Office of air quality management

(2). Section 10.4—Potential Costs to Local and State Agencies

408. Comment: Tax revenues in 2016 would be almost $1 billion less in California, with a cumulative loss of more than $3.2 billion from 2009 to 2016 (See Appendix B at 36.) (Comments of the Alliance of Automobile Manufacturers On the Proposed Rulemaking to Adopt Regulations to Control Greenhouse Gas Emissions from Motor Vehicles, page 25)

Sales tax revenues would be about $600 million lower in 2016, with a cumulative reduction of almost $2 billion over the eight-year period from 2009 to 2016. Income tax revenues would be roughly $400 million lower, with a cumulative loss of nearly $1.3 billion. Overall state sales and income tax revenues in 2016 would be almost $1 billion less in California as a result of the Staff Greenhouse Gas Proposal, with a cumulative loss of more than $3.2 billion from 2009 to 2016. (NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, page 35)

Agency Response: Both the August 6, 2004 Initial Statement of Reasons, (section 10.4), and the September 10, 2004 Addendum to the ISOR, (p.25), estimate potential revenue losses to local and state agencies. The relevant tax revenues come from two sources: (1) excise tax and sales tax on gasoline sales; and (2) sales tax on purchase of new vehicles.

Our revised estimate, which appears in the Addendum, projects that, if tax rates remain the same, gasoline excise and sales tax revenues to state and local agencies would decline by $1.3 billion in 2020, compared to a no-regulation scenario. We also estimate that $131 million of this revenue loss would be offset by increased vehicle sales tax revenues, for a net potential revenue loss of approximately $1.1 billion in the year 2020.

However, as we also state in both the ISOR and the Addendum, these potential net losses may be further offset as buyers of regulated vehicles spend some percentage of their increased personal income on goods subject to local sales tax.

409. Comment: Government purchases constitute roughly three percent of passenger car sales. Because we assume that the government has a balanced budget constraint, the increased expenditure on new covered vehicles must be offset by either an increase in taxes or a decrease in other expenditures. For this analysis, we assume that the government pays for its share of increased vehicle prices by increasing taxes to households. The amount of this increase is equal to the total consumer welfare effect multiplied by the proportion of total vehicle purchases by the government (0.03). (NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, attachment B7-6)

Agency Response: Staff disagrees with the commenter’s assumption that the government pays for its share of increased vehicle prices by increasing taxes to households. This assumption tends to exaggerate the adverse impacts of the proposed regulations on the government and households. The government, like households, would experience significant savings from reduced vehicle operating costs. The government would also experience increased tax revenue associated with increased vehicle prices. Reduced vehicle operating costs tend to induce the overall economic performance, as a result increasing the government tax revenues. These savings would significantly offset increased vehicle prices in any year, thus reducing the government need to increase taxes to consumers. Staff believes that savings from vehicle operating costs would more than offset increased vehicle prices to government over time.

410. Comment: These consumer welfare gains due to fuel efficiency are divided among households, businesses and governments using the same approach as described above for the price increases. For households, these effects increase consumer surplus and thus have the opposite sign of the price effects described in the previous section. (NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, attachment B7-7)

Agency Response: ARB Staff agrees with the commenter that a decline in vehicle operating costs would increase consumer surplus. Other purchasers of vehicles such as governments and businesses would also benefit.

(3). Section 10.5—Potential Impact on Individual Consumers

411. Comment: Nothing in the text of AB 1493 permits the Board to ignore any category of costs. The statute’s requirement to consider what is “economical” to the consumer must be understood as a mandate to consider all relevant costs to the consumer, not just the costs of additional hardware. Consumers who buy a vehicle to meet standards that they would not otherwise prefer, incur costs that are real and quantifiable, as demonstrated by the literature reviewed above. In view of the staff report’s failure to consider all the relevant types of costs of the proposed regulation to vehicle purchasers, the Board lacks sufficient analysis, and the public has not been properly advised of the total estimated costs of the proposed rule. On such a record ARB cannot adopt the proposed rule. (Comments of the Alliance of Automobile Manufacturers On the Proposed Rulemaking to Adopt Regulations to Control Greenhouse Gas Emissions from Motor Vehicles, pages 12­13)

Agency Response: Staff disagrees with the comment. In determining what regulations would be economical to the consumer, staff considered all relevant potential costs to the consumer.

Staff did not ignore the possibility that manufacturers could choose to comply with the regulation by compromising vehicle drivability or performance. The ISOR identifies, in detail, a variety of technology-based approaches to compliance which conserve for consumers not only current levels of vehicle performance and drivability, but anticipated levels of vehicle performance and drivability.

While it is possible that automakers might choose to achieve compliance by making drastic pricing adjustments or reducing existing levels of vehicle performance or drivability, such approaches are highly unlikely. It is more likely, in staff’s view, that competitive imperatives will motivate auto manufacturers to achieve compliance by integrating improved technologies, while maintaining or improving upon vehicle performance.

Staff’s estimate of 2009 vehicle attributes incorporates vehicle performance improvements that are expected to occur whether or not the proposed regulation is approved. Hence, compliance need not be achieved at the expense of existing or anticipated performance attributes historically demanded by consumers. Moreover, the regulation is specifically designed to ensure that the full range of vehicle sizes and styles now available will continue to be available to California’s next generation of new-car buyers. Sierra-NERA’s prediction that consumers will incur “real and quantifiable” costs due to compliance-related vehicle attribute changes relies on speculation about both the automakers’ approach to compliance and consumer preferences from the year 2009 forward.

