Home / Development of Options for a Vehicle Feebate in Canada / Feebates – Modeling

Feebates – Modeling

DEVELOPMENT OF OPTIONS FOR A VEHICLE FEEBATE IN CANADA

Previous - Contents - Next

4. MODELING OF FEEBATE OPTIONS

4.1 DESCRIPTION OF MODELS

separate models were used in this analysis:  

• TransportCanada Variant of Greene et al. Vehicle Purchase Model.  This is a spreadsheet-based, nested multinomial logit,consumer choice model that estimates the effect of feebates on consumer purchasing behaviour and manufacturer investment in fuel economy technologies. Each individual vehicle model is included (830 in total North American market), as well as its purchase price and fuel economy. The model solves for fuel economy changes that maximize consumer surplus given a defined feebate. Consumer surplus is maximized in response to the feebate “price” by:
  • shiftingvehicle demand towards more fuel-efficient carsand reducing overall vehicle demand
  • inducingmanufacturers to maintain or improve market shareby improving fuel economy, and
  • providingfuel savings to the consumer through lower overallfuel consumption.
  The
  model produces aggregate national results for a single future year (15 years in the future), representing the new long-run equilibrium impact of the feebate. Transport Canada modified the US model to use aggregated 2003 Canadian and US sales data, updated the technology cost curves based on a 2005 literature review (necessary to track manufacturer responses to the feebate), and added redesign thresholds provided by EEA. (These EEA redesign thresholds are used to determine if sales are adequate in Canada to induce manufacturers to make manufacturing investments that produce fuel economy improvements.) A key assumption in the model is how consumers respond to the price signal introduced by the feebate. The model used the following elasticity assumptions to model the price response of consumers:  
  • -10@ market share of 15 percent within a class 
  • -5@ market share of 10 percent between classes 
  • -1.0for overall sales 
  Some
  limitations of these assumptions are discussed below.  
• NRCanVehicle Stock Model. This is a simple  representation of vehicle turnover and usage overa 15-year period. The model also incorporates the technology redesign schedule provided by EEA for the US Energy Information Administration in order to estimate the timing of technology investments. Outputs include the path of annual fuel savings and GHG emission reductions leading up to the fifteenth year.

4.2 BASE CASE AND ASSUMPTIONSTo

isolate the effect of the feebate, a base case is assessed. This case is generated by the model and based on allowing consumer and manufacturer behaviour to evolve naturally in response to the availability of cost saving-technology with no feebate. The base case has not been calibrated to NRCan's reference case and it does not incorporate macroeconomic or demographic factors. However, in evaluating the incremental impact of feebates, these are not major limitations. key assumptions that remain constant for all scenarios are:  

• Asingle fuel-efficiency cost curve for each class (providedby EEA)
• 2000minimum threshold for redesign of import vehicles 
• 20,000minimum threshold for redesign of domestic vehicles 
• Vehiclelife of 15 years 
• Averageyearly distance traveled of 23,500 km, declining 4percent per year
• 15percent adjustment of fuel consumption rating to approximateon-road conditions
• 23percent rebound effect (where decreased fuel costsinduce more driving, thus negating a portion of the feebate gains)
• 10percent discount rate. 

LIMITATIONS

all modeling exercises, the current one is a crude approximation of reality designed to provide policy guidance. While the model has limitations that we highlight below, it is our view that the model can be used to inform policy. Limitations include:  

