The transportation sector is now the largest source of greenhouse gas (GHG) emissions in the U.S., and nearly 49% of these emissions are from light-duty vehicles (LDVs) 1. Policy strategies to mitigate these emissions have involved picking technology winners that have shifted over time from a focus on low-carbon biofuels in the mid 2000’s to battery electric vehicles (BEVs) more recently. Biofuels went through a phase of hype after being promoted by the Renewable Fuel Standard in 2007 that mandated 36 billion gallons of biofuels by 2022 and then disappointment as the targets for advanced cellulosic biofuels were unmet and concerns about food vs. fuel grew 2,3. Further areas for growth in biofuels are now being considered a solution only for the hard-to-electrify sectors, such as aviation. The recent US National Blueprint for Transportation Decarbonization 1 has now put the spotlight on the electrification of the LDV fleet as the primary approach for reducing their emissions. The current administration has set ambitious goals for increasing the market share of BEV sales to 50% and to reduce the economy-wide GHG emissions by 50% in 2030 relative to 2005 levels4. Various policy incentives, including a Clean Vehicle (EV) tax credit for new EV vehicle purchases as well as for increasing public EV charging stations, are being offered to achieve these targets.
Here we analyze the likely effectiveness of current policies for electrification of the LDV fleet and their GHG outcomes and its implications for the costs of GHG abatement. We quantify the effects of these policies on GHG emissions and their economic costs for the multiple sectors directly affected by these policies, which include the transportation sector which will experience a change in the mix of vehicles and fuels and the electricity and agricultural sectors that provide electricity and biofuels for transportation. These effects will depend on various factors, include consumer willingness to adopt BEVs in response to policy incentives, the net benefits for consumers of various vehicle types, the effect of large-scale adoption of BEVs on the demand for electricity and biofuels and on their consumers and producers and the GHG intensities of alternative liquid fuels and electricity.
In analyzing the effects of EV policy incentives for adoption we recognize that BEV adoption is likely to be affected not only by the costs of the vehicle, fuel and maintenance but also by intangible costs and unobserved differences in behavioral preferences, as suggested by previous studies 6,7. The incentives for new vehicle adoption will also change over time depending on the age distribution of existing LDV stock and the number of vehicles that are reaching the end of their life and need replacement; thus, the dynamics of vehicle stock turnover will influence the demand for new vehicles and the rate at which the share of BEVs can grow.
We consider the effects of the mix of electricity sources and the share of renewable energy in electricity generation, which vary across regions in the US, in quantifying the GHG mitigation benefits of EVs. The overall impact of vehicle electrification on GHG emissions over time will depend on the extent to which the current blend of gasoline and biofuels is displaced and the mix and quantity of electricity that replaces it. The GHG intensity of electricity is expected to decline over time in the US due to policies such as the Renewable Portfolio Standard as well as declining costs of renewable sources.
Existing studies analyzing forward-looking scenarios using energy system optimization models have typically assumed a representative consumer whose decisions are based on deterministic costs6,17–19. This can lead to knife-edge behavior and unrealistic representation of consumer choices with a single (least cost) technology being chosen as optimal in the aggregate. Recent studies have advanced the behavioral realism of these optimization models by introducing unobserved heterogeneity in consumer preferences so that the aggregate outcome includes a mix of vehicle choices 6,20,7 However, their approach is limited to a single sector (vehicle sector) at the national level and does not consider the interactions between fuel supply and demand, mileage driven and vehicle stocks and their effects on vehicle choices or the implications of increased demand for BEVs on the electricity sector and the spatial variability in these across regions in the U.S.
We examine the effectiveness of each of two policies (individually and combined): Clean Vehicle (EV) Tax Credit and increased access to public charging stations, in increasing the share of EVs in new vehicle sales, in the vehicle stock in the US over the 2022–2030 period and GHG emissions, relative to a baseline scenario in which the existing EV tax credit is phased out. We also determine the present value of the welfare costs of GHG abatement, defined as the difference in the social welfare (economic surplus for producers, consumers, and government) between a policy scenario and the baseline scenario, with each of these EV policies. We compare the welfare cost to the social cost of carbon to examine the economic efficiency of these policies. In quantifying the effect of EV policies on GHG emissions, we consider the effects of these policies on the electricity sector which supplies additional electricity to meet demand in the transportation sector. We also consider the effects on biofuel consumption in the transportation sector under alternative assumptions about the specification of the biofuel mandate.
