Diamond hybrid working paper rev Dec06

The Determinants of Hybrid-Electric Vehicle Adoption: Insights from State Registration Data Working Paper December 2006 David Diamond PhD Student School of Public Policy George Mason University Fairfax Virginia Home address: 12848 Tewksbury Drive, Oak Hill VA 20171 ddiamond@gmu.edu (703) 476-4030 Faculty Advisor: Philip Auerswald Abstract This paper examines the effect of tax incentives, gasoline prices and other socioeconomic factors on the demand for Hybrid Electric Vehicles (HEVs) in different U.S states As hybrid sales increase, it is important for policymakers to understand how these factors influence demand in order to judge the effectiveness of competing HEV incentive policies The paper develops a demand model for per-state market-share, and uses crosssectional time-series data on new Hybrid Electric Vehicle (HEV) registrations in different U.S states in 2003 and 2004 to evaluate the significance of difference factors In 2003, a number of predictors were significant, suggesting different policy alternatives for promoting adoption In 2004, when demand for the Toyota Prius exceeded supply, HEV registrations were explained almost entirely by dealer location Introduction Consumer interest in Hybrid Electric Vehicles has risen steadily in recent years in response to rising fuel costs and increased concern about pollution and greenhouse gas emissions Hybrid vehicles utilize the same gasoline fuel infrastructure as conventional Internal Combustion Engine (ICE) vehicles, yet represent a distinct technology improvement that can provide greater fuel economy and reduced emissions for equivalent vehicle performance HEVs face the same barriers to diffusion as any new technologies such as lack of knowledge by potential adopters as well as variation in the factors that affect different consumers’ individual utility calculations, such as price, income, discount rate and risk tolerance (Jaffe and Stavins 1994) (Stoneman and Diederen 1994), as well as high initial costs (Argote and Epple 1990) As an energy efficiency technology, HEVs also address positive externalities associated with resource management, the environment and energy security that are not taken into account by the market (Jaffe and Stavins 1994) The presence of these externalities presents a strong economic argument for government intervention, yet it leaves open the question of the most efficient and effective avenue(s) for government action Traditional government intervention policies into the market for new technologies include taxes or subsidies to account for externalities (Baumol 1972), regulation to force adoption of beneficial technologies (Norberg-Bohm 1999) and taxes on resource inputs to induce innovation and adoption of more resource efficient technologies (Newell, Jaffe, and Stavins 1999) In the case of hybrid vehicles, Federal, State and Local Governments have primarily chosen the first option, offering tax deductions, credits, fee reductions and HOV lane privileges to HEV owners (DOE State Incentives for Hybrid Electric Vehicles 2004) Despite the publicity of these subsidy programs, it is still unclear how effective they actually are in promoting hybrid vehicle demand compared to other economic and social factors As hybrid sales increase, it is important for policymakers to understand how incentives and other factors affect adoption patterns, in order to judge the effectiveness and efficiency of HEV incentive policies In section of this paper, I discuss the background and policy implications of HVEs In sections and 4, I introduce a demand model using cross-sectional time-series data on new HEV registrations in U.S states in 2003 and 2004 to examine the effects of tax incentives and gas prices on the number of HEVs registrations in different U.S., states, while controlling for income, average vehicle mileage and the availability of car dealerships In section 5, I discuss the results In a year that automobile makers can distribute supply to accurately meet the demand for hybrid vehicles is in relative equilibrium (an assumption whose accuracy I discuss for each time period), the results suggest that tax incentives, gas prices, income and driving habits are all significant In a year of constrained demand, however, the results suggest that the location of car dealerships dominates In section 6, I discuss the implications of both cases for policymakers In section 7, I discuss future research stemming from this paper Background Hybrid-electric vehicles combine a gasoline engine with an electric motor and battery to provide power to a car’s drive train Despite the additional weight of the electric motor and battery, a