• Decisionbasis. In the real world, manufacturers  would redesign based on a number of complex considerationsand consumers would be driven by a variety of considerations that are not easily represented by a simple elasticity function. In this simulation, the model assumes that manufacturers redesign their vehicles to maintain market share and not to maximize profits. Similarly, societal costs may be fairly represented by consumer surplus changes, but it is not possible to estimate the share of the burden that would fall on producers and not be passed on to consumers.
• Airquality benefits. Important co-benefits  from reduced fuel consumption like improved air quality,a reduction in adverse health outcomes and the monetary value of those outcomes are also not modeled.
• Valuationof fuel savings. A central assumption  of the model is that consumers undervalue fuel savings.A key limitation is that there is no data on the extent to which this is true of Canadian consumers. As a result, we use the same assumption as Greene et al. (three years undiscounted) and we undertake a sensitivity analysis by examining the effect of assuming that consumers already fully value fuel savings. To the extent that consumers do undervalue fuel savings, a feebate would be justified on efficiency grounds alone, with carbon mitigation benefits additive.
• Consumerelasticities. These values determine  the extent to which consumers respond to price signals.Greene et al. used short-run elasticities, which are assumed to vary with make/model market shares, but are not estimated by manufacturer or vehicle class. Given the function used, classes with larger market shares have smaller elasticities, whereas models with small market shares have high elasticities (in some cases, unrealistically high). Furthermore, cross-price elasticities are not differentiated by make/model or class. This means that consumers diverted from buying a large SUV are assumed to be equally likely to buy a subcompact as a minivan or mid-size SUV. It also means that cross-price elasticities are higher across the board for classes with large market shares; this issue matters more for Canada, since the market shares are less evenly distributed than in the US. Since there is no comparable data on Canadian elasticities, our approach has been to use the Greene et al. elasticities but to halve them, as a way of approximating long-run responses and to mitigate some of the effects described above. We also undertake a sensitivity analysis by examining the effects of assuming the full elasticities prescribed by Greene et al.
• Staticdesigns. The model uses a database of  vehicles, which is based on 2003 models and assumesno changes in makes and models over the 15-year period. This is unrealistic but neutral in terms of costs. Furthermore the model uses a static technology cost curve that assumes no progress in available fuel economy technologies over the period—this assumption tends to overestimate costs. Finally, the model assumes that all technology investments are used to improve fuel economy and that weight and performance remain constant. (Given past experience, this tends to overestimate the fuel economy effectiveness of the investments.)
• Hybridand diesel technologies are not included.  As discussed in Section 2, hybrids and diesels areexpected to play a significant role in improving fuel economy. Unfortunately the current version of the model lacks the information necessary to include these options. (Our understanding is that Transport Canada is working to add them to the model soon.) As a result, the model overestimates costs and underestimates fuel economy gains.
• Effectson used vehicle markets are not modeled.  As discussed in Section 2, the used vehicle marketrepresents more than half of vehicle sales. If the price of certain models rises because of feebates, it is likely that some of the demand will be filled from the used vehicle market. However, the model assumes that consumers primarily respond by shifting purchases to other new vehicles. As a result, the effectiveness of fuel economy gains is overestimated.
• Thresholdapproach to redesign. The cost curve  approach does not account for economies of scale beyondthe pre-set threshold, as the thresholds are only set to limit the access to technology in a Canadian-only scenario.
• Spillovereffects of unilateral Canadian policy.  The model assumes that unilateral Canadian policieshave no effect on US policy. Thus any spillover effects that might occur are discounted. For example, should Canadian policies increase the potential for voluntary or mandatory fuel economy improvements in the US, the effect of this on the overall size of the market for redesign is not included. (As a result, costs are potentially overestimated and effectiveness is underestimated.)
• Reboundeffect. “Rebound” refers  to the increase in distance travelled that accompaniesreduced driving expenses (in this case, fuel savings). The model assumes a rebound effect of 23 percent—meaning 23 percent of fuel savings are lost to this effect—based on research of past experience in the US. The potential future rebound effect in Canada has not been studied, but 23 percent almost certainly overestimates the effect and therefore underestimates the GHG savings.

