3.0 Synthesis of the State of the Practice
This chapter summarizes the state of the practice for evaluating the travel and environmental impacts of congestion pricing projects. The findings presented here are drawn primarily from a review of published literature associated with the eight study projects described in Chapter 2.0. This chapter is divided into three sections. The first highlights some of the fundamental differences and similarities in the evaluation approaches among the eight study projects. The second section focuses on travel impacts. The final section focuses on environmental impacts.
All eight of the study projects include before-after (or “with/without”) evaluations of deployed congestion pricing schemes: either short-term demonstrations with simulated congestion charges or long-term deployments with real charges. The Oregon, Puget Sound and Commute Atlanta mileage-based value pricing investigation are examples of the former (short-term, simulated) and the other five projects are of the latter type (long-term, non-simulated).
Two of the projects—London and Singapore—have been evaluated in an on-going manner over many years, encompassing a number of expansions and other changes in the congestion pricing programs. Those projects, to varying extents, have produced a series of results reports which include comparisons and which draw conclusions based on many years of post-deployment data. The six other projects generally focus on shorter, finite, post-deployment evaluation periods of between a few months and, in the case of I-15, up to three years. Although the Minnesota I-394 HOT lanes project is subject to on-going monitoring, the comprehensive evaluation effort was, essentially, a one-time effort.
London demonstrates the most extensive evaluation effort among the study projects in terms of evaluation period (over five years) and the breadth of impacts, performance measures and data sources. The next most extensively studied projects are Stockholm and Singapore.
The simulated pricing field deployment projects—Oregon, Puget Sound and Commute Atlanta—focused more narrowly on individual traveler behavior than the other projects, that is, they did not consider the traffic or environmental ramifications of changes in travel behavior. Further, to varying degrees (Commute Atlanta less so), those three projects focused more narrowly on driving behavior and associated measures and less on other modes than did the other study projects.
This section contains two parts. The first summarizes travel impacts, performance measures and data collection methods and the second part summarizes travel findings and study limitations.
Table 3‑1 compiles the travel impacts, performance measures and associated data collection methods for each of the eight study projects. Among those projects, three broad types of travel impacts were found most commonly:
The travel impact most commonly considered in the eight study projects is traffic. Transit and traveler behavior impacts are less common, but far from unusual. Few study projects examined safety impacts.
Tables 3-2 through 3-4 summarize the state-of-the-practice for travel impacts in terms of the relative prevalence of various performance measures and associated data collection methods. This information is based primarily on the eight study projects. Greater variety was found in the area of traffic impacts than in transit and traveler behavior and therefore Table 3‑2 uses three categories to assess prevalence and Tables 3-3 and 3-4 use only two. Assessments of the relative rarity of various measures are subjective and there are a few close calls. For example, categorizing travel times and speeds as “very common” versus “common.” Assessments for the performance measures are absolute in the sense that they reflect how commonly each measure appears in the evaluation reports that were reviewed. Assessments of methods are relative in the sense that they do not describe how often a particular data collection method appears in the literature overall, but rather, how frequently that method is used among those projects that include the associated performance measure. For example, a “very common” data collection method for an “uncommon” measure means that, in the literature overall, that data collection method is not common.
1 Rare as an end measure, commonly collected to support environmental analysis (an input to emission rates)
2 This measure appears both as a traffic impact measure as well as a traveler behavior impact measure (Table 3-4) to reflect the two different approaches to collecting these data (counts for traffic vs. surveys for travel behavior) and the different focus in traffic analysis (impact, e.g., person throughput, on roadways) versus travel behavior (impact on people).
As indicated in Tables 3-2 through 3-4, and as one would expect with evaluations of deployed projects, all of the traffic and transit measures utilize objective data rather than modeling, simulation or other estimated or derived sources. In contrast, travel behavior measures always rely on travelers’ self reports of their behavior, although in the case of travel diaries—where travelers are asked to very accurately and specifically record the details of individual trips using paper or electronic forms—reported behavior is expected to be a fairly accurate reflection of actual behavior.
Overall, explicit consideration of mode choice changes and the change in mode splits or shares—the percentage of person trips made by various modes—is not extremely rare, but is far from standard. Although some evaluations do not consider mode changes at all, most of them do, with the smaller, less comprehensive studies often inferring some mode change effects based on changes in traffic volumes and transit ridership. The most comprehensive evaluations, such as London, do explicitly consider mode changes and attempt a complete accounting, e.g., of the X number of car trips that were eliminated, Y percent went to transit, etc. However, even among the more robust evaluations, comprehensive treatments which capture both the total change in person trips and trace all of the mode-to-mode changes are very uncommon. In London, the disposition of the car trips that were eliminated was examined, but a complete accounting of total person trips before and after by mode was not performed.
