Transportation Emissions Response to Carbon Pricing Programs
Craig Raborn
Duke University September 2009
CCPP 09-05
Nicholas School of the Environment
Nicholas Institute for Environmental Policy Solutions Center on Global Change
Response to Transportation Emissions
g Programs Carbon Pricin
BACKGROUND PAPER
Craig Raborn
Climate Change Policy Partnership Duke University
September 2009
Contents
Executive Summary ... 4
Background ... 7
Model Description and Data ... 10
Model outputs ... 11
Primary data ... 12
Gasoline price effect estimates ... 13
Energy Information Administration ... 14
Environmental Protection Agency ... 14
Nicholas Institute/RTI International ... 14
Price elasticity of demand for travel (VMT) ... 14
Carbon Pricing Scenario Model Results ... 17
Gasoline price changes are minor ... 17
Vehicle miles traveled changes only minimally ... 19
Emissions only slightly lower than baseline ... 20
Conclusion ... 23
Appendix A: Complementary Transportation Policies ... 24
Improved vehicle fuel economy ... 24
Tailpipe GHG emissions regulations ... 25
Surface transportation reauthorization ... 25
Transportation‐specific carbon tax or increased gas tax ... 26
Low‐carbon fuel standard ... 26
Feebates ... 26
Pay‐as‐you‐drive insurance ... 27
Appendix B: Technical Model Details ... 28
Model description ... 28
Methodology summary ... 28
Gasoline prices in the model ... 29
Vehicle miles traveled ... 29
Emissions calculations ... 30
CCPP‐STEM Model Equations (Model v 1.1) ... 30
Price of gasoline for year y ... 30
Vehicle miles traveled for year y ... 30
Vehicle fleet fuel economy ... 31
Fuel consumption ... 31
Carbon intensity ... 31
Carbon dioxide emissions ... 32
Bibliography ... 33
Executive Summary
Transportation accounts for approximately one‐third of greenhouse gas (GHG) emissions in the United States, and how it is treated in federal climate change policy will have a significant impact on the nation’s efforts to reduce GHG emissions. Most leading proposals for U.S. climate change policy would implement an economy‐wide cap‐and‐trade program that restricts the amount of GHG emissions and allows trading of permits between emitters, allowing them to either reduce emissions or purchase permits. Because the number of permits available each year declines, the price of carbon is expected to increase. The result is a market force effect in which the increasing price on carbon encourages users to find various ways to reduce emissions and avoid paying these costs. An economy‐wide cap‐and‐trade program allows the market to find the lowest‐cost and most efficient emissions reductions. An inclusive cap‐and‐trade program that covers all sectors of the economy ensures that the cap can be met with reductions from any potential cost‐effective source. Cap‐and‐trade is not intended to result in an equal distribution of efforts across sectors. Because emissions from different sectors have different costs of reduction, sectors will reduce emissions at different rates. As the transportation sector is a significant source of emissions, understanding how a cap‐and‐trade program affects this sector can help
policymakers design optimal policies.
Cars and trucks, called “light‐duty vehicles” (LDVs), account for more than half of GHG emissions from the transportation sector and about 18% of total U.S. GHG emissions. Because LDVs make a significant contribution to U.S. GHG emissions, understanding how drivers will respond to cap‐and‐trade‐induced changes in gasoline prices will help policymakers incorporate transportation into a climate change policy. Transportation policymakers can also benefit from better understanding the potential effects on travel demand from climate change policy; changes in travel demand (vehicle miles traveled) because of higher gasoline prices might indicate different long‐term transportation spending needs. If consumers’
demand for travel changes little in response to small increases in gasoline prices stemming from a price on carbon, the effectiveness of carbon pricing for reducing GHG emissions from LDVs is likely limited.
A new model developed by Duke University’s Climate Change Policy Partnership estimates the effects of a cap‐and‐trade‐based price of carbon on gasoline prices, overall vehicle miles traveled, and CO2
emissions from LDVs. For this paper, the CCPP model builds a carbon price scenario based on estimates from a range of sources of the extensively modeled Lieberman‐Warner Climate Security Act of 2008 (S.
2191 in the 110th Congress).1 The scenario uses carbon price–induced increases in gasoline prices of 6%
to 9% in 2015 and 10% to 17% in 2030. The carbon price scenario in this paper is based on previous estimates and analyses, and is not a modeled prediction of the carbon prices that would result under any specific cap‐and‐trade system. The carbon price scenario in this model covers the period to 2030;
1 Although more recent climate change policies have been proposed (notably the American Clean Energy and Security Act, H.R.
2454), few reliable estimates of the carbon price from these proposals were available at the time this paper was written.
Extensive modeling results from the Lieberman‐Warner bill were available and are used for this paper. This selection was made to allow the CCPP model to develop a carbon price scenario based on a wide range of potential price effects, and is not an endorsement of any policy proposal.
this is an important consideration because the cap‐and‐trade program extends to 2050. Some of the emissions‐reducing effects of an economy‐wide cap‐and‐trade system on the transportation sector may not be achieved until after 2030 because other sectors can initially reduce emissions at lower costs. By 2030 and beyond, however, the lowest‐cost reductions should be achieved, leaving the large amount of relatively more expensive transportation emissions as the primary remaining source for emissions reductions.
The overall result of the CCPP model and the carbon price scenario is that by 2030 the transportation sector does not significantly reduce travel demand and resulting emissions in response to the cap‐and‐
trade program simulated by the carbon price scenario. This background paper also provides policymakers and stakeholders the following specific key results:
• Gasoline prices: Gasoline prices until 2030 are not dramatically affected by the scenario’s carbon price. The likely price changes under the carbon price scenario would not exceed the regular price fluctuations most consumers frequently see under current conditions. The expected gasoline price in 2030 is $4.40, while the expected price without cap‐and‐trade is
$3.90. The expected gasoline price difference of 13.4% in 2030 is only slightly outside the range of historic price variations, and is not unusual; from 1993 to 2008, year‐to‐year price variations exceeded the expected carbon price effect (13.4%) six times.
