Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

An integrated assessment of climate mitigation policy, air quality and traffic safety for passenger.. Mazzi, Eric 2010

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
24-ubc_2010_spring_mazzi_eric.pdf [ 926.27kB ]
Metadata
JSON: 24-1.0069914.json
JSON-LD: 24-1.0069914-ld.json
RDF/XML (Pretty): 24-1.0069914-rdf.xml
RDF/JSON: 24-1.0069914-rdf.json
Turtle: 24-1.0069914-turtle.txt
N-Triples: 24-1.0069914-rdf-ntriples.txt
Original Record: 24-1.0069914-source.json
Full Text
24-1.0069914-fulltext.txt
Citation
24-1.0069914.ris

Full Text

AN INTEGRATED ASSESSMENT OF CLIMATE MITIGATION POLICY, AIR QUALITY AND TRAFFIC SAFETY FOR PASSENGER CARS IN THE UK  by Eric Mazzi B.S. Mechanical Engineering, California State Polytechnic University, Pomona, 1987 M.S. Mechanical Engineering, University of Southern California, 1990 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Resource Management and Environmental Studies) The University of British Columbia (Vancouver) April, 2010 © Eric Mazzi, 2010  ABSTRACT  Climate change mitigation policies applied to passenger cars can be effective in reducing tailpipe CO2 rates by changing vehicle mass, fuels, and drive-train technology. However, these same factors can lead to changes in vehicle emissions, vehicle safety, and, consequently, changes in health outcomes from air pollution and traffic collisions. These relationships are examined using the UK as a case study where tax regimes based on tailpipe CO2 emission rates have been in place since 2001. Policymakers are tasked to design CO2 policies for passenger cars, but the effectiveness of new policies will depend on how well climate mitigation is balanced with other relevant risks. I examine the rationale and introduce the basic framework for an Integrated Assessment approach to quantitatively assess passenger car CO2 policies. As industrialized countries transition to more heterogeneous fleets with increasing uptake of alternative fuels and technologies, the importance of decision criteria choices, risk metrics, system boundaries, and inclusion of all relevant risks using an Integrated Assessment framework will be increasingly critical. Since 2001, there has been a strong growth in diesel car registrations in the UK. For 2001-2020, I estimate that switching from gasoline to diesel cars reduces CO2 emissions by 0.4 mega-tonnes annually. However, current diesel cars emit higher levels of PM10 and the switch from gasoline to diesel cars is estimated to result in 90 additional deaths annually (range 20-300) from 2001-2020. The UK has also had an increase in registrations of lighter vehicles. The relationship between tailpipe CO2 emission rates, vehicle mass, and traffic safety risks were examined. The two-car “first law” fatality risk ratio for drivers of lighter cars relative to drivers of heavier cars was estimated to be the mass ratio raised to the power 5.3. Independent estimates of driver killed or serious injury risk in two-car collisions were found to be inversely related to vehicle CO2 emission rates. Scenario analyses show that policies combining incentives for lighter cars with a 1,600 kg upper limit for new cars should simultaneously achieve traffic safety and climate mitigation goals more effectively than policies with no upper limit on mass.  ii  TABLE OF CONTENTS  ABSTRACT...................................................................................................................................... ii TABLE OF CONTENTS ..................................................................................................................... iii LIST OF TABLES............................................................................................................................ vii LIST OF FIGURES ........................................................................................................................... ix LIST OF ACRONYMS ..................................................................................................................... xiv GLOSSARY ................................................................................................................................... XV PREFACE ..................................................................................................................................... xix ACKNOWLEDGMENTS .................................................................................................................... xx DEDICATION ................................................................................................................................ xxi CO-AUTHORSHIP STATEMENT ..................................................................................................... xxii 1  INTRODUCTION, LITERATURE REVIEW, OBJECTIVES AND HYPOTHESES ...................................... 1 1.1  INTRODUCTION ................................................................................................................ 1  1.2  RESEARCH OBJECTIVES BASED ON LINKS BETWEEN PASSENGER CAR CO2, AIR QUALITY, AND TRAFFIC SAFETY....................................................................................................... 3  2  1.3  RESEARCH QUESTIONS, HYPOTHESES, AND TASKS ........................................................... 6  1.4  CRITICAL HIGHLIGHTS FROM THE LITERATURE ................................................................... 9  1.4.1  Climate Mitigation Policy: the Importance of Multiple Risk Assessment .................. 9  1.4.2  Air Quality and Ancillary Benefits of Climate Mitigation Policy............................... 12  1.4.3  Vehicle Mass, Traffic Safety, and CO2 Emissions from Passenger Cars............... 13  1.5  HOW THE THESIS CHAPTERS FIT TOGETHER ................................................................... 17  1.6  REFERENCES ................................................................................................................ 18  INTEGRATED ASSESSMENT OF MULTIPLE RISKS TO ASSESS CURRENT AND FUTURE CLIMATE MITIGATION POLICIES FOR PASSENGER CARS .......................................................................... 27  2.1  INTRODUCTION .............................................................................................................. 27  2.2  IDENTIFYING RELEVANT RISKS AND CHOOSING DECISION CRITERIA.................................. 29 iii  2.2.1  Decision Criteria Based on Marginal External Costs ............................................. 30  2.2.2  Decision Criteria Based on Public Health Metrics.................................................. 31  2.2.3  Decision Criteria Based on New Public Policy Priorities ........................................ 32  2.2.4  Choosing the Appropriate Decision Criteria........................................................... 33  2.3 2.3.1  Linking Policies to Risks ........................................................................................ 33  2.3.2  Climate Mitigation Policy Options and Potential Influences on Risks..................... 36  2.3.3  Linking Policies to Risks with Alternative Fuels and Technologies ........................ 41  2.4  3  4  PATHWAYS FROM POLICY TO RISKS ................................................................................ 33  ESTABLISHING CAUSAL LINKS BETWEEN POLICIES AND OUTCOMES.................................. 42  2.4.1  Basic issues in Establishing Causal Links Between Policy and Outcomes............ 42  2.4.2  What Caused the Rapid Rise in Diesel Cars in the UK?........................................ 43  2.5  CONCLUDING REMARKS ................................................................................................. 45  2.6  ACKNOWLEDGEMENTS ................................................................................................... 45  2.7  REFERENCES ................................................................................................................ 46  AIR QUALITY IMPACTS OF CLIMATE MITIGATION: UK POLICY AND PASSENGER VEHICLE CHOICE54 3.1  INTRODUCTION .............................................................................................................. 54  3.2  METHODS...................................................................................................................... 56  3.3  RESULTS ....................................................................................................................... 60  3.4  DISCUSSION .................................................................................................................. 64  3.5  ACKNOWLEDGEMENTS ................................................................................................... 66  3.6  REFERENCES ................................................................................................................ 67  REGULATING CAR MASS FOR CONCURRENT TRAFFIC SAFETY AND CLIMATE MITIGATION BENEFITS ............................................................................................................................. 71  5  4.1  INTRODUCTION .............................................................................................................. 71  4.2  DATA AND METHODS ...................................................................................................... 74  4.3  RESULTS ....................................................................................................................... 79  4.4  DISCUSSION .................................................................................................................. 84  4.5  ACKNOWLEDGEMENTS ................................................................................................... 88  4.6  REFERENCES ................................................................................................................ 89  CONCLUSIONS ...................................................................................................................... 92 5.1  RELATIONSHIP BETWEEN PPOLICY INTEGRATED ASSESSMENT, AIR QUALITY, AND TRAFFIC SAFETY RESEARCH........................................................................................................ 92  5.2  RESEARCH RELATIONSHIP TO CURRENT WORKING HYPOTHESES IN THE FIELD OF STUDY AS REFLECTED IN THE LITERATURE ..................................................................................... 94 iv  5.2.1  Air Quality and CO2 Reduction in Transportation................................................... 94  5.2.2  Traffic Safety and CO2 Reduction in Transportation .............................................. 94  5.3  STRENGTHS AND WEAKNESSES OF THIS THESIS .............................................................. 96  5.3.1  Strengths ............................................................................................................... 96  5.3.2  Weaknesses .......................................................................................................... 97  5.4  SIGNIFICANCE AND POTENTIAL APPLICATIONS OF THIS THESIS ......................................... 98  5.5  RECOMMENDATIONS FOR FUTURE RESEARCH ................................................................. 99  5.5.1  Mitigating Air Quality and Health Impacts from Pre-Euro V Diesel Cars in Europe 99  5.5.2  Technology and Policy Assessment of Imposing a 1,600 kg Cap on Passenger Car Mass ............................................................................................................. 100  5.5.3  Traffic Safety Risks Due to Variation in Vehicle Mass and Size: Do Consumers Understand and Internalize the Relative Risks of Purchasing Smaller, Lighter Vehicles? ............................................................................................................. 100  5.5.4  Public Health Risks from Environmental Noise and Hybrid Technology .............. 101  5.5.5  Development of Quantitative, Integrated Policy Models for Passenger Car Choice and Risks............................................................................................................. 101  5.6  REFERENCES .............................................................................................................. 102  APPENDIX A - SUPPORTING INFORMATION FOR AIR QUALITY IMPACTS OF CLIMATE MITIGATION: UK POLICY AND PASSENGER VEHICLE CHOICE .......................................................................... 107 A.1  INTRODUCTION ............................................................................................................ 108  A.2  METHODS.................................................................................................................... 109  A.3  RESULTS ..................................................................................................................... 114  A.4  DISCUSSION ................................................................................................................ 116  A.4.1  NUMBER AND EMISSION CLASS OF ADDITIONAL DIESELS ......................................... 117  A.4.2  ANNUAL KILOMETERS TRAVELLED .......................................................................... 118  A.4.3  SPATIAL DISTRIBUTION OF VEHICLES ...................................................................... 121  A.4.4  PM10 EMISSIONS AND AMBIENT CONCENTRATIONS .................................................. 121  A.4.5  HEALTH EFFECTS .................................................................................................. 124  A.5  REFERENCES .............................................................................................................. 126  v  APPENDIX B - SUPPORTING INFORMATION FOR CHAPTER 3 “TAILPIPE CO2 EMISSION REGULATIONS AND AUTO COLLISION RISKS: A UK CASE STUDY” ................................................................ 129  B.1  INTRODUCTION ............................................................................................................ 130  B.2  METHODS.................................................................................................................... 138  B.2.1  METHODS: DATA SOURCES .................................................................................... 138  B.2.2  METHODS: SUMMARY TRAFFIC SAFETY STATISTICS................................................. 140  B.2.3  METHODS: USING SURROGATE DATA TO ASCERTAIN CURB MASS ............................ 146  B.2.4  METHODS: “FIRST LAW” RR ANALYSIS .................................................................... 152  B.2.5  METHODS: ABSOLUTE RISK ANALYSIS .................................................................... 155  B.2.6  METHODS: STATISTICAL MODELS FOR CURB MASS AND CO2/KM.............................. 160  B.3  RESULTS ..................................................................................................................... 165  B.4  DISCUSSION ................................................................................................................ 166  B.5  REFERENCES .............................................................................................................. 170  vi  LIST OF TABLES  Table 1.1  Summary of research tasks, data sources, and analysis methods. ..........................8  Table 2.1  Approximate ranking of risks in the UK for different decision criteria, based on the fleet dominated by spark-ignition gasoline and compression-ignition diesel passenger cars. ......................................................................................................30  Table 4.1  Comparison of key parameters for the “first law” RR analysis data set: heavier cars and their drivers cf. lighter cars...............................................................................76  Table 4.2  Results of absolute risk analyses for single-car, car-pedestrian, and two-car collisions. The analysis simulated three alternate scenarios over the time period 2000-2005. Values shown are point estimates of the mean for year 2005. Time series of results including standard errors are plotted in Figure 4.6. ......................82  Table 4.3  Results of simulation of two-car collision relative risk fleet calculation for the year 2005. Relative risk is the driver fatality risk of lighter cars divided by driver fatality risk of heavier cars in the relationship RR = .  is calculated in the simulation assuming randomized collision events sampled from a mass distribution representative of the UK on-road car fleet.  is examined parametrically ranging from 2 to 6. .............................................................................................................84  Table A1  Comparison of the cost of ownership for matched pairs of petrol and diesel 2005 car models (all models shown meet Euro IV Emission Standards).......................108  Table A2  European Union “Euro” emission limits [10] and weighted average emission factors [11] for passenger vehicles in grams per kilometer (g/km). ..................................113  Table A3  Total estimated changes in emissions due to additional diesels 2001-2020. Intervals are defined by dates when new EU emission standards apply as shown in Figure 2.3. Diesels emit higher amounts of PM10, NOx, and 1,3 butadiene but lower amounts of CO, HC, benzene, and CO2. ..............................................................115  Table A4  Summary morbidity and mortality results..............................................................116  Table A5  Emission test failure rates for petrol and diesel passenger cars in the UK. ..........124  Table B1  Summary of selected studies which estimated the effect of vehicle mass on fatality and injury risk. ......................................................................................................131 vii  Table B2  Summary description of electronic databases received........................................138  Table B3  Variation in curb mass and tailpipe CO2 emission rate for model year 2007 Ford Focus versions offered in the UK..........................................................................147  Table B4  Variation in curb mass and tailpipe CO2 emission rate for model year 2007 BMW Series 3 versions offered in the UK. BMW Series 3 was the 10th most newly registered UK car in 2006. ....................................................................................147  Table B5  Range of engine sizes, body types, and propulsion types associated with a model year 2003 Peugeot model 307S. ..........................................................................148  Table B6  Number (1,000’s) of registered private and light goods vehicles by engine size. .151  Table B7  “First law” RR analysis: comparison of data set statistics. ....................................153  Table B8  Single car Absolute risk analysis: comparison of data set statistics. ....................156  Table B9  Comparison of data set statistics for car-pedestrian absolute risk analysis. .........158  Table B10  Summary results of regression models relating curb mass (kg) and gCO2/km to explanatory variables using JATO data. ...............................................................161  viii  LIST OF FIGURES  Figure 1.1  UK trend in CO2 emissions for the total economy, transportation end use, and passenger cars [16]. .................................................................................................2  Figure 1.2  The relationship between vehicle mass, tailpipe CO2/km, fuel type, PM10 emissions, and relative fatality risk. CO2/km was regressed on curb mass for model year 2007 UK cars [21]. PM10 emission factors are for gasoline and diesel cars subject to Euro III, IV, or V emission standards [19]. Fatality relative risk (RR) is calculated based on Evans’ “first law” for two-car crashes [20]: RR = μλ (λ = 3.8), where μ is defined as the mass ratio. RR is the ratio of driver fatality risk for cars at a given mass divided by the fatality risk for cars with a mass of 1,456 kg, the mean of all 2007 models [21]..............................................................................................4  Figure 1.3  UK passenger car market trends in size and fuel choice 1995-2007. The legend includes % change in new registrations over 10 years from 1998-2007. Vehicle size and mass are strongly correlated (see Chapter 4), so “large” also means “heavy,” and “small” also means “light”. In terms of fuel choice, alternative-fuelled vehicles have grown dramatically in recent years but still comprise less than 1% [29]............................................................................................................................5  Figure 2.1  Marginal external social costs (pence per kilometer) in Great Britain adapted directly from reference [6]. Error bars represent the estimated range of the marginal external costs for each risk. .....................................................................31  Figure 2.2  Influence diagram of relationships between passenger car policies and risks in a market dominated by spark ignition gasoline and compression ignition diesel cars.34  Figure 2.3  Diagram illustrating passenger car that policy can influence the makeup of on-road vehicle stocks by regulating new cars, scrapped cars, or both. ..............................40  Figure 2.4  Effect of scrapping age and VKT on life cycle gCO2/km for 2006 models of gasoline and diesel cars in Europe [98]. .................................................................40  Figure 2.5  Mean noise rating of some 2006 models of alternative fuelled vehicles relative to conventional gasoline or diesel models [80]. The mean for gasoline models is the same for diesel models (72.3 dB(A) for both groups). CNG = compressed natural gas. LPG = liquefied petroleum gas........................................................................42 ix  Figure 3.1  Diesel share of new car registrations in the European Union (EU) and the UK. While aggregate EU demand for diesels began increasing in 1995, UK demand continued to decline until the first CO2 policy incentive came into effect in 2001 and has been increasing continuously...........................................................................55  Figure 3.2  Total number and percentage market share of new registrations of private and company diesel cars in the UK 1994-2005. During this period, the ratio of the price of petrol to diesel was remarkably stable averaging 0.98 (range 0.95-1.00) and fuel price advantages experienced elsewhere do not provide a plausible explanation for the observed changes in diesel registrations in the UK. .........................................56  Figure 3.3  Integrated framework for assessing emissions from additional diesels (i.e., diesels substituted for petrol vehicles). Actual diesel share of new registrations from 19902005 is based on industry data. Projected shares from 2006-2007 are based on industry forecasts, and from 2008-2020 based on authors’ projections. The focus of this study is on the area between the actual/projection curve and the “no growth” curve which is split into three time intervals defined by the applicable emission standard: Euro III, Euro IV, and post-Euro IV. ........................................................60  Figure 3.4  Summary results of the impact of additional diesels in the UK from 2001-2020 .....61  Figure 3.5  Estimates of additional diesels in the UK 2001-2020 disaggregated by Euro emission class. “Additional diesels” are defined as the number of petrol vehicles switched to diesel beyond the “no growth” estimate. Euro III and Euro IV emission standards apply in 2001 and 2006, respectively. Early adoption of some Euro IV diesels is incorporated into our estimates. Legislation to harmonize diesel and petrol particulate matter emission limits is proposed by 2009, described as “postEuro IV” in this study...............................................................................................62  Figure 3.6  Estimated changes in emissions from 2001-2020 due to additional diesels. The solid lines (y-axis to left) show estimated changes in emissions of common air contaminants in kilo-tonnes, while the dashed line (y-axis to right) shows CO2 in mega-tonnes. Diesels emit higher rates of PM10 and NOx, and lower rates of HC, CO, and CO2. Emissions of common air contaminants are assumed to be harmonized for diesel and petrol vehicles beginning 2009, so differences in all emissions except CO2 approach zero from 2009-2020 as higher polluting diesels are scrapped...........................................................................................................63  x  Figure 4.1  1997-2006 time series of fatalities (left Y-axis) and KSI (right Y-axis) per billion passenger km for key road user groups in the UK. This illustrates the variation of casualty rates between road users (e.g., motorcycle occupant rates are 40-50 times car occupants), and that fatality rates do not always parallel injury (e.g., bicycle fatal and KSI rates 2003-2006). ..................................................................72  Figure 4.2  Distribution of 2,946 fatalities and 2,714 fatal crash events in the UK for 2007. There were also 30,720 KSI and 27,036 KSI events in 2007 with a similar distribution as fatalities. ..........................................................................................73  Figure 4.3  Baseline plus three alternative scenarios for years 2000-2005 used in the absolute risk analysis and the RR fleet composition simulation. “Lighter” group is comprised of engine size ranges 701-1,000 cubic centimeter (CC), 1,001-1,200 CC, and 1,201-1,500 CC. “Mid-mass” includes 1,501-1,800 CC and 1,801-2,000 CC. “Heavier” includes 2,001-2,500 CC, 2,501-3,000 CC, and 3,000 CC and over. 700 CC and under, were not included because there were too few vehicles, annual fatality counts were zero or near zero, and it was not possible to estimate risks. Scenario descriptions are provided in the text. .......................................................78  Figure 4.4  “First law” RR risk of driver fatality in two-car collisions and ratio of CO2 emission rates for UK cars 1995-2005. RR of driver fatality in two-car collisions is shown on the left vertical axis (mean +/- one standard error). Ratio of tailpipe CO2 emission rate (mean +/- one standard deviation) is shown on the right vertical axis. A horizontal line is drawn showing where CO2/km ratio =1.0. ....................................79  Figure 4.5  Relationship between driver two-car conditional KSI risk and vehicle CO2 emission rate in the UK. Model year passenger cars 1995-2004 for crash events during calendar years 2000-2004 are included. Increased CO2/km is a modest but significant predictor of decreased risk. Car make and model size key as defined by DFT: L/S = low/sports, S = small, S/M = small/medium, M = medium, L = large, MPV = multipurpose, 4WD = four wheel drive. .......................................................81  Figure 4.6  Single-car, car-pedestrian, and two-car absolute risk results, shown from top to bottom. Changes in annual fatalities are plotted as percentage change relative to the baseline. A horizontal line is plotted at 0% (no change). Error bars represent +/- one standard error. Points are plotted slightly offset in the time scale (x-axis) to make error bars visible. ..........................................................................................83  xi  Figure 4.7  Comparison of RR calculated in this study (from Figure 4.4), to previous results from U.S. data sets. Both the horizontal and vertical axes are logarithmic scales. The effect of point of impact strongly affects the RR. When the front of a striking car crashes into the driver side of the struck car, the RR is much larger. While the RR for front-to-front collisions has been observed to pass through the origin (i.e., RR ≈ 1 for mass ratio ≈ 1), the RR for purely front-to-driver side impacts has been observed to pass through 10 at the origin (i.e., RR ≈ 10 for mass ratio ≈ 1) [27]. Because of this relationship, the RR for side impact collisions is commonly fit to the equation RR = A * μλ, where statistical models reveal A ≈ 10.................................86  Figure A1  Model for estimating annual number of scrapped vehicles, based on UK deregistration statistics. The annual rate of vehicles scrapped peaks at 10.7%, 14 years after the year the vehicle was initially registered. The area under the curve through 20 years is 85.7%. ...................................................................................109  Figure A2  Fleet average difference in CO2 emission factors (gCO2/km) between petrol and diesel passenger vehicles in the UK from 1997-2020...........................................111  Figure A3  Average annual travel distance for the first two years of ownership for diesel and petrol-fuelled passenger cars. Data are from the UK National Travel Survey. Sample sizes for individual years range from 12 to 563. Both company and privately owned vehicles are included. .................................................................119  Figure A4  Annual travel distance for diesel and petrol cars for NTS survey year 2004. This plot shows the effect of vehicle age on annual travel distance. ............................120  Figure B1  Historical traffic fatality rates for the UK................................................................140  Figure B2  Number of vehicles involved in fatal crashes 1994-2005......................................140  Figure B3  Road types for all fatal crashes 1994-2005. .........................................................141  Figure B4  Road class for all fatal crashes 1994-2005...........................................................142  Figure B5  Road speed limits for all fatal crashes 1994-2005. ...............................................142  Figure B6  Casualty types for all fatalities 1994-2005. ...........................................................143  Figure B7  Sex for all fatalities 1994-2005. ............................................................................143  Figure B8  Age and sex for all fatalities in 2005. ....................................................................144 xii  Figure B9  Crash mode for all single vehicle crashes 1994-2005. .........................................144  Figure B10  Objects struck off carriageway for all single vehicle crashes 1994-2005. .............145  Figure B11  Point of impact for two vehicle fatalities 1994-2005 ..............................................145  Figure B12  Curb mass (kg) histogram for model year 2007 Ford Focus versions. .................146  Figure B13  Diagram of the UK Data Archive data sorting process..........................................152  Figure B14  Regression model #2 line fit and residual plot. .....................................................163  Figure B15  Regression model #9 line fit and residual plot. .....................................................164  Figure B16  Line fit plot of “first law” RR regression model that relates mass ratio to RR fatalities in two car-collisions. Top panel models the relationship as RR = µλ. Bottom panel models the relationship as RR = α + βµ. µ = mass ratio. ………………………….165  Figure B17  Line fit plot of conditional risk model that relates two-car collision % risk of serious injury or fatality to CO2 emission rate....................................................................166  Figure B18  Mean annual travel distance by engine size category in 1998 and 2004 for cars in UK National Travel Survey [42]. Includes vehicles coded as “cars” and “Landrover/Jeep”, but excludes “light vans”. Error bars are +/- one standard deviation. n=8207 for year 2004; n=8692 for year 1998. ......................................167  Figure B19  A comparison of the proportions of registered cars and recorded crash involvements for year 2005, disaggregated by engine size [32, 33]. ....................168  Figure B20  Change in annual new car registrations in 2006 as compared to 1997 in the UK.169  xiii  LIST OF ACRONYMS  ACEA  European Automobile Manufacturers’ Association (www.acea.be)  AIS  Abbreviated Injury Scale  CO  carbon monoxide  DALY  Disability Adjusted Life Years  DFT  UK Department for Transport  EU  European Union  FARS  Fatality Analysis Reporting System (U.S.)  gCO2  grams of CO2  GHG  greenhouse gases (CO2, CH4, N2O)  HC  hydrocarbon emissions (also known as volatile organic compounds)  KSI  killed and seriously injured  IPCC  Intergovernmental Panel on Climate Change  NCAP  New Car Assessment Program  NHTSA  U.S. National Highway Safety Administration  NOx  nitrogen oxides  PM10  Particulate matter less than 10 microns aerodynamic diameter  PM2.5  Particulate matter less than 2.5 microns aerodynamic diameter  RR  Relative risk  VED  Vehicle Excise Duty  YLL  Years of Life Lost  xiv  GLOSSARY  Aggressivity: the risk imposed on others (drivers/passenger/pedestrians) due to the combined effect of driver behavior and their vehicles. Cars: four-wheel passenger vehicles including all market segments. Vehicles that would be classified as “light trucks” in the U.S. and Canada are included. In this study, cars strictly includes four-wheel passenger vehicles coded as vehicle type 9 in the UK Data Archive. All two-wheel vehicles, taxis, heavy goods, bicycles, etc,. are not included as “cars.” Case car: see definitions for striking car and struck car. Casualty: slight injury, serious injury, or fatality Compatibility: design features that result in equalizing risks of fatality or injury between two cars; the goal of aggressivity programs are to make cars compatible. Aggressivity is generally considered to result from three design parameters: mass compatibility; structural or stiffness compatibility, and geometric compatibility (matching height of structural members that contact in collisions – e.g., average height of force in crash tests). CO2 emission rate: average tailpipe greenhouse gas emissions per unit distance traveled, quantified as grams CO2 per kilometer (gCO2/km) where “CO2” includes CH4 and N2O based on equivalent global warming potential conversions. The gCO2/km emission factors employed in this study do not include other emissions that affect global warming such as black carbon, nor are life cycle emissions included. Crash, collision, or accident: an event where a vehicle strikes anything (e.g., another vehicle, pedestrian, or object). The terms “collision” or “crash” are generally preferred over the term “accident.” Crash (or collision) type: in this study (and some other published studies, although the use of terminology is not universal) this phrase refers to basic crash groupings such as two-car, multiplecar, single-car, car-pedestrian, car-bicycle, car-heavy goods, car-motorcycle, and heavy goodspedestrian. Crashworthiness: relating to physical design or technology of vehicles aimed at minimizing injuries or fatalities when collisions occur (sometimes described as “secondary safety”). Crash prevention or avoidance: relating to the ability of the driver or vehicle technology to avoid a collision (sometimes described as “primary safety”). xv  Delta V (or ∆V): velocity change for two cars in a collision resulting from conservation of momentum. For the simple case of two cars colliding head-on (12 o’clock point of impact) it can be expressed as: ∆V1 = (V1 + V2) * M2 / ( M1 + M2) or ∆V1 = SQRT ( 2 * Ea * M2 / [ M1 * ( M1 + M2) ] ) ∆V2 = (V1 + V2) * M1 / ( M1 + M2) or ∆V2 = SQRT ( 2 * Ea * M1 / [ M2 * ( M1 + M2) ] ) V ≡ velocity; M ≡ mass; Ea ≡ total kinetic energy absorbed in the crash subscript 1 is case car, subscript 2 is other car ∆V is computed for crashes in the U.S. Crashworthiness Data System (CDS), but no known ∆V estimates exist for UK data sets. Driver behavior: how, where, and when people drive, and how vehicles are maintained. Driver: person operating the vehicle. Euro NCAP: European New Car Assessment Programme (www.euroncap.com) which crash tests new cars to determine crashworthiness with four basic protocols: front impact, side impact, pole test, and pedestrian protection. Fatality or Mortality: death following a vehicle collision within 30 days (30 days in the UK, the number of days can vary in different countries and jurisdictions). Fuel economy: fuel consumption per unit distance (e.g., L/100km) or its inverse (miles per gallon). In this thesis, the term refers to the rated fuel economy such as listed in the JATO database. Onroad fuel economy for UK cars is generally considered to be 10% less efficient than rated (tested) fuel economy. Haddon matrix: a Haddon matrix in traffic safety is a general way of modeling causes of traffic casualties and interventions that uses a 3 by 3 matrix. The horizontal row of the matrix is comprised of (i) human factors, (ii) vehicle, and (iii) environment. The vertical column is comprised of (i) precrash, (ii) during crash, and (iii) post-crash factors. Induced exposure: methods that estimate absolute risk of injury or fatality using indirect estimates of exposure based on collision data. Risk is quantified with the numerator being the count of outcome of interest (e.g., fatalities or injuries for chosen types of events and driver characteristics). The denominator is estimated based on various assumptions using the collision data set, such as assuming mathematical relationships between proportions of drivers involved in single-car events and multiple-vehicle events. “Quasi-induced exposure” methods make use of data where one driver xvi  is recorded as being responsible for the event, while another driver is designated non-responsible (the non-responsible group is assumed to be representative of all drivers of a given type). Injury: The UK Data Archive uses a three-tier injury rating: slight injury (cuts, bruises, strains), serious injury (e.g., severe cuts, internal injury, any injury requiring hospitalization, injuries that lead to fatalities after 30 days or more), and fatality (traffic collision injury leading to death in less than 30 days). Globally the Abbreviated Injury Scale (AIS) is often used. The AIS is a 6 point scale ranging from AIS1 for minor injury to AIS6 for fatality. Make: the manufacturer of a vehicle (e.g., Ford, Renault, Vauxhall) Market segment: passenger car groups as commonly reported by the UK Society of Motor Manufacturers and Traders Ltd. (www.smmt.co.uk): mini, supermini, lower-medium, upper-medium, executive, luxury, specialty sports, 4X4/SUV, multi-purpose. Mass or curb mass: static vehicle mass with a full tank of fuel and other fluids. In Europe, curb (spelled “kerb” in the UK) mass data beginning in 1996 includes 75 kg to represent the mass of the driver. In this study the words “light” or “lighter,” and “heavy” or “heavier” refer to lesser or greater mass, respectively. The U.S. EPA classifies cars for purposes of dynamometer testing according to their “inertia weight” which is the curb mass plus 300 lbs. Curb mass in the U.S. does not include driver or cargo (ANSI D16.1-1996). Although weight and mass are distinctly different quantities from Physics (weight being the force exerted by gravity acting on mass), these terms are considered synonymous for the purposes of this study. Model or model range: the basic name of a car design such as Volkswagen Golf, Ford Focus CMax, or BMW 318. Model Year: the year a vehicle is manufactured and sold as per the manufacturer. In this study, the “year first registered” data field is considered equivalent to model year, even though this is unlikely to be correct for imported cars. Occupant: either the driver or passenger. Passenger: person in the vehicle other than the driver. Risk: three types are commonly quantified to measure traffic safety: 1. Absolute risk is used in this study to mean risk measured in fatalities (or injuries) per year (e.g., annual car-pedestrian fatalities in the UK), or fatalities per year per unit quantity of vehicles (e.g., millions registered) or vehicle-distance (e.g., billion vehicle-km). In general xvii  other risk metrics are used such as casualty counts per unit driving time. Induced exposure methods are also used to estimate absolute risks. 2. Relative risk or risk ratio is used in this study to measure the risk of one group divided by the risk of another group. The “first law” relationship described in this study is an example of relative risk because it quantifies the risk of one group of drivers divided by the risk of another group of drivers. 3. Conditional risk is used to measure the risk (e.g., probability of fatality or injury) given that a crash occurs. The conditional risk metric is perhaps the best indicator of crashworthiness (or “secondary safety”). Road user: any persons using public roads who are subject to potential injury or fatality from car collisions; includes pedestrians, cyclists and occupants of cars, two-wheel vehicles, taxis, buses, heavy goods vehicles etc. Size: a physical measure of the vehicle that is commonly based on a characteristic length such as wheelbase (the distance between front and rear axles), length (also called “overall length”), width (also called “overall width”), or track width (the side-to-side distance between wheels). Many other direct measures of size have been assessed. Other indirect measures for size have been used such as the lateral distance needed to perform a 180° turn (“turn distance”). In this study, the words “small” or “smaller” and “large” or “larger” refer to size, not mass. Stiffness: generally, the slope of the force (y-axis) versus displacement (x-axis) for a car as it crushes in a collision. Multiple definitions of stiffness have been employed and typically, but not consistently, stiffness has been strongly correlated with mass. Striking car: the second car of interest in a risk calculation, also called the “other,” “partner,” or “bullet” car. Struck car: the first car of interest in a risk calculation, also called the “case”, “subject”, “self”, “own”, “target”, or “driven” car. Traffic safety: relating to human health impacts as a result of vehicles operating on public roadways. Although property damage is generally included as a traffic safety impact, this study only considers human health. Version or trim: descriptors that define specific features of a car model or model range. Examples: Volkswagen Golf 2.0 TDI 140 PS sport, Ford Focus C-Max 2.0I Zetec auto, and BMW 318D M sport touring. xviii  PREFACE  This thesis is written following the manuscript thesis format of the University of British Columbia (UBC). UBC’s manuscript thesis is comprised of one or more manuscripts suitable for journal publication (either as published or as intended for publication), with an Introduction at the beginning and a Conclusion at the end1. In order to fulfill the manuscript thesis requirements there is some redundancy in the Introduction and Conclusion chapters. Each journal requires an introduction to frame the problem, which is also required in the Introduction chapter. In general this thesis is written in the first person singular (“I”). The exception is the first person plural (“we”) is used in Chapters 3 and 4, (as well as Chapters 1 and 5 when referring to Chapters 3 and 4) because this material is based on journal articles for which I am the principal author, but there are one or more co-authors. Because this thesis employs multiple disciplines, it is critical for readers to understand the use of terminology. I have created a Glossary, mainly for the traffic safety research because I found no single source suitable to define terminology in this field. For terminology rooted in human health science, I make use of terms as defined in “The Dictionary of Epidemiology” by John Last (4th edition, Oxford University Press, 2000). As examples, the phrases “ecologic fallacy,” “relative risk,” and “disability adjusted life years” are all defined by Last. For terminology rooted in economics, I make use of the text “Economics of the Public Sector” by Joseph Stiglitz (3rd edition, W.W. Norton and Company, 2000). As examples, the phrases “rebound effect” and “social cost” are defined by Stiglitz. For general words, I have used Merriam-Webster’s online dictionary at www.merriamwebster.com.  1  See http://www.grad.ubc.ca/students/thesis/index.asp?menu=002,002,000,000 for more information.  xix  ACKNOWLEDGMENTS Nothing is accomplished without the help of others, and this thesis is no exception. I have learned a lot and owe gratitude to many. I doubt I will remember everyone at this time of writing, so I’ll focus on my family, student colleagues, and faculty mentors. Acknowledgments of funding and reviewers for specific articles are provided at the end of chapters 2, 3, and 4. On a personal level, I have received patient support and inspiration from my wife Theresa, and my daughters Alex and Angeline – thank you so much for all that you are. To Theresa I owe more gratitude than I can say, personally and academically (even going back to the old Chaffey College days!). Alex and Angeline, you have given me the gift of appreciating both the value of living in the moment and investing in our collective future, not to mention helping me learn from my many mistakes. I have learned and been inspired a great deal from my student colleagues at UBC at the Institute for Resources, Environment, and Sustainability (IRES), and the Bridge Program. I won’t list specific student colleagues to avoid generating an exhaustive list. But I will say that I have learned from and been inspired and impressed by many, including those I’ve frequently collaborated with and even some that I’ve seldom spoken with by virtue of seeing their work. I also owe a great deal to my thesis committee. I have truly been privileged with a world-class faculty committee. I would like to thank Dr. Milind Kandlikar, Dr. Michael Brauer, and Dr. Douw Steyn from whom I’ve learned a great deal through my exams, directed studies, and thesis support. I would also like to thank Dr. Kay Teschke for her outstanding teaching support through the Bridge Program. I also want to thank the university examiners, Dr. Steve Rogak and Dr. Karin Mickelson, for their helpful comments and questions. I also thank my external examiner, Dr. Lester Lave for his helpful, detailed comments on my thesis. Last, but first, I owe a great deal to Dr. Hadi Dowlatabadi. Beginning with our introduction eight years ago in his Integrated Assessment course, it has been quite a journey and I thank him for his tireless support. I will share one brief story about Hadi for the record. I had the chance to chat briefly with a world-renowned economist about my thesis to gain some insights. The subject of equity in public policy came up, to which my famous economist commenter said (paraphrase) “oh yes, Hadi cares about equity.” I don’t think it was meant as a compliment in particular. But it struck me right away – yes, I thought, he does care about equity which is one reason I’m privileged to work with him!  xx  DEDICATION  . I dedicate this thesis to my family: Theresa, Alex, and Angeline.  