Surface Transportation Innovations Newsletter

Surface Transportation Innovations #23

Topics include: paying for new lane construction out of toll revenue; the continued decline of car-pooling; tolling existing free lanes; incident-related congestion; and inaccuracies in rail and road traffic forecasting.

In this issue:

How Much Can Tolls Pay For?

One of the great question marks about proposed new market-priced congestion-relief lanes is the extent to which new-lane construction can be paid for out of toll revenues. (Or the flip side: to what extent will such projects produce “surplus revenues” that could be used for other purposes such as subsidizing mass transit?)

The two pioneering California HOT lane projects represent the two extremes of this question. By converting already existing HOV lanes on I-15 to a form of HOT lane, SANDAG was able to generate significant surplus revenues which it is using to subsidize express bus service in the corridor. By contrast, the 91 Express Lanes in Orange County represented over $100 million in all-new construction, and while toll revenues are fully supporting the capital and operating costs of the HOT lanes, there is nothing left over for other purposes.

Recent studies of building networks of HOT lanes-in Atlanta, Denver, and Minneapolis/St. Paul-have reached dramatically different conclusions about the extent to which capital costs can be covered by toll revenue financing. I reviewed these three studies in my May 2005 column in Public Works Financing, noting that each used a somewhat different methodology and made somewhat different assumptions about pricing. The Atlanta study used much-too-low figures for HOT lane pricing (and did not adjust them for inflation as at least a proxy for the need to keep increasing them over time to keep traffic flowing smoothly). So not surprisingly, it concluded that toll revenues could at most cover operating and maintenance costs, but not capital costs. The Twin Cities study, with more realistic pricing assumptions, estimated that toll revenues could cover just 22% of capital costs. But the Denver study estimated that tolling could support 50-60% of capital costs. To be sure, all three studies were of basic networks of tolled lanes, so it may be unfair to compare them with studies of individual corridors. Certainly it would make sense for investors to look first at the most promising corridors, and both the Denver and Twin Cities studies found some corridors that looked like they could be self-supporting.

I mention this to put in context the most recent news on the private-sector proposal to add HOT lanes to the southwestern quadrant of the Beltway around Washington, DC Last month’s announcement by VDOT and the Fluor/Transurban team was that their revised financing model will permit this $900 million project to be done without any tax money. What has changed from the previous model (which would have required 15-20% of the cost to be paid for by VDOT) is the shift from an all-debt/30-year financing model to a debt-plus-equity model, financed over as much as 60 years. Australian developer/operator Transurban would provide at least 15% of the project cost as an equity investment, and would be willing to make much of its return on investment in the out-years, after the debt is paid off. So it appears that the choice of financing method can make a significant difference in terms of how much of the cost can be recovered from toll revenues. That means enabling legislation should welcome long-term (50+ year) franchise arrangements such as that now contemplated by VDOT.

The other key factor in how much toll revenues can finance is getting the pricing right. My colleague Peter Samuel ( in March did a quick analysis of the proposed financing of the Inter County Connector (ICC), a new congestion-relief highway in suburban Maryland . Given the level of congestion in the area, the ICC would be essentially an all-HOT-lanes facility. Samuel notes that the Draft Environmental Impact Statement (DEIS) from last December estimated that toll revenues could finance only a small fraction of the ICC—s $2.2 billion cost. That study assumed 17 cents/mi. peak tolls and 13 cents/mi. off-peak—reasonable for a new (conventional) toll road these days. Drawing on income-level data for the area, and estimates that the ICC will save about 30 minutes over its full 18 miles during peak periods, Samuel suggests that peak toll rates should be more like 55 cents/mile, more than three times what the DEIS used. That suggests the annual toll revenues could be something like three times the $60 million cited in the DEIS. And that, in turn, suggests that close to $2 billion in capital costs could be financed—nearly the entire capital cost of the ICC.

These are early days in the development of serious urban congestion-relief toll lanes. Transportation planners clearly have a lot to learn about how these projects differ, fundamentally, from conventional toll roads.

» return to top

The Continued Decline of Car-Pooling

For the past 25 years or so, America’s principal strategies for dealing with urban traffic congestion have been expanded mass transit and car-pool (HOV) lanes. The 2000 census provided the grim news that the fraction of commuters using either of these modes had reached new lows, despite all the billions spent developing them. But the latest news on car-pooling is even more discouraging.

