Real-world charging behavior and preferences of electric vehicles users in Germany

Abstract As energy consumption and greenhouse gas emissions in the transport sector have increased continuously in recent years, electric vehicles have become a potential solution to achieve environmental goals in this sector. The uptake of electric vehicle is strongly dependent on sufficient charging infrastructure. While much research has been conducted on electric vehicle charging simulations and projections, there is a research gap on real-world charging by actual users. Therefore, we conducted a survey on charging behavior and preferences in Germany in 2020 with around 4,000 electric vehicle users. The survey included a stated choice experiment and a willingness-to-pay-analysis. We found that participants’ charging behavior is dominated by home charging and that public infrastructure is perceived to be insufficient. Next to charging prices, measures of comfort through the occupancy rate of charging infrastructure and additional waiting time are driving factors when making the decision to charge. Participants specified an acceptable time for additional walking distances when parking at charging facilities between 5 and 10 minutes. High abstention rates in answering some questions indicate the need for sufficient education strategies considering existing and new charging technologies. Drivers with high annual mileage and battery electric vehicle users are willing to pay more for additional charging power. Users are open to smart charging, as long as flexibility for spontaneous trips is maintained. This research offers timely insights into real-world charging behavior and preferences, while providing recommendations for the future roll out of charging infrastructure in Germany.


Introduction
The aim of this research paper is to understand real-world charging behavior and preferences of electric vehicle (EV) users. Rather than simulating or modeling EV charging, we directly question users in order to get insights from a user-oriented perspective. The paper accounts for both battery electric vehicles (BEVs) and plug-in hybrid electric vehicles (PHEVs). BEVs only have an electric motor, whereas PHEVs have a small battery and an internal combustion engine. Including both types of vehicles and differentiating them in the analysis is critical as they have fundamentally different charging profiles. To date there is a research gap on actual charging behavior and preferences in a real-world setting. This paper and the findings fill this research gap and provide timely and relevant insights on charging behavior and preferences. As we look into charging behavior and user preferences, it is important to differentiate between relevant user groups and charging locations. The most prominent charging locations will either be home charging, meaning charging at privately owned locations (e.g., garages or private parking spaces) or public charging, meaning charging at locations such as supermarkets, hotels, restaurants etc.

Motivation
The transportation sector is critical to achieve climate change objectives. In 2020 there were 739 million tons of CO 2 -equivalent (CO 2 -eq.) emissions in Germany of which 146 came from the transportation sector, thus making up approximately 20% of total emissions (Umweltbundesamt, 2021). Within the transportation sector over 97% of emissions result from road transportation with air, rail, and ship transportation making up the remaining three percent of transportation related emissions (Umweltbundesamt, 2021). While Germany managed to reach its reduction goals for 2020 of 150 million tons CO 2 -eq. in the transport sector, this was due to the extraordinary impact of the COVID-19 pandemic. The goal for 2030 (95 million tons CO 2 -eq.) may prove difficult to achieve (Umweltbundesamt, 2021).
In response to the particular challenge of the transportation sector, the Federal Ministry of Transport and Digital Infrastructure launched the National Platform Future of Mobility in 2018 to provide concrete recommendations to achieve emission reductions (Nationale Plattform Zukunft der Mobilit€ at, 2018). The transition of the passenger vehicle fleet to EVs is key to this strategy. By 2030, between 10.5 and 11.8 million new EVs are anticipated for the passenger vehicle fleet in Germany (Nationale Plattform Zukunft der Mobilit€ at -Arbeitsgruppe 2, 2021) and in more recent scenarios even up to 14.8 million new EVs by 2030 are projected (Nationale Leitstelle Ladeinfrastruktur, 2020; Nationale Plattform Zukunft der Mobilit€ at -Arbeitsgruppe 1, 2019). Figure 1 shows the development of the EV stock and charging infrastructure density in Germany.
Despite the increase in the EV fleet, the continued success of EVs relies on users' perceptions of charging infrastructure. While the build-up of charging infrastructure is progressing, users have concerns due to the lack of range of EVs and the amount of infrastructure (Pevec et al., 2019). In order to build infrastructure efficiently and to meet demands, it is critical to understand the charging behavior, preferences, and needs of actual EV users. In the following section we provide a literature review of what is known to date about EV users' charging behavior and preferences.

