Travel Time Perception Errors: Causes and Consequences

Congratulations to soon to be Dr. Carlos Carrion (shown in the center of the picture, between alums Nebiyou Tilahun and Pavithra Parthasarthi), who recently defended his Ph.D. Thesis “Travel Time Perception Errors: Causes and Consequences” (a draft of which is linked). He is working as a post-doctoral researcher at MIT/SMART in Singapore.

Travel Time Perception Errors: Causes and Consequences
This research investigates the causes, and consequences behind travel time perception. Travel times are experienced. Thus, travelers estimate the travel time through their own perception. This is the underlying reason behind the mismatch between travel times as reported by a traveler (subjective travel time distribution) and travel times as measured from a device (e.g. loop detector or GPS navigation device; objective travel time distribution) in collected data. It is reasonable that the relationship between subjective travel times and objective travel times may be expressed mathematically as: Ts = To + ξ. Ts is a random variable associated with the probability density given by the subjective travel time distribution. To is a random variable associated with the probability density given by the objective travel time distribution. The variable ξ is the random perception error also associated with its own probability density. Thus, it is clear that travelers may overestimate or underestimate the measured travel times, and this is likely to influence their decisions unless E(ξ) = 0, and Var(ξ) ≈ 0. In other words, travelers are “optimizing” (i.e. executing decisions) according to their own divergent views of the objective travel time distribution.
This dissertation contributes novel results to the following areas of transportation research: travel time perception; valuation of travel time; and route choice modeling. This study presents a systematic identification of factors that lead to perception errors of travel time. In addition, the factors are related to similar factors on time perception research in psychology. These factors are included in econometric models to study their influence on travel time perception, and also identify which of these factors lead to overestimation or underestimation of travel times. These econometric models are estimated on data collected from commuters recruited from a previous research study in the Minneapolis-St. Paul region (Carrion and Levinson, 2012a, Zhu, 2010). The data (surveys, and Global Positioning System [GPS] points) consists of work trips (from home to work, and from work to home) of subjects. For these work trips, the subjects’ self-reported travel times, and the subjects’ travel times measured by GPS devices were collected. Furthermore, this dissertation provides the first empirical results that highlight the influence of perception errors in the valuation of travel time, and in the dynamic behavior of travelers’ route choices. Last but not least important, this dissertation presents the most comprehensive literature review of the value of travel time reliability written to date.

Traffic Congestion and Job Accessibility

KFAI Community Radio in Minneapolis

I was interviewed a few weeks ago by Dale Connelly for KFAI Community Radio. The edited interview (aired April 22, 2013) is below as an MP3 (5:51)

Every so often we get a discouraging report on traffic congestion in metropolitan areas.
The latest rating from Texas A & M’s Transportation Institute gave Washington DC the worst rating for congestion, followed by Los Angeles, San Francisco, New York and Boston. No big surprise there.
But one University of Minnesota professor says counting stationary cars is only part of the story. David Levinson is a professor of civil engineering and author of the Access Across America Study. He told KFAI’s Dale Connelly there’s more to consider when looking at the problem of traffic congestion.
David Levinson holds the Richard P. Braun/Center for Transportation Studies Chair in Transportation at the University of Minnesota. His study, Access Across America, says some of the cities regularly identified as most congested actually have transportation networks that provide good access to jobs. You can see the study online at


Street Improvement Fees

The League of Minnesota Cities writes:

Briefing paper—2013
Minnesota cities and street improvement districts
League position
The League supports HF 745 (Erhardt, DFL-Edina) and SF 607 (Carlson, DFL-Eagan), legislation that would allow cities to create street improvement districts. This authority would allow cities to collect fees from property owners within a district to fund municipal street maintenance, construction, reconstruction, and facility upgrades. If enacted, this legislation would provide cities with an additional tool to build and maintain city streets.

Sounds like a good idea to me. To be fair, there are opponents. The stated opposition seems odd. They oppose this tool because it is not voter approved, yet I don’t ever recall voting on property tax hikes, or sales taxes for stadia which are imposed on me. The real opposition is because it shifts the burden from one class of taxpayers to another, hopefully so that it is more closely aligned with benefits.
At any rate, our research on the similar Transportation Utility Fees is:

Drew Kerr has a pay-walled article at Finance & Commerce here.

