If EPA WIFIA wants to keep offering cancellable, UST rate-locked, one-time resettable loan commitments to sophisticated Aa3/AA- public-sector water agencies with excellent tax-exempt financing alternatives, there will be systemic funding losses in the portfolio. Over the long project construction periods, the borrowers will rationally draw on ‘in-the-money’ loans but cancel or request a reset on ‘out-of-the-money’ loans. Like credit losses, the funding losses won’t be precisely predictable for any given loan or year. But on a portfolio basis over time, they are certain. Some estimate of likely funding losses can be made based on interest rate cyclical patterns, predictable rational borrower behavior, general loan nature and terms, etc.
That expected funding cost estimate will require discretionary appropriations, just like expected credit cost estimates, which now receive about $60m per year for about $6b in loan commitments, based on typical investment-grade loan credit losses. The discretionary appropriations can then be apportioned to specific loans based on specific characteristics like borrower credit rating, their financing alternatives, detailed loan terms, project risk, etc. Again, exactly as is done for credit loss loan subsidies.
Once the funding losses have been estimated, using all available information and the best analytical practices, and discretionary appropriations for such losses granted by Congress and apportioned by OMB for specific loan commitments, FCRA re-estimate machinery and permanent indefinite authority will again work as intended — to smooth out non-systemic budget ‘noise’, leaving the best possible ‘signal’ of program cost for policy review and development.
The Cost of Optionality and Correlation
Let’s review the theory in a bit more detail. There are several elements to WIFIA’s predictable funding losses.
- Optionality in WIFIA loan terms: A UST rate-locked loan commitment that’s optionally drawn and cancellable without penalty can be used as financing with an embedded interest rate option. The value of this option is enhanced by the likelihood of downward reset of the commitment rate, a non-statutory but established practice of the program.
- Optionality in WIFIA loan drawdown: As long as the eligible capex has occurred, WIFIA allows the borrower to draw anytime in any amount during a long construction period (5-10 years, the latter involving a loan commitment for a multi-project capex program) or all at once up to one year after completion (as permanent financing), or not at all. This allows the borrower to use short-term financing for construction advances whenever their short-term rates are below the loan commitment rate lock. Depending on yield curve slope, if UST rates are rising this drawdown optionality may delay draws — and result in even higher UST rates (and WIFIA funding losses) at disbursement.
- Correlation with UST rates in borrower financing alternatives: A high borrower credit rating is not enough to utilize a WIFIA loan’s full optionality. For Aa3/AA- public water agencies, however, their main infrastructure financing alternative, tax-exempt bonds, is both highly correlated with, and close in yield to, the UST yield curve. If UST rates fall after loan commitment, the WIFIA alternative is quickly ‘out-of-the-money’ relative to the tax-exempt bond alternative and it will be cancelled or (to avoid political embarrassment) reset, as occurred for about a dozen loan commitments originally executed 2018-2020.
By offering optionality (per loan terms and by program practice) and correlation (per the selected class of borrowers), funding losses on a portfolio basis over time are inevitable. Back in 2021, I did a hypothetical model (see Appendix 1) to illustrate how this works, using a real-world interest rate pattern over 20 years and loan commitments that either were (1) required to be drawn, per construction schedule, regardless of changing UST rates, or (2) optionally drawn or reset, as with WIFIA loans. Below are a couple pages from the model summary — comments in yellow block added for this post:
Note that in the Optional/Reset case, an average funding loss (about 4% in this model) could be calculated across all loans for the 20-year period. Again, the specific loss for any particular loan was not predictable (only that there are no gains, as with credit losses) but the portfolio average can be the basis of discretionary appropriations to cover the predictable cost of WIFIA loans that include optionality being offered to a class of borrowers with UST-correlated financing alternatives.
Estimating Predictable Portfolio Funding Losses — About 5%?
Now back to WIFIA’s current reality. The $2.1b of funding losses to date vividly demonstrates that slight changes in UST interest rates have a huge impact on the NPV of loan value — finance people know this, but maybe the EPA and OMB bureaucrats did not. They do now, and they should also take on board that there will never be $2.1b of funding gains, or maybe any material gains at all, going forward to balance the portfolio’s losses.
FCRA law and OMB practice, as detailed in Circular A-11, would seem to require modelling the expected portfolio net funding cost. [1] This doesn’t strike me as being too difficult — it would look something like the illustrative model I did in 2021, but with a lot more data, which I assume OMB has or could obtain. There’s an immense amount of data about UST rate patterns over time, as well as tax-exempt bonds and short-term financing rates, and EPA and OMB will have the precise details on WIFIA loan terms and borrower characteristics. Yes, the initial funding loss models won’t have the depth of OMB’s current credit models, but the modelling methodology and estimates can be improved over time, with FCRA re-estimate machinery automatically covering the initial errors. What’s the alternative? To ignore the losses? [2]
I’d guess the basic average funding loss per loan commitment, when the smoke clears, will be about 5%. WIFIA results to date suggest an average closer to 9% ($2.1b loss on $23b loan commitment portfolio) but that reflects only interest rate patterns 2018-2025. I think there’ll be long stretches where the losses are much lower or zero, as borrowers cancel or wait for a reset when rates are falling. The 4% average loss from the illustrative model was based on real UST patterns after all, and you’d expect refinements (e.g., delayed draws in rising rate environment from borrowers using ST financing) to add about another 1%. So about 5% is probably in the ballpark. [3]
For Policy and Congress, Maybe Not So Bad?
