RESEARCH ARTICLE Why is the market for hybrid electric vehicles (HEVs) moving slowly? Djamel Rahmani 1 , Maria L. Loureiro 2 * 1CREDA-UPC-IRTA, Edifici ESAB, Parc Mediterrani de la Tecnologia, C/Esteve Terrades, Castelldefels, Barcelona, Spain, 2Departamento de Fundamentos da Ana ´lise Econo ´mica, Facultade de C. Econo ´micas e Empresariais, U. Santiago de Compostela, Santiago de Compostela, Spain *[email protected] Abstract Hybrid electric vehicles (HEVs) could be a good short term option to help achieve global targets regarding road transport greenhouse gas emissions. Several common and countryspecific public policies based on price or tax rebates are established in order to encourage the adoption of HEVs. The present research empirically assesses market preferences for HEVs in Spain, looking at the role of subsidies. An interactive internet-based survey was conducted in a representative sample (N = 1,200) of Spanish drivers. Drivers are willing to pay an extra amount of €1,645 for a HEV model compared to a conventional vehicle, premium which is well below the price markup for these cars. Therefore, current levels of economic subsidies applied in isolation to promote these types of vehicles may have a quite limited effect in extending their use. Overall, it is found that drivers have clear misconceptions about HEVs, which affect their purchasing choices and perceptions. Therefore, a policy mix of various incentives (including informational campaigns) may be required in order to stimulate the demand for HEVs. 1. Introduction In December 2015, a total of 195 countries ratified a universal agreement to combat climate change at the Climate Summit in Paris (COP 21), where they expressed their willingness to move together towards a low carbon economy. The European Union (EU) has announced its plan to achieve its ambitious challenges, including the reduction of its greenhouse gas emissions by 40% by 2030 from the 1990 level; improving energy efficiency by 40%; and increasing the contribution of renewable energy in its energy consumption by 27% [1]. One of the priorities is the transport sector, due to its significant contribution to global warming, and air pollution (Directive 2009/28/EC; 2009/30/EC; Directive 2009/33/EC). In particular, road transport is a major source of greenhouse gas emissions in European cities, being responsible for one fifth of the EU’s total emissions of carbon dioxide (CO2), the main greenhouse gas. Despite a slight decrease in the last few years, these emissions are still 20.5% higher than in 1990 [2]. The consequences of air pollutants generated by the transport sector on PLOS ONE | https://doi.org/10.1371/journal.pone.0193777 March 21, 2018 1 / 14 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Rahmani D, Loureiro ML (2018) Why is the market for hybrid electric vehicles (HEVs) moving slowly? PLoS ONE 13(3): e0193777. https://doi.org/10.1371/journal.pone.0193777 Editor: Xiaosong Hu, Chongqing University, CHINA Received: October 23, 2017 Accepted: February 16, 2018 Published: March 21, 2018 Copyright: ©2018 Rahmani, Loureiro. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: Study data is available at figshare (DOI:10.6084/m9. figshare.5789586). Funding: This work was supported by Fundacio ´n Ramo ´n Areces (http://www.fundacionareces.es), Dr. Maria Loureiro; Secretarı ´a de Estado de Investigacio ´n, Desarrollo e Innovacio ´n, ECO201679446-R, Dr. Maria Loureiro. Competing interests: The authors have declared that no competing interests exist.

