Can Carpooling Clear the Road and Clean the Air?

Abstract

Proponents of High Occupancy Vehicle (HOV) lanes claim the lanes reduce vehicle-trips by encouraging more people to carpool, but the evidence is mixed. This article reviews studies on the impacts of HOV lanes with a focus on behavioral models. This research makes a case for performance measures with direct welfare, congestion, or air pollution effects and focuses on papers that explicitly model carpool formation and allow for induced demand. Papers on individual regions find that HOV lanes may increase or decrease welfare, while more general papers find that the impact of HOV lanes depends on underlying parameters. The Clean Air Act assumes that HOV lanes reduce traffic volume and improve air quality and recommends HOV lanes to areas with poor air quality. This research finds there is no consensus on HOV lane impacts and, that using HOV lanes as a travel control measure may be misguided, but that further research is warranted.

Keywords

travel demand management traffic control measures high occupancy vehicle lanes carpooling transportation economics

Introduction

The United States has built over 2,500 lane-miles of High Occupancy Vehicle (HOV) lanes to encourage carpooling and thus reduce congestion and automobile emissions. Proponents claim HOV lanes reduce the total amount of person delay because they give priority to higher occupancy vehicles. Proponents may additionally argue that HOV lanes can relieve congestion on adjacent general purpose lanes by encouraging people to carpool, thus moving more people in fewer cars. Detractors claim they are ineffective at reducing traffic and may produce worse outcomes than a general purpose lane or even not building an additional lane at all. Theory suggests and empirical work supports the notion that HOV lanes provide an incentive to carpool where there is a large difference between speeds on HOV and general purpose lanes. However, as HOV lanes encourage more drivers from the general purpose lanes to carpool, increased use of the HOV lanes and decreased use of the general purpose lanes reduce the time difference. Additionally, reduced congestion on HOV and general purpose lanes draws drivers who were previously taking other routes, not driving, driving off-peak, or taking transit, a behavioral response known as induced demand. The net impact of HOV lanes on traffic volume rests on the ability of the mode shift to carpooling to outweigh induced demand.

Studies on HOV lanes typically have one of the following limitations: a focus on performance metrics without clear welfare or environmental consequences, ad hoc assumptions about how carpools form, unrealistic assumptions about inelastic demand (thus ignoring induced demand), or results that only apply to a particular HOV lane. This article first explains these drawbacks, reviews the literature on carpooling, and finds that the evidence on HOV lanes is mixed. While some studies find HOV lanes can improve traffic and increase welfare in a study area, in other study areas HOV lanes have been found to worsen congestion and decrease welfare. More general studies find the success of HOV lanes depends on local conditions but these studies have shortcomings that are reviewed in this article.

Recently, residents of Los Angeles experienced “Carmageddeon,” a 42-hour highway closure to build HOV lanes on a 10-mile section of I-405. Proponents called the $1 billion expansion an investment in California that would encourage “the right type of growth” (Smith 2011) while detractors cite a correlation between the number of HOV lanes been built and a decrease in nationwide carpooling rates (Orski 2001; Ganos 2011).¹ Similar arguments have been made about new HOV lanes elsewhere and the research does not support either side. There is no evidence on the impact of HOV lanes on “smart growth” nor have HOV lanes been causally connected to the nationwide decline in carpooling. These questions are not impossible to answer, but they are difficult to answer adequately as they cross many disciplines and require the construction of a complicated counterfactual. Some evaluation projects ask the wrong questions while others use overly simplified methods to model travel demand. This article discusses which performance metrics are compatible with welfare analysis and makes a case for including endogenous carpool formation and induced demand to model HOV lanes. Better understanding of HOV lanes will have implications for HOV lanes and related programs such as the push to allow alternative fuel vehicles to use HOV lanes² and conversion to High Occupancy Toll (HOT) lanes.

Performance Measures

The reasons cited for building HOV lanes are vast, but like most public infrastructure projects they should be subjected to a cost–benefit analysis that includes benefits that may not be traded on the market such as additional leisure time, better air quality, and reduced greenhouse gas emissions. Concerns about equity can also be evaluated by looking at the distribution of costs and benefits, similar to the analyses in Rodier and Johnston (1997) and Safirova et al. (2003). When HOV lanes are used to improve traffic conditions, a careful accounting of the costs and benefits should be used to evaluate effectiveness. As an air pollution and greenhouse gas mitigation tool, a study on the impact of HOV lanes on vehicle miles traveled (VMT) and local air pollutants would be appropriate. Instead, many studies ask if HOV lanes are “effective” by focusing on performance metrics such as the number of people they move, the average vehicle occupancy (AVO), the impact of HOV lanes on parking costs, the violation rate in the HOV lane, and the public’s attitude toward HOV lanes.³ These statistics describe HOV lanes, but cannot tell a manager if congestion and air quality are improving due to the HOV lane. Chang et al. (2008) report the most cited reason for building HOV lanes is to “maximize person throughput” while PB Study Team (2002) determined HOV lanes were a success whenever they moved more people with fewer cars than a general purpose lane and provided an alternative to single occupancy vehicles (SOV). These outcomes do not necessarily translate into improvements in welfare, the reasons for which are further explained in the second section.

The Clean Air Act Amendments assume HOV lanes improve air quality by controlling traffic. Section 179(b)(1)(B) restricts nonattainment areas⁴ from expanding general purpose highway capacity but specifically does not place restrictions on new HOV lanes (Leman, Schiller, and Pauly 1994; Johnston and Ceerla 1996). Areas in severe or extreme nonattainment for ozone are required to enact transportation control measures such as HOV lanes. States that do not comply can face sanctions including the withholding of federal highway funds. Casey (2000) reports tension between highway and environmental managers on the impacts and goals of HOV lanes. Are these HOV lanes truly reducing travel demand from what it otherwise would be or are they simply a loophole for nonattainment areas to expand capacity without running afoul of the Clean Air Act? Even if traffic and air quality managers believe HOV lanes reduce congestion and improve air quality, they may be working against their stated goals by building additional HOV lanes. In these cases, congestion and air quality may be improved if the manager were to use other traffic control measures or even do nothing.⁵ Focusing on performance metrics such as AVO or violation rates can lead managers to believe HOV lanes are effective without asking fundamental questions about changes in welfare, congestion, and air quality.

Carpools Should be Endogenous

Many studies focus on the technical potential of carpooling but make ad hoc assumptions about the behavioral responses required to achieve the technical potential. For instance Noland, Cowart, and Fulton (2006) analyze “the impact of adding one person to every car motorway trip” to conclude that carpooling is one of the most effective ways to decrease fuel consumption. The authors list ways to theoretically achieve this increase in carpooling, but do not analyze whether carpooling incentives would be enough to get every commuter to give up the convenience, flexibility, and comfort of a single occupancy vehicle. Using a range of carpooling rates does not fix this error. A traffic manager who wants to justify an HOV lane but knows it will only be justified if carpooling increases by 5 percent can assume that carpooling will increase by 10 percent and run sensitivity analyses on what happens when it increases in a range of 5–15 percent. Ben-Akiva and Atheron (1977) assume a fixed time differential between HOV and general purpose lanes; however, time differentials between HOV and general purpose lanes are a result of an equilibrium between the costs of carpooling and the benefits. A large time differential will attract more carpoolers, which will congest the HOV lane and lead to a reduced time differential. A model is needed that allows individuals to decide whether to carpool based on the time and monetary cost of each lane where the time differentials are also determined by the actions of every other commuter. Wachs (1991) provides an early review of behavioral models to understand travel behavior, but these recommendations have not always been applied to carpooling behavior.