The 2003 data from the Census Bureau’s American Community Survey were released recently. They show that car-pooling has continued to decline, reaching a new low of 10.4% in 2003, down from 11.2% in 2000. Even more disturbing is new data about who is actually doing the car-pooling. In Commuting in America II, Alan Pisarski raised the issue of “fam-pooling”—i.e, family members who would be traveling together anyway simply taking advantage of the HOV lanes. If the purpose that justifies spending billions of dollars on HOV lanes is to change travel behavior by reducing the number of cars on the road at rush-hour, giving HOV space to a mom taking junior to school simply doesn’t cut it. Data on the degree of fam-pooling have been rather sketchy, but estimates from surveys done in several metro areas have ranged from one-third to as much as two-thirds.

But at a recent TRB conference on Census Data for Transportation Planning, Nancy McGuckin and Nandu Srinivasan presented a paper called “The Journey-to-Work in the Context of Daily Travel.” Using data from the National Household Travel Survey (NHTS) and the National Passenger Transportation Survey (NPTS), they found that in 2001 some 83% of carpools could be classified as fam-pools (Table 12).

This is a devastating finding. It means that four out of five of the vehicles to whom we are giving away very costly-to-build space in HOV lanes are not cars that have eliminated another car from the road at rush hour. They are cars whose drivers are able to opportunistically get a faster trip due to having a family member in the car.

Now obviously, there are exceptions. Some HOV corridors, as in Houston, the Washington, DC suburbs, and Los Angeles, are heavily used by buses, fulfilling the original intent of getting more people to use fewer vehicles to get to work. But it now appears that most of America’s HOV-2 lanes serve mostly fam-pools. It—s high time we started re-thinking this policy, figuring out how to put these lanes to higher and better use. A forthcoming Reason policy study will address this issue.

» return to top

Tolling Existing Free Lanes?

One of the biggest controversies in the endless saga of the federal reauthorization process has been whether or not the remaining portions of the ban on putting tolls on the existing Interstate system would remain or be repealed. Last year’s Kennedy Amendment in the House, which passed with trucking industry support, was driven by concern over this issue. The Administration’s SAFE-TEA bill, which the Senate largely adopted last year, would have removed all federal restrictions on tolling existing lanes for congestion management, leaving this question up to the states.

This year the picture is different. The House defeated a much milder Kennedy Amendment (which would have continued the Value Pricing program and the existing pilot program to rebuild up to three Interstate facilities with tolls). As passed, the House bill would permit up to 25 congestion pricing projects, three Interstate rebuilds with tolls, and three new Interstates funded with tolls. But the Senate bill adopted this week would limit value pricing to new lanes and converted HOV lanes, while eliminating the existing Interstate rebuild program (except for Virginia, presumably to let the I-81 continue). The pricing/tolling provisions will now have to be worked out in the conference committee.

The issue of tolling existing lanes has always been troubling for me (and I’ve been quoted not quite accurately by advocates on both sides of the issue). On one hand, I think value pricing is the most powerful tool we have for managing traffic flow and relieving congestion. And, as a libertarian, I’m opposed to the federal government telling road owners how they should manage and pay for their roadways. Plus, over the long term, I’m convinced we need to shift from a highway funding system based on fuel taxes to one based on direct payment for the highway services people use. But—and it’s a big but-getting from here to there (where “there” means a more customer-friendly highway system) should not be done by forcing new approaches down the customers’ throats. Change is more likely to come about, and to have long-term support, if we can craft win-win opportunities. That’s why Reason’s work has stressed measures such as HOT lanes and Toll Truckways, that give highway users a choice and, in particular, give them significantly better service in exchange for paying a price.

To the best of my knowledge, Texas is the only state whose tolling and public-private partnership law (HB 3588) permits the conversion of existing free lanes to tolled lanes. The first attempt to do that, in Austin, created a huge political firestorm. In response, the original sponsor of HB 3588, Rep. Mike Krusee, has proposed state legislation to require that when a regional mobility authority proposes such a conversion, it would have to be approved by a vote of the county commission. A state senate bill by Sen. Todd Staples, would require a local referendum in addition to county commission approval.

My assessment is that if Congress adopts something like the House provisions, we will see very few conversions of free highways to toll roads. But we will see much-needed lane additions being developed as premium-priced lanes offering better services in exchange for payment. That’s also what we will get if the Senate provisions prevail.