CARBITS consumer response modeling, on the other hand, indicates that from 2009­2013, Californians will buy more, not fewer, new compliant vehicles despite retail price increases. By 2009, consumers may well prefer the attributes of regulated vehicles to those of unregulated vehicles. Such a response could result from a preference for lower vehicle operating costs. Other factors not included in the model, such as heightened awareness of the costs of climate change and dependence on foreign oil, could increase consumer preference for the attributes of compliant vehicles. See also the response to comment 568.

412. Comment: The economic analysis in the staff report ignores opportunity costs, which are important in mainstream economic analysis of measures like that proposed in this rulemaking. These opportunity costs include the value of the foregone opportunity to purchase a vehicle which may be less fuel efficient but has other features that a consumer desires more than enhanced fuel efficiency, such as vehicle performance, safety, capacity, comfort, and aesthetics. (Alliance of Automobile Manufacturers)

Agency Response: See response to comment 411.

413. Comment: The economic analysis in the August 2004 staff report and its supporting materials assume that the major cost that would be incurred by a consumer who purchases a vehicle certified to the proposed CO2 requirements would be the cost of the hardware needed to meet those requirements. The staff ignores opportunity costs, which are important in mainstream economic analysis of measures like that proposed in this rulemaking. The total costs of a rule like that proposed in the staff report include, for example, the value of the foregone opportunity to purchase a vehicle which may be less fuel-efficient but has other features that a consumer desires more than enhanced fuel efficiency. Such features obviously include vehicle performance, safety, capacity, comfort and aesthetics. (Comments of the Alliance of Automobile Manufacturers On the Proposed Rulemaking to Adopt Regulations to Control Greenhouse Gas Emissions from Motor Vehicles, page 8)

Agency Response: The ARB’s staff technical evaluation of technologies and fuels that can reduce climate change emissions in passenger vehicles shows that automotive manufacturers have a number of technological options to meet the requirements of the climate change regulations. Most of these technologies are currently available or would be available in the 2009 timeframe. The application of these technologies is not expected to alter other vehicle features that a consumer desires. Some vehicle models are currently equipped with these technologies. The ARB staff believes that a more widespread use of these technologies would not come at the expense of compromising other vehicle features that a consumer desires such as vehicle performance, safety, capacity, comfort and aesthetics. In other words, the staff expects no loss of value to consumers due to the foregone opportunity. On the contrary, the staff believes that the use of climate change emissions reducing technologies in new vehicles would enhance their marketability, and that these technologies pay for themselves.

In determining what regulations would be economical to the consumer, staff considered all relevant potential costs to the consumer. Staff analyzed at great length the opportunity cost that consumers might incur under the proposed regulation. The findings of that analysis are shown in Tables 12.1-6 and 12.1-7 and discussed on page 179 of the ISOR.

Consumer response modeling results indicate that, in the early years of the regulation, consumers will value the attributes of regulated vehicles as much or more than those of unregulated vehicles. Despite increased prices, new car sales during the period from 2009 to 2013 are expected to be higher in the regulated scenario than in the unregulated (baseline) scenario.

In later years, (2013-2020), as per-vehicle compliance costs climb, the regulation scenario projects slightly slower new-car sales. But the projected decline in 2013-2020 sales is in line with normal consumer response to price increases in the absence of product attribute changes. This implies that, in general, consumers will be as satisfied with the attributes of regulated vehicles as with unregulated alternatives. Therefore the cost of any frustrated demand for higher operating cost vehicles, as evoked by the commenter, is considered negligible.

h. ISOR Section 11—Impacts on Minority and Low Income Communities

(1). Section 11.1—ARB Environmental Justice Policy

414. Comment: We want to recognize that this is only a first step in our commitment for just climate policies, and there is a long way to go. We applaud the efforts of ARB staff and the leadership of the Board and we strongly support the recommendations. (Bahram Fazeli, Communities for a Better Environment)

Agency Response: ARB staff agrees with this comment, and the regulations approved by the Board will achieve the maximum feasible and cost effective reductions. No further response needed.
415. Comment: This regulation is a major step forward for EJ Communities and for California and I hope other States follow. (Dr. Clark, West County Toxics Coalition)

It is important for the Board to continue in its leadership role by looking at this in a very serious manner, adopting this role, working within the communities that would be impacted to find mutual solutions and working with all stakeholders to make sure that we can move forward and gain progress on this. It's very important. (Carlos Porras, private citizen)

Agency Response: Staff agrees with the comments.

(2). Section 11.3—Potential Environmental Impacts

416. Comment: Increases in heat waves would most affect people of color and the urban poor. Increases in air pollution are likely to exacerbate existing health problems and raise health exposure. Unhealthy exposures occur unequally among vulnerable communities and low-income people. Low-income and people of color are disproportionately employed in industries most sensitive to climate change; thus leading us to conclude that the economic impacts are also likely to be quite disparate. (Michel Gelobter, Executive Director of Redefining Progress)

Agency Response: The comment is supportive of the staff analysis and the proposed regulation. No further response needed.