human health are an increasing cause for concern. According to the European Commission [3], air pollution causes the premature death of more than 400,000 people in Europe every year. However, and despite the “Dieselgate” scandal, diesel vehicles represented still 52% of all Europe’s new registrations in 2015. While in United States, Chinese, and Japanese, the new registrations in the same year were dominated by gasoline vehicles [4]. The registrations of alternative fuel vehicles (AFVs) have increased in EU member states and European Free Trade Association (EFTA) countries. In particular, Italy (14.1%), Norway (12.6%), and Poland (8.1%) registered the highest growth in the AFV new registrations in the period from 2013 to 2015 [5]. However, the growth rate of the AFV new registrations was very low in most member countries. The rise of the AFV diffusion is due to several factors, including the availability of a wide range of models in the market, but also in part to different government incentives based on direct rebates in purchase prices, tax reductions, free parking, and access to priority lanes, among others [5]. The plug-in hybrid electric vehicles (PHEV), hybrid electric vehicles (HEVs), and electric vehicles (EVs) accounted for only about 2.6% of 2015 new vehicle registrations in the EU. Thanks to the CO2-based vehicle taxation scheme, Netherlands led the PHEVs and EVs sales in 2015 with contributions of 8.8% and 0.9%, respectively. Among the EFTA countries, Norway is a clear example in terms of fiscal incentives to promote the adoption of AFVs, especially PHEVs and EVs. In fact, 22% of 2015 new vehicle sales in Norway were PHEVs and EVs [4]. Among the three models, HEVs are the most sold in EU, being the highest sales in 2015 those registered in the Netherlands (3.3%) and France (2.2%). HEVs represented 25% of 2015 Toyota vehicles sold in the EU. The HEVs are also the bet of many countries like Japan and U. S. where their 2015 market shares were around 22% and 5%, respectively [4]. Nevertheless, up to now, and despite the existence of various stimuli, the market penetration for EVs, HEVs and PHEVs is still quite low in most countries. This study explores the potential reasons behind such a low adoption rate. In particular, it explores the relevant factors that drive people’s vehicle choices, especially those that play a key role in preferring HEVs over conventional vehicles, looking not only at economic incentives but also at perceptions and knowledge about HEVs. A HEV (non-plugin) is an AFV which uses internal combustion engines and electric batteries. It uses braking energy, which is normally wasted, to recharge the battery. HEVs offer economic and environmental advantages over conventional cars, ceteris paribus. The engineering literature has produced very relevant references providing accurate and technical information about the advantages in terms of current efficiency, environmental performance and future possibilities of HEVs. [6,7,8,9]. In general, HEVs are cheaper than EVs and PHEVs; they do not suffer from battery problems or lack of infrastructures, and benefit from public incentives in many countries. Therefore, they should be very competitive with respect to diesel and gasoline vehicles. It has not been clear why drivers have avoided switching to HEVs. Is it primarily related to their price? And if so, what types of incentives are needed to encourage drivers to switch to HEVs? Or is it based on other misconceptions and concerns associated with HEVs that may be more important than the price markup? Based on a discrete choice experiment (DCE) included in an extensive online questionnaire, the present paper aims to provide some insight into these questions, as well as the type of incentives that are required in order to galvanize the HEV market. Specifically, it explores preferences towards car attributes, including fuel consumption and CO2 emissions (improved in HEVs). In addition, it tests whether drivers’ perceptions towards subsidies may encourage demand for HEVs. Finally, individual heterogeneity in preferences for car attributes, including the price–a factor that is often overlooked–is considered by specifying a random parameter logit (RPL) model. The market for hybrid electric vehicles PLOS ONE | https://doi.org/10.1371/journal.pone.0193777 March 21, 2018 2 / 14