xxi  CO-AUTHORSHIP STATEMENT  Chapter 2. The idea for this study was jointly developed by me and Hadi Dowlatabadi. I developed the scope of the research, selected the methodology, performed the analyses, and prepared the manuscript. Chapter 3. Hadi Dowlatabadi proposed the idea for the study. I developed the scope of the research, selected the methodology, obtained all data, performed the analyses, and prepared the manuscript. Hadi helped make decisions on the scope of the analysis and contributed to the manuscript preparation. Chapter 4. Hadi Dowlatabadi proposed the idea for the study. I developed the scope of the research, selected the methodology, obtained all data, performed the analyses, and prepared the manuscript. Hadi helped make decisions on the scope of the analysis and contributed to the manuscript preparation. Hadi also created the initial fleet simulation model in AnalyticaTM, which I modified. Milind Kandlikar helped make decisions on the scope of the analysis and contributed to the manuscript preparation. Additional contributions to Chapters 2, 3, and 4 are provided in the acknowledgements for each article.  xxii  1  INTRODUCTION, LITERATURE REVIEW, OBJECTIVES AND HYPOTHESES  1.1  INTRODUCTION  Climate policies applied to passenger vehicles can be effective in reducing CO2 emissions by changing vehicle mass, fuel choice, and technology [1]. However, these same factors that reduce CO2 emissions, such as reducing vehicle mass and switching from gasoline cars to diesel cars, can lead to changes in common air pollutant emissions, vehicle safety, and, subsequently, changes in health outcomes. This is an applied research study on the human health risks resulting from climate mitigation policies aimed at the passenger transportation sector. By necessity, most research to date has been based on scenarios or hypothetical changes in transportation choices [2]. Where it is feasible, this research emphasizes how actual changes in passenger car choices in the United Kingdom (UK) may have changed human health outcomes resulting from air quality and traffic safety. From both scientific and public policymaking perspectives, climate change is widely described as one of the most challenging and urgent problems in society today. The scientific evidence of climate change has been thoroughly evaluated by the Intergovernmental Panel on Climate Change (IPCC). The IPCC’s current finding is that “observational evidence from all continents and most oceans show that many natural systems are being affected by regional climate changes” and that “it is likely that anthropogenic warming has had a discernible influence on many physical and biological systems.” [3] Examples of policymakers’ acceptance of the need to address climate change are ubiquitous, spanning all levels of government. To cite a municipal example, London has committed to reduce CO2 emissions by 60% from 1990 levels by the year 2020 [4]. At a national level, the UK government has stated that “The 2008 Climate Change Act made Britain the first country in the world to set legally binding ‘carbon budgets’, aiming to cut UK emissions by 34% by 2020 and at least 80% by 2050” [5]. In principal there are three basic interventions that policymakers can target to address climate change [2, 6]: mitigation, adaptation, and geoengineering. This research specifically focuses on mitigation via reductions in CO2 emissions from passenger cars. In the UK public policymakers at multiple levels of government have targeted reductions of CO2 emissions [7]. In 2008 the UK created the Department of Energy and Climate Change with a mandate to “bring together … energy policy … and … climate mitigation policy.” [8] Because of the importance of transportation in emissions, with passenger cars as a major contributor, multiple government agencies in the UK have identified climate mitigation for the 1  road transportation sector as a high priority [7, 9-12], including creating an Office for Low Emission Vehicles [13]. The UK was also an early adopter of climate mitigation policies for passenger cars. In 2001 it adopted a vehicle excise duty (VED) with annual fees based on a scale of certified CO2/km for car makes and models [14]. Approximately half of all new cars in the UK are sold as company cars. In 2002 a tax on the benefit-in-kind income from employee use of company cars was adopted, again based on certified CO2/km [15]. While the UK has reduced overall CO2 emissions, transportation remains an exception. Figure 1.1 illustrates trends in CO2 emission in the UK. From 1990 to 2007, CO2 emissions were reduced for every major end use except transport where emissions have risen by 8% [16]. Passenger cars have comprised 58 to 61% of transport emissions over this time period. The UK is the second largest new car market in the European Union (EU) where the European Automobile Manufacturers’ Association (ACEA) established a voluntary agreement with the European Commission in 1998. The ACEA agreed to reduce EU fleet-average passenger from 185 gCO2/km in 1995 to 140 gCO2/km by 2008, but the target was not met as ACEA’s 2008 average was 152 gCO2/km [17]. Moreover, reductions from passenger cars appear likely to remain difficult as car ownership continues to rise. The UK Department for Transport (DFT) forecasts ownership levels relative to the year 2000 of 122% by 2015, and 133% by 2025, a growth rate that is in excess of the population growth rate [18]. Figure 1.1  UK trend in CO2 emissions for the total economy, transportation end use, and passenger cars [16]. Million tonnes CO2 equivalent  900  180  800  160  700  140  600  120  500  100  400  80  300  60 total (left Y-axis)  200 100 0 1989  40  total transport (right Y-axis)  20  passenger cars (right Y-axis)  0 1991  1993  1995  1997  1999  2001  2003  2005  2007  2  1.2  RESEARCH OBJECTIVES BASED ON LINKS BETWEEN PASSENGER CAR CO2, AIR QUALITY, AND TRAFFIC SAFETY  Figure 1.2 illustrates the essential relationships that led to the research objectives for this thesis. Empirical relationships for UK cars confirm that tailpipe CO2 emission rates increase linearly with curb mass for gasoline and diesel cars, and that diesel cars emit less CO2 than gasoline cars. Compared to gasoline cars, PM10 emission factors from diesel cars are 25, 12, and 2 times greater for cars meeting Euro III, IV, and V emission standards, respectively [19]. Theoretical and empirical “first law” relationships for U.S. cars reveal that the relative risk of fatalities for drivers of lighter cars versus heavier cars in two-car collisions rises non-linearly with the mass ratio [20]. There have been two noteworthy trends in the UK as shown in Figure 1.3: (1) the growth in market share of diesel cars at the expense of gasoline cars, and (2) the growth in market share of both large (heavy) and small (light) cars at the expense medium class cars. Diesel cars registrations have increased 182% from 1998 to 2007, notably since 2001 when a sharp and sustained rise in diesel new registrations has occurred. New registrations of large cars, particularly sport utility and multipurpose, and small cars have increased over the same time period by 55% and 30%, respectively. Concurrently, registrations of medium class cars have declined 15%. Based on visual inspection of these time series trends, the rise in small cars is more temporally related to the ACEA CO2 agreement, while the diesel growth is more temporally related to the UK’s VED and company car tax regimes. However there are no known studies that have rigorously quantified the impact of these policies on new registrations such as with econometric methods [22]. There are other potential factors that have contributed to the growth in diesel and small cars, such as technological change, fuel taxation, and global oil prices. Nonetheless, the UK CO2 policies and ACEA CO2 agreement are commonly credited with making substantial contributions in reducing UK fleet-averaged emissions of newly registered cars from 190 gCO2/km in 1997 to 165 gCO2/km in 2007 [15, 17, 23, 24]. Learning from the UK experience is important because improved vehicle efficiency through changes in vehicle design (including fuel and mass) are promoted as viable strategies globally to mitigate climate change [1, 25-27].  3  The relationship between vehicle mass, tailpipe CO2/km, fuel type, PM10  Figure 1.2  emissions, and relative fatality risk. CO2/km was regressed on curb mass for model year 2007 UK cars [21]. PM10 emission factors are for gasoline and diesel cars subject to Euro III, IV, or V emission standards [19]. Fatality relative risk (RR) is calculated based on Evans’ “first law” for two-car crashes [20]: RR = μλ (λ = 3.8), where μ is defined as the mass ratio. RR is the ratio of driver fatality risk for cars at a given mass divided by the fatality risk for cars with a mass of 1,456 kg, the mean of all 2007 models [21].  RR ≡ fatality relative risk RR = (risk at variable mass) / (risk at mean mass) 10.0 where mean car mass = 1456 kg  500 450  grams/km CO2 or PM *10000  350 300 RR = 1.0  250  1.0  200 150  RR, fatality relative risk  400  100 50 0 500  0.1 1000  1500 2000 Curb mass (kg)  2500  3000  3500  gCO2/km Gasoline = -23.6 + 0.162 * kg gCO2/km Diesel = -28.4 + 0.132 * kg gPM/km Diesel Euro III gPM/km Diesel Euro IV gPM/km Diesel Euro V gPM/km Gasoline Euro III/IV/V RR, fatality relative risk (Y axis at right)  The intent of the first research objective was to examine how research on air quality and traffic safety risks could inform development of CO2 policies for passenger cars. However, there are many risks associated with passenger car use and thus the first research objective was chosen to be an examination of the rationale and basic framework for quantitative Integrated Assessment models of passenger car CO2 policy incorporating air quality, traffic  4  safety, and other relevant risks. Additional research context to support selection of the first research objective is provided in Section 1.4.1. The ACEA, VED, and company car tax policies provided incentives to manufacturers and consumers to reduce the average CO2/km of newly registered cars, including diesel [14, 15, 17]. Yet these policies were adopted 7 to 12 years prior to 2009 Euro V standards that were planned to harmonize the emissions of NOx, PM10, CO, and VOC’s for gasoline and diesel cars [19]. This mismatch in the timing of CO2 and air pollutant emission standards led to the second research objective, which was to assess the impact of diesel growth on CO2 emissions, air quality, and human health. Additional research context that supports the second research objective is provided in Section 1.4.2. The growth of larger (heavier) and smaller (lighter) cars at the expense of medium class cars (Figure 1.3) led to development of the third research objective. Given the prior research that vehicle mass is a determinant of both annual fatalities as well as the equity of risk in two car collisions [20, 28] and that vehicle mass directly influences CO2 emissions regardless of fuel type (Figure 1.2), the third research objective was chosen to examine relationships between tailpipe CO2 emissions and traffic safety risks, using vehicle mass as an intermediate variable. Section 1.4.3 provides an expanded discussion of the research context that supports the third research objective. Figure 1.3  UK passenger car market trends in size and fuel choice 1995-2007. The  legend includes % change in new registrations over 10 years from 1998-2007. Vehicle size and mass are strongly correlated (see Chapter 4), so “large” also means “heavy,” and “small” also means “light”. In terms of fuel choice, alternative-fuelled vehicles have grown dramatically in recent years but still comprise less than 1% [29]. Small +30% 1998-2007  70%  Medium -15% 1998-2007 Large +55% 1998-2007  60% New registrations  Diesel +182% 1998-2007  50% 40% 30% 20% 10% 0% 1994  1996  1998  ACEA voluntary CO2 agreement  2000 UK CO2-based VED  2002  2004  UK CO2-based company car tax  2006  2008  5  1.3  RESEARCH QUESTIONS, HYPOTHESES, AND TASKS  For each of the three research objectives, research questions, hypotheses and specific tasks for each research question are described below. This is followed by Table 1.1 which summarizes the data sources, methods, and limitations to achieving each objective RQ1 Question: In quantitative analysis of policies to achieve reductions in CO2 emissions from passenger cars, how important are Integrated Assessment methods and multiple risk frameworks, as opposed to separate, independent analyses of policies and risks (like air quality or traffic safety)? RQ1 Hypothesis: A broadening of the domain of mitigation policies for passenger cars can be more effective in reducing annual CO2 emissions, and also improve public health and other public risks associated with passenger car use. RQ1 Tasks: a. Review and summarize policy options that are designed to reduce annual CO2 emissions from passenger cars in the UK. b. Discuss the potential effects of choice of decision criteria and risks included or excluded from policy analyses. c. Assess the potential effects of choice of risk metrics. d. Outline an Integrated Assessment framework for analyzing CO2 reduction policies for passenger cars demonstrating how policies are linked to multiple public risks. RQ2 Question: How has the steep growth of diesel cars changed CO2 and PM10 emissions, and mortality related to PM10? RQ2 Hypothesis: Substitution of diesel cars for gasoline cars has produced savings in tailpipe CO2 emissions, with a resultant tradeoff in chronic exposure mortality due to traffic PM10 emissions. RQ2 Tasks: e. For the 2001-2020 study period, calculate annual CO2 reduction due to “additional diesels” based on the difference in average CO2 emission factors for gasoline and diesel cars, annual travel distance, and a UK-specific model for car scrap rates.  6  f.  For the 2001-2020 study period, calculate reduction in petroleum fuel consumption due to additional diesels based on the fuel economy (L/100km) for the average diesel and average gasoline cars.  g. For the 2001-2020 study period, estimate annual counts of additional diesels, and changes in air contaminant emissions (PM10, NOx, CO, and HC). h. Use the published results of existing air quality modeling studies to estimate the annual, population-weighted change in PM10 concentration. i.  Quantify changes in annual mortality and hospitalizations using previously developed concentration-response functions most applicable to the UK.  RQ3 Question: Climate mitigation policies such as the ACEA CO2 agreement and UK tax policies have provided incentives in favor of cars with lower certified CO2/km tailpipe emission rates, which are generally smaller and lighter cars. Simultaneously, consumer preferences for larger, heavier, higher CO2/km cars have kept these segments gaining market share. What is the relationship between tailpipe CO2/km emission rates of passenger cars and traffic safety risk? RQ3 Hypothesis: Occupants of smaller and lighter (lower CO2/km) will be subject to relatively larger risks of fatality or injury as compared to occupants of larger and heavier cars (higher CO2/km). Reducing on-road shares of heavier cars while increasing shares of lighter cars in the UK, compared to business-as-usual growth of heavier cars, will result in improvements in traffic safety. RQ3 Tasks: j.  Replicate Evans’ “first law” driver fatality risk for two-car crashes using UK data for recent crash years and model years [20].  k. The UK Department for Transport (DFT) has published the conditional risk of serious injury or fatality for makes and models of cars for model years 2000-2004 [30]. Using emissions data from independent sources, examine how the conditional risk relates to car CO2/km. l.  Estimate changes in annual traffic fatalities in the UK during 2001-2005 if large and heavy cars had declined in market share, instead of risen. Estimate changes in fatalities for car-pedestrian, single-car, and two-car crash modes.  7  m. Estimate changes in the “first law” fatality risk for two-car crashes in 2005 if large and heavy cars had declined in market share, instead of risen, and assuming randomized crash events.  Table 1.1 Research Task  1. Quantitative Integrated Assessment modeling of passenger car CO2 policy incorporating air quality, traffic safety, and other relevant risks  2. Assess the impact of diesel growth on CO2 emissions, air quality, and human health  Summary of research tasks, data sources, and analysis methods. Parameters or variables Demonstrate the importance of choice of decision criteria using UKspecific parameters Demonstrate the importance of a multiple risk framework using UK-specific parameters  Sources of data and information resources  Use the results of research task 1 (air quality and diesels), task 2 (traffic safety and curb mass), and other published UK-specific risk estimates [18, 31-33].  Outline a framework for quantitative policy analysis of multiple risks  Selected literature on quantitative policy analysis [38, 39].  Emission factors  National Atmospheric Emission Inventory [19] Society of Motor Manufacturers and Traders Ltd. [24, 29, 40]  Baseline mortality and morbidity rates, demographics  Dept. of Health [42] UK Statistics Office Government Actuary's Department [43]  Ambient concentrations  Data archive studies [44-48] Dept. for Environment, Food, Rural Affairs  Concentrationresponse factors  Peer reviewed literature [49, 50] and UK-specific studies [30, 51]  Analysis methods Tabulated relative ranking of risks using three decision criteria: i) economic social cost [34], ii) public health [35], and iii) social-political priorities [36, 37]. Showed how the set relevant risks can change and how the system boundary for risk analysis can change with choice of fuel and vehicle technology. Developed an influence diagram showing basic linkages between policies and risks.  Impact pathway analysis. Mortality for ages ≥ 30. Quantify health outcomes as annual counts.  Data gaps; research challenges; limitations; other comments  The importance of Integrated Assessment, multiple risk quantitative modeling of passenger car CO2 policies has been demonstrated with UK-specific quantities and discussion. Building a quantitative policy model would be a separate research project, and beyond the scope of this thesis.  Only PM10 mortality and partial morbidity is quantified. Work is completed and published [41].  8  Research Task  3. Assess the correlation between tailpipe CO2 emission rates and various traffic safety measures in the UK, using curb mass as the primary variable  1.4  Parameters or variables  Sources of data and information resources  Baseline traffic fatality rates, vehicle populations  UK Data Archive [52, 53] Department for Transport [18]  Curb mass  JATO Dynamics [21] and available online, open source databases [58-60] Surrogate measures of mass: make, model, engine size, and fuel type  Fatality risk rates and coefficients  Statistical estimates using individual collision data from the UK Data Archive, and both direct and indirect measures of vehicle curb mass  Analysis methods 1. Replicated Evans’ method for calculating “first law” of relative fatality risk using UK data [20, 54]. 2. Graphically and numerically compared CO2/km vs. conditional fatality risk using DFT’s results [55]. 3. Using Mengert’s method [56] and policyrelevant scenario of varied fleet mass composition, estimated changes in total fatalities for: single-car, two-car, and carpedestrian crashes. 4. Using the same fleet composition scenarios for which changes in total fatalities are estimated, calculated “first law” relative risk using Latin Hypercube methods [57] to simulate two-car collisions.  Data gaps; research challenges; limitations; other comments  Data on casualties, vehicles, and crash modes for every individual UK car crash reported to police from 19942005 (inclusive) have been obtained from the UK Data Archive. A principal limitation is that curb mass data are not part of the UK Data Archive databases. This was overcome by manually entering curb mass, or using surrogate measures for curb mass. This reduced the precision of the estimates.  CRITICAL HIGHLIGHTS FROM THE LITERATURE  Highlights of the literature emphasizing critical findings that are most relevant to my research objectives are provided here. Additionally, each chapter provides a review of the relevant literature. 1.4.1  CLIMATE MITIGATION POLICY: THE IMPORTANCE OF MULTIPLE RISK ASSESSMENT  The first research objective was intended to set the research framework for the two public health case studies, to examine the lessons learned in designing climate mitigation policy for passenger cars to simultaneously address air quality and traffic safety risks. However, the first research question and tasks were proposed on the premise that a policy analysis  9  framework to incorporate additional risks (i.e., more than just air quality and traffic safety) is necessary. Policymakers in industrialized countries are increasingly tasked to design and implement climate mitigation policy for passenger cars [27, 61, 62]. Car fleets in developed countries are currently dominated by conventional gasoline and diesel technology2 [27, 63], and risk assessments incorporating social costs commonly assume 100% conventional technology [34, 64]. Integrated Assessment of multiple risks is growing in importance because industrialized countries are presently at a crossroads in fuel and drive-train choices. Scenarios for the near to distant future commonly project a much more heterogeneous mix of fuels and drive-train technologies. Examples from the International Energy Agency’s (IEA) current suite of transportation scenarios include steep uptake of plug-in hybrids beginning in 2015, 90% electric or fuel cell vehicles by 2050, and a range of advanced conventional cars (defined as 50-70% better fuel consumption) as high as 90% to as low as 10% [26]. While greenhouse gas reduction is a principal driver of these transitions, climate change is just one of many risks associated with passenger car use, and the effectiveness of new climate policies will depend on how well they balance climate change mitigation with other risks [2]. In designing policies to achieve transitions such as envisioned by IEA, it is clear that risk assessment models will need to adapt. In this study I examine the rationale, and present a basic framework, for Integrated Assessment models to incorporate multiple risks from a societal perspective in the analysis of climate mitigation policies for passenger cars. Using the UK as a case study, I illustrate the importance of decision criteria, risks included (or excluded), and how the choice of fuel and technology options influences system boundaries and relative ranking of multiple risks. The UK is a useful case study because UK circumstances are generally applicable to other developed countries (albeit not completely). The UK was an early adopter of mitigation policies for passenger cars with nationwide tax regimes in place since 2001-2002 [14, 15]. As in other industrialized countries passenger car ownership and road transport CO2 emissions continue to rise [35]. The UK Department for Transport (DFT) forecasts ownership levels relative to the year 2000 of 122% by 2015, and 133% by 2025, a growth rate that is in excess of the population growth rate [18]. Currently the UK car fleet is dominated by sparkignition gasoline and compression-ignition diesel cars, comprising 99.3% of all new  2  I use the term “conventional diesel technology” to include diesel cars with or without particle filters.  10  registrations [24]. However, alternative drive-trains, such as grid-independent hybrids and alternative-fuelled vehicles, such as ethanol-powered cars (E85), are already rapidly gaining market share [24]. Government policies such as the UK Renewable Transport Fuels Obligation [65] and voluntary and regulatory CO2 measures in Europe are expected to continue to stimulate the proliferation of alternative vehicle technologies and fuel types [66]. The UK is similar to other developed countries in its struggle to limit CO2 from transportation. From 1990 to 2005 CO2 emissions were reduced for every major sector, except transportation where emissions have risen by 12% [67]. Because of the importance of transportation in emissions, with passenger cars as a major contributor, multiple government agencies in the UK have identified climate mitigation for the road transportation sector as a high priority [7, 9-12]. The pitfalls of focusing on a single problem in policy design and analysis for passenger cars have long been established [68]. The array of regulations and policies affecting passenger cars is often in tension, and not always consistent with green design principles [69, 70]. A critical issue in Integrated Assessment of multiple risks is that analysis of fuels and vehicle technology alone is insufficient because ownership (cars per capita), annual vehicle kilometers traveled per car (VKT) and patterns of use (e.g., time of day and location) are essential factors that influence risks. A recent example from the UK is that steep growth of diesel cars has not yielded expected savings in fuel consumption and CO2 due to consumer choice of larger cars, driver behavior (speed), and VKT rebound effects [71]. Another example is the ongoing debate about fuel economy standards in the U.S. Considering only the principal environmental risks (air quality and climate change) and energy security, more stringent fuel economy standards are strongly beneficial. However, when considering traffic safety risks via disparate vehicle mass and economic risks of traffic congestion via VKT rebound effects, the likelihood of realizing net social benefits with more stringent fuel economy standards is less clear. While the need for Integrated Assessment in transportation policy is often acknowledged [2], much research has employed Integrated Assessment of multiple risks but applied one decision criterion such as social cost-benefit [35, 72, 73]. Other research has focused on integration of near-term public health risks but has not incorporated other risks such as climate change or congestion [35, 74]. Stakeholders involved in passenger car policy development are increasingly sophisticated in recognizing the interdependencies between fuels, technology, and vehicle ownership and VKT as well as the relevance of a multiple risk framework. For example, U.S. car manufacturers argued that fleet turnover effects (the rate of change of in-use technology over time) and rebound effects (i.e., VKT) might nullify welfare gains from California’s 11  proposed CO2 standards for passenger cars [75]. In Europe, proposals by the European Commission for car CO2 standards differentiated by mass classifications were refuted by a leading environmental organization on the basis of traffic safety concerns [76]. Therefore, robust Integrated Assessment of multiple risks in policy development can reasonably be expected to improve the social and political acceptability of new policies. There are inherent limitations in focusing primarily on Integrated Assessment of multiple risks. A truly comprehensive Integrated Assessment would explore the best strategy for meeting transportation needs within the context of all relevant local, social, and physical conditions. Additional domains that are relevant include organizational and institutional domains [77], interactions with other energy-intensive sectors [78], integration of transportation and land use planning [79], and integration of passenger car travel with other travel modes [35]. In the larger picture, the integration of climate change with other drivers of global change is relevant [80]. Another limitation of my research is that it focuses on risks and does not examine the positive attributes of consumer choice [81]. As an example, the relative benefits of gasoline-powered cars and all-electric cars are not equal. The travel range for a single refueling of a gasoline-powered car is much greater than an all-electric car, an attribute strongly valued by consumers. On the other hand, there could be traffic safety benefits due to more alert driving due to drivers taking a short-rest every few hours to charge batteries. Thus the relative benefits of fuels and technologies to consumers are excluded [63]. 1.4.2  AIR QUALITY AND ANCILLARY BENEFITS OF CLIMATE MITIGATION POLICY  The second research objective was motivated by research literature on the ancillary benefits of climate mitigation policies. CO2 reduction policies have been promoted on the basis that reducing fossil fuel use provides dual benefits in terms of long-term climate change attenuation and short-term air quality improvements. Models predict that climate policies will reduce fossil fuel combustion and lower air pollutant emissions, and therefore provide substantial public health benefits [82-86]. For example, a study in Science [82] concluded “that GHG mitigation can provide considerable local public health benefits from air pollution reduction alone to countries that choose to abate GHG emissions by reducing fossil fuel combustion.” This and similar studies were cited by the IPCC in the 2001 Assessment that featured a chapter on ancillary air pollution benefits of climate policy, without any mention of the possibility of disbenefits [85] Similar findings were more recently published for the European context [86] which found that “Substantial ancillary benefits were found for regional air pollution (SO2, NOx, VOC and particulate matter) of implementing the Kyoto Protocol (intended to control greenhouse gas emissions) in Europe.” The European Environmental 12  Agency also published findings that “Action to combat climate change will deliver considerable ancillary benefits in air pollution abatement” [84]. Neither of these European studies mentioned the potential air pollution disbenefits of using diesel technology for CO2 reduction from passenger cars. There is some acknowledgement in the literature of the potential impacts of diesel cars, in particular one economic study [87] which examined taxation policy for gasoline and diesel cars, including accounting of the “environmental costs” because “diesels have high emissions of particulate matter”. Another editorial acknowledged the need for research into the climate mitigation and air quality effects of diesel car technology [88]. One study assessed the upstream CO2 and energy implications of increasing refinery production of diesel relative to gasoline fuels [89]. Studies in the U.S. context examined ambient ozone implications based on a hypothetical transition from gasoline cars to modern diesel technology [90], and the global warming implications of black carbon emissions from diesels [91]. While there are a handful of studies that partly fit the aim of the proposed research on diesels in the UK, the need for this research is still clear. A majority of research has omitted mention of the social costs of diesel cars in studies of transportation, CO2, and air pollution emissions. Moreover the studies that contain this omission were published in the highest impact publications such as Science [82], The Lancet [83], Environmental Health Perspectives [92], and IPCC’s Assessment Reports [85]. Where research has attempted to quantify CO2 and air quality tradeoffs of diesel cars, it has by necessity been based on scenarios, or hypothetical changes, in transportation choices. Where it is feasible, my research is based on actual changes in passenger car choices in the UK after enactment of climate mitigation policies targeting passenger cars. 1.4.3  VEHICLE MASS, TRAFFIC SAFETY, AND CO2 EMISSIONS FROM PASSENGER CARS  The third objective was formulated on the premise that vehicle mass, CO2 emissions, and traffic safety are interrelated. The relationship between vehicle mass and tailpipe CO2 emissions is relatively straight forward [93, 94]. For any given vehicle design, a heavier vehicle consumes more fuel and results in greater CO2 emissions directly out of the tailpipe and indirectly upstream of the vehicle in the energy supply chain. The relationship between vehicle mass and traffic safety is less clear [28, 95-102]. Appendix B, Table B1 summarizes the methods and findings of several critical studies, and some of the critical findings will be discussed here. One of the earliest landmark studies of these 13  relationships assessed the linkage between U.S. CAFE standards and traffic safety with U.S. FARS data, using vehicle mass as the intermediary variable [96]. It was concluded that the CAFE standard would “be responsible for several thousand additional fatalities over the life of each model year’s cars.” Another landmark study by the U.S. National Highway Safety Administration (NHTSA) using FARS data found that reducing average vehicle mass by 45 kg (100 pounds) would result in a net increase in annual fatalities [28]. The NHTSA findings have been questioned by others who also used FARS data, but separately assessed the effects of mass and size to show that reducing average mass while holding size constant would decrease annual fatalities in the U.S. [103]. Other research examined model-specific fatality rates and argued that vehicle “mass may not be fundamental to safety” [104, 105]. While these dominated the policy debates in the U.S., particularly around the issue of fuel economy standards [106], there are other studies that also reveal important relationships with respect to vehicle mass and traffic safety. For example, one study examined the independent effects of vehicle, driver, and collision variables for single vehicle crashes using FARS data and concluded that increased mass and size together reduce fatality risk [104]. Another examined the role of mass, size, and energy absorption in head-on two-car collisions for multiple casualty types and in multiple contexts (Germany, Japan, U.S.) and found size to be dominant, but mass still an important factor [107]. Yet another study examined driver fatality odds ratio in two-car U.S. collisions, separately assessing different crash modes and independent variables for mass, multiple size metrics, and driver characteristics [108]. It was found that mass ratio in two-vehicle collisions affected fatality risk more than any other vehicle variables, and that equalizing mass across the on-road fleet would reduce overall fatality risk. While the majority of research on the role of vehicle mass in traffic safety has used U.S. data, the relationship has been examined in the UK as well. One study examined the effect of a uniform 10% vehicle mass reduction in the UK and concluded this would reduce fatalities in single-car, two-car, and car-pedestrian crashes [95]. A critical review of the literature therefore, does not lead to a broad consensus on the role of vehicle mass in traffic safety. The basic underlying reason for this is that traffic safety is the net result of a large number of important variables and complex relationships that have constantly changed over time, and virtually all studies have important limitations. While every study makes some contribution, every study also possesses one or more of the following limitations:  14  1. Aggregation. Data is aggregated in one or more important ways: o  Crash event aggregation: quantitative relationships are not based on individual crash data, but aggregated into groups leaving open the possibility of spurious findings due to the ecological fallacy (e.g., [102]).  o  Spatial aggregation: crash data are aggregated at a state (in the U.S.) or national level which combines data across important dimensions such as differing regulations for licensing and traffic enforcement, urban and rural road types, or other variables (e.g., [96]).  o  Crash type aggregation: crash data are aggregated across multiple types of crash events (e.g., car-heavy goods, single-car, two-car, car-pedestrian). For example, two frequently cited U.S. studies both aggregated crash data for car collisions with pedestrians, motorcycles, and bicyclists, even though these three types of crashes are very different [28, 103].  o  Time aggregation: multiple crash years are aggregated leaving many potential confounders such as changes in roads, speed enforcement, or vehicle safety and technology. For example, one study that analyzed the role of mass and size included FARS data over the time period of 1975-1998 [28, 103, 107].  2. Choice of risk metric. Two important risk metric choices influence the ability to generalize traffic safety research: o  Risk type: absolute risk, relative risk (including risk ratio), or conditional risk are available options. Absolute risk (e.g., annual fatalities) is a common choice, but does not provide insights into the distribution of risk such as the relative risk between drivers of heavier (higher CO2/km) cars and lighter (lower CO2/km) cars. Risk equity between two groups can be measured by ratios, but ratios can vary without changing absolute risks. Conditional risk studies provide valuable insights into crashworthiness, but not crash avoidance.  o  Casualty type: fatalities and various degrees of injury are the available choices. Fatalities tend to be the metric of choice most often used, yet this choice excludes injury rate which is also critical to achieving traffic safety goals.  15  3. Vehicle and crash factors. Basic vehicle parameters such as mass, size, stiffness, and height affect safety, as well as specific safety equipment such as side air bags and electronic stability control [94, 97, 109-112]. For example, none of the aforementioned studies on the role of vehicle mass and traffic specifically adjusted risk estimates based on electronic stability control, even though this technology has been selectively adopted over time since the early 1990s and has been found to reduce fatal single-car crashes by 30-70%, and fatal rollover crashes by 70-90% [111]. Collision speed has been shown to be a critical crash factor in determining casualty severity [101, 113-115], yet most databases have only indirect indicators of collision speed such as road speed limit. Crash mode (points of impact, rollovers, hit objects) is yet another factor that substantially influences risks, but is not always adjusted in risk estimates [28, 102, 103, 116]. 4. Driver behavior. While there is wide agreement in the traffic safety literature on the critical importance of driver behavior as a determining factor in risk [102, 103, 105, 117-129], in practice it is difficult to adjust risk estimates for behavior and this correction factor is seldom used. Seat belt use, previous traffic citations, intoxication, and related variables have been employed in various analyses. While two U.S. studies [28, 103] have been compared [28, 103, 130], the comparisons have failed to mention that one study incorporated a relatively detailed sensitivity analysis using 9 behavioral variables [28], while the other did not adjust for behavior beyond basic age and sex indicators [103]. Others have claimed to have ruled out “non-subtle” behavior effects in their analysis [102], but this conclusion was derived by comparing aggregate descriptive statistics for a relatively small set of behavior variables. 5. Factors other than road environment, vehicle, and driver. In general, traffic safety risks can be viewed as a Haddon matrix which incorporates factors other than what is normally associated with the crash environment, vehicle, and driver [131]. For example, it has been argued that one-third of the reduction in fatalities in the UK may be due to more rapid response to serious collisions and new medical technologies [132].  16  A critical question in the development of policies to improve traffic safety and climate mitigation from passenger cars is: can vehicle mass be reduced without compromising safety? The balance of the research literature indicates the answer is “yes.” But it is a conditional “yes” because it matters how mass is reduced. For example, the mass of individual vehicles can be reduced by eliminating safety equipment such as side air bag systems and electronic stability control components, but the safety tradeoffs would likely be negative. Average, on-road fleet mass can be reduced by introducing large numbers of the smallest car models while allowing numbers of large cars to grow rapidly, but this likely would have negative implications for the distribution of risk. While safety, design and material use is rapidly diffused through all vehicle types, and given similar rates of fleet turnover in different classes, vehicle mass will continue to be one of many important variables in traffic safety, and also directly influence vehicle CO2 emission rates. Thus further research into the relationships between vehicle mass, CO2 emissions, and traffic safety is warranted.  1.5  HOW THE THESIS CHAPTERS FIT TOGETHER  The overall objective of this thesis is to assess the relationships between climate mitigation policy for passenger cars with case studies of the two largest public health risks: air quality and traffic safety. Before embarking on the two case studies, it was important to set the broader policy context first, emphasizing Integrated Assessment methods as the cornerstone of policy analysis and design; this is accomplished in Chapter 2. Subsequently Chapter 3 provides a case study on air quality and CO2 emissions, while Chapter 4 provides the traffic safety case study. Chapter 5 provides conclusions including relationship of the policy, air quality, and traffic safety analyses, discussion of my thesis research in relation to current research hypotheses, strengths and weaknesses, significance, and recommendations for further research.  17  1.6 1.  REFERENCES Pacala, S. and R. Socolow, Stabilization wedges: solving the climate problem for the next 50 years with current technologies. Science, 2004. 305: p. 968-972.  2.  Ribeiro, K., et al., Transport and its infrastructure, in Climate Change 2007: Mitigation. Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. 2007, Cambridge University Press.  3.  Parry, M., et al., Technical Summary. Climate Change 2007: Impacts, Adaptation and Vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. 2007: Cambridge, UK.  4.  London. Mayor's Priorites. 2009 [cited September 20, 2009]; Available from: http://www.london.gov.uk/mayor/priorities/environment.jsp.  5.  HM-Government, The UK Low Carbon Transition Plan. 2009.  6.  Keith, D., Geoengineering. Nature, 2001. 409: p. 420.  7.  DTI, Meeting the energy challenge: a white paper on energy (May 2007). 2007, Department of Trade and Industry.  8.  DECC. UK Department of Energy & Climate Change. 2009 [cited; Available from: www.decc.gov.uk.  9.  CFIT, Transport and climate change. 2007, Commission for Integrated Transport (www.cfit.gov.uk).  10.  DEFRA, The UK Government Sustainable Development Strategy. 2005, UK Department for Environment, Food and Rural Affairs (DEFRA).  11.  DFT, Delivering better transport: priorities for 2007-08 and beyond. 2007, UK Department for Transport.  12.  HM-Treasury, Budget 2007 - Building Britain’s long-term future: prosperity and fairness for families. 2007, HM Treasury (UK).  13.  HM-Government. Office for Low Emission Vehicles. 2009 [cited; Available from: http://interactive.bis.gov.uk/lowcarbon/.  14.  MORI, Assessing the impact of Graduated Vehicle Excise Duty: quantitative report. 2003, UK Department for Transport (www.dft.gov.uk) and MORI (www.mori.co.uk).  15.  IR, Report on the evaluation of the company car tax reform: stage two. 2006, UK Inland Revenue (www.inlandrevenue.gov.uk).  18  16.  DECC, Statistics about climate change. 2009, Department of Energy & Climate Change.  17.  Fontaras, G. and Z. Samaras, A quantitative analysis of the European Automakers’ voluntary commitment to reduce CO2 emissions from new passenger cars based on independent experimental data. Energy Policy, 2007. 35: p. 2239-2248.  18.  DFT, UK Transport Statistics online database (www.dft.gov.uk/pgr/statistics). 2008, UK Department for Transport.  19.  NAEI, National Atmospheric Emission Inventory online database (www.naei.org.uk). 2008.  20.  Evans, L. and M. Frick, Mass ratio and relative driver fatality risk in two-vehicle crashes. Accident Analysis and Prevention, 1993. 25(2): p. 213-224.  21.  JATO. Jato Dynamics new car database for Great Britain. 2007 [cited June 3, 2007].  22.  Verboven, F., Quality-based tax discrimination and tax incidence: evidence from gasoline and diesel cars. The RAND Journal of Economics, 2002. 33(2): p. 275-297.  23.  Lane, B., Car buyer research report: consumer attitudes to low carbon and fuel-efficient passenger cars. 2005, Low Carbon Vehicle Partnership (UK).  24.  SMMT, New car CO2 report 2008. 2008, Society of Motor Manufacturers and Traders Ltd. (www.smmt.co.uk).  25.  DeCicco, J., F. An, and M. Ross, Technical options for improving the fuel economy of U.S. cars and light trucks by 2010–2015. 2001, American Council for an Energy-Efficient Economy www.aceee.org.  26.  Fulton, L., P. Cazzola, and F. Cuenot, IEA Mobility Model (MoMo) and its use in the ETP 2008. Energy Policy, 2009. 37: p. 3758-3768.  27.  Sperling, D. and D. Gordon, Two Billion Cars: Driving Toward Sustainability. 2009: Oxford University Press.  28.  Kahane, C., Vehicle weight, fatality risk and crash compatibility of model year 1991-99 passenger cars and light trucks. 2003, National Highway Traffic Safety Administration (U.S.).  29.  SMMT, Motor industry facts - 2008. 2008, Society of Motor Manufacturers and Traders (SMMT) www.smmt.co.uk.  30.  COMEAP, Quantification of the effects of air pollution on health in the United Kingdom. 1998: The Stationary Office London (www.tsonline.co.uk).  31.  Dargay, J. and M. Hanly, Volatility of car ownership, commuting mode and time in the UK. Transportation Research Part A: Policy and Practice, 2007. 41: p. 934-948.  19  32.  Sorrell, S., The rebound effect: an assessment of the evidence for economy-wide energy savings from improved energy efficiency. 2007, UK Energy Research Centre (www.ukerc.ac.uk).  33.  WHO, Environmental burden of disease: Country profiles (United Kingdom). 2007, World Health Organization.  34.  Parry, I., M. Walls, and W. Harrington, Automobile externalities and policies. Journal of Economic Literature, 2007. XLV: p. 373-399.  35.  Woodcock, J., et al., Energy and Health Series: Energy and Transport. The Lancet, 2007. 370: p. 1078-1088.  36.  Jacoby, H., R. Prinn, and R. Schmalensee, Kyoto's unfinished business. Foreign Affairs, 1998. July/August: p. 54-66.  37.  Levy, D. and S. Rothenberg, Heterogeneity and Change in Environmental Strategy: Technological and Political Responses to Climate Change in the Global Automobile Industry, in Organizations, Policy and the Natural Environment: Institutional and Strategic Perspectives, A. Hoffman and M. Ventresca, Editors. 2002, Stanford University Press.  38.  Morgan, M.G. and M. Henrion, Uncertainty: a guide to dealing with uncertainty in quantitative risk and policy analysis. 1992: Cambridge University Press.  39.  Morgan, M.G., et al., Why Conventional Tools for Policy Analysis are Often Inadequate for Problems of Global Change. Climatic Change, 1999. 41: p. 271-281.  40.  SMMT, New vehicle registrations in the UK. 2009, Society of Motor Manufacturers and Traders (www.smmt.co.uk).  41.  Mazzi, E. and H. Dowlatabadi, Air quality impacts of climate mitigation: UK policy and passenger vehicle choice. Environmental Science & Technology, 2007. 41: p. 387-392.  42.  DOH, Health statistics online data. 2006, UK Department of Health.  43.  GAD, Government Actuary's Department (GAD) online statistics. 2006, United Kingdom Government Actuary's Department (GAD).  44.  Hayman, G., et al., Modelling of Tropospheric Ozone (AEAT/ENV/R/1858 issue 2). 2005, Department for Environment, Food and Rural Affairs, the Scottish Executive, the National Assembly of Wales and the Northern Ireland Department of the Environment.  