• Vehicle miles traveled (VMT): Under the carbon pricing scenario, VMT will continue to increase until 2030, but at a lower rate than the Energy Information Administration’s (EIA) projected baseline trends. The expected VMT in 2030 under carbon pricing is 3.54 trillion miles while the EIA reference case predicts VMT will reach 3.81 trillion miles under business‐as‐usual conditions, a difference of 7.0%. The expected annual growth rate in VMT with the carbon pricing scenario decreases from 1.63% (EIA baseline) to 1.27%. Recent research and estimates have suggested that future VMT will be lower than the EIA baseline; if correct, the scenario’s carbon price‐
induced effect on VMT is merely within the range of expected values. Some decrease in VMT is expected with higher gasoline prices, and this is reflected in the model.
• Carbon dioxide (CO2) emissions: Carbon dioxide emissions from LDVs under the carbon pricing scenario are expected to decrease by 2030.2 The CCPP model estimates LDV emissions in 2010 of 1170.2 million tons3 (Mt) of CO2 and expected emissions in 2030 of 1065.5 MtCO2, a decrease of 8.9% over 20 years (0.45%/year).
• Disproportional CO2 reductions from transportation: The carbon price scenario does not reduce CO2 emissions from LDV transportation at the same proportion as overall reductions. The 3%
expected reduction in CO2 emissions from LDVs is much smaller than expected reductions from other sectors during the early years of a cap‐and‐trade program. The electricity sector is expected to supply 88% of total CO2 reductions, and the industrial sector another 9% (United
2 Reductions from LDV transportation in subsequent years might be larger after other sectors have made more significant reductions in the earlier years of the program.
3 All tons (t) in this report are metric tons (1 metric ton = 1,000 kg = 2,204.62 lbs).
States Energy Information Agency 2008). Most CO2 emissions reductions will initially come from other sectors, and the transportation sector will purchase disproportionately more allowances.
• The probability that a carbon price will result in lower LDV CO2 emissions than business‐as‐usual is about 55%. There is roughly a 50% probability that carbon pricing, such as that represented in the model, will have no measurable effect on LDV transportation emissions.
These results should help policymakers understand the limitations of relying solely on a market‐based carbon pricing program if the policy goal is to achieve immediate GHG emissions reductions from LDV transportation. Some of these limitations are due to barriers or market failures that prevent consumers from responding to higher gasoline prices. These barriers include (but are not limited to) land‐use planning resulting in a built environment that encourages (or effectively mandates) LDV‐only driving, high costs of lower‐carbon vehicles that make them unattainable for many drivers, and lack of access to lower‐carbon modes. If policymakers want to achieve significant GHG emissions reductions from transportation during the early years of a cap‐and‐trade program, they may need to consider
alternative or supplemental policies that address these other barriers and result in emission reductions without relying on price signals. Although not modeled for this paper, over the lifespan of a cap‐and‐
trade program, significant emissions reductions will almost certainly come from the transportation sector, but those reductions will likely occur in the later years of the program (after 2030), once lower‐
cost reductions have been achieved from other sectors.
In addition to a cap‐and‐trade policy, a number of complementary options are available to reduce GHG emissions from the LDV transportation sector. There will be many policy components of the upcoming surface transportation reauthorization—expected in 2009—that have the potential to affect GHG emissions. These components include focusing on new performance standards, the potential expansion of road pricing (including congestion pricing or variable rate pricing), the development and consideration of new or revised funding mechanisms, and improved support for alternative travel modes. Additional options include further development of revised Corporate Average Fuel Economy (CAFE) standards, adoption of tailpipe GHG emissions regulations, a transportation‐specific carbon tax or gas tax increase, low‐carbon fuel standards, feebates, and pay‐as‐you‐drive insurance programs. Implementing these policies in concert with a long‐term program such as cap‐and‐trade would require developing more permanent versions than currently implemented. This background report provides an overview of these policy options; future CCPP research will examine specific policy options (potential benefits, costs, and barriers) in more detail.
Background
Transportation accounts for approximately one‐third of greenhouse gas (GHG) emissions in the United States (United States Environmental Protection Agency 2006), and how it is treated in federal climate change policy will have a significant impact on the nation’s efforts to reduce GHG emissions.4 Most leading proposals for U.S. climate change policy would implement an economy‐wide cap‐and‐trade program that requires participants to submit an allowance for each ton of CO2 emitted. A cap‐and‐trade program restricts the number of allowances and, by extension, emissions. If a participant has fewer allowances than needed to cover its emissions, it can reduce emissions or purchase allowances from another participant who can reduce emissions at a lower cost. In this way, participants find the lowest‐
cost reductions among themselves. In order to achieve emission reduction goals, the number of allowances available each year declines. Fewer allowances mean greater scarcity and higher allowance prices in outer years.
An inclusive cap‐and‐trade program that covers all sectors of the economy ensures that the cap can be met with reductions from any potential cost‐effective source. Excluding sectors hinders the ability of a cap‐and‐trade program to achieve the most efficient reductions. Further, although an entire sector may not provide significant reductions, individual firms within that sector may be able to achieve reductions at costs lower than the prevailing carbon price. Equal distribution of effort and reductions across sectors and firms is not a goal of a cap‐and‐trade program. Over the lifespan of a cap‐and‐trade program
different sectors will reduce emissions at different rates; in early years, sectors that can reduce
emissions at lower costs will provide most reductions, while in later years—after the low‐cost reductions have been achieved—reductions will come from sectors of the economy with higher costs for
reductions. The transportation sector is one of the sectors with higher costs to reduce emissions, so it may not provide many early‐year emissions reductions.
Cars and trucks, called “light‐duty vehicles” (LDVs), account for more than half of GHG emissions from the transportation sector and about 18% of total U.S. GHG emissions (United States Environmental Protection Agency 2006). Because LDVs make a significant contribution to U.S. GHG emissions, understanding how drivers will respond to cap‐and‐trade‐induced changes in gasoline prices will help policymakers incorporate transportation into a climate change policy. If drivers’ travel behavior changes only a small amount in response to increases in gasoline prices stemming from a price on carbon, then the effectiveness of carbon pricing for reducing GHG emissions from LDVs is likely limited.
This background paper identifies the expected effects of a carbon price scenario (based on a cap‐and‐
trade program) on emissions from LDV travel. Doing so requires addressing the uncertainties of the interactions between carbon pricing, gasoline prices, driving behavior, fuel consumption, and carbon emissions. The uncertainty surrounding policy development for climate change is a major challenge for
4 Similarly, the manner is which climate change measures are treated by other transportation policies – specifically the upcoming surface transportation reauthorization – will affect the success of climate change policies. Some of these are discussed in Appendix A, and future CCPP research will examine the climate change impacts of many of these policies.
policymakers. To aid the development and implementation of policy decisions that have the highest potential for achieving desired outcomes, policymakers need accurate quantitative descriptions of the uncertainty in emissions outcomes under various possible policies.