» return to top

How Much of Congestion Is Due to Incidents?

The latest annual Urban Mobility Report has just been released by our colleagues at the Texas Transportation Institute ( As usual, it makes pretty grim reading. Nearly all the key numbers reached new highs, including the total hours of delay (3.7 billion) and the estimated annual cost of congestion ($63.1 billion).

Also noteworthy is the latest update on that handful of metro areas that still seek to “build their way out” of congestion, by adding lane-miles of capacity to keep pace as much as possible with the growth in vehicle miles traveled (VMT). Only four urban areas show a narrow gap between traffic growth and capacity (keeping capacity growth within 10% of demand growth): Anchorage, New Orleans, Pittsburgh, and Tulsa . The report also includes a chapter that crunches the numbers to assess how much additional roadway capacity would be needed in each of the other metro areas to match the growth of travel.

But that calculation is somewhat misleading when put in perspective, based on what we are learning about the causes of traffic congestion. A growing number of studies find that incident-related congestion accounts for half or more of the total hours wasted, with only the balance considered “recurring” congestion, due to demand exceeding capacity. The largest source of incident congestion is crashes, followed by construction work zones, vehicle breakdowns, and weather effects (rain, fog, snow, ice).

The Federal Highway Administration has been conducting a several-phase study called The Temporary Loss of Capacity Study. The Phase 2 report, which goes into considerable detail, is S. M. Chin, et al., “Temporary Losses of Highway Capacity and Impacts on Performance, Phase 2” (Oak Ridge National Laboratory, ORNL/TM-2004/209, November 2004). It’s available online at:

A growing number of large metro areas now operate incident-response programs, such as SAFEClear in Houston, Freeway Service Patrol in Los Angeles, and the Incident Response Program in Seattle. TTI estimates that these kinds of programs have benefits that are from three to ten times greater than their costs. While I have not reviewed any benefit-cost studies of such programs, that finding seems quite plausible.

It’s important that we realize that pricing and capacity additions are not the sole answer to congestion. Given the huge role of incidents in the overall picture, programs targeted to clearing incidents quickly are an essential tool in the tool-box.

» return to top

Over-Hyped Projections Are a Big Problem

A new paper by colleague Bent Flyvbjerg, lead author of Megaprojects and Risk, is shaking up the world of transportation planners. It’s called “How (In)accurate Are Demand Forecasts in Public Works Projects?” It—s noteworthy not only for its content but because it was published in the Spring 2005 issue of the respected Journal of the American Planning Association. You can download it from their website:

What Flyvbjerg and two colleagues have done is to look at inaccuracies in demand forecasts for road and rail projects, using a database of 210 projects in 14 countries, with an aggregate value of $59 billion. The most notable finding is that rail passenger forecasts are over-estimated by an average of 106%. By contrast, about half of road project traffic forecasts are off by more than 20%–but about as many are under-estimated as over-estimated. The authors note that it would be interesting to compare toll roads with non-toll roads, but their data did not permit this. Another finding is that the accuracy of forecasts has not improved over time; their database goes back to the 1960s.

What accounts for the huge inaccuracy of rail traffic forecasts? The authors hypothesize that because of the strong competition for funds for rail projects, there is “an incentive for rail promoters to present their projects in as favorable a light as possible-that is, with overestimated benefits and underestimated costs.” They also speculate that where there is a political or ideological desire to shift people from road to rail, this may create pressures for over-optimistic rail forecasts. They then investigated the cause of inaccuracy for each of the projects in their database, both by asking project managers and by seeing what researchers concluded. The second largest stated cause for inflated rail forecasts was “deliberately slanted forecast.” They conclude that “Rail forecasts are systematically and significantly overestimated to a degree that indicates intent and not error on the part of rail forecasters and promoters.”

As to what can be done about this, for public-sector projects they recommend the use of “reference cases” as part of forecasting, so that a new project’s forecasts are put in the context of actual results on other projects of the same type. And for projects done via public-private partnerships, they recommend that the decision to proceed with a project should be made contingent on the willingness of private capital to provide at least one third of the total cost without a sovereign guarantee (i.e., true capital at risk). And they charge the planning profession with treating the production of deliberately deceptive forecasts as “a violation of their code of ethics-that is, malpractice.”

» return to top