2. Literature review Many contributions explored consumers’ preferences for different AFVs including EVs, HEVs, PHEVs, liquefied petroleum gas, compressed natural gas, biofuel, and hydrogen powered vehicles [10,11,12,13,14]. The results detected heterogeneous individual perceptions for different AFVs, with conventional cars remaining the most attractive option. From the extensive existing literature on car choices, few studies [15,16,17,18,19] have specifically investigated consumers’ preferences for hybrid cars (HEVs or PHEVs). Erdem et al. [17] used a contingent valuation method to estimate the willingness to pay (WTP) for HEVs in Turkey. The results showed that people were willing to pay an average premium of US$ 858 to change to a HEV. Thatchenkery and Beresteanu [19] explored HEV demand in the USA using the United States 2006 Polk New Vehicle Registration Cross-sectional Data. They showed that people were sensitive to fuel efficiency, but were more sensitive towards horsepower and weight. Axsen et al. [15] combined the revealed preferences (recent car purchases) and stated preferences of Canadian and Californian car owners to explore how consumer preferences for HEVs have shifted (specifically focusing on the neighbor effect) as HEV market penetration increased. The results showed that the WTP for HEVs rose with the market share. Chua et al. [16] employed scales and items to compare HEV and conventional car buyers in Australia. The results from factor analysis showed that while preferences for conventional cars were more sensitive to variations in quality and performance, and less sensitive to image and social influence, HEV buyers placed a great amount of importance on their ‘green’ image and social influence, and little importance on quality and appeal. Heffner et al. [18] used informal face-to-face interviews to investigate whether the ‘green’ social image influenced United States households to adopt HEVs. The results showed that all HEV owners placed some importance on the ‘green’ image of their cars, although they did not adopt HEVs by only focusing on their image. The present research joins this line of studies by adding a number of contributions. First, it assesses the heterogeneity of preferences towards car attributes across drivers. Second, this study investigates the importance of incentives when buying efficient cars, and explores whether these incentives increase the demand for HEVs. Third, results are contextualized in the current market conditions. 3. Case study and policy context The present research was conducted in Spain. The current economic crisis has resulted in the Spanish vehicle fleet being one of the oldest in Europe, currently with an average car age of 11.3 years, and with emissions that affect seriously the air quality in cities [20]. Vehicles older than 10 years account for 50% of all cars circulating in Spain [21]. Driving vehicles of this kind multiplies the environmental damage caused by road transport. In this context, various strategies have been promoted in Spain, including the Movele and the Pive public programs, both aimed at promoting the market adoption of efficient cars. The Pive program is designed to encourage the acceptance of HEVs, PHEVs, EVs, and extended range electric vehicles. Currently, the Pive program [22] offers a discount of 1,500 on the purchase of a new vehicle, after turning in a private car over 10 years of age, or any commercial vehicle over 7 years of age. The vehicle purchased must be new and an efficient model (EVs, HEVs, PHEVs, or using alternative fossil fuels). As a result of this program, the Institute for the Diversification and Energy Saving (IDAE) estimated that from 2012 to 2014 the Pive plan has led to the replacement of 715,000 old vehicles, saved 248 million liters of fuel per year, and reduced greenhouse gas emissions by 513,000 tons of CO2 per year [22]. The Spanish government also reformed the car registration tax (Law 34/2007 of 15th of November on air quality and protection of the atmosphere), making it inversely proportional to the amount of CO2 emissions (0% for emissions lower than 120 g/km, 4.75% of the The market for hybrid electric vehicles PLOS ONE | https://doi.org/10.1371/journal.pone.0193777 March 21, 2018 3 / 14

value of the car for emissions between 120 and 160 g/km, 9.75% for emissions of between 160 and 200 g/km; and finally, 14.75% for emissions of 200 g/km or higher). In spite of these current public policy efforts to encourage drivers to adopt HEVs, these are still not particularly popular in the Spanish market. In 2016, HEVs only accounted for 2.70% of new passenger car sales [23]. The Japanese Toyota brand led the sales of HEVs in Spain, with a market share of more than 70%. Together with its premium brand Lexus, they accounted more than 80% of the total HEVs sold in Spain in 2016. While the vast majority of HEVs sold were gasoline, diesel hybrid cars only represented 6% of the total units sold [23]. 