45.  Hayman, G., et al., Modelling of Tropospheric Ozone Formation. 2002: prepared by AEA Technology for UK Department for Environment, Food and Rural Affairs (DEFRA).  46.  Stedman, J., et al., Revised PM10 projections for the UK for PM10 objective analysis. 2002: prepared by AEA Technology for UK Department for Environment, Food and Rural Affairs  20  (DEFRA) and The National Assembly for Wales, The Scottish Executive and the Department of the Environment in Northern Ireland. 47.  Stedman, J., et al., Quantification of the health effects of air pollution in the UK for revised PM10 objective analysis. 2002: AEA Technology and UK Department of Health.  48.  Stedman, J., et al., Baseline PM10 and NOx projections for PM10 objective analysis. 2001: prepared by AEA Technology for UK Department for Environment, Food and Rural Affairs (DEFRA) and The National Assembly for Wales, The Scottish Executive and the Department of the Environment in Northern Ireland.  49.  Dockery, D., et al., An association between air pollution and mortality in six U.S. cities. New England Journal of Medicine, 1993. 329: p. 1753-1759.  50.  Pope, C.A., et al., Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. Journal of the American Medical Association, 2002. 287(9): p. 11321141.  51.  COMEAP, Statement on long-term effects of particles on mortality. 2001: UK committee on the medical effects of air pollutants (www.advisorybodies.doh.gov.uk/comeap).  52.  DFT, Make-Model crash database linkable to UK Data Archive data (electronic database). 2006, UK Department for Transport.  53.  DFT, Road Accident Data [computer files] 1994 SN3461, 1996 SN3627, 1997 SN3817, 1998 SN4027, 1999 SN4080, 2000 SN4356, 2001 SN4501, 2002 SN4588, 2003 SN4979, 2004 SN5042, 2005 SN5244, 2006 SN5493. 2007, UK Data Archive (Colchester, Essex).  54.  Evans, L., Traffic Safety. 2004: Science Serving Society.  55.  DFT, Cars: make and model: the risk of driver injury in Great Britain: 2000-2004. 2006, Department for Transport (UK).  56.  Mengert, P. and S. Borener, Overall fatality risk to the public at large related to national weight mix of passenger cars (DOT-TSC-HS070-PM-89-27). 1989, National Highway Traffic Safety Administration (NHTSA).  57.  Lumina, Analytica Professional for Windows 4.1.2.4 User Guide. 2008.  58.  Globalcar, Online car specifications (www.globalcar.com). 2007.  59.  Street-car, Online car specifications (www.street-car.net). 2007.  60.  Whatcar, Online car specifications (www.whatcar.co.uk). 2007.  61.  Cambridge.Systematics.Inc., Moving Cooler: An Analysis of Transportation Strategies for Reducing Greenhouse Gas Emissions. 2009, Urban Land Institute: Washington, D.C.  21  62.  Kopp, R., Transport policies to reduce CO2 emissions from the light-duty vehicle fleet (Issue Brief 12), in Assessing U.S. Climate Policy Options. 2007, Resources for the Future: Washington, DC.  63.  Keefe, R., J. Griffin, and J. Graham, The Benefits and Costs of New Fuels and Engines for Light-Duty Vehicles in the United States. Risk Analysis, 2008. 28(5): p. 1141-1154.  64.  Sansom, T., et al., Surface transport costs and charges: Great Britain 1998. 2001, Department of the Environment, Transport and the Regions (DETR).  65.  DFT, Carbon and sustainability reporting within the renewable transport fuel obligation (RTFO). 2008, UK Department for Transport.  66.  Onoda, T., IEA policies - G8 recommendations and an afterwards. Energy Policy, 2009. 37: p. 3823–3831.  67.  DEFRA, Online database of GHG emissions (http://www.defra.gov.uk/environment/statistics/globatmos/alltables.htm). 2008, Department for Environment, Food and Rural Affairs.  68.  Lave, L., Conflicting objectives in regulating the automobile. Science, 1981. 212(893-899).  69.  Lave, L., Controlling contradictions among regulations. The American Economic Review, 1984. 74(3): p. 471-475.  70.  Maclean, H. and L. Lave, Life cycle assessments of automobile/fuel options. Environmental Science & Technology, 2003. 37(5445-5452).  71.  Bonilla, D., Fuel demand on UK roads and dieselisation of fuel economy. Energy Policy, 2009. 37: p. 3769-3778.  72.  Kohler, J., Y. Jin, and T. Barker, Integrated modelling of EU transport policy. Journal of Transport Economics and Policy, 2008. 42(1): p. 1-21.  73.  Ogden, J., R. Williams, and E. Larson, Societal lifecycle costs of cars with alternative fuels/engines. Energy Policy, 2004. 32(7-27).  74.  Litman, T., Integrating public health objectives in transportation decision-making American Journal of Health Promotion, 2003. 18(1): p. 103-109.  75.  ARB, California state motor vehicles pollution control standards; Request for waiver of federal preemption; Opportunity for Public Hearing, Docket ID #EPA-HQ-OAR2006-0173; 72 FR 21260 (April 30, 2007). 2007, California Air Resources Board.  76.  T&E, Danger ahead: Why weight-based CO2 standards will make Europe's car fleet dirtier and less safe. 2007, European Federation for Transport and Environment: Brussels, Belgium.  22  77.  Hull, A., Policy integration: what will it take to achieve more sustainable transport solutions in cities? Journal of Transport Economics and Policy, 2007. 15: p. 94-103.  78.  Lund, H. and E. Munster, Integrated transportation and energy sector CO2 emission control strategies. Transport Policy, 2006. 13: p. 426–433.  79.  Chatterjee, K. and A. Gordon, Planning for an unpredictable future: Transport in Great Britain in 2030. Transport Policy, 2006. 13: p. 254–264.  80.  Dowlatabadi, H., On integration of policies for climate and global change. Mitigation and Adaptation Strategies for Global Change, 2007. 12: p. 651–663.  81.  Maclean, H. and L. Lave, Evaluating automobile fuel/propulsion system technologies. Progress in Energy and Combustion Science, 2003. 29: p. 1–69.  82.  Cifuentes, L., et al., Hidden Health Benefits of Greenhouse Gas Mitigation. Science, 2001. 293: p. 1257-1259.  83.  Davis, D., et al., Short-term Improvements in Public Health from Global-Climate policies on Fossil-Fuel Combustion: An Interim Report. The Lancet, 1997. 350: p. 1341-1349.  84.  EEA, Air quality and ancillary benefits of climate change policies (EEA Technical report No 4/2006). 2006, European Environment Agency (EEA): Copenhagen, Denmark.  85.  IPCC, Climate Change 2001: Mitigation. 2001: Intergovernmental Panel on Climate Change (www.ipcc.ch).  86.  van Vuuren, D., et al., Exploring the ancillary benefits of the Kyoto Protocol for air pollution in Europe. Energy Policy, 2006. 34(4): p. 444-460.  87.  Mayeres, I. and S. Proost, Should diesel cars in Europe be discouraged? Regional Science and Urban Economics 2001. 31: p. 453–470.  88.  Swart, R., et al., A Good Climate for Clean Air: Linkages between Climate Change and Air Pollution. Climatic Change, 2004. 66: p. 263-269.  89.  Kavalov, B. and S. Peteves, Impacts of increasing automotive diesel consumption in the EU, I.f.E. Directorate General Joint Research Centre (DG JRC), Editor. 2004, European Commission  90.  Jacobson, M., et al., The effect on photochemical smog of converting the U.S. fleet of gasoline vehicles to modern diesel vehicles. Geophysical Research Letters, 2004. 31: p. L02116.  91.  Jacobson, M., Correction to ‘‘Control of fossil-fuel particulate black carbon and organic matter, possibly the most effective method of slowing global warming’’. Geophysical Research Letters, 2005. 110: p. D14105.  23  92.  Cifuentes, L., et al., Assessing the Health Benefits of Urban Air Pollution Reductions Associated with Climate Change Mitigation (2000-2020): Santiago, Sao Paulo, Mexico City, and New York City. Environmental Health Perspectives, 2001. 109(Supplement 3): p. 419425.  93.  Moon, P., A. Burnham, and M. Wang. Vehicle-cycle energy and emission effects of conventional and advanced vehicles (2006-01-0375). in 2006 World Congress. 2006. Detroit, Michigan: Society of Automotive Engineers (www.sae.org).  94.  Tolouei, R. and H. Titheridge, Vehicle mass as a determinant of fuel consumption and secondary safety performance. Transportation Research Part D: Transport and Environment, 2009. 14: p. 385-399.  95.  Broughton, J., The likely effects of downsizing on casualties in car accidents. 1999, Transport Research Laboratory (UK).  96.  Crandall, R. and J. Graham, The effect of fuel economy standards on automobile safety. Journal of Law & Economics, 1989. 32: p. 97-118.  97.  Evans, L. and M. Frick, Car size or car mass: which has greater influence on fatality risk? American Journal of Public Health, 1992. 82(8): p. 1105-1112.  98.  Noland, R., Motor vehicle fuel efficiency and traffic fatalities. The Energy Journal, 2004. 25(4): p. 1-22.  99.  Robertson, L., Blood and oil: vehicle characteristics in relation to fatality risk and fuel economy. American Journal of Public Health, 2006. 96(11): p. 1906-1909.  100.  Thomas, P. and R. Frampton, Car size in UK crashes: the effects of user characteristics, impact configuration, and the patterns of injury. Traffic Injury Prevention, 2002. 3: p. 275-282.  101.  Toy, E. and J. Hammitt, Safety impacts of SUVs, vans, and pickup trucks in two-vehicle crashes. Risk Analysis, 2003. 23(4): p. 641-650.  102.  Wenzel, T. and M. Ross, The effects of vehicle model and driver behavior on risk. Accident Analysis and Prevention, 2005. 37: p. 479-494.  103.  van Auken, R. and J. Zellner, A further assessment of the effects of vehicle weight and size parameters on fatality risk in model year 1985-98 passenger cars and 1985-97 light trucks. 2003, Dynamic Research Inc.  104.  Bedard, M., et al., The independent contribution of driver, crash, and vehicle characteristics to driver fatalities. Accident Analysis and Prevention, 2002. 34: p. 717-727.  105.  Evans, L., Comment: the dominant role of driver behavior in traffic safety. American Journal of Public Health, 1996. 86(6): p. 784-786.  24  106.  NHTSA, Final regulatory impact analysis: corporate average fuel economy and CAFE reform for MY 2008-2011 light trucks. 2006, National Highway Traffic Safety Administration.  107.  Wood, D., C. Simms, and C. Glynn. Two car frontal collisions: the role of car mass, collision speed distribution, and frontal stiffness on occupant fatality and injury. in 2005 International IRCOBI Conference on the Biomechanics of Impact. 2005. Prague: International Research Council on the Biomechanics of Impact (IRCOBI).  108.  Padmanaban, J. Influences of vehicle size and mass and selected driver factors on odds of driver fatality. in 47th Annual Proceedings of the Association for the Advancement of Automotive Medicine. 2003.  109.  Austin, R., Vehicle aggressiveness in real world crashes. 2005, National Highway Traffic Safety Administration (U.S.).  110.  Evans, L. How to make cars lighter and safer (2004-01-1172). 2004: Soceity of Automotive Engineers.  111.  Ferguson, S., The effectiveness of electronic stability control in reducing real-world crashes: a literature review. Traffic Injury Prevention, 2007. 8: p. 329-338.  112.  Robertson, L., How to save fuel and reduce injuries in automobiles. Journal of Trauma, 1991. 31(1): p. 107-109.  113.  Joksch, H., Velocity change and fatality risk in a crash - a rule of thumb. Accident Analysis and Prevention, 1993. 25: p. 103-104.  114.  Mosedale, J. and A. Purdy, Excessive speed as a contributory factor to personal injury road accidents. 2003, United Kingdom Department for Transport.  115.  Patterson, T., et al., The effect of increasing rural interstate speed limits in the United States. Traffic Injury Prevention, 2002. 3: p. 316-320.  116.  Evans, L., Causal influence of car mass and size on driver fatality risk. American Journal of Public Health, 2001. 91: p. 1076-1081.  117.  Andersson, A., O. Bunketorp, and P. Allebeck, High rates of psychosocial complications after road traffic injuries. Injury, 1997. 28(8): p. 539-543.  118.  Dobson, A., et al., Women drivers’ behaviour, socio-demographic characteristics and accidents. Accident Analysis and Prevention, 1999. 31: p. 525-535.  119.  Evans, L., Safety-belt effectiveness: the influence of crash severity and selective recruitment. Accident Analysis and Prevention, 1996. 28(4): p. 423-433.  25  120.  Fernandes, R., R.F. Soames-Job, and S. Hatfield, A challenge to the assumed generalizability of prediction and countermeasure for risky driving: different factors predict different risky driving behaviors. Journal of Safety Research, 2007. 38: p. 59-70.  121.  Klauer, S., et al., The impact of driver inattention on near-crash/crash risk: an analysis using the 100-car naturalistic driving study data. 2006, National Highway Traffic Safety Administration.  122.  Krahe, B. and I. Fenske, Predicting aggressive driving behavior: the role of macho personality, age, and power of car. Aggressive Behavior, 2002. 28: p. 21-29.  123.  Langford, J. and S. Koppel, Epidemiology of older driver crashes – identifying older driver risk factors and exposure patterns. Transportation Research, 2006. Part F 9: p. 309-321.  124.  Lourens, P., J. Vissers, and M. Jessurun, Annual mileage, driving violations, and accident involvement in relation to drivers’ sex, age, and level of education. Accident Analysis and Prevention, 1999. 31: p. 593-597.  125.  Redelmeir, D., R. Tibshirani, and L. Evans, Traffic-law enforcement and risk of death from motor-vehicle crashes: case-crossover study. The Lancet, 2003. 361: p. 2177-2182.  126.  Shope, J., Influences on youthful driving behavior and their potential for guiding interventions to reduce crashes. Injury Prevention, 2006. 12(Suppl I): p. i9-i14.  127.  Vingilis, E. and S. Macdonald, Review: drugs and traffic collisions. Traffic Injury Prevention, 2002. 3: p. 1-11.  128.  Warner, H. and L. Aberg, Drivers' decision to speed: a study inspired by the theory of planned behavior. Transportation Research, 2006. Part F 9: p. 427-433.  129.  Yun, J., Offsetting behavior effects of the corporate average fuel economy standards. Economic Inquiry, 2002. 40(2): p. 260-270.  130.  Wenzel, T. and M. Ross, Increasing the fuel economy and safety of new light-duty vehicles. 2006, U.S. Department of Energy under Contract No. DE-AC03-76SF00098.  131.  O'Neill, B. and D. Mohan, Reducing motor vehicle crash deaths and injuries in newly motorized countries. British Medical Journal, 2002. 324(7346): p. 1142-1145.  132.  Noland, R. and M. Quddus, Improvements in medical care and technology and reductions in traffic-related fatalities in Great Britain. Accident Analysis and Prevention, 2004. 36: p. 103113.  26  2  INTEGRATED ASSESSMENT OF MULTIPLE RISKS TO ASSESS CURRENT AND FUTURE CLIMATE MITIGATION POLICIES FOR PASSENGER CARS3  2.1  INTRODUCTION  Policymakers in industrialized countries are increasingly tasked to design and implement climate mitigation policy for passenger cars [1-3]. Car fleets in developed countries are currently dominated by conventional gasoline and diesel technology [3, 4], and risk assessments incorporating social costs commonly assume 100% conventional technology [5, 6]. Integrated Assessment of multiple risks is growing in importance because industrialized countries are presently at a crossroads in fuel and drive-train choices. Scenarios for the near to distant future commonly project a much more heterogeneous mix of fuels and drive-train technologies. Examples from the International Energy Agency’s (IEA) current suite of transportation scenarios include steep uptake of plug-in hybrids beginning in 2015, 90% electric or fuel cell vehicles by 2050, and a range of advanced conventional cars (defined as 50-70% better fuel consumption) rising as high as 90% to as low as 10% [7]. While greenhouse gas reduction is one principal driver of these transitions, climate change is just one of many risks associated with passenger car use, and the effectiveness of new climate policies will depend on how well they balance climate change mitigation with other risks [8]. In designing policies to achieve transitions such as envisioned by IEA, it is clear that risk assessment models will need to adapt. In this study I examine the rationale, and present a basic framework, for Integrated Assessment models to incorporate multiple risks from a societal perspective in the analysis of climate mitigation policies for passenger cars. Using the UK as a case study, I illustrate the importance of decision criteria, risks included (or excluded), and how the choice of fuel and technology options influences system boundaries and relative ranking of multiple risks. The UK is a useful case study because UK circumstances are generally applicable to other developed countries (albeit not completely). The UK was an early adopter of climate mitigation policies for passenger cars with nationwide tax regimes in place since 2001-2002 [9, 10]. As in other industrialized countries passenger car ownership and road transport CO2 emissions  3  A version of this chapter had been submitted for publication. Mazzi, E. “Integrated Assessment of Multiple  Risks to Assess Current and Future Climate Mitigation Policies For Passenger Cars,” 2009. Based on various reviewer feedback, it is being substantially revised for re-submission to an appropriate journal.  27  continue to rise [11]. The UK Department for Transport (DFT) forecasts ownership levels relative to the year 2000 of 122% by 2015, and 133% by 2025, a growth rate that is in excess of the population growth rate [12]. Currently the UK car fleet is dominated by spark-ignition gasoline and compression-ignition diesel cars, comprising 99.3% of all new registrations [13]; however, alternative drive-trains, such as grid-independent hybrids and alternative- fuelled vehicles, such as ethanol-powered cars (E85), are already rapidly gaining market share [13]. Government policies such as the UK Renewable Transport Fuels Obligation [14] and voluntary and regulatory CO2 measures in Europe are expected to continue to stimulate the proliferation of alternative vehicle technologies and fuel types [15]. The UK is similar to other developed countries in its struggle to limit CO2 from transportation. From 1990 to 2005 CO2 emissions were reduced for every major sector, except transportation where emissions have risen by 12% [16]. Because of the importance of transportation in emissions, with passenger cars as a major contributor, multiple government agencies in the UK have identified climate mitigation for the road transportation sector as a high priority [17-21]. The pitfalls of focusing on a single problem in policy design and analysis for passenger cars have long been established [22]. The array of regulations and policies affecting passenger cars is often in tension, and not always consistent with green design principles [23, 24]. A critical issue in Integrated Assessment of multiple risks is that analysis of fuels and vehicle technology alone is insufficient because ownership (cars per capita), annual vehicle kilometers traveled per car (VKT) and patterns of use (e.g., time of day and location) are essential factors that influence risks. A recent example from the UK is that steep growth of diesel cars has not yielded expected savings in fuel consumption and CO2 due to consumer choice of larger cars, driver behavior (speed), and VKT rebound effects [25]. Another example is the ongoing debate about fuel economy standards in the U.S. Considering only the principal environmental risks (air quality and climate change) and energy security, more stringent fuel economy standards are strongly beneficial. However, when considering traffic safety risks via disparate vehicle mass and economic risks of traffic congestion via VKT rebound effects, the likelihood of realizing net social benefits with more stringent fuel economy standards is less clear. While the need for Integrated Assessment in transportation policy is often acknowledged [8], much research has employed Integrated Assessment of multiple risks but applied one decision criterion such as social costbenefit [5, 26, 27]. Other research has focused on integration of near-term public health risks but has not incorporated other risks such as climate change or congestion [11, 28]. Stakeholders involved in passenger car policy development are increasingly sophisticated in recognizing the interdependencies between fuels, technology, and vehicle ownership and VKT as well as the relevance of a multiple risk framework. For example, U.S. car manufacturers 28  argued that fleet turnover effects (the rate of change of in-use technology over time) and rebound effects (i.e., VKT) might nullify welfare gains from California’s proposed CO2 standards for passenger cars [29]. In Europe, proposals by the European Commission for car CO2 standards differentiated by mass classifications were refuted by a leading environmental organization on the basis of traffic safety concerns [30]. Therefore, robust Integrated Assessment of multiple risks in policy development can reasonably be expected to improve the social and political acceptability of new policies. There are inherent limitations in focusing primarily on Integrated Assessment of multiple risks. A truly comprehensive Integrated Assessment would explore the best strategy for meeting transportation needs within the context of all relevant local, social, and physical conditions. Additional domains that are relevant include organizational and institutional domains [31], interactions with other energy-intensive sectors [32], integration of transportation and land use planning [33], and integration of passenger car travel with other travel modes [11]. In the larger picture, the integration of climate change with other drivers of global change is relevant [34]. Another limitation of this study is that it focuses on risks and does not examine the positive attributes of consumer choice [35]. As an example, the travel range for a single refueling of a gasoline-powered car is much greater than an all-electric car, which is a positive attribute strongly valued by consumers. Thus the relative benefits of fuels and technologies to consumers are excluded [4]. In this study I first review the importance of identifying the risks of concern and choice of decision criteria in public policy affecting passenger car fuels, technology, ownership, and use. Next, I describe the key linkages between policies and the chosen set of risks, and illustrate how the pathways linking policies and risks depend on the types of technologies or fuels being analyzed. I conclude with summary remarks on development of policy.  2.2  IDENTIFYING RELEVANT RISKS AND CHOOSING DECISION CRITERIA  Following Morgan and Henrion [36], it is emphasized that in the early phases of policy analysis the choice of decision criteria must be made and the principal risks of concern identified, and that these choices must be made iteratively throughout the policy analysis. When preparing to assess new or modified passenger car policies, identifying relevant risks is strongly dependent upon choice of decision criteria. I discuss three different approaches to choosing decision criteria and the associated risk priorities that result, primarily using the UK as a case study: (1) economic, utility-based approach based on marginal external costs, (2) public health approach based on various mortality and morbidity metrics, and (3) criteria based on public agency new policy  29  priorities or rights-based constrained risks. Each of these decision criteria approaches is valid, yet each can lead to different policy and risk mitigation priorities as summarized in Table 2.1. Table 2.1  Approximate ranking of risks in the UK for different decision criteria, based on the fleet dominated by spark-ignition gasoline and compression-ignition diesel passenger cars. Economic External  Public Health  Social Costs Primary  Congestion  New Public Policy Priorities  Traffic Safety, Air  Climate Change  Quality Secondary  Air Quality, Traffic  Environmental Noise  Safety  Congestion, Energy Security, Climate Change, Air Quality  Tertiary  Environmental Noise,  Congestion, Energy  Environmental  Climate Change,  Security, Climate  Noise  Energy Security  Change  Water Security, Water Quality, Land  often excluded  Use, Food Security  2.2.1  DECISION CRITERIA BASED ON MARGINAL EXTERNAL COSTS  I illustrate the relationship between relevant risks and decision criteria based on the current UK passenger car fleet that is dominated by spark ignition gasoline and compression ignition diesel cars. However, it is noted that the set of risks is also strongly affected by the choice of fuels and technology included in the analysis. For example, when considering biofuel-powered vehicle technology on a large scale, food security, water security, land use, and water quality risks rise in importance [37] while improvements in air quality and climate change remain uncertain [38, 39]. Scenarios that include coal-to-liquids fuels substantially mitigate energy security risks for countries that have large coal reserves but can exacerbate other risks such as air quality, water quality, ecosystem destruction and climate change [40]. The first comprehensive estimate of marginal external costs of transportation for the UK was published by Peirson and colleagues [41]. The most recent UK estimates are summarized in Figure 2.1 [6], and are generally similar to figures for the U.S. except that external costs of congestion, air quality, and traffic collisions are closer [5]. The most striking result of using social costing as the decision criterion in the UK is that the costs of congestion dominate all 30  environmental and health risks, including climate change which is assessed to be one order of magnitude lower. This is precisely why the VKT rebound effect is critical from an economic externality perspective; a relatively small increase in VKT occurring in urban centers at peak times may overwhelm estimated public welfare benefits from reducing CO2, noise, air pollutants or traffic safety risks. In fact, if one adopts economic externalities as the sole decision criterion, then arguably climate change risks could be omitted in the analysis of passenger car policies. Figure 2.1  Marginal external social costs (pence per kilometer) in Great Britain adapted directly from reference [6]. Error bars represent the estimated  marginal external cost, pence per km travel .  range of the marginal external costs for each risk.  12 p/km 10 p/km 8 p/km 6 p/km 4 p/km 2 p/km 0 p/km global warming  noise pollution  local air quality  traffic congestion collision risks  Like all single-metric approaches, externality costing has its limitations and criticisms such as sensitivity to assumptions about internalized risk, inadequate accounting of equity, and large uncertainties particularly due to choice of value of statistical life and discount rates [6, 42-46]. It is noted that motor vehicle transport also imposes physical inactivity risks [46], but quantifying this risk is only relevant when comparing modal shifts such as to cycling and walking. Physical inactivity risks do not change when comparing passenger car technology options against each other because all options equally promote physical inactivity. 2.2.2  DECISION CRITERIA BASED ON PUBLIC HEALTH METRICS  Developing risk priorities from a public health perspective using mortality and morbidity produces a different ranking compared to externality costing. The top three priorities are clearly traffic safety, air quality, and noise. Climate change health risks within the UK have not yet been 31  quantified, and there is no indication the direct risks will be of similar magnitude to traffic safety or air quality in the near future, although these risks are expected to grow more rapidly than others over time [47]. Traffic congestion could be expected to induce some health risks such as increased exposure to air pollution, but congestion impacts on collision injuries and noise is difficult to assess. Where there is congestion, there is a slowing of traffic [48] and when collisions occur they are likely to be less injurious [49]. However, congestion can also lead to more aggressive driving and excess speeds to escape the congested area. Similarly mixed effects are expected in environmental noise health effects by increasing time of exposure but reducing peak noise levels because of slower speeds [50]. The relative ranking among the top three health risks in the UK depends upon the choice of metric. Choosing annual mortality as the metric, air quality is the top priority. In 2005, annual mortality in the UK due to urban air pollution (of which transport represents over 80% [51]) was estimated to be 12,400 due to particulate matter and 700 due to ozone [52, 53]. Traffic fatalities in 2005 were 3,201 [54]. For environmental noise, no UK-specific burden of disease estimates are yet available [55]; however, scaling estimates made for all EU-25 countries of 50,000 premature deaths annually [50] to the UK indicates roughly 6,000 deaths brought forward annually. Choosing years of life lost (YLL) as the metric, traffic safety becomes the top priority due to a high proportion of young casualties. Using UK life tables [56], traffic safety resulted in 127,000 YLL in 2005. Assuming 9 years of life lost per statistical death due to particulate matter [57, 58] and 1 year due to ozone [59], air pollution results in 112,000 YLL. A rough estimate for noise is 60,000 YLL, assuming 10 years of life lost per statistical death [60]. A further extension of YLL is to add morbidity effects to quantify disability adjusted life years (DALYs) [61]. I found no estimates comparing DALYs for all environmental health risks in the UK; however, estimates for the Netherlands show that air pollution DALYs are twice that of traffic safety, while traffic safety DALYs are twice that of environmental noise [60]. 2.2.3  DECISION CRITERIA BASED ON NEW PUBLIC POLICY PRIORITIES  While climate change tends to subordinate to other risks when applying economic externality or public health decision criteria, indications are that climate change ranks highest from the perspective of new public policy priorities. Multiple UK government agencies identify climate change as a high priority while interrelated risks such as air quality and traffic congestion remain important but are emphasized less than climate change [17-21]. The CO2 reduction targets set by government agencies can be considered to be largely rights-based, constrained-risk criteria.  32  2.2.4  CHOOSING THE APPROPRIATE DECISION CRITERIA  To summarize, Table 2.1 shows the relative rankings of risks based on choice of decision criteria. The example of the UK demonstrates there can be substantial differences in the ranking of risks and subsequently policy priorities. When choosing decision criteria, it is critical to recognize the distinction between policy analysis for research and public policymaking purposes. In research, it is common to adopt a single decision criterion (e.g., economic externality) or a specific metric (e.g., mortality as the health metric). However policy analysts in practice explicitly or implicitly are forced to adopt multiple decision criteria in order to accommodate the divergent interests of stakeholders[61]. Inevitably, agencies from different levels of government or agencies mandated to manage specific public risks typically employ different criteria [62]. In the UK, any transportation proposal that requires the funding or approval of the DFT must be appraised using economic, utility-based criteria [63], but different decision criteria prevail in constraining individual risks. For example, the Kyoto Protocol targets are an important determinant of CO2 emission constraints in the UK, and these targets were the outcome of a political process where circumstances dictated that an arbitrary constrained risk approach be employed[64]. Passenger car tailpipe CO2 emission rates are strongly influenced by the cost-effectiveness-based automobile manufacturers’ EU-wide voluntary targets [64], modified by the influence of UK-specific taxation policies on fuels and vehicle purchases [9, 10]. UK traffic safety goals are set according to various constrained risk casualty targets, partly influenced by technology-based criteria, and without any explicit mention of utility-based criteria [65]. Fuel economy standards are set based on technological criteria at the EU level [66]. Noise standards are determined by a hybrid of utility and technology-based EU standards for vehicle and tire noise [50, 67]. Clearly in the UK, where circumstances are generally similar to other industrialized countries, public policymakers aiming to reduce CO2 from passenger cars will have to design policies aimed to balance multiple risk targets developed from multiple decision criteria. Ultimately there is no clear, objective answer as to which decision criterion is best. The aim of public policymakers should be to design policy portfolios that perform acceptably well regardless of the decision criteria [68].  2.3 2.3.1  PATHWAYS FROM POLICY TO RISKS LINKING POLICIES TO RISKS  A basic influence diagram illustrating the relationship between policies and risks for current generation gasoline and diesel cars is shown in Figure 2.2. Linking policies and risks, I present an influence diagram comprised of a framework with three basic economic actor groups: 33  consumers, manufacturers, and fuel providers, and four basic factors: ownership, annual VKT and patters of use, vehicle technology, and fuel properties [2, 5, 69, 70]. Figure 2.2  Influence diagram of relationships between passenger car policies and risks in a market dominated by spark ignition gasoline and compression ignition diesel cars.  passenger cars usage, fuels & technology economic actors VKT: km per car consumers ownership: cars/person passenger car policies  car manufacturers  vehicle technology  economic, ecological, and human health risks climate change  air quality  energy security  traffic safety  congestion fuel providers  fuel properties noise  In response to policies, manufacturers and consumers determine vehicle ownership and technology choice as shown in Figure 2.2. However, beyond manufacturer pricing and consumer income, there are a variety of complex factors involved such as urban design, employment dynamics, expected annual VKT, congestion, and choice of alternative modes [71-73]. VKT is principally a consumer response, influenced by the marginal cost of driving through factors such as commuting times, fuel costs and road pricing [74]. Fuel properties (i.e., the choice of diesel or gasoline in this case) are determined principally by taxes, fuel providers, and auto manufacturers (through fuel specific changes in drive-train technologies) [75]. Ownership and VKT (including patterns of use) are linked to all risks in Figure 2.2 because more cars and more usage change emissions of contaminants and noise, congest roads, and raise the probabilities of collisions with other road users. It is noted, however, that these simple relationships can be complex and do not always change risks in the expected direction. Congestion charging in London has been estimated to increase average vehicle speeds by 2.1 34  km/hour, reducing NOx and PM10 in diesel cars, but increasing NOx in gasoline cars [48]. Air pollutant emissions vary at micro-timescales due to changes in vehicle speeds and use patterns (e.g., cold starts) [76], and also at macro-timescales due to fleet turnover effects that change the fleet composition mix of older and newer cars [77]. Increased vehicle speeds could also be expected to worsen traffic safety due to the power law influence of speed on casualty severity [49], although research on London’s congestion charge system concludes no obvious change in collision frequency. Increased speed could also possibly worsen noise because in the speed range of 30-50 km/hour tire noise begins to dominate motor and exhaust noise [50]. The fact that ownership and VKT influence all risks highlights the importance of modal shift (i.e., rail, bus, bicycle, and walking) in passenger transport policy. As illustrated in Figure 2.2, the combination of vehicle technology and fuel properties may be strongly linked to air quality and weakly linked to energy security (petroleum consumption) and CO2 emissions [25, 78]. Vehicle technology such as crash avoidance systems and occupant protection features, as well as fleet mass composition, all influence traffic safety risks [79]. Fuel properties alone are shown in Figure 2.2 as only influencing climate change, air quality, and energy security because the properties of fuel consumed do not alter congestion or traffic safety risks. I also note when considering conventional vehicle technologies that fuel and drive-train choices are not linked to changes in noise because the average diesel and gasoline cars in the UK are both 72.3 dB(A) [80], and both have been subject to the same 74 dB(A) regulation since 1998 [50]. One of the tasks in Integrated Assessment is to determine first-order and second-order effects, to ascertain which risks are of most concern and which are less important in evaluating policy and technology options[43]. Such decisions depend on research analyst choice and public agency targets and priorities. For example, research for the UK estimated a mortality increase of roughly 100 per year due to substitution of diesel cars for gasoline models [78]. Out of roughly 13,000 air pollution deaths in the UK annually, is 100 (0.8%) a first-order or second-order effect? Research on fleet mass reduction scenarios and traffic safety in the UK projected that carpedestrian fatalities would be reduced 18% [79]. Again, is this a first-order or second-order effect? In general, iteration in the analysis and scrutiny of the results (e.g., to compare to published research and government priorities) may be required to ascertain whether a risk result is first-order or second-order. The direction of change estimated for a particular risk, whether things are improving or worsening, may itself trigger a risk to be classified as first-order or second-order even if the magnitude of change is small.  35  2.3.2  CLIMATE MITIGATION POLICY OPTIONS AND POTENTIAL INFLUENCES ON RISKS  Table 2.2 provides a comprehensive list of policies that have the potential to reduce CO2 emissions from passenger cars with examples from the UK where applicable. A key concept is that every policy option listed, if sufficiently strong in its technical requirements or the price signal it induces or by the way policy mechanisms are structured, can potentially influence to a measurable degree ownership, VKT and patterns of use, vehicle technology, and fuel choice. Therefore all links to risk shown in Figure 2.2 are potentially affected by the policies in Table 2.2. Table 2.2 was developed based on the following assumptions.   Climate change. Most policies, by definition, should produce direct Climate change mitigation benefits. Mandating biofuels could reduce tailpipe CO2, but indirectly increase CO2 through land use. Policies aimed at VKT and usage can be expected to have indirect climate change effects, including potential disbenefits through increased vehicle ownership and life cycle increases in CO2.    Air quality. Most CO2 mitigation policies hold the potential to produce both indirect benefits and disbenefits depending on the technology and policy conditions such as degree of coordination of air emission and CO2 emission regulations for gasoline and diesel cars. Introducing biofuels or electricity to power cars could potentially improve or exacerbate ozone and airborne particles depending on meteorology and emissions from other sectors.    Energy security. Most policies are expected to improve (or at least not worsen) energy security via reduction in petroleum consumption or diversifying vehicle fuel mix (e.g., biofuels and electricity).    Traffic safety. Vehicle oriented policies that might induce bimodal vehicle fleets (like fleet average CO2 mandates, vehicle-specific cap and trade, or technology mandates) could result in traffic safety disbenefits. Conversely, policies that might minimize distribution of fleet mass, like feebates, might produce indirect benefits. Fleet turnover and traffic safety are related in that any policy that might retard or accelerate scrapping of the oldest cars could produce disbenefits or benefits, respectively. Induced changes in VKT and traffic safety are difficult to predict (e.g., reducing on road time, but increasing average speeds could cancel out).    Congestion. Broad, multi-sector policies could result in congestion benefits if vehicle usage is reduced, or disbenefits if non-transport sectors reduce CO2 to the extent that vehicle usage is allowed to increase. Road pricing and per-kilometer taxes should 36  produce congestion benefits as these are aimed at vehicle use; however road pricing that differentiates by time and place will be more effective than general per-kilometer taxes. Rebound effects from policies that reduce (e.g., technology policies) or increase (e.g., fuel policies) the marginal cost of driving could produce congestion disbenefits or benefits, respectively.  Table 2.2  Policies with potential to reduce CO2 emissions from passenger cars. Every policy option has the potential to influence ownership, VKT, vehicle technology, and fuel properties.  Broad, multisector policies  Policy Category  Policy Description [2, 8284] economy-wide trade  cap  and  Upstream carbon tax [86]  fuel tax  Fuel-oriented policies  company car free fuel tax  fuel-specific mandate  low carbon fuel standard (lifecycle CO2 basis)  fuel feebate (↑ fee for ↑ life cycle gCO2/L; rebates below gCO2/L pivot point)  Example in the UK  CC = climate change, AQ = air quality, ES = energy security, TS = traffic safety, C = congestion, N = noise Benefit4 Disbenefit  EU ETS [85], but CC, ES not applicable to cars  AQ, C  AQ, C  AQ, C  AC, C  CC, C  AQ  CC, ES  gasoline and CC, ES diesel tax [21, 87, 88] income tax CC, ES through employer [21]  AQ, C  Renewable CC, ES Transport Fuel Obligation [14]  AQ  AQ, CC  AQ  AQ  AQ  AQ  CC, ES  CC, ES  direct 4  Benefits and disbenfits are classified as direct or indirect as follows.  indirect  37  Policy Category  Policy Description [2, 8284] fleet gCO2/km emission standards (option for tradable credits)  fleet CO2/km CC target per ACEA voluntary [64]  gCO2/km acquisition tax  CO2/km CC registration taxes flat fee of £38 [21]  vehicle technology- or ownership-based policies  gCO2/km circulation tax  vehicle  vehicle  CO2/km company car tax  green labeling  scrappage bounty[90]  CO2/km vehicle CC taxes VED tax bands [10] UK benefit-in-kind CC CO2/km tax bands [9]  AQ, ES, N  AQ, TS, C, N.  AQ, ES, C, N  AQ, C, N  AQ, ES, C, N  AQ, C, N  AQ, ES, C, N  AQ, C, N  AQ, ES, N  AQ, N  EC Directive CC 1999/94/EC [89]  General Motors AQ, TS EcoFLEX [90] CC, E  vehicle feebates (↑ fee for ↑ gCO2/km; rebates below gCO2/km pivot point) [91] gCO2/km vehicle-specific cap and trade  technology mandate (e.g., minimum sales of zero emission vehicles) gCO2/km emissions tax  VKT or usage oriented policies  CC = climate change, AQ = air quality, ES = energy security, TS = traffic safety, C = congestion, N = noise Benefit4 Disbenefit  Example in the UK  road pricing  per-km tax or insurance premiums [81, 93, 94]  CC AQ, ES, TS, C, N  AQ, C, N  AQ, ES, C, N  AQ, TS, C, N  CC  CC, AQ  CC ES, C, N  CC, AQ, TS, C, N  AQ, ES, C, N  AQ, C, N  CC, AQ, ES, N  CC, AQ, ES, TS  CC, AQ, ES, N  AQ  CC  London C congestion charge[48, 74, 92] C  38   Noise. Any policy that increases uptake of relatively quiet hybrids might produce health benefits. However, it is unknown to what degree quiet cars might affect pedestrian and bicycle safety The far reaching influence of policy is illustrated by the London congestion charging scheme. While the main aim of the charge is to reduce congestion (and hence travel times) within the charging zone at peak times, the scheme is also expected to affect ownership. Various rules, such as exempting cars certified at less than 120 gCO2/km, penalizing cars over 225 gCO2/km, and exempting certain vehicle types (e.g., all electric cars or motor tricycles), are intended to influence consumers’ purchasing choices [81]. The scheme could also have the effect of increasing ownership if consumers purchase an exempt vehicle to be used for local travel. Market responses to the range of policies shown in Table 2.2 can influence risks via fleet turnover effects that are not always considered in policy analysis. The fleet turnover effect is illustrated in Figure 2.3 for gasoline and diesel cars to emphasize that public risks due to passenger car use are a function of the on-road vehicle stock, not just the new cars entering the stock. For example, air quality policies have been criticized on the basis that stringent emission standards may worsen public health due to consumers retaining older, higher polluting cars [95, 96]. Fleet turnover effects are also relevant to traffic safety as newer cars are shown to exhibit significantly improved occupant safety protection than models just a few years older [97]. Fleet turnover effects can also substantially change life cycle CO2 emissions as shown in Figure 2.4 [98]. For the average European car driven 14,000 km/year and average age 14 years, scrapping a car 3 years earlier increases average life cycle emissions by 12 gCO2/km, while scrapping 3 years later decreases life cycle emissions by 12 gCO2/km. For low travel distance cars, life cycle CO2/km is more sensitive to scrappage timing, while high travel distance cars (e.g., taxis, company cars) are less sensitive. Finally, it should be noted that fleet turnover rates can change over time. In the U.S. the mean lifetime of a car was 11 years for a 1966 model compared to 15 years for a 1990 model [99].  39  Figure 2.3  Diagram illustrating passenger car that policy can influence the makeup of on-road vehicle stocks by regulating new cars, scrapped cars, or both.  Policies affecting rate and type of new cars  on road vehicle stock and patterns of use  Policies affecting rate and type of scrapped cars  Figure 2.4  Effect of scrapping age and VKT on life cycle gCO2/km for 2006 models of gasoline and diesel cars in Europe [98].  600  life cycle gCO2/km  500  400  300  200  100  7,000 km - Gasoline  14,000 km - Gasoline  28,000 km - Gasoline  7,000 km - Diesel  14,000 km - Diesel  28,000 km - Diesel  0 0  5  10  15  20  years  40  2.3.3  LINKING POLICIES TO RISKS WITH ALTERNATIVE FUELS AND TECHNOLOGIES  This description of pathways from policies to risks has focused primarily on current gasoline and diesel cars. When evaluating alternative fuels or vehicle technologies, the risks and relationships shown in Figure 2.2, the impact of policies shown in Table 2.2, and the importance of life cycle CO2 emissions all change substantially. A complete description of these effects is beyond the scope of this study but I will mention a few examples. Figure 2.5 shows that, relative to current gasoline or diesel cars, CNG or hybrids would effect non-trivial changes in noise exposures of +1.7 dB(A) and -2.5 dB(A), respectively [80]. A noise reduction of 2.5 dB(A) could have a favorable impact on cardiovascular health, estimated to reduce the odds ratio of incidence of myocardial infarction from 1.20 to 1.14 relative to a no-effect level of 60 dB(A) [50]. Europe’s emissions trading scheme (ETS) does not include passenger cars but does include production of fuels and electricity [85]. Because the well-to-tank CO2 emissions per unit energy of diesel fuel is roughly 25% less than gasoline [35, 100], the ETS effect on fuel prices should moderately favor diesel cars over gasoline. However, the ETS could potentially be a much stronger influence on vehicle choice if plug-in hybrid technology is available, depending on the carbon intensity and cost of the electricity5 [101, 102]. In terms of life cycle CO2 emissions, comparing a future spark ignition car fuelled by lignocellulosic ethanol to a current generation spark ignition gasoline car could result in roughly one order of magnitude reduction from 300 gCO2/km to 30 gCO2/km [24]. The example of plug-in hybrids illustrates how rapidly impacts can change, because when consumers plugs in their car they have immediately and dramatically changed the system boundary of their energy use and associated impacts depending on the design of the electricity supply system [103].  