(3). Section 11.4—Potential Economic Impacts

417. Comment: The proposal also provides strong evidence that the regulations would benefit disadvantaged communities and low-income and minority residents of California, and would allow the state to continue to attract businesses to these and other communities throughout the state. As discussed in our July 6, 2004, comments on the draft staff proposal that preceded the ISOR, the other aspects of the economic analysis developed by CARB staff are sound and amply demonstrate that the proposed regulations meet all of the economic stipulations of the authorizing legislation. (John M. DeCicco, Ph.D., and Kate M. Larsen, Environmental Defense; letters of support also received from Natural Resources Defense Council, Bluewater Network, Environment California, Communities for a Better Environment, Union of Concerned Scientists, Sierra Club, Coalition for Clean Air, Conservation Law Foundation, Alliance for a Clean Waterfront, As You Sow, The David Brower Fund, Clean Water Action, Coalition of Concerned National Park Retirees, Community Clean Water Institute, National Parks Conservation Council, Neighborhood Parks Council, Rainforest Action Network, San Francisco Bicycle Coalition, San Francisco Tomorrow, Santa Barbara Channelkeeper, Vote Solar Initiative, Community Action to Fight Asthma)

Agency Response: The comment is supportive of the staff analysis and the proposed regulation. No further response needed.

i.	ISOR Section 12—Other Considerations
(1).	Section 12.1—Consumer Response Effects on Emissions and State
	Economy
(a).	General Effect

418. Comment: Thus, the design changes made in response to CARB's proposed GHG standards will be transparent to consumers, in terms of vehicle functionality, and the price impacts will most likely be unobservable. In other words, the technology improvements induced by the GHG standards will play out very similarly to what has occurred in response to past air pollution emissions control standards, and we can expect to see the benefits of reduced GHG emissions even as cars and light trucks continue to improve in other ways, without any appreciable impacts on either consumer acceptance or overall sales. (John M. DeCicco, Ph.D., and Kate M. Larsen, Environmental Defense; letters of support also received from Natural Resources Defense Council, Bluewater Network, Environment California, Communities for a Better Environment, Union of Concerned Scientists, Sierra Club, Coalition for Clean Air, Conservation Law Foundation, Alliance for a Clean Waterfront, As You Sow, The David Brower Fund, Clean Water Action, Coalition of Concerned National Park Retirees, Community Clean Water Institute, National Parks Conservation Council, Neighborhood Parks Council, Rainforest Action Network, San Francisco Bicycle Coalition, San Francisco Tomorrow, Santa Barbara Channelkeeper, Vote Solar Initiative, Community Action to Fight Asthma)

Agency Response: Staff agrees with the comment.

419. Comment: My concern with this regulation is the direct and harmful effect it will have on my business and the businesses of other automobile dealers in California. Based on the cost estimates in the ARB staff’s latest review, it is clear that the proposed regulations are very expensive. By model year 2012, the estimated costs of the first phase of the regulation are between $280 per vehicle (larger light-duty trucks) and $370 per vehicle (passenger cars and small light-duty trucks). The estimated costs then skyrocket and reach levels of $1030 per vehicle (larger light-duty trucks) and $1060 per vehicle (passenger cars and smaller light-duty trucks) by model year 2016.

The estimated costs by the staff represent huge cost increases. The cost increases will result in much higher vehicle prices for my customers. Cost increases of this magnitude will immediately lead to lower vehicle sales. Automobile sales are highly dependent on price—if the price of automobiles increases—the number of automobiles will decrease. This is simply a result of the decreased market for a more expensive product. This result is not idle speculation, it is based on my 30 years of experience selling vehicles. Whenever the cost of automobiles has increased faster than the income available to purchase vehicles, the result is less vehicles sold.

Reduced vehicle sales will directly reduce the income from my dealership and its economic value as a continuing enterprise. I was therefore startled to see that the staff’s assessment of the impact on automobile dealerships assumes that there will be no net impact. The staff claims that the “dealers’ loss of sales volume was roughly compensated by the increase in vehicle prices”. This assumption is directly contrary to my 30 years of experience selling automobiles.

The higher vehicle prices will result in lower sales, and the overall effect will be a drop in total revenue. Further, because the costs of all vehicles increase, people will not be able to purchase either the same type of vehicle they previously purchased or will not be able to purchase the same trim level that they previously purchased. My customers will have to settle for a vehicle with less options, a smaller engine, or a smaller vehicle. All of these changes will result in lower net revenue for my dealership.

I do not usually comment on emission regulations that are issued by the Air Resources Board. However, because the costs estimated by the ARB staff are so high, I feel compelled to urge you to consider the actual economic impact on independent California automobile dealers. The only prudent thing to do in the face of the extremely high costs estimated by the ARB staff is to not implement such extreme measures. I cannot believe that the people of California envisioned such costly and extreme measures would ever be promulgated. (John W. Gardner, Central Valley Automotive; similar letters received from Sturgeon and Beck Inc. Buick Pontiac GMC, Hansel Toyota, Galpin Motors, 3 Way Chevrolet, George Chevrolet, Toyota of San Bernardino, Turlock Auto Plaza, Hoblit Motors, Courtesy Automotive Center, Steves Chevrolet-Buick, Fireside Dodge, Hemborg Ford, Pearson Ford, Bob Williams Chevrolet Geo , Hendrick’s Hallowell Chevrolet, Surroz Motors Inc.).

Agency Response: Staff disagrees with the comment. Vehicles meeting the greenhouse gas standards will have lower operating costs, which offsets the up-front purchase price increase. Staff used the CARBITS model to quantify the consumer response to this combination of price increase and lower operating cost. In 2009-2012, during the phase-in of near-term technologies to reduce climate change emissions, the modeling projected that sales of new vehicles increase due to the regulation. This increase comes about because consumers are willing to pay for the lower operating cost. During the phase-in of the mid-term technology, sales growth of new vehicles drops modestly due to the regulation, but the lower operating cost is a mitigating factor that blunts the effect of the price increase. Furthermore, manufacturers have historically used marketing strategies to soften the impacts of any price increase and are likely to do so for this regulation to maintain market share and sales. Their efforts will mitigate any potential impact on their dealers.