4. Survey design Data from drivers were collected using an online survey directed to a representative sample of drivers over the age of eighteen. The survey was administered to 1,200 residents in Spain. The number of fully completed and useful questionnaires was 1,016. The survey asked drivers to provide information about several car related issues, including current car(s) ownership, brand preferences, awareness of energy consumption issues, and their environmental attitudes. Next, the survey provided information about HEVs, asking about their intentions and plans for future car-purchases, including a DCE to elicit preferences to buy a future car. It concluded with the socio-demographic characteristics of the driver. Participants were asked about what size (small, medium or large) they would prefer to have their next car? And thanks to the interactive aspect of the questionnaire, this information was received immediately, and automatically, depending on their answer, they were assigned to one of the two possible versions of the DCE survey. In particular, one was designed for drivers interested in buying small or medium-sized cars and the other for those who were willing to buy large cars. A total of 875 drivers (86.12% of the completed surveys) expressed their desire to buy a small or midsize car in the future, while only 138 drivers (13.58% of the completed survey) stated their wish to adopt a large size car in the future. The survey questions were common to the participants. The only difference between the two versions was the levels of the attributes included in the DCE. In this paper, data from the survey completed by those drivers willing to buy a small and medium-size car are analyzed. 4.1 Experimental design and DCEs A DCE is used as it is the more appropriate way for measuring consumer welfare, and its results are more consistent with the economic theory than a traditional conjoint analysis [24]. In addition, HEVs have a small market share, and revealed preference data sources are still scarce. The DCE method is based on the assumptions of economic rationality and utility maximization [24]. It consists of presenting drivers with several car alternatives, and asking them to choose one of them based on their preferences. Each individual is expected to choose the alternative that maximizes his/her utility. Moreover, the utility derived from an alternative is assumed to depend on the marginal utilities associated with its attributes [25]. As a HEV is a quasi-public good, both economic attributes and environmental (non-economic) attributes are included. In the survey, and prior to the DCE exercise, participants were familiarized with HEVs and the expected consumption and emissions for a mid-size car. They were also required to assume that all non-specified attributes remained constant across alternatives. A DCE was then carried out, in which the participants could select between a regular vehicle and a HEV, or just remain with the status quo option (neither car). Focus groups, pilot surveys and previous studies were used in order to identify the most relevant attributes and suitable levels for our DCE exercise. Previous studies [13] summarized the determinant factors of a car choice process mainly into economic attributes The market for hybrid electric vehicles PLOS ONE | https://doi.org/10.1371/journal.pone.0193777 March 21, 2018 4 / 14

(purchase price, fuel cost), non-economic attributes (refueling or recharging time, availability of fuel or recharging opportunities, technological performance), and environmental attributes (emissions). Besides the type of vehicle, two economic attributes have been included: price and fuel consumption, factors that are highly and primarily valued by drivers when considering the purchase of AFVs [26]. Apart from the monetary attributes, each choice set included two non-monetary or environmental attributes. The environmental attributes included were carbon dioxide (CO2) emissions, which were found to be significant in earlier studies [13], and the option of biofuel adaptation (flex-fuel), which is a recent trend in carmakers. In fact, European legislation (Directive 2003/30/EC) and national legislation (Spain’s Royal Decree 61/2006) allow carmakers to incorporate bio-fuel directly into conventional fuel without the need for specific labeling, unless the proportion exceeds 5%. Some existing studies [10,27] have explored preferences for biofuel cars, although it has never been investigated as an additional attribute to conventional and HEVs. The attribute levels are based on information obtained from car suppliers in the Spanish market for small and midsize cars. This information is used to determine 2 levels of vehicle type (regular or HEV) and 3 levels of prices used in the analysis: a low price level (12,000), a medium price level (16,000) and a high price level (20,000). The mid-price level considered corresponds to the average price of new cars sold in Spain in 2012. From 2009 to 