A prior post described an analysis of the estimated economic cost of WIFIA’s $16 billion portfolio of loans and loan commitments at federal FYE 2022. For various reasons outlined there, a simple estimate of the cash cost of funding the portfolio is probably a more pragmatic approach. This post starts to sketch that out.

WIFIA has built its $16 billion portfolio with 92 loans over the last five fiscal years. I assume that the average WIFIA loan has a weighted-average life (WAL) of about 20 years and therefore is executed or re-executed with an interest rate corresponding to then-current 20Y UST yield. My portfolio model starts with the reported loan amounts and execution dates. The model assigns each loan an interest rate equal to the 20Y UST on its execution date.

My portfolio model shows a simple average interest rate of 1.96% for the 92 loans. WIFIA’s website reports a 2.00% average, which I assume is also simple, given its purpose.  This seems to confirm my estimates of the portfolio’s basic characteristics.

The weighted-average interest rate (WAIR) is a more useful metric for estimating future portfolio interest revenue because it weights individual loan amounts. I estimate the WAIR at FYE 2022 to be 1.86%, a little lower than the simple average due to loan volume in the second half of 2021.

The chart below summarizes the data I’m using on a monthly basis FY 2018-2022.

The portfolio model at FYE 2022 is the basis for projecting future loan interest revenues and Treasury funding cost. Assumptions for three additional primary factors are necessary, (1) schedule of drawdown of remaining loan commitments and overall loan portfolio principal amortization, (2) the WAIR each year, and (3) Treasury’s average interest cost to fund the outstanding loan balance each year.

Simplified Primary Assumptions

In a detailed model, each loan’s drawdown, amortization schedule and interest rate would be considered, and projections would be based on the aggregate. The WIFIA program certainly has that data, and even from publicly available information (e.g., press releases about a project progress, borrower financial reports, etc.) a lot could be estimated. For this exercise, however, I’m going to make some highly simplified assumptions about the portfolio’s characteristics as a whole. This will of course limit the precision of the results, but I think they’ll still be accurate enough for the purposes discussed in more detail below.

For drawdown and amortization, I start with the assumption that the portfolio is 75% undrawn at FYE 2022. This appears to be consistent with EPA’s estimated numbers in the 2023 White House budget. The portfolio is assumed to be fully drawn in 2028, with 35-year straight-line principal amortization starting in 2029. The portfolio is fully amortized in 2064.

I assume that the portfolio’s WAIR in any year will remain at 1.86%. This certainly won’t be the case in reality, as various loans amortize in different ways, and the annual WAIR will fluctuate. But I think on average over the portfolio’s full term, it should be accurate enough.

Assumptions about Treasury’s interest cost to fund the portfolio are the main difference between economic and cash projections. Obviously, Treasury does not actually match fund its ‘investments’ in the WIFIA portfolio by issuing zero-coupon bonds, the methodology used in FCRA economic evaluations. Instead, portfolio funding is sourced from overall debt issuance, and will be reflected in an amount of additional federal debt equal to the portfolio’s outstanding at any point. Needless to say, the size of WIFIA’s portfolio is completely immaterial to the overall level of federal borrowing, so it won’t impact anything at Treasury. In that context, the cash interest cost of funding the portfolio each year will simply be a very tiny slice of the average overall cost of federal borrowing.

The average interest rate on federal debt depends on a huge number of factors and projecting it for the next forty years is a difficult mission, to say the least. Fortunately, CBO provides exactly this projection in its Long-Term Budget Outlook, at least for years to 2052. In their 2022 Extended Baseline Projection, they assume as a primary economic variable that average nominal rates on federal debt outstanding will be 2.5% for 2022-2032, 3.4% for 2033-2042 and 4.0% for 2043-2052. I use these points to create a simple linear pattern that peaks at 4.0% and declines back to 2.5% by 2064.

