Transportation Improvements and More Driving

I am working on a post for my Scaling project on the rebound effect with transportation. That is, when a new road is built or an old road expanded, when remote work becomes more prevalent, when access to destinations becomes more convenient with mixed use development, or with other kinds of improvements, do people drive more in response? Or, is the reduction in driving less than what one might naively expect? The answer is generally ‘yes’. We’ll take a partial look at why this is the case and what it means in practice.

Unlike my previous two posts, I am writing the material freshly tonight, as what I have planned for the Scaling site is not done and probably won’t be done until next week at the earliest. Consequently, I will revert to a more informal style.

The basic issue to keep in mind is this. We know that, in general, road construction or widening occurs when there is heavy traffic congestion, with the goal of relieving congestion. However, the congestion relief makes driving more attractive and encourages people to drive more, which (at least partially) causes the system to revert to a congested state. Does this mean that road expansion is a fool’s errand? Think about it for a minute, and then we’ll look at some research.

Induced and Latent Demand

The most common term in the popular literature for the phenomenon just described is induced demand. When driving becomes more attractive, people want to do more of it, and thus demand increases.

Every economist would, I hope, immediately raise objections about what we mean by “demand”. In any market situation, for which road capacity no exception, demand is not just a static value, but rather a function of cost—hence the phrase “demand curve”. In the case of driving, costs are not necessarily direct monetary costs, but also include the opportunity cost of the time spent driving. Most readers probably do not rank driving at the top of their favorite list of activities, especially if that time is spent being stuck in traffic. Thus the demand curve is a function for which the input is cost and output is an amount of driving.

For this reason, the phrase latent demand indicates that increased roadway capacity does not create a demand for driving, but rather allows motorists to fulfill demand that was previously too costly. Clifton and Moura (2017) offer a more through discussion of the issue.

I am not aware of any papers that make a serious attempt to delineate between induced and latent demand for driving (if anyone is aware of such a paper, I would be interested to know). Many use the terms interchangeably.

Rather than offer a sharp, binary distinction, Clifton and Moura present a hierarchy of six types of demand: realized travel (Level 1); scheduled travel (Level 2); tentative and planned travel (Level 3); aspirations and intentions (Level 4); dreams, desires, and possibilities (Level 5); and unimagined travel (Level 6). Parthasarathi, Levinson, and Karamalaputi (2003) and Næss, Nicolaisen, and Strand (2012) both split the hairs of the terms, yet their analyses leave no indication as to whether we should regard increased driving as induced or latent demand.

I find the distinction to be unhelpful. I am not aware of any serious work that convincingly parses the distinction between induced and latent demand, let alone quantifies it. The issue is not merely semantic; it proxies for some rather fundamental assumptions about urban planning will be interrogated later in this post.

For two reasons, I prefer to use the term rebound effect. The first reason is as described above. The second reason is that the rebound effect is a well-known principle in energy economics, and I find it useful to analogize between these two things. And so let’s take a brief digression into the rebound effect for energy.

The Rebound Effect

We expect energy efficiency to save energy, and at first glance it seems straightforward to estimate how much energy is saved.

The United States uses about 13 million barrels of oil per day to drive cars. Automakers are always working on ways to make cars more fuel-efficient, and so let’s say that over the next few years, technological innovations allow a 20% savings in fuel consumption That works out to 2.6 million barrels per day, and so that much oil is left in the ground. Right?

Well, no; there are many mechanisms by which the amount saved is less. Because driving is now less expensive, consumers may prefer larger or more powerful cars, which are less fuel efficient. They may choose to drive more. Or, maybe there will be savings on the oil used for driving, which lowers the price of oil and makes flying more attractive, increasing the amount of aviation. By these and other mechanisms, a portion of the expected 2.6 million barrel savings goes back into additional consumption. That portion is called the rebound. If, for instance, only 1.3 million barrels are saved and 1.3 million barrels are used for additional consumption, we say that the rebound is 50%. There might be no savings at all, in which case the rebound is 100%.

It is theoretically possible that rebound is greater than 100%, a situation that the British economist William Stanley Jevons posited in 1865 as applying to energy efficiency in coal consumption for steam engines. See also Saunders (1992) for a more recent treatment. He refers to the commonality of >100% rebound as the Khazzoom-Brookes postulate and regards the phenomenon as a primary driver of economic growth under neoclassical growth theory.

