With healthcare reform stalling, President Trump’s administration seems ready to shift the focus to infrastructure.

Good infrastructure, especially highways, bridges, and airports, can certainly improve economic mobility and lower costs by reducing travel time between locations. This, however, says nothing about the kinds of institutions most likely to produce good infrastructure or who should fund it.

Here’s a handy guide to some of the bad economic reasoning you will likely hear as the debate about infrastructure spending heats up.

1. Past benefits don’t mean future benefits

In his joint address to Congress, President Trump declared that “the time has come for a new program of national rebuilding.” The implication was clear: building new infrastructure was a success in the past, so it would be good for the economy today.

Past experience and the experience of other countries lead to mixed conclusions about the value of public infrastructure project. Highway construction can substantially boost productivity for industries associated with road use, but the same research finds those benefits to be largely one-offs. More recent research has found that too many new highways were built between 1983 and 2003. It has also found that marginal extensions to the highway system are unlikely to increase social welfare because the cost savings from reduced travel times are relatively small.

We should judge new projects on their own merits, not against old examples or countries in different circumstances.

2. Don’t ignore opportunity costs

“Traffic Congestion Costs Americans $124 Billion A Year” is a headline from 2014. As legislation for infrastructure is pushed, we will hear plenty about the costs of delays to the economy.

These costs are undoubtedly very real, but so are the costs of building new infrastructure, and that money can’t then be spent on other things that we might have preferred to spend them on. Without the aid of clear market signals, it’s very difficult and maybe impossible for governments to determine the optimal amount and nature of infrastructure spending. It would obviously be prohibitively expensive to eliminate all congestion by expanding every freeway to 15 lanes but building no new highways would also be problematic. How far should a road expansion go? How often should it be repaired? How much transportation should go by train? How much money should be spent on research and development for completely new ways of meeting transportation demand? Markets are good at finding the optimal mix over time and rewarding those who are better at satisfying demand. Governments, even with the best of intentions, lack the necessary knowledge about each of our individual opportunity costs to find that mix. They certainly lack the incentive structure to improve over time.

3. But what about that one bridge…?

Individual catastrophic events can lead to concern about the physical conditions of infrastructure. The I-35W bridge collapse in Mississippi in 2007 is a recent example. Even more recently, commentators have used the I-85 bridge fire and collapse in Georgia as justification for more infrastructure investment. But these are rarities that tell us little about the quality of infrastructure overall.

You often hear that 58,791 bridges are structurally deficient, for example, which sounds kinda scary. Less often will you hear that, according to the Federal Highways Agency, “structurally deficient does not mean that it is likely to collapse or that it is unsafe.” You also won’t hear that the proportion of bridges labeled structurally deficient has fallen from 24.1 percent in 1990 to 9.6 percent in 2015.

When you hear individual statistics on the dire state of U.S. infrastructure, ask questions like, “compared to what?,” “how has this changed over time?,” and “is there a demand for this to be replaced?”

4. Cheap debt doesn’t make everything a bargain

In 2015, Nobel Laureate Robert Shiller argued that “the government should be borrowing, it would seem, heavily and investing in anything that yields a positive return.” The Brookings Institution recently spelled out similar logic, suggesting that low interest rates should also be inducing private sector investment.

The mistake here is to conflate a less costly time to invest with a “good time” to invest. Take the example of a toll road. If the long-term growth and population outlook for an area has seriously slowed, then the expected use of that toll road would fall, as would demand for investment opportunities. This would cause lower interest rates, all else being equal.

Those lower interest rates, however, would not indicate that it was a good time to invest. They would be signalling lower expected future revenues.

Similar logic applies to government investment in transport infrastructure without user fees — if there are structural reasons why demand for transportation use is falling, then any investment would yield far fewer economic benefits.

Infrastructure decisions should be judged by robust estimates of costs and expected benefits, not just how cheap it is to borrow.

5. How stimulus actually works

Leaked documents show that the Trump administration is likely to prioritize “shovel ready” projects and those that are “direct job creators.” But previously on his campaign website Trump’s team had suggested the goal of infrastructure development was “more  rapid productivity gains.”

This conflates two well-known arguments for infrastructure investment. The first is that government investment spending can be used to “stimulate” the economy and put people back to work. The second is that smart, efficient investments can help enhance long-term productivity growth.

These two ambitions often conflict. Attempts to stimulate quickly and get people back to work will likely result in sloppy project selection and the hiring of more labor than would be most efficient. And since government is, well, government, it’s a pretty good bet that infrastructure funds will go preferentially to the well-connected.