The Importance of Induced Demand

A new HOV lane expands capacity. If a converted HOV lane makes a highway able to transport more people, then it may also expand capacity. Expansions in capacity cause demand to increase, which may result in a worsening of traffic. This increase in demand may come from people who were using other models (transit, off-peak, or nonmotorized transportation), people taking more trips because congestion has cleared up, and long-run changes (e.g., people building houses farther away from city centers). Duranton and Turner (2011), Bento et al. (2011), Cervero (2002), and Johnston and Ceerla (1996) stress the importance of incorporating short- and long-term demand responses to capacity expansions. Duranton and Turner look at road expansions and find that VMT increases in “exact proportion to highways.” Bento et al. measure the impact of short-term induced demand in the context of HOV lanes to find a smaller effect, but one that is consistent with Duranton and Turner’s work. Building additional capacity or increasing the physical capacity without addressing demand for travel will not relieve congestion. Studies that ignore induced demand ignore an important empirical finding. Assuming inelastic demand biases results to favor HOV lanes because elastic demand reduces the gains from HOV lanes. In models without induced demand, the only changes that can occur are moves from SOVs to HOVs of two (HOV2), three (HOV3), or more (HOV4 +). Any decision about whether to expand capacity needs to account for short-term induced demand impacts such as mode shifts and additional trips as well as longer term impacts such as changing land use patterns.⁶ HOV lanes are no different.

Thus far, the evidence on HOV lane effectiveness is mixed, partially due to the difficulty of the problem. Realistic models are costly, difficult to solve, and data intensive. Ad hoc assumptions are often necessary since little is known about carpooling behavior, and researchers have yet to develop a series of best practices for modeling carpooling behavior. When studies have included realistic behavioral assumptions about carpooling, they have focused on one geographical area or road, making generalizations difficult.

Modeling techniques used to study HOV lanes are reviewed in the subsequent sections. The second section explains the economics of carpooling and discusses papers that examine alternative performance measures. The focus is on peer-reviewed literature, although unpublished reports from engineers, transportation agencies, and consulting firms are included as evidence on the practice of HOV lane modeling. The second section is not an exhaustive list of papers that focus on alternative performance measures, but instead an explanation for why measures of welfare, VMT, and air pollution are preferable. The third section discusses engineering models of transportation demand, particularly the demand portions of the four-step model (FSM). The fourth section covers discrete choice models and models of utility maximization. The fifth section discusses models based on cost-minimizing behavior. The results of these models are summarized in Table 1. The sixth section documents additional studies that model carpooling behavior but do not evaluate HOV lanes. The seventh section discusses HOV lanes and equity. The focus of this article is looking at studies with a behavioral component. Papers that focus on structural features without a behavioral component are mentioned but not the focus of this work.

Table 1.

Summary of Traffic-Based Models that Evaluate the Impact of HOV Lanes on Welfare, VMT, and Air Pollution

Authors	Type of Model	Endogenous Carpools?	Induced Demand?	Area	Impact on Welfare	Impact on VMT or Air Pollution
Mannering and Hamed (1990)	FSM	No	No	State College, Pennsylvania	Increases with high vanpool and HOV percentages	—
Burris and Lipnicky (2009)	FSM	No	No	Houston	—	Decreases fuel consumption and total travel time
Johnston and Ceerla (1996)	FSM	Yes	Yes	Sacramento	—	Increases VMT by 4 percent
Rodier and Johnston (1997)	FSM	Yes	Yes	Sacramento	Decreases	Increases VMT and emissions of TOG, CO, NOx and PM
Cambridge Systematics (2002)	FSM, DCM	No	No	Twin Cities, MN	—	Decreases VMT but increases fuel consumption and emissions
Ben-Akiva and Atheron (1977)	DCM	No	Yes	Washington, DC	—	Decreases
Small, Winston, and Yan (2006)	DCM	Yes	Yes	SR 91, Southern California	Increases	Decreases
Dahlgren (1998)	DCM	Yes	No	General	—	Increases except where general purpose lanes are congested
Yang (1988)	DCM	Yes	No	General	Increases, but no sensitivity analysis	Decreases
Konishi and Mun (2010)	Cost Min	Yes	No	General	Depends on Parameters	Depends on parameters

Defining Success

Unpriced and underpriced transportation systems are inefficient because the marginal private cost of driving (time, fuel, tolls) is lower than the marginal social cost of driving, which includes externalities (congestion, emissions). Drivers do not take into account their contributions to congestion, air pollution, and accidents. This results in overutilization and excessive pollution. In this section, I argue that an efficient system would involve more carpooling and less driving than an unpriced general purpose lane, but that not all systems with higher carpooling rates or higher person throughput are more efficient (or welfare enhancing) than unpriced general purpose lanes. An economically efficient allocation of road space would maximize the difference between aggregate consumer benefit and total cost including both social and private costs (Small and Verhoef 2007). For policies such as HOV lanes that presume to be an improvement over the status quo, welfare should be calculated with and without HOV lanes to determine their effectiveness. This welfare measure could be constructed by modeling mode choice through a discrete choice model and calculating social surplus under each scenario (Small, Winston, and Yan 2006).

Maximizing Social Surplus

This section uses a numerical example to show the difference between the social optimum and decentralized traffic equilibrium. This example is derived from the classic explanation for congestion reviewed in both Small and Verhoef (2007) and Rouwendal and Verhoef (2006) but is meant to be purely illustrative and not representative of all traffic systems.

In Table 2, there are seven types of roadway users, A through G, who derive heterogeneous benefits and costs of travel. The costs of driving reflect both fuel and time costs and depend on how much congestion exists on the road. More cars on the road raise travel costs for everyone. Carpooling costs differ from driving solo as carpoolers save money on the line haul portion of the trip but spend more time assembling the carpool and incur costs from a loss of flexibility.

Table 2.