5  Currently, substantial taxes are levied on transportation fuels. Should electric vehicles grow in numbers, the  failure to impose similar taxes on electricity for transportation would represent very large public subsidies.  41  Figure 2.5  Mean noise rating of some 2006 models of alternative fuelled vehicles relative to conventional gasoline or diesel models [80]. The mean for gasoline models is the same for diesel models (72.3 dB(A) for both groups). CNG = compressed natural gas. LPG = liquefied petroleum gas  mean dB(A) relative to conventional gasoline or diesel  2.0 1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5  G CN  G LP g  ol in as  e  d br i y h  .  -2.0 -2.5 -3.0  2.4  ESTABLISHING CAUSAL LINKS BETWEEN POLICIES AND OUTCOMES  In Section 1.3 I discuss linkages between policies and risks. This leads to an important question in the analysis and design of public policy, which is to ascertain to what extent specific policies caused outcomes of interest. For example, in the analysis of tradeoffs between vehicle emissions of air pollutants and CO2 in the UK, an important question is: to what extent did climate mitigation policies cause the growth of diesel cars? I will illustrate the challenges in establishing a causal link between climate policy and outcomes of interest by briefly examining this question. First I will summarize basic issues in establishing causation, followed by a discussion of the various factors that may have caused the growth in diesel cars. 2.4.1  BASIC ISSUES IN ESTABLISHING CAUSAL LINKS BETWEEN POLICY AND OUTCOMES  Employing a traditional scientific approach [104], I suggest there are three basic elements in establishing causation between a specific public policy and an outcome of interest:  42  1. Temporal relationship. Did the policy precede changes in the outcome of interest? In establishing causal links, it is important to attend to timing. As an example, in DFT’s evaluation of its VED policy impact on CO2 emissions, researchers were careful to choose subjects (e.g. private consumers and fleet operators) who made decisions after the VED was adopted in March, 2001[10]. 2. Quantitative and qualitative evidence linking the policy to the outcome. A general approach to examining causal evidence in research is to use visual diagrams to illustrate relationships between dependent variables (outcomes of interest) and independent variables, including the effects of intervening variables [105]. Quantitative evidence may come in the form of a statistical model such as done in econometric modeling of vehicle ownership (see [106] for a review of such models, and [107] for a UK-specific modeling example). Qualitative evidence may come in various forms such as survey interviews or focus groups to ascertain awareness of policies, for example as done in an examination of the VED tax in the UK [108]. 3. Reasoning and peer review. The presence of valid temporal relationships and quantitative or qualitative models cannot be divorced from reasoning in establishing causal linkages to policies. In practice, this means conducting a critical examination of all evidence guided by selected criteria. In health research, Bradford Hill’s eight criteria [109] form the reasoning that is often used to guide examinations of causal arguments (see [110] as an example). Are there confounding factors that have been overlooked in the analysis? To what degree are statistical relationships consistent with those found previously by other researchers, perhaps in other contexts? Subjecting the analysis of policy causation to peer review is a practical way of answering such questions and bolstering causal arguments in policy analysis [36]. 2.4.2  WHAT CAUSED THE RAPID RISE IN DIESEL CARS IN THE UK?  As described in Section 3.4, there are multiple factors that may have contributed to rise in diesel market share in the UK. Three specific CO2 policies potentially contributed to the rise in diesel cars including the UK VED tax [10, 108], the UK company car tax [9], and the European-wide ACEA agreement[64]. Technological change by the introduction of turbo-charged direct injection (TDI) is another factor [111], as well as fuel prices [112]. All three CO2 policies could have plausibly contributed to the rise in diesels in the UK. The rise in diesel market share began in the UK in the year 2000, while a rise in the rest of the EU began in 1997 (see Figure 3.1). In terms of temporal relationships, the ACEA agreement was implemented as of 1997 and thus was more temporally related to the rise in the EU average diesel share than the UK. At first glance, neither the VED nor the company car tax could have 43  caused the rise of UK diesels because the rise initiated prior to policy implementation dates. However, it should be noted that these policies were announced in advance of the implementation dates. For example the company car tax had been announced in 1999, followed by release of the details of the CO2-based tax regime in 2000, with implementation beginning in 2002 [113]. The government’s analysis of the policy cited strong awareness amongst company decision-makers and that the policy led to direct increases in diesel cars [9, 113]. As well, an independent study attributed growth in diesels to the company car tax [114]. In contrast, it was determined that purchasers of private (non-company) cars had relatively low awareness of the VED policy, and that the VED had low impact on fleet composition based on qualitative [108] and quantitative [10] evidence. A research study of the ACEA agreement implicitly assumed that all changes in fleet composition affecting CO2 emissions were due to the manufacturer association’s voluntary agreement [64]. However, the objective of that research was to quantify CO2 emission rates due to changes in fleet composition rather than assess policy causation. In terms of modeling fleet composition, none of the CO2 policy impact assessments included econometric models. Amongst the aforementioned policy assessments [9, 10, 108, 113, 114], only the ACEA study was published in a peer-reviewed journal [64]. Fuel prices are another factor cited for the rise in diesel market share in the UK [112, 114]. In terms of the relative price of diesel versus gasoline, this is not a factor that could have contributed to the rise in diesel market share in the UK because the price ratio has been kept at unity over the time period where diesels have grown (see Figure 3.2 caption). Another hypothesis could be that a rise in the real price of both petroleum fuels led to consumers purchasing vehicles that consume less fuel per kilometre – i.e. diesel cars and smaller gasoline cars. Examining year 2005 prices of fuels relative to 1996, average fuel costs rose 55%, while relative to the year 2000 prices rose 9% (UK fuel duties increased in the late 1990’s) [12]. Therefore it is plausible that the rise in the real price of fuels partly contributed to the rise in diesels, but again there is no known analysis of this factor. Technological change in the form of advanced TDI technology is yet another factor hypothesized to contribute to the rise in diesel cars. TDI advances for diesel cars are credited with improving efficiency, reducing noise, increasing acceleration and power, and reducing smoke and particle emissions [111]. TDI was first introduced to the European market in 1988 [115], although gains in TDI technology continued such as specific power of engines rising from roughly 40 to 60 kW per litre from 1991 to 2001 [116]. One of the temporal issues in examining causal linkages for TDI is that there is no indication of a difference in TDI technology offered in the UK compared to the rest of the EU [111], yet the UK lagged the EU by three years in terms of rising diesel shares.  44  My conclusion with regard to climate policy causal links to rising diesel share in the UK is that there are multiple explanatory factors that could have potentially contributed to this trend. The temporal relationships and available qualitative and quantitative evidence indicate that the company car tax clearly had some non-negligible causal influence on the growth of diesel cars. However the degree to which this tax regime influenced diesel growth as compared to other direct and indirect public policies has yet to be rigorously analyzed in either peer- or non-peerreviewed literature.  2.5  CONCLUDING REMARKS  I have argued for the growing importance of Integrated Assessment of multiple risks in the design of climate mitigation policies for passenger cars in industrialized countries. Using the UK context for illustration purposes, it is shown that three important choices in policy analysis can substantially change policy design to mitigate CO2 from passenger cars: (1) decision criteria, (2) choice of fuels and technology, and (3) the scope of risks included or excluded. These choices should be made iteratively during the policy analysis and the choices differ for researchers as compared to public policymakers. Employing Integrated Assessment it becomes clear that the direct implications of fuels and technology alone are insufficient to estimate changes in risks, because how policies affect ownership and VKT and patterns of use matters. While passenger car stocks in industrialized countries are presently dominated by spark ignition gasoline and compression ignition diesel cars, substantial growth of alternative fuels and vehicle technologies appears likely in the near future. The proliferation of these alternatives can drastically expand system boundaries affected by new policies in terms of energy use and CO2 emissions. With climate change as a growing policymaking priority and the concomitant changes in fuels and technologies, Integrated Assessment of multiple risks resulting from passenger car climate mitigation policy is useful and increasingly necessary. It can inform development of climate mitigation policy to show what is possible and, improve the social and political acceptance of policies, minimize unintended consequences, and improve the likelihood of realizing net public benefits.  2.6  ACKNOWLEDGEMENTS  I thank Hadi Dowlatabadi, Michael Brauer, Milind Kandlikar, and Douw Steyn for comments on drafts of this article. I thank the University of British Columbia Bridge Program, AUTO21: B06 BLC, and Carnegie Mellon University’s NSF-supported Center for Integrated Study of the Human Dimensions of Global Change (SBR-9521914), and the Center for Climate Decision Making (SES-0345798). I am also grateful for generous support from the Exxon-Mobil Education Foundation. All errors are my responsibility. 45  2.7 1.  REFERENCES Cambridge.Systematics.Inc., Moving Cooler: An Analysis of Transportation Strategies for Reducing Greenhouse Gas Emissions. 2009, Urban Land Institute: Washington, D.C.  2.  Kopp, R., Transport policies to reduce CO2 emissions from the light-duty vehicle fleet (Issue Brief 12), in Assessing U.S. Climate Policy Options. 2007, Resources for the Future: Washington, DC.  3.  Sperling, D. and D. Gordon, Two Billion Cars: Driving Toward Sustainability. 2009: Oxford University Press.  4.  Keefe, R., J. Griffin, and J. Graham, The Benefits and Costs of New Fuels and Engines for Light-Duty Vehicles in the United States. Risk Analysis, 2008. 28(5): p. 1141-1154.  5.  Parry, I., M. Walls, and W. Harrington, Automobile externalities and policies. Journal of Economic Literature, 2007. XLV: p. 373-399.  6.  Sansom, T., et al., Surface transport costs and charges: Great Britain 1998. 2001, Department of the Environment, Transport and the Regions (DETR).  7.  Fulton, L., P. Cazzola, and F. Cuenot, IEA Mobility Model (MoMo) and its use in the ETP 2008. Energy Policy, 2009. 37: p. 3758-3768.  8.  Ribeiro, K., et al., Transport and its infrastructure, in Climate Change 2007: Mitigation. Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. 2007, Cambridge University Press.  9.  IR, Report on the evaluation of the company car tax reform: stage two. 2006, UK Inland Revenue (www.inlandrevenue.gov.uk).  10.  MORI, Assessing the impact of Graduated Vehicle Excise Duty: quantitative report. 2003, UK Department for Transport (www.dft.gov.uk) and MORI (www.mori.co.uk).  11.  Woodcock, J., et al., Energy and Health Series: Energy and Transport. The Lancet, 2007. 370: p. 1078-1088.  12.  DFT, UK Transport Statistics online database (www.dft.gov.uk/pgr/statistics). 2008, UK Department for Transport.  13.  SMMT, New car CO2 report 2008. 2008, Society of Motor Manufacturers and Traders Ltd. (www.smmt.co.uk).  14.  DFT, Carbon and sustainability reporting within the renewable transport fuel obligation (RTFO). 2008, UK Department for Transport.  46  15.  Onoda, T., IEA policies - G8 recommendations and an afterwards. Energy Policy, 2009. 37: p. 3823–3831.  16.  DEFRA, Online database of GHG emissions (http://www.defra.gov.uk/environment/statistics/globatmos/alltables.htm). 2008, Department for Environment, Food and Rural Affairs.  17.  CFIT, Transport and climate change. 2007, Commission for Integrated Transport (www.cfit.gov.uk).  18.  DEFRA, The UK Government Sustainable Development Strategy. 2005, UK Department for Environment, Food and Rural Affairs (DEFRA).  19.  DFT, Delivering better transport: priorities for 2007-08 and beyond. 2007, UK Department for Transport.  20.  DTI, Meeting the energy challenge: a white paper on energy (May 2007). 2007, Department of Trade and Industry.  21.  HM-Treasury, Budget 2007 - Building Britain’s long-term future: prosperity and fairness for families. 2007, HM Treasury (UK).  22.  Lave, L., Conflicting objectives in regulating the automobile. Science, 1981. 212(893-899).  23.  Lave, L., Controlling contradictions among regulations. The American Economic Review, 1984. 74(3): p. 471-475.  24.  Maclean, H. and L. Lave, Life cycle assessments of automobile/fuel options. Environmental Science & Technology, 2003. 37(5445-5452).  25.  Bonilla, D., Fuel demand on UK roads and dieselisation of fuel economy. Energy Policy, 2009. 37: p. 3769-3778.  26.  Kohler, J., Y. Jin, and T. Barker, Integrated modelling of EU transport policy. Journal of Transport Economics and Policy, 2008. 42(1): p. 1-21.  27.  Ogden, J., R. Williams, and E. Larson, Societal lifecycle costs of cars with alternative fuels/engines. Energy Policy, 2004. 32(7-27).  28.  Litman, T., Integrating public health objectives in transportation decision-making American Journal of Health Promotion, 2003. 18(1): p. 103-109.  29.  ARB, California state motor vehicles pollution control standards; Request for waiver of federal preemption; Opportunity for Public Hearing, Docket ID #EPA-HQ-OAR2006-0173; 72 FR 21260 (April 30, 2007). 2007, California Air Resources Board.  47  30.  T&E, Danger ahead: Why weight-based CO2 standards will make Europe's car fleet dirtier and less safe. 2007, European Federation for Transport and Environment: Brussels, Belgium.  31.  Hull, A., Policy integration: what will it take to achieve more sustainable transport solutions in cities? Journal of Transport Economics and Policy, 2007. 15: p. 94-103.  32.  Lund, H. and E. Munster, Integrated transportation and energy sector CO2 emission control strategies. Transport Policy, 2006. 13: p. 426–433.  33.  Chatterjee, K. and A. Gordon, Planning for an unpredictable future: Transport in Great Britain in 2030. Transport Policy, 2006. 13: p. 254–264.  34.  Dowlatabadi, H., On integration of policies for climate and global change. Mitigation and Adaptation Strategies for Global Change, 2007. 12: p. 651–663.  35.  Maclean, H. and L. Lave, Evaluating automobile fuel/propulsion system technologies. Progress in Energy and Combustion Science, 2003. 29: p. 1–69.  36.  Morgan, M.G. and M. Henrion, Uncertainty: a guide to dealing with uncertainty in quantitative risk and policy analysis. 1992: Cambridge University Press.  37.  Rajagopal, D., et al., Challenge of biofuel: filling the tank without emptying the stomach? Environmental Research Letters, 2007. 2(044004): p. 1-9.  38.  Farrell, A., et al., Ethanol can contribute to energy and environmental goals. Science, 2006.  39.  Jacobson, M., Effects of ethanol (E85) versus gasoline vehicles on cancer and mortality in the United States. Environmental Science & Technology, 2007. 41: p. 4150-4157.  40.  Williams, R. and E. Larson, A comparison of direct and indirect liquefaction technologies for making fluid fuels from coal. Energy for Sustainable Development, 2003. VII(4): p. 103-129.  41.  Peirson, J., I. Skinner, and R. Vickerman, Estimating the external costs of UK passenger transport: the first step towards an efficient transport market. Environment and Planning A, 1995. 27: p. 1977-1993.  42.  Lindberg, G. Recent progress in the measurement of external costs and implications for transport pricing reforms. in Implementing Reform on Transport Pricing: Identifying ModeSpecific issues. 2002. Brussels, Belgium.  43.  Morgan, M.G., et al., Why Conventional Tools for Policy Analysis are Often Inadequate for Problems of Global Change. Climatic Change, 1999. 41: p. 271-281.  44.  Naess, P., Cost-benefit analyses of transportation investments - neither critical nor realistic. Journal of Critical Realism, 2006. 5(1): p. 32-60.  48  45.  Nash, C., Marginal cost and other pricing principles for user charging in transport: a comment. Transport Policy, 2003. 10: p. 345–348.  46.  Sturm, R., Economics and physical activity. American Journal of Preventive Medicine, 2005. 28(2S2): p. 141-149.  47.  HPA, Health Effects of climate change in the UK 2008: an update of the Department of Health report 2001/2002. 2008, UK Health Protection Agency (HPA).  48.  Beevers, S. and D. Carslaw, The impact of congestion charging on vehicle speed and its implications for assessing vehicle emissions. Atmospheric Environment, 2005. 39: p. 6875– 6884.  49.  Evans, L., Traffic Safety. 2004: Science Serving Society.  50.  den Boer, L. and A. Schroten, Traffic noise reduction in Europe: Health effects, social costs and technical and policy options to reduce road and rail traffic noise. 2007, CE Delft (www.ce.nl) commissioned by T&E (http://www.transportenvironment.org).  51.  DFT, Sources of Particulate Matter in Urban Areas: TRAMAQ Project UG 250. 2002, UK Department for Transport (www.dft.gov.uk).  52.  COMEAP, Quantification of the effects of air pollution on health in the United Kingdom. 1998: The Stationary Office London (www.tsonline.co.uk).  53.  WHO, Environmental burden of disease: Country profiles (United Kingdom). 2007, World Health Organization.  54.  DFT, Road Accident Data [computer files] 1994 SN3461, 1996 SN3627, 1997 SN3817, 1998 SN4027, 1999 SN4080, 2000 SN4356, 2001 SN4501, 2002 SN4588, 2003 SN4979, 2004 SN5042, 2005 SN5244, 2006 SN5493. 2007, UK Data Archive (Colchester, Essex).  55.  WHO-Europe, Burden of disease from environmental noise (http://www.euro.who.int/Noise/activities/20021203_3 downloaded April 3, 2008). 2008, World Health Orgainization.  56.  GAD, Government Actuary's Department (GAD) online statistics. 2006, United Kingdom Government Actuary's Department (GAD).  57.  Cohen, A., et al., The global burden of disease due to outdoor air pollution. Journal of Toxicology and Environmental Health, Part A,, 2005. 68: p. 1301–1307.  58.  Kunzli, N., et al., Assessment of deaths attributable to air pollution: should we use risk estimates based on time series or cohort studies? American Journal of Epidemiology, 2001. 153: p. 1050-1055.  49  59.  Levy, J., et al., Assessing the public health benefits of reduced ozone concentrations. Environmental Health Perspectives, 2001. 109(12): p. 1215-1226.  60.  Knol, Trends in the environmental burden of disease in the Netherlands 1980 – 2020. 2005, RIVM (www.rivm.nl/en) commissioned by Netherlands Environmental Assessment Agency.  61.  Murray, C.J.L., and Lopez, A.D., ed. The Global Burden of Disease: A Comprehensive Assessment of Mortality and Disability from Diseases, Injuries, and Risk Factors in 1990 and Projected to 2020. Global Burden of Disease and Injury Series. 1996, Harvard University Press.  62.  Marsden, G. and P. Bonsall, Performance targets in transport policy. Transport Policy, 2006. 13(191-203).  63.  CFIT, A Review of Transport Appraisal. 2004, Commission for Integrated Transport (www.cfit.gov.uk).  64.  Fontaras, G. and Z. Samaras, A quantitative analysis of the European Automakers’ voluntary commitment to reduce CO2 emissions from new passenger cars based on independent experimental data. Energy Policy, 2007. 35: p. 2239-2248.  65.  Broughton, J., Monitoring progress toward the 2010 casualty reduction target - 2005 data. 2007, Transport Research Laboratory (www.trl.co.uk) for the UK Department for Transport (www.dft.gov.uk).  66.  Kwon, T., The determinants of the changes in car fuel efficiency in Great Britain (1978–2000). Energy Policy, 2006. 34: p. 2405–2412.  67.  FEHRL, Study SI2.408210 tyre/road noise - volume 1: final report. 2006, Forum of European National Highway Research Laboratories (FEHRL).  68.  Gouge, B., et al., An Integrated Approach to Transportation Policy in BC - assessing greenhouse gas reductions opportunities in freight transportation. 2008, Pacific Institute for Climate Solutions.  69.  Greening, P., European vehicle emission legislation – present and future. Topics in Catalysis, 2001. 16/17(1-4): p. 5-13.  70.  Schipper, L., Determinants of automobile use and energy consumption in OECD countries. Annual Review of Energy and Environment, 1995. 20: p. 325-386.  71.  Dargay, J., The effect of prices and income on car travel in the UK. Transportation Research Part A: Policy and Practice, 2006. 41(949-960).  72.  Dargay, J. and M. Hanly, Volatility of car ownership, commuting mode and time in the UK. Transportation Research Part A: Policy and Practice, 2007. 41: p. 934-948.  50  73.  Whelan, Modelling car ownership in Great Britain. Transportation Research Part A: Policy and Practice, 2007. 41: p. 205-219.  74.  Prud'homme, R. and J. Bocarejo, The London congestion charge: a tentative economic appraisal. Transport Policy, 2005. 12: p. 279–287.  75.  Verboven, F., Quality-based tax discrimination and tax incidence: evidence from gasoline and diesel cars. The RAND Journal of Economics, 2002. 33(2): p. 275-297.  76.  Boulter, P. and I. McCrae, The links between micro-scale traffic, emission and air pollution models. 2007, Transport Research Laboratory (www.trl.co.uk).  77.  Singh, R., A. Huber, and J. Braddock, Sensitivity analysis and evaluation of MicroFacPM: a microscale motor vehicle emission factor model for particulate matter emissions. Journal of the Air & Waste Management Association, 2007. 57: p. 420-433.  78.  Mazzi, E. and H. Dowlatabadi, Air quality impacts of climate mitigation: UK policy and passenger vehicle choice. Environmental Science & Technology, 2007. 41: p. 387-392.  79.  Mazzi, E., H. Dowlatabadi, and M. Kandlikar, Regulating Car Mass for Concurrent Traffic Safety and Climate Mitigation Benefits. 2010 (DRAFT).  80.  VCA, Online data of vehicle emissions for UK cars model years 2000-2006. 2007, Vehicle Certification Agency (www.vcacarfueldata.org.uk).  81.  Greenberg, A., Designing pay-per-mile auto insurance regulatory incentives. Transportation Research Part D: Transport and Environment, 2009. 14(6): p. 437-445.  82.  Acutt, M. and J. Dodgson, Policy instruments and greenhouse gas emissions from transport in the UK. Fiscal Studies, 1996. 17(2): p. 65-82.  83.  Glaister, S., UK transport policy 1997-2001. Oxford Review of Economic Policy, 2002. 18(2): p. 154-186.  84.  Potter, S. and G. Parkhurst, Transport policy and transport tax reform. Public Money & Management, 2005(June): p. 172-178.  85.  Grubb, M. and K. Neuhoff, Allocation and competitiveness in the EU emissions trading scheme: policy overview. Climate Policy, 2006. 6: p. 7-30.  86.  Nordhaus, W., After Kyoto: Alternative Mechanisms to Control Global Warming. The American Economic Review, 2006. 96(2): p. 31-34.  87.  COWI, Fiscal measures to reduce CO2 emissions from new passenger cars. 2002, European Commission.  51  88.  Kunert, U. and H. Kuhfeld, The diverse structures of passenger car taxation in Europe and the EU Commissions proposal for reform. Transport Policy, 2007. 14(4): p. 306-316  89.  DFT, New passenger cars - information on fuel consumption and CO2 emissions. 2005, UK Department for Transport.  90.  Just-Auto, UK: GM adds green scrapping incentive (June 6), in Just-Auto (www.justauto.com). 2006.  91.  Greene, D., Feebates, footprints and highway safety. Transportation Research Part D: Transport and Environment, 2009. 14: p. 375-384.  92.  TFL. Congestion charging (www.tfl.gov.uk/roadusers/congestioncharging/default.aspx). 2008 [cited April 14, 2008].  93.  Bordoff, J. and P. Noel, Pay-As-You-Drive Auto Insurance: a Simple Way to Reduce DrivingRelated Harms and Increase Equity, in The Hamilton Project. 2008, The Brookings Institution.  94.  Parry, I., On the costs of policies to reduce greenhouse gases from passenger vehicles (RFF DP 06-14). 2006, Resources for the Future.  95.  Gruenspecht, H., Zero emission vehicles: a dirty little secret. Resources, 2001(142): p. 7-10.  96.  Pickrell, D., Cars and clean air: a reappraisal. Transportation Research Part A: Policy and Practice, 1999. 33: p. 527-547.  97.  Broughton, J., Casualty rates by type of car (version 3). 2007, TRL Limited.  98.  Kimmo, K., Life-cycle energy consumption and carbon dioxide emissions of world cars (www2.lut.fi/~kklemola/dontfly/carsof2006.htm). 2006.  99.  Schafer, A. and H. Jacoby, Vehicle technology under CO2 constraint: a general equilibrium analysis. Energy Policy, 2006. 34: p. 975-985.  100.  Heywood, J., et al., The performance and future ICE and fuel cell powered vehicles and their potential fleet impact, in 2004 SAE World Congress. 2004, Society of Automotive Engineers.  101.  Golbuff, S. Design optimization of a plug-in hybrid electric vehicle (2007-01-1545). in 2007 World Congress. 2007. Detroit, Michigan, USA: Society of Automotive Engineers (www.sae.org).  102.  Samaras, C. and K. Meisterling, Life cycle assessment of greenhouse gas emissions from plug-in hybrid vehicles: implications for policy. Environmental Science & Technology, 2008. 42: p. 3170–3176.  52  103.  Bergerson, J. and L. Lave, Should We Transport Coal, Gas, or Electricity: Cost, Efficiency, and Environmental Implications. Environmental Science & Technology, 2006. 39(16): p. 5905-5910.  104.  Giere, R., Understanding Scientific Reasoning. 4th ed. 1998: Harcourt Brace College Publishing.  105.  Creswell, J., Research design: qualitative, quantitative, and mixed method approaches (2nd ed.). 2003: Sage Publications.  106.  De Jong, G., et al., Comparison of car ownership models. Transport Reviews, 2004. 24(4): p. 379–408.  107.  Page, M., G. Whelan, and A. Daly, Modeling the factors which influence new car purchasing, in European Transport Conference 2000. 2000: Cambridge, UK.  108.  MORI, Assessing the impact of Graduated Vehicle Excise Duty: qualitative report. 2003, UK Department for Transport (www.dft.gov.uk) and MORI (www.mori.co.uk).  109.  Last, J., A dictionary of epidemiology. 4th ed. 2001: International Epidemiological Association.  110.  Vedal, S., Ambient Particles and Health: Lines that Divide. Journal of the Air & Waste Management Association, 1997. 47: p. 551-581.  111.  Schipper, L., C. Marie-Lilliu, and L. Fulton, Diesels in Europe: Analysis of Characteristics, Usage Patterns, Energy Savings, and CO2 Emission Implications. Journal of Transport Economics and Policy, 2002. 36(2): p. 305-340.  112.  FleetNews, Diesel sales rocket with fuel price rise, in Fleet News. 2006.  113.  IR, Report on the evaluation of the company car tax reform. 2004, UK Inland Revenue (www.inlandrevenue.gov.uk).  114.  Lane, B., Car buyer research report: consumer attitudes to low carbon and fuel-efficient passenger cars. 2005, Low Carbon Vehicle Partnership (UK).  115.  Sher, E., ed. Handbook of Air Pollution from Internal Combustion Engines: Pollution Formation and Control. 1998, Academic Press.  116.  Scheid, E. Development trends in passenger car DI engines. in AVEEC 2001 (presentation downloaded from www.meca.org). 2001.  53  3  AIR QUALITY IMPACTS OF CLIMATE MITIGATION: UK POLICY AND PASSENGER VEHICLE CHOICE6  3.1  INTRODUCTION  Climate mitigation policies have been promoted on the basis that reducing fossil fuel use provides dual benefits in terms of long-term climate change attenuation and short-term air quality improvements. Models predict that climate policies result in reduced fossil fuel combustion, lower air emissions, and subsequently provide public health benefits [1-4]. In Europe, diesel cars are viewed as a promising option to reduce greenhouse gas emissions from personal transportation. Diesel fuel has a higher energy and carbon density than petrol (38.5 MJ/liter gross heating value and 778 gC/liter versus 34.9 MJ/liter and 659 gC/liter), but diesels have 25% better fuel economy based on matched pair vehicle models and thus emit 1520% less CO2 per kilometer [5-6]. Beginning March, 2001 a new vehicle excise duty (VED) was introduced in the UK whereby vehicles were taxed annually based upon certified CO2 emissions [7]. In April, 2002 the UK’s company car benefit-in-kind tax was changed to a CO2based system as well [8]. Although a surcharge for diesel vehicles was applied to reflect their impact on air quality, the cost of owning diesel cars was and is lower (see illustrative cost comparison in Table A1 in Appendix A). These CO2 policies are credited with contributing to the UK’s success in reducing CO2 emissions [8-10]. Emissions of CO2 from new passenger vehicles registered in the UK between 2000 and 2005 have fallen from a fleet average of 181grams to 169 grams of CO2 per kilometer [10]. Diesel vehicles have made a contribution to the CO2 reductions as their market share has grown non-linearly at an annual compounded rate of 21% from 2001-2005. As shown in Figure 3.1, diesel new registrations in the UK had declined from 1995 onwards while new registrations of diesel cars were increasing in the rest of the European Union (EU) [11]. The changes in the tax regime are demonstrably the turning point for the resurgence of diesels in the UK.  6  A version of this chapter has been published. Mazzi, E. and H. Dowlatabadi, “Air quality impacts of climate  mitigation: UK policy and passenger vehicle choice”. Environmental Science and Technology, 2007. 41: p. 387392.  54  On a parallel track and lagging the CO2 policies, new EU emission standards for passenger vehicles are being promulgated to converge emissions from diesels and petrol engines to the best that either technology can achieve. Thus, 2001 Euro III standards and 2006 Euro IV standards both drive down the higher emissions of PM10 (particulate matter less than 10 microns) and NOx (nitrogen oxides) from diesels, and force gasoline engines to lower their CO (carbon monoxide) and HC (hydrocarbons) emissions. Here we examine consumer switching from petrol-fuelled to diesel-fuelled passenger cars in the UK (all vehicles designed to carry passengers, but excluding freight vehicles). The tradeoffs between greenhouse gas reductions and air pollution impacts are assessed over a 20 year study period from 2001-2020. While diesel substitution for petrol cars scenarios have been assessed elsewhere [12], this is an empirical study in the context of an actual CO2 policy and targets an acknowledged gap in the climate policy literature [13,14]. These research findings have direct relevance to the design of climate policies in the transportation sector throughout the developed world. Figure 3.1 Diesel share of new car registrations in the European Union (EU) and the UK. While aggregate EU demand for diesels began increasing in 1995, UK demand continued to decline until the first CO2 policy incentive came into effect in 2001 and has been increasing continuously.  Diesel % of total new registrations  60% 50% 40%  1988 Fiat introduces EU's 1st turbocharged direct injection (TDI) diesel  EU without UK UK  30% 20% 10%  2002 CO2 based company car tax 2001 CO2 based vehicle excise duty  0% 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 Year 55  3.2  METHODS  We define “additional diesels” as the number of newly registered diesel vehicles additional to an estimated “no growth” diesel market share. Growth in new diesel registrations is estimated relative to historical average new registrations for private and company car registrations (roughly half of all newly registered cars in the UK are for company fleets). Figure 3.2 shows 1994-2005 data for new diesel registrations for private and company vehicles. To quantify the overall growth in diesel vehicles, we assign 15% as the “no growth” diesel market share from 2001-2020. This is based on the average diesel share for 11 years from 1990-2000 leading up to the CO2 policies, and agrees closely with government projected diesel share [15]. Factors that likely have contributed to the growth in diesel market share, including CO2 policy, are described in the discussion section. Figure 3.2  Total number and percentage market share of new registrations of private and company diesel cars in the UK 1994-2005. During this period, the ratio of the price of petrol to diesel was remarkably stable averaging 0.98 (range 0.95-1.00) and fuel price advantages experienced elsewhere do not provide a plausible explanation for the observed changes in diesel registrations in the UK. number of diesel company car registrations number of diesel private car registrations  50%  diesel % of total company cars  45%  diesel % of total private cars  40%  600 500  35% 30%  400  25% 300  20% 15%  200  10% 100  Percentage of new diesel registrations  Number of new diesel registrations, 1000's  700  5%  0  0% 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Year  56  An annual time series of additional diesels for 2001-2020 is developed by adding in new registrations and subtracting estimates of scrapped cars. An estimate of the yearly number of scrapped cars is made using a model developed and calibrated using UK deregistration data as described in Figure A1 of Appendix A [16]. The annual time series of additional diesels is used to estimate changes in fuel consumption, emissions of CO2 and common air contaminants, and human health effects. Changes in fuel consumption and CO2 emissions are estimated based on the difference between petrol and diesel fleet average emission factors shown in Figure A2 of Appendix A. These emission factors use UK-specific, certified CO2/km [17] and are based on actual fleet average CO2 emissions from 2001 to 2005 [10], and projected fleet average emissions through 2020. Changes in fuel consumption are estimated using fleet CO2/km and conversion factors of 2,763 gCO2/liter for diesel and 2,504 g/liter for petrol [18]. We highlight that fleet average diesel CO2/km is only 4% lower than petrol as of model year 2005, because diesel technology’s inherent advantage has been offset by consumer preference for larger cars [10]. Changes in emissions of common air contaminants are estimated based on UK-specific emission factors for PM10, NOx, CO, HC, benzene, and 1,3 butadiene as provided in Table A2 in Appendix A. As with CO2, changes in emissions are quantified based on the difference between emission factors for petrol and diesel cars. Emissions are estimated for each year from 20012020 based on the applicable EU emission class (Euro III, Euro IV, post-Euro IV), and using the UK-specific emission factors derived based on measurements under actual driving conditions [15]. Automobile usage is a constant 15,800 km annually in our analysis, based on average driving distances for all UK cars [19]. We assumed the same pattern of vehicle usage regardless of type (diesel or petrol) because of the high market share of diesels, and did not include a rebound effect (higher usage) due to the lower operating cost of diesels. A sensitivity estimate of the rebound effect is provided in the discussion. Further data and discussion of the annual travel distance are contained in Section A.4.2 of Appendix A. Human health impacts resulting from changes in common air contaminants are estimated solely on changes in particulate matter emissions employing a conventional impact pathway method [20, 21]. The rationale for using particulate matter is that it tends to dominate human health impacts from air pollution based on current science [20, 22-24]. Moreover, a recent study of transportation and air pollution in the UK concluded that reduction of PM10 should be the top priority if the goal is to reduce air pollution impacts on public health in the UK [25]. Another recent source apportionment study of London also concluded that controls on particulate matter 57  emissions from vehicles are most likely to result in the greatest improvements in ambient particulate matter concentrations [26]. The impact of excluding ambient ozone, NOx, CO, HC, benzene, and 1,3 butadiene in the quantification of health effects is discussed in the discussion section. Published emission factors and modeling studies for the UK are based exclusively on PM10; however, in general, more than 99% of particulate emissions by mass from diesel cars are PM2.5 [27]. Hence, for all practical purposes, changes in PM10 and PM2.5 emissions are equal. This allows the use of the health coefficients that are based on either PM10 or PM2.5. We use concentration-response coefficients based on PM10 and on PM2.5 where supporting evidence is available. To maintain clarity, for the remainder of this article we will only refer to particulate matter emissions or ambient concentrations as PM10, except where referring to original study findings that were based on PM2.5. We quantify changes in ambient PM10 concentrations by employing the results of atmospheric modeling studies of vehicle emission reduction measures in the UK [28-30]. The analysis scenario most applicable to our research estimated that an annual UK-wide reduction of 3.76 kilo-tonnes of particulate matter due to traffic emission controls would result in a UK-wide, population weighted change in average annual ambient concentration of 0.277 µg/m3. This is a ratio of 0.0737 µg/m3 change in annual mean ambient PM10 concentration per 1 kilo-tonne change in annual PM10 emissions, which includes a component of secondary nitrate particles as described in Section A.4.4 Appendix A [29]. This ratio was multiplied by our estimated annual change in PM10 emissions to estimate changes in population-weighted ambient PM10 concentrations. This PM10 ambient concentration estimate is equivalent to an intake fraction of 20 grams per million grams emitted – in general agreement with estimates of vehicle emissions found elsewhere [31,32]. For the association between changes in ambient PM10 and mortality, we employ low, central, and high concentration-response coefficients. For our low mortality coefficient, we use 0.75% change in mortality annually per 10 µg/m3 change in PM10, applied to the entire UK population. This coefficient was derived through meta-analyses by the UK Department of Health expert Committee on the Medical Effects of Air Pollution (COMEAP) applying the results of time series epidemiological studies [35]. We use this coefficient because it has been endorsed specifically for the UK; however, there are arguments that time series results are inappropriate for estimating long-term death rates [36, 37].Our central mortality coefficient uses the results of the American Cancer Society cohort study which found a 4% increase in chronic, all-cause exposure mortality per 10 µg/m3 increase in PM2.5 for a cohort of subjects age 30 and older in the U.S. [33]. We 58  chose this study because it is by far the largest cohort study published. It estimated chronic rather than acute mortality, used the same exposure metric as we estimate in our study (annual average concentration), and is the preferred study for chronic exposure mortality estimates according to the latest World Health Organization guidelines [34]. For our high mortality coefficient, we use the results of the Harvard Six Cities study which found a 13% increase in chronic exposure mortality per 10 µg/m3 change in PM2.5 for a cohort of about 8,000 subjects age 25 and older in the U.S. [38]. More recent research on intra-urban variation in PM2.5 and mortality in Los Angeles indicates that the true mortality coefficient may be somewhere between the central and high coefficients we employ in this study [39]. We quantify morbidity by estimating respiratory and cardiovascular hospitalizations using rate coefficients adopted by COMEAP [28]. Changes in both hospitalization rates are estimated at 0.8% per 10 µg/m3 change in PM10 applied to the entire UK population. To tie together the linked steps in this analysis, an integrated framework is provided in Figure 3.3 that illustrates the timing of CO2 policy incentives, EU emission standards, and changes in diesel market share over time. To assess timing of the CO2 policies and EU vehicle emission standards, we divide the study time horizon into three intervals: a “Euro III interval” from 20012005 when Euro III emissions standards applied, a “Euro IV interval” from 2006-2008, and the “post-Euro IV interval” from 2009-2020 when progressively higher emission standards apply to all new vehicle purchases. Early adoption of Euro IV diesels is incorporated into our estimates as described in Section A.2 of Appendix A.  59  Figure 3.3  Integrated framework for assessing emissions from additional diesels (i.e., diesels substituted for petrol vehicles). Actual diesel share of new registrations from 1990-2005 is based on industry data. Projected shares from 2006-2007 are based on industry forecasts, and from 2008-2020 based on authors’ projections. The focus of this study is on the area between the actual/projection curve and the “no growth” curve which is split into three time intervals defined by the applicable emission standard: Euro III, Euro IV, and post-Euro IV.  50% 45%  Euro IV standards apply in 2006  actual diesel share  Diesel % of total new registrations  projected diesel share  40%  Post-Euro IV diesels (with PM trap) in 2009  Most Euro III and Euro IV diesels scrapped by 2020  “no growth” diesel  35% CO2 tax and Euro III standards apply in 2000  30% 25%  Euro IV interval  post-Euro IV interval  Euro III interval  20% 15% 10% 5% 0% 1990  1995  2000  2005  2010  2015  2020  Year  3.3  RESULTS  Figure 3.4 summarizes the main results of this study. The estimated additional diesel vehicle counts (Box A), emissions and fuel consumption (Box B), ambient concentrations (Box C), and health effects (Box D) are shown. 60  Figure 3.4  Summary results of the impact of additional diesels in the UK from 2001-2020 box A: number of cars petrol to diesel (2001-2020 total) Euro III  +0.7  million diesel cars  Euro IV  +1.6  million diesel cars  post-Euro IV  +7.0  million diesel cars  box B: change in emissions & fuel 2001-2020 total  per 106 Euro III vehicles  per 106 Euro IV vehicles  per 106 post-Euro IV vehicles  PM2.5, kt  +12  8.3  3.7  0  NOx, kt  +93  66  30  0  HC, kt  -73  -34  -31  0  CO, kt  -204  -126  -73  0  CO2, Mt  -7.0  -2.3  -1.3  -0.9  -20  -7  -4  -3  6  fuel, 10 bbl oil  box C: change in air quality (annual average) PM2.5  +0.043  ↑  NO2 ozone CO  ↑  or  g/m  3  g/m3  ↓  ↓  g/m  3  g/m  3  box D: change in public health mortality (2001-2020 total)  mortality per 106 vehicles  mortality per Mt CO2 abated  Euro III  +910  +1,320  570  Euro IV  +940  +590  870  post-Euro IV  0  0  0  The estimated number of additional diesels over the 20 year study period is 0.7 million for Euro III, 1.6 million for Euro IV, and 7 million for post-Euro IV. Figure 3.5 shows the cumulative time series of additional diesels disaggregated by emission class, which is quantified by adding newly registered cars less the estimated number of scrapped cars.  61  Figure 3.5  Estimates of additional diesels in the UK 2001-2020 disaggregated by Euro emission class. “Additional diesels” are defined as the number of petrol vehicles switched to diesel beyond the “no growth” estimate. Euro III and Euro IV emission standards apply in 2001 and 2006, respectively. Early adoption of some Euro IV diesels is incorporated into our estimates. Legislation to harmonize diesel and petrol particulate matter emission limits is proposed by 2009, described as “post-Euro IV” in this study. 7 post-Euro IV  Cumulative additional diesels, millions  Euro IV Euro III  6 CO2 based Vehicle Excise Duty (VED)  5  Proposed harmonization of petrol and diesel car emission limits (post-Euro IV)  4  Euro IV CO2 emission based limits Company apply Car Tax  3  2  1  0 2001  2003  2005  2007  2009  2011  2013  2015  2017  2019  Year  Over the 20 year study period, additional diesels are estimated to increase PM10 and NOx emissions by 12 kilo-tonnes and 93 kilo-tonnes, respectively. HC and CO emissions are estimated to decrease by 73 kilo-tonnes and 204 kilo-tonnes, respectively. CO2 emissions are estimated to decrease by 7 mega-tonnes. These total changes in emissions were obtained by integrating the annual time series of emissions shown in Figure 3.6 over the 20 year study  62  period. It is important to note that the lower polluting emission classes provide progressively lower CO2 and fuel saving benefits. Figure 3.6  Estimated changes in emissions from 2001-2020 due to additional diesels. The solid lines (y-axis to left) show estimated changes in emissions of common air contaminants in kilo-tonnes, while the dashed line (y-axis to right) shows CO2 in mega-tonnes. Diesels emit higher rates of PM10 and NOx, and lower rates of HC, CO, and CO2. Emissions of common air contaminants are assumed to be harmonized for diesel and petrol vehicles beginning 2009, so differences in all emissions except CO2 approach zero from 2009-2020 as higher polluting diesels are scrapped.  