A multi-model approach: international electric vehicle adoption Alan Jenn Postdoctoral Researcher Gil Tal Professional Researcher Lew Fulton STEPS Director Sustainable Transportation Energy Pathways Institute of Transportation Studies University of California, Davis
What will electric vehicle adoption look like in the future? The primary objective of this research is to investigate future international scenarios of electric vehicles (EVs) to 2030/2050. The broad set of questions we are interested in studying: How many electric vehicles will be on the road in 2030? 2040? 2050? When will we see 100 million EVs cumulatively in the world? Can this be achieved by 2030? What sorts of vehicle pricing and attributes are required to boost adoption? Which policies can be implemented in different countries to promote EVs in the most effective manner?
What is the future of electric vehicle adoption? 3 1.0 CARB2009 0.8 Market Share 0.6 0.4 CET2009 FORECASTS NATACA2013 EPRI.NRDC2007 PNNL2008 0.2 AEO2013 0.0 2010 2020 2030 2040 2050 Year
Using three different models for scenario 4 How are different estimates effected by their choice of modeling technique as opposed to assumptions about inputs? We employ three very different models to test for the robustness of outcomes: Discrete choice model Technology diffusion model Regression of trends
Discrete choice model 5 We simulate consumers decision making process about selecting a product among a set of discrete choices An individual is more attractive based on its comparative value compared to other vehicles Vehicle attributes inform attractiveness Fuel Efficiency Vehicle Brands Engine Size, Power Vehicle Classes Vehicle Sales
Technology diffusion model 6 Diffusion of innovation theory argues that most technologies are adopted under similar trends Vehicle projections are made based on how the trends can be fit against historical sales of electric vehicle technologies Vehicle Sales
Regression of trends approach 7 Regression of trends looks at how vehicle sales are effected across a number of factors that change over time The model includes both variables intrinsic to the vehicles (prices, efficiency) as well as external factors (gas prices, GDP) Improvements in efficiency Vehicle price changes Gas prices Vehicle Sales
Our project leverages an extremely detailed dataset 8 We are using a large dataset of new vehicle registrations from IHS Automotive. The data were cleaned and expanded in cooperation with the International Energy Agency. A quick summary of the dataset Number of countries: 39 Years: 2005, 2008, 2010-2015 Unique vehicle models: 4,771 Unique vehicle manufacturers: 503 Total registrations: 509,194,651 Vehicle attributes: Axles, drive, engine size, # of cylinders, engine power, fuel type, transmission, turbo, price, segment, curb weight, footprint, fuel efficiency (NEDC/WLTP), and emissions (NEDC/WLTP)
International vehicle registrations by fuel technology (cumulative 2010-2015) 9 120 Total Registrations (millions) 90 60 30 Fuel Type BEV CNG Diesel Flexfuel HEV Hybrid Hydrogen LPG Petrol PHEV 0 USA China Japan Brazil Germany Russia India United Kingdom France Canada Indonesia Italy Mexico South Korea Australia Argentina Malaysia Philippines South Africa Thailand Turkey Peru Egypt Chile Spain Ukraine Belgium Netherlands Sweden Switzerland Austria Denmark Portugal Norway Ireland Finland Greece Luxembourg Macedonia Country
Worldwide sales of electric vehicles (EVs per 10000 Worldwide conventional sales of vehicles electric vehicles sold) (EVs per 10000 conventional vehicles sold) Worldwide sales of electric vehicles (EVs per 10000 conventional vehicles sold) 10 0 10 30 50 100 250 10 30 50 100 250
Number of electric vehicle models available 11 Number of electric vehicle models available Number of electric vehicle models available 0 3 10 25 40 80 3 10 25 40 80
Assumptions for forecasting We examine three basic scenarios for forecasting, each scenario assumes different prices and model availabilities for electric vehicles. Price reductions are exogenous, their mechanisms are not explicitly stated but include learning-by-doing, competition, and policy incentives. Model availability refers to the coverage of EVs in the market of vehicle models. Range of EVs are increased over time to allay issues of range anxiety 0.5 Average Price Reduction ($) 120000 80000 40000 Proportion of models as EVs 0.4 0.3 0.2 0.1 Range Increase (km) 120 80 40 2020 2025 2030 2035 2040 Year 0.0 2020 2025 2030 2035 2040 Year 2020 2025 2030 2035 2040 Year
Low Medium High Sample of vehicle scenario projections using discrete choice modeling approach 13 Belgium Chile Germany Japan Portugal Switzerland Market Share 2020 2025 2030 2035 2040 2020 2025 2030 2035 2040 2020 2025 2030 2035 2040 2020 Year 2025 2030 2035 2040 2020 2025 2030 2035 2040 2020 2025 2030 2035 2040 Fuel Type BEV CNG Diesel Hybrid Petrol PHEV Flexfuel LPG HEV
Sample of vehicle scenario projections using Bass diffusion modeling approach 14 Saturation of Market Potential (BEV) Australia Canada Chile China 0.3 1e 03 0.2 5e 04 0.1 0.0 0e+00 France Germany Italy Japan 0.3 075 050 0.2 0.1 0.0 Russia Ukraine United Kingdom USA 025 000 2010 2020 2030 2040 2050 2010 2020 2030 2040 2050 2010 2020 2030 2040 2050 2010 2020 2030 2040 2050 Year
How do the different model projections compare? 15 125 Worldwide stock of EVs (millions) 100 Discrete choice model yields the lowest adoption, 75 this is sensible as the models reflect current 50 consumer preferences Shape of diffusion 25 curve due to addition of demand due to exceeding vehicle 0 lifetime Different models vary in Year estimation across the same inputs by an order of 2 (from 60 million vehicles in 2040 up to 120 million vehicles) Worldwide stock of EVs (millions) 125 100 75 2020 2025 2030 2035 2040 50 25 0 Method Bass Diffusion Choice Model Linear Regression 2020 2025 2030 2035 2040 Year Me
Discussion 16 Despite consistency in model inputs, there are significant differences in outputs. The choice of model matters! 100 million vehicles by 2040 is an optimistic projection for choice and regression models but is achieved in the diffusion model. International factors are vital to consider and lead to stark differences in adoption potential across all three models.
Ongoing work 17 Further calibration of models is needed: Data cleaning Investigating combinations of different variables Incorporating more variables and scenarios (e.g. charging infrastructure) Focusing on policy impacts and integrating existing policies Investigate uncertainty in each modeling scenario