(b). CARBITS Modeling Issues

420. Comment: If the model were to receive a review that would be typical in deciding whether to use the model to evaluate the potential impacts of a regulation, the following additional information would be required:

1. 
   The survey data and computer programs that were used to derive the data on which basis the Sheng model was estimated;
   
2. 
   The computer program(s) used to estimate the Sheng model;
   
3. 
   The computer program(s) used to simulate household demographic changes;
   
4. 
   The computer program(s) used to run the CARBITS micro simulations; and
   
5. 
   Any other computer program(s) or data that are necessary to fully replicate CARBITS. (NERA Economic Consulting, Reviews of Studies Evaluating the Impacts of Motor Vehicle Greenhouse Gas Emissions Regulations in California, page 8).
   

ARB staff apparently has justified not providing the full set of CARBITS information on the basis that CARBITS has already been peer reviewed. The fact that two published papers cited by ARB Staff have been peer reviewed, however, does not mean that CARBITS itself has been subject to peer review or that peer review would imply that the underlying data and methodologies are valid. (NERA Economic Consulting, Reviews of Studies Evaluating the Impacts of Motor Vehicle Greenhouse Gas Emissions Regulations in California, page 8).

Agency Response: Staff disagrees with the comment. ARB provided NERA with all the information about CARBITS in our possession, in July and August 2004. ARB does not have the information listed above nor did we try to gather it at any time while UC Davis ITS was developing CARBITS.

ARB has posted on its web site, for the public including NERA to view, these peer-reviewed articles:

• 
  • David Brownstone, David S. Bunch, Kenneth Train, Joint mixed logit models of stated and revealed preferences for alternative-fuel vehicles, Transportation Research Part B, volume 34, pages 315-338.
  
  0. 
     • David Brownstone, David S. Bunch, Thomas F. Golob, and Weiping Ren, A Transactions Choice Model for Forecasting Demand for Alternative-Fuel Vehicles, Research in Transportation Economics, Volume 4, pages 87-129, 1996.
     
  1. 
     ARB has provided NERA with this peer-reviewed article:
     
• 
  • David S. Bunch, David Brownstone, and Thomas F. Golob, A Dynamic Forecasting System for Vehicle Markets With Clean-Fuel Vehicles, Seventh World Conference on Transport Research, 1995.
  

ARB also provided two dissertations of graduate students who performed some of the original work. Each student had a committee of professors who rigorously reviewed the dissertations.

• 
  • Hongyan Sheng, A Dynamic Household Alternative-fuel Vehicle Demand Model Using Stated and Revealed Transaction Information, 1999.
  
• 
  • Camilla Kazimi, A Microsimulation Model for Evaluating the Environmental Impact of Alternative-Fuel Vehicles,1995.
  

ARB’s description of peer-reviewed papers produced by University of California researchers relates to the fact that multiple researchers have done a substantial amount of work in this area covering a multi-year period. Clearly, the methodology has been thoroughly and rigorously peer-reviewed. CARBITS development rested on both the substance of the previous work and on the experience and insights gained while working in this area.

During the development of CARBITS, ITS would send a preliminary version of CARBITS to ARB. ARB staff would test the model and have issues with it. ARB would report the issues and ITS would make improvements. ARB and ITS continued this iterative process until ARB had a version of CARBITS that worked properly. ARB compared CARBITS results with EMFAC and with vehicle data obtained from CALTRANS. ARB insisted on consistency with sales counts, vehicle population, and vehicle age distribution. ARB also insisted on internal consistency of consumer response to changes in vehicle attributes. ARB is satisfied that CARBITS works as it should. ARB staff did not consider it necessary to obtain the FORTRAN source code and look through it searching for bugs. This would be equivalent to asking Microsoft for the source code underlying the software used to compose this document.

ITS calibrated CARBITS, at ARB’s request, to be consistent with EMFAC. (EMFAC is ARB’s model for mobile source emissions, approved by U.S. EPA.) That is, the household weights were set such that for the base year, 1995, the vehicle age distribution of the CARBITS baseline scenario matched that of EMFAC for 1995. CARBITS output for 2000, after five simulated years of household vehicle transactions, still shows excellent agreement with EMFAC, in the vehicle age distribution. Because of the good agreement between the two models in spite of their very dissimilar methods of calculating vehicle populations, we have confidence that CARBITS is a useful tool for the Staff Report’s supplemental analysis.

421. Comment: The Sheng dissertation model uses a multinomial logit (MNL) specification for the revealed preference (RP) portion of the model. The MNL specification imposes what is known as the “independence of irrelevant alternatives” (IIA) assumption. This assumption substantially lessens the computational complexity and burden of estimating the vehicle choice model. However, as is well established in the economics literature, the IIA assumption also imposes restrictions on households’ patterns of substitution between choices.

Under the IIA assumption, for example, the cross-price elasticities of demand with respect to a particular product’s price are forced to be equal to each other. In the current context, IIA would force the cross price elasticity of minivans with respect to sports cars to be equal to the cross price elasticity of luxury cars with respect to sports cars. These restrictions imposed by IIA are often unrealistic – the cross price elasticity between luxury cars and sports cars might be expected to be larger than the cross price elasticity between minivans and sports cars as purchasers of sports cars would view luxury cars as closer substitutes for sports cars than minivans.