typical hybrid is still considerably more fuel efficient in overall driving performance than an Internal Combustion Engine (ICE) car of equivalent size and performance Although most automobile manufacturers either offer HEV vehicles or plan to offer them by the end of the decade, a price premium for HEVs compared to equivalent ICE automobiles may limit demand Although Ogden (2004) demonstrates the potential for HEV technologies to have lower societal lifecycle costs then ICE and many fuel cell technologies for automotive applications, the market does not currently factor the externality costs of HEVs Canes (2003) uses total lifecycle costs to compare equivalent hybrid and gasoline models, and found that lifecycle costs for hybrid vehicles exceed those for equivalent ICE vehicles based on vehicle price, and fuel/maintenance expenses over an expected lifetime, even when he takes pollution costs into account Several researchers and firms have used survey data to draw preliminary conclusions about factors affecting HEV demand These studies generally conclude that hybrid owners tend to be in the highest income demographics and are more sensitive to gas prices than environmental benefits in purchasing their vehicles (Changewave Research 2005: Year of the Hybrid 2004) Burke (2004) and Jaffe, et al (1994) also cite the high discount rate that consumers place on new cost-saving technologies, expecting payback for their investments in as little as three years Cao and Mokhtarian analyzed national-level hybrid sales data, and found lagged gasoline prices to be significant in explaining aggregate diffusion patterns (2004) There is a fairly well developed economic literature on the pricing and demand for automobiles in the U.S aggregate market using national level aggregate data over time, which I adapt to study demand for HEV vehicles Chow (1957) proposed a linear demand equation for nationwide automobile demand over time, as a function of income, price and prior sales Wetzel and Hoffer (1982) and Irvine (1983) develop demand models and use regression analysis to determine consumer preferences for individual features, including fuel economy Wetzel and Hoffer focus on disaggregating based on vehicle class, while Irvine focuses on cross-price elasticities for specific car models with different attributes Hedonic techniques focus on the supply and demand equilibrium in the market to determine quality adjusted prices for unique attributes Bajic (1988) uses hedonic regression to calculate quality adjusted prices for different attributes, and then calculate variance in those attribute prices among specific models for use in a regression of market share By this logic, once the market has determined a quality adjusted price for a given option, the variance from that price for individual models or markets would result in changes in consumer demand for that particular model, as evidenced by a higher or lower market share compared to similar models Berry, Levinsohn and Pakes (1995) generalize previous models of automobile demand The authors express the indirect consumer utility function, U, for a given automobile as a function of consumer characteristics and product characteristics, and allow for multiple higher level interactions between consumer and product attributes A flexible discrete choice model such as theirs provides the most accurate estimation specification, but it requires a large number of data points across model and time space to achieve sufficient degrees of freedom to estimation A Model for HEV Demand While established automobile demand models are useful for the general automobile market, their application to HEVs is limited by the relatively limited selection of HEV models and their short time on the market, which reduces the number of modelyear sales data points available for analysis As a new technology, HEVs are also not in an equilibrium market, so any model for demand must allow for the effect of diffusion over time Also, it is anticipated that local subsidies and incentives will have a significant effect on demand by, in essence, changing the quality adjusted prices for the HEV option, so it is important not to aggregate out these factors by examining only national level data that is used by most automobile market analyses To overcome these limitations, I focus not on variance in characteristics between different HEV models, but on the variance in tax incentives and consumer characteristics among a cross-sectional dataset of U.S states To develop a model that is suitable for HEV vehicles, I will start with the same basic behavioral principles of Berry, Levinsohn and Pakes [BLP] (1995) In an equilibrium market, the