Evaluation of the Minnesota Road Fee Test

MnDOT has posted the evaluation of the recent Mileage Based User Fee test: Connected Vehicles for Safety, Mobility, and User Fees: Evaluation of the Minnesota Road Fee Test:

“In 2007 Minnesota legislature approved a $5,000,000 project in order to demonstrate technologies which will allow for the future replacement of the gas tax with a fuel-neutral mileage charge. The Minnesota Department of Transportation (MnDOT) organized a study to examine the implementation and operation of a mileage based user fee program (MBUF), which might allow for the supplementation or replacement of traditional gas taxes. The primary objectives of the study were to: assess the feasibility of using consumer devices for implementing Connected Vehicle and MBUF applications. These applications included localized in-vehicle signing for improving safety, especially for rural areas, and the demonstration of the proposed Connected Vehicle approach for providing location-specific traveler information and collecting vehicle probe data. The study consisted of 500 voluntary participants, equipped with an in-vehicle system comprised of entirely commercially available components, primarily a smartphone using an application capable of tracking participant vehicle trips. Successfully meeting its primary objectives, the system was capable of assigning variable mileage fees determined by user location or time of day, as well as presenting in-vehicle safety notifications which had measureable effect on the participants driving habits. MnDOT contracted Science Applications International Corporation (SAIC) to perform research for the project and an evaluation of its findings. This document is the final report from SAIC, providing a summary of the study, its findings and an evaluation of the project as a whole.”

The rates tested were:
1. Outside Minnesota miles – $0.00 per mile;
2. Inside Minnesota miles – $0.01 per mile;
3. Twin Cities (“Metro Zone”) – Peak miles – $0.03 per mile; and
4. “Non-Technology” miles – $0.03 per mile.
It looks like there were some technology problems in the experiment (having worked with GPS and in-vehicle devices for research, I believe this is still emerging technology with imperfect reliability, insufficient for mainstream application):

As mentioned previously, trip data was only available for 57 percent of trips generated by the system. Of the 43 percent of trips where trip data was not available, 69 percent of the trip data loss was due to a vehicle detection failure. Trip data was only recorded if the system could both detect the device was in the correct vehicle and a valid GPS signal was found. Therefore, the remaining 31 percent of the trip data loss can likely be attributed to poor GPS signal during trips. Although the log messages associated with GPS availability cannot be extrapolated to measure the number of trips or miles impacted, the loss of trip data resulting from vehicle detection failures or lack of GPS signal during trips clearly identifies GPS availability as a significant system issue. The deployment team’s report provides additional insight into the accuracy of the system as it relates to GPS connectivity and accuracy. Intermittent GPS signal was reported as a contributing factor to lower device miles compared to odometer miles collected.

Just as a random statistic, which probably doesn’t mean a lot, the report includes the word “success*” 27 times and “fail*” 33 times.
I believe MBUF or an equivalent will come eventually, but here and now the gas tax (or wholesale fuel tax, if we want to hide it), properly indexed to inflation and fuel economy, is where we need to be focusing for the revenue required to operate road systems in the US.

Property Tax on Privatized Roads

Recently published:

  • Junge, Jason and David Levinson (2013)
    Property Tax on Privatized Roads.
    Research in Transportation Business and Management. [doi]
    Roads cover a significant fraction of the land area in many municipalities. The public provision of roads means this land is exempt from the local property tax. Transferring roads from public to private ownership would not only remove maintenance costs from city budgets, but increase potential property tax revenue as well. This paper calculates the value of the land occupied by roads in sample cities and determines the potential revenue increase if they were subject to property tax. Further calculation computes the extent to which the property tax rate could be reduced if the land value of roads were added to the tax base.

    JEL code: R40, R11, R14

    Keywords: tax, land value, locational analysis, transportation finance

The case for (and against) public subsidy for public transport

Crossposted from David King and I argue The case for (and against) public subsidy for public transport 


“In most of the United States and much of the world, public transport is publicly subsidized. Everyone in an area pays for transit whether or not they use it. This was not always the case, and need not everywhere be the case. Once mass transportation was provided to the public for profit (in Minneapolis and St. Paul as well as most other US cities) from the late 1800s through the first half of the 1900s. While rights-of-way were often publicly provided, the companies operating transit paid for the maintenance of those rights-of-way above and beyond what was required for transit.”


The case for (and against) public subsidy for public transport

This post is co-authored with David King (a displaced Minneapolitan who lives in New York, and who blogs at Getting from here to there)

In most of the United States and much of the world, public transport is publicly subsidized. Everyone in an area pays for transit whether or not they use it. This was not always the case, and need not everywhere be the case. Once mass transportation was provided to the public for profit (in Minneapolis and St. Paul as well as most other US cities) from the late 1800s through the first half of the 1900s. While rights-of-way were often publicly provided, the companies operating transit paid for the maintenance of those rights-of-way above and beyond what was required for transit.