0.5  10  0.0  0  -5  NOx  -10  -0.5  PM2.5  Change in CO 2 emissions, mega-tonnes/year  Change in PM 2.5, NOX, HC, and CO emissions, kilo-tonnes/year  5  HC CO  -15  CO2 -20 2000  -1.0 2002  2004  2006  2008  2010 2012 Year  2014  2016  2018  2020  As a result of the 12 kilo-tonne increase in particulate emissions, the average annual change in ambient PM10 concentration was estimated to be 0.043 µg/m3 over the 20 year study period. Ambient concentration of NO2 is determined to increase by an un-quantified amount, while CO decreases. Changes in ambient ozone concentrations were not estimated because of the complexity and uncertainty in atmospheric chemistry that would result from an increase in NOx and a decrease in HC, the two principal precursors. No known UK studies have modelled such a 63  scenario. The available studies conclude that ozone formation might be limited by NOx or HC under different conditions [40]. The most recent study indicated that reductions in HC usually improved ozone while “NOx emission control gave a more complex response, which was metric and region specific. Generally NOx emission control had an adverse effect on ozone air quality.” [41]. Therefore we show that ambient ozone levels may increase or decrease as indicated in Figure 3.4. Over the 20 year study period, the average annual mortality is estimated at 90 deaths per year applying the central concentration-response coefficient. The low and high mortality estimates are 20 and 300 deaths per year, respectively. Annual average hospitalizations total 32 per year. All mortality and morbidity impacts are associated with the Euro III and Euro IV additional diesels, based on the assumption of harmonization of emissions for diesel and petrol in the postEuro IV standards. Included in Box D of Figure 3.4 are the ratios of the central mortality estimate per million additional diesels and per mega-tonne of CO2 abated, averaged over the 20 year study period. The additional mortality per million Euro III diesels and Euro IV diesels is estimated at 1,320 deaths and 590 deaths, respectively. The average mortality per mega-tonnes of CO2 abated is 570 for Euro III, 460 for Euro IV, and 0 for post-Euro IV.  3.4  DISCUSSION  There are many uncertainties and limitations associated with this analysis. Section A.4 of Appendix A contains a detailed discussion of how these uncertainties and limitations are likely to have over- or underestimated fuel savings, emissions, exposure, and health effects estimates in the following five areas: 1.  Number and emission class of additional diesels. The number of registered vehicles subject to successive EU emission standards is not stated in the data. We used the range of manufacturers’ offerings meeting the different standards as the guide to apportion registrations to each emission class. Our assumptions with respect to emission class likely result in a low bias of the estimated air quality and health impacts. Additional discussion is provided in Section A.4.1 of Appendix A.  2.  Annual kilometers travelled. The assumption of equal annual travel distance for petrol and diesel vehicles does not account for variation in driver types (company car and high travel distance drivers), nor an economic rebound effect. Our assumption of driver types is consistent with a saturation of high travel distance drivers choosing diesels as described by Schipper [5]. In general there is a 64  measured rebound effect elasticity of 0.1-0.3% increase in annual travel distance for every 1% decrease in fuel costs per kilometer [42]. Therefore our annual travel distance assumptions likely underestimate air quality impacts and overestimate CO2 reductions and fuel savings. .Additional discussion is provided in Section A.4.2 of Appendix A. 3.  Spatial distribution of vehicles. Available evidence indicates that our assumptions of spatial distribution of vehicles (urban/rural/motorway) does not substantially bias our estimates, either high or low. Additional discussion is provided in Section A.4.3 of Appendix A.  4.  PM10 emissions and ambient concentrations. Our assumptions of PM10 emissions and ambient concentrations, including meteorology, secondary particulate matter (sulphate, nitrates, and organic), fuel quality, brake and tire wear, and vehicle age likely bias our estimates high in some cases and low in others. Overall, the air pollutant emissions and ambient concentrations are likely underestimated. Additional discussion is provided in Section A.4.4 of Appendix A.  5.  Health effects. Health effects are likely underestimated, specifically the limitation of estimating health outcomes solely from PM10. The likely impacts on ambient levels and subsequently health effects of ozone, SO2, HC, NOx, CO, benzene, and 1,3 butadiene are discussed further in Section A.4.5 of Appendix A.  Given these various sources of uncertainty, we think our basic findings are robust. Overall, the study may be biased in overestimating fuel savings and CO2 emissions and underestimating the health damaging emissions and associated health impacts. Additional disbenefits may also be present that have not been considered in this study, namely black carbon emissions from diesel engines and their exacerbation of climate change [43, 44]. We estimate that consumers switching from petrol to diesel cars in the UK over the time period of 2001-2020 will reduce CO2 by 7 Mt and saves 20 million barrels of oil. However, ancillary air quality effects hinge upon the fuel properties and conversion technology, not just the quantity of fuel consumed, and adverse air quality is estimated to result in 90 additional deaths annually (range 20-300). The CO2 reductions, fuel savings, and additional mortality are not necessarily all attributable to the CO2 policies. Econometric models are a valuable tool to explain changes in consumer choice of fuel and vehicle types [45], but to our knowledge no econometric models have been developed to accurately quantify UK CO2 policy influence on diesel growth. Various reports attribute no impact to the VED CO2 tax [7], while estimates of the impact of the company 65  car CO2 tax on company car diesel growth range from 33% [8] to 100% [9]. Other factors such as the European manufacturer voluntary CO2 program, oil prices, and technological change are potential influences as well [8, 10, 46], but not fuel prices or taxes (Figure 3.2). To the extent that CO2 policies contributed to diesel growth, coordinating CO2 controls with tightening of emission standards would save lives. Because of the uncertain CO2 policy impact, our estimates of emissions, fuel savings, and health impacts per unit quantity of additional diesel cars (Figure 3.4) are emphasized. As the rest of the EU and other developed countries prepare to integrate transportation into climate mitigation programs, the lessons learned from the UK experience can help inform the design of climate policies in the transportation sector to better balance near-term health effects with long-term climate mitigation.  3.5  ACKNOWLEDGEMENTS  We thank Lester Lave, Michael Brauer, Julian Marshall, and two anonymous reviewers for insightful comments on drafts of this article. We thank the University of British Columbia Bridge Program, AUTO21: B06 BLC, and Carnegie Mellon University’s NSF-supported Center for Integrated Study of the Human Dimensions of Global Change (SBR-9521914), and the Center for Climate Decision Making (SES-0345798). We are also grateful for generous support from the Exxon-Mobil Education Foundation. All errors are a measure of the fallibility of the authors and of how much more we need to learn before making sound policy.  66  3.6 1.  REFERENCES Davis, D.; Kjellstrom, T.; Sloof, R.; McGartland, A.; Atkinson, W.; Barbour, W.; Hohenstein, W.; Nagelhout, P.; Woodruff, T.; Divita, F.; Wilson, J.; Deck, L.; Schwartz, J., Short-term Improvements in Public Health from Global-Climate policies on Fossil-Fuel Combustion: An Interim Report. The Lancet 1997, 350, 1341-1349.  2.  Cifuentes, L.; Borja-Aburto, V.; Gouveia, N.; Thurston, G.; Davis, D., Hidden Health Benefits of Greenhouse Gas Mitigation. Science 2001, 293, 1257-1259.  3.  IPCC, Climate Change 2001: Mitigation. Intergovernmental Panel on Climate Change: 2001.  4.  van Vuuren, D.; Cofala, J.; Eerens, H.; Oostenrijk, R.; Heyes, C.; Klimont, Z.; den Elzen, M.; Amann, M., Exploring the ancillary benefits of the Kyoto Protocol for air pollution in Europe. Energy Policy 2006, 34, (4), 444-460.  5.  Schipper, L.; Marie-Lilliu, C.; Fulton, L., Diesels in Europe: Analysis of Characteristics, Usage Patterns, Energy Savings, and CO2 Emission Implications. Journal of Transport Economics and Policy 2002, 36 part 2, 305-340.  6.  IEA Energy Statistics Manual; International Energy Agency (www.iea.org): 2005.  7.  MORI Assessing the Impact of Graduated Vehicle Excise Duty: Quantitative Report; UK Department for Transport (www.dft.gov.uk) and MORI www.mori.co.uk: 2003.  8.  IR Report on the evaluation of the company car tax reform: stage two; UK Inland Revenue (www.inlandrevenue.gov.uk): 2006.  9.  Lane, B., Car buyer research report: consumer attitudes to low carbon and fuel-efficiency passenger cars.; Low Carbon Vehicle Partnership (UK): 2005.  10.  SMMT UK New Car Registrations by CO2 Performance 2005 Annual Report; The Society of Motor Manufacturers and Traders Ltd. (www.smmt.co.uk): 2006.  11.  SMMT, New vehicle registrations in the UK. In Society of Motor Manufacturers and Traders (www.smmt.co.uk): 2005.  12.  Jacobson, M.; Seinfeld, J.; Carmichael, G.; Streets, D., The effect on photochemical smog of converting the U.S. fleet of gasoline vehicles to modern diesel vehicles. Geophysical Research Letters 2004, 31, L02116.  67  13.  Krupnick, A.; Burtraw, D.; Markandya, A., In Ancillary Benefits and Costs of Greenhouse Gas Mitigation Strategies, Expert Workshop on the Ancillary Benefits and Costs of Greenhouse Gas Mitigation Strategies, Washington, D.C., March 27-29, 2000; Washington, D.C., 2000.  14.  Swart, R.; Amann, M.; Raes, F.; Tuinstra, W., A Good Climate for Clean Air: Linkages between Climate Change and Air Pollution. Climatic Change 2004, 66, 263-269.  15.  NAEI National Atmospheric Emission Inventory online database (www.naei.org.uk): September 22, 2006, 2005.  16.  TRL Data required to monitor compliance with the End of Life Vehicles Directive; PR SE/483/02; Transport Research Laboratory, UK: 2003.  17.  VCA, Vehicle Database. In Vehicle Certification Agency (www.vcacarfueldata.org): 2005.  18.  Delcan Strategies To Reduce Greenhouse Gas Emissions From Passenger Transportation In Three Large Urban Areas - Final Report (by Delcan Corporation, KPMG, and A.K. Socio-Technical Consultants Ottawa Inc.); Transport Canada (www.tc.gc.ca): 1999.  19.  DFT Transport Statistics (online data); UK Department for Transport (www.dft.gov.uk): September 15, 2006, 2005.  20.  Martuzzi, M.; Krzyzanowski, M.; Bertollini, R., Health impact assessment of air pollution: providing further evidence for public health action. European Respiratory Journal 2003, 21, (Supplement 40), 86s-91s.  21.  Künzli, N.; Kaiser, R.; Medina, S.; Studnicka, M.; Chanel, O.; Filliger, P.; Herry, M.; Horak Jr., F.; Puybonnieux-Texier, V.; Quénel, P.; Schneider, J.; Seethaler, R.; Vergnaud, J.-C.; Sommer, H., Public-health impact of outdoor and traffic-related air pollution: a European assessment. The Lancet 2000, 356, 795-801.  22.  Smith, K., Why Particles? Chemosphere 2002, 49, 867-871.  23.  WHO, Comparative Quantification of Health Risks: Global and Regional Burden of Disease due to Selected Major Risk Factors. World Health Organization: Geneva, 2004.  24.  McCubbin, D.; Delucchi, M., The health costs of motor vehicle related air pollution. Journal of Transport Economics and Policy 1999, 33, (3), 253-286.  25.  Watkiss, P.; Jones, G.; Kollamthodi, S., An Evaluation of the Air Quality Strategy Additional Analysis: Local Road Transport Measures. prepared for UK Department for Environment, Food and Rural Affairs (DEFRA): 2004. 68  26.  CERC, Source Apportionment for London using ADMS-Urban. prepared by Cambridge Environmental Research Consultants Ltd. for: UK Department for Environment, Food and Rural Affairs (DEFRA) and The National Assembly for Wales, The Scottish Executive and the Department of the Environment in Northern Ireland: 2004.  27.  Ntziachristos, L.; Giechaskiel, B.; Pistikopoulous, P.; Fysikas, E.; Samaras, Z., In Particle Emission Characteristics of Different On-Road Vehicles, PM Characterization in Diesel and Gasoline Exhaust Gas, Yokohama, Japan, 2003; Society of Automotive Engineers (SAE): Yokohama, Japan, 2003.  28.  Stedman, J.; Bush, T.; Murrells, T.; Hobson, M.; Handley, C.; King, K., Quantification of the health effects of air pollution in the UK for revised PM10 objective analysis. AEA Technology and UK Department of Health: 2002.  29.  Stedman, J.; Bush, T.; Murrells, T.; Hobson, M.; Handley, C.; King, K., Revised PM10 projections for the UK for PM10 objective analysis. Prepared by AEA Technology for UK Department for Environment, Food and Rural Affairs (DEFRA) and The National Assembly for Wales, The Scottish Executive and the Department of the Environment in Northern Ireland: 2002.  30.  Stedman, J.; Bush, T.; Murrells, T.; King, K., Baseline PM10 and NOx projections for PM10 objective analysis. Prepared by AEA Technology for UK Department for Environment, Food and Rural Affairs (DEFRA) and The National Assembly for Wales, The Scottish Executive and the Department of the Environment in Northern Ireland: 2001.  31.  Marshall, J.; Teoh, S.; Nazaroff, W., Intake fraction of nonreactive vehicle emissions in US urban areas. Atmospheric Environment 2005, 39, 1363-1371.  32.  Marshall, J.; Riley, W.; McKone, T.; Nazaroff, W., Intake fraction of primary pollutants: motor vehicle emissions in the South Coast Air Basin. Atmospheric Environment 2003, 37, 3455-3468.  33.  Pope, C. A.; Burnett, R.; Thun, M.; Calle, E.; Krewski, D.; Ito, K.; Thurston, G., Lung, Cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. Journal of the American Medical Association 2002, 287, (9), 1132-1141.  34.  Ostro, B., Outdoor air pollution: Assessing the environmental burden of disease at national and local levels; World Health Organization: 2004.  35.  COMEAP, Quantification of the effects of air pollution on health in the United Kingdom. The Stationary Office London (www.tsonline.co.uk): 1998. 69  36.  Kunzli, N.; Medina, S.; Kaiser, R.; Quenel, P.; Horak, F.; Studnicka, M., Assessment of deaths attributable to air pollution: should we use risk estimates based on time series or cohort studies? American Journal of Epidemiology 2001, 153, 1050-1055.  37.  Roosli, M.; Kunzli, N.; Brauen-Fahrlander, C.; Egger, M., Years of life lost attributable to air pollution exposure in Switzerland: dynamic exposure-response model. International Journal of Epidemiology 2005.  38.  Dockery, D.; Pope, C.; Xu, X.; Spengler, J.; Ware, J.; Fay, M.; Ferris, B.; Speizer, F., An association between air pollution and mortality in six U.S. cities. New England Journal of Medicine 1993, 329, 1753-1759.  39.  Jerrett, M.; Burnett, R.; Ma, R.; Pope, C.; Krewski, D.; Newbold, K.; Thurston, G.; Shi, Y.; Finkelstein, N.; Calle, E.; Thun, M., Spatial Analysis of Air Pollution and Mortality in Los Angeles. Epidemiology 2005, 16, (6), 727-736.  40.  Hayman, G.; Jenkin, M.; Pilling, M.; Derwent, R., Modeling of Tropospheric Ozone Formation. prepared by AEA Technology for UK Department for Environment, Food and Rural Affairs (DEFRA): 2002.  41.  Hayman, G.; Abbott, J.; Thomson, C.; Bush, T.; Kent, A.; Derwent, D.; Jenkin, M.; Pilling, M.; Richard, A.; Whitehead, L., Modeling of Tropospheric Ozone (AEAT/ENV/R/1858 issue 2); Department for Environment, Food and Rural Affairs, the Scottish Executive, the National Assembly of Wales and the Northern Ireland Department of the Environment: 2005.  42.  Greening, L.; Greene, D.; Difiglio, C., Energy efficiency and consumption - the rebound effect - a survey. Energy policy 2000, 28, 389-401.  43.  Jacobson, M., Correction to ‘‘Control of fossil-fuel particulate black carbon and organic matter, possibly the most effective method of slowing global warming’’. Geophysical Research Letters 2005, 110, D14105.  44.  Sato, M.; Hansen, J.; Koch, D.; Lacis, A.; Ruedy, R.; Dubovik, O.; Holben, B.; Chin, M.; Novakov, T., Global atmospheric black carbon inferred from AERONET. Proceedings of the National Academy of Sciences 2003, 100, (11), 6319-6324.  45.  Verboven, F., Quality-based tax discrimination and tax incidence: evidence from gasoline and diesel cars. The RAND Journal of Economics 2002, 33, (2), 275-297.  46.  Fleet News Diesel sales rocket with fuel price rise; March, 2006.  70  4  REGULATING CAR MASS FOR CONCURRENT TRAFFIC SAFETY AND CLIMATE MITIGATION BENEFITS7  4.1  INTRODUCTION  We examined the relationship between vehicle mass, traffic collision health risks, and tailpipe CO2 emissions for passenger cars using the UK as a case study. Traffic collisions continue to present substantial public health risks in the UK [1] and globally [2]. Transportation is a large and growing source of greenhouse gases, and CO2 mitigation policies are being designed to encourage consumers to choose cars with better fuel efficiency and lower CO2 emission rates [3, 4]. Technical factors that determine tailpipe CO2 emission rates are: a) vehicle mass, b) drivetrain efficiency, and c) carbon intensity of the fuel. Mass reduction is potentially a win-win strategy with dual benefits of climate mitigation and nearterm public health because mass is a determinant of both traffic safety and tailpipe CO2 [5]. Mass is not an attribute desired by consumers [6], but is a tangible parameter that is amenable to regulation [7], and technology exists to reduce mass without compromising vehicle size [7, 8]. Amongst the myriad of technologies to improve vehicle fuel economy and CO2 emission rates, vehicle mass reduction via lightweight technology is likely the only option that also has the potential for traffic safety benefits without compromising consumer preferences. An additional consideration is that mass reduction synergizes well with other traffic safety and climate mitigation strategies. For any improvements in traffic safety technology (e.g., pedestrian safety) or management (e.g., enforcement [9]), reducing mass is expected to provide positive (albeit uncertain) incremental benefits [7]. Similarly, reducing mass in combination with vehicle technology to reduce tailpipe CO2 (e.g., biofuels, diesel, or hybrid) is expected to provide incremental positive benefits [8]. When examining car mass reduction as a policy strategy, we hold that it is critical to recognize two issues at the outset. First, recognize that traffic safety is not just measured by fatality counts for car occupants [10]. Traffic safety encompasses risk of both fatality and injury for multiple vehicle types, road users, injury rates, and crash events as illustrated in Figures 4.1 and 4.2 [11].  7  A version of this chapter has been submitted for publication: Mazzi, E., H. Dowlatabadi, and M. Kandlikar,  “Regulating car mass for concurrent traffic safety and climate mitigation benefits” 2010.  71  Moreover, traffic safety risks can be quantified as risk ratios [12], conditional risks [5], or absolute risks [13]. UK traffic safety goals include reductions in total killed or seriously injured (KSI), total children KSI, the slight injury rate, and vulnerable road user casualties such as bicyclists and pedestrians [11]. Second, it is critical to recognize that mass reduction is not just about uniform mass reductions, but also about the fleet distribution of mass [14]. Uniform mass reduction has not been conclusively shown to reduce overall fatalities [13, 15, 16]. However, the risks of fatality between specific mass groups are measurably impacted by uniform mass reduction, leading to a change in the incidence of fatality and injury risk across the public [14, 16, 17].  Figure 4.1  1997-2006 time series of fatalities (left Y-axis) and KSI (right Y-axis) per billion passenger km for key road user groups in the UK. This illustrates the variation of casualty rates between road users (e.g., motorcycle occupant rates are 40-50 times car occupants), and that fatality rates do not always parallel injury (e.g., bicycle fatal and KSI rates 2003-2006).  140  1,600 pedestrian, fatal  1,400  120  100  1,000 80 800 60 600 40  400  KSI per billion passenger-km  Fatal per billion passenger-km  cars, fatal  1,200  bicycles, fatal  motorcycles, fatal  pedestrian,KSI  cars, KSI  bicycles, KSI  motorcycles, KSI  20  0 1996  200  0 1998  2000  2002  2004  2006  72  Figure 4.2  Distribution of 2,946 fatalities and 2,714 fatal crash events in the UK for 2007. There were also 30,720 KSI and 27,036 KSI events in 2007 with a similar distribution as fatalities.  Fatalities  Fatal crash events  other 5%  ≥ 3 veh 16%  1 car 18%  bicycle 5% motorcycle 20%  cars 48%  2 veh (1 not car) 25%  2 car 14% car ped 13%  pedestrian 22%  1 veh (not car) 7%  other veh ped 7%  A substantial body of research has found that vehicle mass is a key predictor of various measures of traffic safety risk [12, 15, 16]. However, mass is also strongly correlated with other traffic safety factors including physical metrics such as size and power [7], and behavioral metrics such as selling price [18]. Some authors have hypothesized that increased car mass itself (or its physical correlates like size and power) is partly causal in increasing risk taking behavior [19], thus illustrating the complexity of interactions between physical and behavioral factors in traffic safety. It has been argued that mass is not fundamental to traffic safety [18], while other research implies that mass offers a protective effect [12, 15]. We highlight that research consistently shows delta-V is a significant predictor of risk of injury and fatality [7, 20], and that the mass of vehicles is fundamental in calculating delta-V applying conservation of momentum. A more detailed listing of factors that influence traffic safety, along with references, is provided in Appendix Section B.1. Table B1 provides a list of important studies that evaluated the role of vehicle mass in traffic safety. 73  The scope of this research is strictly passenger cars, and CO2 emission rates as measured by equivalent gCO2/km (including N2O and CH4) emitted at the tailpipe. We do not examine all factors that determine annual CO2 emissions, such as annual travel distance. We primarily examine vehicle attributes independent of roads, behavior, and other variables. We also do not examine the relative merits of traffic safety policies aimed at road (e.g., congestion controls) or behavior (e.g., speed enforcement). Therefore we do not address, nor do we argue, that one of these domains is more or less important than the other. The contribution of this research is to quantify multiple traffic safety risk metrics and relate them to CO2 emissions, using UK data. The significance is to quantify potential benefits of regulating vehicle mass for concurrent benefits in traffic safety and climate mitigation. The UK is important as a global early adopter of climate mitigation policies for passenger cars with tax regimes in place since 2001-2002 [21, 22], as well as being the second largest market subject to the European automobile CO2 policies. Although every country (and regions within countries) will have different traffic safety priorities, we expect the UK results can be generalized to other developed countries.  4.2  DATA AND METHODS  Our principal data source was the UK Data Archive which has three databases for each calendar year: collisions (“accidents”), casualties, and vehicles [23]. We obtained these databases for calendar years 1994-2005, which have 5,354 driver fatalities. The UK Data Archive databases included only generic specification of vehicles such as cars, motorcycles, and bicycles. The UK Department for Transport (DFT) provided additional databases for years 1994-2005 with car make, car model, engine size, fuel type, body type, and other parameters [24] which we linked to the UK Data Archive fields. Using UK data we quantified traffic collision risks and car CO2/km rates in four different ways. Appendix Section B.2.1 provides more a detailed description of the data used in this study. Section B.2.2 with Figures B1 through B9 provides graphical descriptive statistics that is a useful reference for understanding specific trends in traffic safety in the UK. The UK Data Archive data bases do not have curb mass, which is a critical limitation. We were able to make use of independent databases and proxy indicators of vehicle mass as described in Appendix Sections B.2.3. First we calculated the relative risk (RR) of driver fatality for lighter cars versus heavier cars, described as the “first law” of two-car collisions [12], then compared the RR and CO2 emission 74  rates [25, 26]. The “first law” is a power law relationship between vehicle mass and RR of driver fatality of the following form [12]: RR ≡ μλ where: RR ≡ n1/n2 = (number driver fatalities in lighter cars) / (number of driver fatalities in heavier cars) μ ≡ (M2/M1) = (mass of heavier car) / (mass of lighter car) For the RR analysis, two-car collisions were included for years 1996-2005 and where both vehicles were also of model years 1996-2005. Consequently, the final dataset was comprised of 280 cars, 140 crash events, and 144 driver fatalities. Neither the UK Data Archive databases nor the supplemental databases provided by DFT included vehicle mass. We were able to determine curb mass using an independent data source 26]. Appendix Section B.2.4 provides a more detailed description of the study methods for estimating RR. RR is a robust metric as many confounding factors cancel out, and it is considered to be independent of behavior [12, 27]. In interpreting RR, the main consideration is whether the numerator and denominator groups differ substantially in major confounding factors such as car crashworthiness, seat belt use, age and sex of drivers, collision speed, and collision points of impact [12, 20]. In this study we have no data on seat belt use. We have surrogate data on car crashworthiness (model year) and collision speed (road speed limits). We have direct data on age and sex of drivers, and points of impact. Overall, based on comparisons summarized in Table 4.1, we expect confounding factors to be minimal in the RR calculation. Our second analysis is a comparison of DFT’s conditional KSI risk estimates for two-car collisions[28] and CO2 emission rates [25]. As they are only for makes and models, DFT’s estimates do not relate risk to more fundamental parameters of interest such as mass, size, stiffness, and safety equipment. For each make and model of car model years 1995-2004 in DFT’s analysis, we quantified the CO2 emission rate [25, 26]. We first examined the relationship between risk and CO2 graphically, then by regressing two-car conditional risk on the CO2 emission rate.  75  Table 4.1  Comparison of key parameters for the “first law” RR analysis data set: heavier cars and their drivers cf. lighter cars. Variable  Heavier Car  Lighter Car  140  140  1998.8  1998.1  Average car curb mass, kg  1339  1044  Average car length, m  4.38  3.98  Average CO2 emission rate, gCO2/km  201  172  Average driver age  44.5  46.4  Driver % male  75.0%  73.6%  Slightly injured  39.6%  10.7%  Seriously injured  34.5%  12.1%  Fatal  25.7%  77.1%  KSI  60.2%  89.2%  Number of drivers (vehicles) Average car model year  For our third analysis we estimated changes in absolute risk, measured by fatalities per year, as a result of mass distribution scenarios for the UK on-road car fleet over the time period 20002005. We computed CO2 emission rates and changes in absolute fatality risk for three collision types: single-car, car-pedestrian, and two-car. We estimated fatality risk using a method originally developed by Mengert and Borener [29] because it satisfies the study objectives and is feasible given our available data fields. The U.S. National Highway Traffic Safety Administration has replicated this method and concluded results were consistent with its detailed statistical analysis of vehicle, road, and driver behavior variables [15]. Changes in fatality risk are estimated based on changes in the mass distribution for on-road car fleets. We derived estimates of curb mass using a database provided by JATO Dynamics Ltd [30] which included size (length, width, wheelbase), curb mass, fuel economy, and other parameters. We computed ordinary least squares regression relationships between mass, size, CO2 emission rate, and engine size. We were able to use engine size categories as a suitable proxy to create a mass category field to link to the UK Data Archive. Standard errors for fatality risk were estimated 76  assuming a poisson distribution [12]. We highlight that a key limitation of UK data, as compared to the U.S., is sample size. Total fatalities in the U.S. are over one order of magnitude greater, therefore standard errors as a percentage of the mean fatality counts are typically 3 times larger for UK data. Appendix Section B.2.5 provides a more detailed description of the study methods for the absolute risk analysis. Scenarios were developed for the absolute risk analysis to compare baseline fatality risk to alternate scenarios as shown in Figure 4.3. In our scenarios, the year 1999 is the initial condition and we simulate plausible changes in fleet mass composition over the years 2000-2005. All mass reduction strategies will reduce tailpipe CO2, but there are three options to reduce mass in a way that might produce beneficial traffic safety outcomes: impose a lower mass limit, manipulate the distribution without upper or lower limits based on traffic safety risks by group and collision type, or impose an upper mass limit [7]. A lower mass limit conflicts with climate and energy security goals. Manipulating the distribution by groups and collision types is infeasible due to large uncertainties and unrealistic administrative requirements. Hence an upper limit on mass is the most feasible option and is built into our scenarios. The “Increase lighter car” scenario was developed to simulate policies that induce rapid growth of lighter cars, the “Increase lighter cars and prohibit heavier cars” to simulate prohibition of heavier cars (approximately 1,600 kg mass and larger), and the “Constant mass” scenario to simulate changes in fatality risk independent of fleet average mass. A realistic upper limit on mass is 1,600 kg based on research that shows that 33% curb mass reduction for pickups, minivans, and sport utility can be achieved at vehicle price premiums well under 10% [8]. Our fourth risk analysis was an estimation of fleet average RR for two-car collisions using the same 2000-2005 scenarios as the absolute risk analysis shown in Figure 4.3. A distribution of mass for on-road cars for each scenario was developed using fleet distribution by engine size, and then mass distributions were estimated using the statistical relationships using the JATO data. Appendix Section B.2.6 provides a more detailed description of these statistical relationships. Two-car RR was estimated assuming fatal collisions occur randomly, with the power law exponent (λ) varied parametrically at values 2, 4, and 6. Analytica™ version 4.1 software was used to simulate randomized crashes using the Latin Hypercube sampling algorithm.  77  Baseline plus three alternative scenarios for years 2000-2005 used in the absolute risk analysis and the RR fleet composition simulation. “Lighter” group is comprised of engine size ranges 701-1,000 cubic centimeter (CC), 1,001-1,200 CC, and 1,201-1,500 CC. “Mid-mass” includes 1,501-1,800 CC and 1,801-2,000 CC. “Heavier” includes 2,001-2,500 CC, 2,501-3,000 CC, and 3,000 CC and over. 700 CC and under, were not included because there were too few vehicles, annual fatality counts were zero or near zero, and it was not possible to estimate risks. Scenario descriptions are provided in the text. 1501-1800 (M)  35%  1201-1500 (L)  30%  1801-2000 (M)  25%  1001-1200 (L)  20%  701-1000 (L)  15%  2001-2500 (H)  10%  2501-3000 (H)  40%  1995  1997  1999  2001  2003  2005  1501-1800 (M) 1201-1500 (L) 1801-2000 (M)  25%  1801-2000 (M)  30% 1001-1200 (L)  25% 701-1000 (L)  20% 2001-2500 (H)  15% 2501-3000 (H)  10%  3000 & over (H)  5%  700 & under  1995  40%  Scenario: Increase light cars and prohibit heavy (decrease mid-mass cars)  30%  1201-1500 (L)  35%  0% 1993  700 & under  35% Registered (on road) car shares  40%  3000 & over (H)  5% 0% 1993  1501-1800 (M)  Scenario: Increase light cars (BAU heavier cars, decrease mid-mass cars)  Baseline (actual)  1001-1200 (L)  20% 701-1000 (L)  15% 2001-2500 (H)  10%  1997  1999  2001  2003  2005  Scenario: Constant mass  1501-1800 (M)  35% Registered (on road) car shares  Registered (on-road) car shares  40%  Registered (on-road) car shares  Figure 4.3  1201-1500 (L)  30%  1801-2000 (M)  25% 1001-1200 (L)  20% 701-1000 (L)  15% 2001-2500 (H)  10%  2501-3000 (H)  2501-3000 (H)  5%  5% 3000 & over (H)  0% 1993  1995  1997  1999  2001  2003  2005  700 & under  3000 & over (H)  0% 1993  1995  1997  78  1999  2001  2003  2005  700 & under  4.3  RESULTS  The results of the “first law” two-car RR analysis are summarized in Figure 4.4. The data set was suitable to develop five groups with curb mass ratio ( ranging from 1.06-1.55. The regression analysis for the relationship RR =  resulted in = 5.31 (95% CI 3.9-6.7; R2 = 0.71, p <0.0005), and a line fit plot is provided in Figure B16). Figure 4.4  “First law” RR risk of driver fatality in two-car collisions and ratio of CO2 emission rates for UK cars 1995-2005. RR of driver fatality in two-car collisions is shown on the left vertical axis (mean +/- one standard error). Ratio of tailpipe CO2 emission rate (mean +/- one standard deviation) is shown on the right vertical axis. A horizontal line is drawn showing where CO2/km ratio =1.0. 1.6  1.4  12  1.2 10 1.0 8 0.8 6 0.6 4 0.4  2 2-car RR data +/- 1 std error (left Y-axis) RR fitted line, R^2 = 0.71, p<0.0005 (left Y-axis) CO2/km: heavy car / light car +/- 1 std dev (right Y-axis) CO2/km ratio = 1.0 (reference line)  0 1.0  1.2  1.4  CO2/km ratio: heavier cars / lighter cars  RR ≡ fatality risk lighter car drivers / fatality risk heavier car drivers  14  0.2  0.0  1.6  mass ratio (μ): heavier car / lighter car  79  Based on visual inspection of the data in Figure 4.4, we also attempted to fit the data to a linear statistical model with the following results (plot in Figure B16): RR = α + β * µ (R2 = 0.96) α = - 14.1 [95% CI -20.9 to -7.4] p < 0.007 β = 14.5 [95% CI 9.1 to 19.9] p < 0.004 Appendix Section B.3, Figure B16 includes line fit plots for both the “first law” and linear statistical equations. The results of the conditional risk analysis are provided in Figure 4.5. DFT-derived conditional KSI estimates are reported as a standardized percentage, rounded to whole numbers. Hence the reported risks can be viewed as a conditional risk category ranging from values 1-10. Car models are plotted in order of decreasing conditional risk category. Regressing percent risk on CO2/km revealed that decreasing CO2/km is a statistically significant predictor for increasing driver KSI risk: Driver % KSI risk = 8.4 – 0.020 * CO2/km (R2 = 0.22; p<0.0002; intercept 95% CI = 6.5-10.3; slope 95% CI = 0.010-0.030) Appendix Section B.3, Figure B17 provides a line fit plot for the conditional risk statistical model. The results of the absolute risk analysis are presented in Table 4.2 and Figure 4.6. The “Increase lighter cars” scenario predicts declines in single-car, car-pedestrian, and two-car fatalities of 10%, 14%, and 13%, respectively by 2005 if rapid growth in lighter cars had begun in 2000 (with BAU heavier car shares). Under this scenario, our model predicts a reduction of 126 deaths per year could have been achieved by 2005 for these three collision types combined. Results are significant at the level of one standard error, which is an accepted level of significance commonly applied in traffic safety [12] and also described as “likely” in climate policy analysis[3]. The “Increase lighter cars and prohibit heavier cars scenario” predicts that rapid growth of lighter cars combined with prohibition of heavy cars entering the fleet beginning in 2000 produces the same 10% reductions in single-car fatalities, and steeper reductions in car-pedestrian and two-car fatalities of 18% and 17%, respectively. Under this scenario, our model predicts a reduction of 154 deaths per year could have been achieved by 2005 for these three collision types combined. Single-car and carpedestrian results are significant at one standard error, while two-car results are significant at two standard errors.  80  gCO2/km +/- 1 standard deviation (right Y-axis)  15 200  10 150  100  5 50  0 0  81  g CO2 / km  25  Daewoo Matiz (S 95-03) Fiat Seicento (S 98-03) Daewoo Lanos (S/M 97-99) Hyundai Coupe (L/S 95-01) VW Lupo (S 99-02) Peugeot 206 (S 98-04) Mitsubishi Carisma (M 95-03) Daewoo Nubira (M 97-03) Honda CRV (4WD 97-01) MG MGF (L/S 95-03) Rover 25/45 (S/M 99-04) Skoda Felicia (S/M 95-00) Toyota Yaris (S 99-04) Fiat Punto (S 99-03) Vauxhall Corsa (S 99-03) Hyundai Atoz (S 98-00) Renault Clio B (S 98-04) Mini (S 98-03) Citroen Xsara (S/M 97-00) Audi TT (L/S 99-01) Jaguar S Type (L 99-04) Toyota Avensis (M 97-03) Skoda Octavia (M 98-04) Vauxhall Astra (S/M 98-04) Seat Leon (S/M 00-03) Fiat Bravo (S/M 95-01) Renault Megane (S/M 95-04) Ford Ka (S 96-04) Peugeot 307 (S/M 01-04) Skoda Fabia (S/M 00-04) VW Polo (S 98-04) Mercedes A CL (S 98-04) Ford Puma (L/S 97-01) Ford Focus (S/M 98-03) Audi A3 (S/M 96-03) BMW Z3 (L/S 96-01) Ford Galaxy (MPV 95-03) Lexus IS200 (M 99-03) Volvo V70 (L 97-02) Citroen Synergie (MPV 95-00) VW Beetle (S/M 99-03) Peugeot 406 (M 95-03) Suzuki Baleno (S/M 95-00) Nissan Almera (S/M 95-03) Citroen Picasso (MPV 00-04) Citroen Xsara (S/M 00-03) Citroen C5 (M 01-03) Audi A4 (M 95-04) Saab 9-3 (M 98-03) BMW 300C (M 98-04) Audi A6 (L 97-00) LR Freelander (4WD 97-03) Saab 9-5 (L 97-03) Alfa 156 (M 00-03) Rover 75 (L 98-04) Honda Accord (98-03) Renault Scenic (MPV 96-02) Citroen C3 (S 02-04) Jaguar X Type (M 96-03) Mercedes ML (4WD 98-03)  Conditional risk of driver serious injury or fatality, %  Figure 4.5 Relationship between driver two-car conditional KSI risk and vehicle CO2 emission rate in the UK. Model year passenger cars 1995-2004 for crash events during calendar years 2000-2004 are included. Increased CO2/km is a modest but significant predictor of decreased risk. Car make and model size key as defined by DFT: L/S = low/sports, S = small, S/M = small/medium, M = medium, L = large, MPV = multipurpose, 4WD = four wheel drive. Standardized, conditional risk of serious injury or f atality, with biased 95% CI (lef t Y-axis)  350  20 300  250  The “Constant mass” scenario predicts that prohibition of new models of heavy cars entering the fleet beginning in 2000, while increasing mid-mass shares and decreasing light cars to maintain baseline average mass, would increase fatalities for all three collision types. Fatality increases for single-car, car-pedestrian, and two-car collisions were estimated at 6%, 4%, and 7%, respectively. Under the “constant mass” scenario, our model predicts an increase of 61 fatalities per year by 2005 for these three collision types combined. However, results are not significant, even at one standard error. Table 4.2  Results of absolute risk analyses for single-car, car-pedestrian, and two-car collisions. The analysis simulated three alternate scenarios over the time period 2000-2005. Values shown are point estimates of the mean for year 2005. Time series of results including standard errors are plotted in Figure 4.6.  Scenario: Scenario: Increase Increase Year 2005 lighter cars lighter cars fleet Baseline (BAU for (simulation ∆ from and prohibit fleet runs heavier cars, baseline heavier cars (actual) decrease (decrease 2000mid-mass mid-mass 2005) cars) cars). average mass, kg std dev mass, kg average gCO2/km new car gCO2/km single-car fatalities carpedestrian fatalities two-car fatalities sum of fatalities  ∆ from baseline  Scenario: Constant mass (prohibit heavier cars while maintaining baseline mass)  ∆ from baseline  1,341  1,270  -71  1,225  -115  1,341  0  0  286  286  240  240  192  192  185  173  -12  166  -19  185  0  169  162  -7  143  -26  191  22  426  384  -10%  382  -10%  452  6%  306  263  -14%  251  -18%  317  4%  325  283  -13%  270  -17%  348  7%  1,057  931  -126  903  -154  1,118  61  82  Figure 4.6  Single-car, car-pedestrian, and two-car absolute risk results, shown from top to bottom. Changes in annual fatalities are plotted as percentage change relative to the baseline. A horizontal line is plotted at 0% (no change). Error bars represent +/- one standard error. Points are plotted slightly offset in the time scale (x-axis) to make error bars visible. Baseline (actual) Scenario: Increase light cars (BAU for heavier cars, decrease mid-mass cars) Scenario: Increase light cars and prohibit heavy cars (decrease mid-mass cars). Scenario: Constant mass (prohibit heavy cars while maintaining baseline mass)  single car collision annual fatalities, % change relative to baseline  15% 10% 5% 0% -5% -10% -15% -20% -25% 1999  2000  2001  2002  2003  2004  2005  2006  Baseline (actual) Scenario: Increase light cars (BAU for heavier cars, decrease mid-mass cars) Scenario: Increase light cars and prohibit heavy cars (decrease mid-mass cars). Scenario: Constant mass (prohibit heavy cars while maintaining baseline mass)  car-pedestrian collision annual fatalities, % change relative to baseline  15% 10% 5% 0% -5% -10% -15% -20% -25% 1999  2000  2001  2002  2003  2004  2005  2006  two car collision annual fatalities, % change relative to baseline  Baseline (actual) Scenario: Increase light cars (BAU for heavier cars, decrease mid-mass cars) Scenario: Increase light cars and prohibit heavy cars (decrease mid-mass cars). Scenario: Constant mass (prohibit heavy cars while maintaining baseline mass)  15% 10% 5% 0% -5% -10% -15% -20% -25% 1999  2000  2001  2002  2003  2004  2005  2006  83  Results for two-car relative risk (RR) simulation are summarized in Table 4.3. The fleet simulation for the “Increase lighter car” scenario estimates that RR increases 14%, 36%, and 69% relative to the baseline scenario for λ values of 2, 4, and 6 respectively. For the “Increase lighter cars and prohibit heavier cars” scenario, the estimated RR increases by 4%, 11%, and 22% for the same λ values. For the “Constant mass” scenario, the estimated RR decreases by 10%, 21%, and 33%. Table 4.3  Results of simulation of two-car collision relative risk fleet calculation for the year 2005. Relative risk is the driver fatality risk of lighter cars divided by driver fatality risk of heavier cars in the relationship RR = .  is calculated in the simulation assuming randomized collision events sampled from a mass distribution representative of the UK on-road car fleet.  is examined parametrically ranging from 2 to 6.  Baseline  Scenario: Increase  Scenario: Increase  lighter cars (BAU for  lighter cars and prohibit  heavier cars, decrease heavier cars (decrease mid-mass cars)  mid-mass cars).  Scenario: Constant mass (prohibit heavy cars while maintaining baseline mass)  λ=2  λ=4  λ=6  λ=2  λ=4  λ=6  λ=2  λ=4  λ=6  λ=2  λ=4 λ=6  Mean  1.5  2.7  5.6  1.8  3.7  9.6  1.6  3.0  6.9  1.4  2.1  3.8  Std dev  0.6  2.6  11.2  0.8  4.1  19.9  0.7  3.3  16.1  0.5  1.8  6.8  --  --  --  14%  36%  69%  4%  11%  22%  -10%  % of baseline  -21% -33%  mean  4.4  DISCUSSION  We examined reductions in vehicle mass that might best achieve the dual goals of lowering CO2 emissions and traffic casualties. The “first law” RR analysis for model year 1996-2005 cars in the UK resulted in a value of λ of 5.3. This implies, for example, that in multiple fatal collision events between 1,000 kg (135 gCO2/km) and 2,000 kg (200 gCO2/km) cars, the ratio of fatalities for drivers of lighter cars to drivers of heavier cars will be 39. The key finding is that drivers of linearly less CO2/km cars are 84  subject to non-linearly increased risk of fatality relative to drivers of higher CO2 /km cars in twocar collisions. An important secondary finding from our results in Table 4.1 is that there is a shift in the partition of risk in comparing lighter car/heavier car drivers from a 77%/26% split in fatalities, to a 12%/35% split for seriously injured. A λ value of 5.3 is higher than values found in the U.S. which ranges from 2.7-3.8 for primarily frontal collisions [12, 27]. This result is likely explained by the point of impact as illustrated in Figure 4.7. Compared to the complete UK Data Archive data set, our analysis data set had 9 percentage points fewer frontal impacts and 2 percentage points more driver-side impacts, but eliminating all side impact collisions was not feasible due to the aforementioned sample size limitation. We did attempt to fit our data to the same side-impact equation shown in Figure 4.7 (RR = A * μλ), but the results were invalid (Appendix Section B.4). Increased collision speed is yet another factor that increases λ[12]. The only variable we have to indicate collision speed is road speed limit, which shows 69% and 15% of collisions occurring on 60 mph and 70 mph roads, respectively, compared to values of 63% and 12% in the complete Data Archive. Driver age and gender are additional factors that may have increased the value of λ, since older people and females tend to be more vulnerable to injury [17]. In lighter cars, the average driver was 1.9 years older and the percentage male was 1.4% lower as compared to heavier cars, trends that would be expected to increase λ. Finally, lighter cars were 0.7 years older on average than heavier cars that could make a small influence to increase λ. The linear statistical model that relates RR to mass ratio was also found to be significant. If this is the best approximation, then it indicates that in multiple fatal collision events between 1,000 kg (135 gCO2/km) and 2,000 kg (200 gCO2/km) cars, the ratio of fatalities for drivers of lighter cars to drivers of heavier cars will be 15; this is much lower than the value of 39 produced by the “first law” model. We have no way of definitively determining which model is best approximation; however, based on the results found in the U.S. with larger data sets [12], we expect it is more likely that the power law relationship is a more accurate approximation. In either case, the fatality risk ratio, RR, is much greater than the mass ratio in two-car collisions.  85  Figure 4.7  Comparison of RR calculated in this study (from Figure 4.4), to previous results from U.S. data sets. Both the horizontal and vertical axes are logarithmic scales. The effect of point of impact strongly affects the RR. When the front of a striking car crashes into the driver side of the struck car, the RR is much larger. While the RR for front-to-front collisions has been observed to pass through the origin (i.e., RR ≈ 1 for mass ratio ≈ 1), the RR for purely front-to-driver side impacts has been observed to pass through 10 at the origin (i.e., RR ≈ 10 for mass ratio ≈ 1) [27]. Because of this relationship, the RR for side impact collisions is commonly fit to the equation  100  UK 2-car RR data +/- 1 std error UK RR fitted line, R2=0.71, p<0.0005 U.S. front-front crash modes (1991-95 data)  RR ≡ fatality risk light car drivers / fatality risk heavy car drivers  U.S. front-driver side crash modes (1991-95 data)  10  .  1 1  mass ratio: heavy car / light car  2  RR = A * μλ, where statistical models reveal A ≈ 10.  