SCENARIOS

primary purpose of the modeling is to explore the effects of various feebate options. In this respect, the model is primarily designed to examine the effects of varying rate designs as opposed to implementation options, such as approaches to revenue neutrality, phase-in periods, or annual feebates at vehicle registration. As well, the model is designed to use fuel consumption as the rate basis.   our scenarios are selected to include variations on the following options:  

• Formof the function: Fully linear, $5,000 cap, or zerofeebate within 1.0 litre per 100 km of pivot (deadband)
• Rate:$250, 500, or $1000 per litre per 100 km 
• Numberof classes: single, car and truck, or 11 classes. 

secondary purpose of the modeling is to examine the sensitivity of the results to differences in key assumptions. In this respect, the following alternatives are considered:  

• Elasticities
  • Baseassumption – represents long-run elasticities(i.e. half of the values used by Greene et al. presented above)
    • -5@ market share of 1.5 percent within a class 
    • -2.5@ market share of 10 percent between classes 
    • -0.5for overall sales 
  • Sensitivityanalysis for double these figures (i.e. the fullvalues used by Greene et al.).
• PolicyScope 
  • Baseassumption: feebates are implemented in Canadaonly
  • Sensitivityanalysis for Canada–US policy. 
• ConsumerPerception of Value of Fuel Savings 
  • Baseassumption: three years of undiscounted savings 
  • Sensitivityanalysis for fuel discounted valuation. 
• FuelPrice 
  • Baseassumption: C$0.90 per litre 
  • Sensitivityanalysis at C$1.20 per litre. 

alternative options and assumptions are grouped into 12 scenarios, as shown in Table 4.1.  

RESULTS

Appendix

B contains detailed result for each of the scenarios. Selected results are presented in Table 4.2. Note that Scenario 7 could not be modeled because of problems with the specification of the discontinuity. 4.2: Scenario Results  

Scenario			Total Transfer ($M)	Fuel Economy (l/100km) 2018		Unval. Fuel Savings ($M)	GHG Emission Reduction (Mt)		Societal Cost ($M)*Benefit shown as neg.			Sales (2010)
#	Policy Options	Assumptions	2010	Cars	Trucks	2018	2018	2003-2018	2010	2018	2003-2018	Total Change (#veh.)	Big 3 (%)	Other (%)	Cars (%)	Trucks (%)	Avg. Car Price ($)	Avg. Truck Price ($)
1	Base Case	First 3 yrs valuation Base Elasticities	0	7.1	9.8	0	0	0	0	0	0	0	62	38	56	44	24,600	32,600
2	$250/l/100 km	Canada only First 3 yrs valuation Base Elasticities	290	6.9	9.7	180	1.5	13	(80)	(180)	(540)	(260)	61	39	58	42	24,200	32,500
3	$500/l/100 km		570	6.8	9.5	370	3.0	26	(120)	(310)	(800)	(1,200)	60	40	59	41	23,800	32,400
4	$1000/l/100 km		1,100	6.5	9.2	760	6.2	53	(90)	(460)	(460)	(6,300)	58	42	62	38	23,100	32,200
5	$500/l/100 km	North America	550	6.4	8.8	730	6.0	42	(160)	(650)	(1,340)	(1,800)	60	40	59	41	23,900	32,600
6	$500/l/100km Cap@$5,000	Canada only First 3 yrs valuation Base Elasticities	570	6.8	9.5	370	3.0	26	(120)	(310)	(800)	(1,200)	60	40	59	41	23,800	32,400
8	$500/l/100km Sep. Cars & Trucks		450	6.8	9.5	250	2.1	16	(70)	(220)	(480)	(860)	61	39	56	44	23,800	31,400
9	$500/l/100km 11 Classes		280	6.8	9.6	200	1.6	12	(50)	(180)	(380)	(520)	61	39	56	44	23,900	32,400
10	$500/l/100km	Full Valuation	590	6.6	9.2	0	3.0	26	20	(20)	292	(47))	61	39	58	42	23,700	32,500
11	$500/l/100km	Double Elasticities	560	6.7	9.3	1,200	5.4	49	(540)	(1,070)	(3,540)	(5,700)	58	42	62	38	23,000	32,100
12	$500/l/100km	$1.20/l	580	6.7	9.4	400	2.4	21	(160)	(360)	(990)	(30)	60	40	59	41	23,700	32,400