To meet the specific needs of accurately estimating LDV emissions, this background paper presents a new model that analyzes LDV transportation behavior and emissions through 2030 based on gasoline price changes (as induced by carbon pricing), changes to travel behavior, and changes to the carbon content of transportation fuel. It then uses the model to examine the likely effects of an economy‐wide carbon pricing program on CO2 emissions from LDVs to estimate the likelihood that recent policy proposals will attain proportional reductions from LDV emissions. The model does not estimate effects beyond 2030. Although most carbon pricing proposals extend to 2050, most baseline data used for the model do not extend beyond 2030.5 As described earlier, some of the emissions‐reducing effects of an economy‐wide cap‐and‐trade program on the transportation sector may not be achieved until after 2030 because other sectors can reduce emissions at a lower cost. In effect, the CCPP model estimates the effects for the first half of a carbon pricing program. By 2030 and beyond, however, the lowest‐cost reductions should be achieved, leaving the large amount of relatively more expensive transportation emissions as the primary remaining source for emissions reductions.
The CCPP model incorporates the uncertainties that arise from estimates of carbon price effects, regular variations in gasoline prices, and uncertainties about how drivers respond to changes in gasoline prices.
The model conducts a Monte Carlo simulation (i.e., it generates a large number of estimated values based on values repeatedly selected at random from a range of empirical data) of how carbon prices might affect national levels of vehicle miles traveled (VMT). The Monte Carlo method selected for this report is particularly well‐suited for estimating long‐run LDV emissions because it can include historically observed and empirically estimated variations in factors that will affect LDV travel and emissions.
For this report, a generic cap‐and‐trade system (based generally on the Lieberman‐Warner Climate Security Act of 2008, described below) is used as the basis for a carbon pricing scenario. Although the model and carbon price scenario are based on estimates from one policy proposal, because the model includes more factors than just the carbon price, the results should indicate the general trend of emissions and other effects from any program that establishes a similar cap on GHG emissions, or develops comparable prices for carbon allowances.
Previous analyses of cap‐and‐trade legislation highlight the need for further examination of how carbon pricing will affect driving and carbon emissions from cars and trucks. The Environmental Protection Agency’s analysis of the Lieberman‐Warner legislation (S. 2191) noted that transportation “provides a relatively small proportion of CO2 emissions abatement” under the proposal and concluded that “the price signal provided by S. 2191 . . . is not high enough to cause large changes in the demand for transportation or changes in how transportation services are provided” (United States Environmental
5 The CCPP‐STEM model uses a variety of data sources, but a significant amount of baseline data comes from the Energy Information Agency’s Annual Energy Outlook (AEO), which projects to 2030.
Protection Agency 2008). The Congressional Budget Office (CBO), in examining expected effects of fuel prices on travel speeds and vehicle purchasing decisions, concluded that gasoline prices would not significantly change these behaviors, and identified low elasticity and “the much smaller effect [carbon pricing] would have on gasoline prices relative to the recent increase in those prices” as a main reason for the small expected effects (Congressional Budget Office 2008).
Press reports from summer and fall 2008 closely linked reduced driving (i.e., VMT) with higher gasoline prices (Kraus 2008), but in reality the relationship is much more complex. Nationally, VMT actually results from a number of factors, including economic measures, such as gross domestic product and employment rates; demographic factors, such as household size and cars per household; geographic factors, such as housing and population density; and variable cost measures, such as gasoline prices (see, e.g., Polzin 2006 and Memmot 2007). These different factors help explain recent VMT trends: prior to recent gasoline price fluctuations and economic conditions, demographic changes had slightly
reduced the annual rate of growth in VMT.6 Then, as increasing gasoline prices in 2008 dramatically increased the cost of driving, VMT actually began to decline. The dramatically lower gasoline prices that followed in late 2008 did not result in similar increases in VMT, however. Deteriorating economic conditions counteracted these effects: the expected increased VMT from lower gasoline prices was probably negated by higher unemployment resulting in fewer people driving. Determining the precise changes to driving that result from these recent price changes is impractical until the long‐term patterns have become more established, but the CCPP model developed for this background paper indirectly incorporates these effects.
6 These demographic changes include a large number of factors, such as a decrease in the average number of cars per
household, fewer young drivers as a result of restrictions for new licenses, etc. Perhaps the most important demographic factor was that—until the late 1990s—large percentages of households switched from single‐worker to having two workers. But in recent years, the number of two‐worker households appears to be reaching its maximum potential. Because fewer households can add a new worker, fewer households are adding new drivers (see, e.g., Polzin, 2006 and Memmot, 2007).
Model Description and Data
The CCPP Stochastic Transportation Emissions Model (CCPP‐STEM) generates an estimated range of emissions reductions from various policies by conducting a Monte Carlo simulation of the likely effects of carbon price increases on gasoline prices, driving behavior, and subsequent emissions from 2010 to 2030. A Monte Carlo process uses repeated random sampling to generate the model’s outcome. The CCPP model repeats the modeling cycle 10,000 times7; these multiple simulations allow CCPP‐STEM to reflect real‐world variations and present a realistic range of potential outcomes.8 By comparison, a deterministic model calculates and presents a single outcome that misleadingly can appear to be the only possible outcome. Monte Carlo methods have been used in a wide range of settings, and regularly deliver results that turn out to be more accurate than those predicted by deterministic models or expert predictions. (A detailed description of CCPP‐STEM is provided in Appendix B.)
One effective method to estimate the effect of gasoline prices on VMT (and therefore carbon emissions) is to use the gasoline “price elasticity of demand” for travel. Simply put, this measure describes how gasoline price changes result in changes in VMT. As described above, many factors influence VMT. CCPP‐
STEM addresses this issue by looking at more than one estimate of gasoline price elasticity of demand.