The analysis of two-car conditional KSI risk and CO2 emission rate shows that, for example, the average driver choosing a car that is 50 gCO2/km higher would be subject to a 1% point statistically significant reduction in KSI risk. Previous research has concluded that vehicle mass is an important determinant of both conditional KSI risk and fuel economy for these same car 86  models and risk estimates [5]. A limitation is that describing a car only by its make and model is imprecise, as there can be a large variation in important determinants of traffic safety and CO2 emission rate for a given make and model. For example, a BMW Series 3 model includes versions ranging in mass of 1,320-1,810 kg and emission rates of 135-310 gCO2/km[30]. We expect that the result of these blunt metrics is to reduce the magnitude of the effect of mass and its correlates on estimated risk. Our absolute risk analysis shows that rapid, yet realistic reductions in fleet mass over the years 2000-2005 in the UK could have reduced single-car, car-pedestrian, and two-car fatalities up to 10%, 18%, and 17%, respectively. Simultaneous reductions in fleet average CO2 emission rate of 12 gCO2/km and 19 gCO2/km could have been achieved based on our “Increase lighter cars” and “Increase lighter cars and prohibit heavier cars” scenarios, respectively. A key finding is that prohibiting entry of the heaviest cars into the fleet, vehicle mass roughly 1,600 kg and over in our scenario, was estimated to provide incremental reductions in three types of fatalities as well as tailpipe CO2. A strength of our analysis is that the results are not skewed by improving safety technology over time because our scenarios simulated changes in proportions of the actual UK fleet over the years 2000-2005, and estimated changes in fatality risk based on empirical rates for the same time period. A limiting assumption is that drivers shifting from heavier to lighter cars would not have altered their behavior, which conflicts with the “offsetting behavior” hypothesis [19], but there is no known data or methods to quantify such behavior shifts. The fleet simulation of randomized collisions and RR shows that it makes a difference in how fleets are transitioned to lighter, lower CO2 cars. From 2000-2005, mean “first law” RR in the UK would increase 14-69% (for λ range 2-6) under the “Increase lighter cars” scenario, but increase only 4-22% under the “Increase lighter cars and prohibit heavier cars” scenario. Therefore policy regimes that simultaneously increase shares of lighter cars and decrease heavier cars can be expected to both reduce CO2 and mitigate the equity of two-car fatality risk. The “Constant mass” scenario demonstrates that transitions towards a more homogeneous fleet, even without any reductions in average mass or CO2, improves the equity of fatality risk. These findings are consistent with other research that reducing the mass range of on-road car fleets is the most beneficial way to transition to a lower mass fleet[14, 16]. An examination of the assumption of randomized collisions is provided in Appendix Section B.4. It is important to highlight that vehicle size and mass correlate in our data [30] (n=3298) according to the following relationship: β  curb mass, kg = α * (wheelbase, m)  87  where: α = 109.9 +/- 8.2 (95% CI) and β = 2.62 +/- 0.07 (95% CI) Thus our results do not estimate the real effects of adopting technology that reduces mass without reducing size. With such technology it is plausible that KSI risk with decreasing mass as shown Figure 4.5 may not hold, that our estimates of fatality risk benefits would be larger, and that the value of fleet mean λ would be lower. The relationship between mass and size for current UK cars is statistically no different than 1980-1990 U.S. cars [12], which is clear evidence that potential benefits of mass reduction technology remain untapped. We conclude that reducing vehicle mass including phase in of an upper mass limit can yield public benefits simultaneously for traffic safety and climate mitigation, and that our UK results can be generalized to other developed countries.  4.5  ACKNOWLEDGEMENTS  We thank the UK Department for Transport and JATO Dynamics Ltd. for provision of high quality data sets. We thank the University of British Columbia Bridge Program, AUTO21: B06 BLC, and Carnegie Mellon University’s NSF-supported Center for Integrated Study of the Human Dimensions of Global Change (SBR-9521914), and the Center for Climate Decision Making (SES-0345798). We are also grateful for generous support from the Exxon-Mobil Education Foundation.  88  4.6 1.  REFERENCES Broughton, J., “Monitoring progress toward the 2010 casualty reduction target - 2005 data.” 2007, Transport Research Laboratory (www.trl.co.uk) for the UK Department for Transport (www.dft.gov.uk).  2.  Ameratunga, M.H. and R. Norton, “Road traffic injuries: confronting disparities to address a global health problem.” The Lancet, 2006. 367: p. 1533-1538.  3.  Ribeiro, K., S. Kobayashi, M. Beuthe, J. Gasca, D. Greene, D. Lee, Y. Muromachi, P. Newton, S. Plotkin, D. Sperling, R. Wit, and P. Zhou. “Transport and its infrastructure, in Climate Change 2007: Mitigation. Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change.” 2007, Cambridge University Press.  4.  Sperling, D. and D. Gordon, “Two Billion Cars: Driving Toward Sustainability.” 2009: Oxford University Press.  5.  Tolouei, R. and H. Titheridge, “Vehicle mass as a determinant of fuel consumption and secondary safety performance.” Transportation Research Part D, 2009. 14: p. 385-399.  6.  Maclean, H. and L. Lave, “Evaluating automobile fuel/propulsion system technologies.” Progress in Energy and Combustion Science, 2003. 29: p. 1–69.  7.  Elvik, R. and T. Vaa, “The Handbook of Road Safety Measures.” 2004: Elsevier.  8.  DeCicco, J., F. An, and M. Ross, “Technical options for improving the fuel economy of U.S. cars and light trucks by 2010–2015.” 2001, American Council for an Energy Efficient Economy (www.aceee.org).  9.  Redelmeir, D., R. Tibshirani, and L. Evans, “Traffic law enforcement and risk of death from motor vehicle crashes: case-crossover study.” The Lancet, 2003. 361: p. 21772182.  10.  Elvik, R., “Dimensions of road safety problems and their measurement.” Accident Analysis and Prevention, 2008. 40: p. 1200–1210.  11.  DFT, “UK Transport Statistics online database (www.dft.gov.uk/pgr/statistics)”. 2008, Department for Transport (UK).  12.  Evans, L., “Traffic Safety.” 2004: Science Serving Society.  89  13.  Van Auken, R. and J. Zellner, “A further assessment of the effects of vehicle weight and size parameters on fatality risk in model year 1985-98 passenger cars and 1985-97 light trucks.” 2003, Dynamic Research Inc.  14.  Buzeman-Jewkes, D., D. Viano, and P. Lövsund, “Occupant Risk, Partner Risk and Fatality Rate in Frontal Crashes: Estimated Effects of Changing Vehicle Fleet Mass in 15 Years.” Traffic Injury Prevention, 2000. 2(1): p. 1-10.  15.  Kahane, C., “Vehicle weight, fatality risk and crash compatibility of model year 1991-99 passenger cars and light trucks.” 2003, National Highway Traffic Safety Administration (U.S.).  16.  Broughton, J., “The likely effects of downsizing on casualties in car accidents.” 1999, Transport Research Laboratory (UK).  17.  Padmanaban, J. “Influences of vehicle size and mass and selected driver factors on odds of driver fatality.” in 47th Annual Proceedings of the Association for the Advancement of Automotive Medicine. 2003.  18.  Wenzel, T. and M. Ross, “The effects of vehicle model and driver behavior on risk.” Accident Analysis and Prevention, 2005. 37: p. 479-494.  19.  Yun, J., “Offsetting behavior effects of the corporate average fuel economy standards.” Economic Inquiry, 2002. 40(2): p. 260-270.  20.  Wood, D., Veyrat, N., Simms, C., and Glynn, C. “Limits for survivability in frontal collisions: Theory and real life data combined.” Accident Analysis and Prevention, 2007. 39: p. 679-687.  21.  MORI, “Assessing the impact of Graduated Vehicle Excise Duty: quantitative report.” 2003, UK Department for Transport (www.dft.gov.uk) and MORI (www.mori.co.uk).  22.  IR, “Report on the evaluation of the company car tax reform: stage two.” 2006, UK Inland Revenue (www.inlandrevenue.gov.uk).  23.  DFT, “Road Accident Data [computer files] 1994 SN3461, 1996 SN3627, 1997 SN3817, 1998 SN4027, 1999 SN4080, 2000 SN4356, 2001 SN4501, 2002 SN4588, 2003 SN4979, 2004 SN5042, 2005 SN5244, 2006 SN5493. 2007,” UK Data Archive (Colchester, Essex).  24.  DFT, “Make-Model crash database linkable to UK Data Archive data (electronic database).” 2006, UK Department for Transport.  90  25.  VCA, Online data of vehicle emissions for UK cars model years 2000-2006. 2007, Vehicle Certification Agency (www.vcacarfueldata.org.uk).  26.  Whatcar, “Online car specifications (www.whatcar.co.uk).” 2007.  27.  Joksch, H., D. Massie, and R. Pichler, “Vehicle aggressivity: fleet characterization using traffic collision data (DOT HS 808 679).” 1998, National Highway Traffic Safety Administration (NHTSA).  28.  DFT, “Cars: make and model: the risk of driver injury in Great Britain: 2000-2004.” 2006, Department for Transport (UK).  29.  Mengert, P. and S. Borener, “Overall fatality risk to the public at large related to national weight mix of passenger cars (DOT-TSC-HS070-PM-89-27).” 1989, National Highway Traffic Safety Administration (NHTSA).  30.  JATO. “JATO Dynamics new car database for Great Britain.” 2007 [cited June 3, 2007].  91  5  CONCLUSIONS  Here I summarize and discuss my thesis contributions to the design of climate mitigation policy for passenger cars. First I summarize the relationship between my three thesis articles, and then comment on the results of my research relative to current hypotheses in the study of public policy. I summarize key strengths and weaknesses of my research, and summarize what I believe is the significance and potential application of my research. Specific recommendations for future research are provided.  5.1  RELATIONSHIP BETWEEN PPOLICY INTEGRATED ASSESSMENT, AIR QUALITY, AND TRAFFIC SAFETY RESEARCH  Chapter 2 describes the rationale and basic framework for Integrated Assessment models of passenger car risks. Air quality and traffic safety are just two of the risks that are relevant in development of policy to reduce CO2 from passenger cars. The key issues and relationships in Integrated Assessment of passenger car CO2 policy (Chapter 2), air quality (Chapter 3), and traffic safety (Chapter 4) are summarized here.   The choice of decision criteria and risks included or excluded (see Table 2.1) influences the relative importance of air quality and traffic safety as compared to other risks. Risk rankings can change depending on the analyst’s choice of decision criterion and public risk metric. Moreover, available choices of decision criteria are different if the analyst is a researcher (more choice) as compared to a public policymaker (less choice). o  For the current fleet dominated by gasoline and diesel cars, it was found that air quality and traffic safety outcomes rank highly regardless of whether monetized social costs, public health, or public agency new policy priority is chosen as the decision criterion (see Table 2.1).  o  In terms of health risk metrics, if annual fatalities is chosen as the metric, then air quality is dominant; if YLL is chosen, then traffic safety is the highest priority (see Section 2.2.2).    Public risks, including air quality and traffic safety, are influenced by policy through pathways (see Figure 2.2) that include three principal economic actors (consumers,  92  manufacturers, and fuel suppliers) and four basic factors: VKT and patterns of use, ownership, vehicle technology, and fuel properties.   Four categories of policy that could be leveraged to reduce passenger car CO2 emissions were identified (see Table 2.2): multi-sector policies, fuel-oriented policies, vehicle technology policy, and VKT and usage policy. Considering current gasoline and diesel cars, specific policies in all four categories are identified as potentially influencing air quality. However only vehicle technology and ownership, and vehicle VKT and usage categories are identified as potentially influencing traffic safety.    Compared to today’s car fleets dominated by conventional gasoline and diesel cars, the direction and magnitude of change in air quality, traffic safety, and other risks can be expected to change substantially with future fleets that are projected to be more heterogeneous. Future fleets are projected to have substantial uptake of one or more of the following [1]: biofuelled cars, plug-in hybrids, all-electric, and hydrogen fuel cell cars. The relative importance of air quality and traffic safety risks in more heterogeneous fleets is uncertain and there could be substantial shifts. For example, all-electric car air quality impacts would shift to the electricity generation sector which, in general, can vary greatly depending on the design of the electricity system [2]. The air quality impacts of current generation biofuels is regionally dependent and unclear. This was illustrated by research on ethanol versus gasoline in the U.S., which concluded that health impacts from ambient ozone varied across the country where it was sometimes positive and sometimes negative [3].  Therefore, my key conclusion about how the three parts of my thesis research relate is that the air quality and traffic safety research articles are risk-specific health and CO2 emission case studies that are essential elements to building the capacity to develop Integrated Assessment models of public risks due to passenger cars.  93  5.2  RESEARCH RELATIONSHIP TO CURRENT WORKING HYPOTHESES IN THE FIELD OF STUDY AS REFLECTED IN THE LITERATURE  5.2.1  AIR QUALITY AND CO2 REDUCTION IN TRANSPORTATION  Approximately one decade ago, the working hypothesis in research on air quality and CO2 reduction was captured in the “ancillary benefits” literature. The prevailing research hypothesis was that reducing CO2 from transportation (and other sectors) would yield simultaneous benefits in air quality improvements, and no exceptions to this hypothesis were noted [4-6]. IPCC even sponsored a workshop to examine this hypothesis8 titled “Workshop on Assessing the Ancillary Benefits and Costs of Greenhouse Gas Mitigation Strategies” which, by its title, implied that only benefits could be realized. In its 2001 Working Group III report, IPPC described these ancillary benefits without mentioning the possibility of disbenefits [7]. Our research on diesel substitution for gasoline cars in the UK has helped inform the evolution of the “ancillary benefits” research towards a more careful approach which considers that both benefits and disbenefits might be possible. While our research has caught some media attention, implying that tradeoffs were surprising [8], research that identified potential tradeoffs between CO2 reduction and air quality predates the ancillary benefits literature, particularly for biomass utilization [9]. Currently the literature in this area appears to have a more cautious approach such as the aforementioned potential ambient ozone disbenefits from ethanol-powered cars [3]. In its 2007 Working Group III report IPCC specifically acknowledged the potential air quality disbenefits of diesel cars [10]. 5.2.2  TRAFFIC SAFETY AND CO2 REDUCTION IN TRANSPORTATION  Currently the prevailing research indicates there is a relatively weak link between traffic safety and CO2 from passenger cars on the basis that vehicle mass has uncertain impacts on traffic safety (see discussion in Section 1.4.2 and Section 3.1). However, in my view, this conclusion is strongly influenced by two factors: (1) a lack of clarity of what is meant by “traffic safety,” and (2) incomplete recognition of the changes in the mix of road users that researchers and policymakers alike are aiming to achieve.  8  See http://www.oecd.org/document/59/0,3343,en_2649_34361_1914811_1_1_1_1,00.html  94  “Traffic safety9” and related terms as implied by researchers is often oversimplified to a degree that traffic safety agencies cannot accept in practice. Some examples were previously described in Sections 1.4.2 and 3.1. Another example is a recently published article in a journal issue on “The interaction of environmental and traffic safety policies” [11], one article summarizes the research on “highway safety” and draws conclusions about “the safety issue” citing only findings on fatalities and largely based on two-car collisions [12]. Traffic safety encompasses risk of both fatality and injury for multiple vehicle types, road users, injury rates, and crash events (see Figures 4.1 and 4.2 [13]). Moreover, traffic safety risks can be quantified as risk ratios [14], conditional risks [15], or absolute risks (annual fatality count is a common measure of absolute risk) [16]. The UK DFT has specific traffic safety goals that include reductions in total KSI, total children KSI, the slight injury rate, and vulnerable road user casualty reductions such as bicyclists, motorcyclists, and pedestrians [13]. As such, concluding that traffic safety and passenger car CO2 are weakly linked largely based on inferences from two-car fatalities is not congruent with the needs of traffic safety authorities. Researchers and policymakers alike are aiming to dramatically change the mix of users that will share the roads with passenger cars. From a land use and public health perspective, researchers have identified steep growth in walking and bicycling (active transport) as a high priority [17]. Again using the DFT as an example, traffic safety authorities have recognized the need to improve road safety of multiple users that are expected to change over time due to CO2 reduction policies [18]. “A move towards a low carbon economy could also lead to a rise in the number of vulnerable road users. Potential increases in walking, cycling and use of powered two-wheelers mean that protection of these categories of road user should continue to be a priority in terms of vehicle engineering developments. In addition, an increase in car sharing could lead to an increase in rear seat passenger numbers, who do not currently enjoy equivalent safety protection to those in front seats.” In my view, there remains a gap between research on traffic safety and passenger cars CO2 reduction where research has yet to keep up with the complexity of current and future traffic safety needs that public agencies are mandated to address. The scope of our research (Chapter  9  The definition used in this thesis is “relating to human health impacts as a result of vehicles operating on public  roadways” as per the Glossary.  95  4) was designed to make progress towards closing this gap to help meet the needs of policymakers.  5.3  5.3.1  STRENGTHS AND WEAKNESSES OF THIS THESIS  STRENGTHS  In my view there are two key strengths of this thesis research: (1) an emphasis on actual changes as compared to completely hypothetical scenarios, and (2) methodological strength that raises the bar on the ability to replicate integrated policy research results. The air quality and traffic safety studies both were developed with the goal to emphasize actual changes using the UK as a case study, while minimizing the use of hypothetical changes. While the use of hypothetical changes or scenarios is essential in policy research [19], deliberate choices were made to emphasize actual changes. In the air quality study, changes in emissions, ambient concentrations, and health outcomes were based on actual growth of diesel cars through the year 2005, with forecasted growth through the year 2020. While this is a long forecast period, all air quality mortality estimates were based on industry diesel projections just until the year 2009 when Euro V emission standards were expected to be in full compliance; forecasts from 2010 to 2020 only factored into the CO2 estimates. The traffic safety research was also developed to increase the emphasis on actual changes. An important choice was in the absolute risk analysis to use empirical fatality rates (car-pedestrian, single-car, and two-car) over the years 1999 to 2005 for realistic scenarios in how the fleet mix could have been changed. While we could have chosen to examine major changes in vehicle composition such as larger on-road shares of lighter cars and complete elimination of heavier cars, this would have drifted further from a realistic scenario. Instead, we chose to examine the elimination of new heavier cars (roughly 1,600 kg) combined with realistic increased shares of lighter cars beginning in the year 2000. In practice, policymakers have no way to simply eliminating all heavier cars (or any cars for that matter) in a short time frame because cars remain on the road until scrapped through normal retirement or by accelerated scrappage policies [20]. Projecting traffic safety outcomes into the future due to changes in vehicle design is also subject to much legitimate criticism because safety technology has been shown to improve substantially over time periods of as little as 3 years [21], thus such projections are weaker as compared to our approach.  96  The ability to replicate policy research has long been identified as a challenge [19]. A methodological strength of this thesis research is that analysis assumptions, methods, and data are sufficiently documented such that independent researchers would have a fair chance at replicating and validating the results. For the air quality research, careful attention was made to document basic vehicle counts, vehicle types, emission factors, ambient concentration estimates, and health risk coefficients. Similarly for the traffic safety analysis, careful attention was paid to document data sources and analytical methods. Although replication would be more challenging for the traffic safety based on the large database sizes (see Table B2), data sorting and reduction procedures were documented with sufficient detail to enable replication. It is my experience that other similar research is not as thoroughly documented to enable replication; one example is the aforementioned integrated study that examined electricity system design including CO2 and air quality [2]. 5.3.2  WEAKNESSES  Chapters 2, 3, and 4 each have a number of weaknesses and limitations that are discussed in the respective chapters. I will highlight a primary weakness of each, and suggest how these might be overcome in the future. A principal weakness of the air quality study in Chapter 3 was the simplistic treatment of changes in CO2 emissions resulting from diesel substitution10. Annual changes in CO2 were estimated assuming average CO2/km emission rates based on certified emission factors [22], multiplied by assumed average annual distance traveled. This method overlooked factors that were known to be relevant, particularly VKT rebound effects [23]. In fact, since publication of Chapter 3, research on diesel substitution in the UK has concluded that CO2 savings were likely small due to the combined effects of consumer choice of larger cars, VKT rebound effects, and driver behavior (mainly speed) [24]. Even in the absence of specific research to make such estimates, this limitation could have been better overcome by employing quantitative techniques [19] to estimate a range of changes in annual CO2. This could have been done by using general estimates of rebound effects [25] and a deterministic or probabilistic sensitivity analysis by making reasonable assumptions about upsizing and driver behavior.  10  The two anonymous peer reviewers for the Chapter 2 articles both emphasized the weakness of the global  warming benefits of diesel cars. One emphasized the annual travel distance issues, while the other emphasized black carbon effects.  97  A principal weakness of the traffic safety study in Chapter 4 is the absence of vehicle mass in the UK Data Archive data sets. Because curb mass (and its correlates) was the primary intermediate variable to examine changes in CO2/km and various traffic safety risks, this represented a substantial limitation. We were able to make adequate estimates of curb mass in two ways. Using the JATO data we found that engine size was a satisfactory proxy for curb mass (it is noted that our statistical model and coefficients are almost identical to those published by others – see Figure 1 of [15]). For the “first law” RR analysis, we were able to find curb mass for individual cars based on make, model, engine size, and body type using independent online databases [26-28]. Given the continued relevance of curb mass as a determinant of fuel consumption, CO2 emissions, and traffic safety, it would be useful if curb mass data were recorded for all vehicles involved in police-reported collisions in the UK; this is actually the case in the U.S. where the FARS database includes curb mass [14]. The Integrated Assessment methodology described in Chapter 2 is inherently limited by focusing primarily on Integrated Assessment of multiple risks for passenger cars. A truly comprehensive Integrated Assessment would explore the best strategy for meeting transportation needs within the context of all relevant local, social, and physical conditions [29]. Additional domains that are relevant include organizational and institutional domains[30], interactions with other energyintensive sectors [31], integration of transportation and land use planning[32], and integration of passenger car travel with other travel modes [17]. In the larger picture, the integration of climate change with other drivers of global change is relevant [33]. Another limitation of this study is that it focuses on risks and does not examine the positive attributes of consumer choice [34]. As an example, the travel range for a single refueling of a gasoline-powered car is much greater than an all-electric car, which is a positive attribute strongly valued by consumers. Thus the relative benefits of fuels and technologies to consumers are excluded [35].  5.4  SIGNIFICANCE AND POTENTIAL APPLICATIONS OF THIS THESIS  The significance of this thesis is to demonstrate the potential public benefits (or, from another view, avoided consequences) of performing Integrated Assessment of multiple risks in the development of climate mitigation policies for passenger cars. The significance stems from the fact that public policymakers globally are currently tasked to develop these policies [1, 10, 36, 37]. Using the UK as a case study enhances the significance of this research because CO2 reduction policies for passenger cars have been implemented for many years. The first such policy that affected the UK was the manufacturers’ voluntary agreement to reduce fleet average  98  CO2/km for new sales in Europe from 185 gCO2/km in 1995 to 140 gCO2/km in 2008 [38], followed by the VED tax regime in 2001 [39], and the company car tax regime in 2002 [40]. The potential application of this thesis research is in the analysis of real world policies. In the UK, government agencies have been tasked analyze the effectiveness of its policies. Examples include the DFT’s analyses of the VED tax regime done in 2003 which included quantitative analysis of CO2 and air pollution emissions as factors in consumer decisions [39]. The UK Customs and Revenue department also performed similar analyses of the company car tax regime [40]. A local example is the London Transport Authority’s assessment of its transportation policy that focused on congestion control and the side effects on road safety (specifically identifying motorcycles, pedestrians, and bicyclists), CO2, air quality, and noise [41]. An international example is provided by analyses of transportation CO2 policies by the European Commission [42]. Non-government organizations could also utilize this research in its independent assessments of government policies. For example, the European Federation for Transport and Environment regularly assesses the effectiveness of government policies in Europe. This organization assessed proposed fuel economy standards with respect to traffic safety [43], and also recently released report on the effectiveness of the manufacturers’ CO2 agreement [44]. Lastly, researchers can build upon this thesis research to advance the analysis of real world policies. Integrated Assessment modeling continues to be a research priority [29], and the recent analysis of CO2 emissions from diesel cars is an example of the insights that can be revealed from such research [24].  5.5  5.5.1  RECOMMENDATIONS FOR FUTURE RESEARCH  MITIGATING AIR QUALITY AND HEALTH IMPACTS FROM PRE-EURO V DIESEL CARS IN EUROPE  The steep growth of pre-Euro V diesel cars (i.e. where particle filters were not required, and NOx limits differed greatly) experienced in the UK has occurred elsewhere in Western Europe where diesel registrations increased from 14% of all registrations in 1990 to 46% in 2009 [45]. It is likely that disbenefits, such as we estimated for the UK, are occurring in many other European cities. Evidence from the UK and elsewhere shows that diesel cars tend to remain in use longer than their gasoline counterparts [23], and that diesels also fail tailpipe emission tests at higher rates (see Table A5). For these reasons, it is likely that the trajectory of public health impacts from ambient particles (both primary and secondary particles) from pre-Euro V diesels will continue to persist, if not grow. Research into technology and policies to induce retrofit of particle filter technology and/or early retirement of higher polluting diesels is warranted. 99  5.5.2  TECHNOLOGY AND POLICY ASSESSMENT OF IMPOSING A 1,600 KG CAP ON PASSENGER CAR MASS  We have contributed to the body of evidence that shows reducing vehicle mass can be accomplished while meeting traffic safety goals, but that it matters how on-road fleets are transitioned to lower vehicle mass. Specifically, Chapter 4 indicates that reducing the mass of the largest vehicles and narrowing the range and variation in vehicle mass of the on-road fleet is the most effective method to improve both absolute and relative traffic safety risks, while moving toward a less carbon-intensive fleet. Capping the mass of any new car is a method of achieving this, and current engineering research indicates that use of lightweight materials could enable retention of large size vehicles which are highly sought by consumers while keeping mass within a 1,600 kg limit [46-48]. Moreover, mass is not an attribute included in lists of consumer preferences [34, 49], and the impact of adopting lightweight technologies on vehicle price are estimated to be less than the current premium paid for hybrids [46-48]. Further research is needed to more definitively quantify changes in public health, lifecycle greenhouse gas emissions, and policy design to implement an upper limit on the mass of passenger cars. 5.5.3  TRAFFIC SAFETY RISKS DUE TO VARIATION IN VEHICLE MASS AND SIZE: DO CONSUMERS UNDERSTAND AND INTERNALIZE THE RELATIVE RISKS OF PURCHASING SMALLER, LIGHTER VEHICLES?  Externality cost research commonly assumes that consumers internalize differential traffic safety risks when purchasing smaller, lighter passenger cars [50]. This assumption contributes to economic externality research findings that questions public welfare loss associated with car fleets with bimodal distributions in terms of mass and it correlates [51]. I could not locate any empirical validation of the assumption that consumers internalize the risk differential, and it appears to be rooted in expert opinion [52]. However, even the collision safety rating system used for informing consumers provides rankings within vehicle classes rather than across all vehicles. To quote the European NCAP [53]: “Euro NCAP’s frontal impact test simulates a car crashing into another of similar mass and structure. In real life, when two cars collide the vehicle with the higher mass has an advantage over the lighter one. Generally speaking, vehicles with higher structures tend to fare better in accidents than those with lower structures. Therefore, ratings are comparable only between cars of similar mass and with broadly similar structures. Euro NCAP groups cars into the following structural categories: passenger car, MPV, off-roader, roadster and  100  pickup. Within each of those categories, cars which are within 150 kg of one another are considered comparable.” Research is needed to examine this assumption that consumers are aware and have internalized (or voluntarily accepted) the differential risk, and explore alternatives for addressing the gap between subjective risk perceptions and objective risk estimates. 5.5.4  PUBLIC HEALTH RISKS FROM ENVIRONMENTAL NOISE AND HYBRID TECHNOLOGY  Hybrid cars, whether grid independent or plug-in, produce substantially less environmental noise than conventional, internal combustion technology with approximately 2.5 dBA less for 2006 UK models as shown in Figure 2.5. Recent estimates of environmental noise indicate it is a growing public health risk of roughly similar order to traffic safety and air quality [54, 55]. While environmental noise from traffic is identified as a risk factor for cardiovascular disease [56], questions remain as to whether noise effects have been separated from other factors such as air pollution [57]. However, in terms of traffic safety, at least during the period of transition to a quieter fleet, pedestrians and bicyclists might be subjected to increased risk of traffic collisions with noiseless motor vehicles. These potential risks are also in need of quantification as well, and could inform the design of cars to compensate for noiseless operation such as incorporating awareness-level sound emissions into the vehicle design [58]. Subject to rigorous consideration of confounding factors and technology adaptations, the potential public health benefit through noise reduction from uptake of hybrid cars is recommended for further research. 5.5.5  DEVELOPMENT OF QUANTITATIVE, INTEGRATED POLICY MODELS FOR PASSENGER CAR CHOICE AND RISKS  Policies are in place in industrialized countries to induce a transition from current fleets dominated by conventional gasoline and diesel passenger cars, to more heterogeneous fleets comprised of alternative fuels (e.g. biofuels) and technologies (e.g., plug-in hybrids) [1]. With this transition, the set of relevant risks, system boundaries, and relationships between policies and risks grow in size and complexity necessitating expansion of policy modeling capability to estimate risks. This research recommendation is essentially to move forward with actual, working quantitative models using the framework introduced in Chapter 2.  101  5.6 1.  REFERENCES Fulton, L., P. Cazzola, and F. Cuenot, "IEA Mobility Model (MoMo) and its use in the ETP 2008". Energy Policy, 2009. 37: p. 3758-3768.  2.  Bergerson, J. and L. Lave, "Should We Transport Coal, Gas, or Electricity: Cost, Efficiency, and Environmental Implications". Environmental Science and Technology, 2006. 39(16): p. 5905-5910.  3.  Jacobson, M., "Effects of ethanol (E85) versus gasoline vehicles on cancer and mortality in the United States". Environmental Science and Technology, 2007. 41: p. 4150-4157.  4.  Cifuentes, L., V. Borja-Aburto, N. Gouveia, G. Thurston, and D. Davis, "Assessing the Health Benefits of Urban Air Pollution Reductions Associated with Climate Change Mitigation (2000-2020): Santiago, Sao Paulo, Mexico City, and New York City". Environmental Health Perspectives, 2001. 109(Supplement 3): p. 419-425.  5.  Cifuentes, L., V. Borja-Aburto, N. Gouveia, G. Thurston, and D. Davis, "Hidden Health Benefits of Greenhouse Gas Mitigation". Science, 2001. 293: p. 1257-1259.  6.  Davis, D., T. Kjellstrom, R. Sloof, A. McGartland, W. Atkinson, W. Barbour, W. Hohenstein, P. Nagelhout, T. Woodruff, F. Divita, J. Wilson, L. Deck, and J. Schwartz, "Short-term Improvements in Public Health from Global-Climate policies on Fossil Fuel Combustion: An Interim Report". The Lancet, 1997. 350: p. 1341-1349.  7.  IPCC, "Climate Change 2001: Mitigation". 2001: Intergovernmental Panel on Climate Change (www.ipcc.ch).  8.  Anonymous, "Carbon tax turns into a health risk". New Scientist, 2007. 17(26).  9.  Borjesson, P. and L. Gustavsson, "Regional production and utilization of biomass in Sweden". Energy, 1996. 21(9): p. 747-764.  10.  Ribeiro, K., S. Kobayashi, M. Beuthe, J. Gasca, D. Greene, D. Lee, Y. Muromachi, P. Newton, S. Plotkin, D. Sperling, R. Wit, and P. Zhou, Transport and its infrastructure, in Climate Change 2007: Mitigation. Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. 2007, Cambridge University Press.  11.  Noland, R., "The interaction of environmental and traffic safety policies". Transportation Research Part D, 2009. 14: p. 373-374.  102  12.  Greene, D., "Feebates, footprints and highway safety". Transportation Research Part D: Transport and Environment, 2009. 14: p. 375-384.  13.  DFT, "UK Transport Statistics online database (www.dft.gov.uk/pgr/statistics)". 2008, UK Department for Transport.  14.  Evans, L., "Traffic Safety". 2004: Science Serving Society.  15.  Tolouei, R. and H. Titheridge, "Vehicle mass as a determinant of fuel consumption and secondary safety performance". Transportation Research Part D: Transport and Environment, 2009. 14: p. 385-399.  16.  van Auken, R. and J. Zellner, "A further assessment of the effects of vehicle weight and size parameters on fatality risk in model year 1985-98 passenger cars and 1985-97 light trucks". 2003, Dynamic Research Inc.  17.  Woodcock, J., D. Bannister, P. Edwards, A. Prentice, and I. Roberts, "Energy and Health Series: Energy and Transport". The Lancet, 2007. 370: p. 1078-1088.  18.  DFT, "A Safer Way: Consultation on Making Britain's Roads the Safest in the World". 2009, UK Department for Transport.  19.  Morgan, M.G. and M. Henrion, "Uncertainty: a guide to dealing with uncertainty in quantitative risk and policy analysis". 1992: Cambridge University Press.  20.  Alberini, A., W. Harrington, and V. McConnell, "Fleet turnover and old car scrap policies (RFF DP 98-23)". 1998, Resources for the Future.  21.  Broughton, J., "Casualty rates by type of car (version 3)". 2007, TRL Limited.  22.  VCA, "Online data of vehicle emissions for UK cars model years 2000-2006". 2007, Vehicle Certification Agency (www.vcacarfueldata.org.uk).  23.  Schipper, L., C. Marie-Lilliu, and L. Fulton, "Diesels in Europe: Analysis of Characteristics, Usage Patterns, Energy Savings, and CO2 Emission Implications". Journal of Transport Economics and Policy, 2002. 36(2): p. 305-340.  24.  Bonilla, D., "Fuel demand on UK roads and dieselisation of fuel economy". Energy Policy, 2009. 37: p. 3769-3778.  25.  Greening, L., D. Greene, and C. Difiglio, "Energy efficiency and consumption - the rebound effect - a survey". Energy Policy, 2000. 28: p. 389-401.  26.  Globalcar, "Online car specifications (www.globalcar.com). 2007.  27.  Street-car, "Online car specifications (www.street-car.net). 2007. 103  28.  Whatcar, "Online car specifications (www.whatcar.co.uk). 2007.  29.  Rotmans, J., "Tools for Integrated Sustainability Assessment: a two-track approach". Integrated Assessment, 2006. 6(4): p. 35-37.  30.  Hull, A., "Policy integration: what will it take to achieve more sustainable transport solutions in cities?" Journal of Transport Economics and Policy, 2007. 15: p. 94-103.  31.  Lund, H. and E. Munster, "Integrated transportation and energy sector CO2 emission control strategies". Transport Policy, 2006. 13: p. 426–433.  32.  Chatterjee, K. and A. Gordon, "Planning for an unpredictable future: Transport in Great Britain in 2030". Transport Policy, 2006. 13: p. 254–264.  33.  Dowlatabadi, H., "On integration of policies for climate and global change". Mitigation and Adaptation Strategies for Global Change, 2007. 12: p. 651–663.  34.  Maclean, H. and L. Lave, "Evaluating automobile fuel/propulsion system technologies". Progress in Energy and Combustion Science, 2003. 29: p. 1–69.  35.  Keefe, R., J. Griffin, and J. Graham, "The Benefits and Costs of New Fuels and Engines for Light-Duty Vehicles in the United States". Risk Analysis, 2008. 28(5): p. 1141-1154.  36.  Kopp, R., Transport policies to reduce CO2 emissions from the light-duty vehicle fleet (Issue Brief 12), in Assessing U.S. Climate Policy Options. 2007, Resources for the Future: Washington, DC.  37.  Sperling, D. and D. Gordon, "Two Billion Cars: Driving Toward Sustainability". 2009: Oxford University Press.  38.  Fontaras, G. and Z. Samaras, "A quantitative analysis of the European Automakers’ voluntary commitment to reduce CO2 emissions from new passenger cars based on independent experimental data". Energy Policy, 2007. 35: p. 2239-2248.  39.  MORI, "Assessing the impact of Graduated Vehicle Excise Duty: quantitative report". 2003, UK Department for Transport (www.dft.gov.uk) and MORI (www.mori.co.uk).  40.  IR, "Report on the evaluation of the company car tax reform: stage two". 2006, UK Inland Revenue (www.inlandrevenue.gov.uk).  41.  GLA, "The Mayor’s Transport Strategy Revision". 2004, Greater London Authority.  42.  EC, "Results of the review of the Community Strategy to reduce CO2 emissions from passenger cars and light commercial vehicles". 2007, European Commission (http://europa.eu/). 104  43.  TandE, "Danger ahead: Why weight-based CO2 standards will make Europe's car fleet dirtier and less safe". 2007, European Federation for Transport and Environment: Brussels, Belgium.  44.  TandE, "Reducing CO2 Emissions from New Cars: A Study of Major Car Manufacturers' Progress in 2008". 2009, Transport and Environment.  45.  ACEA, "New Passenger Car Registrations - Breakdown by Specification". 2009, European Automobile Manufacturers Association (www.acea.be).  46.  An, F., D. Friedman, and M. Ross. "Near-term fuel economy potential for light duty trucks (2002-01-1900)". in 2002 Future Car Congress. 2002. Arlington, Virginia: Society of Automotive Engineers (www.sae.org).  47.  DeCicco, J., F. An, and M. Ross, "Technical options for improving the fuel economy of U.S. cars and light trucks by 2010–2015". 2001, American Council for an Energy Efficient Economy www.aceee.org.  48.  Moon, P., A. Burnham, and M. Wang. "Vehicle cycle energy and emission effects of conventional and advanced vehicles (2006-01-0375)". in 2006 World Congress. 2006. Detroit, Michigan: Society of Automotive Engineers (www.sae.org).  49.  Lane, B., "Car buyer research report: consumer attitudes to low carbon and fuel efficient passenger cars." 2005, Low Carbon Vehicle Partnership (UK).  50.  Lindberg, G., "Measuring the marginal social cost of transport: accidents". Research in Transportation Economics, 2005. 14: p. 155-183.  51.  Parry, I., M. Walls, and W. Harrington, "Automobile externalities and policies". Journal of Economic Literature, 2007. XLV: p. 373-399.  52.  Lindberg, G., "Calculating Transport Accident Costs - Final Report of the European Commission Expert Advisors to the High Level Group on Infrastructure Charging (Working Group 3)". 1999, European Commission.  53.  NCAP. “European New Car Assessment Programme - Comparable Cars.” 2009 [cited September 25, 2009]; Available from: http://www.euroncap.com/Content-WebPage/0f3bec79-828b-4e0c-8030-9fa8314ff342/comparable-cars.aspx.  54.  den Boer, L. and A. Schroten, "Traffic noise reduction in Europe: Health effects, social costs and technical and policy options to reduce road and rail traffic noise". 2007, CE Delft (www.ce.nl) commissioned by TandE (http://www.transportenvironment.org).  105  55.  WHO-Europe, "Burden of disease from environmental noise (http://www.euro.who.int/Noise/activities/20021203_3 downloaded April 3, 2008)". 2008, World Health Orgainization.  56.  Babisch, W., "Transportation Noise and Cardiovascular Risk: Review and Synthesis of Epidemiological Studies; Dose-effect Curve and Risk Estimation". 2006, Federal Environmental Agency: Berlin, Germany.  57.  Beelen, R., G. Hoek, D. Houthuijs, P. van den Brandt, R. Goldbohm, P. Fischer, L. Schouten, B. Armstrong, and B. Brunekreef, "The joint association of air pollution and noise from road traffic with cardiovascular mortality in a cohort study". Journal of Occupational and Environmental Medicine, 2009. 66: p. 243-250.  58.  Shim, S., A. Miller, and A. Marsh, "The Emergence of New types of Powertrain and the Impact on the Insurance Industry". 2009, KART/KIDI, Thatcham (Motor Insurance Repair Research Centre).  106  APPENDIX A - SUPPORTING INFORMATION FOR AIR-QUALITY IMPACTS OF CLIMATE MITIGATION: UK POLICY AND PASSENGER VEHICLE CHOICE  107  This Appendix contains Supporting Information published as part of Chapter 2 in Environmental Science and Technology.  A.1  INTRODUCTION  Table A1 illustrates the cost of ownership of matched pairs of petrol and diesel cars for model year 2005. Diesels have a higher purchase price but generally lower operating costs.  Table A1  Comparison of the cost of ownership for matched pairs of petrol and diesel 2005 car models (all models shown meet Euro IV Emission Standards) 1st Year 1st Year Company Company Company Car Tax Car CO2 Car Tax 3 22% (40% Tax % Bracket) Bracket)  gCO2 per km  VED CO2 Band  VED Cost  Fuel Liter per 100 km  Purchase 1 Price  Fuel Cost 15,800 km  Insur2 ance  2.0 Duratec HE Ford Mondeo Saloon 4-door LX petrol manual  187  F  £165  7.5  £16,067  £955  £910  23%  £821  £1,493  2.0 Duratorq Ford Mondeo TDCi (115PS) 4-door LX diesel manual  159  D  £135  5.8  £16,594  £750  £910  17%  £626  £1,138  -£30  23%  £527  -£204  £0  -6%  -£196  -£356  Manufacturer and Model  Engine and Transmission  diesel – petrol difference  -15%  Ford Focus 4-door Ghia petrol  2.0i Duratec 16V Manual  173  E  £150  6.9  £15,803  £883  £781  20%  £702  £1,276  Ford Focus 4-door Ghia diesel  2.0i 16V TDCi (136PS) Manual  145  C  £115  5.2  £17,082  £684  £843  15%  £568  £1,032  -£35  24%  £1,280  -£199  £63  -5%  -£134  -£244  diesel – petrol difference  -16%  1.4i SXi 5 Door Vauxhaul Corsa 16V petrol Manual  142  C  £115  5.7  £13,266  £725  £642  15%  £442  £803  1.3CDTi SXi 5 Vauxhaul Corsa Door 16V diesel Manual  115  B  £85  4.2  £13,699  £543  £597  15%  £455  £827  -£30  -27%  £433  -£182  -£45  0%  £13  £24  diesel – petrol difference  -19%  1  Includes 17% value added tax (VAT), £25 for plates, and £38 for first registration cost. Per person annual insurance cost for London, 2 drivers aged 40, no accident history, £250 deductible, private use, nominal theft protection. 3 Beginning January 1, 2006 Euro IV diesel cars incurred a 3% surcharge, which is not included here. Note that lower medium segment and larger diesels, even with the 3% surcharge, still pay less company car tax. The company car benefit-in-kind tax bottoms out at 15% regardless of how low the CO2/km, and tops out at 35% regardless of how high the CO2/km or the diesel surcharge. 2  108  A.2  METHODS  Here we present additional tables of estimated results as cited in the main article and some additional commentary on the study methods. Estimates for scrapped vehicles were made using the deregistration model shown in Figure A1 [1].  Figure A1  Model for estimating annual number of scrapped vehicles, based on UK deregistration statistics. The annual rate of vehicles scrapped peaks at 10.7%, 14 years after the year the vehicle was initially registered. The area under the curve through 20 years is 85.7%.  deregistered vehicles as a % of total registrations  12% 11% 10% 9% 8% 7% 6% 5% 4% 3% 2% 1% 0% 0  5  10  15  20  25  Age (years)  Reductions in CO2 due to additional diesels are estimated using UK fleet wide average differences between petrol and diesel passenger vehicles shown in Figure A2. Actual differences in gCO2/km are applied through the year 2005 based on industry data which use conventional greenhouse gases (CO2, CH4, N2O) [2]. Global warming effects of black carbon are under review in the UK, but are not included in the CO2-emission factors at this time [3]. For the years 2006-2020, we estimated the differential by assuming a declining difference in the CO2-emission factor by 5% per year until 2015, and thereafter a constant difference of 109  4.4 gCO2/km as shown in Figure A2. The basis for this assumption for the years 2006-2020 is:   The CO2 advantage of diesel passenger vehicles increased rapidly through 2001 due to technological improvements achieved by auto manufacturers [4].    