2013, most of the new cars sold in Spain (80% of the total) were priced below 20,000, due in part to the decrease of purchasing power of consumers caused by the economic crisis. For these reasons, and given the focus of this work (analyzing the demand for small and medium HEVs), the upper price level is set at 20,000 and the lower price level at 12,000. The fuel consumption attribute was expressed as fuel cost () per 100 kilometers [10,28]. This unit is used because drivers tend to remember how much fuel their car consumes in terms of euros/kilometers. The fuel cost was computed as the product between the numbers of liters of fuel the vehicle would require to travel 100 kilometers, and the average fuel price in Spain (1.35 per liter at the time of the study). Similarly, the CO2 emissions were expressed as grams of CO2 per kilometer [10,28]. Again, for simplicity, and for the purposes of this research, only two emission levels are included: a more efficient level (100gr per kilometer) and an inefficient level (150gr per kilometer). Finally, the presence or absence of the potential of biofuel adaptation corresponded with the two dichotomous levels specified for the corresponding attribute. The combination of these five attributes and their levels, using SPSS orthogonal main effects design and then the procedure of Street and Burgess [29] (vector of differences = 12111), generated an optimal orthogonal design (OOD). The OOD is constructed so as to maximize the differences in the attribute levels across alternatives, and therefore, maximize the information from each respondent, forcing the tradeoffs of all attributes in the experiment [30]. It should note that this design fits best choices where each alternative has the same number of attributes, and each attribute has the same number of levels. The final design contained 8 choice cards with a design efficiency of 98%. Each respondent was presented with a total of 8 choice cards, a reasonable number that does not affect data quality [31]. Fig 1 shows an example of a choice card. The no-choice alternative (neither car) was provided in order to make the choice decisions very similar to market decisions (or more realistic). 4.2 Choice modeling specification Assuming utility maximizing behavior, the empirical applications based on discrete choice models make possible to estimate the probability that an individual chooses a given car alternative, among a set of available alternatives. The utility that an individual i derived from choosing The market for hybrid electric vehicles PLOS ONE | https://doi.org/10.1371/journal.pone.0193777 March 21, 2018 5 / 14

a car alternative j among a set of J alternatives (conventional car, HEV or no-choice option) in each choice situation t may be expressed [32,33] as a sum of an observable utility component (X0 ijtb) and unobservable component or error term (ε ijt ): Uijt¼X0 ijtbþεijt ð1Þ where, X0 ijt: is a vector of specific car attributes and specific individual characteristics. β: is a vector of parameters associated with the explanatory variables. The multinomial logit model (MNL) [34] is derived assuming that the error terms (ε ijt ) are independently and identically extreme value type I distributed (IID). The MNL probability of choosing an alternative a among a set of J alternatives is given by [32,33]: Liat¼expðX0 iatbÞ X J j¼1 expðX0 ijtbÞð2Þ The MNL is based on the assumption of the independence of irrelevant alternatives (IIA). The MNL imposes homogeneity in tastes, inflexible substitution patterns in preferences between different alternatives and independence in unobserved factors over time [35]. An alternative model which is much more flexible and which overcomes the limitations of MNL is the RPL. In addition to the fact of not complying with the IIA property, the RPL allows for: a) random heterogeneous preferences across individuals, b) unrestricted substitution patterns, and c) correlation in unobserved factors over time [35]. The RPL model probability (unconditional probability) is the integral of the conditional probability over all the random parameters [32,33]: Piat¼ZexpðX0 iatbÞ X J j¼1 expðX0 ijtbÞ fbjyÞdbð3Þð where, Fig 1. Choice experiment question and card example. https://doi.org/10.1371/journal.pone.0193777.g001 The market for hybrid electric vehicles PLOS ONE | https://doi.org/10.1371/journal.pone.0193777 March 21, 2018 6 / 14