Empirically based estimates of VMT elasticity9 have varied widely over the past 30 years, and these differences probably encompass similar variations in the other factors (GDP, employment,
demographics, etc.) described above (see, e.g., Polzin 2006 and Memmot 2007). By including elasticity estimates derived from data ranging back more than 30 years, CCPP’s model in effect includes these other VMT‐influencing factors that were present during the time periods when each of these studies were conducted.
Each simulation in the model draws randomly selected values from a range of values established from existing literature and analysis for gasoline price variations, modeled estimates of carbon price effects, and elasticity of VMT based on gasoline price changes. These values are applied to existing estimates of baseline gasoline prices from 2010 to 2030, on‐the‐road fuel efficiency of the existing and future fleet of U.S. cars and trucks, and baseline trends in VMT from 2010 to 2030. The model allows scenarios to include changes in CAFE standards, overall carbon intensity of gasoline, or generalized reductions in VMT. The model cannot reliably examine scenarios that include multiple policy options and does not include a feedback process for how changes to gasoline prices or VMT might affect each other or drive demand for new technologies. These effects, however, are addressed in the model by using the Monte Carlo process to randomly vary factors that would usually generate feedback effects.
The model generates estimated values for each year (2010–2030) for the following measures:
7 There is no absolute rule for the number of simulations that should be conducted, but 10,000 is commonly used. 10,000 simulations are also sufficient to make results subject to the law of large numbers, so that they can be statistically analyzed as if based on real‐world results.
8 Details on stochastic models and processes can be found in Papoulis and Pillai, 2001; Hayek, 2009; and Malliaris and Brock, 1982. Details about Monte Carlo methods can be found in Fishman, 1995 and Robert and Casella, 2004.
9 A list and description of these studies is provided in the “Price elasticity of demand for travel (VMT)” section later in this paper.
• Gasoline price
• Vehicle miles traveled (Light‐Duty Vehicles)
• CO2 emissions (from LDVs)
CCPP‐STEM can adjust the following effects to evaluate different policy scenarios:
• Gasoline price changes
• VMT changes
• Vehicle fuel efficiency
• Changes to the carbon intensity of fuel
For the purposes of this initial paper, only changes to gasoline prices based on a proposed carbon pricing program are modeled. No changes to VMT trends, vehicle fleet fuel efficiency, or carbon intensity of fuel are modeled. Results of modeling variations in these policies will be presented in subsequent CCPP research.
Model outputs
There are two primary comparative measures from the CCPP’s stochastic model. The first set of useful measures from the CCPP model is “expected values.” The model’s large number of simulated results can be treated as a representation of real outcomes and analyzed as if they were real. Results from the model can be averaged to develop an expected value that is the mean of all simulations. This value can then be compared to other estimates (from other analyses, or other scenarios) of the same measure.
For example, the expected value of gasoline prices in 2020 from the model is $3.89 per gallon. This expected value can be compared to other estimates, such as EIA’s baseline, to suggest how much prices will be different under the model’s assumed carbon pricing. The CCPP model also estimates a range of potential values, as well as the highest and lowest simulated values. To reduce extreme variations, the CCPP model uses results within 2 standard deviations (2SD) from the mean, which encompasses 95% of all results, to estimate the range of values for any measure.
Second, the model generates “success rates,” which are the percentages of simulations that attain certain targets. For example, if a policy seeks to reduce emissions by 400 MtCO2, and 850 of the model’s 10,000 simulations result in emissions reductions of that scale or larger, the model has an 8.5% success rate. To simplify understanding these results, the success rate for any given policy can be treated as the probability that the policy will achieve whatever outcome is being compared.10
10 By generating a large number of simulated results based on the range of real world observations for model inputs, a Monte Carlo process basically mimics the distribution of results that would be expected in the real world. With a sufficiently large number of simulations, the results can be treated like a probability distribution.
Primary data
The model starts each of its simulations with the following data sources and values:
• The carbon price scenario’s effects on gasoline price are supplied from three analyses of the Lieberman‐Warner Climate Security Act of 2008 (S. 2191) (EPA 2008b; EIA 2008; Murray and Ross 2007). These price effects can be replaced with updated values as analyses of current policy proposals become available. These analyses and their probability of inclusion in the model are described below.
• Baseline gasoline prices are taken from the EIA AEO 2009 (Early Release) report (EIA 2009a).
These start at $2.721 in 2010 and increase to $3.896 in 2030 (2007$ values).
• Annual gasoline price variations are based on historical EIA data from 1993 to 2008 (EIA 2009c).
During that time, gasoline prices varied an average of 7.7%, with a standard deviation of 11.1%.
• Baseline vehicle miles traveled are also from EIA’s AEO 2009 (Early Release) report (EIA 2009a).
Starting annual VMT in 2010 is estimated at 2.752 trillion miles, increasing to 3.807 trillion miles in 2030.
• Gasoline price elasticity effects are supplied from four different studies of price elasticity of demand for gasoline conducted from the late 1970s through 2006 (Small and Van Dender 2007;
Goodwin et al. 2004; Graham and Glaister 2004; Schimek 1996). These studies are described below.
• Initial vehicle fleet fuel efficiency is taken from values reported by the National Transportation Safety Board (NHTSA) in 2008 (Yacobucci and Bamberger 2009), and adjusted to reflect actual on‐the‐road mileage (e.g., Mintz 1993, EIA 2009b, and EIA 2007). The data include different mileage rates for cars and small trucks. These values include vehicle fleet data going back 14 years. The initial fleet fuel economy in 2010 used in the model is 20.7 miles per gallon (mpg).
• Long‐term fleet fuel efficiency values are taken from NHTSA’s regulations implementing the 2007 Energy Independence and Security Act (EISA).11 These values establish a pathway to average fleet fuel efficiency for new cars of 31.6 mpg by 2015, although actual on‐road mpg for all LDV vehicles for that year is estimated at 21.8 mpg. EISA continues only through 2020, and the model assumes that fuel efficiency requirements will continue to increase at the EISA rate each year until 2030.
• Carbon content of fuel is assumed to be 19.4 pounds of CO2 emissions per gallon of gasoline (EPA 2006).