In recent years more gains have been made in petrol vehicle design, and consumer preferences for larger and heavier diesel cars have resulted in a declining CO2 advantage for diesels based on fleet averages [2].    The tightening of emission controls on diesel vehicles to harmonize with petrol vehicles is another factor that will likely close the gap between diesel and petrol vehicles, which is reflected in our projections. The addition of particle filter technology is expected to result in a fuel economy4 penalty of 3-5% [5, 6].  4  In this study the term “fuel economy” refers only to liters per kilometer, not energy efficiency. 110  Fleet average difference in CO2 emission factors (gCO2/km) between petrol and diesel passenger vehicles in the UK from 1997-2020. 18 Petrol minus Diesel unit CO2 emissions (grams per kilometre)  Figure A2  16  actual g/km CO2 projected g/km CO2  14 12 10 8 6 4 2 0 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 Year  111  Fuel savings due to additional diesels were estimated as equivalent barrels (bbl) of crude oil saved. This was calculated directly from the fleet average differential CO2/km, then using emission factors (2,504 gCO2/L petrol and 2,763 gCO2/L diesel5), with a 25% additional energy savings expected in the upstream refining of diesel relative to petrol. Barrels of oil saved per unit quantity of additional diesels as shown in Figure 2.4 are based on the average over the vehicle lifetime (approximately 13 years and 205,000 km). Total barrels of oil saved over the 2001-2020 study period as shown in Figure 2.4 are not necessarily the unit fuel savings multiplied by the number of additional diesels, because the 20-year analysis period does not include the lifetime of every car, particularly post-Euro IV models (e.g., consider that some post-Euro IV cars enter the vehicle stock in our model in 2019, only 1 year before the end of the analysis period). Emissions estimates for common air contaminants are made by applying standard vehicle emission factors as derived by the UK National Atmospheric Emission Inventory (NAEI), which are shown in Table A2 along with the applicable emission limit. Euro III emission factors applied from 2001-2005 and Euro IV from 2006-2008. Beginning in 2009, “post-Euro IV” emissions of PM10 are assumed to be equal for diesel and petrol based on the European Commissions proposal to harmonize emission standards. Assumptions were made regarding early adoption of Euro IV vehicles registered in the UK prior to 2006 at 33% in 2004 and 66% in 2005. We assumed a proportion of kilometers driven on road types typical for the average of all passenger cars: 41% rural, 41% urban, and 18% motorway [7]. The calculation of intake fraction described in the Chapter 2 is based on a population of 60 million, an average breathing rate of 12.2 m3 per person-day, and the ratio of 0.0737 g/m3 change in annual mean ambient PM10 concentration per 1 kilo-tonne change in annual PM10 emissions [8, 9].  5  These emission factors are CO2 equivalent that include CO2, CH4, and N2O tailpipe emissions. 112  Table A2  European Union “Euro” emission limits [10] and weighted average emission factors [11] for passenger vehicles in grams per kilometer (g/km)6.  Limit or CO Standard Fuel Factor g/km (Date)  Euro III (2001)  Euro IV (2006)  Petrol Diesel Petrol Diesel  Proposed Petrol (2009) Diesel  6  HC7 g/km  NOx8 g/km  HC+ NOx g/km  PM109 g/km  1,3 Benzene Butadiene g/km g/km  limit  2.30  0.20  0.15  no limit  no limit  no limit  no limit  factor limit  0.730 0.64  0.205 no limit  0.246 0.50  0.451 0.56  0.002 0.05  0.00330 no limit  0.00026 no limit  factor  0.085  0.029  0.582  0.611  0.044  0.00073  0.00036  limit  1.00  0.10  0.08  no limit  no limit  no limit  no limit  factor  0.487  0.196  0.128  0.324  0.002  0.00291  0.00020  limit  0.50  no limit  0.25  0.30  0.025  no limit  no limit  factor limit  0.085 1.0  0.027 0.075  0.291 0.06  0.318 none  0.022 0.00510  0.00067 no limit  0.00033 no limit  limit  0.5  no limit  0.20  0.25  0.005  no limit  no limit  CO= carbon monoxide, HC= hydrocarbons, NOx= nitrogen oxides, PM10= 10 micron particulate  matter. 7  Petrol HC emission factors include evaporative emissions applying NAEI equations for diurnal,  hot soak, and running loss emissions. Equation inputs: diurnal rise = 9oC, maximum daily average = 15oC, annual mean = 11oC, and Reid Vapour Pressure = 90 kPa. Carbon canister control assumed for all vehicles. 8  NOx and PM10 emission factors include cold start emissions (grams per trip) averaged into the  emission factor based on UK average 8.4 kilometer per trip, and assuming half of all starts are cold starts. 9  Emission standards are actually stated in terms of PM10. However as described in the Methods  section of the paper, vehicles emit approximately 100% PM10. 10  Applicable only to petrol vehicles with lean-burn, direct-ignition engines. 113  We quantified morbidity by estimating respiratory and cardiovascular hospitalizations using rate coefficients adopted by COMEAP. Changes in both are estimated at 0.8% per 10 g/m3 change in PM10 applied to the entire UK population. We use a baseline rate for respiratory hospitalizations of 925 per 100,000 (ICD10 codes J00-J99, 2003), and 632 per 100,000 (I20I52, 2003) for cardiovascular [12]. Baseline mortality estimates were the actual 2003 rates (www.gad.gov.uk). The baseline all cause mortality for all ages was 995 per 100,000 (used for the low-mortality estimate). For the central- and high-mortality estimates, we used a baseline mortality of 1,424 per 100,000 for ages 30 and older. The UK population was assumed to be a constant 60 million in our analysis, and a constant 37.7 million for those aged 30 and older. A more complete mortality and morbidity analysis would include additional outcomes such as acute effects, exacerbation of asthma, and chronic exposure-mortality for people under the age of 30 year [13]. However, estimating this additional suite of health outcomes would not, in our view, contribute to the study findings and could distract attention from other parts of the analysis.  A.3  RESULTS  Here we present additional tables to support material included in the Results section of Chapter 2. During a given year, the number of additional diesels is the sum of new registrations, plus those vehicles remaining on the road from previous years. Annual time series for new vehicle registrations (total vehicles, number of diesel vehicles, and number of petrol vehicles) through 2005 are based on industry statistics [2]. Projections from 2006-2007 are based on industry forecasts. We estimate total new passenger vehicle registrations from 2008-2020 based on a 1% (compounded) annual growth rate. We estimate diesel market share for new vehicles to increase 2% points annually from 2008-2010, and with no further growth from 2011-2020. The rationale for no further growth at that time is based on industry and government expectations [14], which also seems feasible because compliance with proposed Euro V emission standards to install particulate traps are expected to increase the cost of diesels substantially [10]. 114  Table A3 provides the estimated changes in common air contaminants over the 20-year study period. Table A4 provides the estimated changes in health outcomes.  Table A3  Total estimated changes in emissions due to additional diesels 2001-2020. Intervals are defined by dates when new EU emission standards apply as shown in Figure 2.3. Diesels emit higher amounts of PM10, NOx, and 1,3 butadiene but lower amounts of CO, HC, benzene, and CO2. PM10 (kt)  NOx (kt)  CO (kt)  HC (kt)  CO2 (Mt)  Benzene (kt)  1,3 butadiene (kt)  Zone A : Euro III  5.7  45  -87  -24  -1.6  -0.35  0.013  Zone B : Euro IV  6.0  47  -117  -49  -2.1  -0.65  0.038  0  0  0  0  -3.3  0  0  Total 2001-2020  11.7  93  -204  -73  -7.0  -1.00  0.051  Annual average  0.58  4.6  -10  -3.7  -0.35  -0.05  0.003  Zone C: post Euro-IV  115  Table A4  Summary morbidity and mortality results. Hospitalizations Respiratory  Zone A: Euro III Zone B: Euro IV Zone C: post Euro-IV Total 20012020 Annual average  A.4  Mortality  Central per Central per Cardio- Central11 106 Diesel Mt CO2 total vascular Vehicles Reduced  Low12 Total  High13 Total  190  130  910  1,320  570  190  2,950  190  130  940  590  460  200  3,060  0  0  0  0  0  0  0  380  260  1,850  200  270  390  6,010  19  13  90  200  260  20  300  DISCUSSION  Here we discuss the uncertainties and limitations associated with our methods and assumptions for each step of the analysis, and how these uncertainties and limitations are likely to have over- or underestimated emissions, fuel savings, exposure, and health outcome estimates. The issues discussed here are: 1) number and emission class of additional diesels, 2) annual travel distance, 3) spatial distribution of vehicles, 4) PM10 emissions and ambient concentrations, and 5) health effects.  11  Concentration response coefficient of 4% ∆mortality per ∆10 µg/m3 ambient PM10 (>age 30).  12  Concentration response coefficient of 0.75% ∆ mortality per ∆10 µg/m3 ambient PM10 (all  ages). 13  Concentration response coefficient of 13% ∆ mortality per ∆10 µg/m3 ambient PM10 (>age 30). 116  A.4.1  Number and Emission Class of Additional Diesels  The number of diesels assumed to be in each EU emission class affects the exposure and health impact estimates. As stated in the methods section, we assumed all additional diesels registered through 2003 were Euro III, declining to 67% in 2004, 34% in 2005, and 0% (i.e., 100% Euro IV) through 2008. Particulate traps (i.e., filters) on diesels are required to equalize emissions of PM10 with petrol (but not necessarily other emissions) [15, 16]. No diesel vehicles were assumed to have traps. In reality, there are three exceptions to these assumptions which we comment as follows [17]: Euro II vehicles in 2001-2002. PM10 emissions from Euro II diesel vehicles average 0.08 g/km (excluding cold start emissions), or about double the Euro III emission rate and quadruple the Euro IV rate. Our analysis assumes that no additional diesels were Euro II. However, in 2001 there were 24 manufacturers offering 131 different Euro II diesel models, compared to 26 manufacturers offering 328 Euro III models. In 2002 there were 10 manufacturers offering 31 different Euro II diesel models, compared to 31 manufacturers offering 459 Euro III models. Moreover, Euro II diesels would have been cheaper than Euro III models.   Early adopted Euro IV with particulate traps prior to 2009. We assume no early adoption of diesel vehicles equipped with particulate traps in our analysis. In 2004, 3 of 33 manufacturers (20 of 910 different diesel models) offering diesels equipped with traps and these were higher priced vehicles. There are currently no known projections for substantial increases in early adoption of diesel vehicles with particulate traps.    Sales of alternative fuel vehicles. The CO2 tax regime may have been at least partly responsible for sales of electric, electric hybrid, or liquefied petroleum gas (LPG) vehicles. However, these vehicles comprised only 0.3% of new registrations or less each year from 2001-2005 and there are currently no known projections for substantial increases. Our projected CO2/km emission factors (Figure A2) were developed in consideration that petrol hybrid vehicles would erode the CO2/km advantage of diesel over time.  In our assessment, the assumptions regarding the number and emission class of diesel vehicles are more likely to underestimate rather than overestimate the PM10 emission estimates.  117  A.4.2  Annual Kilometers Travelled  There are three main phenomena that affect annual travel distance of passenger cars: (1) the type of driver – e.g., a high-travel distance driver versus a low-travel distance driver, (2) changes in vehicle travel distance with age, and (3) an economic rebound effect. Schipper and colleagues [18] have described how diesel vehicles in Europe have been used differently than petrol vehicles, including higher annual travel distance. For example diesels averaged 25,000 km annually compared to 14,000 km petrol in the UK for 1995 [18]. Schipper also described a saturation effect where “as diesels gain market share in each country, the number of naturally high mileage drivers switching to diesel eventually is exhausted, and the later switchers are more likely to represent persons closer to average in their travel patterns.” Our analysis assumptions are consistent with this saturation effect for all diesel growth. To examine this assumption, we examined data from the UK National Travel Survey (NTS) in Figure A3 [19]. Plotted in this figure are the average annual travel distance values for the first two years14 of ownership using data from NTS survey years 1998-2004. The data show that average annual travel distance for newly registered vehicles has been declining for both petrol and diesel (a trend that also appears in the average vehicle stock data reported by the UK Department for Transport). Moreover, the average difference between annual travel distance for diesel and petrol has been declining. For the years 1998-2000 (i.e., prior to the CO2 policies, which had been announced in 1999 and implemented in 2001-02) the average difference for the data plotted in Figure A3 was 13,200 kilometers, compared to a difference of 11,600 kilometers for 2001-04 (i.e., 12% less). These data provide evidence of a trend towards this saturation effect. It should be noted that these data provide only indirect insights with limited sample sizes; a direct examination of this assumption would be obtain data for newly registered diesel cars for those owners that previously had purchased petrol cars, and then compare the annual travel distance, but identifying such owners is not possible with the NTS data.  14  Data is averaged over the first two years for two reasons: (1) in consideration of the small  sample sizes, it provides a more robust average than a single year, and (2) it minimizes capturing changes in annual travel distance due to vehicle age (see Figure A4) or change of ownership (e.g., company cars are sold to non-company car drivers every 2.6 years on the average). 118  Figure A3  Average annual travel distance for the first two years of ownership for dieselfuelled and petrol-fuelled passenger cars. Data are from the UK National Travel Survey. Sample sizes for individual years range from 12 to 563. Both company and privately-owned vehicles are included.  Average annual kilometers for the first two years of ownership  35,000  kilometers per year  30,000 25,000 20,000 15,000  diesel 10,000  petrol Linear (diesel)  5,000  Linear (petrol)  0 1998  1999  NTS = UK national travel survey  2000  2001  2002  2003  2004  NTS survey year  Data on the effect of vehicle age on annual travel distance are presented in Figure A4. These data show that both diesel and petrol vehicles are driven less as the vehicle ages. Since we use the annual travel distance of the average vehicle stock in the UK, which includes old cars and new cars, then the effect of travel distance with age is averaged into our estimate. The higher travel distance for diesel shown in Figure A4 is likely the effect of different drivers – i.e., high travel distance or company car drivers. Again, our assumption is that the driver types that switch from petrol to diesel represent the average driver in terms of annual travel distance. The third phenomenon is the economic rebound effect (i.e., consumers driving more kilometers annually due to reduced marginal cost of driving), which we do not include in our baseline analysis results as described in the main article. While actual, UK-specific fuel price differences could not induce a rebound effect (because fuel price ratio has not changed; see Figure 2.2), the fuel economy advantage of diesels could. UK rebound elasticities are not available, to our knowledge, and developing an econometric model of the rebound effect would be a research effort that would warrant a separate study. However, as a sensitivity analysis, we consider results from a review of rebound effect on vehicle kilometers travelled 119  (VMT) in the U.S. [20]. This review found that long-term, direct rebound effect on VMT to range from 10-30% (fuel cost per kilometer elasticities). Fuel-economy savings between diesel and petrol are proportional to the CO2/km savings, which vary year to year in our analysis (at least until year 2015). For example, a fuel savings of 0.36 liter/100km (equivalent to 10 gCO2/km diesel advantage – see Figure A2) would be realized for a driver switching from petrol to diesel, amounting to a 5% savings in fuel cost per km. Using the general, longterm rebound effect elasticities cited above, this would result in 0.5-1.5% increase in VMT (85-255 km/year in our analysis using 17,000 km/yr average).  Figure A4  Annual travel distance for diesel and petrol cars for NTS survey year 2004. This plot shows the effect of vehicle age on annual travel distance.  Annual kilometers for survey year 2004 (effect of vehicle age on annual mileage) 30,000  kilometers per year  25,000 20,000 15,000 10,000  diesel Linear (diesel)  5,000  petrol Linear (petrol)  1990  1992  1994  1996  1998  2000  2002  2004  year car first registered  Overall, we expect that our assumptions with respect to annual travel distance underestimate PM10 exposure due to diesel growth potentially on the order of 20% or less; however, this level of uncertainty is very small compared to the uncertainty in the PM10 mortality health coefficient which is a factor of 15 (1,500%) difference between the low and the high. In terms  120  of fuel and CO2 savings, our annual travel distance assumptions may overestimate the actual benefits of diesels by 10-30%, but we lack the data to make a better estimate15. A.4.3  Spatial Distribution of Vehicles  The spatial distribution of vehicles in our study is the same as the assumptions of the modelling studies used to estimate ambient PM10 [9, 21]. Distribution of emission sources spatially (and temporally) does influence ambient particle (and ozone) episodes. Higher or lower proportions of vehicles in urban areas will increase or decrease the population weighted ambient PM10 exposure estimate. At a smaller scale, proximity to road traffic by 100 meters or less has been shown to significantly increase cardiopulmonary mortality [22]. We have no evidence to expect the uncertainty in spatial distribution would substantially over- or underestimate PM10 exposure. A.4.4  PM10 Emissions and Ambient Concentrations  Here we discuss assumptions and uncertainties in the PM10 emissions and ambient concentration estimates. The ambient model studies used in our study were complex analyses which applied many assumptions (e.g., 1996-1999 base years for emissions and meteorology, non-gravimetric ambient data converted to gravimetric using 1.3 scaling factor) and assessed many uncertainties (e.g., effect of roadside dust, meteorology, alternate secondary particulate assumptions). Readers are referred to the cited studies for discussions of these assumptions and uncertainty assessments [9, 23]. Contributions of secondary PM10 are included in our study, because they were included in the ambient model studies by way of scaling actual ambient data rather than modelling the atmospheric chemistry [9, 21]. Associated with 3.76 kilo-tonnes PM10 reductions estimated in the model studies were 23 kilo-tonnes of NOx reduction, which is roughly proportional to increases of 0.58 kilo-tonnes PM10 and 4.6 kilo-tonnes NOx in our study. So while NOx contributions to secondary PM10 are incorporated in an inexact manner in our results, the contribution of UK NOx emissions (as secondary nitrates) to mean annual PM10 concentrations is small. One of the reasons for this is that roughly half of the secondary  15  Schipper et al (2002) similarly noted the lack of robust estimates for rebound effects in Europe,  and emphasized that fuel price differences (which exist in various EU countries, but not the UK) had the strongest effect in eliminating diesel’s efficiency advantage. 121  nitrates in the UK are attributed to atmospheric transport of emissions from mainland Europe [24]. Stedman and colleagues [9] found that 23 kilo-tonnes NOx reduction from illustrative traffic-control measures was predicted to reduce secondary PM10 an average of 0.015 g/m3, which represented 5.3% of total PM10 reduction. It should also be noted that total nitrate particle contributions may be small in terms of UK annual means but assessments of ambient data from local episodic events found nitrate particles to contribute substantially [24]. In terms of HC, our analysis shows a reduction of 3.7 kilo-tonnes per annum over the 20year study period. What about the potential reductions in ambient PM10 from the secondary particles? UK ambient-modelling studies assumed the organic fraction of secondary particles to be negligible (i.e., that secondary particulate is comprised only of nitrates and sulphates) [9, 21]. The basis for this assumption is that organic aerosols were found, based on “preliminary models”, to comprise a maximum of 15% of total secondary particles, with a large portion from natural sources [9]. On the other hand, source-apportionment estimates using receptor models found secondary organic aerosols from urban traffic sources to contribute as much as 40% to specific episodic events in summer months [24]. Nonetheless, the studies generally agree that secondary organic particles are still relatively small contributors to total secondary particles based on annual means, and that primary particles still dominate ambient concentrations. Hence we reason that neglecting the reduction in HC due to diesel substitution for petrol vehicles results in a small overestimate of PM10 exposure. There are a few additional exclusions from our analysis which should be mentioned for completeness:   Upstream/downstream emissions and life cycle effects are excluded (e.g., differential emissions of PM10 and CO2 in diesel and petrol production, vehicle manufacturing, and vehicle scrap and salvage). The sole exception is that fuel savings (bbl crude oil) estimates include an adjustment for upstream differences in the production of diesel and petrol.  122    Changes in diesel fuel sulphur content are ignored16, as this has been found to be very small or negligible in terms of reductions in ambient PM10 concentrations by 2010 [9].    Changes in brake and tire wear in PM10 emission estimates are ignored. Combined, these comprise less than 0.01 g/km based on NAEI emission factors [11]. Estimating differences in brake and tire wear between petrol and diesel is highly uncertain, and is likely to be negligible compared to changes in tailpipe emission.    The assumption of constant annual travel distance means that modelled air-quality disbenefits occur later than the real situation, since vehicles are driven more kilometers in the early years of use (Figure A4).    We have not included any explicit increase in emission factors with vehicle age or emission control failures. Research has shown that a small percentage of vehicles can contribute a large proportion of fleet emissions [25]. On-board diagnostics and manufacturer legal requirements to maintain emissions were adopted with the goal of minimizing this phenomenon [26]. Euro III standards are required to be met for 80,000 km and Euro IV standards for 100,000 km. Table A5 provides failure rates for emission tests for petrol and diesel vehicles, although it is not stated which pollutant failed the test. Thus some increase in emissions is still very likely (e.g., even if emissions stay below the legal limit), so this leads to an underestimate of the emissions and exposure, which is expected to be small by comparison to the uncertainty in the mortality risk coefficients.  16  “Ultra low sulphur” diesel and petrol not-to-exceed 50 mg/kg (50 ppm) are to be available as of  2005, while “sulphur free” fuels not-to-exceed 10 mg/kg (10 ppm) are to be available by 2009 (Department for Transport www.dft.gov.uk; [9]). The stated aim of the directive is to enable optimization of vehicle technology for low CO2 emissions, with secondary benefits by way of air pollutants such as NOx and PM10 such as through improved catalyst performance. 123  Table A5  Emission test failure rates for petrol and diesel passenger cars in the UK. 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  Petrol  6.0%  7.4%  7.9%  7.1%  6.4%  5.6%  4.3%  3.1%  2.4%  1.8%  1.4%  Diesel  8.5%  10.6%  9.6%  8.5%  7.3%  5.9%  6.1%  5.5%  5.0%  4.4%  3.8%    Our assumption that post-Euro IV emissions of petrol and diesel vehicles are fully harmonized is not fully realistic. The proposed new standards for Euro V are shown in Table A2. Diesel cars will still be allowed to emit more than three times as much NOx as petrol cars (0.20 g/km diesel compared to 0.06 g/km petrol). Although the same PM10 standard is proposed, the emission factor for petrol vehicles is 0.002 g/km, which is already 60% below the proposed limit; hence it is possible that diesel cars will still emit double the mass emissions of petrol cars and remain within the proposed Euro V limits.  A.4.5  Health Effects  The principal, quantifiable air pollutants of concern on the basis of public health in the UK are PM10, ozone, and SO2 [27]. Given that vehicles contribute relatively little to ambient SO2, and PM10 was assessed quantitatively, the absence of quantitative estimates of changes in ozone health effects remains a substantial limitation of this study, in particular because ozone has been confirmed to be linked to mortality (i.e., not just morbidity) [28]. Ozone health effects were not estimated here because we determined it was not feasible using published studies to make direct quantitative estimates of changes in ambient ozone concentrations based the estimated changes in NOx and HC emissions, which is discussed in the main article. Separate from the contribution to secondary PM10 and ozone, NOx is itself associated with respiratory hospitalizations. An assessment of NOx health effects, independent of ozone or PM10, found that 23 kilo-tonnes annual reduction in NOx by traffic measures would result in 171 fewer respiratory hospitalizations (based on COMEAP dose-response factor of increase in respiratory hospital admissions perg/m3 NO2) [23]. Scaled to our 4.6 kilotonnes annual increase in NOx over the 20-year study period, this translates to roughly 40 additional respiratory hospitalizations per annum. Hence our hospitalization projections are underestimates.  124  Carbon monoxide is another contaminant excluded from our health effect estimates. The estimated reduction due to additional diesels is 10 kilo-tonnes per annum over the 20-year study period. Quantifying the health effects of outdoor carbon monoxide in the UK has been discouraged for reasons of insufficient information [27]. Nonetheless carbon monoxide remains below the UK ambient objective of 10 mg/m3 (maximum daily running 8-hour mean). In 2004 there were no exceedances, and the maximum value at any monitoring station was 5.9 mg/m3. Reductions in carbon monoxide from vehicles has also been implicated in saving lives from accidental and suicidal poisonings in the U.S., which would be an ancillary benefit of diesel cars [29]. Hence, excluding carbon monoxide does overestimate the health effect impacts in our study, although the influence is expected to be minimal by comparison to the PM10 mortality effects. Finally, we discuss the exclusion of health-effect estimates for changes in toxic emissions. Changes in benzene due to additional diesels were estimated to decrease by 0.05 kilotonnes per annum over the study period, and 1,3 butadiene emissions were estimated to increase by 0.003 kilo-tonnes per annum. The health risk associated with these two contaminants are considered “likely to be exceedingly small,” although no safe minimum level has been identified [30]. In terms of the UK air-quality objectives, actual measured levels in 2004 were [11]:   Benzene running annual mean objective is 16.25 µg/m3. There was only one UK monitoring site in 2004, which recorded an annual mean of 10 µg/m3.    1,3 butadiene running annual mean objective is 2.25 µg/m3. There was only one UK monitoring site in 2004, which recorded an annual mean of 0.5 µg/m3.  125  A.5 (1)  REFERENCES TRL, "Data required to monitor compliance with the End of Life Vehicles Directive," Transport Research Laboratory, UK, 2003.  (2)  SMMT, "UK New Car Registrations by CO2 Performance 2005 Annual Report," The Society of Motor Manufacturers and Traders Ltd. (www.smmt.co.uk), 2006.  (3)  DEFRA, "Air quality and climate change: a UK perspective (third report of the Air Quality Expert Group)," UK Department for Environment, Food and Rural Affairs (DEFRA), 2006 draft.  (4)  Sher, E., Ed. Handbook of Air Pollution from Internal Combustion Engines: Pollution Formation and Control; Academic Press, 1998.  (5)  Pischinger, S., Topics in Catalysis 2004, 30/31, 5-16.  (6)  Sluder, C.; B. West, "Catalyzed diesel particulate filter performance in a light-duty vehicle (SAE # 2000-01-2848)," Society of Automotive Engineers, 2000.  (7)  DFT, "Transport Statistics (online data)," UK Department for Transport (www.dft.gov.uk), 2005.  (8)  Marshall, J., W. Riley,T. McKone, W. Nazaroff, Atmospheric Environment 2003, 37, 3455-3468.  (9)  Stedman, J.T. Bush, T. Murrells, M. Hobson, C. Handley, K. King, Revised PM10 projections for the UK for PM10 objective analysis; prepared by AEA Technology for UK Department for Environment, Food and Rural Affairs (DEFRA) and The National Assembly for Wales, The Scottish Executive and the Department of the Environment in Northern Ireland., 2002.  (10)  TandE, "EURO 5 Emission Limits for Passenger Cars and Light Duty Vehicles: Position Paper – September 2005," European Federation for Transport and Environment, 2005.  (11)  NAEI, "National Atmospheric Emission Inventory online database (www.naei.org.uk)," 2005.  126  (12)  DOH, UK Department of Health, 2005.  (13)  Ostro, B., "Outdoor air pollution: assessing the environmental burden of disease at national and local levels," World Health Organization, 2004.  (14)  IR, "Report on the Evaluation of the Company Car Tax Reform: Stage One," UK Inland Revenue (www.inlandrevenue.gov.uk), 2004.  (15)  Walker, A., Topics in Catalysis 2004, 28.  (16)  TandE, Waiting for Euro 5 and Euro 6: New Emission Standards for Passenger Cars, Vans and Lorries (fact sheet); European Federation for Transport and Environment: Brussels, Belgium, 2004.  (17)  VCA; Vehicle Certification Agency (www.vcacarfueldata.org), 2005.  (18)  Schipper, L., C. Marie-Lilliu, L. Fulton, Journal of Transport Economics and Policy 2002, 36, 305-340.  (19)  DFT, "National Travel Survey, 2002-2004 [computer file]. Colchester, Essex: UK Data Archive [distributor] SN: 5340.," UK Department for Transport, 2006.  (20)  Greening, L., D. Greene,C. Difiglio, Energy policy 2000, 28, 389-401.  (21)  Stedman, J.; T. Bush, T. Murrells, K.King, Baseline PM10 and NOx projections for PM10 objective analysis; prepared by AEA Technology for UK Department for Environment, Food and Rural Affairs (DEFRA) and The National Assembly for Wales, The Scottish Executive and the Department of the Environment in Northern Ireland., 2001.  (22)  Hoek, G. B., B; Goldbohm, S; Fischer, P; and van den Brandt, P The Lancet 2002, 360.  (23)  Stedman, J.; T. Bush, T. Murrells, M. Hobson, C. Handley,K. King, Quantification of the health effects of air pollution in the UK for revised PM10 objective analysis; AEA Technology and UK Department of Health, 2002.  (24)  DETR, Source Apportionment of Airborne Particulate Matter in the United Kingdom (report of the Airborne Particles Expert Group); Department of the Environment, Transport and the Regions (DETR) the Welsh Office, the Scottish Office and the Department of the Environment (Northern Ireland), 1999.  127  (25)  Beaton, S., G.; Bishop, Y. Zhang, L. Ashbaugh, D. Lawson, D. Stedman, Science 1995, 268, 991-993.  (26)  EU Directive 98/69/EC relating to measures to be taken against air pollution by emissions from motor vehicles and amending Council Directive 70/220/EEC; European Parliament and the Council of the European Union, 1998.  (27)  COMEAP, Quantification of the effects of air pollution on health in the United Kingdom; The Stationary Office London (www.tsonline.co.uk), 1998.  (28)  Bell, M., F. Dominici, J. Samet, Epidemiology 2005, 16, 436-445.  (29)  Mott, J., M. Wolfe, C. Alverson, S. Macdonald, C. Bailey¸ L. Ball, J. Moorman, J. Somers, D. Mannino, S. Redd, Journal of the American Medical Association 2006, 288, 988-995.  (30)  COMEAP, Statement on Transport and health in London; UK committee on the medical effects of air pollutants (www.advisorybodies.doh.gov.uk/comeap), 1999.  128  APPENDIX B - SUPPORTING INFORMATION FOR CHAPTER 3 “REGULATING CAR MASS FOR CONCURRENT TRAFFIC SAFETY AND CLIMATE MITIGATION BENEFITS”  129  This appendix documents various details for the Introduction, Methods, Results, and Discussion sections of Chapter 3. It will be submitted to The Lancet as an online supporting information document.  B.1  INTRODUCTION  The key premise of the analysis of traffic safety in relation to car CO2 emission rate emissions is that there is an association between traffic safety and car mass. Attempts to quantify the relationship between any parameter and traffic fatalities must consider many interrelated (and often correlated) factors that can confound apparent relationships. Here we list the key factors, followed by Table B1 summarizing several studies of the relationship between mass and traffic safety. 1. Vehicle design described in two categories: (1) basic features: structural stiffness, geometry, car size, engine size, power; and (2) safety technology: seat belts/restraints, airbags, braking systems, stability control, and others [1-6]. 2. Environment and road: urban/rural, number of lanes/separation, speed limits, weather, speed enforcement [6-9]. 3. Driver behavior: seat belt use, choice of speed, choice of environmental conditions to drive under, distractions (e.g., cell phones), alcohol and drug use, sleepiness [1, 2, 5, 10-17] . Examples of Indicator variables for behavior are: previous license suspensions, number of speeding violations, number of previous police-reported collisions, and previous drinking driving violations. 4. Occupant vulnerability: age, sex, health status, alcohol use [7, 11, 15, 18, 19]. 5. Crash event described in two categories: (1) crash mode: single-car, multiple-car, car-other vehicle types (heavy trucks, motor cycles, mopeds, bicycles), carpedestrian and (2) crash type: points of impact, rollovers, hit objects [3, 17, 20, 21]. Note that some parameters are listed under more than one category. For example, seat belts are clearly a design feature of the vehicle with large variability in efficacy such as simple designs in many 1980’s vehicles to advanced tensioning systems in many new vehicles. However seat belt use is a behavior which has a very strong influence on the risk of fatalities [2, 22].  130  Table B1 Study  Summary of selected studies which estimated the effect of vehicle mass on fatality and injury risk. Description  Key findings  Data requirements  Strengths  Limitations  U.S. National  Quasi-induced exposure methods  100 lbs (45 kg)  Individual-level crash  Careful sensitivity  Does not assess the  Highway  are employed, although the main  reductions in  data for driver,  analysis for many  effects of size or other  Transportation  results reported are absolute  vehicle mass (and  environment, and  factors, except size.  correlates such as power  Safety  fatality risk using vehicle  size, which is  cars. Curb mass is  Statistical  independent of mass.  Administration  registration years for cars, and  correlated to mass)  the principal  adjustment of risk  Hence all results are to be  (NHTSA) by  vehicle travel distance for “light  increases overall  independent variable  estimates for “bad”  interpreted as the  Kahane[20]  trucks.” Regression analysis to  fatality risk.  of interest. Driver  driver behavior  combined effect of mass,  estimate the change in fatality risk  behavior data such  using 9 metrics.  size, power, and other  for “case” cars for a 100 lbs (45 kg)  as alcohol use, drug  Large sample sizes  correlated attributes that  decrease in mass. Compares cars  use, suspended  and relatively low  were not separately  of mass M to cars M-100 lbs,  license, crash  uncertainty.  accounted for in the  statistically adjusted for age, sex,  frequency history,  Includes risk  analysis. Collision speed  road environment, impact point,  and various law  estimates for all  controlled based on speed  etc.; includes adjustment for driver  violations.  major crash modes  limit of 55 miles/hour, an  behavior variables. U.S. fatality  and types, except  imprecise measure of  analysis reporting system (FARS)  that motorcycle,  speed. Evaluates only  and other U.S. data sources for  mopeds, and  fatalities, excludes injuries.  calendar years 1995-2000, and  pedestrians are  model years 1991-1999. Excludes  combined as one  2-door cars because of the  category.  documented association between risk taking drivers and 2-door cars.  131  Study  Description  Key findings  Data requirements  Strengths  Limitations  Mengert and  Estimates absolute fatality risk.  Decreasing fleet  Individual-level crash  Data requirements are  Absolute risk estimates are  Borener [23]  Analyzed changing mass mix of  mass increases  data for cars and  minimal. Risk  not adjusted for many  cars in the U.S. Directly calculates  fatality risk for  fatality counts  estimates using this  important variables (size,  fatality risk factors for collisions of  single and two-car  including curb mass  method were  stiffness, seat belt use,  cars in various mass groups  crashes, reduces  as the principal  compared to results  behavior), other than mass  without adjustment for variables  risk for pedestrians,  independent variable  using NHTSA’s  and its correlates. Risk  describing casualties, road, or  with an overall net  of interest. Driver  detailed statistical  estimates are characterized  vehicle. These fatality risk factors  increase in risk.  data (age, sex, etc.)  models and found to  by high degree of uncertainty.  are used to estimate changes in  and road  compare reasonably  Evaluates only fatalities,  risk due to changes in mass  environment data are  well, thus validating  excludes injuries.  distribution for on-road cars.  not required.  this method [20].  Toy and Hammit  Estimates conditional fatality and  Increasing delta V,  Study used U.S.  Relatively unique  Body type, size, vehicle  [24]  Abbreviated Injury Scale (AIS) risk,  a measure of the  Crashworthiness  study in that it uses  safety equipment, and other  based on police-reported crash  combined effect of  Data System with  delta V as an  vehicle attributes are not  events. Uses “delta V” as an  mass and crash  detailed data on  independent variable,  included. Assessed only two-  independent variable, estimated for  severity (including  driver, road  thus accounting for the  car crash modes. Being a  each crash based on detailed  speed),  environment, and  combined effect of  conditional risk analysis,  crash event parameters.  significantly  crash variables for  crash energy  crash avoidance (including  increases fatality  roughly 6,500  (surrogate for “crash  driver behavior) is not  risk in two car  fatalities.  severity”) and mass.  included.  collisions. But  Adjusts for restraint  vehicle type and  use.  other factors are also important.  132  Study  Description  Key findings  Data requirements  Strengths  Limitations  Wood and  Estimates relative fatality and AIS  For two-car frontal  German, Japanese,  Detailed accounting  Crash avoidance (including  Simms [25]  risk. Model is developed to  collisions, length  and U.S. FARS  of vehicle attributes  driver behavior) is not  estimate the risk in car-to-car  ratio explains injury  empirical data for  to separate the  included. Only two-car frontal  collisions considering three basic  and fatality risk  fatalities are used.  effects of mass,  collisions are assessed. Not  ratios: mass, length (size), and  distributions better  size, and crash  clear if age and gender of  energy absorbed.  than mass ratio or  severity.  casualties are adjusted in the  collision energy. Broughton [26]  risk estimates.  Estimates conditional fatality risk  Uniform 10%  Individual-level crash  Estimates risks for  Method has been criticized for  for UK model year 1991-94 cars.  reduction in mass  data for drivers, road  most crash modes  using mass ratio instead of  Vehicle mass is the fundamental  reduces fatality risk  environment, and  (single-car, car-  mass difference. Size,  independent variable. Conditional-  in single-car, two-  cars using UK Data  pedestrian, car-heavy  stiffness, power, safety  fatality risks are used to estimate  car, and car-  Archive. Curb mass  goods, and car-car)  equipment, and other  changes in downsizing of on-road  pedestrian crashes.  was input from an  and vulnerable road  important variables not  fleets by assuming frequencies of  independent source.  users (vehicle  available. Crash avoidance  crash events remain unchanged  Detailed data for on-  occupants,  (including driver behavior) is  with changes in mass.  road cars in the UK  pedestrians).  not included. Evaluates only  are used.  fatalities, excludes injuries.  133  Study  Description  Key findings  Data requirements  Strengths  Limitations  Wenzel and  Risk defined as fatalities per  Mass is ‘not  Crash events and  Use of absolute fatality  Mass and risk relationship  Ross [6, 17]  quantity of registered cars. Risk to  fundamental” traffic  fatality rates  risk, divided into risk to  described as a “popular  drivers, risk to others, and  safety.  disaggregated by  drivers and risk to  belief” without supporting  combined risk (sum of driver and  make and model.  others, is a meaningful  assessment of previous  others) are quantified for many  Driver behavior  risk metric directly  research. High level analysis  makes and models of cars in the  variables. Vehicle  linked to common  aggregating the effect of  U.S., using primarily FARS data.  design attributes.  traffic safety policy  many factors into one result  Driver behavior and annual travel  Annual travel  agendas, including  (e.g., crash avoidance and  distance statistics are compared  distance data.  occupant protection  crashworthiness, multiple  and car compatibility.  crash modes combined into  Provides convincing  one risk-to-driver estimate).  evidence linking  Behavior assessed with  observed crash data  descriptive statistics, not as  and underlying crash  control variables (e.g., as with  physics for the effect  NHTSA [20]). Choice of be  of some design  behavior risk metric can affect  features on driver risk  the ability to generalize  such as unibody cf.  findings [27], therefore non-  body on frame, track  subtle effects may not be  width, and center of  ruled out. The “design quality”  mass.  and country of origin effects  amongst different car groups.  explain risk variation for some important car groups but not all, and no connection to underlying physics (e.g., delta V) is provided. Evaluates only fatalities, excludes injuries.  134  Study Evans [3]  Description  Key findings  Data requirements  Strengths  Limitations  Estimates relative fatality risk for  Adding 75 kg  Uses individual-level  Makes use of new  Wide range of model years is  U.S. drivers of lighter cars versus  mass, holding size  crash data for cars  equations to separate  employed, 1975-1998,  drivers of heavier cars in head-on  constant, in two-car  based on fatality  the effects of mass  therefore confounding by  crashes. New equations are  head-on crashes  counts, crash  and size.  improved safety technology  employed to separate the effects of  reduces driver  involvements, and  and vehicle design (e.g.,  size and mass. Makes use of  fatality risk 8%,  vehicle mass. Age,  Evans showed elsewhere that  added passenger mass to  while increasing  sex, or other data are  relationship between size and  separate the effects of size and  risk to collision  not used.  mass for 1975-1979 cars is  mass. Mass is the fundamental  partner driver by  significantly different than  parameter assessed.  about 8%.  model year 1980-1990 cars [28]), yet these data points are all combined in the analysis. Moreover, one relationship between mass and size is used develop new equations to separate the effects of mass and size, even though the relationship between mass and size changes over the analysis period. No accounting for driver behavior. Evaluates only fatalities, excludes injuries.  135  Study Van Auken [21]  Description  Key findings  Data requirements  Strengths  Research objectives were to  Reducing mass,  Individual-level crash  Relatively unique in  Driver behavior variables  separate the effects of mass, size,  while holding  data for driver,  quantitatively  excluded. Some important  crashworthiness, crash avoidance,  wheelbase and  environment, and  separating the effects  vehicle features such as  and compatibility. Uses data from 7  track width  cars.  of vehicle mass and  stiffness and electronic  U.S. states for calendar years  constant,  size. Includes multiple  stability control, were not  1995-1999 crashes, and model  decreases fatality  crash modes: two-car,  accounted for. Used 3  years 1985-1998 cars and 1985-  risk.  single-car (rollovers,  relatively blunt variables for  1997 light trucks. Quasi-induced  Reducing  hit objects), car-  age and sex (young drivers,  exposure methods were employed,  wheelbase or track  pedestrian/bicycle/  old males, old females).  with the main results reported  width, holding  motorcycle, car-heavy  Evaluates only fatalities,  based on vehicle registration years  mass constant,  goods/bus. Vehicle  excludes injuries. Collision  increases fatality  control variables  speed controlled based on  risk.  included airbags and  speed limit being 55  antilock brakes,  miles/hour, an imprecise  vehicle age, four/all  measure of speed.  as the measure of exposure.  Limitations  wheel drive, and two doors. Crandall [29]  Landmark study of the relationship  CAFE responsible  Aggregate data on  Robust assessment of  Aggregate data. Correlates of  between U.S. CAFÉ standards and  for ~ 225 kg (500  fuel economy, vehicle  the effect of CAFE  vehicle mass (e.g., size,  traffic fatalities. Authors argued  pounds) reduction  mass, engine size,  policy on fuel economy  power) could confound effect  that CAFÉ had strong influence on  in 1989 model cars,  fuel prices, steel  and vehicle mass.  of mass. Effects of safety  vehicle mass (assessed with  which in turn  prices, etc. Based on  technology over time not  economic models), then assessed  increased fatalities  model years 1970-  necessarily captured in the  effect of vehicle mass on total  2,200-3,900 for  1987.  