Hypothetical Costs of Driving Alone and Carpooling—No HOV Lanes

User	A	B	C	D	E	F	G
Benefit	6.00	4.00	4.00	1.40	1.40	1.00	1.00
Congestion-dependent cost of driving alone when traffic volume is:
1 car	1.00	1.00	1.00	1.00	1.00	0.88	0.88
2 cars	2.00	1.50	1.50	1.16	1.16	1.00	1.00
3 cars	3.00	2.00	2.00	1.32	1.32	1.12	1.12
4 cars	4.00	2.50	2.50	1.48	1.48	1.24	1.24
5 cars	5.00	3.00	3.00	1.64	1.64	1.36	1.36
Congestion-dependent cost of carpooling when traffic volume is:
1 car	3.00	1.80	1.80	1.10	1.10	0.84	0.84
2 cars	3.50	2.05	2.05	1.18	1.18	0.90	0.90
3 cars	4.00	2.30	2.30	1.26	1.26	0.96	0.96
4 cars	4.50	2.55	2.55	1.34	1.34	1.02	1.02

If the benefits of driving outweigh the costs, a user will drive. Once the user has decided to drive, the user will pick the mode with the lowest costs. For instance, if there is only one car on the road, user A will receive $6 of benefits from driving and incur $1 in costs to drive alone or $3 to carpool. Since the net benefits of driving are positive and driving alone is the lowest cost mode, user A will drive alone and receive a net benefit of $5. Equilibrium occurs when the total number of people who want to drive alone plus half the carpoolers is equal to the number of vehicles on the road. Using numbers from Table 2, this occurs when users A, B, and C drive alone while D and E carpool. This results in four cars total, which raises costs such that F and G receive a negative net benefit from both solo driving and carpooling, which keeps F and G from using the road. The total social benefits are $5.12, which can be calculated by summing the benefits of driving for A, B, C, D, and E and subtracting the costs by mode for all four users.

The decentralized equilibrium can be improved upon. Users D and E raise the total costs of driving for A, B, and C by $2 and only receive a net benefit of $0.12. A, B, and C could pay D and E between $0.12 and $2 to use transit or stay off the road and all five parties would be better off. This is a Pareto optimal improvement and would result in a social surplus of $7.00. An even higher social surplus of $7.90 could be achieved by having A drive alone while B and C carpool. Either of these Pareto improvements could be achieved through a toll. Depending on how the revenues from tolls are redistributed, this could lead to a more equitable situation than the status quo.

Road pricing is politically unpopular and not an option for many transportation managers, but is still a useful concept to understand economically efficient systems. One important thing to note is that the most efficient system is not necessarily the one that maximizes person throughput. A strict maximization of person throughput could be achieved by building new lanes and subsidizing all vehicle trips, even those for which the marginal benefit is lower than the marginal cost.⁷ Induced demand is a combination of people forsaking discretionary trips and using alternative modes such as transit, walking, or cycling (Cervero 2002). Thus, the people who are induced to carpool by HOV lanes may belong to one of three counterfactual groups: driving solo, coming from other modes, or simply not traveling in the absence of HOV lanes. An economically efficient system provides a superior experience for those on the road because it has less congestion and allocates that space to those with the highest value for it. An economically efficient system may not be equitable, but the toll revenue collected may be able to fund alternative transportation or used for other equity goals.

An HOV lane can be modeled as a capacity expansion for carpools that lowers the costs of carpooling. The HOV lane will not change the capacity of the general purpose lane if it is an additional lane, but it will raise the costs of driving on the general purpose lanes if the HOV lane is a conversion of one of the general purpose lanes. This is shown in Table 3 where I have presented hypothetical costs for users A through G when there is an additional (Panel 1) and a converted HOV lane (Panel 2). The additional lane does not change the capacity of the general purpose lane, but the converted lane does.

Table 3.

Hypothetical Costs of Driving with HOV Lanes

User	A	B	C	D	E	F	G
Benefit	6.00	4.00	4.00	1.40	1.40	1.00	1.00
Panel 1. Costs of driving with an additional HOV lane
Congestion-dependent cost of driving alone when GP traffic volume is:
1 car	1.00	1.00	1.00	1.00	1.00	0.88	0.88
2 cars	2.00	1.50	1.50	1.16	1.16	1.00	1.00
3 cars	3.00	2.00	2.00	1.32	1.32	1.12	1.12
4 cars	4.00	2.50	2.50	1.48	1.48	1.24	1.24
5 cars	5.00	3.00	3.00	1.64	1.64	1.36	1.36
Congestion-dependent cost of using HOV lane when HOV traffic volume is:
1 car	3.00	1.80	1.80	1.10	1.10	0.84	0.84
2 cars	4.30	2.45	2.45	1.31	1.31	1.00	1.00
3 cars	5.60	3.10	3.10	1.52	1.52	1.15	1.15
4 cars	6.90	3.75	3.75	1.72	1.72	1.31	1.31
Panel 2. Costs of driving with a converted HOV lane
Congestion-dependent cost of driving alone when traffic volume is:
1 car	2.60	1.80	1.80	1.21	1.21	1.21	1.21
2 cars	4.20	2.60	2.60	1.37	1.37	1.33	1.33
3 cars	5.80	3.40	3.40	1.53	1.53	1.45	1.45
4 cars	7.40	4.20	4.20	1.69	1.69	1.57	1.57
5 cars	9.00	5.00	5.00	1.85	1.85	1.69	1.69
Congestion-dependent cost of using HOV lane when HOV traffic volume is:
1 car	3.00	1.80	1.80	1.10	1.10	0.84	0.84
2 cars	4.30	2.45	2.45	1.31	1.31	1.00	1.00
3 cars	5.60	3.10	3.10	1.52	1.52	1.15	1.15
4 cars	6.90	3.75	3.75	1.72	1.72	1.31	1.31

Table 4.

Comparison of Outcomes in No HOV and HOV Scenarios

	No HOV Lanes		Additional HOV Lane	Converted HOV Lane
	Decentralized	Social Optimum	Additional HOV Lane	Converted HOV Lane
Solo drivers	A, B, C	A	A, B, C	A
Carpoolers	D, E	B, C	D, E, F, G	B, C, D, E
Traffic volume	4	2	5	3
Average vehicle occupancy	1.25	1.5	1.4	1.67
Social surplus	5.12	7.90	7.18	6.68

Building an additional HOV lane results in five cars on the road: A, B and C drive solo, while D, E, F, and G carpool in the HOV lane. The total social surplus in this situation is $7.18 which is higher than the decentralized equilibrium without HOV lanes but of course extra capacity has been added to the transportation network. A converted HOV lane results in a total of three cars on the road: A drives along while B and C carpool on the HOV lane. The total social surplus in this case is $6.68. The additional lane increases welfare and traffic volume by adding capacity, while the conversion of a general purpose lane decreases welfare and traffic volume relative to the social optimum but is still an improvement to the decentralized no HOV lane scenario. This is specific to our particular example, but if we were to use performance metrics such as the number of carpoolers or average vehicle occupancy, the HOV lanes would be labelled a success. Using a different set of plausible costs and benefits of driving, an HOV lane may be welfare decreasing. However, this welfare-decreasing HOV lane may be observationally equivalent to a welfare increasing HOV lane if we are only observing the number of carpoolers. Thus it is essential to be thoughtful on what constitutes a success in transportation policy.

Other Performance Measures

Distinguishing between a welfare-increasing lane and a lane that decreases welfare is difficult without a model that incorporates many different concepts and parameters from economics to civil and transportation engineering.

A careful accounting of transportation costs could provide the same information as a welfare metric.⁸ Transportation costs would need to account for time and monetary costs, assembly costs for forming carpools, utility/disutility for changing modes, pollution and other social costs, and a reservation cost for people who do not use single or HOVs because they have been priced off the road by congestion. Differences in the cost of building and maintaining HOV lanes should be included. Most studies rely on crude estimates of transportation cost by looking only at fuel usage, changes in time spent traveling, accidents, and pollution.