indirect utility of consumer i for any vehicle j can be expressed as: U i j  f  p j , x j ,  j ,  i ;  Equation where p j price of vehicle j x j  observed product characteristics of vehicle j (size, features, power, etc)  j unobserved product characteristics (quality, styling, brand reputation, etc)  i  consumer preferences and socioeconomic characteristics of consumer i and  is a vector of parameters to be estimated Consumer i will choose to purchase vehicle j if and only if: U i j  p j , x j ,  j ,  j ;  U ir  p r , xr ,  r ,  r ;  for r = 0, 1, 2, …., J1; r ≠ j For a given population, the aggregate demand Aj for vehicle j is given by: The choice r = corresponds to the outside alternative of not purchasing any vehicle   A j   : U i j  p j , x j ,  j ,  j ;  U ir  p r , x r ,  r ,  r ;  for r = 0, 1, 2, …., J; r ≠ j and the market share sj of a given model is now a function of: s j  f  p j , x j ,  j ,  ;  Equation It is important to note that in Equation 2, the market share is now a function of the price and characteristics of an individual product j, but of the characteristics of the overall population, so that average consumer preferences and characteristics can be assumed for any market share data point The characterization of market share in equation 2, adapted from BLP (1995) is fairly straightforward, and is a generalized form of earlier specifications for automobile demand At this point, however, I depart from previous studies and adapt equation to analyze HEV market share observations by state First, I will assume that observed and unobserved product characteristic vectors x and ξ not vary among states, because automakers sell the same car models with nearly equivalent prices and specifications in each U.S state.2 Thus, I can reduce equation to: s s j  f  p sj ,  s ;  Equation where the subscript s denotes an observation for an individual state Next, I will assume that for the years of analysis, 2003 and 2004, the differences in characteristics of each HEV model offered are small enough that they can be treated as In reality, there are differences in emissions standards from state to state, with California being the most notable For the purposes of this analysis, however, I will assume that these differences are transparent to consumers and are not factored into buying decisions This may bias the results most significantly in California, where hybrid models for sale have lower GHG and pollution emissions than in other states a homogenous single model I am forced to make this assumption due to a lack of sales data broken down by individual model But, as I explain later in the description of my data set, this assumption is not unrealistic for HEV models in those years This allows me to express equation as: s s h  f  p sh ,  s ;  Equation where the subscript for the model, j, has been replaced by h to represent only HEV market share and prices in a given state, s I can now look at the remaining terms in equation 4, psh and ζs to determine which determinates of these terms will vary among states for HEV vehicles On an individual level, consumers’ preferences for different auto characteristics, ζ, are affected by a number of factors that vary, on average, by state One of the most important is income The demand for any product is generally given as a function of individual income, with the individual demand of person i for product j expected to vary as a function of (yi - pj ) Individuals’ utility for any type of vehicle is also expected to be influenced by benefits or status derived from a particular observed or unobserved product characteristic There are an endless number of ways in which these preferences could change by state or region – such as a greater preference for American-made cars in Detroit or for trucks in rural areas For HEVs, certain states and municipalities provide benefits such as HighOccupancy-Vehicle (HOV) lane privileges, free parking or exemption from emissions testing (DOE State Incentives for Hybrid Electric Vehicles 2004) Additionally, in states where the population tends to be more environmentally friendly, HOV ownership may afford the owner a greater social standing among peers For this analysis, I will lump all of these factors into one variable, B, to indicate the monetary and non-monetary benefits over time of owning an HEV Next, I focus on the price term in equation to bring in a number of other factors that vary by state I decompose the total ownership price into two different components, the upfront cost plus the discounted lifetime fuel costs3: Ph  Pupfront  Pfuel This is where I am able to introduce several other factors that vary by state: tax incentives, gasoline prices