Consider the 5%, or 6% when credit losses are added in, from a policy perspective. Plenty of federal credit programs are much bigger and cost a lot more. Yes, a loan subsidy rate of 6% only supports a 17:1 program leverage ratio, not the current 100:1 oft touted by the water agency lobbyists [4], but it’s not so bad. If Congress wanted to fund a WIFIA program that could offer about $4b of loan commitments (a realistic level of demand) with embedded options to Aa3/AA- water agencies each year it would only cost about $235m, vs. about $40m if the funding losses were ignored. Is that really a problem? If WIFIA really is so ‘critical’ for US water infrastructure, as the standard narrative would have it, would less than $200m per year be a justification to end or curtail the program? Surely, the water lobbyists could get that through, just as they got $65m of completely unneeded funding for WIFIA in the FY26 spending bill.
If there is a problem, it’s not about the money, which is trivial. It’s about the questions that might be asked about why discretionary appropriations are required for expected funding losses. The answers would need to explain, at least a little, what WIFIA actually does. But isn’t that the point of correct federal budgeting?
Apportionment from Discretionary Funding
Once the reality of predictable funding losses at WIFIA and the need for discretionary appropriations is acknowledged, as it is for credit losses, the next step is to develop a methodology to apportion the discretionary funding to specific loan commitments.
Again, this doesn’t strike me as difficult. Much will be similar to credit loss apportionment — consideration of borrower credit quality, specific loan terms and intercreditor factors, nature and risks of the project being financed, etc.
Each loan commitment can be evaluated with respect to the circumstances in which the loan might be drawn. A fairly simple loan commitment to a large, highly sophisticated AAA public water agency that is frequently issuing in the tax-exempt bond market and has extensive short-term financing facilities will likely only be drawn when the commitment rate is ‘in-the-money’ relative to these alternatives. This type of commitment will, on average, result in the highest funding loss per dollar and should require the apportionment of highest level of subsidy.
In contrast, a complicated BBB- project financing commitment to a multi-owner, special-purpose company financing a relatively hi-tech project with some revenue risk will have few cost-effective financing alternatives and can be expected to be drawn as construction progresses. Funding losses on this type of loan will likely, on average, be balanced with funding gains and may not require any apportionment for funding loss.
There will be a spectrum of funding loss risk, as there is for credit loss. Below is a quick sketch of how the spectrum could be defined by risk category:
The sketch uses the same type of categorization that I used in a prior post to illustrate OMB’s actual credit apportionments FY18-FYTD26. Assuming that funding risk and credit risk are inversely correlated, as shown above [5], and WIFIA’s pattern of loan selection continues the same FY27-FY34, total apportionments would require about 5% of loan commitments in discretionary appropriations. It would look something like this:
Note that this type of apportionment is intended to be applicable for loan commitments at any point in the interest rate cycle. A refined version might include adjustments for interest rate conditions at the time of execution (e.g., degree at which commitment rate is above or below long-term UST averages). A loan commitment with high average expected funding loss executed a low point in the cycle would require higher apportionment, etc. Such a ‘cyclical’ adjustment would, I think, need to be quite complex to ensure that only expected losses were included and not distort the long-term averages — maybe a topic for another post.
From everything I can see, EPA WIFIA and OMB have plenty to work with to at least acknowledge the issue and start developing solutions. Further delay might suggest indifference, or worse, intention — not a good look with respect to the ADA.
_______________________________________________________________________________
Notes
[1] If WIFIA offered sub-UST interest rates at loan commitment (as was proposed in the 2018 SRF-WIN bill), funding losses would most certainly be included in the estimates of loan subsidy. How is offering a loan with an embedded option that makes the disbursement of sub-UST loans (relative to then-current funding rates) a near-certainty any different?
[2] Well, yes — simply ignoring the losses is likely the plan of water lobbyists who serve the interests of the ‘exclusive option club’ of water agencies that can use WIFIA as an interest rate management tool. No judgement intended here — I get that everyone is just doing their job in the narrative neoliberal system we find ourselves in.
The plan would probably succeed if the losses were less eye-popping (say, under a billion) and a plausible case for ADA violations wasn’t involved. This area of federal budgeting is very arcane, support for water infrastructure is strong and bipartisan, and a ‘narrative’ story about how ‘critical’ WIFIA is for thirsty children and all that is easy to promulgate. But at $2.1b of losses and rising, what’s happening is hard to hide, and the federal bureaucrats involved in apportioning discretionary funding for WIFIA loan subsidies might feel that specific blame via the ADA could be headed their way. Still, a plan to camouflage the losses by even more intense ‘narrative’ and (of course) lobbying might work — we’ll see.
[3] OMB is not shy of apportionments from discretionary appropriations in the 5% range for WIFIA loans. One loan, not closed but apparently well along in the process, $51m for Big Bear Area Regional Wastewater Agency, was apportioned at 4.95% in FYTD26. A few others are in the 3-4% range. I don’t see how that works with the BBB- minimum rating requirement, but I guess they think it will. More to the point — given OMB’s evident willingness to predict complex credit situations, why should be they be reluctant to predict complex funding cost situations and apportion 5-6% for funding losses for a loan to a large AAA public water agency that will only require 0.30% for credit losses?
[4] Also re SRF-WIN in the context of funding losses, this letter opposing the original SRF-WIN bill from the water lobbyists highlighting the ‘100-1’ ratio hasn’t aged at all well: In Retrospect, An Ironic Criticism of the SRF-WIN Act
[5] There will be exceptions to this correlation. SWIFIA loans to state-backed SRFs are of the highest credit quality, but funding losses will be limited by the provision in SWIFIA law that requires drawdown in a year after loan commitment. Or, if as a condition for resetting an existing loan commitment, a Aa3/AA- borrower agrees to draw the entire commitment within a month or so. Fast drawdowns mean less time for UST rates to rise, and lower expectations of option value. Final apportionment can take these factors into account, just as specific credit risk mitigating factors are considered in the credit loss apportionment.