The Sheng model uses an MNL specification and thus it imposes the IIA assumption. However, the Sheng dissertation does not indicate that any test of the validity of the IIA assumption was conducted. This raises the question of whether the Sheng model has incorrectly imposed the IIA assumption and its restrictive substitution patterns. If IIA has been incorrectly imposed, the model (and thus CARBITS) would be biased and unreliable.

Brownstone, et al. estimates a vehicle choice model using data from the same underlying survey as that used in the Sheng dissertation. They determine that the MNL model with its IIA assumption is appropriate in their context. However, the Brownstone, et al. model differs in a significant respect from the Sheng model. The Brownstone, et al. model addresses a household’s choice among vehicles, conditional on having purchased a vehicle. The Sheng model, in contrast, addresses both the household’s decision as to whether to purchase a vehicle at all as well as the choice among vehicles for those households who choose to purchase. In the context of the Sheng model, IIA is imposed over all choices, including the “no purchase” option. Thus, even if the IIA assumption is valid within the vehicle choice “branch” of the decision tree (as Brownstone, et al. appears to show), there is no reason to believe that IIA applies over the whole decision tree. (NERA Economic Consulting, Reviews of Studies Evaluating the Impacts of Motor Vehicle Greenhouse Gas Emissions Regulations in California, pages 10-11).

Agency Response: Staff disagrees with the comment. The IIA issue is primarily relevant for aggregate level models estimated on aggregate level data. Specifically, when using disaggregate data it is possible to include many economic and demographic factors in the utility function that drive household decision-making. Aggregate data, by its nature, does not capture detailed effects that distinguish one household from another.

For an aggregate level model, as NERA points out, it would indeed be unrealistic to force the cross price elasticity of minivans with respect to sports cars to be equal to the cross price elasticity of luxury cars with respect to sports cars. On the other hand, in a disaggregate model that takes household characteristics into account, the choice of vehicle depends on the size of the household, the age of the drivers, and household income. By taking these economic and demographic factors into account, each household makes a decision that makes sense for itself.

In other words, MNL can be a reasonable approximation for individual level behavior when a more complex utility function is estimated. CARBITS is a household-level model using disaggregate data, with many household-level factors in the MNL utility function. Therefore the IIA issue associated with aggregate level models is insignificant. NERA even acknowledges that IIA is appropriate for the Brownstone et al. model, which is very similar to the Sheng model underlying CARBITS.

422. Comment: The TSD states that the Sheng model was estimated on a sample of households, which were selected in part on the basis of having “three or fewer vehicles after transacting”. Under this definition, the sample used to estimate the Sheng model was selected in part on the outcome being studied.

For example, consider two households that are otherwise identical in terms of their observed characteristics. Both households have three vehicles during the first wave of the survey. Between waves, one household chooses to add a fourth vehicle, while the second household undertakes no transactions. The first household would be excluded from the estimation sample because it would have four vehicles after the transaction. The second household would be included because it would have only three vehicles. Thus, the only reason the second household appears in the sample while the first does not is the difference in respective outcomes of the two households’ purchase decisions.

This type of endogenous sample selection leads to bias in the estimated coefficients if uncorrected. (NERA Economic Consulting, Reviews of Studies Evaluating the Impacts of Motor Vehicle Greenhouse Gas Emissions Regulations in California, pages 11-12).

Agency Response: Staff disagrees with the comment. Of the 749 households with valid information on replacement or addition, only 84 (11%) had four or more vehicles. This sub-sample size is statistically too small given the number of estimated parameters and decision points for an accurate estimation of transitions for a four-vehicle decision tree that models such transactions. Likewise, whatever bias results from leaving out the 84 households would be small. Therefore the exclusion has no meaningful impact on the results."

423. Comment: The TSD describes the data on which CARBITS is based. These data sets are seven to ten years old, in which period much has changed in both the Californian and national vehicle markets. For example, CARBITS uses the National Personal Transportation Survey data to calibrate benchmarks of the overall transactions and sales rates, and sales rates as a function of vintage. Insofar as these parameters have changed over the past decade, CARBITS simulations will not be a good description of the current vehicle market. (NERA Economic Consulting, Reviews of Studies Evaluating the Impacts of Motor Vehicle Greenhouse Gas Emissions Regulations in California, page 12).

Agency Response: The CARBITS approach is based on theoretical principles that are widely accepted in the travel demand field. ARB wanted to take advantage of the previous work done by ITS. Model development takes a long time. Data collection is expensive and slow. Model construction and calibration and peer review takes time. The effort to make CARBITS useful to ARB took another two years.

ITS calibrated CARBITS, to be consistent with EMFAC. CARBITS output for 2000, after five simulated years of household vehicle transactions, still shows excellent agreement with EMFAC in the vehicle age distribution. This shows that despite using the seven to ten year old data the model correlates well with current conditions. Because of the good agreement between the two models in spite of their very dissimilar methods of calculating vehicle populations, we have confidence that CARBITS is a useful tool for the Staff Report’s supplemental analysis.

424. Comment: An important consequence of using these outdated data, as well as the weighting procedure used to incorporate them into CARBITS, is that CARBITS contains too many cars relative to other vehicle groups, as compared with more recent data. As Table 6.1-4 in the August Staff Report shows, 53 percent of vehicles sold by the largest car manufacturers in California in 2002 were in classes PC and LDT1. However, in CARBITS baseline output, the corresponding figure is 83 percent. CARBITS therefore misrepresent the structure of the vehicle market. Insofar as owners of vehicles in different categories have different preferences and characteristics, consumer responses forecast by CARBITS will be similarly inaccurate. (NERA Economic Consulting, Reviews of Studies Evaluating the Impacts of Motor Vehicle Greenhouse Gas Emissions Regulations in California, pages 12-13).