Within the area of traveler behavior surveys, variations can be observed based on the following major variables:
The state-of-the-practice in regard to these parameters can be summarized from the eight study projects as follows:
Considering the full range of travel impacts (traffic, transit, traveler behavior, safety) the state of the practice in regard to instrumented vehicles, including traffic probes, can be summarized as follows:
Appendix A includes a table summarizing the main travel related findings of the eight study projects. Results indicate that a variety of congestion pricing projects have been shown to be effective in reducing traffic congestion. Highlights of reported findings include the following:
In reviewing the findings of congestion pricing evaluations, the emphasis in this study was to understand how the evaluation methodologies and challenges impacted the ability of researchers to draw definitive conclusions. The table in Appendix A includes a column that summarizes major caveats or limitations relevant to specific findings reported for the eight study projects. These caveats and limitations can be summarized as follows:
This section contains two parts. The first part summarizes environmental impacts, performance measures and data collection methods and the second part summarizes environmental findings and study limitations.
A 2008 FHWA lessons-learned report on their Value Pricing Pilot Program noted that congestion pricing evaluations have paid less attention to equity and environmental impacts than traffic impacts, project operations, and public and customer satisfaction.51 That is also generally the case among the eight study projects.
Table 3‑5 identifies the environmental impacts, performance measures and associated data collection methods for each of the eight study projects. Among those projects, three broad types of environmental impacts were common:
Pre-deployment estimation of the environmental impacts of congestion pricing projects—such as environmental assessments (EAs) and environmental impact statements (EISs) developed in the U.S. in compliance with the National Environmental Policy Act of 1969 (NEPA)—do consider a much wider range of potential impacts, including land use. However, the present study focuses on measurement of the actual, post-deployment impacts of congestion pricing and, with one exception, there were no examples of those other types of environmental impacts among the study projects. The one exception—the one other environmental impact that was found in the study projects (London and Stockholm)—is the impact of congestion pricing on business and the economy. However, in neither London nor Stockholm were these impacts included within the “environmental” impact category. This state of the practice summary focuses on the more common impacts: air quality, noise and environmental justice.
Some of the eight congestion pricing study projects—the simulated pricing field demonstrations in Oregon, the Puget Sound region, and Atlanta—did not consider any environmental impacts whatsoever. However, with the exception of the seeming omission of noise and the addition of pedestrians perceptions of safety in Singapore, all of the other study projects considered air quality, noise and—although under various names and varying emphases—environmental justice. In several cases, “environmental justice” impacts were termed “equity” impacts or were part of broader investigations such as “social impacts.”
In regard to performance measures and data collection methods, air quality and noise environmental impact analyses show less variation from study to study and fewer measures and methods overall than do travel impact analyses. The most common air quality performance measure—used in each air quality analysis that was reviewed—is the volume of pollutant emissions from roadway traffic. The only other performance measure that was found, and which is much less commonly considered, is ambient pollutant concentrations. Only one data collection method was found for vehicular pollutant emissions: calculating emissions by multiplying project-attributable changes in vehicle miles traveled on roadway links by model-derived emission factors (e.g., grams emitted per mile) corresponding to the observed speeds on the roadway links. Likewise, there was only one data collection method found for ambient pollutant concentrations: roadside pollution sensors. None of the study projects included in-depth considerations of greenhouse gases. Rather, when greenhouse gases are addressed at all, it is typically done by including the precursor CO2 among the calculated vehicle emissions. No examples of the sort of atmospheric modeling necessary to actually estimate green house gases were found.
There is even less variation in evaluation practices related to noise impacts. Noise impacts in all of the study projects were measured in terms of ambient sound levels, in decibels, and in all cases roadside noise monitors were used to collect the data.
Pedestrian safety was only considered as an environmental impact and evaluated in Singapore. Pedestrian safety was measured in terms of pedestrians’ perceptions which were gathered through surveys, interviews or focus groups.
In contrast to air quality and noise, analyses of environmental justice or equity impacts vary more from project to project. This may be because environmental justice analyses typically include more qualitative or subjective elements and, therefore, or more of an art than a science, or it may be that as a relatively new area of interest, fewer standardized approaches have thus far emerged. Although specific approaches vary, environmental justice analyses generally focus on two main areas. The first is to examine how travel impacts distribute geographically and, using geographic socioeconomic databases, infer whether projects differentially impact (both benefits and disbenefits) areas with concentrations of low income and/or minority populations. The second element employs public surveys and or focus groups to gauge the perceptions of low income and/or minority populations.