Subsidy should be considered two ways: capital subsidy and operating subsidy. These are related, but different enough that they should be considered separately.

Capital subsidy can be direct or indirect (such as assistance with land acquisition), and these monies come from federal, state, metropolitan, local and sub-local sources. Traditionally capital subsidy has largely come from federal and state sources, though recently local sources through sponsorship (see the Emirates Airways gondola in London, for instance) or value capture have been used. Capital subsidy for transit expansion rarely, if ever, considers the effects capacity and network expansion have on operating subsidy, however. Since every transit system in the United States requires an operating subsidy, every service expansion increased the required operating subsidy and makes the financial position of transit agencies worse over the medium and long term.

Operating subsidies are from local, regional and state sources. The federal government placed severe limits on using federal money for operations in the 1970s, in part because most of the increases in subsidy went to total wages without any increase in productivity. The primary reason for operating subsidy for US systems now seems to be “that’s the way we do it here,” which is not a proper justification. Many of the cities around the world—and in North America if we look to Canada, where the Toronto system is required to maintain 75% farebox recovery in order to receive provincial subsidy for the remaining costs—have much higher farebox recovery, fewer operating subsidies and much higher ridership, which suggests a justification for less subsidy and higher fares: planning without prices leads to bad planning.


Yes No
Rivalry Yes Private “Congesting”
No Club Public

Economics defines 4 types of goods: Public, Private, Congesting (or Common Pool Resource), and Club. Public goods are, by this definition, neither excludable (to use it, you must pay for it) nor rivalrous (the good is scarce and only one person can use it at a time). Yet public transport is both excludable (at the cost of validating payment), and rivalrous (when congested). Under those conditions it satisfies the definition of a private good. Many private goods are privately provided, hence the name.

However sometimes transit is operated non-excludably, for instance the Campus Connectorat the University of Minnesota, or an elevator in your nearest multi-story building. Similarly, sometimes transit operates with an honor system payment with lax (or no) enforcement.

Often transit is not rivalrous, (non-rivalry implies my consumption does not affect yours, by increasing its cost or diminishing its quality) e.g. in off-peak times. The off-peak Campus Connector is a public good. It is paid for as a Club Good by the University of Minnesota and its students, since they are the primary beneficiaries (almost no one outside the U community would bother riding, so even it if is technically non-excludable, it is functionally excludable in that no one who isn’t going between campuses would bother riding, and the people going between campuses have something to do with the University).

The case for subsidy for some public goods is obvious. In the absence of excludability and rivalry, one needs to get revenue from somewhere to operate a service that provides public benefits. The classic example is national defense. I can’t just “not subscribe” to national defense, it protects me whether I want it or not. We can of course debate the amount of public good we want.



  1. Transit often operates on the left-hand side of theU-shaped cost curve. Fixed costs are spread over more and more users as the quantity demanded increases, while marginal costs remain small if not zero. If we charge riders e.g. an average cost for a service with near zero marginal cost (which is an approximation of the situation in transit in the absence of crowding, certainly in terms of the short run marginal cost, ignoring a few things like the delays which boarding imposes on other passengers), we get under-consumption and under-supply compared to the social optimum. That means if we charge more than the marginal cost of the ride, we get a less than socially optimal number of passengers (there is a deadweight loss). Somebody who would ride at a lower price that was still at least as high as their marginal cost cannot. The social benefit (consumer’s surplus) of that unmade trip is foregone. Unfortunately because of high fixed costs, this implies that fares at marginal cost will not recover total costs. Thus the natural monopoly / economies of scale or density / declining fixed cost is one aspect that might warrant subsidy.
  2. There are network externalities associated with public transit. The more users of transit there are on a system, the more useful the system is for everyone. This is sometimes called the Mohring Effect, but the basic idea is that if 50 people want a ride each hour, you send one bus. If 100 people want a ride each hour, you send two buses, each a half-hour apart, and the average rider only has to wait half as long, (reducing wait times, and over a network, reducing transfer times) benefitting everyone. Similarly, the more riders, the more spatial coverage that can be provided (reducing access and egress times).
  3. Transit helps the transportation disadvantaged. Equity or welfare has often been an argument in favor of subsidy, that we do it to provide benefits for people unable to afford otherwise, or transportation for the disadvantaged. This gets more into values than economics, but there are some people who would be employed but for their ability to access jobs, so some subsidy on the transportation front is at least partially repaid by more economic productivity.
  4. Transit subsidy helps poor jurisdictions. This has also been argued at the macroscopic level, e.g. Yonah Freemark justifies the federal transit program by arguing in favor of spatial cross-subsidies, i.e. benefitting poor jurisdictions rather than poor people.
  5. Transit reduces congestion on other modes, by taking cars off the road, and therefore benefits drivers (who should thus pay for it).
  6. Transit stimulates economic development.
  7. Cars are subsidized, therefore transit should be subsidized.