per gallon and travel habits The upfront cost for the vehicle will be the purchase price minus any state tax incentives, T, for purchasing HEV vehicles Fuel costs are proportional to the product of the price of gas per gallon, pg and the number of miles traveled during a given period, VMT While there are other factors that influence total ownership costs of hybrid vehicles, such as maintenance costs, battery replacement, registration, etc, I will assume that these terms either not vary by state, or that they affect all vehicles equally and therefore would not affect the total market share distribution Thus, price and individual preference per state in equation can be expressed as:  s  f ( y s , B) p sh  f (T , p gas  VMT ) so that equation becomes: s s h  f  y s , B, T , p gas * VMT ;  Equation There is considerable debate over how consumers factor fuel costs into automobile purchases, and how discount rates and expectations of future prices factor into personal calculations For this paper, I introduce the fuel costs only to the extent that they vary by state and would make consumers in one state more or less likely to purchase a HEV Additionally, I will follow Irvine (1983), and transform market share into a logarithmic odds form This transformation avoids the potential for the model forecasting shares outside the (0,1) interval (Irvine 1983) The resulting equation is:  s ls sh log sh   s sh    f  y s , B, T , p gas * VMT ; t ;   Equation Before taking equation to the data however, I must address the effect of time on the diffusion of HEV technology If HEVs were a mature technology for which consumers had fairly well established preferences, market share would be expected to vary over time only as the prices, performance characteristics and generic options of these vehicles changed HEVs are a new technology, however, which suggests that for any given state, the change in market share over time should follow a classic diffusion pattern to an equilibrium market share value Consumers take time to respond to price and market signals in varying their consumption habits and demand for new technologies Barriers to rapid adoption of HEVs include lack of knowledge by potential adopters as well as variation in the factors that affect different consumers’ individual utility calculations, such as risk tolerance or desire to be an early adopter While some technically savvy or environmentally conscious consumers derive greater utility from their status as early adopters, the utility of other consumers (often called imitators) increases only after they see their friends or neighbors adopting In general, this diffusion pattern over time follows a classic sigmoid or “s-shaped” curve (Jaffe, Newell, and Stavins 2002) Adoption occurs slowly at the onset, increases exponentially, and then tapers asymptotically to a steady state in which the technology Variable T ys HDT, HDH Data blends, including Federal and Local taxes) State-wide tax and other incentives Source DOE EERE and www.electricdrive.org Income per capita US Census Bureau (Table 1: Annual Estimates of the Population for the United States and States, and for Puerto Rico: April 1, 2000 to July 1, 2005 (NST-EST2005-01) 2005) Number of Toyota and Honda dealerships o Toyota Motor Sales, USA per state as a percentage of all new car (Toyota dealerships per state) dealerships o Honda Motor Corporation of America (Honda dealerships per state) o National Automobile Dealers Association (NADA) (total number of new-car dealerships per state I calculated the value of tax incentives as the combined value of Federal plus State incentives, assuming a Federal tax deduction of $2000, a 25% average Federal tax rate, and an average purchase price for State calculations of $20,000 The Federal deduction value was essentially constant for each state and could have been omitted from the linear specification However, I included it in order to provide non-zero values in each observation for use in a log-linear regression The number of available observations is limited to the list of the top 20 States for 2003 and top 15 states for 2004 for total HEV registrations provided by RL Polk in annual press releases While a full list of 50 states would have provided a better number of observations, the 20 states for 2003 accounted for 85% of total US HEV registrations that year, while the 15 states for 2004 accounted for 77% of total registrations The data set also shows a good deal of variance in each of the variables, with the exception of the 13 HOV dummy This dummy variable is positive only for Virginia since no other states allowed an HOV exemption for hybrids that year Because the available data does