model. Evaluates primarily  fatalities.  model year 1989  fatalities, with sensitivity  cars over 10 years  analysis for injuries..  (i.e., ~ 300 per year)  136  Study Bedard [30]  Description  Key findings  Data requirements  Strengths  Limitations  Estimated the independent effects  Increased mass  Individual-level crash  Accounts for some  Independent effects of mass  of vehicle, driver, and collision  and size together  data for driver,  important behavior  and size were not estimated  characteristics using U.S. FARS  reduces fatality risk  environment, and  variables (alcohol and  (wheelbase and weight  data for single vehicle crashes with  in single-car, fixed-  cars.  seat belt restraint use),  correlated r = 0.82, so  fixed objects.  object crashes.  collision speed, and  wheelbase dropped). Ability  vehicle characteristics  to generalize is limited as  (mass, model year,  many crash modes excluded  and age). Internal  (e.g., multi-vehicle,  validity high based on  pedestrians, motorcycles,  assessment of single-  single-car rollovers).  vehicle, fixed object  Evaluates only fatalities,  crashes.  excludes injuries.  Padmanaban  Estimated driver-fatality odds  Mass ratio of two-  40 different vehicle  Relatively unique in  Evaluates only fatalities,  [31]  (conditional risk) in two-vehicle  car collisions  parameters, plus seat  assessing 12 different  excludes injuries. Effect of  collisions (car-car and car-light  explains fatality risk  belt use, drinking  size metrics for size  other critical vehicle variables  truck) in the U.S. based on various  more than any  driving, and age/sex.  such as overall  omitted such as age and  vehicle size metrics, controlled for  other vehicle  length/width/height,  stiffness. Crash avoidance  mass. Used 1990-2000 calendar  variables, including  volume, and front  not assessed.  year data for model years 1981  various size  overhang.  and later. Evaluated separately  metrics.  frontal collisions, left-side impact,  “Equalizing” fleet  and right-side impact.  mass will reduce overall fatality risk.  137  B.2  METHODS  B.2.1  Methods: Data Sources  Details for electronic databases received are provided in Table B2. At the end of this appendix is a data dictionary for this study. Figures B1 through B11 contain a summary of some important UK statistics and time trends for general reference. Table B2  Summary description of electronic databases received1.  Database Name (for This Study)  Database Names (Source)  Number of Records  Description/Comments  UK Data  acc94, acc95,  2,719,913 total  data sets for year 1994-2005 (inclusive)  Archive crash  …acc05  for 12 years  with data fields describing each police-  database [32]  (UK DFT via the UK  reported crash event in the UK  Data Archive) UK Data  veh94, veh95,  4,967,866 total  data sets for year 1994-2005 (inclusive)  Archive vehicle  …veh05  for 12 years  with data fields describing vehicle types  database [32]  (UK DFT via the UK  for all police-reported crash event in the  Data Archive)  UK; specific make/model/version data is not included  1  Complete database dictionary can be provided upon request.  138  Database Name (for This Study)  Database Names (Source)  Number of Records  Description/Comments  UK Data  cas94, cas95,  3,698,606 total  data sets for year 1994-2005 (inclusive)  Archive  …cas05  for 12 years  with data fields describing each casualty  casualty  (UK DFT via the UK  (slight injury, serious injury, or fatality) for  database [32]  Data Archive)  all police-reported crash events in the UK; excludes behavioral data such as previous motoring offences  DFT  makemodel94-05  3,9168107  provides individual data on passenger  make/model  (direct from UK DFT  total for 12  cars linkable to the UK Data Archive data  crash database  December, 2006)  years  sets (linking fields = accref and vehref)2.  [33]  Vehicle data is provided for roughly ¾ of all crash records (excludes heavy goods vehicles, bicycles, motorcycles, and others). Model years 1990 and later included (yr1stregi code ≥1990) 23,452 different make/model/versions (all casualties) 6,489 different make/model/versions (fatalities only)  DFT  070611-D-make  293,902 total  make, model, make code, model code,  make/model  model 2006 (direct  records  body type, propulsion, year first  version  from DFT June,  registered, engine size group, and  database [33]  2007)  number of cars registered year end 2006  JATO database  JATO Dynamics  4,304 total for  Make, model, and version data for UK  [34]  Ltd. (latest version  years 8 years  cars. Years 2000-2007 are included, but with very few make/model/versions for earliest years (3 records for 2000; 186 records for 2004; 1,990 records for 2007).  2  Database variables: accref = accident reference number; vehref = vehicle reference number.  139  B.2.2  Methods: Summary Traffic Safety Statistics  Figure B1  Historical traffic fatality rates for the UK.  14.0 per billion vehicle km 12.0  per 100,000 population  Fatality Ratio  10.0 8.0 6.0 4.0 2.0 0.0 1990  1992  1994  1996  1998  2000  2002  2004  Year  Figure B2  Number of vehicles involved in fatal crashes 1994-2005.  Number of vehicles involved (numveh) for all fatal crashes 1994-2005  1994 1995  60%  1996 1997  50%  1998  40%  1999  30%  2001  2000 2002  20%  2003 2004  10%  2005  0% 1  2  3  4  ≥5  140  80% 70% 60% 50%  Road types for all fatal crashes 1994-2005.  Road types (roadtype) for all fatal crashes 1994-2005 1994 1995 1996 1997 1998  40% 30% 20% 10% 0%  1999 2000 2001 2002  single carriageway: 4+ lanes  single carriageway: 3 lanes  single carriageway: 2 lanes  single carriageway: single track  dual carriageway: 3+ lanes  dual carriageway: 2 lanes  one way street  2003  unknown  Figure B3  2004 2005  141  Road class for all fatal crashes 1994-20053.  Figure B4  Road class (1rdclass) for all fatal crashes 1994-2005  1994  70%  1995  60%  1996 1997  50%  1998  40%  1999 2000  30%  2001  20%  2002 2003  10%  2004  0% d si cl un  M  Figure B5  fie  C  B  A  as  ot or  A( M  )  w ay  2005  Road speed limits for all fatal crashes 1994-2005.  Road speed limits (speedlim) for all fatal crashes 1994-2005 50% 45%  1994 1995  40% 35% 30%  1996 1997 1998  25% 20% 15% 10%  1999  5% 0%  2003  2000 2001 2002 2004  ≤20 mph  3  30 mph  40 mph  50 mph  60 mph  70 mph  2005  Database variable: 1rdclass is the first road class type (there can be more than one road involved, namely an  intersection). Road class descriptions can be found at www.dft.gov.uk/matrix/forms/definitons.aspx .  142  Figure B6  Casualty types for all fatalities 1994-2005.  Casualty types (typecas) for all fatalities 1994-2005  60% 50%  1994  40%  1995 1996 1997  30%  1998  20%  1999  10%  2001  2000 2002 2003  0% pedestrian  Figure B7  cyclist  car (excludes taxi)  taxi  motorcycle minibus, bus, or coach  other  2004 2005  Sex for all fatalities 1994-2005.  Sex for all fatalities 1994-2005 80%  heavy goods vehicle  male female  70% 60% 50% 40% 30% 20% 10% 0% 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005  143  Figure B8  Age and sex for all fatalities in 2005.  Age and sex for all fatalities in 2005 (n = 3,201)  % of all 2005 fatalities  4.5% 4.0%  male  3.5%  female  3.0% 2.5% 2.0% 1.5% 1.0% 0.5% 0.0% 0  10  20  30  40  50  60  70  80  90  100  age Figure B9  Crash mode for all single vehicle crashes 1994-2005.  Single vehicle fatal crashes: skid, overturn, & jackknife 70%  1994  60%  1995 1996 1997  50% 40%  1998  30%  1999  20%  2000  10%  2001  0%  2002 no skid, overturn, or jackknife  skid  skid & overturn  overturn  jackknife & missing values  2003 2004 2005  144  Figure B10  Objects struck off carriageway for all single vehicle crashes 1994-2005.  Single vehicle fatal crashes: objects struck off the carriageway  1994 1995  70%  1996  60%  1997  50%  1998  40% 1999  30%  2000  20%  2001  10%  2002  Figure B11  ditch & submerge water bus stop & other permanent  central or side barrier  tree  utility pole  lamp post  sign or signal  no hit object out carriageway  0%  2003 2004 2005  Point of impact for two vehicle fatalities 1994-2005  Two vehicle fatalities (numveh=2) and impact points (1stptimpac)  1994 1995 1996  70%  1997  60%  1998  50%  1999  40%  2001  2000 2002  30%  2003  20%  2004 2005  10% 0% did not impact  front  back  offside  nearside  145  B.2.3  Methods: Using Surrogate Data to Ascertain Curb Mass  A critical limitation of the UK Data Archive and DFT databases is the lack of curb mass data. While there are independent data sources for curb mass that can be manually linked to our analysis data set, make/model alone is insufficient to look up a precise mass. As an example, Figure B12 shows the variation of curb mass amongst versions of the 2007 Ford Focus[34], which has consistently been the most registered make/model range in the UK in recent years. The overall standard deviation and range of curb mass are 62 kg and 268 kg, respectively. Table B3 similarly shows wide variation in curb mass for the 10th most registered make/model car, the BMW Series 3 [34]. Considering that a 45 kg (100 lb) change in mass has been shown to significantly affect injury and fatality risk [20], this limits the ability to account for changes in mass from the UK Data Archive data.  Figure B12  Curb mass (kg) histogram for model year 2007 Ford Focus versions4.  Ford Focus ‐ curb mass histogram 35 30  Frequency  25 20 15 10 5 0 1206 1241 1277 1312 1347 1383 1418 1453 1489 1524 1559 1595 More  kgBin  4  Ford Focus was the 1st most newly registered UK car in 2006. (note that Ford Focus C-MAX is a different model  range than the Ford Focus, and is not included in these data.)  146  Table B3  Variation in curb mass and tailpipe CO2 emission rate for model year 2007 Ford Focus versions offered in the UK. number of versions  min kg  max kg  range kg  mean kg  std dev kg  min gCO2/ km  max gCO2/ km  range gCO2/ km  mean gCO2/ km  std dev gCO2/ km  diesel  62  1,323 1,630  307  1,401  54  130  163  33  144  54  gasoline  88  1,206 1,547  341  1,315  64  160  233  73  179  64  overall  150  1,206 1,630  424  1,351  73  130  233  103  165  23  Table B4  Variation in curb mass and tailpipe CO2 emission rate for model year 2007 BMW Series 3 versions offered in the UK. BMW Series 3 was the 10th most newly registered UK car in 2006. number of versions  min kg  max kg  range kg  mean kg  std dev kg  min gCO2/ km  max gCO2/ km  range gCO2/ km  mean gCO2/ km  std dev gCO2/ km  diesel  30  1,415 1,810  395  1,594  109  135  213  77  173  20  gasoline  39  1,320 1,810  490  1,549  149  158  310  153  207  30  overall  69  1,320 1,810  490  1,568  134  135  310  175  192  31  For all analyses, except the “first law” RR, we dealt with the lack of curb mass data in our study by using engine size as a surrogate measure for mass. Table B10 and Figure B14 illustrate with model year 2007 cars that engine size alone provides statistically significant explanatory power (R2 = 59%) for curb mass, independent of make/model or any other trim level variable. For the “first law” RR analysis, the combination of make/model, engine size category, body type, and propulsion provided enough data to manually look curb mass for every make and model car, using mass data from various independent sources [34-37]. The DFT make/model version database (see Table B2) includes engine size (a categorical variable), body type (e.g., 5-door hatchback, saloon, estate), and propulsion (e.g., gasoline, diesel, electric). For reasons of determining a curb mass as well as other analysis tasks, we needed to link the UK Data Archive databases with the DFT make/model crash database and the DFT make/model version database. However, there was no code to specify the unique combinations of versions for a given 147  make, model, and model year car. The best available option was to concatenate the codes for make, model, and model year (year first registered assumed to be model year). However, this presented challenges. For example, consider the Peugeot (make code = L1) 307S (model code =866) with year first registered = 2003 (presumed to be model year). When we concatenate these fields (to create a new code = makmodregi), we get the following 13 version or trim combinations, all associated with the same make, model, and model year car as shown in Table B5. So the 2003 Peugeot 307S is associated with at least 3 different engine size ranges, 3 different body types, and 3 different propulsion types. Table B5  Range of engine sizes, body types, and propulsion types associated with a model  year 2003 Peugeot model 307S. makmodregi5  engsizeDFT  body type  propulsion  L1-866-2003  1,201-1,500  3 Door Hatchback  Petrol  L1-866-2003  1,501-1,800  Estate  Petrol  L1-866-2003  1,801-2,000  5 Door Hatchback  Petrol  L1-866-2003  1,201-1,500  5 Door Hatchback  Petrol  L1-866-2003  1,501-1,800  5 Door Hatchback  Petrol  L1-866-2003  1,501-1,800  3 Door Hatchback  Petrol  L1-866-2003  1,801-2,000  3 Door Hatchback  Petrol  L1-866-2003  1,801-2,000  Estate  Petrol  L1-866-2003  Over 30006  3 Door Hatchback  Petrol  L1-866-2003  1,801-2000  5 Door Hatchback  Heavy Oil  L1-866-2003  1,201-1500  5 Door Hatchback  Heavy Oil  L1-866-2003  1,201-1500  5 Door Hatchback  Gas/Bi-Fuel  L1-866-2003  1,801-2,000  Estate  Gas/Bi-Fuel  5  Database variable that is concatenate of make, model, and year first registered.  6  This engine size is likely a coding error in the database for this particular record.  148  In terms of the complete study databases, there were 13,777 unique makmodregi codes which were part of a crash involving a fatality for years 1994-2005. All of these makmodregi codes needed to be matched with an engine size, yet 83% of these 13,777 makmodregi codes were associated with more than one engine size code. To assign a unique engine size (code = engsizeDFT) to each makmodregi, we used the following procedure: 1. Sort makmodregi codes: assign as "unique" or "duplicate." a. "unique" for which only one engine size category is listed (2,310 codes) b. "duplicate" for which more than one engine size category is listed (11,467 codes or 83%) 2. For “duplicate,” we used the number of registered cars in 2006 (code = numcars7) for a given makmodregi as the criteria to assign an engine size category. In other words, the engine size of the most populous makmodregi on the roads in 2006 was assumed to be the engine size for all cars in the analysis data set for a given makmodregi code. The details of this procedure follow: a. Data field named “max” was created to identify the most populous engine size code for each makmodregi code. b. Of the 11,467 makmodregi codes, 8,164 (71%) codes had only value for “max,” thus it was clear which engine size code to assign. c. Of the 11,467 makmodregi codes, 3,303 (29%) codes had more than value for “max” d. If “max” included more than four engine sizes, or four engine-size categories that were not continuous (e.g., 1,201-1,500, 1,501-1,800, 1,801-2,000, 2,501-3,000) we coded these as “error” for the engine size and these records were excluded from our analysis data set. This is because it is very unlikely that any given makmodregi spans over four different engine size categories (e.g., the Ford Focus spans only three, the Peugeot 307S spans four). The result is that 86 makmodregi codes were assigned as “error” for engine size. e. If “max” included four or less engine sizes, we used the total, on-road fleet numbers to assign an engine size. This created the following hierarchy in selecting the engine  7  Database variable: number of cars involved in the collision event.  149  size in these cases: 1,501-1,800, 1,201-1,500, 1,801-2,000, 1,001-1,200, 2,0012,500, 701-1,000, 2,501-3,000, >3,000, <701. If 1,001-1,200, 701-1,000, and 1,2011,500 were all coded as “max,” then 1,201-1,500 was the assigned engine size on the basis of having a larger % of all on-road cars in 2006. This procedure for assigning engine size would be unnecessary if a version code were included in the databases, in addition to first registered (i.e., model year), make, and model. The DFT has advised that a version or trim code can now be made available only for cars, and only cars initially registered after March, 2001 (Paul Syron personal communication, 8 August 2007). It should be noted that the scope for all of our analyses included cars first registered beginning in either 1996 (the “first law” RR analysis) or 1994 (the absolute risk analysis), so having these codes would help somewhat, but would not be expected to make large differences in the precision of our study. It is noted here in the interest of documenting data availability for future research. To supplement the graphs of annual time series for registered cars used to develop scenarios, we provide here Table B6 of annual registrations for UK on-road numbers of cars, which is only available disaggregated by engine size (CC). This comprises the “baseline” scenario in our analysis.  150  Table B6  Number (in 1,000s) of registered private and light goods vehicles by engine size8. 1994  1995  1996  1997  1998  1999  2000  2001  2002  2003  2004  2005  54  46  42  37  29  18  19  23  29  37  47  52  700 < CC ≤ 1,000  1,905  1,757  1,678  1,564  1,459  1,435  1,415  1,368  1,314  1,237  1,199 1,153  1,000 < CC ≤ 1,200  2,261  2,258  2,327  2,336  2,293  2,275  2,228  2,244  2,252  2,221  2,210 2,139  1,200 < CC ≤ 1,500  5,337  5,225  5,321  5,418  5,497  5,600  5,677  5,819  5,894  5,939  6,089 6,181  1,500 < CC ≤ 1,800  6,276  6,345  6,540  6,655  6,766  6,922  6,992  7,124  7,241  7,284  7,405 7,439  1,800 < CC ≤ 2,000  3,088  3,274  3,550  3,828  4,090  4,389  4,604  4,869  5,166  5,398  5,686 5,929  2,000 < CC ≤ 2,500  759  791  851  925  1,003  1,094  1,159  1,275  1,400  1,520  1,639 1,725  2,500 < CC ≤ 3,000  486  494  524  548  574  608  630  666  704  762  841  918  CC > 3,000  313  315  340  371  403  443  473  510  543  587  638  671  CC ≤ 700  All private and light goods, excluding: “other vehicles” and motorcycles,  20,479 20,505 21,173 21,682 22,114 22,784 23,197 23,898 24,543 24,985 25,754 26,207  scooters, and mopeds other vehicles  2,192  2,217  2,267  2,317  2,362  2,427  2,469  2,544  2,622  2,730  2,900 3,019  630  594  609  626  684  760  825  882  941  1,005  1,060 1,075  Motor cycles, scooters and mopeds  8  “Private and light goods” includes all vehicles used privately. The bulk of this group consists of private cars (whether  owned by individuals or companies) and vans and light goods vehicles. The group also contains a number of important minority groups including private buses and coaches, private heavy goods vehicles, and some vehicles not exceeding 3,500 kg which, before 1st July 1995, were taxed in specialized taxation classes. “Other vehicles” includes three-wheeled cars and vans not exceeding 450 kg unladen mass, recovery vehicles and general-haulage vehicles, as described above. Motorized tricycles are included but motor cycle combinations are included with motor cycles.  151  B.2.4  Methods: “First Law” RR Analysis  The process of data sorting was an important part of the analysis and is illustrated in Figure B13. For various reasons, actual totals of casualties in the UK are believed to be less than recorded in the UK Data Archive [38, 39]. Table B7 is a comparison of critical statistics for the complete UK Data Archive, the “first law” RR analysis data set, and various intermediate linking and sorting steps.  Figure B13  Diagram of the UK Data Archive data sorting process.9 Analysis data set casualties UK Archive casualties All road casualties  Table B7 provides a comparison of critical statistics for the complete UK Data Archive, the “first law” RR analysis data set, and various intermediate linking and sorting steps. The second column shows statistics for the complete UK Data Archive set for crash years 1996-2005 including only driver fatalities (5,354 in total) involved in crashes between two cars (i.e., excluding taxis, motorcycles,  9  While the casualty totals in the analysis data set and the complete UK Data Archive have specific counts, the actual  quantity of all road casualties in the UK is unclear as illustrated by the dashed line.  152  heavy goods, etc.). The next column shows that matching with the “make/model” database reduced the basic data set to 4,801 driver fatalities. The next step was to limit the data set to model year cars 1996-2005, resulting in 1,062 driver fatalities. Linking to the “version” database was necessary to provide engine size category and related data fields in order to accurately lookup curb weight, which reduced the driver fatalities in the data set to 946. Thus far, the data set includes 946 drivers who died in two-car collisions driving a case car of model year 1996-2005, but for most of these collisions the other car involved was 1995 or earlier. Limiting the data set further to include only driver fatalities in which the other cars was also model year 1996-2005 produced the analysis data set with 144 driver fatalities. Table B7  “First law” RR analysis: comparison of data set statistics.  UK Data Archive  UK Data Archive linked with makemodel data  UK Data Archive linked with makemodel data  UK Data Archive linked with makemodel 10 and engsize  "first law" RR analysis data set  numveh11  2  2  2  2  2  typecas  9  9  9  9  9  classcas  1  1  1  1  1  severcas  1  1  1  1  1  crash years  1996-2005  1996-2005  1996-2005  1996-2005  1996-2005  case car model year (yr1stregi)  all (no data)  <1978-2005  1996-2005  1996-2005  1996-2005  other car model year (yr1stregi)  any  any  any  any  1996-2005  median yr1stregi  no data  unknown  1998  1998  1998  mean yr1stregi  no data  unknown  1998.3  1998.1  1998.2  std dev yr1stregi  no data  unknown  2.0  1.9  1.8  number of driver fatalities  5,354  4,801  1,062  946  144  driver/fatality age, mean  44.7  44.7  45.5  45.7  49.8  driver/fatality age, std dev  21.5  21.6  21.4  21.5  22.6  10  makemodel is a database variable that is concatenate of make and model; engsize is variable for engine size.  11  Database variables: numveh = number of vehicles involved in collision event; typecas is the casualty type (9 = car  occupant); classcas = casualty class (1 = driver); severcas is severity of casualty (1 = fatal). yr1stregi is year first registered, assumed to be the model year of the car.  153  UK Data Archive  UK Data Archive linked with makemodel data  UK Data Archive linked with makemodel data  UK Data Archive linked with makemodel 10 and engsize  "first law" RR analysis data set  driver/fatality, % male  73.6%  73.5%  70.6%  70.5%  72.2%  % front impact  59.3%  58.9%  56.1%  54.5%  50.0%  % rear impact  3.6%  3.6%  3.0%  3.3%  2.1%  % off-side impact  22.2%  22.6%  25.0%  25.7%  30.6%  % near-/driver-side impact  14.9%  14.8%  15.8%  16.5%  16.7%  speed limit 30 mph, %  13.9%  13.9%  12.1%  11.6%  6.9%  speed limit 40 mph, %  7.4%  7.5%  7.2%  6.8%  4.2%  speed limit 50 mph, %  3.7%  3.9%  3.8%  4.1%  4.2%  speed limit 60 mph, %  62.8%  62.4%  62.2%  61.9%  69.4%  speed limit 70 mph, %  12.1%  12.3%  14.8%  15.6%  15.3%  55.2  55.3  56.0  56.3  58.2  78.6%  78.5%  75.7%  74.8%  75.7%  speed limit, average single carriageway, 2 lanes  We had intended in this study to quantify what Evans describes as a “second law” for two cars of equal mass involved in a collision12. However, we had too few data points for equal mass collisions to perform this analysis. The “second law” relationship follows. RRb ≡ k / M where: RRb ≡ (risk of matched pair at given curb mass) / (risk of 1400kg13 matched pair) k = constant determined empirically M = mass of car 1 = mass of car 2 This “second law” has direct implications for uniform downsizing of vehicle fleets, such as in response to policies mandating fleet-wide improvements in fuel economy. This relationship shows  12  RRb is used here to distinguish it from the first law, which was denoted as RR.  13  Evans’ choice of 1,400 kg is arbitrary, as one could choose another reference mass.  154  that uniform downsizing will, ceteris paribus, result in increased fatality rates even though the mass of both vehicles remains equal. This observed relationship is consistent with well known laws of physics (conservation of energy and momentum) in that less combined mass means less physical protection for vehicle occupants. B.2.5  Methods: Absolute Risk Analysis  This method uses actual fatality risk rates for cars grouped by engine size (a surrogate measure for mass), and then predicts changes in total annual fatalities for hypothetical changes in the fleet composition. The basic data requirements, rationale for selecting this method, and on-road fleet scenarios for this method are provided in the main manuscript. Here we provide a description of the essential details of this method for the three crash modes included in our study: single-car, carpedestrian, and two-car. Uncertainties were quantified assuming crash events and fatalities adhere to a poisson process as has been shown by Evans, and that standard errors for single quantities can be estimated as n½ as long as n is roughly 7 or larger [7]. The uncertainties for combined quantities such as ratios or differences of single quantities, then conventional error propagation formulas are employed. The method applied to single-car fatalities and car-pedestrian fatalities follows:   risk factor ≡ Ki = Fi / Ri    i = mass categories    Fi = number of total single-car or car-pedestrian fatalities involving mass category i    Ri ≡ baseline proportions of registrations for mass category i    scenarios are developed to produce new distributions of fleets ≡ Ri’    finally, new fatality counts are estimated Fi’ ≡ Ki * Ri’    note: Ki factors are proportional to the basic fatality rate per quantity of registered cars  Example: F1 = 50 fatalities in a given year for mass group #1 R1 = 0.15 (15% of on road registrations for mass group #1) 155  K1 = 50 / 0.15 = 333 now assume a scenario with new distribution of cars: R1’= 0.10 F1’ = 333 * 0.10 = 33 fatalities (decrease of 17, or 34% for mass category #1) Statistics for the single-car and car-pedestrian analysis data sets as compared to the complete UK Data Archive are provided in Tables B8 and B9, respectively.  Table B8  Single-car Absolute risk analysis: comparison of data set statistics. single-car analysis data set (linked with engsizeDFT14)  UK Data Archive linked with makemodel data  only UK Data Archive  number of cars (numveh=1 )  1  1  1  type of casualty (typecast=9)  car  car  car  class of casualty (classcas=1or2)  driver or passenger  driver or passenger  driver or passenger  severity of casualty (severcas=1)  fatal  fatal  fatal  crash years, range  1994-2005  1994-2005  1994-2005  car year 1st registered, range  1990-2005  1990-2005  all (no data)  car year 1st registered, median  1991  1991  no data  car year 1st registered, mean  1993  1993  no data  3.5  3.4  no data  fatalities, total number  5,104  5,135  5,371  fatalities, mean age  31.0  31.0  31.2  fatalities, std dev age  17.3  17.3  17.4  fatalities, % male  78.1%  78.1%  78.2%  drivers, mean age  31.6  31.7  31.8  car year 1st registered, std dev  14  Variable from the make and model database provided by DFT: engsizeDFT is engine size field.  156  single-car analysis data set (linked with engsizeDFT14)  UK Data Archive linked with makemodel data  only UK Data Archive  16.4  16.5  16.5  drivers, % male  86.1%  86.1%  86.1%  no skid, no overturn (skidoturn15 = 0)  38.0%  38.0%  38.0%  skid only (skidoturn = 1)  29.6%  30.2%  30.2%  skid and overturn (skidoturn = 2)  20.3%  19.7%  19.6%  overturn only (skidoturn = 5)  12.1%  12.1%  12.1%  hit object in carriageway: none  79.2%  79.0%  79.1%  hit object out of carriageway: tree  35.7%  35.3%  35.1%  road: single carriageway16, 2 lanes  71.0%  71.2%  70.9%  speed limit 30 mph, %  22.9%  23.4%  23.1%  speed limit 40 mph, %  8.7%  8.6%  8.3%  speed limit 50 mph, %  2.7%  2.8%  3.0%  speed limit 60 mph, %  49.7%  49.5%  49.6%  speed limit 70 mph, %  15.9%  15.8%  16.0%  52.7  52.6  52.7  drivers, std dev age  speed limit average mph  15  Database variable: skidoturn defines the single car collision event.  16  “single carriageway” is defined as a road type without a physical barrier separating each direction of travel; “dual  carriageway” is a road type with some form physical barrier separating each direction of travel; single and dual carriageway roads can have multiple lanes in each direction (e.g., 3, 4, 5, 6 or more total lanes).  157  Table B9  Comparison of data set statistics for car-pedestrian absolute risk analysis. car-pedestrian analysis data set (linked with engsizeDFT)  UK Data Archive linked with makemodel data  only UK Data Archive  1-16  1-16  1-16  86.2%  86.5%  86.6%  type of casualty (typecast=9)  car  car  car  class of casualty (classcas=3)  pedestrian  pedestrian  pedestrian  fatal  fatal  fatal  crash years, range  1994-2005  1994-2005  1994-2005  car year 1st registered, range  1990-2005  1990-2005  all (no data)  car year 1st registered, median  1991  1991  no data  car year 1st registered, mean  1993  1993  no data  3.5  3.4  no data  fatalities, total number  5,192  6,385  7,275  fatalities, mean age  49.9  50.0  50.0  fatalities, std dev age  27.1  27.2  27.2  fatalities, % male  64.9%  64.6%  64.7%  drivers, mean age  36.9  36.9  36.8  drivers, std dev age  15.9  15.9  15.8  drivers, % male  76.6%  77.3%  77.6%  pedestrian location: crossing in carriageway outside of crossing  55.3%  55.3%  54.6%  pedestrian physical facility: no crossing facility within 50m  77.3%  77.2%  77.2%  road: single carriageway, 2 lanes  69.8%  70.1%  69.8%  number of cars (numveh ) percentage involving just 1 car  severity of casualty (severcas=1)  car year 1st registered, std dev  158  The two-car analysis procedure is as follows. As there are two groups involved rather than one, a multiplicative risk model is employed.   risk factor ≡ Ki-j = Fi-j / (Ri * Rj)    Fi-j ≡ the number of fatalities between cars in mass categories i and j    Ri and Rj ≡ baseline proportions of registrations for mass categories i and j    Scenarios are developed to produce new distributions of fleets ≡ Ri’ and Rj’    Finally, new fatality counts are estimated Fi-j’ ≡ Ki-j * (Ri’ * Rj’)  For example: F1-2 = 45 fatalities in a given year R1 = 0.10 (mass group #1 = 10% all cars) R2 = 0.05 (mass group #2 = 5% all cars) K1-2 = 70 / (0.10 * 0.05) = 9000 now assume a scenario with new distribution of cars: R1’= 0.15 and R2’ = R2 = 0.05 F1-2’ = 9000 * (0.15 * 0.05) = 68 (increased from 45) fatalities involving two-car collisions between cars of mass categories #1 and #2 For all three absolute risk analyses, the same 8 categories were included as described for the scenarios, using engine size as a surrogate for mass (i.e., 701-1,000 CC as group 1, … 3,000 and over CC as group 8.). These groupings functioned well for the single-car and car-pedestrian analyses, but not for the two-car analysis. For two-car collisions, there were 64 K factors (i.e., K1-1, K1-2, … K8-8). It was found that statistically significant estimates of fatality risk could be made in only 13 of the 64 grouped pairs of collisions due to low fatality counts. This was strictly an outcome due to limited number of observations in the two-car data set. A work around solution was developed by creating just 3 categories: “lighter” (701-1,500 CC), “mid-mass” (1,501-2,000 CC), and “heavier” (2,001 to over 3,000 CC). With these coarser groupings, then the low fatality count problems were resolved. The fatality risk factors and scenario market shares for two-car collisions were calculated with these groupings, which enabled the use of the same scenario market shares for two-car collisions as used for the single-car and car-pedestrian analyses. 159  B.2.6  Methods: Statistical Models for Curb Mass and CO2/km  Table B10 summarizes statistical models for curb mass and CO2/km. Regression analyses were all developed as ordinary least squares (OLS) fit using R statistical software (v2.5.1) or Microsoft Excel 2003. Regression models #2 and #9 were used in this study as follows.   Model #2 was used in the “first law” RR analysis and the AnalyticaTM simulation analysis to estimate curb mass given engine size in the data set.    Model #9 was used in the “first law” RR analysis to estimate gCO2/km given engine size in the data set.  Additional models are included for general reference only. All OLS regression models are based on natural log transformed variables because this was required to satisfy the normality assumption. Homoskedacity was verified by visual review of residual plots. The ln model form also has the convenient property that the b1 and b2 slope coefficients represent elasticity. E.g., 1% change in diesel car mass results in a 1.113% change in CO2/km.  160  Table B10  Summary results of regression models relating curb mass (kg) and gCO2/km to explanatory variables using JATO data. intercept, a  model regression model  n  R2  p  point  slope, b1  95% CI 95% CI std err  point  95%  95%  CI  CI  residual std err  Comment  std err alternate equation:  #1  ln (kg curb mass) = a + b1 * ln (m wheelbase) + error  3,298 0.611  < 0.0001  4.695  4.625  4.765  0.036  2.619  2.547 2.690  0.036  0.0020  curb mass = exp(error) * exp(a) * (wheelbase)^b1 shows that engine  #2  ln (kg curb mass) = a + b1 * ln (cc engine size) + error  size alone has 3,298 0.587  < 0.0001  3.705  3.603  3.807  0.052  0.469  0.455 0.482 0.0069  0.0021  substantial predictive power for mass  ln (kg curb mass) = a + b1 * ln #3  (cc engine size) + b2 * ln (m  b2 = 1.332 [1.2813,298 0.769  < 0.0001  3.530  3.453  3.607  0.039  0.231  0.217 0.244  0.007  0.0016  overall length) + error  1.383 95% CI and std error=0.026] shows that mass  ln (petrol and diesel gCO2/km) #4  has substantial  = a + b1 * ln (kg curb mass) + 3,242 0.494  < 0.0001  -1.321  -1.549  -1.092  0.117  0.902  0.870 0.933  0.016  0.0028  predictive power for CO2/km, regardless  error  of fuel (diesel/petrol) #5  #6  ln (diesel gCO2/km) = a + b1 * ln (kg curb mass) + error ln (petrol gCO2/km) = a + b1 * ln (kg curb mass) + error  1,294 0.780  < 0.0001  -3.016  -3.252  -2.779  0.121  1.113  1.081 1.146  0.016  0.0027  1,948 0.728  < 0.0001  -2.224  -2.429  -2.020  0.104  1.040  1.011 1.068  0.014  0.0026  161  intercept, a model regression model  #7  #8  #9  ln (diesel kg curb mass) = a + b1 * ln (cc engine size) + error ln (petrol kg curb mass) = a + b1 * ln (cc engine size) + error ln (gCO2/km) = a + b1 * ln (cc engine size) + error  n  R2  slope, b1  p  point  95% CI 95% CI std err  1,317 0.614  < 0.0001  3.005  2.820  3.190  0.094  1,980 0.633  < 0.0001  3.897  3.785  4.009  1,948 0.841  < 0.0001  0.984  0.901  1.068  point  95%  95%  CI  CI  residual std err  std err  0.568  0.544 0.592 0.0124  0.0028  0.057  0.439  0.424 0.454 0.0075  0.0026  0.043  0.568  0.557 0.579 0.0056  0.0020  Comment  162  Because models #2 and #9 were used in one of the fatality risk analyses in this study, we provide the line fit and residual plots in Figures B14 and B15.  Figure B14  Regression model #2 line fit and residual plot.  ln (engine size) Line Fit Plot  9.0 ln (kerb wt) Curb kg  ln (kerb wt) curb kg Predicted ln (kerb wt) curb kg  8.0 7.0 6.0 6.0  7.0 8.0 ln (engine size)  9.0  ln (engine size) Residual Plot  Residuals  0.5 0 6.0  7.0  8.0  9.0  -0.5 -1  outliers are: Caterham & Lotus  -1.5 ln (engine size)  163  Figure B15  Regression model #9 line fit and residual plot. ln(gCO2/km) Line fit plot  6.5  ln(gCO2/km)  6.0 5.5 5.0 ln(gCO2/km)  4.5  Predicted ln(gCO2/km)  4.0 6.0  7.0  8.0  9.0  ln (engine size)  0.8  ln (engine size) Residual Plot  Residuals  0.6 0.4 0.2 0 6.0  6.5  7.0  7.5  8.0  8.5  9.0  -0.2 -0.4 ln (engine size)  164  B.3  RESULTS  Figure B16 is a plot of the “first law” RR analysis for the five mass groupings. While the model is statistically significant, it is limited by the number of observational mass groupings that could be created. Figure B17 is a line fit plot of the conditional risk model that relates two-car collision % risk of serious injury or fatality to CO2 emission rate Line fit plot of “first law” RR regression model that relates mass ratio to RR fatalities in two car collisions. Top panel models the relationship as RR = µλ. Bottom panel models the relationship as RR = α + β * µ. µ = mass ratio. 2.5  ln(RR)  2.0 1.5 1.0 0.5  ln(RR) Predicted ln(RR)  0.0 0.0  9.0  0.1  0.2 0.3 ln(wt ratio) ln(mass ratio)  0.4  0.5  RR Predicted RR  8.0 7.0 6.0 RR  Figure B16  5.0 4.0 3.0 2.0 1.0 0.0 0.0  0.5  1.0  1.5  2.0  mass ratio  165  Figure B17  Line fit plot of conditional risk model that relates two-car collision % risk of serious injury or fatality to CO2 emission rate. 12 10  F or SI risk, %  risk, %  8  Predicted F or SI risk, %  6 4 2 0 100  150  200  250  300  350  CO2/km mean  B.4  DISCUSSION  To adequately examine a power law relationship then data spanning at least two orders of magnitude would be best. However our UK data span less than this. We cannot further subdivide our RR data into more mass categories to increase the span of our mass ratios because our fatality counts would become too low to employ the assumption of a poisson distribution [7]. To examine the effect of point of impact in interpreting the “first law” RR results, we attempted to fit the data to the following model RR = A * μλ with the following result. A = 0.15 [95% CI -0.66 to + 0.93] p < 0.63 λ= 4.84 [95% CI 1.56 to + 8.12] p < 0.018 R2 = 0.88 This model is invalid based on the p-value for the constant A. The ratio of river fatality risk has been found to fit the power law relationship in all published work that was found in the literature [25, 40, 41]. Nonetheless, we attempted to fit our data to a linear model as described in Section 3.3, and these data are plotted in Figure B16. With the linear model and µ = 2, the point estimate of RR is 14.9 (i.e., ratio of driver fatality of lighter cars to heavier cars is 14.9 for a mass ratio of 2).  166  A limitation of our fleet simulation RR analysis (using AnalyticaTM) is that we assumed randomized collisions between two cars. In general, the issue of selective recruitment has been found in traffic safety where drivers exhibiting varied levels of risk-taking behaviors are associated with different uses of technology such as seat belt use [2]. This means that drivers of lighter and heavier cars may not be equally likely to be involved in collisions. Additionally, randomized collisions are based only on vehicle counts in our model, and annual distance traveled can vary by mass of cars. While we have no data on risk taking behavior, we were able to perform some validation of our assumption by comparing annual travel distance data [42] and crash involvements disaggregated by engine size groups [32, 33] as shown in Figures B18 and B19. Although the four largest engine size groups averaged approximately 40% higher annual travel than the four smallest engine size groups (under 700 CC group excluded due to small registration counts), a comparison of crash involvements reveals no obvious deviation from registration counts for all groups. Hence the assumption of randomized crash events appears to be reasonable. Figure B18  Mean annual travel distance by engine size category in 1998 and 2004 for cars in UK National Travel Survey [42]. Includes vehicles coded as “cars” and “Landrover/Jeep”, but excludes “light vans”. Error bars are +/- one standard deviation. n=8,207 for year 2004; n=8,692 for year 1998.  25,000 1998  2004  15,000  10,000  5,000  25 01 -3 00 0 30 01 & ov er  18 01 -2 00 0 20 01 -2 50 0  12 01 -1 50 0 15 01 -1 80 0  0  70 110 00 10 01 -1 20 0  Annual miles  20,000  Engine size group, CC  167  Figure B19  A comparison of the proportions of registered cars and recorded crash involvements for year 2005, disaggregated by engine size [32, 33].  35%  2005 registrations 2005 crash events  registrations or crash events  30% 25% 20% 15% 10% 5% 0%  engine size (CC)  168  This research indicates that the distribution of vehicle mass is important in achieving traffic safety goals. It is therefore important to note that the UK trend in registrations has seen a growing disparity in vehicle mass as evidenced by the changes shown in Figure B20 [43. There has been a declining “middle class” of cars, with the proportion of both larger, heavier cars and smaller, lighter cars increasing substantially as shown. Change in annual new car registrations in 2006 as compared to 1997 in the UK.  200% 150% 100% 50%  -50%  in i Su pe rm Lo in we i rM ed U iu pp m er M ed iu m Ex ec ut Lu iv e xu ry Sa Sp lo ec on ia lis tS po rts 4x 4s /S U M Vs ul ti Pu rp os e  0% M  change in new regitrations 1997-2006  Figure B20  169  B.5  REFERENCES  1.  Austin, R., “Vehicle aggressiveness in real world crashes. “ 2005, National Highway Traffic Safety Administration (U.S.).  2.  Evans, L., “Safety-belt effectiveness: The influence of crash severity and selective recruitment.” Accident Analysis and Prevention, 1996. 28(4): p. 423-433.  3.  Evans, L., “Causal influence of car mass and size on driver fatality risk. “ American Journal of Public Health, 2001. 91: p. 1076-1081.  4.  Joksch, H., “Velocity change and fatality risk in a crash - a rule of thumb. “ Accident Analysis and Prevention, 1993. 25: p. 103-104.  5.  Mosedale, J. and A. Purdy, “Excessive speed as a contributory factor to personal injury road accidents. “ 2003, United Kingdom Department for Transport.  6.  Ross, M., D. Patel, and T. Wenzel, “Vehicle design and the physics of traffic safety. “ Physics Today, 2006(January): p. 49-54.  7.  Evans, L., “Traffic safety. “ 2004: Science Serving Society.  8.  Patterson, T., W. Frith, L. Povey, and M. Keall, “The effect of increasing rural interstate speed limits in the United States.“ Traffic Injury Prevention, 2002. 3: p. 316-320.  9.  Redelmeir, D., R. Tibshirani, and L. Evans, “Traffic-law enforcement and risk of death from motor-vehicle crashes: Case-crossover study. “ The Lancet, 2003. 361: p. 2177-2182.  10.  Dobson, A., W. Brown, J. Ball, J. Powers, and M. McFadden, “Women drivers’ behaviour, socio-demographic characteristics and accidents.“ Accident Analysis and Prevention, 1999. 31: p. 525-535.  11.  Evans, L., “Innate sex differences supported by untypical traffic fatalities.“ Chance, 2006. 19(1): p. 10-15.  12.  Klauer, S., T. Dingus, V. Neale, J. Sudweeks, and D. Ramsey, “The impact of driver inattention on near-crash/crash risk: An analysis using the 100-car naturalistic driving study data. “ 2006, National Highway Traffic Safety Administration.  13.  Lourens, P., J. Vissers, and M. Jessurun, “Annual mileage, driving violations, and accident involvement in relation to drivers’ sex, age, and level of education. “ Accident Analysis and Prevention, 1999. 31: p. 593-597.  14.  Shope, J., “Influences on youthful driving behavior and their potential for guiding interventions to reduce crashes.“ Injury Prevention, 2006. 12(Suppl I): p. i9-i14. 170  15.  Vingilis, E. and S. Macdonald, “Review: Drugs and traffic collisions. “ Traffic Injury Prevention, 2002. 3: p. 1-11.  16.  Warner, H. and L. Aberg, “Drivers' decision to speed: A study inspired by the theory of planned behavior. “ Transportation Research, 2006. Part F 9: p. 427-433.  17.  Wenzel, T. and M. Ross, “The effects of vehicle model and driver behavior on risk.“ Accident Analysis and Prevention, 2005. 37: p. 479-494.  18.  Andersson, A., O. Bunketorp, and P. Allebeck, “High rates of psychosocial complications after road traffic injuries. “ Injury, 1997. 28(8): p. 539-543.  19.  Langford, J. and S. Koppel, “Epidemiology of older driver crashes – identifying older driver risk factors and exposure patterns.“ Transportation Research, 2006. Part F 9: p. 309-321.  20.  Kahane, C., “Vehicle weight, fatality risk and crash compatibility of model year 1991-99 passenger cars and light trucks.“ 2003, National Highway Traffic Safety Administration (U.S.).  21.  Van Auken, R. and J. Zellner, “A further assessment of the effects of vehicle weight and size parameters on fatality risk in model year 1985-98 passenger cars and 1985-97 light trucks.“ 2003, Dynamic Research Inc.  22.  Sen, A., “An empirical test of the offset hypothesis. “ Journal of Law and Economics,, 2001. 44: p. 481-510.  23.  Mengert, P. and S. Borener, “Overall fatality risk to the public at large related to national weight mix of passenger cars (dot-tsc-hs070-pm-89-27). “ 1989, National Highway Traffic Safety Administration (NHTSA).  24.  Toy, E. and J. Hammitt, “Safety impacts of SUVs, vans, and pickup trucks in two-vehicle crashes. “ Risk Analysis, 2003. 23(4): p. 641-650.  25.  Wood, D. and C. Simms, “Car size and injury risk: A model for injury risk in frontal collisions “ Accident Analysis and Prevention, 2002. 34: p. 93-99.  26.  Broughton, J., “The likely effects of downsizing on casualties in car accidents.“ 1999, Transport Research Laboratory (UK).  27.  Fernandes, R., R.F. Soames-Job, and S. Hatfield, “A challenge to the assumed generalizability of prediction and countermeasure for risky driving: Different factors predict different risky driving behaviors.“ Journal of Safety Research, 2007. 38: p. 59-70.  28.  Evans, L. and M. Frick, “Car size or car mass: Which has greater influence on fatality risk? “ American Journal of Public Health, 1992. 82(8): p. 1105-1112. 171  29.  Crandall, R. and J. Graham, “The effect of fuel economy standards on automobile safety.“ Journal of Law and Economics, 1989. 32: p. 97-118.  30.  Bedard, M., G. Guyatt, M. Stones, and J. Hirdes, “The independent contribution of driver, crash, and vehicle characteristics to driver fatalities. “ Accident Analysis and Prevention, 2002. 34: p. 717-727.  31.  Padmanaban, J. “Influences of vehicle size and mass and selected driver factors on odds of driver fatality. “ in 47th Annual Proceedings of the Association for the Advancement of Automotive Medicine. 2003.  32.  DFT, “Road accident data [computer files] 1994 sn3461, 1996 sn3627, 1997 sn3817, 1998 sn4027, 1999 sn4080, 2000 sn4356, 2001 sn4501, 2002 sn4588, 2003 sn4979, 2004 sn5042, 2005 sn5244, 2006 sn5493. “ 2007, UK Data Archive (Colchester, Essex).  33.  DFT, “Make-model crash database linkable to UK data archive data (electronic database). “2006, UK Department for Transport.  34.  JATO. “Jato dynamics new car database for Great Britain. “ 2007 [cited June 3, 2007].  35.  Globalcar, “Online car specifications (www.Globalcar.Com). “ 2007.  36.  Street-car, “Online car specifications (www.Street-car.Net). “ 2007.  37.  Whatcar, “Online car specifications (www.Whatcar.Co.Uk).” 2007.  38.  DFT, “Under-reporting of road casualties - phase 1 (road safety research report no. 69).” 2006, UK Department for Transport.  39.  Mackay, M., “Commentary: Quirks of mass accident data bases.” Traffic Injury Prevention, 2005. 6: p. 308-310.  40.  Joksch, H., D. Massie, and R. Pichler, “Fatality risks in collisions between cars and light trucks (dot/hs 808 802).” 1998, National Highway Traffic Safety Administration (NHTSA).  41.  Ross, M. and T. Wenzel, “Losing weight to save lives: A review of the role of automobile weight and size in traffic fatalities.” 2001, American Council for an Energy-Efficient Economy.  42.  DFT, “National travel survey, 2002-2004 [computer file].” Colchester, essex: UK data archive [distributor], sn: 5340. 2006.  43.  SMMT, “Motor industry facts - 2007.” 2007, Society of Motor Manufacturers and Traders (SMMT) (www.smmt.co.uk).  172  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0069914/manifest

Comment

Related Items