Well? What does it look like?
Yes, you need a little bit of discretionary spending to cover the trivial credit risk of loan commitments and pay for program admin. But after that, the real purpose of the program (free interest rate options for large water systems) operates with FCRA re-estimates and permanent indefinite authority to incur unlimited mandatory appropriations.
POSIWID – the purpose of a system is what it does, no?
Download PDF, including InRecap comments 02112026
As a follow-on to the prior post, here we’ll look at OpenOMB’s reports on the EPA WIFIA Direct Financing account apportionments.
Lending Authority
The Financing account is where overall legal lending authority granted by spending bills is apportioned for specific loans, and subsidy from the Program account is also apportioned for specific loans.
Lending authority is not mysterious. It seems OMB apportions a chunk, generally $3b, at a time. Since in theory WIFIA has been getting annual appropriated subsidy amounts to lend about $6b per year (not even counting carryovers), it would seem that OMB can ‘top up’ the apportioned amount within the fiscal year, if necessary (which it hasn’t been recently). This ‘budget authority’ is then apportioned to specific loans at some stage in the approval process in that fiscal year. Lots of minor adjustments, as you’d expect, including a $222m ‘downward re-estimate’, which I think probably relates to cancellations of prior loan apportionments of lending authority.
Subsidy Apportionments
The subsidy apportionments, which appear in dense footnotes, are a lot more interesting. For some reason, the footnotes in earlier FY22 reports go all the way back to the start of the program in FY218, we have the complete set FY18-FYTD26 for subsidy apportionments for all WIFIA’s loan commitments, including those expected to close soon. The total apportioned amount is $185m of credit subsidy for $25.9b in loan commitments across 155 loans, of which $23b in amount and 149 loans have been executed to date. The weighted average subsidy apportioned FY18-FYTD26 is about 0.71% of loan commitment amount, well below the 1.0% number that’s often cited as shorthand for the low taxpayer cost of WIFIA loans (e.g., the lobbyists’ ‘100-1’ ratio).
Here’s the data by range of subsidy amount and number of loans:
As expected, the biggest group by far is borrowers in the 0.31%-0.60% subsidy range, corresponding roughly to Aa3/AA- credit quality. This level of credit rating is typical for most public water systems in the US.
Putting the volume per range in a pie chart shows who the WIFIA program is really meant for, regardless of all the ‘narrative’. 75% of the loans are in very low risk credit categories, indicating not only low-risk borrowers but fairly straightforward water revenue bond-type loan structures. The next 18% or so probably includes low-risk borrowers but maybe some slight idiosyncrasies in loan structure that add a tiny bit of risk. Less than 10% goes to loans that, although still highly creditworthy (per WIFIA’s investment-grade eligibility requirement), might be tough to cost-effectively place in the bond or other private market (project financing, private-sector ownership, unusual project type, complex security structure, etc.).
In other words, about 90% of WIFIA loans probably don’t have any real-world impact on US water infrastructure improvement, since such low-risk, mostly vanilla loans could be easily and cost-effectively placed elsewhere. Instead, most loans simply provide off-budget transfer payments (primarily through ‘free’ interest rate management options embedded in loan terms) to borrowers. Less than 10% might have a real policy outcome in terms of project feasibility or acceleration.
OMB calculates and approves WIFIA loan subsidies apportionments. OMB also presumably checks WIFIA compliance with their own Circular A-129 requirements each year. Why no cognitive dissonance? Willful disregard? Some sort of ideological agenda held by ‘Deep State’ bureaucrats and fostered by lobbyists? Or did they just not GAF?
Download PDF — InRecap comments 02112026
There is now a website, OpenOMB, that tracks OMB apportionments since FY22, including those pertaining to EPA WIFIA. The site has an interesting backstory [1].
It seems that the EPA WIFIA program has two accounts which receive apportionments from OMB. One is the Program Account where discretionary and mandatory appropriations to cover net loan costs are recorded — the program’s ‘equity base’ provided by taxpayers. The other is the Direct Financing Account, which is where the annual lending authority and specific loan allocations are recorded.
This post will focus on the Program Account. I was able to see what appeared to be OMB’s final apportionment numbers for FY22-FY25. No surprises — the numbers tracked what I’d already seen in the WH Budget Technical Appendices for the same years. But there was an important detail, the clarification of which sheds light on the exact nature of the $2.1b loss.
Clarification: Interest Rate Re-Estimates vs. Technical Re-Estimates
I’ve been using ‘interest rate re-estimates’ as shorthand for the source of funding losses and huge mandatory appropriations. Given the current numbers, and my own analyses starting in 2021 which predicted them, I was sure that the mandatory appropriations surfacing in the Budget Appendices were caused by the effect of UST interest rate changes on program loan commitments. What else could they be?
