can record fraud savings in a future score, and it would give Congress a basis for judging whether the platform is performing.
Doug Criscitello is the executive director of the Center for USA Lending, a nonpartisan organization which advances modern design and delivery of federal credit programs. Robert Shea is CEO of GovNavigators.
Sources
- CBO, Cost of Federal Credit Programs in 2026 (104 programs). https://www.cbo.gov/publication/62030
- CBO, Common Budgetary Terms Explained, publication 57660. https://www.cbo.gov/publication/57660
- GAO, IT modernization of critical legacy systems, GAO-25-107795. https://www.gao.gov/products/gao-25-107795
- GAO, federal data center consolidation savings, GAO-21-212. https://www.gao.gov/products/gao-21-212
- GAO, Fraud Risk Management, 2018-2022 data on annual federal fraud losses, GAO-24-105833. https://www.gao.gov/products/gao-24-105833
- SBA Office of Inspector General, COVID-19 Pandemic EIDL and PPP Loan Fraud Landscape, Report 23-09. https://www.sba.gov/document/report-23-09-covid-19-pandemic-eidl-ppp-loan-fraud-landscape
- SBA, Protecting the Integrity of the Pandemic Relief Programs, June 2023, including applications blocked and post-2020 program fraud rates. https://www.sba.gov/document/report-protecting-integrity-pandemic-relief-programs
- Treasury, Bureau of the Fiscal Service, Do Not Pay, fiscal year 2025 results. https://fiscal.treasury.gov/payment-integrity/do-not-pay-dnp
- Congressional Research Service, Improper Payments, Ongoing Challenges and Recent Legislative Proposals, R48296. https://www.congress.gov/crs-product/R48296
- Federal Credit Reform Act of 1990, title V of the Congressional Budget Act of 1974, including section 504(f) on permanent indefinite budget authority for upward reestimates.
- OMB Circular A-11, section 185, Federal Credit, on subsidy estimates, cohorts, and reestimates.
- Congressional Research Service, Budgetary Treatment of Federal Credit, Concepts, History, and Issues for Congress, R42632. https://www.congress.gov/crs-product/R42632
- GAO, Credit Reform, Current Method to Estimate Credit Subsidy Costs Is More Appropriate for Budget Estimates Than for Cost Analysis, GAO-16-41. https://www.gao.gov/products/gao-16-41
- Allocore, platform performance data from SBA pandemic lending operations, company-reported.
- Education FY2027 congressional justification, student loans overview. https://www.ed.gov/media/document/fy-2027-congressional-justification-student-loans-overview-113549.pdf
- HUD FY2027 congressional justification, Program IT Fund, and GAO-21-459 on FHA Catalyst. https://www.gao.gov/products/gao-21-459
- DOE FY2026 Loan Programs Office justification, and Export-Import Bank FY2026 budget justification. https://www.energy.gov/sites/default/files/2025-06/doe-fy-2026-vol-3-lpo.pdf
- USDA loan modernization announcement, June 2026. https://www.usda.gov/about-usda/news/press-releases/2026/06/04/usda-announces-loan-modernization-effort
- Center for USA Lending, bill background and FAQ, and the June 2026 modernization analysis. https://www.usalending.org