Appendix B includes a table summarizing the main environmental impact findings of the eight study projects. Highlights of reported findings include the following:
Caveats and limitations associated with environmental impacts focus largely on the inability to differentiate project contributions from exogenous factor influences on monitored air quality impacts. Although some state-of-the-practice reviews have identified the failure to test for air quality implications of possible changes in traffic flow (e.g., more cruising and less stop and go) as a weakness, this issue is rarely noted by the evaluators of specific projects. There are limited discussions of caveats and study limitations in regard to environmental justice. When discussed, limitations sometimes focus on measurement issues such as survey sample limitations and the inherently subjective and “less scientific” nature of perception data.
This section identifies gaps in the understanding of the travel and environmental impacts of congestion pricing.
Table 3‑6 summarizes the aspects of travel impacts that have generally been well established through before-after evaluations as well as those areas where understanding is less complete. Each of these gaps is discussed following Table 3‑6.
Long Term Impacts
Long-term impacts are not well understood because there have been relatively few long-term evaluations. Although the Singapore and London evaluations have continued over many years, the ability to draw conclusions is still limited by the fact that not all analyses have been continued and because over these longer periods the masking influence of exogenous factors has made isolation of project impacts especially challenging. In particular, there is not a good understanding of the long-term traveler behavior associated with congestion pricing, e.g., will travelers who, in the short term, are willing to continue driving and pay charges switch to other modes or make other changes, such as moving, switching jobs or telecommuting, over the long term? Conversely, will travelers who initially shift to transit or make other changes to avoid the charge eventually drift back to driving—and if so, how many of them and why? Even when long-term traffic and transit ridership data have been studied, only the aggregate change is usually considered and the components of the change are unclear. For example it is unclear whether static mode shares over time mean that travel behavior changes in response to pricing have ended or whether changes continue but off-set one another, e.g., drivers continue to switch to transit but as transit becomes crowded other travelers switch to driving. This area of uncertainty includes a lack of information on long-term impacts on auto ownership.
Another gap concerns the regional impacts of congestion pricing. Although some evaluations have considered travel impacts over a fairly large area (e.g., London and Stockholm), many evaluations have focused on the priced facilities/zone and the immediately adjacent portions of the transportation system. Of course, it is resource intensive to study larger areas, and most evaluations have focused their attention on the areas where the most significant impacts are expected, which is a logical strategy given constrained resources. Unfortunately, this has created something of a self-perpetuating cycle: evaluators do not examine regional impacts in part because there is not strong evidence that there will be regional impacts but the reason there is no strong evidence is not because evaluations have found few impacts but simply because those impacts are very seldom assessed.
Another significant gap in the understanding of travel impacts concerns the specific influence of the pricing project in observed changes. The combined or cumulative impact of pricing projects—the impact of the project coupled with the influence of a wide range of exogenous factors—has been relatively well established in many evaluations. However, most evaluations are not able to accurately isolate the discrete impact of the project. Most evaluations acknowledge one or two exogenous factors that may be relevant, e.g., gas prices or other transportation system changes, and some go as far as to describe the changes in those exogenous factors and qualitatively consider the potential general impact of those factors as they draw conclusions about the project impacts. But rarely are the impacts of the exogenous factors addressed quantitatively. The 2008 FHWA Value Pricing Pilot Program Lessons Learned Final Report came to similar conclusions, noting that:
Even in the few cases where evaluators have estimated the quantitative impact of certain exogenous factors, results have sometimes been challenged. For example, different researchers disagree about the how much of the traffic reductions were due to fuel increases.52
Use of household travel diaries in conjunction with conventional surveys which focus on the influences on travel behavior can be a powerful means for improved understanding of exogenous factors. However, as the evaluators of the Commute Atlanta mileage-based value pricing program evaluation found, use of household travel diaries is not effective unless changes in households between the before and after periods are carefully considered and unless sample sizes are large enough to balance significant demographic variation among households. The Commute Atlanta researchers concluded that the several well known and frequently cited U.S. congestion pricing evaluations have not sufficiently controlled for such factors.