not break down sales by model, I am forced to treat each HEV vehicle sold as essentially homogenous In 2003, this is a relatively accurate assumption Two HEV models – the Honda Civic Hybrid and Toyota Prius – accounted for 96% of the total sales that year These two models both had an MSRP of around $20,000 and had almost identical fuel economy ratings, size, seating capacity and performance This changed somewhat in 2004, when manufacturers began releasing other HEV models in higher quantities and Toyota redesigned the Prius, creating a larger but more fuel efficient model Toyota dealers in 2004 also began experiencing significant waiting lists for the redesigned Prius, indicating a lack of equilibrium between supply and demand (Adams 2004) While the Prius characteristics for 2004 were different from 2003, the sticker price was still approximately $20,000 (although the transaction prices for many models may have been higher due to dealer premiums and optional features) (Adams 2004) Therefore, I will use the same assumptions for the value of tax incentives, with the understanding that they may be biased low due to higher transaction prices Additional State data broken down by HEV model might allow me to overcome some of these limitations and to obtain a larger number of observations However, such data was not available in the public domain Tables (1) and (2) summarize the data for 2003 and 2004, with states listed in order of decreasing market share Several observations from the raw data are in order First, rankings by market share (hybrid registrations as a fraction of total vehicle registrations) are very different than rankings by overall number of hybrids California is 14 frequently cited as the leading state for hybrid sales, yet from a market share standpoint it ranks only sixth out of the 20 states reported Each of the top four states had a significant statewide tax incentive, and all reported states with tax incentives except for New York were among the top states in market share New York appears to be an outlier, although the effect of its tax credit, income and gas prices are potentially offset by the low miles traveled per year Similar trends among states are observed in the 2004 data, which included only observations of the top 15 states in terms of total vehicles sold This low number of observations tended to bias the data against states with low populations and total automobile sales, so that the top market share state for 2003, Colorado, was not reported in 2004 However, all of the top 15 states for total sales in 2004 also ranked among the top twenty (in total sales) states for 2003 For analysis purposes, I assume that the price that consumers pay for a given hybrid before tax incentives and registration fees is the same from state to state One potential flaw in this assumption would be differences in actual prices charged for vehicles of the same model from state to state due to short-run local variations in supply and demand For example, if local demand for hybrids exceeded short term supply, dealers might have charged premiums over the MSRP or not offered discounts and incentives common on equivalent ICE vehicles For example, when I purchased my own Honda Civic Hybrid in Northern Virginia in late 2002, a dealer told me that short-term local demand was so strong (due to HOV privileges) that he had purchased hybrids from dealers in other parts of the state (with no HOV lanes) that were overstocked and sold them for a markup in Northern Virginia By 2003, however, hybrids appeared to be in stock at almost all dealerships with markdowns from sticker price (as is common in most 15 car models), indicating that manufacturers had adjusted supply shipment patterns to account for local demand Analysis and Results I tested the model in STATA using with the log-odds of marketshare as the dependent variable, and log versions of all independent variables with non-zero values First, I performed OLS regressions on the panel data from 2003 and 2004 separately, to test for the significance of the independent variables in a given year Next, I performed random-effects cross-sectional time series GLS regressions on the combined set of 2003 and 2004 data points The log-linear form allows me to treat the coefficients of each independent variable as the elasticity of demand for that variable with respect to the dependent variable These simple log-linear specifications ignore higher order interactions between variables that would be captured in a more sophisticated regression technique such as translog, Box Cox (Bajic 1988) or BLP (1995), but these techniques generally require a larger dataset than the one available The results