National Academies Press: OpenBook
« Previous: 11 Manufacturing Considerations
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 327
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 328
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 329
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 330
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 331
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 332
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 333
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 334
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 335
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 336
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 337
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 338
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 339
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 340
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 341
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 342
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 343
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 344
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 345
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 346
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 347
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 348
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 349
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 350
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 351
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 352
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 353
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 354
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 355
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 356
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 357
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 358
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 359
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 360
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 361
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 362
Suggested Citation:"12 Costs and Benefits." National Academies of Sciences, Engineering, and Medicine. 2019. Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report. Washington, DC: The National Academies Press. doi: 10.17226/25542.
×
Page 363

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

12 Costs and Benefits 12.1 ESTIMATES OF COSTS AND BENEFITS 12.1.1 Introduction This chapter examines the costs and benefits associated with fuel-saving technology. The chapter begins with a review of the benefits and costs of meeting the Phase II standards as calculated by the Environmental Protection Agency (EPA) and the National Highway Traffic Safety Administration (NHTSA). The chapter uses these agencies’ analysis as a base for examining increases in standards beyond the Phase II rule. The analysis of this chapter then focuses on the marginal cost of increases in the fuel efficiency standards beyond those of Phase II regulations. The approach taken in this chapter does not look at every possible technology application, but rather focuses on technologies central to making advances in fuel efficiency beyond where the industry is likely to be in 2027 when the Phase II regulations have been fully implemented. Following the overall analysis of costs and benefits, the chapter turns to a detailed examination of issues involved in the measurement of social benefits and in the projection of future technology costs. The factors examined with respect to the projection of costs include learning that comes with time and with increased production as well as those technology costs incurred in addition to direct manufacturing expenses such as research and development (R&D). 12.1.2 Agency Estimates of Costs and Benefits NHTSA and EPA estimate total costs and benefits of several scenarios of increased fuel efficiency standards in the Regulatory Impact Analysis (RIA) (NHTSA and EPA, 2016). These estimates provide a starting point for analyzing the costs and benefits of improving fuel consumption in medium- and heavy-duty trucks. For example, NHTSA and EPA’s Phase II rule requires a 2.5 percent reduction in fuel consumption per year in the new truck fleet. In addition, NHTSA and EPA model the effects of a 2 percent, a 3.5 percent, and a 4 percent per year alternative. Table 12-1 below is from the Regulatory Impact Analysis and presents the agencies’ summary estimates of costs and benefits of the Phase II rule— a 2.5 percent annual reduction in fuel consumption. Prepublication Copy – Subject to Further Editorial Correction 12-1

TABLE 12-1 Estimates of Lifetime Cost and Benefits of Rule Requiring 2.5 Percent Per Year Reduction in Fuel Consumption for Model Years 2018-2029 (billion $2013) Discount rate 3% 7% Private costs 25.4 17 Technology and indirect cost, normal 23.7 16.1 Profit on additional investment Additional routine maintenance 1.7 .9 Private benefits 157.5 84.7 Fuel savings 149.1 79.7 Savings from less frequent refueling 3.0 1.6 Benefits from additional vehicle use 5.4 3.4 Total net private benefits 132 67.7 Social benefits 64.4 49.1 Reduced climate change from greenhouse gas (GHG) emissions 33 33 Reduced health damages from reduced non-GHG emissions 27.2 14.5 Increased energy security 7.3 3.9 Congestion, crashes, fatalities, and noise from increased vehicle use -3.1 -1.8 Total net benefits (private + social) 196.4 117 NOTE: The baseline used in the above table assumes Phase I standards are met and some additional technology is adopted in the fleet if it can pay back the capital through fuel savings in 6 months or less. In all cases examined by NHTSA and EPA, a 3 percent discount rate was used to value the benefits of reduced GHG emissions. SOURCE: NHTSA and EPA (2016, Table 8-30, p. 8-73). The improvement in fuel efficiency and the reduction in the greenhouse gas (GHG) and other emissions resulting from increased standards for medium- and heavy-duty vehicles (MHDVs) and engines provides benefits both to those who own and operate the vehicles and also to others. The benefit to owners or operators of the vehicles takes the form of a reduced cost for fuel consumption, which is considered their private (direct) benefit. They bear a cost in the form of a higher purchase (and perhaps maintenance) cost for vehicles complying with the increased standard, which is considered a private (direct) cost to them. The private firms gain benefits in the form of lower fuel costs, less time spent in refueling, and economic gains from increased vehicle use that follows from the lower operating costs associated with improved fuel efficiency. In Table 12-1, private benefits and costs are distinguished from external, or social, benefits and costs, which are those that accrue to other people besides the owners or operators of the vehicles subject to the increased standards. These external benefits include the effects of avoided emissions to society at large, including emissions of both greenhouse gases and of criteria pollutants, and also the national security benefits of reduced reliance on imported fuels. A negative consequence associated with increased vehicle use is increased congestion and accidents, constituting a negative externality or social cost. Table 12-1 shows the benefits and costs of the Phase II rule as estimated by NHTSA and EPA using two different discount rates. 1 Both discount rate cases show a positive net benefit from the rule, ranging from $117 billion to $196 billion, depending upon the interest rate used to discount benefits and 1 In the Regulatory Impact Analysis NHTSA and EPA also present results measured against a slightly different baseline, reflecting different assumptions about technology deployment in response to the Phase I standards. The results differ slightly. The results shown here reflect what NHTSA and EPA call the “dynamic” baseline, which assumes more aggressive migration of light-truck technology to the heavy-duty (HD) segments. For a discussion of these two baselines, see NHTSA and EPA (2016, p. 10-120). There are other scenarios considered in the Regulatory Impact Analysis including a low and a high oil price scenario. As expected, total benefits are lower or higher depending on the fuel price largely since the value of fuel savings to private actors changes accordingly. Prepublication Copy – Subject to Further Editorial Correction 12-2

costs. Interestingly, the bulk of the benefits are private benefits. Net private benefits amount to between 59 and 67 percent of total net benefits, depending upon the discount rate used. 2 The reduced climate change impacts are valued at $33 billion, ranging from 50 to 67 percent of total social benefits, depending on the discount rate. In the next section we build upon the modeling done by NHTSA and EPA and discuss what lies behind these estimates of costs and benefits to approximate the marginal cost and benefits of going beyond the 2.5 percent improvement in fuel consumption. 12.1.3 Identifying the Marginal Cost of Fuel Efficiency in 2027: Heavy-Duty Pickups and Vans (Classes 2b and 3) Other aspects of the analysis done by NHTSA and EPA provide insight into the “marginal cost” and “marginal benefits” of going beyond the rule. Consistent with the task given to the committee to examine post–Phase II standards we assume that the Phase II rule takes effect and that the agencies are considering a possible Phase III rule for the period 2027 and beyond. Under this assumption, the required 2.5 percent per annum improvement would already have been met and NHTSA and EPA would be considering further required improvements. In order to meet the higher standard, additional technology will need to be applied to the new vehicle fleet. The question is what technology will be relied upon to achieve these more aggressive standards. Table 12-2 is again from the Regulatory Impact Analysis (NHTSA and EPA, 2016). The table shows the rates at which specific technologies need to be deployed in the heavy-duty pickup and van segments in order to meet the fuel-saving standards under various hypothetical alternative rules. Recall that the 2.5 percent is the actual Phase II rule. The 3.5 percent rule would achieve the same level of stringency (15.6 percent) as the 2.5 percent rule, only it would achieve it 2 years earlier. The 4 percent rule achieves a greater reduction in fuel consumption (17.9 percent), also 2 years earlier than the reductions achieved by the 2.5 percent rule. The greater fuel consumption reductions consistent with the 4 percent rule are achieved, according to the results shown in Table 12-2, by the more widespread use of electrification technologies—electronic power steering (EPS) and accessories, 12 volt (v) stop-start, and strong hybrids in the segments analyzed. 3 For example, the penetration of stop-start technology rises from 80 percent in the 2.5 and 3.5 percent annual improvement cases to 96 percent in the 4.5 percent per annum case. Similarly, the utilization of strong hybrids in the fleet rises from 2 to 7 percent. Other technologies either plateau at high rates such as the case with eight-speed automatic transmission (AT) or turbocharging. Stop-start and strong hybrids appear to be a key incremental technology necessary to move beyond the proposed rule and can be used to measure the marginal cost—the incremental cost of increasing fuel economy beyond the proposed rule—as well as the incremental benefit of compliance. Mass reduction and aerodynamic improvements also play a role in moving from a 2.5 to a 4.0 percent annual reduction in fuel consumption. 2 Normally, we would expect these potential net private savings to result in firms making the investment in fuel- saving technology. NHTSA and EPA discuss why they think that private firms react insufficiently to these potential private gains. The agencies argue that this represents a market failure. The evidence as to whether this does constitute a market failure is ambiguous, and is discussed in Chapter 13 of this report. 3 In the Preliminary Regulatory Analysis, using an earlier version of the Department of Transportation’s Corporate Average Fuel Economy model, even greater penetration rates were estimated for stop-start and mild (ISG) and strong hybrids. These results are shown in the final Regulatory Impact Analysis (NHTSA and EPA, 2016, p. 10- 122). These results are referred to as “Method B” analysis. This earlier analysis showed a much higher penetration rate for the mild and strong hybrids of 51 percent in the case of the 4 percent annual fuel consumption improvement. Stop-start technology is used in 13 percent of the fleet under the same scenario. Prepublication Copy – Subject to Further Editorial Correction 12-3

TABLE 12-2 NHTSA and EPA summary of the impact of various alternatives for HD pickups and vans versus the dynamic baseline (Alternative 1b). Alternative 2 3 4 5 Stringency of Standards Annual increase in 2.0% 2.5% 3.5% 4.0% stringency beginning in model year (MY) 2021 Increases until MY 2025 MY 2027 MY 2025 MY 2025 Total increase in MY 9.6% 15.6% 15.6% 17.9% 2030 stringency relative to final Phase I standardsa Estimated Average Fuel Economy (miles per gallon) Required in MY 2030 19.03 20.37 20.38 20.95 Achieved in MY 2030 19.20 20.47 20.45 20.98 Average Fuel Consumption (gallons/100 miles) Required in MY 2030 5.25 4.91 4.91 4.77 Achieved in MY 2030 5.21 4.88 4.89 4.77 Estimated Average Greenhouse Gas Emissions (grams per mile) CO 2 required in MY 494 462 462 450 2030 CO 2 achieved in MY 490 460 460 449 2030 Technology Penetration in MY 2030 (percent) Variable valve timing 56 56 56 56 and/or Variable valve lift Cylinder deactivation 4 4 4 4 Direct injection engine 17 27 26 29 Turbo charged engine 59 69 68 68 Eight-speed AT 77 95 94 95 EPS, accessories 52 80 80 96 12V stop-start 0 0 3 11 Strong hybrid 0 2 2 7 Aeodynamic 46 80 80 98 improvements Mass Reduction (vs. No Action) Mass reduction (lb.) 28 240 24 289 Mass reduction (percent 0.43 3.6 3.7 4.3 of curb weight) Technology Costs (vs. No-Action) Average vehicle ($) $500 $1470 $1480 1890 b Payback period (m) 19 30 31 33 a Thisincrease in stringency is based on the estimated percentage change in fuel consumption (gallons per 100 miles) stringency projected by the model for the MY 2030 fleet under the final Phase II standards relative to the continuation of the Phase II standards. b Here payback period is calculated using estimated undiscounted retail fuel savings and the initial technology costs for MY2030. SOURCE: NHTSA and EPA (2016). Prepublication Copy – Subject to Further Editorial Correction 12-4

12.1.4 Identifying the Marginal Cost of Fuel Efficiency in 2027: Vocational Segments NHTSA and EPA expect that by 2027 mild hybrids will be deployed in 14 percent of vehicles certified in the multi-purpose and urban subcategories (6 percent integrated and 8 percent non-integrated). Hybrids are seen to play an increasing role in the vocational fleet as the Phase II rule is implemented, rising from 0 percent in 2021, suggesting that further stringency in the standards would rely on further penetration of hybrid technology. Another view of the role of technology at the margin is provided in a study done by the California Hybrid, Efficient and Advanced Truck Research Center (CalHEAT) for the California Energy Commission (CalHEAT, 2013). Hybrid technology also plays a significant role in the CalHEAT study of the “roadmaps” to CO 2 emissions reduction in the California truck fleet. This study looked at what actions were necessary to reduce truck-related petroleum use and CO 2 emissions to 1990 levels by 2020 and by further reductions post-2020 so that by 2050 emissions levels are reduced by 80 percent from 1990 levels. The CalHEAT analysis takes into account the effects of Phase I, but goes beyond the reductions to be achieved by the Phase II standards. Figure 12-1, taken from that study, shows the important role of hybridization in meeting the California targets. The use of hybridization increases substantially in the post-2020 period, under these assumptions of large reductions in CO 2 emissions, but they are consistent with the results of the RIA in that hybridization begins to play a large role in the post-2020 period (CalHEAT, 2013, p. 28)5 FIGURE 12-1 Technology adoption, all truck categories. Projected adoption of CalHEAT roadmap action items by technology group for all CalHEAT truck categories combined, shown by number of vehicles affected, 2010 through 2050. By 2020, the roadmap action items could result in efficiency and emission improvements in approximately 1.7 million trucks, and by 2050, this impact could increase to 2.4 million trucks. NOTE: xEV refers to electric trucks, plug-in hybrids, and electrified corridors; HEV includes hybrid electric, electrified auxiliaries, and electric power take-off; hydraulic refers to hydraulic hybrids; new comb. includes alternative power plants and combustion cycles. SOURCE: CalHEAT (2013). Prepublication Copy – Subject to Further Editorial Correction 12-5

The CalHEAT study sees hybridization and electrification applied pervasively across a wide range of vehicle classes, with the exception of over-the-road Class 8 vehicles (CalHEAT, 2013, p. 20). Given the large role that hybridization is currently expected to play in improving fuel efficiency in the post-2027 world, we treat it as representative of the marginal technology for the next round of efficiency improvements in a Phase III regulation. We assume either that hybridization is this marginal (incremental) technology or that the marginal technology has the same magnitude of cost and fuel efficiency impact as hybridization. Therefore, in the remainder of this section, we analyze the costs and benefits of a Phase III fuel efficiency target using, as a proxy, data on costs and related efficiency improvement possible by hybridization. 12.1.5 Estimating the Marginal Cost of Fuel Efficiency Improvements Table 12-3 shows estimates of the cost of full and partial hybridization in the heavy-duty van and pickup segments as well as other vocational segments. The capital costs are from Chapter 7 of this report. The capital costs include the direct manufacturing costs as well as the indirect costs such as R&D costs, corporate overheads, and sales and distribution costs. Allowance is also made for a return on capital. In addition, the capital costs are projected forward to the 2027 to 2030 period, taking into account expected learning effects over time as volumes increase. The capital cost of the technology is compared to the value of the fuel savings. Several measures are presented to analyze the value of the technology. The first, and most commonly used in the industry, is the simple payback period. This is the capital cost divided by annual fuel savings, at an assumed fuel price. The committee has frequently heard from industry representatives that truck manufacturers and purchasers use the payback period as a heuristic for the value of new technology. The payback period captures, in a rough approximation, the cost of capital and the uncertainty associated with the adoption of new technologies. These uncertainties include the benefits of the new technology under real-world and particular duty cycles, volatility in fuel prices, unanticipated maintenance costs, etc. The committee has heard from manufacturers and purchasers that they look for 1.5- to 2-year paybacks or, in other cases, for a payback period that is half the expected ownership period of the first owner of the vehicle. 4 For the Class 2b pickups the payback periods in the table are calculated using the gasoline price projected for 2027 by the Department of Energy in its Annual Energy Outlook (AEO) 2017 of $2.88 per gallon. For the other vehicle classes, the payback periods are calculated using the forecast for the diesel fuel price of $3.52 per gallon. 5 The second measure shown in the table is the “break-even” fuel price. This measure calculates what the fuel price would have to be in order for the present discounted value of the fuel savings (over 10 years) to equal the initial capital cost. If the forecasted fuel price exceeds the “break-even price,” the technology offers savings to a firm adopting the technology. The final measure is the “break-even capital cost,” that is, the capital cost that would just equal the present value of the fuel savings at the relevant assumed fuel price. Again this is done for the projected 2027 fuel prices. The “break-even capital cost” can also be interpreted as the cost any fuel-saving technology would have to meet to be attractive to a private firm for the vehicle segment, duty cycle, and efficiency gain shown. It is therefore a proxy for the targets any technology must meet to be attractive to private manufacturers and purchasers. 4 Klemick (2015) confirms this observation in Chapter 13 of this report, citing Roeth (2013), and points out that first ownership periods, while lengthened during the recession of 2008 and 2009, have come down recently to about 4 years, suggesting an acceptable payback period in the 2-year range. 5 These are prices converted from dollars of 2016 to constant dollars of 2013. The cost estimates in Chapter 7 used here assume that hybrid technology is added to a gasoline-fueled vehicle for Class 2b pickups. For the other vocational vehicle classes shown in Table 12-3, hybrid technologies are assumed added to diesel-fueled vehicles. Therefore, the breakeven fuel price is compared to gasoline prices for Class 2b vehicles and to diesel fuel prices for the other vehicle classes shown. Prepublication Copy – Subject to Further Editorial Correction 12-6

The calculations in the table represent the typical use in the segment, as measured by the average vehicle miles traveled (VMT) in the segment. Consistent with the Regulatory Impact Analysis, a 7 percent discount rate for private investment is assumed. The life of the vehicle is assumed to be 10 years.6 Using these metrics for evaluating the net benefits of a fuel-saving technology are only approximations since they exclude other potential costs such as maintenance, variability in duty cycle, and driver convenience. Nevertheless, they are broadly suggestive of the net economic benefits of the technologies. 6 The assumption of a 10-year life, coupled with the assumption of a constant VMT over the 10-year period, also biases the estimates in favor of the technologies shown in Table 12-3. In the case of hybrid vehicles, battery life is the relevant period for calculating the fuel savings. These calculations assume, in essence, that the life of the battery is 10 years and, after that, significant additional capital expenditure would be necessary. Prepublication Copy – Subject to Further Editorial Correction 12-7

TABLE 12-3 2027-2030 Estimated Costs, Payback Period, Break-even Fuel and Capital Cost for Hybrid Technology, Waste-Heat Recovery Technology, and Trailer Skirts (2013 dollars) Capital Break- Vehicle Costa Mpgb Efficiency even Fuel Break-even Payback Type Technology Duty Cycle (2030) (2027) Gainc VMTd Pricee Capital Coste (years) Class 2b Stop-start Urban 425 20.6 4.5% 14,000 $1.98 $619 4.8 pickup Multi-purpose 425 20.6 4.5% 14,000 $1.98 $619 4.8 Mild hybrid Urban 2,000 20.6 17.5% 14,000 $2.39 $2,406 5.8 Multi-purpose 2,000 20.6 12.5% 14,000 $3.35 $1,719 8.2 FHEV Urban 4,650 20.6 22.5% 14,000 $4.33 $3,093 10.6 Multi-purpose 4,650 20.6 17.5% 14,000 $5.57 $2,406 13.6 Class 4 Stop-start Urban 700 8.4 4.5% 14,000 $1.33 $1,854 2.7 delivery Multi-purpose 700 8.4 4.5% 14,000 $1.33 $1,854 2.7 van Mild hybrid Urban 4,600 8.4 17.5% 14,000 $2.25 $7,211 4.5 Multi-purpose 4,600 8.4 13.5% 14,000 $2.91 $5,563 5.8 FHEV Urban 7,550 8.4 23.5% 14,000 $2.74 $9,683 5.5 Multi-purpose 7,550 8.4 17.5% 14,000 $3.69 $7,211 7.4 Class 8 Stop-start Urban 1,400 6 4.5% 25,000 $1.06 $4,636 2.1 refuse Mild hybrid Urban 10,825 6 16.5% 25,000 $2.24 $16,997 4.5 truck FHEV Urban 14,875 6 23.5% 25,000 $2.16 $24,208 4.3 Urban bus Stop-start Urban 1,400 6 4.5% 34,000 $0.78 $6,304 1.6 Mild hybrid Urban 10,825 6 16.5% 34,000 $1.65 $23,116 3.3 FHEV Urban 14,875 6 23.5% 34,000 $1.59 $32,923 3.2 Class 8 WHR Highway 7,025 6.4 2.5% 120,000 $2.13 $11,589 4.3 tractor- Skirts Highway 900 6.4 2.5% 40,000 $0.82 $3,863 1.6 trailer aCapital cost of technologies: Chapter 7 of this report. The estimates reflect the high-cost estimates in Chapter 7, Table 7-6, for all technology/vehicle combinations except for Class 2b. It is assumed that the experience gained in light vehicles will inform the Class 2b segment and the lower cost estimates will be achieved. The estimates have been rounded to the nearest $25 interval. bInitial mpg: a. Class 2b pickups—from (Lutsey, 2015, table 1): estimate for model year (MY) 2010 of 16.4. b. Class 3—15.1 (Lutsey, 2015): estimate for MY 2010 MY. Prepublication Copy – Subject to Further Editorial Correction 12-8

c. Vocational vehicles—from Lutsey (ICCT, June 2010): 6 mpg (from table 5, Lutsey estimate for 2027 gasoline vehicles; note: table 1 in the source document gives an estimate of 2014 mpg for various classes with gasoline and diesel combined). c Efficiency: Midpoint estimate from Chapter 7. dVMT: a. Class 2b and 3: from NRC Phase I report (2010a). b. Refuse truck: from Gordon et al. (2003); Greening Garbage Trucks: New Technologies for Cleaner Air, Inform, Inc. 2003. Cited in www.afdc.energy.gov/data/. c. Urban bus: www.afdc.energy.gov/data. d. Class 8 tractor-trailer: same as b and c. e. VMT for trailer skirts reflects an estimate of three trailers used per tractor. eForecasted fuel prices used in payback calculations and in break-even capital cost calculation are from Annual Energy Outlook, 2017, expressed in 2013 dollars. Prepublication Copy – Subject to Further Editorial Correction 12-9

Table 12-3 suggests that payback periods for hybrid applications are generally long. The payback periods for technologies other than stop-start, sometimes called micro hybrid, range from 3.2 years in the case of a full hybrid deployed on an urban bus to more than 13 years for a full hybrid on a Class 2b pickup used in a multi-purpose duty cycle. Stop-start technology, however, has relatively low paybacks in many applications shown in Table 12-3. Based on what the committee has heard in presentations from industry representatives and survey evidence cited above, other than for stop-start, the estimated paybacks would likely not be attractive to most private purchasers of medium- and heavy-duty trucks or the manufacturers. In other words, to achieve fuel efficiency gains beyond the improvements offered by stop- start technology, the larger gains from mild and strong hybrids likely would not be sufficient by themselves to drive widespread diffusion of the technology, given that the adoption decisions would be made based on relatively short payback period requirements. The break-even fuel price metric used here calculates what price of fuel would produce fuel savings such that the present value of those savings over a 10-year period equals the stated capital cost. Importantly, this metric assumes away the uncertainties that private firms face and that they proxy by their use of the simple payback period. In contrast to the payback criterion, the calculated breakeven fuel price metric suggests that if the projected capital costs and efficiency gains are consistently realized, the application of mild and full hybrids in the 2027-2030 period will offer cost-efficient fuel savings in many vehicle classes and duty cycle combinations. Consistent with its relatively quick payback, stop-start technology in most applications would make economic sense at fuel prices far below the prices projected by the Energy Information Agency. Comparing the calculated break-even fuel prices to the AEO forecasts of $2.88 and $3.52 for gasoline and diesel fuel, respectively, suggests that mild hybrids in most cases also would be economically justified. Only in the case of the Class 2b pickup when used in a multi-purpose duty cycle does the break-even price for the mild hybrid exceed $2.88 per gallon. Again, using the breakeven fuel price metric, full or strong hybrids present a mixed case. Full hybrids appear to be uneconomic in Class 2b pickups, with break-even fuel prices substantially exceeding the AEO price forecast. In a Class 4 delivery van, the breakeven price is below the forecasted diesel price of $3.52 per gallon in the urban duty cycle, but slightly exceeds it in the multi-purpose cycle. Full hybrid electric vehicles (FHEVs) appear to meet that economic test in the heavier vocational vehicles, where VMT is high and the vehicles are utilized in an urban duty cycle. The contrast between the results using payback periods and those using break-even fuel prices highlight a debate in the energy economics literature, often referred to as the “energy paradox.” Some would argue that private trucking firms and manufacturers using relatively short payback periods are “leaving money on the table” by not adopting cost-effective fuel-saving technology. Others would argue that engineering analysis does not capture all the costs involved in utilizing the new technology or that the break-even fuel price measures use too long a vehicle life and do not reflect the varieties of duty cycles as the vehicle ages. 147 To the extent that the use of short payback periods, however, represents uncertainties about the costs and performance of new technologies, one would expect private firms to lengthen the acceptable payback period over time if the technologies demonstrate consistent cost and vehicle performance. The break-even capital cost provides an estimate of how much cost reduction would be necessary to produce a favorable economic case for the technologies and applications shown in Table 12-3. For example, in the case of the full hybrid deployed in Class 2b pickups, capital costs would have to decline by approximately 35 to 50 percent to equate the capital cost to the present discounted value of the fuel savings. In contrast, in the stop-start case in the Class 2b vehicle, the estimated capital cost is about two- thirds of the break-even level, indicating a favorable economic evaluation of the technology. Similarly, many of the other vehicle technology–duty cycle combinations shown in Table 12-3 have estimated capital costs below the break-even capital cost level. 147 This issue is discussed further in Chapter 13. Prepublication Copy – Subject to Further Editorial Correction 12-10

As VMT increases, the value of the fuel savings goes up and the technology becomes more economically attractive. Figure 12-2 below show the sensitivity of the break-even period and the break- even fuel price to VMT. As the graphs show, even at a VMT of 20,000 miles, strong and mild hybrid applications still have relatively high payback periods in the Class 2b and Class 4 van segments, again suggesting that private firms using this metric would not favor rapid adoption of these technologies. 148 When capturing 10 years of potential fuel savings, however, the break-even price of fuel, at 20,000 VMT, falls below the forecasted price of fuel of $3.52 for mild and strong hybrid applications in Class 4 vans. In the case of Class 2b pickups at 20,000 VMT, mild hybrids would be economic at the forecasted fuel price, while break-even fuel prices for full hybrids would still exceed the forecasted fuel price. In contrast, also on the chart are the curves representing waste-heat recovery (WHR) technology in Class 8 vehicles and trailer skirts for trailers that lower aerodynamic drag. The WHR technology is still in development and has not yet been deployed on vehicles. NHTSA and EPA expect that this technology will gain increased application beginning in 2021 and rise to 25 percent of tractor engines in 2027 (NHTSA, 2016, p. 2-82). In the post-2027 period, engine manufacturers as well as the committee expect that WHR technology will be deployed (Cummins, 2015). The curve shown in Figure 12-3 therefore represents a view of its potential cost and efficiency, once the obstacles are overcome and the technology is in production. Given the large number of vehicle miles traveled by Class 8 long-haul vehicles, this technology does offer potentially large absolute fuel savings per vehicle as well for the industry as a whole. The large fuel savings per vehicle, in turn, lead to a low break-even fuel price and a break-even capital cost significantly above the estimated potential cost. Nevertheless, at almost 5 years, the simple payback is still higher than the acceptable payback period to private firms, as discussed above. Trailer skirts, on the other hand, are widely deployed on trailers today and have a very low break-even price of fuel and a very quick payback. 20 18 16 Payback Period (Years) 14 12 10 8 6 4 2 0 0 20000 40000 60000 80000 100000 120000 140000 Annual Vehicle Miles Traveled Class 2b Pickup Mild Hybrid (Multipurpose Cycle) Class 2b FHEV (Multipurpose Cycle) Class 2b FHEV (Urban Cycle) Class 4 Van Mild Hybrid (Multipurpose Cycle) Class 4 FHEV (Urban Cycle) Class 8 Refuse Truck FHEV (Urban Cycle) Class 8 Tractor-trailer WHR (Highway) Class 8 Tractor-trailer Side Skirt (Highway) FIGURE 12-2 Payback period sensitivity with respect to vehicle miles traveled. 148 There is clearly a relationship between VMT per year and life of vehicle. At 20,000 annual miles traveled, the 10- year life of vehicle is less likely an appropriate assumption. The estimates here do not adjust for that and thus are biased toward favorability of the technology. Prepublication Copy – Subject to Further Editorial Correction 12-11

$8.00 $7.00 $6.00 $5.00 $4.00 $3.00 $2.00 $1.00 $0.00 0 20000 40000 60000 80000 100000 120000 140000 Annual Vehicle Miles Traveled Class 2b Pickup Mild Hybrid (Multipurpose Cycle) Class 2b FHEV (Multipurpose Cycle) Class 2b FHEV (Urban Cycle) Class 4 Van Mild Hybrid (Multipurpose Cycle) Class 4 FHEV (Urban Cycle) Class 8 Refuse Truck FHEV (Urban Cycle) Class 8 Tractor-trailer WHR (Highway) Class 8 Tractor-trailer Side Skirt (Highway) FIGURE 12-3 Break-even fuel price sensitivity to vehicle miles traveled. SOURCE: Committee calculations based on Table 12-3. As discussed above, the private fuel savings are only part of the benefits to society. In addition to these private savings, reductions in fuel use lower the environmental, health, and national security externalities. In order to reflect these externalities their cost can be added to the forecasted price of fuel. If the actual price of fuel plus the value of the externalities equals or is greater than the break-even price of fuel, the technology is socially efficient. Even if the private decision maker would not benefit from adopting the technology, society would still gain by forcing the technology adoption through regulation or through taxation. 149 The following sections of this chapter consider how to value the externalities associated with fuel production and consumption. The median estimated value of reducing carbon emissions, based on U.S. government estimates and as discussed in the section “The Benefits of Reduced CO 2 Emissions” below, is the equivalent of about $0.50 per gallon in 2013 dollars, using a 3 percent social discount rate. 150 The value of the avoided health damages associated with lower criteria pollutant emissions offset partially by 149 As noted above on page 12-10, this analysis uses average duty cycles and abstracts from some potential costs such as increased maintenance that society may incur by forcing the technology. The variability in duty cycles also means that forcing the technology does risk higher net social costs in some uses. 150 Section 12.2 below discusses how the social cost of carbon is measured. Shown in Table 12-4 is the median estimate of the Interagency Working Group (IWG) of $46 per tonne in 2025 and $50 in 2030. The midpoint of $48 is used here to proxy for the cost in 2027. These estimates were in 2007 dollars, so they are converted to 2013 dollars using a factor of 1.18. The cost per tonne of CO2 is converted to an equivalent cost per gallon of gasoline using the converter found at http://www.icbe.com/carbondatabase/dollarspertontocentspergallon.asp. Note also that the social cost of carbon is the global cost. In other words, the cost of emitting one tonne of carbon leads to the estimated damage globally. The effect on the United States is estimated by the IWG to be from 7 to 23 percent of the global cost. It is, as discussed in Section 12.2, a policy choice whether to consider the global cost or just the impact on the United States. In the last rulemaking NHTSA and EPA chose to use the global cost. A March 28, 2017, executive order issued by President Trump rescinds the Obama administration guidance on the use of the social cost of carbon in rulemaking and relies instead on previous guidance. One result of this order will be to estimate costs and benefits on a U.S.-only basis. Prepublication Copy – Subject to Further Editorial Correction 12-12

increased congestion and noise costs, based on the estimates in the RIA, is approximately $0.31151 using a 3 percent discount rate. Using a 7 percent discount rate the value is approximately $0.16. The national security cost as discussed in Section 12.2 below has an estimated median value of about $0.17 per gallon. The break-even price of fuel can be compared to the social cost of the fuel by adding the value of the externalities to the forecasted price of fuel. Taking the median value of the cost of carbon of $0.50 per gallon, the median estimate of the national security premium of $0.17 per gallon and the estimate of health benefits using the 3 percent discount rate of $0.31 per gallon results in a total external cost of $0.98 per gallon. The forecast of fuel prices of $2.88 and $3.52 for gasoline and diesel fuel, respectively, results in social costs of fuel of approximately $3.86 and $4.50. A comparison of the estimates of break-even fuel prices in Table 12-3 to the estimated social cost of fuel, suggests that hybridization, is likely to be the “next up” technology in the post-2027 time frame in all the vehicle segments and duty cycles shown in Table 12-3, if the vehicles attain at least the level of VMT shown in the table. Only in the case of full hybridization in the Class 2b pickup does the break-even fuel price exceed the estimated social cost of fuel. It is important to restate that this analysis ignores the variation in duty cycles within each vehicle segment shown and that the estimates of social costs, as discussed in the sections that follow, vary considerably. Nevertheless, the results here are suggestive of the technology likely to be used to meet any potential increase in fuel efficiency standards. 12.1.6 Findings Finding: Waste-heat recovery used in Class 8 over-the-road vehicles potentially offers significant cost- effective fuel savings in the post-2027 period. Expected progress in WHR technology suggests that in the post-2027 period this technology will see increasing penetration in Class 8 tractor-trailers. Finding: In 2027 and beyond, stop-start technology applications are expected to have payback periods that would make them attractive in many applications to private firms. The greater fuel savings, however, that are possible with the application of mild and strong hybrids are expected to be less attractive to private firms, given their relatively long payback periods and the weight placed on payback periods by private firms. Finding: If the anticipated cost reductions and efficiency gains are in fact consistently achieved, in the post-2027 period increases in fuel efficiency beyond the Phase II rule are likely to be met with increasing applications of hybrid technologies including stop-start, mild and strong hybrids. The technology penetration will vary depending on vehicle class and duty cycle. Finding: Considering 10-year fuel savings, break-even fuel prices are below the social cost of fuel in most hybrid applications. These results suggest that hybridization is a technology that is “next up” for meeting more stringent post–Phase II standards. Only in Class 2b does the break-even fuel price for full hybrids exceed the forecasted social cost of fuel. Recommendation 12-1: Since hybrid technologies and WHR potentially could play increasing roles in achieving reductions in fuel consumption in the post-2027 period, developments in the cost and efficiency 151 The value of the other non-CO2 pollutants, discussed below in Section 12.2.1, is calculated most appropriately on a per-VMT basis. A proxy measure is calculated here by using the total value of avoided health effects offset by congestion and noise effects as estimated in NHTSA and EPA’s RIA (Table 8-30). The net present discounted value of these non-CO2 externalities ranges from $12.2 billion at a 7 percent discount rate to $24.1 billion at a 7 percent discount rate using NHTSA and EPA’s dynamic baseline. Dividing these totals by the 77.7 billion gallons of fuel saved yields an estimate of average non-CO2 externalities of $0.16 to $0.31 per gallon of fuel saved (RIA, Table 11- 20). (A slightly different baseline yields a range of $0.18 to $0.34 per gallon.) In Section 12.3 of this report the national security premium is discussed. As shown in that section, the median estimate by NHTSA and EPA in the Phase II rulemaking was the equivalent of about $0.17 per gallon. Prepublication Copy – Subject to Further Editorial Correction 12-13

of these technologies should be monitored and be included in a formal interim review of fuel consumption standards. 12.2 THE BENEFITS OF REDUCED CO2 EMISSIONS CO 2 and most other GHGs cause harm not through direct exposure of humans or wildlife to the pollutant but, rather, because their accumulation changes the energy balance in the atmosphere and the chemistry of the ocean. The change in the earth’s energy balance leads to changes in climate worldwide, including changes in temperature and precipitation, and to melting of sea ice sheets and sea level rise. In the ocean, their accumulation leads to ocean acidification. These changes have many consequences that affect the well-being of humans and natural ecosystems worldwide. In some cases, the effects are beneficial; often they are harmful. Because GHGs emitted anywhere on the earth mix globally in the atmosphere and are then introduced into the ocean through interface with the near-surface atmosphere, they are different from other air pollutants including criteria air pollutants: the emission of a GHG anywhere on earth affects global climate and ocean acidification. A second distinctive feature of GHGs is that they are what is known as stock pollutants: the harm comes from the accumulated stock of GHGs in the atmosphere and the ocean. Conventional air pollutants are generally flow pollutants: the harm from them comes, essentially, contemporaneously with the release of emissions and they do not linger. With GHGs, as a stock pollutant by contrast, harm continues from the time of emission until the stock of pollutant has dissipated, which can take centuries in the case of CO 2 . 152 This has two implications for the assessment of the damages. First, for an emission in any given year, the effect on global climate and the resultant consequences play out over time until that emission has entirely decayed. Second, the emission interacts in the atmosphere and the ocean with subsequent emissions and its impact on climate is influenced by those later emissions. Therefore, accounting for the consequences of GHG emissions requires dynamic modeling of the accumulation and dissipation of GHGs in the atmosphere and the ocean as well as the resultant evolution of the climate system. This is done through what are known as Integrated Assessment Models (IAMs). These are mathematical models that track, in sequence, the emission of GHGs as a function of economic activity; the accumulation and decay of GHGs in the atmosphere, the surface (or upper ocean), and the deep ocean; the link from atmospheric concentration to global temperature via radiative forcing 153; and the impact of temperature change on economic activity and human well-being, measured in monetary terms as the equivalent reduction in GDP. Reduced CO 2 emissions are valued using the social cost of carbon (SCC). This is a monetary measure of the discounted present value of the net damage resulting over time from the emission of an additional unit of CO 2 in a given year, estimated by means of an IAM. We use here the estimate of the SCC developed by the federal government’s IWG in 2010 (IWG, 2010), as subsequently updated in 2013 (IWG, 2013). 154 152 The persistence of GHGs varies greatly. The warming impacts of ozone or contrails last only days or months. Those of methane last for 20-30 years. Those of CO2 persist for hundreds of years. Some GHGs persist even for thousands of years. 153 In tracking the linkage from emissions to changed global temperature, the IAMs offer a simplified and reduced- form representation of the vastly more detailed analysis found in Global Climate Models (GCMs). GCMs project precipitation as well as temperature. Because of their highly simplified representation of the carbon cycle the IAMs project only temperature. The GCMs project temperature on a spatial scale consisting of grids of about 175 km. The IAMs project temperature on the spatial scale of large groups of countries (FUND, PAGE) or the entire globe (DICE). Also, while the GCMs operate on something like an hourly time scale, the IAMs operate on an annual time scale. 154 There was subsequently a small correction to the 2013 analysis in IWG (2015). Prepublication Copy – Subject to Further Editorial Correction 12-14

The IWG ran the three main IAMs used in the literature 155 and averaged the results across them. The models were run over the period through 2300. 156 Rather than using the emissions projection native to each IAM, the IWG standardized the emission projections across the IAMs by using a common set of projections taken from an external source.157,158 Five alternative projections were used. Four assume business-as-usual (BAU) emissions through 2100, with a leveling off and downturn in emissions between 2100 and 2200; the fifth alternative assumes that emissions are reduced (“down-scaled”) during the period 2000-2100 to meet a target of 2 degrees Celsius (°C) warming by 2100 and then remain roughly constant thereafter. The results with each of the emissions projections were weighted equally, thereby assigning a 20 percent probability to the low (2°C) projection and 80 percent collectively to the BAU projections. Given a trajectory of emissions, a key parameter in translating these emissions to subsequent changes in temperature is the climate sensitivity. This parameter measures the long-run equilibrium increase in the global annual average surface temperature resulting from a doubling of atmospheric CO 2 relative to the pre-industrial level. 159 It reflects both the direct warming from the increase in radiative forcing and also the warming resulting indirectly from feedback processes triggered by that warming, such as melting of the permafrost. 160 Uncertainties in the value of the climate sensitivity have received considerable attention in the scientific literature. 161 To account for this, the IWG employed a probability distribution for the value of the climate sensitivity, which had the effect of making the SCC estimate probabilistic. 162 The IWG used a well-known probability distribution for the climate sensitivity, that of Roe and Baker (2007), and applied it across all three IAMs. 163 Ten thousand draws of the value of the climate sensitivity were taken from this distribution and input into each of the three IAMs. Given an 155 These are the DICE model, developed by Professor William Nordhaus in 1990; the PAGE model, developed by Chris Hope in 1992; and the FUND model, developed by Richard Tol in the early 1990s. For its 2010 analysis, the IWG used the 1999 version of DICE (Nordhaus and Boyer, 2000), the 2002 version of PAGE (Hope, 2006), and version 3.5 of FUND (Anthoff and Tol, 2012). Following updating of these models by their developers, the 2013 IWG analysis used DICE 2007 (Nordhaus, 2007), PAGE 2009 (Hope, 2011), and FUND 3.8 (Anthoff and Toll, 2014). 156 The native time horizon in PAGE is 2200; that in DICE is 2595; that in FUND is 3000. The IWG made 2300 a common time horizon across all three models. The IWG did not alter the IAM’s native time steps, which are annual in FUND, and decadal in DICE. The time steps in PAGE are decadal through 2060 and 20 years thereafter. 157 The source was a model intercomparison exercise conducted by the Stanford Energy Modeling Forum, EMF-22 (Clarke et al., 2009), designed to assess the economic cost (and feasibility) of meeting a target of 2°C by 2100. The models compared were all versions of an IAM but, with one exception, they stop at the projection of future warming and they lack a damage function. The IWG used the four best known of the models participating in the EMF exercise. 158 While FUND and PAGE use exogenous projections of emissions, emissions are endogenous in DICE. DICE is a modified model of optimal economic growth in which, in each period, there is choice of how much of what is produced in that period to consume, how much to invest (i.e., to allocate to capital formation for use in future production), and how much to allocate to emission abatement. The allocation to investment determines the future level of production which, in turn, determines future emissions. The IWG disabled the optimizing allocation of consumption and investment in DICE and made emissions exogenous, as in FUND and PAGE. 159 This is distinct from the transient climate response, which determines the response of temperature within a fixed short-term period, and which also is represented in the IAM carbon cycle. 160 The climate sensitivity includes feedback processes that occur over the short to medium term (up to 100 to 200 years), but it does not include long-term feedbacks which occur on a time scale of many hundreds to thousands of years. 161 A significant part of the uncertainty arises from the feedback processes and the warming that they generate. 162 The climate sensitivity is modeled as a random variable in FUND and PAGE, along with several other parameters. DICE, by contrast, is entirely deterministic. 163 This probability distribution has a median value of 3.0 and a mean of 3.5. With this distribution, there is only a 1.3 percent chance that the value of the climate sensitivity is below 1.5°C, and only a 10 percent chance that is it above 5.86. This is broadly consistent with the values used for DICE (a fixed value of 3.0) and for PAGE and FUND (both probabilistic with a mean of 3.0), and it is consistent with the findings of both the Fourth and Fifth Intergovernmental Panel on Climate Change Assessment Reports. Prepublication Copy – Subject to Further Editorial Correction 12-15

emission scenario, for each draw the IAM was then run through year 2300, generating a particular trajectory of warming and damages. With three IAMs and five emission scenarios, this generated 150,000 trajectories of warming and damages. The IWG estimated the SCC associated with an increment in emissions occurring in 2010, 2020, 2040, and 2050. 164 In each case, the procedure was to introduce a one-time pulse in emissions in the reference year, leaving later emissions unchanged, rerun the model, and, for each period following the pulse, calculate the additional damages in that period compared to the base run of the same version of the model without the pulse in emissions. For each reference year this generated 150,000 trajectories of incremental damages associated over time with the pulse in emissions. The trajectories of incremental damages were converted to a discounted present value (an SCC value) using three alternative discount rates: 2.5, 3.0, and 5.0 percent. The result was 450,000 estimates of the SCC, corresponding to the three IAMs, the 10,000 values of the climate sensitivity, the five alternative emissions scenarios, and the three alternative discount rates. The results were summarized by reference year and by discount rate, for the three IAMs combined. 165 The IWG tabulated the average SCC value at each of the three discount rates, together with the 95th percentile value at the 3 percent discount rate (the results are reproduced below in Table 12-4). For its central estimate, the IWG used the average SCC value across the three IAMs at the 3 percent discount rate. TABLE 12-4 Revised Social Cost of CO 2 , 2010-2050 (in 2007 dollars per metric ton of CO 2 ) Discount Rate Year 3.0% Avg 2010 31 2015 36 2020 42 2025 46 2030 50 2035 55 2040 60 2045 64 2050 69 SOURCE: IWG (2015, p. 3). The IWG’s 2010 and 2013 estimates of the SCC have been used in numerous rulemaking proposals by federal agencies, some of which provided opportunities for public comment. In addition, following the publication of the IWG’s 2013 estimate, the Office of Management and Budget (OMB) put out a formal request for comments on the methodology used by the IWG. In July 2015, the IWG issued a formal response to these comments (IWG, 2015). The issues raised by the comments fall mainly into two groups: modeling choices that should be considered as policy decisions being made by the U.S. government, and decisions that represent an exercise of judgment with regard to technical details of the empirical analysis. The policy judgments include the decision to consider global impacts rather than just impacts in the United States, the decision to consider impacts through 2300, and the choice of 2.5, 3, and 5 percent as discount rates when calculating the SCC. 164 The IWG used linear interpolation to generate SCC estimates for the in-between years of 2015, 2025, 2035, and 2045. 165 Compared across reference years, the SCC estimate increases (in constant dollars) over time because, as time passes, global temperature is higher and there is a greater accumulation of CO2 in the atmosphere. This causes both the warming impact of a pulse in emissions to be larger and the extra damage from a given increment in warming to be larger. Prepublication Copy – Subject to Further Editorial Correction 12-16

The IWG estimated that damages in the United States might amount to between 7 and 23 percent of the global climate impact reflected in its SCC estimate (IWG, 2010, p. 11). The IWG justified its adoption of a global perspective because of the global externality caused by U.S. emissions, which contribute to damages around the world. Consequently, “to address the global nature of the problem, the SCC must incorporate the full (global) damages caused by GHG emissions” (IWG, 2010, p. 10). It noted that accounting for the global impacts can encourage reciprocal action by other nations. It further observed that adverse impacts of climate change on other countries can have spillover effects on the United States, “particularly in areas of national security, international trade, public health and humanitarian concerns.” Moreover, “if all countries acted independently to set policies based only on the costs and benefits of carbon emissions, it would lead to an economically inefficient level of emission reductions which could be harmful to all countries, including the US” (IWG, 2015, p. 32). The IAMs used by the IWG have different default time horizons: 2200 (Policy Analysis of the Greenhouse Effect [PAGE]), 2595 (Dynamic Integrated Climate-Economy Model [DICE]), and 3000 (Climate Framework for Uncertainty, Negotiation and Distribution [FUND]). The IWG considered the year 2200 too short a time horizon because it could omit a significant fraction of damages, and chose to run the three IAMs through 2300. Some commenters argued for a shorter time horizon, say 2100, due to the uncertainty associated with projecting impacts out to 2300. IWG (2015, p. 29) rejected this suggestion as inappropriate given the long atmospheric lifetime of CO 2 . The longer the time horizon, the greater the role of the discount rate in estimating the SCC. 166 OMB’s Circular A-4 identifies 3 and 7 percent as appropriate discount rates for regulatory impact analysis conducted pursuant to Executive Order 12866. The 7 percent rate is an estimate of the before-tax, real opportunity cost of capital for business borrowers in the United States. The 3 percent rate is an estimate of the social rate of time preference (the consumption rate of interest) for consumers in the United States. The IWG found that the consumption rate of interest is the relevant discount rate for the SCC. IWG (2015) notes that most regulatory impact analysis is conducted over a time frame in the range of 20 to 50 years. The SCC involves a far longer time horizon. OMB Circular A-4 recognizes that special ethical considerations arise when comparing benefits and costs across generations. “Although most people demonstrate time preference in their own consumption behavior, it may not be appropriate for society to demonstrate a similar preference when deciding between the well-being of current and future generations. … Estimates of the discount rate appropriate in this case, from the 1990s, ranged from 1 to 3 percent” (IWG, 2015, p. 21). The central value, 3 percent, is consistent with estimates provided in the economics literature and OMB’s Circular A-4 guidance for the consumption rate of interest. The upper value of 5 percent represents the possibility that climate damages are positively correlated with market returns, which would suggest a rate higher than the risk-free rate of 3 percent. Additionally, this discount rate may be justified by the high interest rates that many consumers use to smooth consumption across periods. The low value, 2.5 percent, is included to incorporate the concern that interest rates are highly uncertain over time. It represents the average rate after adjusting for uncertainty using a mean-reverting and random walk approach as described in Newell and Pizer (2003), starting at a discount rate of 3 percent. Furthermore, a rate below the riskless rate would be justified if climate investments are negatively correlated with the overall market rate of return. Use of this lower value also responds to the ethical concerns discussed above regarding intergenerational discounting (IWG, 2015, p. 22). Other issues raised in the comments include the projection of future GHG emissions, the adequacy of the IAM damage functions used to assess the damage from future warming, and the averaging of the SCC estimates across the three IAMs. Some commenters argued that the emission projections were too pessimistic and understated the future abatement that would occur; others argued that the projections were too optimistic. In response, IWG (2015) acknowledged the uncertainty in these projections: “The chosen scenarios capture a wide range of potential future states of the world, but were not intended to represent a comprehensive 166 The present value today of $10,000 in 2300, discounted at 2.5, 3, and 5 percent, amounts, respectively, to only $8.78, $2.19, and $0.01. Prepublication Copy – Subject to Further Editorial Correction 12-17

accounting of the full range of uncertainty, and therefore it is possible that future outcomes will fall outside of this range” (IWG, 2015, p. 20). Some commenters argued that it is not possible to meaningfully assess how climate impacts occurring two or three hundred years in the future will be valued by society. Some suggested that the IAM damage functions are arbitrary and unreliable. In response, the IWG acknowledged that none of the three IAMs fully incorporates all climate change impacts, whether positive or negative. Using an ensemble of three different models, it pointed out, “was intended to, at least partially, address the fact that no single model includes all of the impacts” (IWG, 2015, pp. 9-10). Some commenters suggested that the IWG should have used the median rather than the mean of the distribution of SCC estimates, which would produce a lower value. Using the median effectively discards the tails of the distribution. The IWG viewed this as inappropriate in the context of climate change where tail risks—risks associated with low probability but highly damaging outcomes—are of the essence (IWG, 2015, p. 26). Finding: It is a government policy decision as to which impacts should count toward the social cost of carbon, i.e., the impact of climate change on the United States versus the global impact. It is also a policy decision over what time period the impacts should be considered (through 2300, or later, or earlier), and how impacts occurring at different points in time should be discounted (e.g., at 3 percent or some other rate). The U.S. government’s choice of time frame is generally consistent with the time span of the IAMs used. Its choice of a discount rate is consistent with estimates in the literature. 12.2.1 Additional Environmental and Health Costs and Benefits from Different Fuel and Technology Strategies We discussed above the methods used by NHTSA and EPA to value the climate change externalities associated with the reductions in carbon dioxide emissions resulting from the fuel savings from the use of more efficient MHDV technologies and strategies. Additionally, there are two important environmental costs or benefits associated with the likely changes in technologies, fuels, and strategies that will be induced by the Phase II standards. First, there are changes in emissions from criteria air pollutants—and resulting changes in health and environmental damages induced by those emissions. Second, there are also changes in levels of damages from other greenhouse gases besides CO 2 . 12.2.2 Costs and Benefits from Changes in Health and Environmental Impacts from Criteria Air Pollutants The use of medium- and heavy-duty vehicles powered by fossil fuels (in particular gasoline and diesel) results in emissions of criteria air pollutants, such as nitrous oxides (NO x ), particulate matter (PM), sulfur oxides (SO x ), volatile organic compounds (VOCs), and air toxic emissions. A key distinction between these pollutant emissions and greenhouse gases is that if the source of emissions is removed, the overall concentration of these pollutants decreases rapidly, whereas, as was explained above, the lifetime of GHGs is many decades to centuries and can lead to effects that will be faced by future generations. In a life-cycle perspective, it is worthwhile pointing out that the upstream emissions associated with the production and transportation of such fuels will also likely change—but over the lifetime of these vehicles, the use phase is by far the largest contributor in terms of emissions. (See, for example, the NRC [2010b] report Hidden Costs of Energy: Unpriced Consequences of Energy Production and Use. Note that while the use phase is the largest contributor to emissions for MHDVs, the same is not true for light- duty vehicles [LDVs], where the use phase constitutes about only one-third of the damages.) The transportation sector overall (including both light-duty and MHDVs) is a very large contributor to some of these criteria air pollutant emissions: the 2015 Transportation Energy Data Book (Davis et al., 2015) reports that in 2014 the transportation sector accounted for 54 percent of total U.S. carbon monoxide emissions, 58 percent of total U.S. NO x emissions, and 23 percent of total U.S. VOC Prepublication Copy – Subject to Further Editorial Correction 12-18

emissions. The transportation sector is a smaller contributor to overall national PM emissions (accounting for less than 3 percent of national PM 10 emissions and less than 6 percent of PM 2.5 emissions) and also a small contributor to SO 2 emissions (about 2 percent of total U.S. SO 2 emissions are attributable to the transportation sector). Also, just considering the transportation sector, heavy-duty vehicles accounted for about 50 percent of total NO x highway emissions, 55 percent of PM 10 highway emissions, and 66 percent of PM 2.5 highway emissions in 2005. 167 Depending on where these emissions occur and their exposure pathways, they may, in turn, lead to serious health and environmental impacts, such as increased human mortality and morbidity. According to report Hidden Costs of Energy (NRC, 2010b), in year 2005 the vehicle sector overall accounted for $56 billion in health and other non-climate-change damages, with $20 billion of these coming from heavy-duty vehicles. As a result of the mix of strategies, technologies, and fuels used under the Phase II standards the level of emissions from these pollutants and their ambient concentration will change. The emissions of these pollutants have health and environmental effects that are important externalities to consider as part of the overall cost-benefit analysis of the Phase II standards. That task, however, is challenging because— unlike GHG emission damages—the health damages associated with the emissions of criteria air pollutants from MHDVs will be very dependent on where these emissions occur, the ambient baseline concentration of those pollutants, and who is exposed to them. Estimates of the damages associated with these mobile sources generally rely on integrated assessment models that relate the emissions sources to dispersion and reaction models, and to exposure to population, dose-response functions, and a monetized valuation of the health effects. All these modeling and estimation steps have deep uncertainties, and nationally disaggregated data on actual emissions and where these occur are lacking. Indeed, almost all of the existing studies have focused on estimating the damages associated with the light-duty vehicle fleet, and estimates for the MHDVs are lacking (with the possible exception of NRC [2010b], in which point estimates for the air pollutant damages were provided for MHDVs). Despite the deep uncertainties, the existing studies often use only point estimates for these damages. This is problematic, as a few key and uncertain assumptions regarding the monetized valuation of the health effects can completely drive the results (for example, the values assumed for the statistical value of life, or the discount rate). The NRC report Hidden Costs of Energy attempted to estimate these damages, and found that for the medium- and heavy-duty vehicles, depending on the vehicle and fuel considered, damages ranged from about $0.03 to $0.10 per VMT in 2005, and would be reduced to less than $0.03 per VMT for all studied vehicle technologies in 2030 in light of ongoing or proposed policies (see Figures 12-4 and 12-5). 167 These figures are for year 2005. These estimates are computed using the 2015 Transportation Energy Data Book from the Oak Ridge National Laboratory. More recent data with a breakdown at the medium- and heavy-duty disaggregation level is not reported. Nonhighway emissions are also not reported at the heavy-duty level. Prepublication Copy – Subject to Further Editorial Correction 12-19

FIGURE 12-4 Aggregate operation, feedstock, and fuel damages of heavy-duty vehicles from air-pollutant emissions (excluding GHGs) in cents per VMT. Panel (a) shows the estimated damages in 2005 and panel (b) shows the estimated damages in 2030. The figure is reproduced from the NRC report Hidden Costs of Energy (Figure 3-9 from that report). NOTE: HDDV, heavy-duty diesel vehicle; HDGV, heavy-duty gasoline vehicle. SOURCE: NRC (2010b). Prepublication Copy – Subject to Further Editorial Correction 12-20

The 2016 Regulatory Impact Assessment includes the assessment of changes in damages from carbon monoxide (CO), fine particulate matter (PM 2.5 ), SO x , VOCs, NO x , benzene, 1,3-butadiene, formaldehyde, acetaldehyde, and acrolein. Importantly, the RIA includes vehicle emissions (i.e., downstream emissions) as well as emissions from fuel production and distribution (i.e., upstream emissions, using GREET as the source of information for these, updated to account for the most recent National Emissions Inventory). The RIA includes different strategies to model the emissions impacts of the different MHDV classes: for tractor-trailers and vocational vehicles, the EPA Motor Vehicle Simulator model is used, whereas for heavy-duty pickups and vans they used two different methods to estimate the emissions (we refer the reader to the RIA for additional details on these).The RIA then models the emissions consequences in terms of changes in damages using an air quality model, using year 2040 for their analysis where a reference scenario in which no standard is implemented is compared to a scenario with the standards. The overall health and environmental effects from these pollutants associated with the standards are found to be driven by (i) the increased use of auxiliary power units, (ii) a decrease in emissions from the upstream fuel production and distribution, and (iii) a reduction in the damages due to a reduction in emissions associated with “the improved engine efficiency, aerodynamics and tire rolling resistance and absolute changes in average total running weight of the vehicles” (NHTSA and EPA, 2016). 12.2.3 Finding Finding: Some attention has been provided to estimating the changes in health and environmental consequences due to criteria air pollutants and air toxics, as well as other pollutants, to the Phase II standards in the 2015 draft RIA and in the final 2016 RIA, suggesting that the changes in health and environmental damages associated with the standards are nontrivial. However, the magnitude of these effects is still uncertain, and the analysis under the final RIA produces only a first step in that direction. 12.2.4 Recommendation Recommendation 12-2: Since the health and environmental benefits from reductions of criteria air pollutants and air toxics may constitute an important portion of the social benefits from the standards, NHTSA and EPA should more carefully assess its value, and consider doing an assessment that includes the monetization of the costs and benefits associated with the changes in criteria air pollutants and air toxics. The committee also urges NHTSA and EPA to consider doing and reporting a detailed sensitivity and uncertainty analysis for these results of ongoing benefits analyses. Prepublication Copy – Subject to Further Editorial Correction 12-21

FIGURE 12-5 Aggregate operation, feedstock, and fuel damages of heavy-duty vehicles from GHG emissions (cents per VMT): (a) estimated damages in 2005; (b) estimated damages in 2030. SOURCE: NRC (2010b). 12.3 NATIONAL SECURITY EXTERNALITIES 12.3.1 Introduction In addition to the environmental externalities associated with oil production and consumption, since the oil crisis of 1973-1974 there has been concern with potential national security externalities associated with oil consumption. Oil supply disruptions and price shocks have preceded most of the post–World War II recessions (Hamilton, 2009). It was the recession of 1973-1974 following the Arab oil embargo that led to the Energy Policy and Conservation Act (EPCA) passed by Congress and signed by President Ford in December of 1975. Among the provisions of EPCA was the establishment of fuel economy standards for Prepublication Copy – Subject to Further Editorial Correction 12-22

light vehicles. The motivation behind the law was the vulnerability of the U.S. economy to oil supply disruptions originating in the Middle East. 12.3.2 The Oil Security Premium There are several aspects to the economic vulnerability associated with petroleum use. First is the negative impact of unanticipated oil price increases on economic activity. Many economic researchers have explored the link between an oil price shock and an economic downturn. Killian (2014) summarizes the most recent research. The source of the negative impact, while not fully understood, revolves around the adjustment costs consumers and businesses incur in response to unanticipated oil price shocks. 168 The disruptive effect on the economy depends on the extent of the price increase and the level of oil consumption, regardless of where the oil was produced (Brown and Huntington, 2013). Oil markets are world markets and a disruption anywhere in the world affects prices throughout the market. In spite of the fact that disruption causes prices everywhere to increase, the bulk of concern over the years since the first oil shock has been with import dependency. The special concern with imports arises because a large fraction of world production comes from unstable parts of the world. As imports rise, likely so does production in these unstable regions, increasing the effect of a production disruption on the world price of oil. Furthermore, the larger is the imported fraction of consumption, the larger is the transfer of economic resources abroad in the event of a price shock. It is for these reasons that some researchers present separate estimates of the security premium for domestic and foreign production (Brown and Huntington, 2013). It is controversial as to whether the macroeconomic disruption effects are true externalities and represent a market failure. Perry and Darmstadter (2003) point out that to the extent that private actors can take into account future disruption costs and insure against them, there is no market failure. A recent National Research Council Report (NRC, 2010b), in fact, concluded that these disruption costs are not externalities. Other researchers have taken an agnostic point of view and, as pointed out in Perry and Darmstadter (2003), have modeled the external costs parametrically, letting the fraction of instability costs that are anticipated by individuals vary. Another aspect of oil markets has received attention in the economics literature. The United States as a whole is a large consumer of oil. Were the United States to curtail its consumption, the fall in demand would lead to lower world oil prices. This is known in the economics literature as “monopsony power.” The gain from the price reduction would accrue to all U.S. oil imports remaining after the reduction in consumption. The price reduction on nonimported consumption is not a net gain to the United States since the gains to consumers are offset by the losses to domestic producers. The extent of monopsony power and the value of the reduction has changed over time as the United States has increased its domestic production and reduced imports as a share of consumption, and as other regions of the world have increased their relative proportion of world consumption. The recent decline in oil prices also has lowered the value of the monopsony premium. For example, in the Regulatory Impact Analysis that accompanied the first phase of regulations in the medium- and heavy- duty trucks, NHTSA, based on research by Leiby et al. (1997) and updated in 2008 (Oak Ridge National Laboratory, 2008), estimated the monopsony premium for the year 2020 at $12.07 per barrel (2013 dollars). The Regulatory Impact Analysis accompanying the Phase II rule (NHTSA and EPA, 2016), using the same methodology as the earlier estimate, put the 2030 monopsony premium at $2.21 per barrel (2013 dollars) (NHTSA, 2010, p. 9-41, 2015, p. 8-77). It is widely recognized that the monopsony premium does not constitute an externality in oil markets. Rather, the issue is the distribution of gains and losses between oil producers and oil consumers. The gains to the United States from exercising monopsony power would come from losses to oil 168 Killian (2014, p. 144) points out that adjustment costs apply to large price decreases as well as to price increases. However, price decreases result in gains to consumers and resulting increases in consumer spending that serve as at least a partial offset to the adjustment costs and their effect on GDP. Prepublication Copy – Subject to Further Editorial Correction 12-23

producers. While not an externality, the potential distributional gains to the United States have led some to call for the inclusion of the monopsony premium in an analysis of benefits to the United States from reducing oil consumption. Whether or not to introduce measures to exploit this power is a policy choice of the U.S. government. In the recent Regulatory Impact Analysis (NHTSA and EPA, 2016), NHTSA and EPA argue that they are taking a global view when establishing policy to deal with climate change. The agencies argue that climate change is a global phenomenon and calculate policy benefits on a global basis. As a result, it would be inconsistent, in their view, to calculate monopsony benefits on a regional basis. In contrast, a recent National Research Council report argues that the United States should count the monopsony benefits as part of the benefits to fuel economy regulations while maintaining a global focus for calculating climate benefits (NRC, 2015). 169 Finally, it has been observed by many (e.g., NRC, 2010b) that there are potential externalities associated with military expenditures and foreign policy goals associated with protecting oil production and shipping lanes. Nevertheless, most observers conclude that it is very difficult to isolate the increased military expenditures or to quantify the value of the political constraints oil consumption may place on the United States. The NRC (2010b p. 336) pointed out that “a 20 percent reduction in oil consumption, for example, would probably have little impact on the strategic positioning of military forces in the world.” 170 12.3.3 Estimates of the Security Premium There have been many attempts at estimating the security premium associated with oil (see Brown and Huntington [2013] for a list of previous studies). The value of the premium depends on the probability and size of potential disruptions; the impact of the disruption on the world price of oil, which in turn depends on the elasticity of demand and the elasticity of supply of the nondisrupted oil; the elasticity of GDP with respect to oil price increases; the level of the price of oil; and the level of imports. These factors vary over time as do changes in oil market institutions so the premium also varies over time. Recent developments in oil markets likely have lowered the security premium. The most significant of these developments is the rapid expansion of oil production in the United States from tight formations—“fracking.” This expansion has lowered the share of imports in total consumption and has increased the elasticity of supply in the United States.171,172 These production developments together with attempts of the Organization of the Petroleum Exporting Countries (OPEC) to maintain a particular level of world market share led to a dramatic decrease in the level of oil prices, falling from over one hundred dollars per barrel in 2008 to below thirty dollars in early 2016. Also over time the intensity of oil use in the United States economy has declined, likely lowering the impact of any unanticipated increase in price on overall economic activity. Recent research on the relation of oil price changes to GDP changes has also shown a reduction in the impact of oil price changes (Killian, 2014). Of course, oil prices have been 169 The climate benefits to the United States have been estimated to be between 7 and 23 percent of total benefits (see Gayer and Viscusi, 2013, p. 252). The March 28, 2017, executive order issued by President Trump rescinds the Obama administration guidance on the use of the social cost of carbon in rulemaking and relies instead on previous guidance. One result of this order will be to estimate costs and benefits on a U.S.-only basis. It remains to be seen how this will affect the treatment of the monopsony premium in future rulemakings (Dudley, 2017). 170 This opinion is shared by NHTSA and EPA: “US military costs are excluded from the analysis performed by ORNL because their attribution to particular missions or activities is difficult and because it is not clear that these outlays would decline in response to incremental reductions in US oil imports” (NHTSA and EPA, 2015, p. 8-85). 171 Hydraulic fracturing—“fracking”—technology results in big increases in production in the first year of production. That means that the response to price increases can be more rapid than is the case with conventional production methods. Recent reports of producers that have drilled wells are not producing the oil but waiting for prices to rise(see Mills, 2016). 172 The import share of total petroleum products supplied is forecast by the 2015 Annual Energy Outlook to decline from 33 percent in 2013 to 14.7 percent by 2025. Prepublication Copy – Subject to Further Editorial Correction 12-24

volatile over the past decades and looking ahead it is reasonable to expect many of these factors to change. Table 12-5 below shows estimates of the premium made by various authors expressed in 213 dollars. The table also shows the date the estimate was made and the range of the estimates. Where possible, separate premiums are shown for domestically produced oil and for imports. In today’s circumstances the domestic premium is particularly relevant. This is because reductions in consumption in today’s circumstances are as likely to result in the reduction of domestic production as in the reduction of foreign production. OPEC and, in particular, Saudi Arabia have shown determination to maintain production in the face of demand reductions. Looking to the future, there is no guarantee this will continue so the relevant premium in, say, 2025 would be some weighted average of the domestic and import premiums calculated at the market conditions prevailing at that time. TABLE 12-5 Recent Estimates of the National Security Premium (2013 dollars per barrel) Monopsony Year Being Domestic Premium Import Premium Premium Source Forecasted (range) (range) (range) NHTSA and EPA 2020 NA 7.60 12.07 (2010) (5.92-12.72) (4.13-22.79) Brown and 2013-2014 3.07 5.27 NA Huntington (2013) (.37-9.33) (1.16-15.26) NHTSA and EPA 2030 NA 7.26 2.83 (2016) (3.40-11.73) (.83-4.56) NOTE: The estimates in the NHTSA reports are based on an original model by Leiby et al. (1997), updated to conditions of 2014 by Leiby (2015). All estimates are converted to 2013 dollars using the GDP deflator. Table 12-5 demonstrates the wide band of uncertainty that surrounds these estimates. The estimates depend on estimates of various elasticities that vary over time, as well as the price of oil. More recent research, for example, suggests the effect on GDP of oil shocks is smaller than previous estimates that relied on the experience of 1973-1974 and 1979 (Killian, 2014). Earlier estimates also did not capture the full effect of the large expansion of oil production in the United States and the dramatic declines in crude oil and petroleum product prices. The last line of the table shows the latest estimates by NHTSA, used in the Phase II rulemaking. The median estimate of $7.26 per barrel is the equivalent of $0.17 per gallon, which is small relative to the volatility in gasoline prices over the past few years. 12.3.4 Findings Finding: Oil consumption makes the U.S. economy vulnerable to supply disruptions. At least some of the effects of these disruptions are likely to be unanticipated and uninsured against by private firms and consumers. Finding: To the extent that consumption is satisfied through imports, supply disruptions and consequent price increases cause the transfer of economic resources from U.S. consumers to foreign producers. Therefore, the security premium associated with imports is larger than the premium associated with consumption satisfied by domestic production. Finding: Recent developments in world oil markets have reduced the vulnerabilities of the U.S. economy to supply disruptions. Most significant is the development of “fracking” technology and the large increase in U.S. domestic oil production. This has increased the ability of the United States to respond to increased prices and reduced the level of imports. Furthermore, OPEC has shown in the recent past a desire to Prepublication Copy – Subject to Further Editorial Correction 12-25

maintain market share and to accommodate the U.S. production increases and the declines in world demand by maintaining their production rates and allowing prices to fall. Finding: Over the long term it is likely that volatility in oil prices and market institutions will continue to alter the value of the security premium associated with oil consumption and imports. Finding: The U.S. government has made a policy choice not to consider any monopsony benefits from reducing U.S. oil consumption and, for the Phase II regulations, to take a global view when evaluating the costs and benefits of actions aimed at reducing climate change. In any case, the estimates of the monopsony premium have declined substantially in recent years. 12.3.5 Recommendation Recommendation 12-3: The committee believes that a security premium reflecting macroeconomic instability and transfers caused by potential supply disruptions are a part of the social cost of consuming and importing oil. It is important to separate the social costs imposed by consumption from the costs imposed by imports. The total cost will be a weighted average of the two premiums and will vary over time with the proportion of consumption satisfied by imports. 12.4 PROJECTING TOTAL CAPITAL COSTS 12.4.1 Indirect Cost Estimation Cost estimation, whether for regulatory purposes or for other reasons, makes a careful distinction between direct and indirect costs. Estimation of both cost categories is essential both for the manufacturer, which needs to know how to price a product to cover costs and generate an acceptable profit, and for the regulator, who is required to make estimates of the costs and benefits of regulation, including costs of compliance. The distinction between indirect and direct costs is closely related to the relationships between two different types of inputs in the production process. For some inputs—those exhausted or incorporated into a unit of the product at the time of production and therefore not available to produce another unit— there is a strict, quantifiable relationship between inputs and outputs. Examples in vehicle manufacturing would include the production labor hours, parts and materials, etc. Multiplying the quantity of each direct input by its unit cost and summing provide an estimate of the direct costs of the vehicle. 173 However, many inputs to the production process, as well as all other corporate activities with costs that must be covered by sales revenues, do not bear this simple or readily observable relationship to outputs, at least not for individual products. These are the inputs that contribute to indirect costs. Foremost among these input categories are the production inputs, which EPA’s Office of Transportation and Air Quality (OTAQ) calls production overhead, broken out as follows: • Production overhead (0.15) o Warranty (0.04) o R&D (0.05) o Depreciation (0.04) o Maintenance, repair, and operations (0.02) 173 Note that a manufacturer’s direct costs may include the indirect costs of parts purchased from suppliers. The fundamental distinction made above still applies (NRC, 2015). To the manufacturer, the cost of purchased parts is precisely known. Prepublication Copy – Subject to Further Editorial Correction 12-26

As is common in matters of classification, designations of direct or indirect costs are somewhat arbitrary. Warranty costs, for example, may be associated with an individual vehicle, but the relationship is stochastic, not fixed, and the costs are incurred at some indeterminate time after the vehicle is sold. Indirect costs are not limited to production overhead. The costs of essential nonproduction activities generate no revenue, at least not directly, and must also be covered by sales revenues. The nonmanufacturing categories and subcategories that enter into OTAQ’s indirect cost calculations include the following (the numbers in parentheses are explained below): • Corporate overhead (0.09) o General administrative: personnel, legal, etc. (0.07) o Retirement and health care expenses (0.02) • Sales, including delivery of new vehicles to dealers, and advertising (0.08) o Vehicle transport to dealers (0.005) o Marketing (0.01) o Dealer services, including sales, training of sales personnel, and net income to dealers for new vehicle sales (0.07) • Net income to manufacturer (0.05) Of course, total indirect costs are reasonably easy to calculate: just identify the costs that are direct and subtract them from total costs (plus net income). Corporate annual and quarterly reports, as well as filings to the Securities and Exchange Commission, generally contain adequate information to do this. The difficulty comes in apportioning the total indirect costs in each of the categories and subcategories above to a multitude of individual products. That job, in turn, requires a vast amount of detailed cost and production data that are generally proprietary and not available to the regulator. Because of these difficulties, it is common to resort to a simpler calculation of indirect cost that uses readily available data: the “retail price equivalent” (RPE). The RPE is defined as the ratio of total costs (direct plus indirect costs including net income or the cost of capital) to the total direct costs, thus allowing total costs to be estimated from direct costs. The numbers alongside the indirect cost categories and subcategories in the bulleted lists above indicate the contribution of that category to the RPE for medium- and heavy-duty (MHD) trucks. Adding in 1.0 for the production sector, OTAQ’s estimate of the total RPE for MHD trucks is 1.36.174 According to Rogozhin et al. (2010b), the RPE is commonly used in the motor vehicle industry for this purpose. OTAQ has also estimated the RPE for a variety of light-duty vehicle manufacturers; the average RPE is 1.46, with a range from 1.43 to 1.49 (Rogozhin et al., 2010). It has also been used for heavy-duty vehicles. In the MHD truck industry overall, OTAQ calculated a mean RPE of 1.42, using data from the U.S. Census and Supplier Relations LLC, a data firm serving the industry. OTAQ also estimated RPEs for several engine and truck manufacturers, with averages of 1.28 (engines) and 1.36 (trucks). The engine RPEs tend to be lower because, for the most part, engine manufacturers do not sell directly to the end user, which eliminates part of the retail component of the RPE. Despite the ease of calculation and wide use of the RPE in the motor vehicle industry, researchers at OTAQ have argued that it has two problematic features. First, OTAQ claims that the RPE is sensitive to the time frame, as many indirect cost activities are short term in nature and do not figure in long-term cost: Many of the indirect costs are likely to be one-time or short-term activities, such as educating dealers and upgrading mechanics’ equipment. These costs will not appear in the long-term IC multipliers. In addition, incremental R&D expenditures will occur over a short period of time, even though they may be amortized over 5 to 10 years. Thus, we 174 The sum of the individual components does not sum to the 1.36 due to rounding. Prepublication Copy – Subject to Further Editorial Correction 12-27

expect to see higher indirect costs initially and lower impacts in the long term as companies assimilate the new technologies (Rogozhin et al., 2010a). For present purposes, however, the time frame is not an issue. We are primarily interested in the incremental long-term costs of regulation, and, in the long run, all costs are incremental. A more important issue with the RPE, however, is that it implicitly assumes that the ratio of indirect cost to direct cost is identical for every product produced by the firm. While this has to be true on average—the volume-weighted average of the indirect cost multipliers (ICMs) must equal the RPE— since all costs need to be accounted for, nonetheless, it is easy to imagine how a disproportionately outsized demand on overhead could be exerted by products that are new, complicated, or difficult to assemble or, for whatever reason, use up more than their share of management and administrative resources, and similarly, how they could disproportionately affect particular indirect cost (IC) subcategories. To address these issues OTAQ developed a method for producing ICMs that are more flexible than the RPE and in particular more sensitive to both the technology involved and the specific impact on individual corporate subunits. In the end, three levels of technology—low, medium, and high—were examined. Office of Mobile Sources identifies low-level complexity as a technology that introduces only minor changes in component design and does not affect the way various components are linked together. Medium-level complexity either changes the design substantially or changes the way the components are used or fit together. High-complexity technology affects both design and integration of components. Examples of the three complexity levels are single-wide tires, engine turbocompounding, and hybrid- electric powertrains, respectively. To accomplish these goals, a set of “adjustment factors” was generated, one factor for each combination of indirect cost subcategory and complexity level, to be used to adjust the RPE contribution from that subcategory. The adjustment factors were to be calibrated so that that a factor of 0 would indicate that a direct cost would have no effect on indirect cost in that subcategory, while a factor of 1 would indicate an average effect (i.e., the same as the subcategory RPE). Table 12-6 shows the IC multipliers for three types of vehicles—LDVs, MHD engines, and MHD trucks—and three levels of technology complexity. To estimate the adjustment factors to use with the HDV Fuel Consumption Rule, OTAQ convened a panel of engineers with considerable experience in the vehicle manufacturing industry, and tasked the group with the development of such factors in a group setting. Two different group mechanisms were used: a consensus process where the group had to arrive at a single set of values, and a Delphi process where participants were made aware of the estimates of others in the group and then allowed to change their estimates in response to the opinions of others. The Delphi process was concluded after three iterations. Finally, the adjustment factors coming out of the two processes (consensus and Delphi) were averaged to obtain the final estimates for the low- and medium-complexity technologies. For the high-complexity technologies, the results were different enough that it made sense to report them separately, as High1 (for consensus) and High2 (for Delphi). The exercise was carried out for two engine and four truck manufacturers and averaged within the groups. The multipliers are consistently lower for MHD engine manufacturers than truck manufacturers, perhaps due to the greater likelihood that engine manufacturers sell directly to end users. Also, for each vehicle category, the ICMs resulting from the expert elicitations conducted by EPA are smaller than RPEs at all technology levels, and they are particularly low for low- and medium-complexity technologies. (In many of the subcategories listed above, in fact, the adjustment factor is zero.) Upon reflection, this seems strange. Indeed, in the RPE-ICM comparisons shown in Table 12-6, each RPE is greater than all of the component ICMs in each row, which implies that the RPE will be greater than the average ICM, Prepublication Copy – Subject to Further Editorial Correction 12-28

regardless of the weights. As noted above, that cannot happen. 175 A more fundamental question is, just what are the entities that ICMs get attached to? It does not seem quite right that it be attached to a “technology,” as OTAQ has done, because various applications of the same technology could in principle give rise to different ICMs. Instead, it seems more useful to think of the vehicle (or engine, or whatever) as a bundle of components, which include a mix of technologies of varying complexity, and calculate ICMs that can be attached to tangible products or assemblies of products. For the cost-estimating purposes of this NASEM committee, it seems safest to continue to use RPEs and ensure that long-run costs are covered. Like the LDV committee, nonetheless, this committee recognizes the cost estimation problem with RPEs that NHTSA and EPA are trying to deal with, and urges them to continue their research on ICMs, especially to develop more solid empirical support for particular ICM values. TABLE 12-6 Comparison of RPE and IC Multipliers IC, by Technology Complexity Category RPE Low Medium High1 High2 LDV 1.46 1.02 1.05 1.26 – MHD engine 1.28 1.04 1.08 1.14 1.24 manufacturer MHD 1.36 1.05 1.11 1.22 1.31 truck manufacturer MHD sample – Single-wide Engine Hybrid-electric powertrains technology tires turbocompounding 12.4.2 Findings Finding: Accurate calculation of indirect costs is just as important to regulatory and cost-benefit analysis as is direct cost estimation. While total indirect costs, and the average indirect cost associated with each unit of output, are just as easy to calculate as direct costs are, their calculation on a product-by-product basis can be much more difficult because of the lack of a clear relationship between inputs and outputs. The RPE method, a standard method for calculating ICs, ignores this problem by assigning the same IC multiplier to all products. Finding: Recently researchers at EPA and NHTSA have been working to develop a new method, which they call ICM, to estimate indirect costs on a more detailed product- or technology-specific basis. In principle, this makes a good deal of sense, for there are plenty of reasons to expect diversity across products in the use of inputs that give rise to indirect costs. Considerable progress has been made in developing this new methodology; however, work remains to be done. Finding: In order to produce more detailed cost breakdowns, the current ICM methodology relies heavily on expert opinion in the absence of product-specific empirical data. In addition, in some circumstances the current ICM methodology gives results that appear to be inconsistent, as shown in the example above where the RPE is calculated to be greater than each of the constituent ICMs. 12.4.3 Recommendations Recommendation 12-4: The committee encourages NHTSA, in coordination with EPA, to further development of the ICM methodology, which promises to provide more flexible and accurate estimates of indirect cost multipliers. Where possible, these estimates should be validated with observations of actual incremental costs of changing product designs. For particularly important cost regulations or cost 175 This point was also made by the study of light-duty vehicle fuel economy by the National Academies (NRC, 2015). Prepublication Copy – Subject to Further Editorial Correction 12-29

estimates, two or more analytical methods could be used to develop a range of cost estimates that could be compared with one another. For example, expert elicitation could be compared with observed cost changes in similar prior regulations or bottom-up engineering analyses. Such parallel analyses would automatically generate a useful sensitivity analysis. Recommendation 12-5: In order to improve all cost estimates, direct or indirect, NHTSA and EPA should make a habit of commissioning retrospective studies of costs of past regulations. Again, a range of analytical methods is available. One that might be of particular interest for indirect costs is hedonic estimation, a statistical method that allows bundled cost estimates to be separated into individual component estimates. That property makes it potentially useful for producing estimated indirect cost components across a range of product categories. 12.4.4 Learning Effects on Capital Costs 12.4.4.1 Learning Curves Repetition drives learning. In business and economics the importance of this fundamental fact is enshrined in the concept of the learning curve. Simply put, the learning curve asserts a declining relationship between the cost of a new product produced by a firm and the firm’s cumulative output of the product. Thus, estimation of the costs of regulation, which usually implies adoption of new or modified products or processes, need to take the learning curve into account. New and emerging technologies used in medium- and heavy-duty vehicle costs and performance will likely evolve over time as a function of several different factors. Policy makers, modelers, and analysts are generally concerned with the effects of transitions to different MHDV technologies and strategies, and how these may influence the outcomes of policies such as the Phase II standards. To describe how the costs of technologies of strategies will perform in the future, energy economic modelers have frequently used the concept of learning or experience curves. These curves can be estimated in the context of retrospective analysis, where the modeler uses historical costs of a technology, a component or a strategy, and assesses how those costs relate to other factors—most commonly the cumulative number of units produced or accumulated service provided. Learning curves are also used in forward-looking modeling efforts, where the modeler assumes specific learning rates for different technologies (often informed by past historical learning rates when those estimates exist) to understand how the share and adoption of different technologies is likely to evolve in the future. The origin of learning curves dates to developments in the psychology of learning in the late 19th century, especially in the work of the German psychologist Hermann Ebbinghaus. Wright (1936) first showed an empirical log-linear relationship between unit costs and cumulative production, in the context of the production of airplanes, and was further supported by experiences in the vast buildup of munitions industries during World War II (Hindle, 2008). Additional development occurred in the early 1960s at the Boston Consulting Group, based on the firm’s experience advising a semiconductor manufacturer. This extraordinarily simple log-linear relationship has shown to hold for a variety of products and technologies. It has been denoted as the “one-factor learning curve,” and it is by far the most widely used model in the energy economics literature to understand how unit costs have evolved (Rubin et al., 2015). The one-factor learning curve relates the unit costs of production with cumulative production, using a log- linear functional form: Y = axb where Y is the unit cost of production, x is the cumulative experience (often represented by the cumulative number of units produced), a is the unit production cost of the first unit, and b is a constant reflecting the rate of cost reduction (Arrow, 1962). Modelers frequently refer to the learning rate, which represents that fractional cost reduction associated with a doubling of units produced, or 1 – 2b. Assuming the relationship Y = axb holds, then a doubling of capacity leads to the unit costs of Y 2x = a(2x)b and therefore the change in costs associated with a doubling of units produced is Prepublication Copy – Subject to Further Editorial Correction 12-30

𝒀𝒀 𝒙𝒙 − 𝒀𝒀 𝟐𝟐𝟐𝟐 𝒂𝒂(𝒙𝒙) 𝒃𝒃 − 𝒂𝒂(𝟐𝟐𝟐𝟐) 𝒃𝒃 𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆𝒆 𝒆𝒆 𝒆𝒆 𝒓𝒓𝒓𝒓𝒓𝒓𝒓𝒓 = ↔ ↔ 𝟏𝟏 − 𝟐𝟐 𝒃𝒃 𝒀𝒀 𝒙𝒙 𝒂𝒂(𝟐𝟐𝟐𝟐) 𝒃𝒃 Intuitively, the one-factor learning or experience curve means that as more units are produced, the firm learns how to reduce the production costs per unit, and empirically it has been shown that this relationship of decreasing costs resulting from learning as the firm produces more units is well described by a log-linear relationship. For example, a 15 percent learning rate (or an 85 percent progress ratio) states that unit costs decline by 15 percent for every doubling of cumulative output. Studies in various settings have found great variation in the learning rates. A survey by Dutton and Thomas (1984) found average progress ratios of 82 percent (i.e., learning rates of 18 percent), with a range from 55 to 95 percent (learning rates of 5 to 45 percent). More recent work, cited by Taylor and Fujita (2013), found similar results, with an additional finding that within-technology progress ratios varied as much as between-technology ratios. This finding is somewhat unsettling, as it suggests that the accuracy of a learning-curve assumption in a particular context could depend on quite detailed information. It should also be noted that the logarithmic functional form generally employed in learning curves implies that the cost reductions continue indefinitely, approaching zero as an asymptote. As a practical matter, however, the period of learning is in most studies considered finite. For electricity generation, Rubin et al. (2015) reviewed peer-reviewed existing studies on learning curves for 11 electricity-generation technologies and found that there was a wide disparity in one- and two-factor learning rates even within the same technology, time period, or region. Closely related to the learning-curve concept is that of the experience curve. The chief difference between the two is that the learning curve refers only to the relationship between manufacturing costs and the volume of production. The experience curve is broader, relating total costs to cumulative sales (Abernathy and Wayne, 1974). For both, the decline in costs is not automatic, but comes only with application of management expertise and attention to the opportunities available. Thus, the potential sources of cost reduction extend far beyond the growth in speed and efficiency that comes with the simple familiarity with the task at hand. As Abernathy and Wayne (1974) note: The producer cuts costs with a combination of effects; these include spreading overhead over larger volume, reducing inventory costs as the process becomes more rational and throughput time drops, cutting labor costs with process improvements, achieving greater division of labor, and improving efficiency through greater familiarity with the process on the part of the work force and management. Dutton and Thomas (1984) lump the categories of improvement into two broad classes. The first is autonomous change, resulting from learning by doing. The second category is called induced change, resulting from conscious efforts of management, including such measures as worker training and specialization, R&D investments, and streamlining of production processes. In the one-factor learning curves the number cumulative units produced is being used as a de facto surrogate for all the factors that influence that technology’s cost (Rubin et al., 2015). Indeed, while offering the benefits of a simple model, these simple one-factor learning curves provide no explanation for the underlying factors and processes of technological change or the causality between them (Clarke et al., 2006; Ferioli et al., 2009; Gillingham et al., 2008, Nordhaus, 2009). Modelers have tried to circumvent some of these issues by using more detailed models—multifactor learning models that explicitly incorporate other factors that may be determinants of unit cost changes. Rubin et al., (2015) reviewed the studies for electricity supply technologies and found that previous studies have included factors such as R&D spending (Clarke et al., 2006; Cohen and Klepper, 1996; Jamasb, 2007), knowledge spillovers (Clarke et al., 2006), increased capital investments (Cohen, 1995; Klepper and Simons, 2000), economies of scale (Nemet, 2006; Sinclair et al., 2000; Yeh and Rubin, 2007), changes in input prices (Joskow and Rose, 1985; Nemet, 2006; Söderholm and Sundqvist, 2007), labor costs (Joskow and Rose, 1985), efficiency improvements (Joskow and Rose, 1985; Nemet, 2006), and other public policies (Söderholm and Klaassen, 2007; Söderholm and Sundqvist, 2007). However, lack of data on these explanatory factors often prevents their use in more detailed learning models. Prepublication Copy – Subject to Further Editorial Correction 12-31

Researchers have found that when R&D is explicitly included as one of the factors contributing to the decrease in unit costs, it is generally a significant explanatory variable (Jamasb, 2007; Klaassen et al., 2005; Kouvaritakis et al., 2000; Söderholm and Klaassen, 2007; Söderholm and Sundqvist, 2007; Watanabe, 1995). Most often, only public R&D is included in such assessments, given the lack of private R&D data. Several studies also explicitly include the enactment of specific policies that may influence costs as controls (most generally as a dummy variable). Researchers have also used variants of the above-mentioned functional forms. Some authors considered component-level learning (i.e., in the case MHDVs, this would mean looking separately at learning for the engine, battery, etc.), or including also factors such as raw materials and labor. Rubin et al. (2007) used the component-based approach for power plants with carbon capture systems. Similarly Weiss et al. (2010) assessed learning effects using component-level analysis for microcogeneration of heat and power. The logic behind this strategy is that for complex multicomponent technologies, different components are likely at different levels of maturity, and therefore have different potential for cost reductions. Finally, for some technologies, such as nuclear, some researchers have actually found negative learning rates—the unit costs actually increased as the cumulative production increased (see, for example, Grubler [2010] and Cooper [2010]). Even for those technologies (such wind and solar) where there is a substantial literature on learning rates, a systematic understanding of why different studies achieved fundamentally different estimates for learning, as well as the implications of assuming different learning rate values in policy analysis, is still lacking. This issue is even more pronounced in the MHDV sector due to several factors. First, the industry is highly segmented and the technologies and components used by different subsectors of the fleet are highly diversified. Indeed, given the diversity of the MHDV fleet it would likely make more sense to think about learning curves at the firm level. Second, data on historical costs and units produced are lacking for the sector. This makes any estimates of learning rates for MHDV unreliable. Third (and this point holds for any prospective use of learning rates), the past may not be a good predictor of the future, namely, in a sector where the technologies and fuels are likely to change over time as a result of the policy. Learning and experience curves are certainly useful descriptive tools for describing the reduction in unit costs over time. But can they be considered predictive? The long list of potential causes of unit cost reductions, as well as the wide range of learning rates found in empirical studies and the uncertainty surrounding the duration of learning, seems to argue against such a notion (Abernathy and Wayne, 1974; Taylor and Fujita, 2013). Nonetheless, learning and experience curves have been used in RIAs to predict cost reductions of federal regulations for almost 20 years, dating back at least to EPA’s 1997 Heavy-Duty Emissions Standards (EPA, 1997). NHTSA began to incorporate a similar but not identical learning-curve analysis into RIAs of fuel economy standards that paired with EPA’s greenhouse gas emission standards after 2000 (Taylor and Fujita, 2013). The predictive value of learning curves may depend not only on the characteristics of the technology, but on the business strategy pursued by the firm. There is, after all, a seeming paradox at the heart of learning-curve analysis. In principle, the learning curve ought to give a huge advantage to a first mover, whose cumulative production is likely to be considerably greater than its rivals, and whose production costs are correspondingly lower. Thus, the first mover in an industry should frequently be the only firm. And yet, most industries contain numerous competitive firms. To resolve the paradox, note that there are other attributes of a firm’s products besides costs that consumers care about. By adopting a strategy of pursuing rapid cost reduction above all else, however, the firm may foreclose other strategies that prize innovation and feature product quality and consumer satisfaction over price. The very definition of innovation, after all, is to do something different, not to do the same thing better. Abernathy and Wayne (1974) illustrate this point by a short case study of the Ford Model T, where the company’s exclusive focus on cost reduction of a standardized product put the company in a box where it was producing a very low-cost product that consumers did not want. Once it became fully aware of the problem, the company was forced to expend huge sums and considerable time on designing Prepublication Copy – Subject to Further Editorial Correction 12-32

new products and retooling its plants, in particular causing the giant River Rouge plant to sit idle for more than a year. This debacle almost wrecked the company. Fourth, as explained in the draft RIA by NHTSA and EPA (2015), studies on learning or experience curves relate costs to cumulative production only after an initial volume threshold is achieved. Indeed, in early stages of technology development, it may be the case that there is an initial increase, rather than a decrease, in unit costs as the first few units are produced (Rubin et al., 2007). In the 2011 RIA, which accompanied the promulgation of the Phase I rule, EPA’s initial approach to learning curves used a learning rate of 20 percent, close to the average found by Dutton and Thomas (1984). The rate is routinely applied only to the long-term direct manufacturing costs of the technology package required by the regulation; its use for indirect costs is at the user’s discretion. Thus, it is closer to a learning curve than an experience curve. When NHTSA and EPA began to collaborate on regulation, a range of learning rates was employed based primarily on the maturity of the technology employed, with new technologies being assigned “steep” learning rates (around 20 percent) and more mature technologies, widely available commercially, “flat” rates in the range of 1 to 3 percent (Taylor and Fujita, 2013). NHTSA and EPA are now using an algorithm far more complicated than what is generally used in the literature, and arguably with a level of complexity that is not well supported given the overall uncertainties on learning rates. More specifically, NHSTA and EPA are assuming that for emerging and newer technologies, the technology will have what they call a “steep learning/volume-based” curve which will be a 20 percent reduction in cost per year when either there is a doubling of the production capacity or after 2 years of production (which reflects the underlying assumption by NHTSA and EPA of a doubling of production every 2 years). After that, the technologies follow what they call a “time-based” or “flat” learning curve which is also applied to the mature technologies. In the time-based or flat learning curve, NHTSA and EPA assume that in the first 5 years there is a 3 percent cost reduction per year, in the following 5 years there is a 2 percent cost reduction per year, and, after that, for the next 10 years, a 1 percent cost reduction per year. Figure 12-6 shows the year-to-year “learning” cost factors used in the draft RIA. The interpretation of the figure is as follows: when the lines cross the horizontal red line (i.e., the value of 1), that represents NHTSA and EPA’s point-cost estimate that the learning was then applied to. So, for example for “Strong Hybrid” and “Low Rolling Resistance Tires-2,” the cost estimate from the agencies was done for year 2021, and then the “learning” factors for other years were applied (so that the cost for “Strong Hybrid” in 2016, for example, was assumed to be 1.55 times the 2021 cost). Prepublication Copy – Subject to Further Editorial Correction 12-33

FIGURE 12-6 Cost factors assumed in the Draft RIA. Figure produced by the committee, using the lines produced in the RIA. Recommendation 12-6: NHTSA and EPA should employ a simpler and more transparent method and rationale on how the unit costs of technologies will evolve, either over time or as a function of the production capacity, in their future assessment. Two options the agencies should consider include (1) assuming specific costs per unit over time for each of the technologies under study, informed by manufacturers’ and experts’ elicitation of likely costs in the future; and (2) applying simple parametric one-factor learning curves, complemented with a sensitivity analysis to assess how high or low these learning factors would need to be for a technology to be more economically appealing than another. NHTSA and EPA also are encouraged to do ex post assessments of their forecasts of learning effects. 12.5 REFERENCES Abernathy, W.J., and K. Wayne. 1974. Limits of the learning curve. Harvard Business Review (September). Anthoff, D. and R.S.J. Tol. 2012. Climate damages in the FUND model: A comment. Ecological Economics 81:42. https://doi.org/10.1016/j.ecolecon.2012.06.012. Anthoff, D. and R.S.J. Tol. 2014. The income elasticity of the impact of climate change, in Is the Environment a Luxury?An Inquiry into the relationship between environment and income. Ed. By Tiezzi, S. and Martini, C. https://doi.org/10.4324/9781315819594. Arrow, K.J. 1962. The economic implications of learning by doing. Review of Economic Studies 155– 173. Brown, S.P.A., and H.G. Huntington. 2013. Assessing the U.S. oil security premium. Energy Economics 38:118-127. Prepublication Copy – Subject to Further Editorial Correction 12-34

CalHEAT (California Hybrid, Efficient and Advanced Truck Research Center). 2013. Market Transformation Roadmap for Medium and Heavy-Duty Trucks. Prepared for the California Energy Commission. Clarke, L., J. Weyant, and A. Birky. 2006. On the sources of technological change: Assessing the evidence. Energy Economics 28:579-595. Clarke, L., J. Edmonds, V. Krey, R. Richels, S. Rose, M. Tavoni. 2009. International climate policy architectures: Overview of the EMF 22 International Scenarios. Energy Economics 31:S64-S81. Cohen, W.M., 1995. Empirical studies of innovative activity. Pp. 182-264 in Handbook of the Economics of Innovation and Technological Change, edited by P. Stoneman. Oxford: Blackwell. Cohen, W.M., and S. Klepper. 1996. Firm size and the nature of innovation within industries: The case of process and product R&D. Review of Economics and Statistics 232-243. Cooper, M. 2010. Policy Challenges of Nuclear Reactor Construction: Cost Escalation and Crowding Out Alternatives. Institute for Energy and the Environment, Vermont Law School. Cummins Inc. 2015. Comments in Response to Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and Heavy-Duty Engines and Vehicles-Phase 2. NHTSA-2014-0132. Davis, S.C., S.W. Diegel, and R.G. Boundy. 2015. Transportation Energy Data Book: Edition 34. Oak Ridge, TN: Oak Ridge National Laboratory. August. Dudley, S. 2017. President Trump takes on the social cost of carbon. Forbes. www.forbes.com/sites/susandudley/2017/03/30/president-trump-takes-on-the-social-cost-of- carbon/#1110d9f2185e. Accessed April 6, 2017. Dutton, J. M., and A. Thomas. 1984. Treating progress functions as a managerial opportunity. The Academy of Management Review 9(2):235-247. EPA (Environmental Protection Agency). 1997. Final Regulatory Impact Analysis: Control of Emissions of Air Pollution from Highway Heavy-Duty Engines. Washington, D.C.: EPA. Ferioli, F., K. Schoots, and B.C.C. van der Zwaan. 2009. Use and limitations of learning curves for energy technology policy: A component-learning hypothesis. Energy Policy 37:2525-2535. Gayer, T., and W.V. Viscusi. 2013. Overriding consumer preferences with energy regulations. Journal of Regulatory Economics 43:248-264. Gillingham, K., R.G. Newell, and W.A. Pizer. 2008. Modeling endogenous technological change for climate policy analysis. Energy Economics 30:2734-2753. Gordon, D., J. Burdelski, and J.S. Cannon. 2003. Greening Garbage Trucks: New Technologies Cleaner Air. New York: INFORM, Inc. Grubler, A. 2010. The costs of the French nuclear scale-up: A case of negative learning by doing. Energy Policy 38:5174-5188. Hamilton, J.D. 2009. Causes and consequences of the oil shock 2007-2008. Brookings Papers on Economic Activity (Spring):215-261. Hindle, T. 2008. The Economist Guide to Management Ideas and Gurus. London: Profile Books. Hope, C. 2006. The marginal impact of CO2 from PAGE2002: An integrated assessment model incorporating the IPCC's five reasons for concern. Integrated Assessment 6:19-56. Hope, C. 2011. The PAGE09 Integrated Assessment Model: A Technical Description. Cambridge Judge Business School Working Paper Series. ICCT. 2015. Policy Update: United States Efficiency and Greenhouse Gas Emission Regulations for Model Year 2018-2027 Heavy-Duty Vehicles, Engines and Trailers. Washington, D.C.: ICCT. Interagency Working Group on Social Cost of Carbon (IWG). 2010. Social Cost of Carbon for Regulatory Impact Analysis under Executive Order 12866. February. United States Government. Available at: http://www.whitehouse.gov/sites/default/files/omb/inforeg/for-agencies/Social- Cost-of-Carbon-forRIA.pdf. IWG. 2013. Technical Support Document: - Social Cost of Carbon for Regulatory Impact Analysis under Executive Order 12866. United States Government. IWG. 2015. Technical Support Document: - Technical Update of the Social Cost of Carbon for Regulatory Impact Analysis under Executive Order 12866. July. United States Government. Prepublication Copy – Subject to Further Editorial Correction 12-35

Available at: https://obamawhitehouse.archives.gov/sites/default/files/omb/inforeg/scc-tsd-final- july-2015.pdf. Jamasb, T. 2007. Technical change theory and learning curves: Patterns of progress in electricity generation technologies. Energy Journal 28:51-72. Joskow, P.L., and N.L. Rose. 1985. The effects of technological change, experience, and environmental regulation on the construction cost of coal-burning generating units. The RAND Journal of Economics 1-27. Killian, L. 2014. Oil price shocks: Causes and consequences. Annual Review of Resource Economics 6:133-154. Doi: 10.1146/annurev-resource-083013-114701. Klaassen, G., A. Miketa, K. Larsen, and T. Sundqvist. 2005. The impact of R&D on innovation for wind energy in Denmark, Germany and the United Kingdom. Ecological Economics 54:227-240. Klemick, H., E. Kopits, K. Sargent, and A. Wolverton. 2014. Heavy-Duty Trucking and the Energy Efficiency Paradox: Evidence from Focus Groups and Interviews. Paper presented to the NRC Committee on Assessment of Technologies and Approaches for Reducing the Fuel Consumption of Medium- and Heavy-Duty Vehicles, Phase Two. Klemick, H., E. Kopits, A. Wolverton, and K. Sargent. 2015. Heavy-duty trucking and the energy efficiency paradox: Evidence from focus groups and interviews. Transportation Research Part A: Policy and Practice 77:154-166. Klepper, S., and K.L. Simons. 2000. The making of an oligopoly: Firm survival and technological change in the evolution of the U.S. tire industry. Journal of Political Economy 108:728-760. Kouvaritakis, N., A. Soria, and S. Isoard. 2000. Modeling energy technology dynamics: Methodology for adaptive expectations models with learning by doing and learning by searching. International Journal of Global Energy Issues 14:104-115. Leiby, P. 2015. Factors Influencing Estimates of the Energy Security Premium for Heavy Duty Phase II Rule, Oak Ridge National Laboratory. Leiby, P., D. Jones, T.R. Curlee, and R. Lee. 1997. Oil Imports: An Assessment of Costs and Benefits. Oak Ridge, TN: Oak Ridge National Laboratory. Lutsey, N. 2015. Regulatory Considerations for Advancing Commercial Pickup and Van Efficiency Technology in the United States Washington, D.C.: ICCT. Mills, M.P. 2016. After the carnage, shale oil will rise again. Wall Street Journal, January 18. Nemet, G. 2006. Beyond the learning curve: factors influencing cost reductions in photovoltaics. Energy Policy 34:3218-3232. Newell, R. and W. Pizer, 2003. Discounting the distant future: how much do uncertain rates increase valuations?, Journal of Environmental Economics and Management 46(1):52-71. https://doi.org/10.1016/S0095-0696(02)00031-1. NHTSA (National Highway Traffic Safety Administration) and EPA (Environmental Protection Agency), 2010. Final Regulatory Impact Analysis. Final Rulemaking for Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and Heavy-Duty Engines and Vehicles. Washington, D.C.: NHTSA and EPA. Chap. 9. NHTSA and EPA. 2015. Proposed Rulemaking for Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium-and Heavy-Duty Engines and Vehicles-Phase 2: Draft Regulatory Impact Analysis. Washington, D.C.: NHTSA and EPA. NHTSA and EPA. 2016. Regulatory Impact Analysis, Final Rule for Greenhouse Gas Emissions and Fuel Efficiency Standards for Medium- and Heavy-Duty Engines and Vehicles, Phase II. Washington, D.C.: NHTSA and EPA. Nordhaus, W.D., J. Boyer. 2000. Warming the World: Economic Models of Global Warming. Cambridge: The MIT Press. ISBN 0-262-14071-3. Nordhaus, W.D. 2007. The Challenge of Global Warming: Economic Models and Environmental Policy. New Haven: Yale University. Nordhaus, W.D. 2009. The Perils of the Learning Model for Modeling Endogenous Technological Change. NBER Working Paper No. 14638. National Bureau of Economic Research. Prepublication Copy – Subject to Further Editorial Correction 12-36

NRC. 2010a. Technology and Approaches to Reducing the Fuel Consumption of Medium- and Heavy- Duty Vehicles. Washington, D.C.: The National Academies Press. NRC. 2010b. Hidden Costs of Energy: Unpriced Consequences of Energy Production and Use. Washington, D.C.: The National Academies Press. NRC. 2014. Reducing the Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy- Duty Vehicles, Phase Two: First Report. Washington, D.C.: The National Academies Press. NRC. 2015. Cost, Effectiveness, and Deployment of Fuel Economies for Light-Duty Vehicles. Washington, D.C.: The National Academies Press. ORNL (Oak Ridge National Laboratory). 2008. The Energy Security Benefits of Reduced Oil Use, 2006- 2015. March. Perry, I.W.H., and J. Darmstadter. 2003. The Costs of US Oil Dependence. Discussion paper 03-59. Resources for the Future, Washington, D.C. Roe, G.H., and M.B. Baker. 2009. Why Is Climate Sensitivity So Unpredictable? Science 318(5850):629- 632. DOI: 10.1126/science.1144735. Rogozhin, A., M. Gallaher, G. Helfand, and W. McManus. 2010a. Using indirect cost multipliers to estimate the total cost of adding new technology in the automobile industry. International Journal of Production Economics 124:360-368. Rogozhin, A., M. Gallaher, A. Lentz, and W. McManus. 2010b. Heavy Duty Truck Retail Price Equivalent and Indirect Cost Multipliers. Draft report prepared by RTI International for Chi Li, Office of Transportation and Air Quality, USEPA, Ann Arbor, MI. Rubin, E., I.L. Azevedo, S. Yeh, and P. Jaramillo. 2015. How to improve modelling approaches to represent the future costs of energy technologies? Reviewing the findings from learning curves from energy technologies. Energy Policy 86:198-218. Rubin, E.S., S. Yeh, M. Antes, M. Berkenpas, and J. Davison. 2007. Use of experience curves to estimate the future cost of power plants with CO 2 capture. International Journal of Greenhouse Gas Control 1:188-197. Sinclair, G., S. Klepper, and W. Cohen. 2000. What's experience got to do with it? Sources of cost reduction in a large specialty chemicals producer. Management Science 46:28–45. Söderholm, P., and G. Klaassen. 2007. Wind power in Europe: A simultaneous innovation–diffusion model. Environmental and Resource Economics 36:163-190. Söderholm, P., and T. Sundqvist. 2007. Empirical challenges in the use of learning curves for assessing the economic prospects of renewable energy technologies. Renewable Energy 32:2559–2578. Taylor, M., and K.S. Fujita. 2013. Accounting for Technological Change in Regulatory Impact Analyses: The Learning Curve Technique. Lawrence Berkeley National Laboratory Report LBNL-6195E. Watanabe, C., 1995. Identification of the role of renewable energy. Renewable Energy 6:237-274. Weiss, M., M. Junginger, ad K. Blok. 2010. A review of experience curve analyses for energy demand technologies. Technological Forecasting and Social Change 77:411-428. Wright, T.P. 1936. Factors affecting the cost of airplanes. Journal of Aeronautical Sciences 3:122-128. Yeh, S., and E.S. Rubin. 2007. A centurial history of technological change and learning curves for pulverized coal-fired utility boilers. Energy 32:1996-2005. Prepublication Copy – Subject to Further Editorial Correction 12-37

Next: 13 Alternative and Complementary Regulatory Approaches »
Reducing Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: Final Report Get This Book
×
Buy Paperback | $85.00
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

Medium- and heavy-duty trucks, motor coaches, and transit buses - collectively, "medium- and heavy-duty vehicles", or MHDVs - are used in every sector of the economy. The fuel consumption and greenhouse gas emissions of MHDVs have become a focus of legislative and regulatory action in the past few years. This study is a follow-on to the National Research Council's 2010 report, Technologies and Approaches to Reducing the Fuel Consumption of Medium-and Heavy-Duty Vehicles. That report provided a series of findings and recommendations on the development of regulations for reducing fuel consumption of MHDVs.

On September 15, 2011, NHTSA and EPA finalized joint Phase I rules to establish a comprehensive Heavy-Duty National Program to reduce greenhouse gas emissions and fuel consumption for on-road medium- and heavy-duty vehicles. As NHTSA and EPA began working on a second round of standards, the National Academies issued another report, Reducing the Fuel Consumption and Greenhouse Gas Emissions of Medium- and Heavy-Duty Vehicles, Phase Two: First Report, providing recommendations for the Phase II standards. This third and final report focuses on a possible third phase of regulations to be promulgated by these agencies in the next decade.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!