Rebound effects can be found in many areas. York (2012) finds that rebound occurs on the supply side as well as the demand side: he concludes that for every joule of low-carbon energy that is added to the energy system, 0.75 joules go to new demand or displacing other low-carbon energy, and 0.25 joules go to displacing fossil fuels. Greiner, York, and McGee (2018) find that natural gas for electricity generally augments, rather than replaces, coal supplies. Qui, Kahn, and Xing (2019) find a rebound from residential solar panel adoption of 18%, meaning that 18% of the energy generated by the panels goes to fulfilling new demand rather than displacing grid electricity.

Outside of the energy world, Ewers et al. (2009) find a rebound in agricultural land for improved yields. Li and Zhao (2018) and Lock and Adamson (2015) find rebounds for water efficiency in irrigation. Pfaff and Sartorius (2015) and Lifset and Eckelman (2013) find rebounds for material efficiency. And Longo et al. (2019) finds that aquaculture generally augments rather than replaces wild-caught fish.

It should be noted that the results described here are contentious, and in the interest of space, I am forced to gloss over some controversies. But I think it is clear that, whatever the precise magnitude, rebound is a widespread phenomenon. At this point you may be thinking, “Well, no [stuff], that’s the whole point of technological innovation, isn’t it?” Hold that thought.

Rebound from Road Expansion

Let’s start with the most well-studied example of rebound from the transportation world. This blog post finds that the history of the induced demand concept for road construction goes back at least to William Haywood on the Holborn Viaduct in 1866. Perhaps the most famous work is Downs (1962), which argues that the new expressways being built at the time would not relieve traffic congestion because the rebound effect would equal the increased capacity.

Todd Litman of the Victoria Transport Policy Institute has done some of the most extensive work on the subject in recent years. Litman (2017) posits these mechanisms by which rebound may occur.

• Traffic shifts from other roads to the newly expanded road.

• Traffic shifts from off-peak hours to peak hours.

• Travel mode shifts from walking/biking/mass transit to driving.

• Longer trips (see my discussion of Marchetti’s Constant).

• Trips that otherwise would not have occurred at all.

• Land use changes to more spread-out development.

The paper goes on to discuss a bunch of studies on the magnitude of the rebound effect. Most of the studies are based on measuring traffic on roads before and after expansion. Some are theoretical models rather than empirical measurements. Some rebounds are based on the speed of travel, and some are based on the lane-mileage of road capacity (a distinction that I will gloss over here in the interest of space, but for our purposes today is not terribly important). Of the studies that give specific numbers for rebound, those numbers tend to cluster between 50% and 100%, with some outliers in both directions.

By now, an astute reader will have noticed several problems with my presentation. The first problem, discussed in Dunkerley et al. (2018), is that it is unclear that the first two mechanisms of rebound described by Litman should actually be considered rebound. If we consider a single road, then shifting from another road to the newly expanded one is a rebound, but from the perspective of the full road network, it is not. Likewise, time shift is a rebound if we focus on peak hours, but not if we focus on travel across all hours.

Common to all literature reviews, a second problem identified by Dunkerley et al. is publication bias. They contend that studies finding higher rebound effects are more likely to be published that those that find lower effects, thus skewing published numbers upward. Not discussed in the paper, there is also clearly a bias in which projects are done. Those roads with high congestion, and thus more likely to have a greater latent demand, are more likely to be widened. Widening a road that already has little congestion would probably have a smaller rebound, but this would be regarded by most citizens as a waste of money and would probably not be done.

A third problem, discussed in Van der Loop, Haaijer, and Willigers (2016), comes from the methodology of measuring pre- and post-widening traffic to determine rebound. How do we know how much of the increased traffic is due to regional growth that is independent of the widening? An accurate measure of the rebound effect should determine not how much traffic increased, but how much traffic increased as a result of the widening. This, unfortunately, is much harder to determine.

It is clear to me that a rebound effect from road expansion does happen, but the conclusion of Downs (1962), that the rebound is generally close to 100% (the increase of traffic is equal to the increase of road capacity) strikes me as implausibly high and not true in general.