Household Travel Behavior Changes
Use of household traveler surveys and travel diaries is not rare, but it is far from commonplace. Also, when household travel behavior has been investigated, there is often insufficient focus on shifts and trade-offs within the households, such as how a vehicle is used by family member X which was formerly driven by family member Y who has shifted to transit in response to pricing. Likewise, there has been little to no significant investigation of how total household travel budgets—in terms of time and money—factor into these intra-household changes and the connection to specific pricing strategies. As a result, the understanding of household impacts of congestion pricing projects is not complete.
The understanding of the impact on congestion pricing projects on person trips, especially person throughput and overall mode shares is incomplete. Many evaluations focus foremost on traffic impacts, or vehicle trips. Many of those evaluations that have included consideration of other modes have focused on transit ridership. Some of the largest and most robust evaluations, like London, have tracked person trip shifts to some degree—e.g., tracing X number of curtailed driving trips into the pricing zone—but those analyses have not included a truly comprehensive accounting of all person trips and their shifts. There has also been very limited consideration of person throughput. Challenges in cost-effectively collecting average vehicle occupancy data and data from travelers who telecommute, forego trips, bicycle, or walk likely contribute to the under consideration of person trip impacts.
Vehicle Speed Fluctuations (Driving Cycle)
Most evaluations focus on the before-after changes in the average speeds on specific roadway links or across roadway networks and these changes are relatively well understood. However, there is not yet a sufficient understanding of impacts on vehicle driving cycle (proportion of vehicle operation in acceleration, deceleration, idle and cruising modes) that are related to changes in traffic flow characteristics and which may not be accompanied by significant changes in average speeds. A roadway that shows little to no change in average speed may in fact demonstrate improved flow and a significantly different driving cycle profile. For example, pre-pricing, a road may have an average speed of 30 miles per hour (mph) that reflects brief bursts of 45 mph travel punctuated by frequent stops (idle). After pricing, the average speed may still be 30 mph but it may reflect a steady, 30 mph flow with no stops and starts. To date, evaluations have neither clearly identified the significance of traffic flow improvements on driving cycle and emissions nor eliminated the possibility that they may be as or more important than reductions in traffic volumes.53
Variability in Performance (Reliability)
Most evaluations focus on average or typical transportation system performance. There has been far less attention paid to the variability in transportation performance, that is, reliability. Gaps in this area include both objective quantification of reliability as well as thorough understanding of traveler attitudes and responses to varying levels of reliability.
Transit Crowding Implications
In so much as one objective of many pricing strategies is to shift some travel from driving to public transportation, transit services and capacity play a key role in congestion pricing success. Although evaluations often document the net changes to transit ridership, the implications of transit capacity and crowding are not well understood. As many transit agencies across the U.S. are implementing or considering service cut-backs, understanding these issues is especially important now.
This section discusses gaps in the understanding of the air quality, noise, and environmental justice (equity) impacts of congestion pricing projects.
What is currently understood regarding the impact of congestion pricing projects on air quality is based on the analyses that have calculated vehicle emissions. As noted in Section 3.2, roadside monitoring of before and after air quality has not enhanced that understanding.
Overall, the calculated vehicle emissions analyses that have been done have provided an understanding of the approximate or partial air quality impacts of congestion pricing projects. Inaccuracies and incompleteness stem from a failure to accurately specify all of the key input parameters—both traffic and emission rate-related.
Table 3‑7 identifies those determinants of congestion pricing project vehicle emission impact that are usually well represented in analyses as well as those that are not. It is the latter that underlie the gaps or uncertainties in the current understanding of the air quality impacts of congestion pricing projects.
Project-Attributable VMT. As discussed in Section 3.3.1, most analyses acknowledge a number of exogenous factors impacting observed, before-after changes in traffic volumes and average speeds, but very rarely are the traffic inputs to emissions volume quantitatively adjusted to eliminate the portion of variation attributable to exogenous factors. Indeed, analysts very seldom understand (or agree on) exactly how much of the observed variation is due to factors such as fuel price changes or employment changes.
Regional VMT and Speed Changes. Also as noted in Section 3.3.1, the analysis of the traffic impacts of congestion pricing projects typically focus on the priced facility/area and, somewhat less typically, may also include the immediately adjacent roadways. Examinations of potential impacts farther away, throughout the region, are rare. Correspondingly, emissions analyses have been unable to capture any of more distant impacts that may be created.
Hourly Variation. Many congestion pricing projects can be expected to have significantly different impacts during different times of day. It does not appear that hourly variations in VMT and or link speeds are standard considerations in congestion pricing project air quality analysis, and thus the understanding of these impacts is incomplete.