f(β|θ): is the density function of the parameters β. This density function may be assumed to follow any closed-form distribution (normal, log-normal, triangular, uniform) [36,35]; θ: are the parameters (mean and standard deviation) of the distribution. In this application, RPL models are estimated assuming log-normal distribution for the coefficients associated with price (PRICE), fuel consumption (FCONSUMPTION) and CO2 emissions (CO2) in order to force them to be negative (on one side of zero) for all individuals. In the same line as previous studies [10,37,12], positive preferences for these attributes are not allowed, as it is not expected that people would prefer higher prices, higher fuel consumptions or higher CO2 emissions. Several distributions (normal, log-normal, uniform, triangular, etc.) are also tested for the coefficient associated with biofuel adaptation (BADAPTATION) but its standard deviation was not statistically significant. Thus, it is considered as a nonrandom parameter. This valuation exercise also aims to predict respondents choices between the two car-alternatives (conventional, and HEVs) and the no-car option (neither A or B), including in the DCE models a no-choice-specific constant (ASC), denoting the election of the status quo option. It is assumed that the no-choice-specific constant follow a normal distribution because drivers may like or dislike staying or not with their current cars. In addition, it explores how preferences for the no-car option, compared to the car options, shift with the following socioeconomic variables: gender (MALE), age (AGE), and monthly income under 1,800 (LHINC). It also analyzes the heterogeneity in preferences for the no-car option among drivers who reported that incentives such as direct subsidies (SUBSIDY) would be important factors when buying an efficient car. This incentive variable was created from the participants’ ratings, when they were asked to state how important (on a 5-point Likert scale: from 1 “not important” to 5 “very important) this factor would be in their decision to select efficient cars. Furthermore, and after the vector of parameters is obtained, the WTP welfare measures are calculated in order to determine the monetary equivalent of the marginal utilities placed by drivers in each car attribute improvement. This step may provide important information to policy makers regarding the economic efforts that people are willing to make to acquire HEVs and some improvements in car attributes. WTP for a HEV compared to conventional vehicle is generally computed as the difference (MU HEV – MU C ) between the marginal utility obtained for HEV (MU HEV ) and conventional vehicles (MU C ). Such values are obtained substituting in (1) the estimated parameters of our empirical model, and then this difference in utilities is divided by the estimated price coefficient (β PRICE ) [33], as shown by the following formula: WTPHEV¼  ðMUHEVMUCÞ bPRICE ð4Þ In a RPL model, when the numerator and the denominator included in Eq 4 are random, the expression of the WTP ratio shown in Eq 4 becomes a randomly distributed term. In this case, Daly et al. [38] advised to ensure finite moments for the WTPs. In the present application, WTP measures are constructed based on unconditional parameter estimates [39], because they allow for prediction outside of the sample, unlike conditional parameter estimates, which only predict within the sample [40]. Deriving WTP based on unconditional parameter estimates requires the population to be simulated [39]. Both, numerator and denominator of (4) have been simulated employing random draws coming from the log-normal distributions defined by the estimated parameters. Draws were generated from both, the numerator and denominator, computing their respective ratio in each draw, as in Hensher et al [39]. The market for hybrid electric vehicles PLOS ONE | https://doi.org/10.1371/journal.pone.0193777 March 21, 2018 7 / 14

5. Data and results Table 1 summarizes the drivers’ perceptions towards HEVs (prior to the information received in the survey, given that participants could select several statements that they considered correct when thinking about HEVs). When compared to conventional cars, 62% of the drivers perceived HEVs as being more expensive, although 28% stated that HEVs have low running costs. In addition, 14% believed that HEVs are slower, and 18% considered that HEVs have less power. These negative perceptions can be an obstacle to introduce HEVs in a country where drivers are in general “speed lowers”. Furthermore, it is worrisome that 16% reported that they did not know what HEVs are like. Finally, 17% reported that HEVs have limited autonomy, showing a clear misunderstanding of the difference between HEVs and EVs. Table 2 describes some drivers’ socio-demographic characteristics and the variables included in the empirical models, containing basic information about the rated importance of incentives including direct subsidies, registration tax exemption, free parking, access to priority lanes, and social image. In summary, public policies based on direct economic incentives, such as subsidies