Agency Response: Staff disagrees with the comment. The distribution of market shares is due to the weighting procedure, not to the age of the data. CARBITS calculates vehicle stock, including new vehicle sales, for 14 vehicle types. Of these, seven correspond to EMFAC class Light Duty Auto, also known as PC (for “Passenger Car”). The other seven classes consist of various pickups, vans, and SUVs. There is no obvious way to translate from these vehicle classes to EMFAC classes LDT1, LDT2 and MDV. ARB and ITS chose a weighting procedure to allow CARBITS output to be consistent with EMFAC output. Thus, both CARBITS and EMFAC show an 83% to 17% distribution of PC/LDT1 LDT2/MDV. In the 2002 CalTrans Travel Survey of households, the share of 2002 household vehicles that are cars is 63%, and this figure does not include any light-duty trucks. In any case, as previously noted, CARBITS performed well in validation tests based on an “apples to apples” comparison using available household survey data.

425. Comment: The Sheng model includes variables taking account of the interaction between fuel operating cost and transaction-type, but these were eliminated from CARBITS. In other words, the coefficients on these fuel operating cost variables were set to zero in CARBITS. This was done based on “expert judgment” that their inclusion constituted a specification error. Dropping the variables that indicate a specification problem without re-estimating the model does not solve the specification problem; it simply hides the fact that a specification problem exists. Thus, the specification problem to which the TSD alludes is still present in the vehicle choice model in CARBITS. (NERA Economic Consulting, Reviews of Studies Evaluating the Impacts of Motor Vehicle Greenhouse Gas Emissions Regulations in California, page 13).

Agency Response: Staff disagrees with the comment. The modifications to the Sheng model were made on the basis of knowledge and insight obtained through intensive calibration, testing, and evaluation efforts. The implications of the two original coefficients were rather subtle. However, it became clear that finding a reasonable way to drop the coefficients would lead to an improvement based on theoretical considerations. It was also clear that the recalibration of the model compensated for the excluded effects. Continued testing and calibration of the model demonstrated that its behavior was an improvement over the previous version.

426. Comment: The second exception to CARBITS’ use of the Sheng RP coefficient estimates is that CARBITS employs the “SP,” or stated preference, coefficient estimate for the tailpipe emissions variable rather than the RP coefficient because the RP coefficient estimate was positive rather than negative. There are several methodological problems with this procedure. First, the RP coefficient estimate is the best estimate of the effect of the emissions variable in the RP model; no econometric justification exists to change it. Second, as with the fuel cost interaction variable coefficients, if it was determined that the RP emissions coefficient should be constrained to a certain value, the Sheng model should have been re-estimated with this constraint imposed so that the rest of the coefficients in the model could be re-estimated optimally. Because this was not done, the coefficient estimates as used in CARBITS are likely to be biased. (NERA Economic Consulting, Reviews of Studies Evaluating the Impacts of Motor Vehicle Greenhouse Gas Emissions Regulations in California, pages 14-15).

Agency Response: For the consumer response scenarios presented in the ISOR and the Addendum, ARB staff changed only the vehicle price and fuel economy. ARB staff did not alter the emissions variable. Whatever effect the emissions variable may have, it is the same effect for the regulation scenario as it is for the baseline scenario. Thus, in terms of the scenario analysis, the issue is moot. As previously noted, CARBITS produced results that matched historical patterns up through 2000 and beyond.

427. Comment: The SP coefficient estimate is based on respondents’ answers to hypothetical survey questions, whereas the RP coefficient is based on the respondents’ actual transactions. SP survey responses are well known to be subject to bias. The problem arises because surveys ask respondents hypothetical questions. Respondents are not forced to make any real economic decisions as a consequence of their answers. Thus, the survey process cannot be said to require them to take the exercise as seriously as they would take an actual purchase opportunity where they would be paying many thousands of dollars. In addition, respondents may have incentives to give “strategic” answers designed to support a policy outcome they think is “right” rather than answers that describe their true economic preferences for the good in question. (NERA Economic Consulting, Reviews of Studies Evaluating the Impacts of Motor Vehicle Greenhouse Gas Emissions Regulations in California, page 15).

Agency Response: Stated preference approaches provide a valuable source of information for developing models used in policy analysis. The questionnaire methodologies used in the survey take the concerns stated above into account. In addition, there are major problems with using revealed preference data alone. The modeling project mitigated these problems by using conjoint data – combining revealed preference data with stated preference data – as mentioned on page 36 of the Technical Support Document on Other Considerations. Prof. Dan McFadden (the 2000 Nobel Prize winner in Economics, for his development of theory and methods for analyzing discrete choice) has stated that carefully collected conjoint analysis data are on the whole measuring the same preferences as revealed preference data. Specifically he states (Daniel McFadden, Disaggregate Behavioral Travel Demand’s RUM Side, July 2000, page 25):

A second innovation in data has had a major impact on travel demand analysis, and probably receives more attention than any other topic in travel demand research. This is the use of stated preference (SP) data, a shorthand for a variety of data that can be collected from individuals by offering them hypothetical choice tasks, eliciting attitudes and perceptions, and collecting subjective reports on preferences. Most of these variables and the methods used to measure them come from psychology via market research. In particular, conjoint analysis has proven that it can give a much more rounded view of the preferences of an individual than the one-dimension picture provided by revealed preference data.

