New Vehicle Feebates: Theory and Evidence Brandon Schaufele (w/ Nic Rivers) Department of Economics University of Ottawa brandon.schaufele@uottawa.ca Heartland Environmental & Resource Economics Workshop Urbana-Champaign, Illinois November 2-3, 2013
New Vehicle Feebates What are they? Feebates are policies aimed at improving the average fuel economy of the new vehicle fleet Fees or taxes are levied on relatively fuel inefficient vehicles Rebates or subsidies are granted to relatively fuel efficient vehicles Fee + Rebate = Feebate Feebates exist in Canada, France and Austria Proposed in Arizona, California, Connecticut, Maine, Maryland, Massachusetts, New York, Oregon and Wisconsin
New Vehicle Feebates Why use them? Governments are unable or unwilling to pursue first-best pollution policies Canada and US impose corporate average fuel economy standards (CAFE) and greenhouse gas intensity regulations Feebate policies and CAFE standard often viewed as substitute policies (Gillingham, 2013; Klier and Linn, 2013) Revenue from fees is used to finance rebates Almost always treated as revenue neutral Feebates are receiving increasing attention Green et al. (2005); Greene (2009); D Haultfoeuille, Durmeyer, Février (2011); Adamou, Clerides, Zachariadis (2012); Gillingham (2013); Roth (2012)
New Vehicle Feebates Unresolved issues in literature 1. Analysis has focused on ad hoc feebate programs What are implications with optimally designed feebates? 2. Many consumer welfare properties of these programs remain unknown 3. Few empirical estimates of important feebate elasticities No quantitative analysis of Ontario s Tax and Credit for Fuel Conservation Program (Ontario s Feebate)
Ontario s Feebate Program Brief timeline 1989 Gas guzzler tax introduced 1990 Tax rates doubled and expanded to cover more vehicles CAW and manufacturers express concerns 1991 Tax schedule amended and subsidy on fuel efficient vehicles incorporated SUVs category added Highway fuel efficiency 1989 1990 1991-2010 rating (L/100km) Cars Cars Cars SUVs less than 6.0 - - -100-6.0-7.9 - - 75-8.0-8.9-200 75 75 9.0-9.4-700 250 200 9.5-12.0 600 1200 1200 400 12.1-15.0 1200 2400 2400 800 15.1-18.0 2200 4400 4400 1600 over 18.0 3500 7000 7000 3200
Ontario Sales Mix over Time
Unanswered Question that Precedes Policy Design Do modest feebates induce behavioral responses?
Example: 2004 Ford Mustang Redesign Relative market share of Ford Mustang in rest of Canada compared with Ontario Ford Mustang, 4.6 litre, 8 cylinder
Data All private vehicle registrations in Canada by province and FSA for 2000-2010 Data from Desrosiers Consultants (Canadian RL Polk data) Merge this dataset with NRCan fuel consumption ratings Also merge StatCan data on retail gasoline prices, consumer prices, population, income, etc. Three definitions of a vehicle Provincial-level data contains approx. 600 unique vehicles across 10 years Approximately 60,000 observations Mapping data to FSAs yields nearly 3.2M observations
Behavioural Response using Provincial Data Feebates have a large effect when vehicles are defined as make-model Very similar results for make-model-series or make-model-series-displacement-# of cylinders (1) (2) (3) (4) (5) Feebate ($1000) -0.51-0.46-0.44-0.47-0.65 ( 0.13) ( 0.10) ( 0.09) ( 0.12) ( 0.11) N 59,579 59,579 59,579 59,579 59,579 vehicle province-class class-year region-class-year vehicle-province vehicle-year fuel cost
Behavioural Response at FSA-level Disaggregating the data to the FSA-level and controlling for additional unobservables reduces the responsiveness (1) (2) (3) Feebate ($1000) -0.27-0.08-0.07 ( 0.01) ( 0.01) ( 0.01) N 3,201,083 3,201,083 3,201,083 mmy mmsy mmsecy mmf mmsf mmsecf fuel cost
Properties of New Vehicle Feebate Programs Under a revenue neutrality constraint
Properties of New Vehicle Feebate Programs Under a revenue neutrality constraint 1. Optimal Indirect Pigouvian tax is heterogeneous across vehicles (Knittel and Sandler, NBER, 2013; Starrett, 1972) Optimal feebate is a vertical translation of this schedule
Optimal Vehicle Tax v. Optimal Feebate tax Optimal vehicle tax 0 emissions
Optimal Vehicle Tax v. Optimal Feebate tax Optimal vehicle tax Optimal feebate (R = 0) 0 emissions
Properties of New Vehicle Feebate Programs Under a revenue neutrality constraint 1. Optimal Indirect Pigouvian tax is heterogeneous across vehicles (Knittel and Sandler, NBER, 2013; Starrett, 1972) Optimal feebate is a vertical translation of this schedule