When road pricing is unavailable, air quality managers may choose to protect public health using HOV lanes to reduce pollution. Stop-and-go traffic produces more pollution per mile, thus VMT and speed should be considered when accounting for the impact of HOV lanes on local air pollutants such as ozone and particulate matter and to a lesser extent greenhouse gases. Cambridge Systematics (2002), one of the few studies to consider VMT, speed, and air quality, finds that even when HOV lanes decrease VMT, HOV lanes create enough congestion on the general purpose lanes to increase local air pollution and fuel consumption. This may be an exceptional case; more research that integrates behavior with air emissions scenarios is warranted. Table 1 in this article uses VMT as a proxy for environmental quality and trip reduction, although I admit there are problems with this proxy.

Instead of studying changes in welfare or VMT due to HOV lanes, studies focus on scale, utilization, and persistence of HOV lanes (Schofer and Czepiel 2000), AVO (Levine and Wachs 1998; Kwon and Varaiya 2008; PB Study Team 2002), parking costs (Ulberg and Jacobson 1988), passengers per lane (Ulberg and Jacobson 1988; PB Study Team 2002), and public attitudes toward HOV lanes (PB Study Team 2002; Cambridge Systematics 2002; Jou, Went, and Chen 2005; Chang et al. 2008). Even when researchers document a meaningful measure such as welfare or VMT, they focus on an individual road or even an individual lane without considering system-wide impacts on the general transportation system (Wellander and Leotta 2001; Martin et al. 2004). This results in a biased view of the city-wide impacts of HOV lanes (Cervero 2002). Schofer and Czepiel (2000) argue that when HOV lanes carry as many or more people than the general purpose lane next to it, the HOV lane should be considered a success. However, a sorting of carpools onto HOV lanes could have achieved higher AVO and passengers per lane than the general purpose lanes without even one new carpool. Daganzo and Cassidy (2008) use this possibility as a baseline to study whether HOV lanes that do not encourage carpooling can still reduce person hours of travel. Daganzo and Cassidy find that HOV lanes can reduce person hours of travel even without reducing VMT. This finding is encouraging but the authors do not tie their results directly to welfare outcomes. If drivers of single-occupant vehicles value their time more than HOV passengers, or spend large amounts of fuel and time on arterial roads to form carpools, social welfare may be lower with HOV lanes. If HOV lanes induce carpooling that would not be undertaken in the counterfactual case without equivalent reductions in solo vehicles, HOV lanes will increase person hours by increasing trips and may slow down average person hours by increasing congestion.

Kwon and Varaiya (2008) use freeway sensor data from California’s Performance Measurement System (PeMS) to find that HOV lanes in California are both underutilized and degraded. They find that HOV lanes suffer a 20 percent capacity penalty, confer small time savings, and only reduce congestion when the general purpose lane is allowed to become congested. While the authors make use of some of the best data and methods available on HOV lanes, their measurements are not closely related to welfare metrics. If roads are to remain unpriced, then zero congestion or perfect utilization may not be economically efficient (Verhoef, Nijkamp, and Rietveld 1996; Braid 1996). Kwon and Varaiya present these findings as evidence of the failure of HOV lanes. However, an HOV lane with small time savings and capacity penalties may still describe a more efficient system than a highway without HOV lanes. These metrics are not tied to changes in VMT, air pollution, or welfare and so cannot be used to evaluate economic efficiency.

If the private value of transportation is less than the social costs, then maximizing person throughput decreases social welfare. When facility managers say they build HOV lanes to maximize person throughput (Chang et al. 2008), it is no wonder some environmentalists see HOV lanes as a thinly disguised attempt to build more roads (Orski 2001). The California Legislative Analyst’s Office (2000) has called for more research on the impact of HOV lanes on air pollution and traffic volume. Instead of quantifying the impact of HOV lanes on air pollution and traffic volume, subsequent reports on HOV lanes have focused on changes in performance metrics such as AVO (PB Study Team 2002). AVO is easier to measure than welfare, but it will not tell policy makers whether building HOV lanes improves congestion and air quality. To understand these impacts, researchers need to use the models described in the next four sections.

The FSM

To understand the impact of HOV lanes, analysts need to build transportation demand models. The workhorse of transportation demand modeling is the FSM. Planners and transportation engineers use the FSM to examine questions such as how many lanes should a bridge have, what is the financial viability of a project, and what are the potential environmental impacts of a project. The FSM dominates transportation planning, despite well-known inadequacies in terms of realism and its ability to answer many relevant policy questions (McNally 2005; Cervero 2002; McNally and Recker 1986).

The first step in the FSM is trip generation to determine the number of trips taken from each trip origin and the number of trips attracted to each destination. The models use demographic and land use information to generate origin–destination matrices which forecast the number and type of trips coming from and going to each area. Trip generation is not modeled using economic fundamentals, and instead is based off historical associations between trip levels and demographic variables. Some models include friction factors that express the reduction in trips taken with greater travel times, but this is an aggregate measure. The FSM’s trip generation step might be adequate for predicting future demand, but this presents a problem in analyzing structural changes such as an upgraded bus system, new transportation technologies, incentives for carpooling, or congestion pricing.

Step 2 in the FSM is trip distribution. This step uses a gravity model or similar method to connect origins with destinations and hence load the demand generated through the first step onto the transportation network. Steps three and four are mode choice and route choice. Mode choice determines the proportion of trips that travel by each mode where modes usually include transit, HOV2, HOV3, and driving alone. Many FSMs predict mode choice with a discrete choice model, which is reviewed in the next section. Route choice allocates origin-destination pairs by a particular mode to a route. This step relies on Wardrop’s principle of user equilibrium (equivalent to a Nash equilibrium in economics) that states each traveler chooses the path with the lowest travel time subject to the decisions of all the other travelers. The FSM has a significant advantage over other models in its ability to model large metropolitan areas and account for complicated geography. Even with simplistic assumptions of trip generation and route choice, an FSM may take millions of dollars to calibrate and weeks to run.

Instead of building a model of carpool formation in the mode choice step, some travel models embed ad hoc assumptions or rely on historical associations. Mannering and Hamed (1990) assume HOV lanes will increase the percentage of HOV passengers from 17 percent of commuters to 30 percent or 40 percent. Burris and Lipnicky (2009) generate a change in mode share by assuming that between 50 percent and 100 percent of carpoolers will continue to carpool when HOV lanes on the Katy Freeway in Houston are converted to general purpose lanes. Burris and Lipnicky justify this range with surveys finding between 65.5 percent and 58.2 percent of carpools on the Katy Freeway consist of family members who they assume would continue to carpool without an HOV lane. This may be defensible in the short term, but with the loss of time savings in the HOV lane, even family members who carpool may switch to driving alone despite higher monetary costs because driving alone comes with greater flexibility. Using a cutoff of 50 percent is as arbitrary as assuming historical percentages of HOV passengers.

Other models employ more complicated but not necessarily better assumptions about mode share. Mallinckrodt (2003) summarizes some of the FSM or similar transportation planning models used to analyze HOV lanes from 1979 to 2002. Most of the FSMs he cites find HOV lanes increase traffic volume.