428. Comment: CARBITS does not include any modeling of the supply-side of the vehicle market. Yet, understanding the supply-side is crucial to understanding the extent to which the cost changes mandated by the proposed regulation would be passed through to purchasers of new vehicles. It is well established in the economics literature that the amount of cost pass-through depends on the shape of the demand curve, the form of competition between manufacturers, and the degree to which the cost increase falls on each manufacturer. Instead of appropriately modeling the supply-side in addition to the demand-side, which would allow a determination of the extent of cost pass-through, CARBITS simply assumes that pass-through would be 100 percent.

Another important supply-side consideration is the extent to which used vehicle prices would increase in response to an increase in new vehicle prices. ARB Staff’s failure to model the supply-side of the market and to incorporate the supply-side into the simulations is tantamount to assuming that vehicle prices (both new and used) are exogenous. This assumption is not supported by the economics literature generally and is rejected specifically in studies of the automobile industry. (NERA Economic Consulting, Reviews of Studies Evaluating the Impacts of Motor Vehicle Greenhouse Gas Emissions Regulations in California, page 16).

Agency Response: Models used for policy analysis in the public sector and market analysis in the private sector almost always focus on the demand side rather than on the supply side. ARB’s objective was to look at consumer response. For that, it is appropriate to consider the demand side alone. CARBITS provides the right kind of information for its role in the Staff Report’s supplemental analysis.

It is clear that ARB staff was making a conservative assumption in using a 100 percent pass-through. It was a perfectly sensible approach to set this value to 100 percent as part of a long list of assumptions required to provide a basic framework for the analysis. The cost-effectiveness calculations show that the vehicles affected by the greenhouse gas regulation are economical to the owner, even assuming that all of the cost increase gets passed through as a price increase. The vehicles pass the test with this assumption in place. If it turns out that, because of supply side considerations, the price increase is somewhat less than the cost increase, then the consumer is even better off.

429. Comment: We have focused on the patterns of scrappage of vehicles implied by CARBITS, since scrappage results figure prominently in the ARB Staff’s use of CARBITS. Scrappage patterns are a key determinant of the age structure of the vehicle fleet population, which in turn is a key influence on the level of emissions from vehicles. The reasonableness of the CARBITS model’s predictions of scrappage effects is therefore clearly crucial to its usefulness for policy evaluation. (NERA Economic Consulting, Reviews of Studies Evaluating the Impacts of Motor Vehicle Greenhouse Gas Emissions Regulations in California, page 17).

Given the inconsistency between the CARBITS scrappage rates and the actual historical scrappage rates, the reliability of the CARBITS scrappage rates is called into question. Scrappage patterns are a key determinant of the age structure of the vehicle fleet population, which in turn is a key influence on the level of emissions from vehicles. Because the scrappage rates produced by CARBITS were unreliable, the conclusions of the ARB Staff Report would have no reliable basis. (NERA Economic Consulting, Reviews of Studies Evaluating the Impacts of Motor Vehicle Greenhouse Gas Emissions Regulations in California, page 23).

Agency Response: Staff disagrees with the comment. It is in fact the age distribution of the vehicle population, not scrappage rates, that figures prominently in the ARB’s use of CARBITS. The age distribution is important to ARB, because older vehicles emit more criteria pollutant emissions. The age distributions produced by CARBITS are both sensible and reasonable. The baseline age distribution agrees well with the age distribution of vehicles in EMFAC. The difference in CARBITS age distribution caused by a regulation has qualitatively the same shape as the difference in age distribution resulting from the NERA/Sierra New Vehicle Market Model.

430. Comment: ARB staff has understated the cost of compliance with the proposed standards by ignoring the average 8% sales tax that applies in California. (Declaration of Thomas C. Austin, page 3)

Agency Response: The appropriateness of including sales tax in the ARB compliance cost estimates is addressed in comment 243. The treatment of sales tax with respect to net consumer benefit is addressed in the response to comment 407. This response addresses the relevance of sales tax to the CARBITS modeling.

The prices in the CARBITS vehicle attribute database do not include sales tax. Consumers make comparisons between vehicles based on price, not price plus sales tax. Thus CARBITS, as a consumer response model, uses price excluding sales tax. CARBITS was calibrated using baseline prices that do not include sales tax. Therefore it would be inappropriate to include sales tax in the regulation scenario.

If the vehicle prices used in the E-DRAM scenario included sales tax, the impact would not be significantly different than the E-DRAM results reported in Revised Tables 10.2-4 and 10.2-5 of the Addendum.

(c). NERA Consumer Response Model

431. Comment: NERA has developed a New Vehicle Market Model to determine the effects of the Staff Greenhouse Gas Proposal on the vehicle market in California and the rest of the United States. The model uses a nested logit demand framework based on historical transaction price, sales data, and vehicle characteristics for both regions. (NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, attachment B-1.1)

Logit discrete choice analysis provides a method for predicting consumer choices, and therefore demand, based on previously observed consumer behavior and other assumptions about demand. The most basic logit framework, also referred to as the “simple” logit framework, groups all product alternatives together and therefore allows only very limited patterns of own-price and cross-price elasticity between different alternatives. This limitation is often referred to ask the “Independence of Irrelevant Alternatives” (“IIA”) problem. The nested logit framework builds on this simple framework, but provides for a much richer pattern of cross-elasticity between different alternatives through the nesting structure. (NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, attachment B-1.1)

We assume each vehicle’s alternative-specific parameter depends upon the vehicle’s type and attributes according to the following model:

α= Xβ+ Dnest γnest+ δyear Dyear + φmake Dmake + ε,

where

α is the alternative-specific coefficient,

X are vehicle characteristics,

Dnest are dummy variables corresponding to vehicle nests,

Dyear are dummy variables corresponding to vehicle model years,

Dmake are dummy variables corresponding to the vehicle make (manufacturer)

ε is an error term capturing unobserved characteristics, and

β, γnest, and φmake are estimated parameters

(NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, attachment B1-7-8)

We use national and California-specific sales data from JD Power and Associates to determine the market share for each vehicle model, aggregated across trim levels, for the years 2001-2003. (NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, attachment B1-8).