Sure enough, the mandatory appropriations each year were due precisely to ‘WIFIA Upward Re-Estimates’ including internal interest thereon [2] as a use of the appropriated funds. But the re-estimates were not labelled as ‘Interest Rate’, but rather as ‘Technical’ re-estimates. So — back to OMB Circular A-11 to find out what was going on.
I remembered that there were two types of re-estimates but got the impression that the labels were self-explanatory. As usual with federal budgeting, however, it’s more complicated. Here’s my interpretation of the relevant parts of A-11:
The sequence appears to be that (1) a loan disbursement triggers a technical re-estimate, the calculation of which includes the then-current UST funding rate, which (2) in turn requires a re-estimate of the credit subsidy, which if positive or ‘upwards’, (3) activates FCRA permanent indefinite authority to incur mandatory appropriations. Since the disbursement is a ‘real-time, real-world’ allocation of resources within a fiscal year and the accounts with Treasury need to balance, the mandatory appropriations are immediately ‘spent’ — that is, permanently recorded as utilized subsidy to cover the now-certain shortfall in the PV of loan debt service. Hence, no carryovers of this type of funding appear in the accounts, in contrast to the big carryovers of unutilized discretionary budget authority.
That’s what I think is happening. Perhaps in future we’ll see some ‘interest rate re-estimates’ as part of a loan’s final cost numbers, but in any case, the technical re-estimates provide the key information. Importantly, this means that the upwards re-estimates in OpenOMB data all originate from actual disbursements, and the funding losses are not a ‘mark-to-market’ type of estimate on the loan commitment portfolio but realized only on amounts disbursed. An EPA WIFIA press release reported $10b in disbursements in mid-2025, close to the time the re-estimates were first done for the WH budget, so the $2.1b of total funding losses FY22-FY25 represents a 21% loss rate. Not pretty.
Patterns Emerge
Other than the EPA press release mentioned above, I can’t find any direct data on WIFIA loan disbursements. But we can make some informed guesses. For $10b of disbursements to cause $2.1b of funding losses, a 21% NPV loss, the negative spread between the average loan commitment rate and the UST funding rate on a typical ‘maxed out’ WIFIA loan pattern (i.e., 35Y term, 5Y deferral, 5Y interest only) has to be about 1.25%. The weighted average loan commitment rate in the $22b undrawn portfolio ranged from about 2.0-2.5%, but this number would rise as older, lower rate commitments were drawn first, perhaps by 1-1.5%. So, disbursements must have occurred when UST rates were in the range of 3.25-4.75% — exactly the pattern we see emerge in mandatory appropriations:
There are two other factors involved. First, although WIFIA allows completely flexible disbursements per borrower request, that is limited by the progress of construction and eligible expenditures. That schedule is also a bit flexible for financing purposes (e.g., when to pay for equipment purchase, contract advances, etc.) but it might be limiting as to amount, especially in the early years. Perhaps by FY25, the FY18-FY21 loan commitments could mostly be fully drawn.
Second, and much more relevant to how borrowers are using WIFIA loan commitments, we wouldn’t expect to see drawdowns until, on average, WIFIA loan commitment rates were lower than the rates on the borrower’s short-term financing alternatives. This crossover likely happened in FY22. As that spread grew dramatically FY23-FY24, you’d expect an increasing rate of disbursements. Since UST were also rising, there would be corresponding funding losses and mandatory appropriations — again, exactly the pattern we see.
There is also something we don’t see — minor funding losses or even gains (these presumably would show up as ‘downward estimates’ with a corresponding increase in program budgetary resources). OpenOMB only goes back to FY2022, but I didn’t see any from what I could understand about WH Budget Technical Appendices 2018-2021. I’d expect minor adjustments, notwithstanding borrower behavior — maybe we’ll see these in OMB FY26 apportionments.
______________________________________________________________________________
Notes
[1] From the ‘About’ page of the website:
“Congress created the apportionment process in the Anti-deficiency Act to ensure agencies spend within the limits of the law and do not prematurely exhaust the funds Congress has given them. But administrations of both parties have, at times, abused this authority to defy Congress’s spending laws — such as when President Trump’s OMB used the apportionment process unlawfully to withhold security assistance funds that Congress had appropriated for Ukraine.
“Until July 2022, however, OMB exercised its apportionment authority largely out of public view. Neither Congress nor the public had access to apportionments.