We develop a multi-sectoral modeling framework to consider these bi-directional effects of policies in the transportation sector on the sectors supplying fuel (electricity and biofuel) to the transportation sector and affected by the policy-induced demand for BEVs (see SI Figure A1). The model includes the electricity sector 16, transportation and agricultural sectors in the United States (U.S.) (see Methods and SI Section A.1) and incorporates the dynamics of changes in the stock of vehicles. It incorporates spatial heterogeneity in the demand for VKT, availability of charging stations, electricity generation mix and its carbon intensity and consumer behavioral preferences towards vehicle choices. A description of the policy scenarios simulated is provided in SI Table A1. The baseline scenario assumes that the state Renewable Portfolio Standards (RPSs) are binding but can be exceeded and that the mandate for corn ethanol set by the Renewable Fuel Standard (RFS) is maintained over the 2022–2030 period across the baseline and policy scenarios. We analyze the implications of EV policies for corn ethanol consumption and of alternative specifications of the RFS for vehicle choice. Various uncertainties, such as those about the extent to which behavioral preference heterogeneity matter and the changing mix of electricity sources over time, will impact outcomes. However, since our focus is on the effectiveness of EV policies, our analysis assumes that various technology and preference parameters are the same across the policy and the baseline scenarios; this mitigates the impact of these uncertainties.
The model considers the choice among five types of alternative vehicles: Conventional vehicles (CVs), Flex-fuel vehicles (FFVs), Hybrid vehicles (HBVs), Plug-in-hybrid electric vehicles (PHEVs) and BEVs (Supplementary Information (SI) Table A3). It endogenously determines the social welfare maximizing region-specific, market shares of each of these five types of vehicles, as well as their VKT, the derived demand for liquid fuels and electricity, and equilibrium fuel prices simultaneously over the 2022–2030 period. In the case of EVs, we incorporate intangible costs, which include range limitation costs that arise due to the short range of the BEV batteries and the absence of frequently available and compatible charging stations between the origin and destination, detour costs of finding a BEV charging station, and costs of waiting in a queue for a charging station and waiting while charging 10,11.
Both the tangible and intangible costs of BEV adoption vary across consumers for various reasons in the model. Differences in demand for vehicle kilometers traveled (VKT) across consumer segments affect the fuel cost as well as the extent to which they are likely to be affected by range anxiety, or the fear of running out of battery. Consumers are also heterogeneous in their willingness for detours and in their costs of waiting for public charging, due to heterogeneity in the available chargers and number of public BEV charging stations across locations and in the value attached to the waiting time depending on their income. Furthermore, vehicle choices are affected by unobserved preference heterogeneity, arising from idiosyncratic differences, among consumers; these are driven by lifestyle, environmental concerns, technology orientation and other aspects that govern the self-identity of individuals12.
We incorporate 5,760 (= 3×24×20×4) types of heterogeneous vehicle consumers: with three different levels (high, medium and low) of driving demand, each with vehicle ownership aged between 1 to 24 years, in each of 20 electricity marketing regions in the U.S. and each with 4 possible combinations of charger availabilities (Home and Work, Home only, Work only, and Public only, discussed in A.1.1). Additionally, we introduce randomness in the preferences of vehicle consumers by constructing 10 clones for each of the 5,760 types of consumers using a Monte Carlo approach; these clones differ in their idiosyncratic preferences, daily mileage driven, and value of time that are randomly selected from specified distributions. Unlike previous studies7, our model determines the spatially heterogenous optimal vehicle choice by maximizing discounted social welfare which allows us to endogenously determine VKT, fuel quantities and prices in both the electricity and transportation sectors and their implications for GHG emissions across 20 electricity marketing regions.
We validated the model by comparing the model outcomes for several endogenous variables in previous years from 2016 to 2019 with observed data for key variables incorporating tangible costs, intangible costs and idiosyncratic preferences (Table A3 in SI). We found that the model fitness improved with the inclusion of intangible costs and behavioral preferences (see SI section A.4 for more details). Percentage deviations between observed and simulated outcomes were less than 5% for most variables and less than 10% for all variables analyzed. In the electricity sector, we found that model outcomes for the aggregate consumption of electricity and the share of various sources of electricity were within 5–10% of observed levels over the 2016–2019 level. We analyze the implications of excluding intangible costs and idiosyncratic preferences on vehicle choice and other outcomes. We also examine the implications of a more optimistic share of renewables in electricity generation on outcomes. We conclude with a discussion of the ineffectiveness of a policy focused on electrification alone as a mechanism for reducing the GHG emissions of the LDV sector.