CBO itself makes clear that such long-term projections are in effect speculative. But as discussed further below, that doesn’t matter for the primary purpose of this exercise, which is to present WIFIA’s cost in terms of federal cash budgeting concepts. In fact, using CBO’s own numbers (whatever they are) is central to establishing a plausible case that WIFIA’s cash numbers aren’t so bad.

The chart below summarizes the simplified assumptions for our projections.

Reduced Tax Expenditures?

There is one other factor to add to the mix, especially with respect to the purpose of this exercise. If WIFIA loans create future incremental federal revenues, could those be included to offset a part of WIFIA’s interest cost?

It appears that such a ‘dynamic analysis’ is often performed by CBO but “not generally reflected in CBO’s cost estimates”. That’s probably just as well in this case because it might raise some uncomfortable questions. But I think there is one type of dynamic analysis that CBO has included in budget scoring for WIFIA legislation — the impact of WIFIA loans on the issuance of tax-exempt debt.

I’ve written about this several times over the last few years. The most succinct summary is a one-pager in the form of a letter to the Joint Committee on Taxation in 2021. For more background, here’s a long post on the topic and a detailed analysis. One thing to note is that JCT admits that its methodology is somewhat speculative. Like the CBO’s similar admission about its projections, that doesn’t really matter for our purposes — if they’re willing to use it for scoring, we’ll work with it, too. Note also that I’m using JCT’s methodology, not their assumptions, which appear to be incorrect with respect to WIFIA’s actual outcomes.

The basic idea is simply that WIFIA loans to highly rated public-sector agencies displace tax-exempt bonds that would have otherwise been issued, thereby reducing future tax expenditures. It’s a cash concept, so estimates can be included in this cash analysis. For the current case, I make some conservative assumptions: (1) 75% of outstanding WIFIA loans displace tax-exempt bonds, with the balance having no effect, (2) a federal tax rate of 25%, and (3) taxable substitute bonds yield an average of 2.75%. The annual reduction in tax expenditures (in effect, an increase in federal tax revenues) is equal to the annual tax paid on the substitute taxable portfolio’s income.

Annual WIFIA Revenue, Treasury Interest Cost

Putting all those factors together, it’s straightforward to project WIFIA revenue (without and with the tax effect) and Treasury’s cost. The chart below summarizes the results annually. As expected, Treasury’s cost is always higher than WIFIA’s revenues — even on a cash basis.

It’s also straightforward to net the revenues and costs, as shown in the next chart below. WIFIA cash losses are at their worst when the portfolio is fully drawn and CBO projects that average federal interest rates are high. They tail off as the portfolio is amortized and rates stabilize. For the next few years, the picture isn’t so bad either — that’s what we want to work with here.

The final chart summarizes the dollar amounts and percentages of the $16 billion portfolio as of FYE 2022. The undiscounted sums, and even the discounted numbers, don’t look great. I had hoped that a cash analysis, which uses medium-term federal interest costs, might be significantly better than the economic results, which reflect the full Treasury yield curve. CBO’s relatively high baseline projections of federal interest cost derailed that simple story. But there are other ways to look at things.

Estimating the economic cost of something requires that you look at the full extent of its impact, no matter how far in the future, and estimate its present value by appropriate discounting. That’s not necessarily true for cash cost projections, which can be easily sliced & diced to determine whether inflows match outflows for periods of particular interest.

CBO establishes one such a period of interest — the 10-year budget scoring protocol for new legislation, something familiar to all policy (and political) stakeholders in WIFIA’s future, or intended lack thereof. In contrast to the full cash results, the next ten years of WIFIA’s FYE 2022 portfolio don’t look so bad, especially if the correct tax effect is included. These can be further packaged on an annual basis. For example, “The portfolio net funding cost is unfortunate, but it’s only about $70 million a year or half of one percent of the portfolio. That’s not much of a problem for a popular program, is it?”

That’s where this cash analysis is going. The goal is not truth, but narrative. The simplified and preliminary analysis in this post needs a lot of refinement. But it shows promise for its fundamental purpose.