These aspects argue in favor of subsidy. But then the question arises, subsidy from whom? That is, what is the appropriate base for providing subsidy? Here we argue in favor of a Club Goods model. People in the Club should help subsidize the service.

The beneficiaries of transit are relatively local. If I live in Minneapolis, the option of riding transit in Las Vegas or Curitiba is of essentially zero value to me. The option of riding transit within the greater Twin Cities region is of some value, and the option of riding transit in Minneapolis is of high value. The option of riding transit that runs past my house to my desired destination is the highest value. Benefits diminish with distance from the system.

We can define the Club more narrowly as anyone who might want to use transit and is willing to pay (or whose employer or University  is willing to pay or help pay) for a season pass. An advantage of using a season pass model (rather than pay per trip) is the ability that it presents in providing services without excessively under-pricing the transit service. Whoever wants to provide transportation benefits for the transportation disadvantaged can subsidize those whom they want without subsidizing everyone.

We can define the Club a bit more broadly as landowners whose property value is increased by the presence of transit. The option of riding transit sometimes is public good (i.e. the option is neither rivalrous nor excludable), and its value is embedded in locations near transit stations. This appears to justify some form of value capture approach (of which property tax is the most widely used, but certainly not the most direct or efficient mechanism).

Both of these clubs are smaller than the municipalities in which transit operates, and much smaller than higher levels of government, like county, state, or nation.

Though there are clearly some arguments in favor, this post promised arguments against transit subsidy as well.

  1. Transit is basically a private good. Private goods can be privately provided, which aligns incentives of the producer with their revenue model directly, better performance is rewarded, worse performance is punished. When all transit lines—and road networks for that matter—are planned and operated below cost we simply don’t have any idea what the true value of any service is. If fares increase to cover costs or at least come closer to covering costs service can adapt to revealed demand and firms and households can adapt accordingly. Without proper prices we don’t know where to increase capacity or improve service. We can’t identify actual bottlenecks or spread peak demand across more hours in the day by using dynamic pricing. By planning service while blind to the value of the service everybody is a bit worse off and many actual transit riders are substantially worse off.
  2. The network effects might be relatively small (either because they are already played out (high frequency service in a high density city), or because they never will be (low frequency service in a low density city)). The best opportunity is thus low frequency service in a high density city, in which service can be increased. Downscaling may need to occur in places with high frequency and low density. Many technologies have network effects but don’t require public subsidy. From Facebook to your ATM networks the amount of public subsidy is zero, or small. What is usually required is a monopoly (AT&T of yore, airline hubs), some type of lock-in (social networks), collusion (credit cards), or cooperation (the English Language, which readers of this blog all use without the government subsidizing [we can argue about schools separately, if you grew up in the US, you probably learned to speak if not read and write English before you went to school]).
  3. Though there are always returns to density (more riders on the bus always lowers the cost per passenger on the bus), bus systems have approximately constant returns to scale (more buses are not less expensive per bus than fewer buses). Many places have figured out an economic model which does work from a profitability stance. There is little reason economically to run a service with many buses under the auspices of a single monopolistic organization. Constant returns industries don’t warrant the same subsidy as increasing returns industries.
  4. While it is claimed transit should be subsidized to stimulate economic development, it is also claimed that economic development should be subsidized because it stimulates transit. It is not clear which, if either, or both, are driving this cart.
  5. Fixed route transit may be made obsolete by new technologies, especially outside of cities.
  6. One bad subsidy does not deserve another. Just because cars are subsidized is not a reason to subsidize transit. It is an argument to remove the subsidies that exist. Technically (if not politically) it would be relatively easy to charge cars for their full cost (i.e. eliminate their subsidy) via higher fuel taxes (or mileage fees) but the amount of the incremental charge would have a very small effect on total automobile use.
  7. As for the Mohring effect, how much should a transit rider pay for the bus or train not taken (subsidy for options)? Consider a downtown worker who prefers to take transit to work, but sometimes has to work late into the evening. There is lots of service for typical 9-5 employment, but a major reason workers are comfortable on transit is that they know there is adequate service for occasional trips such as when they work late. Let’s say that a optimal fare that covers direct costs for a rush hour bus is $2, but every other week a worker has to stay until 9pm when service is infrequent. Should the regular fare be $2.20 to reflect the required subsidy for the not-full 9pm bus? Or should occasional trips be shifted to taxis or made the employers’ problem? In Manhattan most corporate employers will pay for a taxicab home for employees who work past 9pm, and a recent survey in San Francisco shows that employees are more likely to ride transit to work if they know that they can take a cab for their return trip. As this is off-peak time for taxicabs perhaps this is a more optimal solution than subsidizing increased fixed-route services.
  8. Finally there are lots of reasons not to trust the recent experience in transportation with investment. The costs are too high and the benefits are too low. Giving more funds to existing institutions to build more capital-intensive infrastructure while existing infrastructure deteriorates may not produce the hoped for results.