“In the Consolidated Appropriations Acts of 2022 and 2023, Congress took action on a bipartisan basis to fix this problem. It required OMB to post its apportionments on a public website. OMB does so at apportionment-public.max.gov. But OMB’s website is hard to navigate and has no search function, making it difficult to find apportionments and oversee OMB’s work.
“OpenOMB aims to make oversight of OMB’s apportionments easier for Congress, the press, and the public. Drawing on the data files from apportionment-public.max.gov, this site makes OMB’s apportionments searchable and easier to find.”
I looked at the legally required OMB apportionment site — yes, a bit harder to navigate but the same information as OpenOMB is presenting.
More interesting is what OpenOMB is trying to accomplish — in essence, I think, to stop Russ Vought’s OMB from slow walking the apportionment process to deny previously appropriated funds for purposes that the Trump administration opposes — like Ukraine (OpenOMB’s Advisory Board has several ultra-pro-Ukraine people).
Obviously, my interest is different. But OpenOMB’s stated goals made me think of the story behind The Zeldin Letters. Could the specific slowdown in WIFIA loan processing have been due to some stalling in apportionment, which is a final step? The speed at which loans closed thereafter suggests that was the case. Did the stalling come from some OMB officials raising concerns about WIFIA funding losses, $1b of which had surfaced for the WH FY 2026 Budget around May 2025? And the Zeldin Letters campaign was specifically directed at using political influence to override these concerns? Zeldin himself is not MAGA or onboard with Project 2025 — he seems more like a typical ‘narrative neoliberal’ who would want to make sure that EPA’s own ‘crony loaner’ program continued to deliver free interest rate options. The influence prevailed, probably due to a political cost-benefit calculation for this small issue, but could the concerns remain at OMB?
Is it possible that some at OMB would even welcome consideration of Anti-Deficiency Act violations to use as a ‘shield’ against two-bit political influence? Perhaps we’ll see.
[2] The actual funding loss at the time of disbursement is the technical re-estimate plus the interest re-estimate. The separate interest component is (I think) a kind of backdating to the time the commitment was made, i.e., the technical re-estimate alone is what the subsidy should have been at commitment, and the interest is what should have been ‘earned’ in UST SERs PV terms from that point to actual disbursement. Bottom line — the sum of the two numbers, precisely matched by mandatory appropriations, is the real economic loss in the fiscal year.
OMB analyzes the credit risk of federal loans very carefully. This is as you’d expect — most federal loan programs are intended for borrowers who don’t have many cost-effective private-sector alternatives. Credit loss will be the main factor in the cost of such loans.
But even for large, highly rated public sector entities with excellent access to short- and long-term sub-UST tax-exempt finance alternatives, OMB assesses potential credit losses, as small as they are, just as carefully. OMB apportioned the AAA Indiana Finance Authority SRF Tranche 1 with a credit subsidy rate of 0.30%, while the AA City of Fort Worth loan got a rate of 0.32%. Very precise analyses on very small numbers.
But could EPA WIFIA loans experience significant losses from other, non-credit factors? Like a highly rated borrower drawing a 1.45% loan commitment only when their short-term financing rates are above that level and 20Y UST funding costs are in the 4.25% area? This type of predictable behavior caused funding losses $2.1b on approximately $10b of WIFIA loans disbursements. Why wasn’t OMB as equally careful about this type of loss?
Funding loss on fixed-rate loans is not different than credit loss. Once a fixed-rate loan with scheduled debt service is issued, the principal amount is not relevant to its valuation — only the NPV of future cash flows is, a reality that FCRA law and even OMB’s own FCRA methodology recognizes. It does not matter whether a $2.1b federal loss in NPV value comes from a loan doing a Solyndra with zero recovery or a loan paying much less in interest than the Treasury is paying to fund it. In both cases, in economic terms, the taxpayer has to cover the NPV shortfall of expected borrower cash flows.
No Skin in the [Blame] Game?
Since I’m sure OMB understands funding losses, the question remains as to why they don’t consider this factor in apportionment for WIFIA loans. Perhaps the answer comes, ironically enough, from the ‘The Solyndra Failure’ (House Majority report August 2012), page 17:
OMB definitely has ‘skin in the game’ when it comes to credit risk assessment, something amply demonstrated in this report itself. But maybe they’re not so incentivized to care about possible funding losses — as ever with bureaucrats, a taxpayer loss is only a problem if you can be blamed for it. I think there are two reasons why OMB might expect that WIFIA funding losses, even at the current ‘Solyndra-times-four’ scale, won’t even be considered a ‘problem’, much less a source of blame.
