Norwegian School of Economics Bergen, Autumn 2016 An Empirical Analysis of Drivers for Electric Vehicle Adoption: Evidence from Norway 2010-2014 Mads Fjeld Wold & Sara Ølness Supervisor: Gunnar Eskeland Master Thesis, MSc in Economics and Business Administration, Energy Natural Resources and the Environment, Economic Analysis NORWEGIAN SCHOOL OF ECONOMICS This thesis was written as a part of the Master of Science in Economics and Business Administration at NHH. Please note that neither the institution nor the examiners are responsible − through the approval of this thesis − for the theories and methods used, or results and conclusions drawn in this work. Abstract Weexaminehowgovernmentinterventionintheautomobilemarketa�ectedthebattery- electricvehicleadoptionintheNorwegiancountiesintheperiodbetween2010and2014, and what regional di�erences influenced the adoption rate. Norway has the world’s strongest means of support for electric vehicles and represents a mature market with a significant share of the total car fleet being electric. The government has promoted this growthonthebasisthatelectricvehiclesarepartofthesolutiontotheclimateproblem. It postulates that positive externalities from electric vehicle use are not captured by the market, resulting in a market failure, which necessitates government intervention. This paper explores the e�ects of interventions, such as support for charging-network development and financial incentives, on the development of electric vehicle adoption in the 19 Norwegian counties. We use a panel data approach where econometric methods of fixed e�ects, random e�ects and pooled OLS are applied. The period between 2010 and 2014 is covered on a yearly basis in the analysis. The paper contributes to existing literature by studying regions over time. Through pooled OLS, we found charging in- frastructure to have the strongest predictive power followed by the economic gain from free passes through toll stations. Reduced rates for EVs on ferries were expected to have a positive e�ect on EV adoption, but came out with spurious results in this analysis. Time saved by having access to bus lanes did not turn out to have significant influence. Some county-specific features such as coastline and elevation seem to also play a role in the adoption of battery electric vehicles. Our results are interesting as they only partly support existing literature, and supplement it by adding geographic and climatic factors. The paper gives an indication for policy makers of what incentives are e�cient in driving forward EV adoption. Preface This thesis is written as a part of our Master of Science in Economics and Business Administration at the Norwegian School of Economics and corresponds to one semester of full-time studies. We were two students working on this project during the fall semester of 2016 and it correlates to our two majors: Energy, Natural Resources and the Environment and Economic Analysis. Through our work on this project, we have been able to immerse ourselves into an excitingandhighlytimelytopicinaworldwhereclimatechangeproblemsarebecoming increasingly apparent: the adoption of electric vehicles. We chose to write about this topic due to ourinterest in electric vehicles, public policyand because of our concern for the environment. As two Norwegian students, we are proud to be part of a society that values and encourage environmentally-friendly choices and that has been so successful in promoting them. It has been interesting, insightful and rewarding to explore the Norwegian EV history, examine the role of government and attempting to quantify e�ects of various factors a�ecting EV adoption in Norway. Working on this thesis has introducedustotheworldofacademicwritingandtaughtushowtoworkindependently and structured on such a large and comprehensive project. We would like to especially thank our supervisor, Professor Gunnar S. Eskeland, for his guidance and support throughout the whole process. He has been a valuable discussion partner and motivated us when the writing process was going slow. We also want to show our gratitude to PhD Research Scholar Shiyu Yan for his support in working with the data set and for his help with econometric theory. We also need to thank the all the people who have provided us with the necessary data required for our analysis. Especially Jan Kristian Jensen and Nina Lysefjord at Norwegian Public Roads Admin- istration have been of tremendous help to us in collecting data on toll stations, ferries and bus lanes. Bergen, 20th December 2016. Mads Fjeld Wold Sara Elisabeth Ølness 2 Contents Glossary 7 Acronyms 8 1 Introduction 9 1.1 Research question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 2 Historical development 12 2.1 The market phases for electric vehicles in Norway . . . . . . . . . . . . 12 2.2 The role of the government . . . . . . . . . . . . . . . . . . . . . . . . . 13 3 Literature review 17 4 Theoretical background 21 4.1 Market failure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 4.2 Modelling demand . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 5 Data 28 5.1 Building the data set . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 5.1.1 Spatial and time dimension . . . . . . . . . . . . . . . . . . . . 28 5.1.2 Sales statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 5.1.3 Infrastructure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 5.1.4 Financial incentives . . . . . . . . . . . . . . . . . . . . . . . . . 33 5.1.5 Demographics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 5.1.6 Geography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 5.2 Summary statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 5.2.1 Electric vehicles sales share . . . . . . . . . . . . . . . . . . . . 39 5.2.2 Charging infrastructure . . . . . . . . . . . . . . . . . . . . . . . 41 5.2.3 Ferry expenses, toll expenses and bus lanes . . . . . . . . . . . . 43 5.3 Correlation analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 6 Empirical framework 52 6.1 From single cross-sectional models to advanced panel data methods . . 52 6.1.1 Single cross-section . . . . . . . . . . . . . . . . . . . . . . . . . 52 6.1.2 Panel data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 6.1.3 Pooled OLS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 3 6.1.4 Fixed e�ects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 6.1.5 Random e�ects . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 6.2 Coe�cient of determination . . . . . . . . . . . . . . . . . . . . . . . . 58 6.3 Post regression tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 6.3.1 Model selection tests . . . . . . . . . . . . . . . . . . . . . . . . 59 6.3.2 Heteroskedasticity and serial correlation tests . . . . . . . . . . 60 6.4 Model specification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62 7 Results 66 7.1 Regression analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66 7.1.1 Analysing the model . . . . . . . . . . . . . . . . . . . . . . . . 69 7.1.2 Results from heteroskedasticity and serial correlation tests . . . 71 7.2 Sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 7.2.1 Model without control variables . . . . . . . . . . . . . . . . . . 72 7.2.2 Di�erent measures for population density . . . . . . . . . . . . . 74 7.2.3 Combining ferry and toll expenses . . . . . . . . . . . . . . . . . 77 7.2.4 Substituting year-fixed e�ects with time trends . . . . . . . . . 79 7.3 Summary of the results . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 8 Discussion 83 8.1 Discussion of the results . . . . . . . . . . . . . . . . . . . . . . . . . . 83 8.2 Limitations in the data set and empirical approach . . . . . . . . . . . 85 8.3 Suggestions for future research . . . . . . . . . . . . . . . . . . . . . . . 86 8.4 Implications of the study . . . . . . . . . . . . . . . . . . . . . . . . . . 87 9 Conclusion 91 Appendices 93 A Results 94 Bibliography 98 4 List of Figures 2.1 Development of the EV fleet and EV policies in Norway between 1997 and 2013. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 2.2 Demand as a function of incentive changes in the years between 2000 and 2004. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 4.1 Negative Externality Causing a Market Failure . . . . . . . . . . . . . . 22 4.2 Positive Externality Causing a Market Failure . . . . . . . . . . . . . . 23 5.1 Norwegian counties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 5.2 Regional development of electric vehicle sales share, Norway 2010-2014. 40 5.3 Development of electric vehicle sales share, Norway 2010-2014. . . . . . 40 5.4 Share of electric vehicle sales in the extreme cases of Hordalanad and Finnmark, Norway 2010-2014. . . . . . . . . . . . . . . . . . . . . . . . 41 5.5 Regional development of charging points per capita, Norway 2010-2014. 42 5.6 Development in the sales share of electric vehicles and charging points per capita, Norway 2010-2014. . . . . . . . . . . . . . . . . . . . . . . . 43 5.7 Development in sales of electric vehicles and the cumulative number of charging points, Norway 2010-2014. . . . . . . . . . . . . . . . . . . . . 44 5.8 Development in the sales of electric vehicles per capita compared to the total amount of charging points per capita, Norway 2010-2014. . . . . . 45 5.9 Regional development of toll station expenses per car, Norway 2010-2014. 46 5.10 Regional development of ferry expenses per car, Norway 2010-2014 . . . 47 5.11 Regional development of bus lanes, Norway 2010-2014 . . . . . . . . . . 47 5.12 Graphical correlation matrix for the dependent variable and the key ex- planatory variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 5.13 Graphicalcorrelationmatrixforthedependentvariableanddemographic variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 5.14 GraphicalCorrelationmatrixforthedependentvariableandcounty-fixed control variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 7.1 Identifying heteroskedasticity . . . . . . . . . . . . . . . . . . . . . . . 71 A.1 Fossil fuel and electricity prices . . . . . . . . . . . . . . . . . . . . . . 