So how does this net out? We believe that in most places at most times, transit operates like a club, it is excludable but not rivalrous. Therefore we should do what we can to encourage utilization, and play on the idea that people dislike paying out-of-pocket, which discourages use, but are happy to pay for “unlimited services” so they don’t have to think about everylittle transaction cost.

  1. Systems should over time pay for their own operation and maintenance from usage-derived revenue. Anything less is not likely to be financially sustainable with a shift in the political winds. Like other public utilities, transit can and should be able to cover its operating costs from user revenue. If it can’t, the users perceive insufficient benefit. Yes, we are all for congestion charges and conversion of other auto costs to a better basis (e.g. pay-as-you-drive insurance), and that will help transit in selected areas at selected times, but the problem of transit primarily is in the service it provides to transit users, and it needs to operate in the real environment of today, not an idealized transportation financing system of a decade or two from now. The number of articles with the phrase “transit funding crisis” in quotes from Google is over 40,000, and perpetually stressing the system has adverse health effects. Unlike most businesses, when transit’s ridership increased, it was in even direr straits to believe the advocates. More money may be an answer, but money does not buy happiness, transit agencies would be better off if they could reconsider their scope.
  2. Users should be financially incentivized to get season or annual passes (paid monthly with bank debits) and become “members” of the transit system rather than pay-as-you-go “riders”, which will encourage more usage, and many users to get subscriptions so they have the easy option of taking transit. As with many museums and zoos and other clubs, membership should be reciprocal, so joining the Twin Cities Transit System gets me “free rides” in Chicago or New York. This will increase the perceived ownership that passengers have for the service.
  3. Land value capture should pay for capital costs of infrastructure. But we should only build new infrastructure which has a financial model for recouping operating and other ongoing costs. If the infrastructure generates benefits that accrues to landowners, it is both fair and efficient to use some of those benefits to pay for the infrastructure in the first place. For value capture (not just land value capture) as a tool for subsidy, it needs to be recurring rather than a one-time fee. The largest beneficiaries of transit services in the Twin Cities are downtown employers, the airport and the Mall of America. Downtown Minneapolis employers and employees receive a very large benefit because of reduced parking costs. If true, then a tax on downtown/airport/MOA wages (perhaps one-quarter of one percent) makes sense to subsidize additional service and frequency because of the Mohring effect, and these employers act essentially as a club. We still need a financial model for recapitalizing the system after the land is developed. Value capture is still appropriate, but it requires different mechanisms.
  4. The public should subsidize transportation for the disadvantaged from non-transportation specific revenue sources. Perhaps the biggest problem with current subsidies is that they are place-based and not people- based. Why should the entire system be subsidized? Also, why should a professor pay the same fare as students? Or in New York, why should Mayor Bloomberg, the richest guy in the city, pay the same fare as the cleaning staff of Bloomberg, Inc.?



A Portfolio Theory of Route Choice


Recently published:

Although many individual route choice models have been proposed to incorporate travel time variability as a decision factor, they are typically still deterministic in the sense that the optimal strategy requires choosing one particular route that maximizes utility. In contrast, this study introduces an individual route choice model where choosing a portfolio of routes instead of a single route is the best strategy for a rational traveler who cares about both journey time and lateness when facing stochastic network conditions. The proposed model is compared with UE and SUE models and the difference in both behavioral foundation and model characteristics is highlighted. A numerical example is introduced to demonstrate how such model can be used in traffic assignment problem. The model is then tested with GPS data collected in metropolitan Minneapolis–St. Paul, Minnesota. Our data suggest there is no single dominant route (defined here as a route with the shortest travel time for a 15 day period) in 18% of cases when links travel times are correlated. This paper demonstrates that choosing a portfolio of routes could be the rational choice of a traveler who wants to optimize route decisions under variability.

JEL-Code: R41, R48, D63
Keywords: Transportation planning, route choice, travel behavior, link performance