Driving Cycle Changes (Traffic Flow Change). The “grams per mile” emission rates (“factors”) used in emission calculations are almost always derived from emission factor models such as the U.S. Environmental Protection Agency (EPA) MOBILE6 model or the California Air Resources Board EMFAC model. Those models utilize assumptions—which, to varying degrees, can be manipulated by users—regarding the “driving cycles” of vehicles. Driving cycle refers to the proportion of a vehicle’s travel (VMT) under acceleration, deceleration, idle, and cruise (constant speed). Evaluations of congestion pricing project emissions impacts seldom include adjustments to reflect driving cycle assumptions in the emission factor models based on project-attributable impacts to driving cycles. The likely reasons for omitting such potentially important probably changes include: 1) It is usually unclear whether a project has impacted vehicle driving cycles (most traffic analysis look only at net changes in average roadway link speeds), and 2) Some analysts do not understand the significance of driving cycle in the emission factor estimation process, which is opaque in so much as it occurs internal to the emission factor model.
Vehicle Mix. Vehicle emission rates—grams per mile or grams per hour (idle)—vary significantly by vehicle type. The traffic analyses within most evaluations of congestion pricing projects do not appear to explicitly study project impacts on vehicle mix and, as a result, these changes are usually not included in the air quality analysis. That is, the breakdown of total observed link VMT for the before and after scenarios utilize the same vehicle mix. To the extent that congestion pricing projects do not differentially impact different vehicle types, it is of course appropriate to use the same mix, but to the extent that projects do impact this area, these impacts are seldom reflected in the emissions analyses.
There are no gaps, per se, in the understanding of the noise impacts of congestion pricing. This is because, even allowing for the likely inaccuracies in calculated (modeled) noise levels and, in the case of monitored noise levels, the inability to differentiate project impacts from exogenous impacts, congestion pricing projects have not been shown to, and are unlikely to, generate the magnitude of traffic changes needed to produce a change in noise levels that is perceptible to most people. Hearing sensitivity among humans is non-linear in that very large increases in the noise-generating activity are necessary to produce even the smallest perceptible changes in noise levels. In the case of traffic, at least a doubling (or halving) of traffic volume will be necessary to produce a change in noise levels that is noticeable to most people.54 Unless a congestion pricing project is expected to increase or decrease traffic levels by 50 percent or more, and/or noise is an extremely significant concern in a community, there is little value in even attempting to gauge the noise impacts of congestion pricing projects.
Overall, review of the published literature associated with the eight study projects as well as other documents assessing congestion pricing project effects and evaluation methods yielded fewer insights into knowledge gaps in the area of environmental justice than in the areas of travel impacts and air quality. That may be due in part to the fact that, although there is considerable information on environmental justice (or “equity” as it is often termed), there has been somewhat less focus on this topic than on travel impacts, a conclusion shared by a recent report on lessons learned from the FHWA Value Pricing Pilot Program.55 It may also be a function of the fact that much of the focus on environmental justice topics has been on predicting the impacts as part the design of the scheme so as to maximize public acceptance rather than solely on measuring the impacts of deployed projects. Predicting impacts has been a particular focus among European congestion pricing researchers.
Although there is not extensive information identifying gaps in the environmental justice knowledgebase, four specific areas can be identified:
The importance of continued examination of horizontal equity issues is based in part on evaluation findings that have shown that the vertical equity issues associated with congestion pricing projects may not be as great, or at least not as singular of a focus as had been expected.56 The need to gather more information on the long-term environmental justice impacts of congestion pricing reflects a concern common to essentially all impact areas. The importance of an improved understanding of the impacts of pricing revenue redistribution has been cited by several researchers, including the 2008 FHWA Value Pricing Pilot Program lessons learned report, the CURACAO study and a 2009 study by the RAND Corporation.57
51 K.T. Analytics and Cambridge Systematics, Inc., 2008.
52 Coordination of Urban Road User Charging Organisational Issues (CURACAO), “Deliverable D2: State of the Art Review (FINAL). Prepared by the Institute for Transport Studies, University of Leeds. May 2009.
53 Institute for Transport Studies, University of Leeds, 2009.
54 Institute for Transport Studies, University of Leeds, 2009.
55 K.T. Analytics and Cambridge Systematics, Inc., 2008.
56 K.T. Analytics and Cambridge Systematics, Inc., 2008.
57 Ecola, Lissa and Light, Thomas, “Technical Report – Equity and Congestion Pricing, A Review of the Evidence,” RAND Corporation, 2009.
United States Department of Transportation - Federal Highway Administration