Johnston and Ceerla (1996) use an FSM to understand the impact of adding 206 new freeway lane-miles of HOV lanes to the Sacramento region. Their model includes friction factors that purport to model the reduction in commute trips that result from high travel times. The friction factors are applied individually to each mode considered and thus do not represent decisions to switch modes depending on travel time differentials between HOV and general purpose lanes. The results are sensitive to initial assumptions. The authors find building a new HOV lane increases VMT by 4 percent but decreases person delays relative to the no-build scenario. This is one of the more realistic articles on the impacts of HOV lanes, but the results cannot be generalized outside the Sacramento metropolitan area. They do not model land use or automobile ownership. If HOV lanes promote sprawl and higher automobile ownership the way regular lanes do, this is a lower bound for VMT increases due to HOV lanes. Rodier and Johnston (1997) use a similar model based on newer data which finds HOV lanes decrease welfare, increase VMT, and increase emissions. Again, this cannot be generalized outside of Sacramento.

Cambridge Systematics (2002) uses an FSM to examine HOV lanes in the Twin Cities area. The FSM used has a discrete choice model to predict mode share, although it is not clear what is included as a predictor of mode choice. The model assumes inelastic demand for trips, thus the only changes in traffic volume come from commuters shifting between modes. The report looks at the potential conversion of HOV lanes to general purpose lanes and finds that converting the lanes to a general purpose lane would increase VMT, but overall time spent traveling would decrease. The report states a conversion from HOV to general purpose would decrease emissions of carbon monoxide and the ozone precursors, hydrocarbons and oxides of nitrogen, and decrease fuel consumption. Thus, HOV lanes decrease VMT compared to a general purpose lane. HOV lanes increase the amount of vehicle delays due to congestion results and increase local air pollution and fuel consumption. The long run effects of increasing VMT are not addressed.

Discrete Choice Models

Embedded in some FSMs are discrete choice models for mode choice. Discrete choice models have a long history with the transportation literature, starting with the additive random-utility model of McFadden (1974). In discrete choice models, user n decides between alternatives j = 1, . . . , J by choosing the alternative with the highest utility given by:

U_{j, n} = V (z_{j, n}, s_{n}; β) + ϵ_{j, n} .

Here V (·) is known as the systematic utility, z_j,n is a vector of alternative specific attributes, and s_n is a vector of characteristics specific to the decision maker. The unobservable part of the model captures idiosyncratic preferences and is represented by ∊ _j,n .

An individual is said to choose alternative j if the utility the user receives from alternative j is higher than what he or she would have received from all the other alternatives, U_j,n > U_i,n ∀j ≠ i. This can be rewritten as the probability that decision maker n chooses alternative j:

P_{j, n} = Pr (U_{j, n} > U_{i, n} \forall j \neq i) = Pr (V_{j, n} + ϵ_{j, n} > V_{i, n} + ϵ_{i, n} \forall j \neq i)

The researcher can specify a functional form for V (·) and an error structure for ∊_j,n, then estimate a model with direct applications to welfare analysis. Common modeling decisions are to use a linear structure for V (·) and either a logistic distribution for ∊, resulting in the logit model, or a normal distribution for ∊ that results in the probit model. Discrete choice models can be embedded into four-step transportation models as a simple logit in the mode share step or may be used without an FSM.

An early use of discrete choice for carpooling is in Ben-Akiva and Atheron (1977). This article models carpooling as a multiple-level decision from where to live, whether to buy a car, and how many trips to take. The authors look at the impacts of HOV lanes, employer incentives, preferential pricing, and increases in the price of gas on traffic volume. The authors do not allow for the time differential between HOV and general purpose lanes to be endogenous. They calibrate their model to the Washington, DC, area and the Santa Monica Freeway in Los Angeles but only analyze Washington, DC, to find that HOV lanes reduce traffic volume. This model does include an automobile ownership model but the focus is on short-range travel decisions and as such does not allow land use patterns to be endogenous.

Small, Winston, and Yan (2006) estimate a discrete choice model of route choice for a section of a HOT lane on State Route 91 in Southern California. Travelers have a choice between driving alone, carpooling, or paying a toll to use the HOT lane but medium to long-range decisions such as automobile and location are fixed. To estimate value of time and reliability, the authors estimate a specific type of discrete choice model, a nested logit. The article focuses on HOT lanes, but also run a simulation to understand the impact of HOV lanes on VMT and consumer surplus. Small et al. find that HOV lanes provide improvements for both carpoolers and non-carpoolers by doubling the share of people who choose to carpool. They also find that HOV lanes induce people who were not traveling on the corridor to travel on the corridor. The authors demonstrate that discrete choice models can be used to explore the efficiency considerations and create a behavioral explanation of carpooling. This is the only article found that uses a welfare metric to evaluate HOV lanes, allows carpool formation to be endogenous, and allows for induced demand. However, it only applies to one particular road and the results may not be applicable outside of State Route 91 in Southern California.

Dahlgren (1998) embeds a discrete choice model into a traffic equilibrium model to create a general model that can be used to understand under which sets of conditions HOV lanes reduce traffic and congestion. Dahlgren makes the share of carpoolers a function of the time differentials between HOV and general purpose lanes, which is also determined by the model. This model allows the number of carpools to be endogenous but does not allow for induced demand nor long-term land use changes.⁹ Dahlgren’s analysis recognizes that the success of an HOV lane in inducing people to use it is a function of the general purpose lane’s failure to provide uncongested travel. Dahlgren’s results indicate that HOV lanes perform best when the vehicle delay from congestion is high and the percentage of carpoolers without an HOV lane is close to the percentage of lane space later devoted to the HOV lane.

Yang (1998) uses a discrete choice model to build a general model of HOV lanes and investigate when HOV lanes can reduce travel costs on a multi-lane highway. Yang finds that HOV lanes may reduce travel costs (and hence improve welfare) but Yang does not allow for induced demand. Yang provides a numerical example but does not explore the parameter space or perform a sensitivity analysis.

Cost Minimization

Another way to represent optimizing agents is through a cost minimization model. Only one paper was found to do cost minimization with regard to HOV lanes, but there is a rich literature in transportation that models users as cost minimizing agents (de Palma, Kilani, and Lindsey 2008; Arnott, De Palma, and Lindsey 1993a, 1993b; Vickrey 1969). Konishi and Mun (2010) use a cost minimization framework and allow assembly costs, but not time costs, to vary over individuals. Konishi and Mun assume inelastic transportation demand, thus leaving out induced demand. Consumer cost is modeled as a function of the commute cost which is a function of congestion, C(q_i ), plus an assembly cost, t, that varies across commuters according to the distribution function F: R+ → [0, 1], and finally a toll, τ, that varies by lane and carpooling decision.

C (q_{i}) + e t + τ_{i, e} .

The variable e is an indicator variable that denotes the commuter’s carpooling decision, e = 0 if the commuter is not carpooling and e = 1 if the commuter carpools. Tolls and congestion have a subscript to denote which lane the commuter drives in, general purpose, HOV, or HOT.

