We're right to unpack the assumptions that go into congestion costs, but we can use the analysis for more equitable policies, argues Paul Kishimoto.
Tricia Wood’s recent Torontoist transit column “The Real Cost of Congestion in Toronto” has two theses:
- [Congestion cost] numbers count a lot of “lost productivity” that is not actually lost.
- A) Those numbers focus our attention on people who are already moving but stuck in car traffic.B) We should focus on the lost productivity of those who are not moving at all—that is, those who are unable to participate in the city’s economy because they can’t get where they need to go.
2B is not in dispute. This is the equity and social justice argument for supporting active transport (walking, biking, etc.) and mass transit in cities. Mobility means access to participation in society: employment, education, social activities, leisure, medical care, political engagement, and more. If our transportation system fails to provide this mobility to some people, that is a real problem and a reason to care about finding the right solutions. It’s important to talk about.
2A is also broadly true. Wood quotes two figures from Transport Canada, by way of an OECD report: $3.3 billion in “congestion costs for commuters” and $2.7 billion in “economic costs,” then focuses her critique on the first. Indeed, it’s easy to compute a rough estimate of the cost to single drivers in private cars:
- Measure the extra minutes needed to travel a given stretch of road under congestion, compared to free traffic flow,
- Count cars, and multiply by the number of persons experiencing that delay,
- Multiply by some figure for the value of travel time, in dollars per hour.
Other categories of costs, and the travel time costs of other users (which we’ll come to shortly) are more difficult to estimate. This means they are less often estimated, and we have fewer such estimates; and public conversation, in turn, tends to center on the available figures pertaining to drivers. We should be aware of, and correct for, this bias.
The most important conclusions flow directly from 2A and 2B: Torontonians need more, broader, and more detailed metrics of the costs of congestion and travel (on roads and transit lines) and of an inadequate transport system. We need measurement and analysis that provides this data, information, and knowledge, and we should be willing to pay for it. All three levels of government should invest in robust, reliable, ongoing monitoring of our transport system—including the people who it fails to serve. Both governments and news outlets should rise to the task of helping the public understand the full detail of these measures, instead of condensing them to soundbites. If one looks at other cities, there are plenty of examples of how to do this.
Are congestion costs real?
None of this, however, requires us to throw out the concept of congestion costs. In any case, Wood’s critique (the first point) fails, mainly because it does not engage with the working definition of the concept, or the detailed results of decades of research by transit engineers, planners, and economists.
Economic policy analysis is often hard because it requires a counterfactual—a thing that, by definition, does not exist.
Here are some examples:
- If there were a public transit option between Point A and Point B that took less than 90 minutes, then Alice would take a course at Ryerson University.
- If Bob spent 10 fewer minutes driving home from work, then he would spend that time playing catch with his children.
The challenge is that Bob does spend those 10 minutes in traffic, so we don’t know for certain what he would do with the extra time. We can only accept his word for it.
Likewise, Alice does not have that transit option available; so we can only take her word on what she would use it for.
On an individual, anecdotal level, we may find either Alice’s or Bob’s story more compelling—or more appealing to our sense of justice. But neither is inherently more or less artificial.
What’s more, in order to understand the effect of proposed changes, we ultimately need to work with aggregates. We must determine there are a certain number of people like Alice, some other number of people like Bob, like Carol, Dave, and so on. Then we must assume that the people in each group would all make, on average, the same adjustments to their behaviour in response to a change in the transport options available to them. Finally, we need metrics that help us compare across groups of people and categories of impact.
What are the costs?
Travel time costs are surprisingly complex and varied. There is no single figure for the value of time.
Some of the facts known from past research:
- Costs vary depending on the purpose of a trip. For the same person, a 10-minute delay on a trip to Wonderland has a different value than a 10-minute delay for an important business meeting.
- Costs vary across modes and trip segments. Transit riders are generally more willing to accept five extra minutes on the subway than five extra minutes on the bus; and they are more willing to accept five minutes riding than five minutes waiting for a vehicle.
- Costs also vary across with time of day, day of week, weather, and seasons—those five minutes waiting for a bus are worse in the winter.
