March 1, 2018

On February 22, economists at the Wharton School of Business released a brief analysis (the Penn-Wharton Budget Model analysis) of President Trump’s infrastructure plan. While the study does average a diverse array of economic models of the added economic value of federal infrastructure grants, the analysis also reveals a fundamental misunderstanding of how federal credit programs work that calls into question the overall conclusions of the work.

The analysis (which prompted a devastating Washington Post headline on February 23, “The math in Trump’s infrastructure plan is off by 98 percent, UPenn economists say”, which was quoted by Senator Tom Carper (D-DE) to Transportation Secretary Chao in a hearing earlier today), had the following conclusions:

…based on previous experience reviewed herein, most of the grant programs contained in the infrastructure plan fail to provide strong incentives for states to invest additional money in public infrastructure. Indeed, an additional dollar of federal aid could lead state and local governments to increase infrastructure total spending by less than that dollar since state and local governments can often qualify for the new grant money within their existing infrastructure programs. We estimate that infrastructure investment across all levels of government, including partnerships with the private sector, would increase between $20 billion to $230 billion, including the $200 billion federal investment.

We estimate that the plan will have little to no impact on GDP.

However, buried in the text of the analysis was this note:

The literature, which focuses on how state and local governments respond to federal grants, probably understates the additional infrastructure generated by these types of credit programs.

That turns out to be an understatement. To understand why, you have to understand how the budgetary treatment of federal credit programs differs from that of grant programs.

For a grant program, the amount of money appropriated by Congress for the program (or the amount of non-appropriated contract authority provided) bears a relationship of about 1 to 1 to the size of the check that the Treasury writes to the grant recipient. There may be deductions of a percent or less for program overhead and oversight, but the important thing is that the size of the check written by the Treasury to the grant recipient can never exceed the amount of the appropriation or other budget authority provide by Congress.

Federal credit programs don’t work that way at all. Since the Federal Credit Reform Act of 1990 (FCRA), the face value of federal loans is no longer recorded in the federal budget. Instead, Congress only has to provide money to cover the “subsidy cost” of the loan. These costs frequently total less than 10 percent of the face value of the federal loan. So the check written by the Treasury to the state or local government can be 10 or more times greater than the amount of appropriation or other budget authority provided by Congress.

The Trump plan proposes $14 billion in budget authority for loan credit subsidies. The actual subsidy rates won’t be determined until later (they are done on a case-by-case basis by OMB and the lead agency), but the amount of checks that would be written to state and local governments would vastly exceed that $14 billion even under the most conservative assumptions.

The Penn-Wharton analysis assumes that the $14 billion for credit programs (combined with $6 billion to cover the cost of lost tax revenue for increased risk of private activity bonds) would lead to between $10 billion and $30 billion in net total infrastructure spending (inclusive of the $20 billion in federal appropriations or lost tax revenue). But this is based on the misconception that the size of the appropriation is equal to the size of the check written to state or local government partners, which is true for grant programs but manifestly not true for credit programs.

(Most of the studies cited by the Penn-Wharton model were published before FCRA so they wouldn’t be able to take into account the face value vs subsidy cost differential that now exists.)

For a real-world example, there is no need to look at hypothetical economic models. We can look at actual facts – the amount of subsidy funding used by the TIFIA surface transportation credit program to date versus the size of the loan checks the program has written.

From the inception of the TIFIA program in fiscal 1999 through the end of fiscal year 2017 last October, the TIFIA program has used $1.9 billion in budget authority provided by Congress through Highway Trust Fund contract authority or ARRA stimulus appropriations. But it has made direct loans of $28.75 billion over the same period. That means that every dollar of budget authority used by the program led to over $15 of direct loans for infrastructure projects. And this is in the real world, not in an economic model.

According to the project lists on the TIFIA website, the TIFIA loans were used as part of transportation projects with a total cost of $108.3 billion, so the TIFIA loans would up being about 26.6 percent, on average, of the total cost of the project. The loan proceeds were usually combined with federal highway or transit formula funding and state or local revenues or other borrowing, or a public-private partnership component as well.

TIFIA and other loan programs are not a panacea. For years, the program has had trouble finding enough projects that qualify for the loans in order to utilize all its available subsidy funding. (The Trump plan would make more kinds of projects, like airports, ports, commuter rail, and new kinds of water projects eligible for federal infrastructure credit programs.) And one can’t argue with a straight face that none of these projects would have been build in the absence of a federal loan – often, state or local governments would probably have found some other way to finance the project at a greater cost.

But the Penn-Wharton analysis errs in assuming that $14 billion in subsidy appropriations for a federal credit program behaves anything like an appropriation of similar size for a grant program. The economic effects of the same amount of money appropriated for a credit program could be an order of magnitude higher than the economic effects of an identically sized appropriation for a grant program, because of the accounting rules put in place by the FCRA law in 1990. At a minimum, Penn-Wharton should be using the estimated loan levels made available by the $14 billion, not the $14 billion itself.