4.6OBSERVATIONS The scenarios provide a basis for several observations: GeneralObservations. The most significant findingsare: Mostscenarios result in significant fuel savingsand GHG reductions. Mostscenarios produce a net benefit—meaningfrom a societal perspective the benefits are greater than the costs—mostly in the form of unvalued fuel savings. Incontrast with Greene’s US experience,most scenarios produce a more significant shift in sales (although technology still accounts for more than two thirds of the improvement). As a result, the average cost of vehicles is slightly lower and combined with slightly lower sales; total revenues are reduced by approximately $1.5 billion per year (approximately 4 percent). Effectof Rate Change. See Table 4.3.  The key observations are:  Fueleconomy improves and therefore GHG reductionsincrease relatively linearly with an increasing feebate rate. Theshift in sales also increases linearly.  Netbenefits are positive for all rates butlevel off between $500 and $1000. Table 4.3: Effect of Rate Change   Scenario Change in GHG (Mt) Total Societal Benefit ($M) Change in Car Share $250/l/100km -13 540 1.6% $500/l/100km -26 790 3.1% $1000/l/100km -53 460 6.1% Cap& Deadband. The key observations  are:  Acap removes incentives for highly inefficientvehicles to improve, since they just pay a fixed fee. It also diminishes incentives to shift away from those vehicles. A$5000 cap has no significant effect:only one vehicle (Ferrari Enzo) would be above this threshold. Onlyone other (Chevrolet Silverado) wouldbe above $4000. Only15 models out of 800 would be above$3000. Adeadband removes incentives to improvefuel economy for vehicles near the pivot point. Thus,reductions are fewer, and shiftingto more efficient vehicles is distorted. Unfortunately,the model was unable to simulate thiseffect correctly. ConsumerValuation of Fuel Savings. See  Table 4.4. The key observations are:  AtC$0.90 per litre the full discounted valueof reducing fuel consumption by 1.0 per litre per 100 km would be $1700. If we assume that consumers only value three years, or $700, feebates provide a way to compensate for this. Thus, feebates up to $1000 per litre per 100 km should be cost-effective. Results are consistent with this expectation: Thetotal surplus (net benefit) improvesas the feebate rise. Allscenarios up to $1000 have benefitsinstead of costs. Forrates higher than $1000, the costswould outweigh the benefits on the margin. Fullvaluation means that the base case ismore advantageous and that the benefits of feebates are correspondingly reduced. Fuel economy still improves but there is a net cost per tonne for GHGs. Table 4.4: Effect of Consumer Valuation Assumption     Baseline FC in 2018 (l/100km) Average FC in 2018 (l/100km) Change in Consumer Surplus ($M) Unvalued Fuel Savings ($M) Total Societal Benefit ($M) $500/l/100km, 3yr valuation 8.3 7.9 -510 1300 790 $500/l/100km, full valuation 8.1 7.7 -290   -290 Segmentation.See Table 4.5. The key observations are:  Differentiatingpivot points means a lower fee or evena rebate as larger vehicles are assessed only against their cohorts. This discourages shifting to smaller vehicles (less change in market shares) and means less improvement in consumer surplus and fewer GHG reductions. Results are consistent with this expectation. Segmentationof cars and trucks reduces fuel economy,GHG reductions and benefits, while virtually eliminating the shift between cars and trucks (and reducing the shift between manufacturers). Goingto 11 classes has little additionaleffect. Table 4.5: Effect of Segmentation   Scenario Change in GHG (Mt) Total Societal Benefit ($M) Change in Car Share $500/l/100km, 1 pivot -26 790 3.11% $500/l/100km, 2 pivots -16 480 0.02% $500/l/100km, 11 pivots -12 380 0.00% PolicyIntegration with US. See Table  4.6. The key observations are:  Witha unilateral policy, not all makes/modelsmeet the threshold for retooling. Fuel economy is then determined by the average North American willingness to pay (WTP) and a Canadian feebate raises this WTP according to the Canadian market share. WithNorth American implementation, all vehiclesimprove according to the full change in consumer willingness to pay. As a result, GHG reductions are larger and the change in Canadian surplus is larger. An integrated North American feebate doubles the GHG reductions and benefits for the same shift in sales. Table 4.6: Effect of Policy Integration with the U.S.   