11 EISA specifically requires separate average fuel economy standards for passenger cars and for light trucks manufactured in each model year beginning with model year 2011 “to achieve a combined fuel economy average for model year 2020 of at least 35 miles per gallon for the total fleet of passenger and non‐passenger automobiles manufactured for sale in the United States for that model year.” 49 U.S.C.A. 32902(b)(1), 32902(b)(2)(A)
Gasoline price effect estimates
CCPP‐STEM builds a carbon price scenario based on results from external evaluations and analyses of climate policy proposals. The carbon price scenario is used to estimate the likely gasoline price effect from a carbon pricing program. At the time this paper was prepared, the most extensively modeled climate policy proposal was the Lieberman‐Warner Climate Security Act of 2008 (S.2191 in the 110th Congress). The Lieberman‐Warner bill would have placed declining GHG emissions caps upstream on petroleum and natural gas production, and downstream on coal facilities (Parker and Yacobucci 2008). It also would have established a market‐driven system for auctioning and trading emission allowances.
The CCPP selected S.2191 because it was more extensively modeled and analyzed than other legislative proposals; three public agencies or independent organization modeling projects for S.2191 (each discussed below) are used in CCPP‐STEM. The CCPP model includes results from the core policy scenario from each of the three models selected for this analysis. Because the purpose of this paper is not to evaluate the feasibility of these different scenarios, but to identify the range of likely and potential effects on gasoline prices from carbon pricing, the specific assumptions for these scenarios are not analyzed or evaluated.
In order to address long‐term uncertainty of modeling results, the CCPP‐STEM randomly selects estimated gasoline price effects from an empirical distribution, shown in Table 1. This allows the CCPP model to include price effects from a range of viable models. The probability of selection (column one) indicates how frequently the Monte Carlo process conducted by CCPP‐STEM will randomly select the values from that model. For example, the EPA’s results should be selected and used in approximately 3,500 of the model’s scenarios. The CCPP constructed this probability distribution using an approach similar to Webster et al. (2002), based on reviewing the methodologies for each report and gathering informal expert feedback on the models.
Table 1. Empirical distribution of carbon pricing effects on gasoline.
Gasoline Price Percent Change
Selection
Probability Source 2015 2020 2025 2030
35% Energy Information Agency 0.0577 0.0938 0.1081 0.1697 35% Environmental Protection Agency 0.0902 0.1115 0.1250 0.1561 30% Duke Nicholas Institute/RTI 0.0606 0.0743 0.0680 0.0909
Energy Information Administration
The Energy Information Administration modeled S.2191 using the National Energy Modeling System (NEMS), which is used to develop projections in the Annual Energy Outlook. NEMS models all sectors of the economy through 2030, particularly looking at the energy sector and energy‐related economic impacts. The software projects emissions of energy‐related carbon dioxide emissions resulting from the combustion of fossil fuels, representing about 84% of current total U.S. GHG emissions. In real dollar terms, EIA’s modeling estimated a gasoline price increase in 2030 of $0.43 per gallon in the core case (2007$) (EIA 2008).
Environmental Protection Agency
The Environmental Protection Agency modeled S.2191 using the Applied Dynamic Analysis of the Global Economy (ADAGE) model, which is a dynamic computable general equilibrium model capable of
examining many types of economic, energy, environmental, and climate change policies at the
international, national, and regional level. EPA’s analysis looked at both the national and regional results of S.2191, with a reference scenario and nine alternative scenarios. EPA’s model estimated a gasoline price increase of $0.56 per gallon in 2030 with the reference scenario (2007$) (EPA 2008).
Nicholas Institute/RTI International
The Nicholas Institute for Environmental Policy Solutions at Duke University and RTI International conducted a preliminary assessment of the potential impacts of S.2191, also using the ADAGE model (Murray and Ross 2007). NI‐RTI’s analysis looked only at national‐level results. This analysis developed one baseline scenario, a reference scenario, and two additional policy alternatives. The NI‐RTI ADAGE model estimated a gasoline price increase in 2030 of $0.20 per gallon in the core case (2007$).
Price elasticity of demand for travel (VMT)
As described earlier, the gasoline price elasticity of demand for travel is the change in travel that results from changes in gasoline prices. Elasticity is the change that occurs for every 10% increase in gasoline prices. For example, an elasticity of ‐0.10 means that if gasoline prices increase by 10%, VMT will decrease by 1%. In addition, there are short‐term (roughly one year) and long‐term (up to five years) elasticities. The CCPP reviewed results from five studies of VMT elasticity response to price to identify the empirical high and low range for both short‐run and long‐run elasticities of price demand for gasoline.
Small and Van Dender (2007) compared VMT elasticity for 1966 to 2001 to elasticity for only 1997 to 2001, essentially providing two sets of elasticities in one paper. In addition to documenting a recent decrease in elasticity they calculated a short‐run elasticity (for the 1997 to 2001 period) of ‐0.022 and a long‐run elasticity of ‐0.107.
Goodwin et al. (2004) and Graham and Glaister (2004) were both part of a parallel “blind” meta‐analysis review of 175 elasticity studies. Using slightly different methodologies, the results were broadly
consistent: Goodwin et al. estimated a short‐run elasticity of ‐0.10 and long‐run elasticity of ‐0.29, and Graham and Glaister calculated a short‐run elasticity of ‐0.15 and long‐run elasticity of ‐0.31. (Because Goodwin et al. and Graham and Glaister used overlapping sets of data, their combined selection probability was reduced to 40%.)
Schimek (1996) used time series data for the United States from 1950 to 1994 and pooled state‐level data from 1988 to 1992 to estimate a short‐run elasticity of ‐0.08 and a long‐run elasticity of ‐0.21.
The CCPP model randomly selects a paired short‐ and long‐run estimated gasoline price elasticity of demand values from an empirical distribution, shown in Table 2. As with the selection of gasoline price effects, a random selection of elasticity allows the CCPP‐STEM model to include effects from a range of research results.12 Because the range of elasticities from previous research is so large, there was no established methodology for determining the probability distribution for the CCPP‐STEM model’s random selection procedure. The CCPP therefore constructed this probability distribution (again using the Webster et al. [2002] approach) based on careful review of the methodologies for each study, as well as on expert feedback and peer reviews of these studies. The distribution was also constructed to ensure that elasticities cover a long time period, therefore reflecting the underlying economic, social, and demographic conditions of different time periods.
Table 2. Empirical distribution of price elasticity for travel demand.