- Costs vary according to the age, gender, income, and other attributes of the individual traveller.
- Costs vary according to the attributes of individual modes. People will pay more for a seated subway trip with Wi-Fi, no crowding, and working air conditioning, than for a standing trip with no connectivity and their face in someone’s sweaty armpit. In the future, a long ride in a self-driving car may be more tolerable than one in a driven car, because one can use that time to work, or read a book.
How are these measured? In one of two ways: we can simply ask people what they prefer; or we can look at real-world travel data and see what people chose when faced with different options.
In either case, we can find or introduce an element of cost to these choices. For instance, we can ask: which would you rather take—
- a 22-minute subway ride for $3.25, or
- a 25-minute subway ride for $3.00?
This question implicitly values time at 25 cents for every three minutes, or $5/hour. By varying the time and money increments in the questions posed to different respondents, we can find a value at which people are indifferent—they are equally likely to accept $x lower cost, or one minute of reduced subway travel time. $x/minute becomes our value of travel time for this group.
All the listed facts were discovered by some variant of this method. We may find, for instance, that Bob reports his income as $30 an hour, but he’s only willing to pay $3 for a hypothetical toll road that gets him home 10 minutes quicker, so he values his driving time at $18 per hour. Alice may be unemployed, but will report or show us that the value of bus travel time to her is $15/hour, and the value of bus waiting time $30/hour—that is, she would pay $1 more for a bus trip that involved two fewer minutes of waiting outside in the winter for a transfer.
Particular numbers are context-specific, and definitely tied to the attributes of the people used to estimate them. But they represent the real preferences of real people.
Comparing apples and oranges
Wood’s citation doesn’t support the point for which it’s used. Gössling and Choi’s Table 1 (p. 109) lists parameters used in Copenhagen’s transport cost-benefit analysis; the third entry is “time costs.” More detailed measures of the costs (and benefits) of travel do not exclude time costs (because they are fundamentally unreal, or for any other reason); they augment them by placing them alongside other categories of costs and benefits.
These include injuries and repairs from collisions; the recreational value of active transport; and the (local) health and (global) climate damages of pollution. Comparing across categories is certainly challenging. Each cost is borne by different people. For instance, poor people on the other side of the planet may suffer the climate damages from greenhouse gases emitted by cars (or the electric generators that power transit), but pedestrians on Toronto’s streets are the ones faced with risk of injury from collisions. Sometimes, uncertainty means it is impossible or misleading to replace an impact with a monetary value.
Even within one category, like travel time costs, simple aggregation can be problematic and can obscure important dimensions of policy, such as equity. We might estimate that Policy X would save $3 billion in travel time costs for bus riders like Alice; while Policy Y would save $3 billion for drivers like Bob. The totals are identical. But knowing the travel time valuation of these groups, we can notice that the benefits of Policy Y are concentrated among a relatively small number of people; while the benefits of Policy X are distributed more broadly.
People with different political views can—and must—form their own opinions about whether Alices or Bobs are more deserving of the benefits of new transit spending. Likewise, it must be a public decision whether $1 of benefits to Alice, because of her membership in some disadvantaged group, outweigh $1 of benefits to Bob. And in practice, every policy has a complex mix of short- and long-term effects on every category of costs, as pertain to every group of users and would-be users of the transport system.
Again, Wood is right on the political: when we know that Policy Y would save the Bobs of Toronto $3 billion, yet we lack a corresponding total figure for the benefits of Policy X to Alice and co., then the former has a powerful rhetorical advantage in the battle amongst competing uses of public money.
But that’s no reason to tear up the literature on travel time costs—we need those numbers, as much as other measures of the costs of our travel, to motivate change towards a sustainable, equitable transport system. We also need to defuse opposition by showing that policy changes which reduce driving are beneficial to drivers: they can switch to walking or biking on improved streets, or riding better transit; or the choice of others to leave their cars at home will mean lower congestion costs.
Paul Kishimoto is a PhD candidate in Engineering Systems at MIT, and a longtime Torontoist commenter.