or allowing free parking are perceived as the most important incentives for drivers to buy an efficient car. In terms of socio-demographics, the average age of participants in this sample is 46 years, and 51% of the participants were male. One fifth of the participants were unemployed, and about 50% of all households received a monthly income of less than 1,800. The participants reported that on average they drive a car 4 days a week. The sample was representative of the profile of a Spanish driver at least with respect to some important characteristics, such as age and driving frequency. The Spanish Observatory of Drivers [41] defined, through a representative study, a typical Spanish driver as being a 44-year-old male, who uses a car an average of 5 days a week for work. Table 3 summarizes the results of the estimated models. First, a MNL model is estimated and the assumption of independent irrelevant alternatives (IIA) is tested using the Chi-squared Hausman and McFadden test. The results of this test reject the IIA assumption [being the omitted alternative the regular car: Chi-squared (5) = 156.808; with the omitted alternative being the HEV: Chi-squared (5) = 160.883; the 99%; critical value: Chi-squared (5) = 15,086]. Then, to improve the performance, RPL models have been estimated, allowing for correlation over time (but with uncorrelated parameters) and using NLOGIT.5 software with 2000 replication draws in the estimation processes. In particular, a baseline RPL and a RPL model with heterogeneity in the mean of the random parameter associated with the no-car specific constant are specified. In Table 3, Column 1 shows the results of the MNL; the RPL results are presented respectively in Column 2 (baseline RPL), and Column 3 (extended RPL, with interaction terms with Table 1. Perceptions for hybrid electric vehicles (HEVs). Participants’ hybrid car Perceptions (1 = yes, 0 = no) Mean Std. Dev. Compared to a conventional car: A hybrid car is more expensive .620 .485 A hybrid car has lower running costs .281 .449 A hybrid car is slower .142 .349 A hybrid car has less autonomy .168 .374 A hybrid car is less powerful .176 .381 A hybrid car is less safe .013 .116 I do not know what a hybrid car is .158 .365 https://doi.org/10.1371/journal.pone.0193777.t001 The market for hybrid electric vehicles PLOS ONE | https://doi.org/10.1371/journal.pone.0193777 March 21, 2018 8 / 14

the constant). According to the values of the log-likelihood, adjusted pseudo-R 2 , Akaike information criterion (AIC), the RPL improves the MNL model fit (which results are not directly discussed). The moments of the coefficients associated with PRICE, FCONSUMPTION, and CO2 which are calculated converting the log terms are presented in this table. Table 2. Descriptive statistics of the variables included in the RPL model. Variable Description Mean Std. Dev. PRICE price of car-option divided by 10,000. 1.066 .805 FCONSUMPTION euros spent in fuel consumption per 100km. 4 2.943 CO2 grams of CO2 emitted per 1km. 83.333 62.362 BADAPTATION 1 if car-option is adaptable (flex-fuel) to run with biofuels and 0 otherwise. .333 .471 ASC no-car-option constant. .333 .471 MALE 1 for male and 0 otherwise. .513 .499 AGE age of participants (years). 45.972 13.546 LHINC 1 for monthly income under 1,800 and 0 otherwise. .505 .499 SUBSDY importance (score) attributed to the incentive “direct subsidies”. 4.199 .961 https://doi.org/10.1371/journal.pone.0193777.t002 Table 3. Results of estimated MNL and RPL models. MNL Baseline RPL RPL Parameters in utility functions Coeff. Std. Error Prob. |z|>Z Coeff. Std. Error Prob. |z|>Z Coeff. Std. Error Prob. |z|>Z PRICE -2.034 .055 .000 -2.450 .059 .000 -2.371 .051 .000 FCONSUMPTION -.289 .017 .000 -.336 .024 .000 -.348 .023 .000 CO2 -.009 .001 .000 -.013 .001 .000 -.014 .001 .000 BADAPTATION .157 .033 .000 .154 .041 .000 .148 .041 .000 ASC -6.017 .178 .000 -8.175 .220 .000 -9.487 .372 .000 Standard deviations of random parameters LSPRICE . .832 .063 .000 .621 .035 .000 LSFCONSUMPTION . .150 .023 .000 .130 .019 .000 LSCO2 . .012 .002 .000 .012 .0004 .000 NSASC . 2.193 .109 .000 1.209 .080 .000 Heterogeneity in mean, Parameter Variable ASC MALE . . -.620 .114 .000 ASC AGE . . .019 .004 .000 ASC LHINC . . .199 .111 .074 ASC SUBSDY . . .161 .049 .001 Measures of goodness of fit N7,000 7,000 7,000 GROUPS 875 875 875 NB. OBS./GROUP 8 8 8 L.L. FUNCTION -6,655.621 -5,528.428 -5,284.316 K (factors number) 5 9 13 CHI SQU. [K] SIGNIFICANCE . . 4,323.715 .000 3,933.048 .000 R-SQUARED .126 .281 .271 ADJ. R-SQUARED .125 .281 .270 AIC 13,321.2 11,074.9 10,594.6 https://doi.org/10.1371/journal.pone.0193777.t003 The market for hybrid electric vehicles PLOS ONE | https://doi.org/10.1371/journal.pone.0193777 March 21, 2018 9 / 14