of each regression are shown in Table (3) Observations The results varied dramatically between 2003, when hybrid vehicle supplies were generally adequate to meet market demand, and 2004, when high demand for the redesigned Toyota Prius resulted in dealer waiting lists for vehicles In 2003, average income, effective tax incentives, gas prices, and miles traveled per year are all significant (at least to the 10 level) and have positive coefficients, indicating that an increase in any of these factors increases the odds that a consumer would have purchased a hybrid 16 vehicle over a conventional ICE vehicle Even though tax incentives appear to be significant in predicting the market share of hybrids, the magnitude of the coefficient for tax incentives (.137) is much smaller than the coefficient for the gas price*VMT interaction term If the coefficients are interpreted as elasticities, this result suggests that the elasticity of market share with respect to tax incentives was much less than with respect to gas prices or income7 The relative proportion of Toyota dealerships (but not Honda dealerships) was also significant with a positive coefficient, indicating that HEV market share was likely to be higher in states where Toyota dealerships made up a larger proportion of all new car dealerships In 2004, however, the results changed dramatically The effect of Toyota dealership proportion was the only significant variable, explaining over 75% of the variance in hybrid sales among the 15 state observations in the regression The original predictors were actually significant in an earlier regression (results not shown) that did not include the dealer ratios, but when the Toyota dealership ratio is included it dominates everything else in the regression for 2004 The GLS regression on the combined dataset for 2003 and 2004 showed similar results to 2003 Time was included in the first multiyear regression [column labeled (3) in table 3], but was removed in the second regression [column labeled (4)] due to suspected multi-collinearity between time and several other variables (gasoline prices and percapita income) As expected, the removal of time resulted in higher significance levels and larger-magnitude coefficients for the remaining time-dependent variables The regression results mirrored the findings of 2003, which indicated that tax incentives, gas The regressions shown in table used the log-odds of market share as the dependent variable Additional regressions (not shown) using the log of market share as the dependent variable yielded similar results in terms of significance and relative magnitude of coefficients 17 prices*VMTCP and the Toyota (but not Honda) dealership fraction were all significant, with gas prices having the greatest effect HOV privileges were moderately significant, but there is only one data point – Virginia All other HOV exemptions only went into affect recently when Federal law (Energy Policy Act of 2006) authorized states to grant HOV exemptions to hybrids While it is important to control for HOV privileges, the lack of additional data points makes it difficult to provide a meaningful interpretation of the coefficient value for this dummy variable Therefore, the HOV dummy variable more accurately encompasses all of the other effects for Virgina that influence preference for Hybrids, including the HOV priviledges Because HOV privileges only provide utility if the vehicle owner lives or works close to an HOV lane, a study of the true relationship of HOV privileges on Hybrid sales would require a spatial analysis using highly disaggregated data that was not readily available The change in the regression results from 2003 to 2004 is striking, and may be an effect of the inability for Toyota to meet consumer demand for the redesigned Prius model that was introduced in early 2004 If there were waiting lists in all states and Toyota distributed the available supply of cars evenly between all of its dealerships, it would follow that states that had more Toyota dealerships (relative to the total number of car dealerships in the state) would receive more cars, resulting in a higher market share Once all states had waiting lists, the factors that were significant in 2003 (tax incentives, gas prices, income) might still have affected the size of the waiting list, but not the actual delivery pattern for the vehicles This occurs because the total number of car dealerships in each state is highly correlated to the total number of cars sold and registered in the state, with R2= 0.86 18 There is also the possibility of endogeneity issues relating to state tax incentive policies The analysis assumed that state incentive