Rebound from Remote Work

When a person switches some of their work days from in-person to remote, then they won’t have to drive to the office those days (assuming they previously drove). That’s a savings in driving, and we now know enough about rebound to expect that at least some of the savings will go back to additional driving. How much?

In my discussion of Marchetti’s Constant, I cited a few studies indicating that remote work should reduce overall travel; i.e. that the rebound is less than 100%. I had a fruitful discussion with a reader who suggested that the rebound is in fact at or above 100%, or that remote work should increase overall driving. I found the conversation to be valuable and educational, but I am not convinced that backfire (>100% rebound) generally occurs. Let us consider some additional research.

First off, we should clarify that there are several types of rebound here. For the most part, I have been focusing on the number of miles driven. There is also rebound from an energy perspective. Perhaps remote work reduces the amount of driving, but for whatever reason it causes more energy to be consumed on other activities.

Hook et al. perform a systematic literature review on the energy effects of remote work and find 39 studies that meet their criteria for analysis. Of those, 26 show a clear savings in overall energy consumption. The remaining 13 are either ambiguous or show an increase. However, the authors note that the studies don’t consider the full macroeconomic impacts of remote work. For example, maybe the energy savings from not driving boosts economic growth, which causes increased consumption generally.

Considering more direct rebounds, we would expect remote work to stimulate more travel for purposes other than commuting, as per Marchetti’s Constant. It may also encourage people to move farther from where they work, or take jobs farther from where they live. A long commute may not be acceptable if I have to do it five times per week, but if I only have to commute twice per week and can work remotely the other three days, the long commute may be acceptable. Such is the finding of Zhu (2011), who finds that for long-time telecommuters, the ratio of the commute length between telecommuters and non-telecommuters grew from 2001 to 2009. Helminen and Ristimäki (2007) do find a driving savings among telecommuters in Finland, relative to non-commuters, but the rebound is close to 100% due to telecommuters living farther from work. De Abreu e Silva and Melo (2017) make a similar finding for the UK.

Other Rebounds

Mixed use development is a methodology that plans for higher densities and a variety of land uses in a small area. This makes trips more convenient by shortening distances and making walking feasible more often. Sperry, Burris, and Dambaugh (2012), by conducting travel surveys at a mixed use development in Dallas, Texas, found that the convenience of mixed use also spurs a rebound, in that residents make more trips that they would if destinations were farther away, though they still find an overall saving in driving (i.e. rebound is less than 100%). I have not found many studies with deal with this topic.

Mass transit is a favored solution to traffic congestion, as busses and rail can transport the same number of people with much less space. You have probably seen an image like this one at some point.

Image from Reddit, beyond which I have not traced its origins.

So, we should invest more in public transit, get people out of their cars, and relieve congestion. But you know the drill now. According to Beaudoin and Lawell (2018), who analyze a panel data set of 96 urban areas in the United States from 1991 to 2011,

… In the short run, when accounting for the substitution effect only, we find that on average a 10% increase in transit capacity leads to a 0.7% reduction in auto travel. However, transit has no effect on auto travel in the medium run, as latent and induced demand offset the substitution effect. In the long run, when accounting for both substitution and induced demand, we find that on average a 10% increase in transit capacity is associated with a 0.4% increase in auto travel. …

I’ve glossed over mass transit and induced demand for now, but it should be noted that this study suffers from similar limitations as discussed above in the context of road widening.

What’s the Solution?

Rebound, or induced demand, for driving appears to be a universal phenomenon, and it applies to solutions that urbanists like just as well as it does for those they don’t like. For this reason alone, induced demand makes a poor cudgel to be used against road widening.

I strongly suspect that, while rebound is real, backfire is not something that occurs in general, and that the notion traffic always expands to fully fill the capacity of the road network is simply not true. Thus all the things described above—more roadways, remote work, mixed use development, and public transit—do in fact relieve congestion. The problem isn’t that we are doing these things, but that we aren’t doing them enough.

If driving is caused by the availability of roads, this is rather difficult to explain. Image from the Federal Highway Administration.

Steven Polzin at the Reason Foundation highlights pricing as a solution to the negative effects of driving. The greatest negative externality is congestion, and so congestion pricing, such as Singapore’s Electronic Road Pricing system, is an obvious solution.