Properties of New Vehicle Feebate Programs Under a revenue neutrality constraint 1. Optimal Indirect Pigouvian tax is heterogeneous across vehicles (Knittel and Sandler, NBER, 2013; Starrett, 1972) Optimal feebate is a vertical translation of this schedule 2. Feebates can replicate any binding CAFE standard (Klier and Linn, 2012; Gillingham 2013)
Properties of New Vehicle Feebate Programs Under a revenue neutrality constraint 1. Optimal Indirect Pigouvian tax is heterogeneous across vehicles (Knittel and Sandler, NBER, 2013; Starrett, 1972) Optimal feebate is a vertical translation of this schedule 2. Feebates can replicate any binding CAFE standard (Klier and Linn, 2012; Gillingham 2013) 3. Assuming no change along extensive margin implies that optimal vehicle taxes and optimal feebates yield identical levels of total social welfare Feebates approach the Pigouvian benchmark when only relative prices matter Taxes and feebates have distinct distributional profiles
Marginal Private Benefits Marginal Social Benefits Relatively Inefficient Vehicle Social Optimum Private Equilibrium Relatively Efficient Vehicle Share vehicle 1 Share vehicle 2 Total Vehicle Purchases
Marginal Private Benefits Marginal Social Benefits Relatively Inefficient Vehicle Social Optimum Private Equilibrium Relatively Efficient Vehicle Share vehicle 1 Share vehicle 2 Total Vehicle Purchases
Relatively Inefficient Vehicle Social Optimum Private Equilibrium Relatively Efficient Vehicle B C E F A Share vehicle 1 D Share vehicle 2 Total Vehicle Purchases G
Properties of New Vehicle Feebate Programs Under a revenue neutrality constraint 1. Optimal Indirect Pigouvian tax is heterogeneous across vehicles (Knittel and Sandler, NBER, 2013; Starrett, 1972) Optimal feebate is a vertical translation of this schedule 2. Feebates can replicate any binding CAFE standard (Klier and Linn, 2012; Gillingham 2013) 3. Assuming no change along extensive margin implies that optimal vehicle taxes and optimal feebates yield identical levels of total social welfare Feebates approach the Pigouvian benchmark when only relative prices matter Taxes and feebates have distinct distributional profiles
Properties of New Vehicle Feebate Programs Under a revenue neutrality constraint 1. Optimal Indirect Pigouvian tax is heterogeneous across vehicles (Knittel and Sandler, NBER, 2013; Starrett, 1972) Optimal feebate is a vertical translation of this schedule 2. Feebates can replicate any binding CAFE standard (Klier and Linn, 2012; Gillingham 2013) 3. Assuming no change along extensive margin implies that optimal vehicle taxes and optimal feebates yield identical levels of total social welfare Feebates approach the Pigouvian benchmark when only relative prices matter Taxes and feebates have distinct distributional profiles 4. Welfare gains come from the shift in the mix of vehicles across differentiated fuel consumption ratings
Sales of Vehicle by Model Distribution of Sales without Feebate Distribution of Sales with Feebate Fuel Efficiency of Vehicle by Model
Feebates with Substitution Across Pre-defined Vehicle Classes Nested Logit Structure Data includes 17 vehicle classes E.g., subcompact, compact, luxury, small pickup, etc. (1) (2) (3) Feebate ($1000) -0.50-0.10-0.30 ( 0.08) ( 0.03) ( 0.04) Fuel cost -1.63-0.78-2.33 ( 0.12) ( 0.05) ( 0.15) Share in class 0.22 0.66 0.17 ( 0.06) ( 0.02) ( 0.05) N 59,579 59,579 59,579 vehicleprovince vehicleyear
Comparison: Optimal to Actual Based on 2005 Ontario sales and vehicles characteristics
Counterfactual Simulations Emissions Social Cost Welfare Revenue Reduction of Carbon Gain Scenario ($M2002) (Mt CO2) ($/CO2) ($M2002) Actual 249 1.26 25 31.5 Optimal revenue neutral design: Same welfare 0 2.67 11.8 31.5 Parry and Small upper bound 0 25.8 59.2 1,527 Optimal vs. Ontario feebate Optimal revenue-neutral feebate is continuous in fuel consumption rating; Ontario program differentiates between classes and uses notches" Ontario feebate rates reasonably close to calculated optimal levels
Challenges to Optimal Policy Design Asymmetric responses Ontario consumers are more responsive to rebates than fees Heterogeneity Rural households are less responsive than urban households
Final Remarks Feebates have been proposed in several jurisdictions but few empirical studies exist Potential for revenue neutral design make them attractive policy options Ontario s policy generated net welfare improvements of $31M for SCC of $25/tCO 2 Optimal feebate could have achieved same welfare via greater emissions reductions with a breakeven SCC of $12/tCO 2 Feebates only affect vehicle fleet mix not fleet size or distance traveled (similar to CAFE) Externalities are generated by kilometers driven Optimal gas tax (Parry and Small, AER, 2005) still a better policy option