policies were set independent of consumer preferences for Hybrids in each state However, in states where other factors were likely to increase the demand for HEVs, consumers may have placed additional pressure on lawmakers to enact incentive laws In further research, I may address this problem through qualitative analysis of the deliberation processes that resulted in incentive policies in various states Implications for Policymakers If as the results suggest income and gas prices are both effective at influencing the adoption of HEVs when there are adequate supplies, policymakers have several alternatives to promoting adoption While tax incentives may be significant, if the magnitude of their effect is low they may, in essence, more to reward people who would have purchased hybrid vehicles already than to induce new sales, as discussed previously Gas taxes (or simply the market force of rising gasoline prices) may actually have a greater effect on HEV sales, with the added benefit of shifting sales to more fuel efficient vehicles in general (not just HEVs), as well as inducing the innovation of additional fuel efficiency technologies (Newell, Jaffe, and Stavins 1999) However, recent increases in gas prices due to market forces appear to be greater than any gas taxes that have been seriously considered at the state or Federal level, so the induced efficiency effect will likely occur in the absence of any government imposed taxes or incentives Additionally, attempts at implementing gas taxes may much greater resistance from the public than attempts to increase spending for tax credits This occurs both because of the 19 greater visibility of tax increases to voters, and because the overall costs to a given consumer from gasoline tax payments may exceed the proportional cost of the tax credits when vehicle sales are low The results of the 2004 analysis raise another issue for policymakers When HEV supply is constrained, sales patterns may be determined not by consumer demand signals but by internal distribution policies by automakers In this case, gasoline and tax incentive policies will be ineffective in achieving the overall aim of state HEV policies -placing more HEVs on the road in a given state However, they may have an effect on consumers’ decisions to enter the waiting list for a hybrid, resulting in greater sales in the future if automakers clear their demand backlogs In addition to tax incentives benefiting consumers who would have bought hybrids even without an incentive, these incentives may also act as a subsidy to automakers Without the subsidy, automakers might have otherwise lowered their prices to sell more hybrids However, the effect of subsidies on automaker pricing is the result of a complex interaction of consumer and producer behavior that is beyond the scope of this paper Future Research I conducted my analysis based only on a small sample of data A more comprehensive data set that includes information from all fifty states, broken down by specific hybrid models would allow me sufficient data points to test specifications that allow for greater interaction among independent variables, as well as to explore the effects of specific vehicle models and characteristics on HEV demand I suspect that tax and monetary incentives might be more significant in explaining sales of HEVs that have 20 close conventional engine substitutes (such as the Honda Civic and Ford Escape) than those vehicles (such as the Prius) that are marketed as unique models with a greater degree of “green” status Sales or registration data disaggregated to the county or zip code level might also allow me to test for other factors such as local incentives or consumer preferences that might be aggregated out at the state level Although I have confined my analysis to one particular type of technology – HEV vehicles – the method I use could reasonably be applied to any technology market which demonstrates a large cross-sectional variance in monetary incentives and resource prices As states begin to take a more active role in promoting the adoption of environmentally friendly technologies, there may be a rich source of data to analyze the effectiveness of a variety of policy tools and market intervention mechanisms 21 Figures and Tables Figure 1: The effect of incentives on market share for a notional sigmoid diffusion process 22 Table 1: Data Summary for 2003 23 Table – Summary of Data for 2004 * 2003 data used in place of 2004 data, which was not available 24 Table 3: Regression results Dependent variable = Log(sh/(1-sh)) Number of Data Points, N= 2003 OLS (1) 2004 OLS (2) Multiyear GLS (3) Multiyear GLS (year omitted) (4) Coefficient