Selection
Probability Source Data Years Short‐Run Long‐Run
25% Goodwin et al. 1974–1998 ‐0.10 ‐0.29
25% Small and Van Dender (1) 1997–2001 ‐0.02 ‐0.11
20% Schimek 1950–1994 ‐0.08 ‐0.21
15% Graham and Glaister 1974–1998 ‐0.15 ‐0.31
15% Small and Van Dender (2) 1966–2001 ‐0.04 ‐0.22
12 Two basic ways for selecting values can by used in a time series Monte Carlo process. First, a value can be selected and used for every year of each simulation. Second, a new value can be selected for each year of each simulation. Because the CCPP model does not have a feedback process, the second approach is used. This approach is called a “random walks” method, and it provides the model with varied underlying conditions that more realistically capture real world conditions, which is as close as the model can come to incorporating feedback loops. It allows the consumer response to price changes to vary each year, which also then allows the model to indirectly include variations in underlying economic conditions. A random walks approach tends to reduce extreme values, but also increases overall variation.
For the purposes of this paper, the model uses price elasticity of demand for travel, as represented by annual vehicle miles traveled (VMT). Another measure—price elasticity of demand for gasoline—is also available. In general, the price elasticity of demand for gasoline is about twice the price elasticity of demand for travel. For example, if a gasoline price increase were to result in a 1% decrease in VMT, demand for gasoline would likely decrease by 2%. This is because other alternatives are available for travelers to reduce gasoline use, including using more fuel efficient vehicles. Because the CCPP‐STEM model was developed to provide policy insights for policymakers in both climate policy and
transportation policy, it estimates both the reduction in emissions that follows from changes in VMT and the actual changes in VMT that result from gasoline price changes. To do so, it uses the price elasticity of demand for gasoline measures described above.
This CCPP LDV transportation model takes a new approach to estimating emissions from LDV
transportation. By using a stochastic process, it produces results that reflect the real‐world uncertainties that surround how changes to gasoline price (and other policy variations) affect GHG emissions. The model generates two main types of results, the first is a “success rate,” which is the percentage of simulated outcomes that meet or exceed a policy target. The success rate is roughly equivalent to the probability that any given policy will achieve its desired policy target. Second, the CCPP model estimates a range of values that any of the policies in the model are likely to achieve. These two measures allow the model to provide a framework for policy analysis, specifically addressing the uncertainty that is inherent in estimating long‐term outcomes.
Carbon Pricing Scenario Model Results
Using results from the CCPP model, the expected effects of a carbon price scenario based on the Lieberman‐Warner Climate Security Act of 2008 (S.2191 in 110th Congress) are described below. A cap‐
and‐trade program, such as that proposed by the Lieberman‐Warner bill and most current climate change policy proposals, sets a national cap on GHG emissions that decreases over time. Reducing the supply of allowances (representing the equivalent of one ton of CO2) would cause the price of GHG emissions to increase each year. The higher costs, in turn, should encourage entities covered by the legislation to reduce their emissions, primarily through improvements in technology, but also through conservation and shifting to lower‐carbon energy sources. For the transportation sector, however, the effectiveness of using increases in carbon pricing as a means to encourage less carbon use is uncertain.
The CCPP model examines this relationship by estimating the change in gasoline prices from 2010 to 2030 that result from the carbon price scenario, the change in VMT that should be expected from these gasoline price changes, and the emissions that should result from that level of VMT, given expected vehicle fuel efficiency and fuel carbon intensity levels. These results can then be compared to various policy targets to estimate the potential of carbon pricing to meet those targets.
Gasoline price changes are minor
The carbon pricing scenario used in the model will not send a sufficiently strong price signal by 2030 to significantly change driving behavior and reduce GHG emissions from LDVs. The expected gasoline price in 2030 under the carbon price scenario (based on a Lieberman‐Warner‐type cap‐and‐trade system) is
$4.374 (2007$), which is $0.515 higher than the expected baseline price of $3.859. This is an overall upward shift in expected price, but as discussed below, the 13.4% difference is not an unusual variation when compared to historic observations.
Figure 1. Estimated gasoline prices with carbon pricing scenario.
Figure 1 shows the expected prices of gasoline under the carbon price scenario, and a reference case modeled without the carbon price scenario. High and low values, representing 95% of simulated values (two standard deviations), are also shown. Although the mean expected gasoline price is systematically higher with carbon pricing than without, nearly half of the expected prices from each scenario overlap with the other. This suggests that the carbon pricing scenario has a roughly 50% probability of delivering a gasoline price that could reasonably be expected without the carbon price.
Table 3. Expected gasoline prices, 2015–2030.
Reference Estimated Gasoline Price (2007$)
2015 2020 2025 2030
Carbon Price Effect Estimate 3.551 3.890 4.086 4.374
Modeled Baseline Estimate* 3.378 3.569 3.723 3.859
EIA Baseline Projection* 3.462 3.574 3.722 3.896
* The close alignment of CCPP’s stochastically modeled baseline to EIA’s baseline projection (rows 2 and 3) suggests that the CCPP modeling approach is successful at matching real‐world trends.
The carbon pricing scenario’s effect on gasoline prices is slightly less than twice the historically observed (1993 to 2008) year‐to‐year price fluctuations of 7.7%, but only slightly greater than the 11.1% standard
deviation in those regular fluctuations. The 13.4% increase resulting from the carbon price scenario could, therefore, be expected to occur without carbon pricing in nearly 15% of year‐to‐year variations.
(Plus or minus one standard deviation encompasses 68% of all expected results, and only half of the remaining 32% is larger than the standard deviation.) In fact, the observed annual variation in gasoline price for six of the years (40%) from 1993 to 2008 exceeded the probable carbon price effect of 13.4%.
These results strongly suggest that the 13.4% increase over a 20‐year period could easily be absorbed by regular price fluctuations in gasoline prices. Although the expected price change is small, the CCPP model (see figure 1) does show a consistently higher gasoline price with carbon pricing, and that price increase should still have some type of effect on overall vehicle miles traveled.
Vehicle miles traveled changes only minimally
Nearly all research surrounding how driving levels (VMT) respond to gasoline price changes indicates that only a small response should be expected. The CCPP model estimates that—as a result of the small increases in gasoline prices described above—the expected national VMT in 2030 will be 3.539 trillion miles. This is 7.0% lower than the EIA’s baseline projection of 3.807 trillion miles in 2030.
Figure 2. Carbon pricing scenario effect on vehicle miles traveled (VMT).