Konishi and Mun find HOV lanes can be a welfare improvement over general purpose lanes only under certain sets of parameters, but can aggravate congestion in other cases. HOV lanes encourage carpooling and reduce total traffic, but cause distortions by creating different levels of congestion between general purpose and HOV lanes. HOT lanes can mitigate this by allowing solo drivers on HOV lanes, but they also discourage carpooling.

Additional Models of HOV and HOT Lanes

Travel models are not the only way to study HOV lanes. This section reviews studies that analyze HOV lanes but do not or cannot calculate the impact of HOV lanes on welfare or VMT. Teal (1987) analyzes who is likely to carpool using the discrete choice framework and discounts the notion that a substantial number of commuters can be encouraged to carpool using incentive programs. Brownstone and Golob (1992) address Teal’s concern by building a model that predicts which kinds of incentives favor ride sharing: HOV lanes, guaranteed rides home, preferential parking, subsidies, or an amorphous category of other incentives. Bard (1997) looks at how carpooling decreases the total number of trips taken by a household, but does not extend this to traffic outcomes or welfare calculations. Plotz, Konduri, and Pendyala (2010) put bounds on how many carpools were formed because of HOV lanes. Lee (1984) looks at the optimal number of people in a carpool and Ferguson (1997) studies trends in carpooling. Boriboonsomsin and Barth (2008) examine the air quality impacts of continuous versus limited access HOV lanes. These papers could be used as inputs into a study looking at the effectiveness of HOV and HOT lanes but by themselves do not evaluate effectiveness.

A few papers employ models of carpooling behavior to study other subjects. This is not a review of HOT lanes, but some of the techniques used to study HOT lanes also involve studying HOV lanes. Safirova et al. (2003) and Houde, Safirova, and Harrington (2007) develop a model for the Washington, DC, metro area where carpooling is endogenous and demand for trips is elastic. They do not use their model to study the efficiency of HOV lanes but focus on whether or not to convert existing HOV lanes to HOT lanes. Naga (2007) examines under which conditions it would be preferable to have an HOT versus an HOV lane with considerations for equity and how tolls are set. The behavioral aspect of both models use a logit where each mode is decided based on time cost, operating cost, toll cost, and the cost to assemble a carpool. Naga does not allow for induced demand in this model nor does Naga evaluate HOV lanes. Huang, Yang, and Bell (2000) build a discrete choice model of HOT lanes and derive optimality conditions for tolling structures with carpooling. Yang and Huang (1999) build a model of HOV lanes that uses linear demand functions to examine optimal toll regimes when HOV lanes are present.

Equity and HOV Lanes

While the overall impacts of HOV lanes are understudied, even less is known about the distribution of these impacts or the equity of HOV lanes. The equity issue has been extensively studied with regard to road pricing but less with non-tolled travel demand management policies. This is unfortunate. Changes in access, congestion, travel time, and convenience from transportation policies change the costs and benefits felt by society even if individuals do not face different upfront costs. Sakano, Benjamin, and Ben-Akiva (2001) document that Metropolitan Planning Organization (MPO) managers are concerned about equity. The authors find that MPO managers believe HOV lanes benefit primarily middle-income consumers. Their result is based on MPO managers’ perceptions of public usage, not consumer surveys or models that document changes in consumer welfare.

Equity may be measured across income groups, demographic groups, time, or space. There are many ways to summarize the distribution of costs and benefits across these groups. Ramjerdi (2006) argues that no single measure of equity is sufficient and that instead policy makers should examine a range of measures. Many of these measures, such as the variance of logarithms, Gini coefficient, Theil’s entrophy, Atkinson, and Kolm statistics are technical and counterintuitive.

Applied work has examined equity by calculating changes in consumer welfare by income and demographic groups (Bureau and Glachant 2008; Eliasson and Mattsson 2006; Safirova et al. 2003). Four-step, discrete choice, and cost-minimization models can be used to evaluate changes in the distribution of consumer welfare much as they can be used to evaluate overall welfare. Rodier and Johnston (1997) use compensating variation to measure changes in welfare and find that converting general purpose to HOV lanes in Sacramento makes all consumer groups worse off, but that poor households lose less than rich households. Of course, these results only apply to Sacramento. Leman, Schiller, and Pauly (1994) argue that HOV lanes favor long haul over shorter bus routes. The authors claim long-haul bus routes in Houston divert funding from shorter routes harming urban passengers, especially the low-income. This evidence is anecdotal and the equity impacts will depend on the spatial layout of poor populations.

While the theory on evaluating changes in consumer welfare by income or demographic group is straightforward, in practice, equity is often ignored. Rodier and Johnston is the exception in the HOV literature, although many other papers have examined equity regarding congestion pricing (Bureau and Glachant 2008; Ramjerdi 2006) or the equity impacts of HOT lanes (Weinstein and Sciara 2006; Naga 2007; Burris and Hannay 2003; Safirova et al. 2003). Levinson (2010) reviews the literature on the equity impacts of HOT lanes and finds that while HOT lanes generally benefit the better off more than the poor, acceptability is widespread across groups and many groups make use of HOT lanes. This does not necessarily translate to HOV lanes. Often HOV lanes are the status quo from which HOT lanes are judged. The impact of HOV lanes on equity will depend on initial travel patterns (Eliasson and Mattsson 2006), the distribution of vehicle ownership, assembly costs, and other programs such as transit service, HOT lanes, and hybrid exemptions for HOV lanes. This article argues that HOV lanes are understudied and the literature on the efficiency of HOV lanes is mixed. It is unsurprising that the literature on the equity impacts of HOV lanes is even scarcer.

Conclusion

Table 1 summarizes the evidence on HOV lane effectiveness. Seven of the papers report on a specific area, while the last three look more generally at the conditions under which HOV lanes may be effective. The quality of assumptions and level of detail in each model vary. Making carpools endogenous and including induced demand are important to modeling the behavioral aspects of carpooling and many models do not take this into account. Long-range impacts such as increased sprawl or changes in land use are also often ignored. Cambridge Systematics (2002)’s finding that HOV lanes can increase VMT but still lead to an improvement in air quality highlights the need for a more thorough integration between carpooling behavior and emissions modeling.

Any reduction in traffic from increased carpooling will induce more drivers to use both the general purpose and HOV lanes. Shorter travel times will make outer suburbs more attractive, which Leman, Schiller, and Pauly (1994) argue contributes to sprawl. Omitting these effects by assuming inelastic demand or ignoring changes in land-use and location decisions will bias a model toward HOV lanes. Ben-Akiva and Atheron (1977) and Small, Winston, and Yan (2006) include induced demand but still find HOV lanes decrease VMT. These two area-specific studies of Washington, DC, and State Route 91 in Southern California are in contrast to Johnston and Ceerla (1996) who include induced demand in a study of Sacramento and find HOV lanes increase VMT largely as a result of induced demand. None of the general studies include induced demand, but many of them find HOV lane effectiveness depends on underlying structural parameters which may explain the different answers coming from Southern California, Washington, DC, and Sacramento.

Billions of dollars¹⁰ are spent building HOV lanes with only vague definitions of the benefits and incomplete understanding of the impacts. The general confusion as to HOV lane effectiveness is an area where academics can have an impact. Mainly on the work done by Dahlgren, California Assemblyman Tom McClintock sponsored Assembly Bill 44 (AB 44), which requires a review of all HOV lanes. Whether HOV lanes are more efficient than the alternatives, questions posed by AB 44 and the Legislative Analyst’s Report (2000), still have not been answered. PB Study Team (2002) tried to answer concerns brought by the LAO report, but the team did not address welfare or VMT and did not have conclusive evidence that air quality was improved or worsened with HOV lanes. Even though most of the approaches listed here do not adequately model behavior or discuss welfare metrics does not mean such models are impossible. Modifications of Small, Winston, and Yan (2006) to examine a wider range of parameters or expanding Dahlgren (1998) to include induced demand would answer these important questions. An empirical study of the impact of HOV lanes on general traffic impacts using an instrumental variable such as Duranton and Turner (2011) could examine the empirical evidence.

Creating new models and welfare measures for each lane may be prohibitively expensive and beyond the expertise of practitioners who may not be behavioral specialists. If this is the case, researchers may be able to aid traffic managers by linking favored metrics that are easy to observe (such as AVO, passengers per lane, average speeds, population, and values of time) to welfare and emissions outcomes to provide rules of thumb. Thus far, this has not happened, partially because of the difficulty and expense of integrating cutting edge behavioral research with dynamic traffic models. More definitive studies that follow Dahlgren’s attempt to draw informative, welfare-based conclusions are needed. These studies could then be applied by managers to particular areas as reliable, research-based tools to control pollution and congestion.

Transportation policy needs to be based on empirical and theoretical work that rigorously models travel demand and employs performance metrics with welfare implications. Advocates for replacing HOV lanes with HOT lanes (Poole and Orski 1999), building more HOV Lanes (PB Study Team 2002), or converting HOV lanes to general purpose lanes (Orski 2001), are forced to rely on opinions and ad hoc measures such as whether an HOV exhibits “empty lane syndrome.” The Clean Air Act encourages nonattainment areas to use HOV lanes as a travel-demand management tool; yet the effectiveness of this tool is unknown (Johnston and Ceerla 1996). HOV lanes may worsen air quality but are exempt from the kind of scrutiny general purpose lanes receive in nonattainment areas. This article has shown a need for further research in this area and a need to recognize that despite over forty-two years of building HOV lanes, we still do not know whether they increase welfare, reduce congestion and VMT, or improve air quality.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Arnott

De Palma

Lindsey

1993a. “Economics of a Bottleneck.” Journal of Urban Economics 27:111–30.

Arnott

De Palma

Lindsey

1993b. “A Structural Model of Peak-Period Congestion: A Traffic Bottleneck with Elastic Demand.” American Economic Review 83:161–79.

Bard

1997. “Transit and Carpool Commuting and Household Vehicle Trip Making.” Transportation Research Record: Journal of the Transportation Research Board 1598:25–31. (Paper No. 97657).

Ben-Akiva

Atheron

T. J.

1977. “Methodology for Short-range Travel Demand Predictions: Analysis of Carpooling Incentives.” Journal of Transport Economics and Policy 11:224–61.

Bento

A. M.

Kaffine

Roth

Zaragoza

The Effects of Regulation in the Presence of Multiple Unpriced Externalities: Evidence from the Transportation Sector. In press.

Boriboonsomsin

Barth

2008. “Impacts of Freeway High-occupancy Vehicle Lane Configuration on Vehicle Emissions.” Transportation Research Part D: Transport and Environment 13:112–25.

Braid

1996. “Peak-load Pricing of a Transportation Route with an Unpriced Substitute.” Journal of Urban Economics 40:179–97.

Brownstone

Golob

1992. “The Effectiveness of Ridesharing Incentives: Discrete-choice Models of Commuting in Southern California.” Regional Science and Urban Economics 22:5–24.

Bureau

Glachant

2008. “Distributional Effects of Road Pricing: Assessment of Nine Scenarios for Paris.” Transportation Research Part A 42:994–1007.

10.

Burris

M. W.

Hannay

R. L.

2003. “Equity Analysis of the Houston Quickride Project.” Transportation Research Record: Journal of the Transportation Research Board 1859:87–92.

11.

Burris

M. W.

Lipnicky

2009. “HOV Lanes or General Purpose Lanes?” Public Works Management and Policy 14:130–47.

12.

Cambridge Systematics. 2002. “Twin Cities HOV Study.” Prepared for Minnesota Department of Transportation. http://www.dot.state.mn.us/information/hov/report.htm. Accessed June 8, 2012.

13.

Casey

2000. “HOV Performance Monitoring: Two Reports, Multiple Conclusions.” In 10th International Conference on High-Occupancy Vehicle Systems, pp. 145–168.

14.

Cervero

2002. “Induced Travel Demand: Research Design, Empirical Evidence, and Normative Policies.” Journal of Planning Literature 17:3–20.

15.

Chang

Wiegmann

Smith

Bilotto

2008. A Review of HOV Lane Performance and Policy Options in the United States. Final Report to US DOT FHWA.

16.

Daganzo

C. F.

Cassidy

M. J.

2008. “Effects of High Occupancy Vehicle Lanes on Freeway Congestion.” Transportation Research Part B: Methodological 42:861–72.

17.

Dahlgren

1998. “High Occupancy Vehicle Lanes: Not Always More Effective than General Purpose Lanes.” Transportation Research A 32:99–114.

18.

de Palma

Kilani

Lindsey

2008. “The Merits of Separating Cars and Trucks.” Journal of Urban Economics 64:340–61.

19.

Duranton

Turner

2011. “The Fundamental Law of Road Congestion: Evidence from U.S. Cities.” American Economic Review 101:2616–52.

20.

Eliasson

Mattsson

L. G.

2006. “Equity Effects of Congestion Pricing: Quantitative Methodology and a Cast Study for Stockholm.” Transportation Research Part A 40:602–20.

21.

Ferguson

1997. “The Rise and Fall of the American Carpool: 1970-1990.” Transportation 24:349–76.

22.

Ganos

2011. “Diamond Lanes: Another Failed Energy/Environmental Policy.” Forbes.com, July 28. http://www.forbes.com/sites/toddganos/2011/07/18/diamond-lanes-another-failed-energy-environmental-policy/. Accessed February 7, 2012.

23.

Houde

Safirova

Harrington

2007. “Washington START Transportation Model.” Unpublished, Discussion Paper 07–43.

24.

Huang

H. J.

Yang

Bell

M. G.

2000. “The Models and Economics of Carpools.” Annals of Regional Science 34:55–68.

25.

Johnston

Ceerla

1996. “The Effects of High-occupancy Vehicle Lanes on Travel and Emissions.” Transportation Research A 30:35–50.

26.

Jou

R. C.

Went

M. C.

Chen

C. C.

2005. “The Evaluation of High Occupancy Vehicle Lanes on the Sun Yat-Sen Freeway in Taiwan.” Journal of Advanced Transportation 39:169–92.

27.

Kessler

Schroeer

1995. “Meeting Mobility and Air Quality Goals: Strategies that work.” Transportation 22(3): 241–272.

28.

Konishi

Mun

2010. “Carpooling and Congestion Pricing: HOV and HOT Lanes.” Regional Science and Urban Economics 40:173–86.