Coefficient Coefficient Coefficient (SE) (SE) (SE) (SE) 20 15 35 35 Intercept -26.38275*** (8.822079) 5.150081 (17.15024) -19.27229** (7.784836) -30.82276 (2.529215) Log(total tax incentive) 1368939** (.0537138) 9997385** (.3439703) 8970666 * (.4346294) 2482092+ (.1460412) -.0572118 (.0933414) 5977028*** (.1638254) N/A -.0057312 (.0678471) -.3211936 (.7476541) -.0327093 (.5660131) 3207999* (.150124) 0398707 (.093676) 1.171801*** (.3164756) N/A 1219069** (.0570762) 6817194** (.3270778) 6454412* (.389763) 2432689+ (.1540351) -.0366586 (.0991828) 652556*** (.1748452) 1131209+ (.0720737) 1258088** (.0542676) 1.163796 *** (.1141914) 1.111147*** (.2420704) 2011497+ (.1438903) -.0431437 (.0940568) 5990637*** (.1631853) N/A 0.7377 7702 Log(gas price *VMTPC) Log(per capita income) HOV access dummy Log (Honda Dealer ratio) Log(Toyota Dealer Ratio) Year Adj R2 Significance levels + < 17 * < 10 ** < 05 *** < 01 8480 8450 25 References 2005: Year of the Hybrid 2004 Rockville: ChangeWave Research Adams, Noel 2006 Prius Madness, April 9, 2004 2004 [cited Dec 2006] Available from http://www.evworld.com/view.cfm?section=article&storyid=679 Argote, Linda, and Dennis Epple 1990 Learning Curves in Manufacturing Science 247 (4945):920-924 Bajic, Vladimir 1988 Market Shares and Price-Quality Relationships: An Econometric Investigation of the U.S Automobile Market Southern Economic Journal 54 (4):888-900 Bass, Frank 1969 A New Product Growth Model for Consumer Durables Management Science 15 (5):215-227 Baumol, William 1972 On Taxation and the Control of Externalities The American Economic Review 62 (3):307-322 Berry, Seven, James Levinsohn, and Ariel Pakes 1995 Automobile Prices in Market Equilibrium Econometrica 63 (4):841-890 Burke, Andrew 2004 Present Status and Marketing Prospects of the Emerging Hybrid Electric and Diesel Technologies to Reduce CO2 Emissions of New Light-Duty Vehicles in California Davis: Institute of Transporation Studies, University of California, Davis Canes, Michael E 2003 Economics of Hybrid Electric Vehicles IAEE Newsletter, Second Quarter, 13-17 Cao, Xinyu, and Patricia Mokhtarian 2004 Future Demand for Alternative Fuel Passenger Vehicles: A Diffusion of Innovation Approach: UC Davis - Caltrans Air Quality Project Chow, Gregory C 1957 Demand for Automobiles in the United States: A Study in Consumer Durables Amsterdam: North-Holland Publishing Company Highway Statistics 2003 2004 edited by F H A US Department of Transportation Highway Statistics 2004 2005 edited by F H A US Department of Transportation Hybrid Vehicle Registrations Increase 81% in 2004 2005 Southfield, MI: R L Polk Hybrid Vehicle Registrations up 25.8% in 2003 2004 Southfield, MI: R L Polk Irvine, F Owen Jr 1983 Demand Equations for Individual New Car Models Estimated Using Transaction Prices with Implications for Regulatory Issues Southern Economic Journal 49 (3):764-782 Jaffe, Adam B., Richard G Newell, and Robert N Stavins 2002 Environmental Policy and Technical Change Environmental and Resource Economics 22:41-69 Jaffe, Adam B., and Robert N Stavins 1994 The Energy Efficiency Gap: What does it mean? Energy Policy 22 (10):804-810 Newell, Richard G., Adam B Jaffe, and Robert N Stavins 1999 The Induced Innovation Hypothesis and Energy Saving Technological Change The Quarterly Journal of Economics 114 (3):941-975 Norberg-Bohm, Vicki 1999 Stimulating "Green" Technological Innovation: An analysis of alternative policy mechanisms Policy Sciences 32 (1):13-38 Ogden, Joan M., Robert H Williams and Eric D Larson 2004 Societal lifecycle costs of cars with alternative fuels/engines Energy Policy 32:7-27 26 State Incentives for Hybrid Electric Vehicles 2005 Department of Energy 2004 [cited October 2005] Available from http://www.eere.energy.gov/cleancities/incen_laws.html Stoneman, Paul, and Paul Diederen 1994 Technology Diffusion and Public Policy The Economic Journal 104:918-930 Table 1: Annual Estimates of the Population for the United States and States, and for Puerto Rico: April 1, 2000 to July 1, 2005 (NST-EST2005-01) 2005 Washington, DC: US Census Bureau 27 ... time, total cumulative share of hybrids in a state, scum, and any other factors that affect the knowledge and visibility of hybrids among potential consumers For this paper, data on For a further... buy a hybrid with a monetary incentive than without one In fact, market share for certain hybrid models has fluctuated considerably since their introduction 10 cumulative state shares of hybrids... (hybrid registrations as a fraction of total vehicle registrations) are very different than rankings by overall number of hybrids California is 14 frequently cited as the leading state for hybrid