Figure 2 shows the expected annual VMT with the model’s carbon price scenario. The expected high and low VMT range for 1 standard deviation (encompassing 68% of likely results) and 2 standard deviations (95% of likely results) are shown. The EIA projected VMT is also presented. Values for the model’s estimate and the EIA projection are provided in Table 4.
Table 4. Expected vehicle miles traveled, 2015–2030.
Reference Estimated vehicle miles traveled (billion miles)
2015 2020 2025 2030
Carbon Price Effect Estimate 3,011 3,137 3,352 3,539
Modeled Baseline Estimate 3,032 3,209 3,435 3,648
EIA Baseline Projection 2,877 3,165 3,489 3,807
VMT decreases with the carbon pricing scenario, but only minimally. The EIA baseline rate of growth from 2010 to 2030 averages 1.63% per year; with carbon pricing the rate of growth is 1.27% per year.
The trend is not so clear, however. Until about 2020, the carbon price effect actually exceeds the EIA baseline. This unexpected result is probably due to the effects of large recent variations in gasoline prices that are included in the CCPP model’s initial elasticity calculations (described in the Technical Appendix), but are treated differently by EIA’s process for generating its baseline. Starting in 2020, however, the model generates the expected trend of lower VMT under the carbon pricing scenario.
The CCPP model uses EIA’s baseline for its initial value and growth rate. But recent research from a number of sources (see, e.g., Puentes 2009 and Polzin 2006) has indicated that the rate of growth in VMT has been decreasing over recent years. EIA’s estimated 1.63% annual growth may be the correct trend, but other research has suggested that the actual rate of growth will be closer to 1.1% to 1.4%
(see, e.g., Puentes 2009; AASHTO 2008; and Polzin 2006). The VMT growth rate estimates generated under the carbon price scenario fall within this range of recent research and estimates. If these recent studies are correct, then the carbon price–induced effect on VMT is merely within the range of expected VMT growth. A decrease in VMT—or VMT growth rates—is the expected effect of higher gasoline prices, but the significance of the carbon price scenario’s expected VMT changes is unclear because the range of potential VMT values is very large and other factors—such as fuel efficiency—influence final emission levels. An examination of how the expected VMT results in changes to CO2 emissions indicates how the expected reduction in VMT will help reach specific climate policy goals.
These results may also be informative for policymakers involved in overall transportation policy. The expected rate of growth in VMT has implications for how much and what kind of infrastructure may be needed. Changes in VMT also have implications for how transportation infrastructure might be funded and for what level of funding might be available with different revenue sources.
Emissions only slightly lower than baseline
Expected CO2 emissions with the carbon pricing scenario show increasing LDV transportation emissions until approximately 2014, followed by a gradual decrease until 2030. The model estimates emissions in
2010 of 1,170.2 million metric tons of carbon dioxide (MtCO2), and an expected value in 2030 of 1,065.5 MtCO2, a decrease of 8.9% over 20 years.
Table 5. Expected CO2 emissions from light‐duty vehicles, 2015–2030.
Reference Estimated CO2 Emissions (MtCO2)
2015 2020 2025 2030
Carbon Price Effect Estimate 1187.8 1139.0 1089.0 1065.5
Modeled Baseline Estimate 1222.1 1183.0 1133.9 1115.2
EIA Baseline Projection* ‐‐ ‐‐ ‐‐ ‐‐
*EIA does not generate a specific estimate for LDV‐only CO2 emissions.
The comparison of expected LDV emissions modeled with and without carbon pricing shows the relative expected effectiveness of the two policies. The mean difference between expected emissions is about 3% lower with the carbon pricing scenario, so the CCPP model indicates that carbon pricing would result in slightly lower expected emissions than without carbon pricing.
Although GHG‐reducing policies do not usually establish sector‐specific goals, it is informative to
examine how reductions from transportation compare to other sectors. Analysis of projected reductions from other sectors typically shows much larger expected or projected reductions in CO2 emissions (see, e.g., EIA’s and EPA’s analysis of S.2191 and Paltsev et al. 2008). This indicates that—at least for the years covered by the CCPP‐STEM model (2010 to 2030)—other sectors that can make emissions reductions at lower costs will make proportionally larger emissions reductions than the transportation sector under a carbon pricing program.
Some estimates of the relative emissions reductions from different sectors (from EIA’s analysis of S.
2191) are shown in Table 6, below. This example shows that the transportation sector reduces emissions by 3.8% in 2030, while the electricity sector reduces emissions by 87.9% and the industrial and building sectors reduce emissions by 9% and 9.5%, respectively. These results are generally mirrored in other analyses of carbon pricing proposals.
Table 6. EIA‐estimated CO2 emissions reductions by sector, 2015–2030.
Sector Estimated CO2 Emissions (percent)
2020 2030
Transportation 1.6 3.8
Electricity 25.9 87.9
Industrial 4.7 9.0
Buildings 5.4 9.5
Although outside the scope of this paper, because currently there are few readily available low‐cost alternatives for reducing emissions from transportation, the sector will initially purchase more emissions allowances than other sectors. Over the long run of a carbon pricing program, transportation should provide emissions reductions, but these reductions will probably occur later in the program once the lower‐cost reduction opportunities in other sectors have been achieved.
Additionally, 54.6% of the CCPP model emission simulations with the carbon price scenario have
expected emissions that are lower than the mean baseline emissions without the carbon price scenario, suggesting only a small expectation that carbon pricing will reduce LDV emissions more than taking no carbon pricing action for LDVs. There is roughly a 50% probability, therefore, that a carbon pricing program will have no measurable effect on LDV transportation emissions.
Conclusion
A carbon price resulting from a cap‐and‐trade system will not likely achieve significant GHG emission reductions from passenger vehicles during the early years of the program. The results of a new model developed by the CCPP, using a carbon price scenario based on the proposed Lieberman‐Warner (S2191 in 110th Congress) cap‐and‐trade program, show that the estimated effects on gasoline prices, national vehicle miles traveled (VMT), and resulting CO2 emissions are generally relatively small.