We use data on transaction prices for each model from JD Power and Associates for both California and the rest of the United States. (NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, attachment B1-9).

We use data from EPA’s Fuel Economy statistics to reflect the fuel efficiency of each vehicle model. (NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, attachment B1-9).

We rely on data from Ward’s for information about other vehicle attributes, including engine size, number of cylinders, curb or test weight, horsepower, length, and height. (NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, attachment B1-9).

Consistent with various literature sources, we assume an aggregate elasticity for the new vehicle market of –1.0. We set the own-price elasticity of the “normalized” vehicle model to be –4.0, which is consistent with various other literature estimates of individual model own-price elasticities. (NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, attachment B1-10).

The nesting parameters for nested logit models represent the similarity between choices for vehicles falling within the same nest. For the “Buy” nest we use a nesting parameter equal to 0.9, for vehicle types we use a nesting parameter equal to 0.6 and for vehicle classes we use nesting parameters equal to 0.3. (NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, attachment B1-10).

In general, the model uses vehicle prices and assumptions about consumer response to changes in those prices to calculate marginal costs that are consistent with both profit-maximizing behavior and the observed market share. Marginal costs for each vehicle are calculated based on each vehicle’s calculated elasticity. (NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, attachment B1-10).

Agency Response: The approach taken by NERA is to estimate an aggregate level nested logit model that focuses solely on new vehicle sales, totally ignoring the used vehicle market. Such models are not typically used in policy analysis involving long time horizons. They do not include enough detail on fundamental behavior and preferences to reasonably capture potential dynamic effects. They completely ignore important effects associated with used vehicle holdings that occurred as a result of decision making in previous periods. Moreover, all of these decisions will be affected by decisions they have made in previous periods; these previous decisions are missing in NERA’s model. In addition, it is obvious that other important effects related to demographic trends, changes in household structure, etc., are not captured by this approach.

The NERA New Vehicle Market Model differs from CARBITS in fundamental ways:

• 
  • The NERA model uses aggregate sales data, whereas CARBITS is a household-level microsimulation. That is, CARBITS simulates consumer decision-making at the household level, where vehicle purchase decisions actually get made. The NERA model uses sales totals that aggregate away any information about individual decisions.
  
• 
  • The NERA model is completely devoid of any demographic information. CARBITS simulates the behavior of more than 40,000 households, modeling the number of individuals, their ages, and household income. CARBITS simulates births, deaths, marriages, divorces, and grown children leaving home from 1995 to 2020.
  

Furthermore, NERA has not indicated whether the New Vehicle Market Model has been peer-reviewed at any level, nor whether its baseline outputs agree with any dataset or output from any other model.

In conclusion, ARB continues to regard CARBITS as a wholly appropriate model of consumer response for the Staff Report’s supplemental analysis.

432. Comment: In 2020, new vehicle sales of PC/LDT1’s and LDT2’s combined are about 176,000 lower as a result of the Staff Greenhouse Gas Proposal. In contrast, the number of vehicles in the fleet with model years before the regulations would take effect (i.e., pre-2009 model year vehicles) is more than 1 million greater in 2020 as a result of the Staff Greenhouse Gas Proposal. (NERA Economic Consulting and Sierra Research, Environmental and Economic Impacts of the ARB Staff Proposal to Control Greenhouse Gas Emissions from Motor Vehicles, page ES – 3)

Agency Response: These results rest on these considerations, developed in detail in the appendices submitted by NERA-Sierra:

• 
  • The costs of compliance are about three times as high as ARB estimates
  
• 
  • The fuel savings are about a third a much as ARB estimates
  
• 
  • The rebound effect is about six times as high as ARB estimates
  
• 
  • NERA-Sierra use an aggregate-level model to calculate the fleet turnover effect.
  

Other responses in this document (see section III.A.2.c(4)) explain that the cost increases and rebound effect assumed by NERA-Sierra are significantly overstated. Likewise, the operating cost savings assumed by NERA-Sierra are understated (see section III.A.2.c(5)).

NERA-Sierra developed its own vehicle holdings model to calculate fleet turnover effect. The model uses market-level aggregate data to estimate the coefficients used in the model. By contrast, ARB uses CARBITS, a household-level transactions model. Modeling at the level of households is more realistic than modeling with aggregate data. Also, a transactions model is better than a holdings model for measuring consumer response to changes in vehicle attributes. Thus, the basic conclusion of the AAM/NERA/Sierra documents – that the regulation results in harm to the California economy and environment – does not hold.

The analysis upon which the regulation is based as presented and documented in the Staff Report was subjected to external peer review by UC scientists. To our knowledge the NERA model and its results were not.

433. Comment: The estimated number of reduced new vehicle sales in 2020 ranges from about 53,000 to more than 300,000. The estimated number of increased pre-2009 vehicles in 2020 ranges from about 64,000 motor vehicles to more than 1 million motor vehicles.

Download 2.03 Mb.

Share with your friends: