How not to assess technology costs and benefits: CAFE edition


One of the central arguments in the debate over fuel economy regulations is how much any proposed mpg standard will raise the price of a vehicle versus how much direct benefit it will provide a car buyer in return. Most of the benefit comes in the form of savings at the gas pump, which is simple to calculate over the lifetime of a vehicle if you know the fuel economy. Estimating the cost, or price increase, is more complicated, because it can vary wildly depending on where you start from and what assumptions you make about future technology improvements and other variables. The methodology you choose for estimating cost is a key determinant of where you come out on the cost of a fuel economy standard.

Opponents of new U.S. fuel economy standards for 2017–2025 that could conceivably go as high as 62 mpg have to a very great degree relied on a 2010 National Academy of Sciences study for their starting point and certain pivotal assumptions in estimating the costs of those standards. Specifically, a market and industry forecast by the Center for Automotive Research is frequently cited (see, e.g., here or here), which the ICCT has extensively critiqued.

Unfortunately, using the NAS to assess 2025 technologies and costs introduces a fundamental bias that skews the analysis. Exactly how and why is worth explaining in some detail.

The original charge to the National Research Council committee on light-duty fuel economy that produced the report was to assess technology benefits and costs up to year 2025—the year in which these potential new standards would take full effect. CAR quoted this charge to explain the report’s relevance, but ignored that fact that the final NRC report explicitly repudiated the possibility that its findings could be relevant beyond 2015:


“Tables S-1 and S-2 show the committee’s estimates of fuel consumption benefits and costs for technologies that are commercially available and can be implemented within 5 years. The cost estimates represent estimates for the current (2009/2010) time period to about 5 years in the future.” [NAS report page S-1]


In other words, using the NAS report as a source of technology benefit and cost projections confuses the near term (2010–2015) with the long term (2025). This has two bad effects. One, it completely misses the benefits from technologies not included in the NRC report that will be in use in 2020–2025. Two, it flattens the cost and benefit trendlines at the point they’d reached in 2010.

The internal combustion engine is a mature technology, but that doesn’t mean it’s moribund. Computer simulations and computer-aided design have enabled advances in engine design that weren’t possible before, and have also accelerated the rate at which new technologies can be brought to market. And since vehicle technology benefits are trending upward while costs trend downward, use of the NAS report cuts out fifteen years of rising benefits and falling costs, distorting the final cost-benefit ratio.

Two examples of missing engine technologies will illustrate point one. Turbocharged gasoline direct injection (GDI) systems are already available, deployed by manufacturers including Hyundai and Ford. The next generation of these engines will incoporate cooled exhaust gas recirculation (EGR) and use the turbocharger to boost the amount of EGR, in addition to boosting the amount of intake air. This diluted air/fuel mixture lowers combustion temperatures under all load conditions, which reduces heat losses and pumping losses for higher efficiency and allows higher boost levels without knock for additional engine power. Result: Boosted EGR systems improve efficiency by more than 10%, even compared to turbocharged, downsized, direct-injected engines such as Ford’s Eco-Boost system.

Hyundai, Nissan, VW, BMW, and Mercedes are all introducing parallel hybrid systems with two clutches, commonly referred to as P2 hybrid systems, in 2011 and 2012. Because P2 hybrids use a single electric motor that can be used with a conventional transmission, they cost much less than the power-split hybrid that uses a complex power-split device with two larger electric motors. The fuel-efficiency benefits from the parallel hybrids are expected to be only marginally smaller than the future improved power-split system, but at about one-third lower cost. Another important hybrid cost reduction is the development of Li-ion batteries optimized for high power instead of high energy, which are expected around 2015. These batteries will be able to satisfy the high power demands for acceleration assist and capture of regenerative braking energy with a much smaller battery pack, further reducing the cost of P2 hybrids by hundreds of dollars. As a result, a majority of hybrid vehicles in the 2025 timeframe are expected to be P2 hybrids and the cost is expected to be low enough by 2020 for acceptance by mainstream customers.

Several factors drive vehicle technology benefit trends upward and cost trends downward: technology penetration, continued developments in technology through research and development, cost reductions from learning by doing. The effects of these are hard to predict or accurately quantify—but they do exist and will only accelerate with continued improvements in computer simulation and computer-aided design.

The NRC report took the conservative approach of simply not trying to predict future technology development. As it carefully stated, “except where indicated otherwise, the cost estimates provided are based on current conditions and do not attempt to estimate economic conditions and hence predict prices 5, 10, or 15 years into the future.” The CAR forecast was implicitly based on the assumption that no further technology improvement or cost reductions will occur after 2013—but then ignored that assumption and went ahead to predict economic conditions and prices ten years further on. This not only affects conventional gasoline technology improvement costs, but greatly overestimates the future costs of hybrid systems, where much learning is yet to happen.

It is ironic in one sense that opponents of the most stringent standards should rely so much on the NAS report, because that report specifically recommended that the best way to assess technology costs is to do tear-down analyses of new technology and compare the costs to older technology. This is precisely what EPA has done, contracting with FEV and Monro, and conducting a study with Ricardo to assess future technology benefits using detailed simulation analysis, as recommended by vehicle manufacturers. NHTSA and EPA understand the importance of assessing future technologies and costs and have by far the best data, information, and expertise, relying on extensive input from manufacturers in addition to their own studies. Proponents of the 2010 NAS report, to be consistent, should probably defer to EPA/NHTSA’s conclusions until they’re proven wrong by evidence assembled w/ equal meticulousness.