A cap‐and‐trade system works by increasing the cost of emitting GHGs, encouraging the development of new lower‐emitting technologies. The higher costs also increase conservation and support shifting to lower‐carbon energy sources. Because the results of the CCPP model and carbon price scenario suggest that a carbon price will not significantly reduce emissions from LDV transportation, however, this implies that the costs of new transportation technologies are generally greater than the costs of using existing technologies and paying the additional carbon price. The carbon price scenario results also indicate that other potential responses to a carbon price, such as increased conservation (i.e., less overall LDV use) are not likely to occur. These results are consistent with other research (see, e.g., EIA’s and EPA’s analysis of S. 2191 and Paltsev et al. 2008) that finds that the transportation sector currently has few low‐cost opportunities for reducing CO2 emissions.
These results suggest that there are limitations of relying solely on a market‐based carbon pricing program to achieve immediate GHG emissions reductions from LDV transportation. Some of these limitations are due to barriers or market failures that prevent consumers from responding to higher gasoline prices. These include (but are not limited to) land‐use planning resulting in a built environment that encourages (or effectively mandates) LDV‐only driving, high costs of lower‐carbon vehicles that make them unattainable for many drivers, and lack of access to lower‐carbon modes. If policymakers want to achieve significant GHG emissions reductions from transportation during the early years of a cap‐and‐trade program, they may need to consider alternative or supplemental policies that address these other barriers and that decrease emissions without relying on relatively small cap‐and‐trade‐
induced gasoline price changes to reduce LDV travel. Although not modeled for this paper, over the lifespan of a cap‐and‐trade program, significant emissions reductions will almost certainly come from the transportation sector, but those reductions will likely occur in the later years of the program (after 2030), once lower‐cost emissions from other sectors have been achieved.
Appendix A: Complementary Transportation Policies
As described in the main text of this paper, policymakers may want to look for additional opportunities to reduce emissions from the transportation sector. A few prominent alternative transportation policies are presented and briefly discussed below. These can generally be divided into three categories:
• Improving vehicle technologies and fuel efficiency
• Reducing the carbon intensity of transportation fuels
• Reducing overall demand for travel‐derived fuel consumption
The approaches in this appendix are described only briefly; each has potential advantages, as well as important social, political, and technological barriers that need to be more fully investigated. The potential of some of these has already been established by recent policy announcements by the Obama administration, which has taken action to improve fuel economy standards and implement tailpipe emissions standards. The CCPP plans to further explore and model the effects of these and other transportation policies in future analysis.
One important point for policymakers seeking to use these complementary policies in concert with (or as an alternative to) a carbon pricing mechanism is permanence of the alternate policy. Most of these complementary policies are usually adopted for short time frames, such as 5–10 years, and thus do not offer the same long‐term certainty that a cap‐and‐trade program would. To work as a policy that complements long‐term carbon pricing as a means to reduce CO2 emissions, these complementary policies would probably need to be implemented for the same time period, so that interactions between the two policy mechanisms could be better planned. They could also include automatic adjustment mechanisms triggered if transportation emission targets are under‐ or over‐attained.
Improved vehicle fuel economy
If vehicles become more fuel efficient—i.e., use less fuel per mile of travel—the increase in VMT‐derived GHG emissions might be partially or completely offset. The Energy Independence and Security Act of 2007 (EISA) required an increase in minimum Corporate Average Fuel Economy (CAFE) standards to 35 miles per gallon (mpg) in 2020, although the Obama administration recently announced plans to require compliance with that standard by 2016. The current (2009) standard for cars is 27.5 mpg and 22.2 mpg for light trucks. Fuel economy requirements have two basic objectives: to reduce fuel consumption and to encourage improvements in vehicle technologies (Gallagher, Collantes, et al. 2007). Improved fuel economy standards, however, have the effect of reducing drivers’ cost per mile, so they may actually stimulate an increase in driving (Small and Van Dender 2007). Because the VMT effects of different levels of fuel economy standards are not fully understood, CAFE standards alone cannot provide certainty about GHG emissions. However, in combination with a carbon pricing system, fuel economy standards may spur GHG reductions beyond the price signal sent by a cap‐and‐trade program.
Tailpipe GHG emissions regulations
Tailpipe emissions regulations are restrictions on the per‐mile emission of GHGs from individual vehicles. Similar to fuel economy standards, but focused on GHG emissions rather than fuel
consumption, tailpipe standards are expected to accelerate adoption of new vehicle technologies. By regulating all GHG emissions (compared to CAFE standards, which effectively only restrict CO2
emissions), greater mitigation can be achieved, and manufacturers have greater flexibility in how they choose to reduce emissions than they do with fuel economy standards (Gallagher, Collantes et al. 2007).
Although the Obama administration recently acted to establish a tailpipe emissions standard, policymakers could further regulate tailpipe emissions through some type of performance standard, which would also provide greater certainty about GHG emissions than would fuel economy standards.
Surface transportation reauthorization
Congress reauthorizes surface transportation programs of the Federal government every five years. This process determines funding levels for road building, maintenance, transit systems, and nearly every other form of surface transportation. The upcoming reauthorization of surface transportation programs is expected to dramatically increase funding, and will provide a number of opportunities for
policymakers to address climate change from the transportation sector. These include:
• Shifting the focus of new infrastructure and construction to performance standards, including emphasis on congestion relief. Performance standards, which are intended to ensure that transportation projects have defined construction and operational goals, are likely to become more prominent in project selection and review. A climate change performance standard could look at lifecycle GHG emissions from transportation projects and include a requirement that projects reduce emissions over their functional life. This can be accomplished through improved project selection and planning, so that projects that will encourage new travel, without
otherwise accounting for resulting emissions increases, will be less likely to be selected than projects that reduce congestion or offer alternate travel modes.
• Potential expansion of facility or road pricing, including congestion pricing and increased use of tolls. Road pricing is a market‐based mechanism that applies charges for use of the
transportation infrastructure to more efficiently balance the supply and demand of transportation capacity. Road pricing is particularly attractive as a means of overcoming projected revenue shortfalls as reliance on the gas tax continues to result in diminishing revenues. Road pricing is also effective at managing congestion during peak travel periods.
• Development and consideration of new or revised funding distribution mechanisms. Funding currently flows from the federal level to states, with states currently guaranteed a minimum funding amount based on their gas tax revenues. Revised funding mechanisms, tied to performance standards and clearly defined national transportation goals, might allow direct funding for transportation projects to Metropolitan Planning Organizations, as well as states, in a manner that meets national transportation needs more directly than the current approach of