A new study by the nonpartisan Congressional Budget Office finds that “Under typical conditions, after receiving an additional dollar in federal highway [formula] grants, state and local governments spent 26 cents less of their own funds on highway capital projects than they would have otherwise.”

The study, entitled “Fiscal Substitution in Spending for Highway Infrastructure,” is a follow-up to a 2018 CBO paper which found that, while such “fiscal substitution” effects did exist for federal highway spending, they weren’t ready to assign a percentage estimate to it. The 2018 report noted that, after an extensive literature review, “For an increase of $1 in federal grants, most estimates suggest, state and local governments would reduce spending on highways from their own funds by between $0.20 and $0.80.”

The new study finds a $0.26 state-local spending reduction for every $1 – while significant, it is important to note that this is on the low end of the earlier estimates cited in the 2018 paper.

For the new analysis, the CBO authors looked at the same collection of earlier studies analyzed by the 2018 CBO paper, some of which date back as far as 1987, plus a new 2020 NBER paper released since then. But they also ran a lot of original analysis based on many different potential sets of circumstances.

CBO found that “two-thirds of the time, the reduction from a similar increase in grants would be between 14 cents and 38 cents.” The 26 cent figure splits the difference between those two numbers (which, remember, are much lower than the 20 cent to 80 cent range from the paper three years ago). (It’s actually more complicated than that – the coefficient was -0.24, plus another 2 cents related to state balanced budget requirements.)

The authors of the CBO study make the following points about how their analysis differs from the several other studies done on this topic:

• “First, we include spending by both state and local governments, whereas much of the literature refers only to spending by state governments…Some states keep spending decisions largely centralized at the state level, whereas others transfer more of the responsibility to local governments.”
• “A second way that our analysis differs from analyses in much of the existing literature is that our primary interest is in spending on highway capital projects…By contrast, most of the existing literature considers the effect of federal grants on total state and local government spending for highways—not only capital spending but also operation and maintenance costs and the costs of other highway-related programs, such as educational programs focused on highway safety.”
• “…most prior analyses have treated all federal grants the same. In fact, programs differ in terms of their criteria for distributing funds, rules on project eligibility, state and local matching rates, and how easily funds can be shifted to other transportation-related priorities. We distinguish among three types of federal grants: formula grants, ARRA grants, and allocated grants.”
• Other studies used federal outlays as a measure of federal spending increases, which (for these programs) are the joint federal-state/local reimbursements to contractors made after a project is completed, and are thus a mix of federal and state highway decisions in the same variable. Instead, the new CBO study uses “the amount of money the federal government has made available to a state in a given year but has not yet committed to a project or paid to the state. Those amounts, called obligation limits, are highly correlated with federal grant outlays but are determined independently of states’ spending decisions.”

Once CBO built its model, they continued to tinker with it. For example, the initial model used personal income and state motor fuel tax collections as variables to try and estimate how much money a state had to spend, but the Alaska and North Dakota spending numbers became outliers, so they came up with a variable for certain states that get a large share of their tax dollars from resource extraction to make AK and ND make sense.

In addition, there are always individual one-off things that throw off a trend, and the CBO model created “dummy variables” to deal with those. For example, when a U.S. city is awarded an Olympic Games as a host venue, there are a lot of one-time federal infrastructure dollars awarded, so the new CBO model adjusts for Georgia’s 1996 and Utah’s 2002 Olympic infrastructure awards.

CBO even, somehow, came up with a good adjustment not just for federal October 1 – September 30 fiscal years to state July 1 – June 30 fiscal years, but also a plug-in for the four weirdo states (Alabama, Michigan, New York and Texas) which use other fiscal years.

In 2017 dollars (PPI adjusted, not CPI):