Scenario Change in GHG (Mt)   Total Societal Benefit ($M) Change in Car Share $500/l/100km, Canada -26 790 2.76% $500/l/100km, North America -42 1,340 2.48% Elasticities.See Table 4.7. The key observations are:  Elasticitiesdetermine the effects on total sales andshift in the fleet mix. Greene’s original elasticities represented short-run responses; to better reflect the available information on long-run responses, we halved them to get our baseline elasticities. As a result, the baseline elasticities yield more conservative estimates of the benefits to consumers, but also predict smaller sales shifts. WithGreene’s original elasticities,consumers are more sensitive to price changes. As a result, the sales mix changes more easily but policies are less costly, since consumers take greater advantage of the option to purchase other vehicle types. This doubles the size of the sales shifts and the GHG reductions. At the same time, the change in surplus is five times greater. Table 4.7: Effect of Elasticity Assumptions   Scenario Change in GHG (Mt) Change in Consumer Surplus ($M) Unvalued Fuel Savings ($M) Total Societal Benefit ($M)   Change in Car Share, 2018 Change in Sales, 2018 $500/l/100km Baseline Elasticities -26 -510 1,300 790 2.8% -0.1% $500/l/100km Double Elasticities -49 -680 4,580 3,900 5.7% -0.4% Priceof Fuel. See Table 4.8. The  key observations are:  Ahigher fuel price means that there ismore incentive for fuel economy present in the base case. Thus, the incremental of the feebate on fuel and GHG savings is reduced (by about 20 percent). Onthe other hand, the unvalued fuel savingsare worth more so the overall benefit is higher and the benefit per tonne is greater (almost $50 per tonne). Theprice of fuel has little impact on thesales mix. Table 4.8: Effect of Fuel Price   Scenario Change in GHG (Mt) Change in Consumer Surplus ($M) Unvalued Fuel Savings ($M) Total Societal Benefit ($M) Change in Car Share, 2018 $500/l/100km Fuel @ $0.90/litre -26 -510 1,300 790 2.8% $500/l/100km Fuel @ $1.20/litre -49 -680 4,580 3,900 2.9% 4.7 IMPLICATIONS Modelresults are subject to the limitations andassumptions described above. The main   value of the model is to assist in understandingthe relationships between inputs and various indicators of environmental effectiveness, economic efficiency, etc. Thus, although the results are considered directionally valid, individual values should be treated with caution. Someimplications are clear and relatively robust.Notwithstanding modeling limitations   and assumptions, there are a number of findingsderived from a logical understanding of the mechanisms at work and, therefore, not likely to be in doubt. These findings are: Feebateswill encourage additional investment infuel-efficiency technology and shift the market towards more fuel-efficient vehicles (trucks to cars, large cars to small cars, more fuel-efficient cars in the same class). Overtime, this will improve the fuel efficiencyof the vehicle stock and will reduce GHG emissions. Theinvestment in fuel-efficiency technologywill raise the cost of individual vehicles and reduce consumer surplus accordingly. Tothe extent that consumers undervalue fuelsavings, feebates will capture savings that would otherwise not have been realized. If the undervaluation is significant, over the life of the vehicle fuel savings are likely to exceed the added cost to vehicles, resulting in a net economic benefit to society. Higherprices will depress vehicles sales.  Theshift towards more fuel-efficient vehicleswill also reduce overall revenues. Ina single-class feebate, GM, Ford, andDCX will lose additional market share and will bear a disproportionate share of the adjustment costs. This could be alleviated by adopting separate classes for trucks and cars, though this would reduce GHG savings and economic benefits. Theextent of the shifts is determined bythe elasticities of demand. If elasticities are greater than expected, the environmental and economic benefits will be greater but so will the adjustment costs. Conversely, if elasticities are less than expected, the environmental and economic benefits will be reduced, as will the burden on manufacturers.

« Previous — Contents — Next »