Creating an ‘Indifference Deficiency’
Without ‘skin in the blame game’, EPA WIFIA and OMB can be indifferent to funding losses and the mandatory appropriations required to cover them. This ‘indifference deficiency’ in the apportionment of discretionary appropriations is subtly different from the classic ‘coercive deficiency’ prohibited by the ADA. There doesn’t need to be any active effort or planning to justify a future coercion — just don’t bother to think about the consequences of WIFIA loan commitment terms (“they’re unpredictable!”) and let the FCRA re-estimate machinery automatically do the coercive work a few years later, off-budget and off-stage.
Note that the characterization of funding losses as ‘unpredictable’ is key to the justifying indifference. Even with $2.1b of losses on $10b of disbursements, there’s a lot of scope in this defense. In standard WIFIA ‘narrative’ terms, it’s nearly bulletproof — the whole topic is far too complicated for most people to judge what was predictable or not. But even among the better informed, the extreme movements in interest rates during WIFIA’s operational period 2018-2025 seem to offer some defense. In the chart at the top of this post, the rise in 20Y UST rates off Covid-suppressed historic lows in 2020-2021 back to historical averages 2023-2025 was rapid. Yes, in retrospect the $2.1b loss was predictable, but was it at the time?
As with credit loss, exact funding losses will of course never be predictable. But I think there’s a strong case to be made that some funding loss was predictable when WIFIA was executing and resetting loan commitments in 2020 and 2021, and the failure to apportion anything for this cost is a specific violation of the ADA. Simply ignoring the probability of funding loss at any time in the interest rate cycle is a more general violation. The rationale for specific and general violations of the ADA is different, so I’ll sketch them out separately.
Apportionment for loans executed at a specific point in interest rate cycle: Look again at the chart at the top of this post. If an OMB official had looked at the interest rate data up to mid-2020, the average 20Y rate would have been 4.5% with a standard deviation of 1.65% — not appreciably different than the values through January 2026 (4.4% and 1.63% respectively). Even apart from the fact that low rates were due to active Fed rate suppression during Covid, the low 20Y UST rates in 2020-2021 were outside one standard deviation from the long-term average and might be expected to revert to the mean at some point when loans were being disbursed over the next five years. Not certain by any means, but at least somewhat probable? At least as probable as the tiny chance of a Aa3/AA- borrower defaulting a few years after loan drawdown?
The point is that just as OMB should (and apparently does) use all available data to assess probable credit losses, readily available data for probable funding losses on future disbursements should also be considered in apportionment at loan commitment execution.
Apportionment for loans executed at any point in interest rate cycle: A different type of defense for indifference deficiencies comes from the likely intention of FCRA re-estimate mechanics in the first place. The argument would roughly go: “Yes, loans executed at low points in the interest rate cycle will likely cause funding losses, but these losses will eventually be balanced by funding gains from loans executed at high points. On a net basis, this is budgetary ‘noise’ and ignoring it results in a better budget ‘signal’ for policymakers. In contrast, credit losses are always one-way, and a precise assessment provides a pure ‘signal’ for policymakers”.
This argument has a lot of merit for most federal loans. But not for EPA WIFIA’s Aa3/AA- public-sector borrowers. Consider what borrower behavior would be if the same scale of change in 20Y UST rates had occurred but in the opposite direction. This in fact happened between 2009-2010 to 2020-2021 — a 3.5% fall over ten years as opposed to a 3.5% rise over five years. If EPA WIFIA had been in operation and making loan commitments in 2010, could anybody credibly argue that a $2.1b funding gain on $10b of disbursements would have happened? That Aa3/AA- water agencies would have drawn 4.5% WIFIA loan commitments when 20Y UST rates were in the 2.0% range and tax-exempt bond rates even lower? Of course not. The commitments would have been cancelled or reset, as many 2018-2019 loan commitments were.
To think that EPA WIFIA borrowers will ever be the source of significant funding gains is as unrealistic as assuming that some borrowers will voluntarily be the source of ‘credit gains’ by overpaying their loans. Never happen. For this type of highly rated public sector borrower, funding losses are like credit losses — one-way only — and should be assessed in the same way.
From a ‘discussion’ with Gemini this morning — who (or more precisely in this case, what) else would be willing to ‘discuss’ possible EPA WIFIA ADA violations and FCRA law on a Sunday morning?
I didn’t know GAO provided ‘legal decisions’ on anything [update: apparently they do, including specifically on appropriations], but that’s what Gemini offered as a draft unprompted. The 2-page draft does seem to be succinct summary of a case, so I thought it worth posting. FWIW — usual caveats on unedited AI output apply.