29.

Kwon

Varaiya

2008. “Effectiveness of California’s High Occupancy Vehicle (HOV) System.” Transportation Research Part C 16:98–115.

30.

Lee

L. W.

1984. “The Economics of Carpools.” Economic Inquiry 22:128–35.

31.

Legislative Analyst’s Office. 2000. “HOV lanes in California: Are They Achieving Their Goals?” http://www.lao.ca.gov/2000/010700_hov/010700_hov_lanes.html. Accessed June 8, 2012.

32.

Leman

C. K.

Schiller

P. L.

Pauly

1994. Re-Thinking HOV—High Occupancy Vehicle Facilities and the Public Interest. Report to the Chesapeake Bay Foundation. http://ntl.bts.gov/DOCS/retk.html. Accessed February 26, 2011.

33.

Levine

Wachs

1998. “Factors Affecting Vehicle Occupancy Measurement.” Transportation Research A 32:215–29.

34.

Levinson

2010. “Equity Effects of Road Pricing: A Review.” Transport Reviews 30:33–57.

35.

Los Angeles County Metropolitan Transportation Authority. 2011. “Federal Funds Boost $1 Billion I-405 Carpool Lane Project.” Accessed October 4, 2011. http://www.metro.net/news/works/volume-one/vol1-issue1/federal-funds-boost-1-billion-i-405-carpool-lane-p/

36.

Mallinckrodt

2003. “The Best Effectiveness on HOV Lane Effectiveness.” Unpublished. http://www.urbantransport.org/hov.html. Accessed October 4, 2011.

37.

Mannering

Hamed

1990. “Commuter Welfare Approach to High Occupancy Vehicle Lane Evaluation: An Exploratory Analysis.” Transportation Research A 24:371–79.

38.

Martin

Perrin

Lambert

2004. Evaluation of the Effectiveness of High Occupancy Vehicle Lanes. Report to Utah Department of Transportation.

39.

McFadden

Conditional Logit Analysis of Qualitative Choice Behavior. In: Zarembka

ed. Frontiers in Econometrics. New York, NY: Academic Pres; 1974: 105–142.

40.

McNally

2005. The Four Step Model. In: Hensher

Button

, eds. Handbook of Transport Modeling, edited by. Amsterdam, The Netherlands: Pergamon.

41.

McNally

Recker

1986. On the Formation of Household Travel/Activity Patterns: A Simulation Approach. Final Report to USDoT, University of California.

42.

Naga

2007. “A Mathematical Model for Evaluating the Conversion of High Occupancy Vehicle Lane to High Occupancy/Toll Lane.” Unpublished, M.S. Thesis, University of California, Davis.

43.

Noland

R. B.

Cowart

W. A.

Fulton

L. M.

2006. “Travel Demand Policies for Saving Oil during a Supply Emergency.” Energy Policy 34:2994–3005.

44.

Orski

C. K.

2001. “Carpool Lanes—An Idea Whose Time Has Come and Gone.” TR News 214:24–26.

45.

PB Study Team. 2002. HOV Performance Program. Report to Los Angeles County Metropolitan Transportation Authority prepared by Parsons Brinckerhoff Quade & Douglas, Inc., Kaku Associates, Inc., Texas Transportation Institute, Strategic Consulting & Research and HS Public Affairs.

46.

Plotz

Konduri

Pendyala

2010. “To What Extent Can High-Occupancy Vehicle Lanes Reduce Vehicle Trips and Congestion?” Transportation Research Record: Journal of the Transportation Research Board 2178:170–76.

47.

Poole

Orski

1999. “HOT Lanes: A Better Way to Attack Highway Congestion.” Regulation Accessed February 26, 2010. 23. http://www.cato.org/pubs/regulation/regv23n1/poole.pdf.

48.

Ramjerdi

2006. “Equity Measures and Their Performance in Transportation.” Transportation Research Record: Journal of the Transportation Research Board 1983:67–74.

49.

Rasmussen

2001. Games and Information. Blackwell Publishers.

50.

Rodier

Johnston

R. A.

1997. “Travel, Emissions, and Welfare Effects of Travel Demand Management Measures.” Transportation Research Record: Journal of the Transportation Research Board 1598:18–24.

51.

Rouwendal

Verhoef

E. T.

2006. “Basic Economic Principles of Road Pricing: From Theory to Applications.” Transport Policy 13:106–14.

52.

Safirova

Gillingham

Harrington

Nelson

2003. “Are HOT Lanes a Hot Deal? The Potential Consequences of Converting HOV to HOT Lanes in Northern Virginia.” Unpublished, Urban Complexities Issue Brief 03-03.

53.

Sakano

Benjamin

Ben-Akiva

2001. Evaluating Effects of Travel Demand Management in a Medium-Sized Urban Area. Report to the Urban Transportation Institute.

54.

Schofer

Czepiel

2000. “Success Factors and Decision Issues for High- Occupancy Vehicle Facilities.” Transportation Research Record: Journal of the Transportation Research Board 1711:13–22.

55.

Shewmake

Jarvis

2012. “Hybrid Cars and HOV Lanes.” Working Paper. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1826382. Accessed June 8, 2012.

56.

Small

Verhoef

2007. The Economics of Urban Transportation. New York, NY: Routledge.

57.

Small

Winston

Yan

2006. “Differentiated Road Pricing, Express Lanes and Carpools: Exploiting Heterogeneous Preferences in Policy Design.” Unpublished Working Paper 06-06-02, AEI–Brookings Joint Center for Regulatory Studies.

58.

Smith

2011. “Is the 405 Project Really Worth the Hassle?” Los Angeles DailyNews. July 15. http://www.dailynews.com/politics/ci_18488886. Accessed April, 1, 2012.

59.

Teal

1987. “Carpooling: Who, How and Why.” Transportation Research Part A 21:203–14.

60.

Ulberg

Jacobson

1988. “Evaluation of the Cost-Effectiveness of HOV Lanes.” Transportation Research Record: Journal of the Transportation Research Board 1181:13–24.

61.

Verhoef

Nijkamp

Rietveld

1996. “Second-Best Congestion Pricing: The Case of an Untolled Alternative.” Journal of Urban Economics 40:279–302.

62.

Vickrey

1969. “Congestion Theory and Transport Investment.” American Economic Review 59:251–60.

63.

Wachs

1991. “Policy Implications of Recent Behavioral Research in Transportation Demand Management.” Journal of Planning Literature 5:333–41.

64.

Weinstein

Sciara

G. C.

2006. “Unraveling Equity in HOT Lane Planning: A View from Practice.” Journal of Planning Education and Research 26:174–84.

65.

Wellander

Leotta

2001. “Gauging the Effectiveness of High-Occupancy Vehicle Lanes; Applying Three Criteria to Available Data Reveals Benefits, Viability.” TR News 214:12–19.

66.

Yang

1998. “When Should Carpool Lanes be Introduced in a Multi-lane Highway?” Journal of Advanced Transportation 32:242–52.

67.

Yang

Huang

H. J.

1999. “Carpooling and Congestion Pricing in a Multi- lane Highway with High Occupancy Vehicle Lanes.” Transportation Research A 33:139–55.