What’s the Problem?

It is axiomatic among urbanist activists and modern planners that cars are bad and that the central goal of urban planning should be to reduce car usage. Seldom do they subject this belief to critical scrutiny and see what a poor basis for policy it is.

Many of the papers I cited above, as well as popular articles on the subject, present induced demand as though it is some sort of scandal associated with road widening and as though it is a novel idea. But the whole point of the road system is to facilitate travel. Cities would not be functional without travel. Indeed, if a road widening occurred and there wasn’t an increase in traffic, then I would wonder about the wisdom of public expense on the project.

Increased traffic means more people get where they want to go. In order for the rebound to be a bad thing, it must be that the negative externalities of driving outweigh the benefits. Is there evidence for such a situation? Næss, Nicolaisen, and Strand (2012) find that due to rebound and the negative externalities of driving, principally traffic congestion, the benefit of a road widening project in The Netherlands is reduced by 30% compared to what the benefit would be without considering rebound. But this calculation fails to account for the private benefits of increased travel, and while the paper acknowledges the issue, it offers no way of guessing how the benefits and costs of rebound compare to each other.

To his credit, Litman (2017) does attempt to put numbers on the issue. He argues that the marginal benefit of travel decreases with volume—something that would be expected with any demand curve—and finds that the costs outweigh the benefits. He rather crudely assumes that the benefit of the rebounded travel is worth half of the time savings for existing motorists. This assumption is defensible, but I’ll admit that I need to look at the issue more carefully.

By looking at driving, a fundamental activity in the modern world, strictly as a problem that needs to be curtailed, the whole thing comes across as uncomfortably similar to Malthusian ecologism. Some studies make it explicit, such as Khmara and Kronenberg (2022), which states,

Perhaps the closest [among several urban planning philosophies] to degrowth values is the concept of compact cities, which implies dense urban structure with mixed land use, thus reducing the demand for intra-city travel, land, and infrastructure.

Xue and Kębłowski (2022) write in a similar vein. Not all advocates for car-free development are degrowthers, but there is enough nexus between the ideas that one cannot help but be suspicious of the movement as a whole.

To conclude, induced/latent demand and the rebound effect are very real phenomena in transportation systems, though their magnitude remains poorly understood and quite possibly severely overestimated. In contrast to how the idea is most often portrayed in the popular media and academic literature, rebound is a natural and healthy element of how cities function, not some kind of scandal or a novel idea. The way induced demand has so frequently been portrayed is truly a case of how a little knowledge of economic principles is a dangerous thing.

Quick Hits

The articles in the first issue of the Markets and Society journal are all good reads, at least those I have read so far. A particularly interesting article is Dekker and Quintas (2024), The Night the Line Was (Not) Crossed: The Use of Repugnance for Product Differentiation. The article discusses how Extreme Championship Wrestling, which existed from 1993 to 2001, used repugnance as a marketing tool, and how ECW was able to portray itself as more in violation of social norms than it actually was. For a time, ECW carved out a successful niche, but the problem was that as the novelty of their norm-violating behavior wore off, ECW was forced into more dramatic actions in a cycle that proved to be unsustainable. Several imitators tried unsuccessfully to copy ECW’s formula, demonstrating how repugnance is an inherently limited market niche.

My piece on the ideological origins of 9/11 last September was long, and I was forced to cut many topics, including what I most wanted to write about: the motivations of the attackers themselves. To that end, I recommend the 2001 documentary, A Mission to Die For. It is well-done and surprisingly expansive for a film that came out only two months after the attacks. The film is a biography of Mohamed Atta, the State-side ringleader of the plot and the hijacker-pilot of American 11, which crashed into the North Tower of the World Trade Center. Despite the extensive details of Atta’s life that are presented, I find that how a person would be compelled to perform such an evil act remains beyond comprehension.

Henry Sokolski of the Nonproliferation Policy Education Center has a new, short, and freely available book China, Russia, and the Coming Cool War. The phrase “Cool War” makes the contest sound less serious than it is, but this quibble aside, I find the work insightful, if at times a bit banal. It is a high-level overview of the imperatives for the United States and her allies to come out on top in the emerging geopolitical contest with China, Russia, and the other countries in their orbit.