96 5 List of Tables 5.1 Summary statistics of relevant variables . . . . . . . . . . . . . . . . . . 39 5.2 Yearly means for charging points per capita and relevant variables . . . 43 5.3 Yearly national means for toll and ferry expenses per car . . . . . . . . 44 5.4 Correlation matrix for all variables. . . . . . . . . . . . . . . . . . . . . 48 7.1 Nomenclature for regression variables . . . . . . . . . . . . . . . . . . 67 7.2 Regression results for electric vehicle adoption in Norway . . . . . . . . 68 7.3 Post-regression tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 7.4 heteroskedasticity Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 7.5 Regressions results: no control variables. . . . . . . . . . . . . . . . . . 73 7.6 Regressions results: specifying population density . . . . . . . . . . . . 75 7.7 Regression results: combining ferry and toll expenses . . . . . . . . . . 78 7.8 Regressions results: specifying time trends . . . . . . . . . . . . . . . . 80 A.1 Test for joint significance of time dummies . . . . . . . . . . . . . . . . 94 A.2 Regression results: no time dummies . . . . . . . . . . . . . . . . . . . 95 A.3 Wooldridge test for serial correlation . . . . . . . . . . . . . . . . . . . 96 A.4 Regression results: urban density . . . . . . . . . . . . . . . . . . . . . 97 6 Glossary Battery Electric Vehicle Vehicle using a battery as its only fuel source. Hybrid Electric Vehicle Vehicle using a combination of fossil fuel and electricity as fuel. The electricity is generated as the vehicle is in motion. Internal Combustion Engine Vehicle Vehicle using either gasoline or diesel as fuel. Plug-in Hybrid Electric Vehicle Vehicle using a combination of fossil fuel and electricity with batteries that can be recharged. 7 Acronyms BEV Battery Electric Vehicle. EV Electric Vehicle. FE Fixed E�ects. GHG Greenhouse Gas. GSL Generalised Least Square. HEV Hybrid Electric Vehicle. ICEV Internal Combustion Engine Vehicle. MASL Metres Above Sea Level. MPC Marginal Private Cost. MSC Marginal Social Cost. OLS Ordinary Least Square. PHEV Plug-in Hybrid Electric Vehicle. POLS Pooled Ordinary Least Square. RE Random E�ects. SSB Statistics Norway. NPRA Norwegian Public Roads Administration. TØI Institute of Transport Economics. VKT Vehicle Kilometres Travelled. WMO World Meteorological Organisation. YED Income Elasticity of Demand. 8 1 Introduction Climate change has been on the political agenda of a majority of nations around the world for an extensive period of time, and countless solutions have been proposed to diminish the problems caused by global warming. In recent years, climate change issues have become exceedingly pressing: global temperature records are being consecutively beaten, wildlife is going extinct en mass, coral reefs are dying and huge masses of people are beingdisplaced fromtheirhomes andlabelledenvironmentalrefugees. Inthe aftermathoftheParisAgreementof2015, thedebateonwhatspecificnationalmeasures eachcountrymustundertaketocombatclimatechangehasbecomeincreasinglyrelevant as countries are trying to reduce their emission levels in the most e�cient way. Several countries have targeted the transportation sector in their pursuit of lower na- tional emission levels. According to the U.S Environmental Protection Agency (2015), the transportation sector accounted for 26% of the total GHG emissions in the United States in 2014. Light-duty vehicles were responsible for 61% of these emissions. In the EU, the European Commission (2016) reports that around 12% of EU’s total CO 2 emissions come from passenger cars. The Norwegian Environmental Agency reveal that emissions from road tra�c accounts for 19% of Norway’s total emissions (Miljødirek- toratet, 2016). These emissions can most easily be cut by reducing the amount of transportation needed. In many developed countries, e�orts to increase urbanisation are being made to reduce the overall need for transportation. Improvements in the e�ciency of modes of transportation are being undertaken as fossil fuel vehicles are progressively becoming more e�cient and their carbon footprint mitigated. By devel- oping a well-functioning public transportation system and making biking a more viable transportation option, many countries are attempting to reduce emissions from road tra�c by moving people over to less polluting transportation modes. Encouraging the adoption of zero-emission vehicles is yet another way governments attempt to scale down emissions from private transportation. The electrification of passenger vehicles has generated a lot of interest over the past years. This has been due to the vehicles’ prominence in peoples everyday life and the tremendous technological developments they have undergone over the past years. Im- provementsinbatterytechnologyhaveenabledBatteryElectricVehicles(BEV)to, ona single charge, drive distances comparable to those ICEVs drive on a full tank. Car man- ufacturers worldwide are investing large sums into the development of technologies for zero-emission vehicles in order to reduce the carbon foot print of their car fleet. There 9