Literature review of charging behavior and preferences
In this section we review the literature published within the last five years concerning charging behavior and preferences of EV users. To date, there is a research gap on actual charging behavior and preferences in a real-world setting. This paper and its findings fill this research gap and provide timely and relevant insights on actually charging behavior and preferences based on users' charging in Germany. In addition, we conduct a stated-experiment survey to determine which factors are important for given charging scenarios. Furthermore, we go beyond prior work in characterizing the perceptions of EV availability, typical charging behavior, and quantitative modeling of charging choices among actual German EV owners. This contribution is necessary for the successful build-up of charging infrastructure, which is a central challenge in transitioning to an EV mass market. Most studies are focused on BEV users and were conducted using surveys, interviews, data collection from tracking technologies such as GPS or from charging stations, simulation of charging behavior (Pagani et al., 2019;van der Kam et al., 2019;Yang et al., 2020), or with a combination of multiple methods (Hardman et al., 2018;Wang et al., 2021). Most studies focused on countries in Europe (i.e., the Netherlands, Switzerland, Germany, UK, and Ireland), the USA, or China. Current research on this topic covers a wide number of topics ranging from charging needs and behavior profiles to peak load measurements and economy of charging infrastructure with diverse results.
The current findings on charging behavior do not deliver a cohesive image. Some studies conclude that charging behavior is very individual and depends on charging infrastructure availability along routes and on destinations as well as on socio-economic and socio-demographic attributes (e.g. Ashkrof et al., 2020;Lee et al., 2020;Zhang et al., 2020). Other studies showed that there are various types of user groups which take different aspects into account when making a charging decision, such as charging point location, charging price, satisfaction, environmental aspects and more (Wang et al., 2021;Wen et al., 2016;Wolff & Madlener, 2019).
A detailed review of individual studies reveals the variety of findings on charging behavior and preferences. In a review article, Hardman et al. analyze consumer preferences of charging infrastructure (Hardman et al., 2018). The data analyzed was from questionnaire surveys, interviews, modeling, GPS data, and infrastructure data. Of particular interest, the authors found significant uncertainty in present trends and usage. They found even more uncertainty in projecting future behavior. Thus, illustrating the challenges in understanding users' preference and behavior is crucial in researching their acceptance and the sufficient ramp-up of EVs.
Identification of user types is seen to be a common framework when assessing charging behavior and preferences. Wolff and Madlener examine users' preferences for six charging attributes: place, duration, power, wait time, renewable electricity, and cost (Wolff & Madlener, 2019). From this they define five categories and their associated preferences as follows: environmentalists, EV owners, EV experts, athome chargers, and home owners. Similarly, Wen et al. identify different EV charging types of behavior (Wen et al., 2016). Using a stated preference study, they classify three charging behavior types: those dependent on price and necessity, those who charge whenever there is an opportunity, and those with multiple factors (e.g., charging power, time, costs). Finally, Wang et al. also analyze factors influencing charging and create user groups (Wang et al., 2021). The two user groups identified are service focused and pragmatic focused (accounting for multiple factors such as range, time, and costs).
In addition to literature focusing on identifying user groups, the importance of different factors on charging has been examined. Wolbertus and Van den Hoed compare fast charging usage patterns in cities and along corridors (Wolbertus & Van den Hoed, 2019). Focusing on densely populated areas in Europe, they combine the analysis of 1.4 Figure 1. Development of the electric vehicle fleet and charging infrastructure (CI) from 2011 to 2020 in Germany according to the Bundesnetzagentur 1 and Lades€ aulenverordnung 2 (Bundesnetzagentur, 2021;Kraftfahrtbundesamt, 2021a million charging sessions with survey data as well as stated choice experiment data from 100 EV drivers at fast charging stations. Findings indicate that differences in charging behavior are mainly due to technological restrictions. In another paper, Wolbertus et al. used stated and revealed preference data to quantify the effects of daytime-parking and free parking on charging behavior (Wolbertus et al., 2018b). Their results show that a daytime charging policy leads to a minor decrease in charging occupancy. Free parking is shown to increase the length of charging sessions.
According to Morrissey et al. (2016), EV users prefer to charge their vehicles at home typically during periods of the highest demand in the electrical grid. Zhang et al. also found that charging load profiles vary with demographic and social attributes (gender, age, education level) (Zhang et al., 2020). Lee et al. examined charging behavior of almost 8,000 EV owners in California (Lee et al., 2020). Factors of significance for charging behavior were socio-demographics (gender and age), vehicle attributes, commuting behavior, and ability to charge at work. Globisch et al. used a rating-based conjoint analysis for over 1,000 drivers in Germany to understand charging preferences (Globisch et al., 2019). They found that most drivers are unwilling to pay for public charging and that charging duration has a strong influence on the evaluation of public infrastructure. The authors find a weaker influence from station coverage.
Other studies have used real-world charging behavior to provide insights into charging. Helmus et al. (2020) use charging data from the Netherlands to define charging behavior by session types and user types (Helmus et al., 2020). They found thirteen charging session types and nine user types. In another paper, Sun et al. (2016) use real-world data from Japan to determine where users charge at fastcharging infrastructure and the acceptable detour for these events (Sun et al., 2016). The paper found that station choice is diverse among the users and that private users will accept a detour of 1.75 km on working days. Furthermore, also for public charging points, the occupation time of a charging point is far longer than the charging process itself (Wolbertus et al., 2018a). To prevent issues in the power grid, it is necessary to shift charging processes with long occupation times to time slots with lower grid loads with the help of smart charging processes. It was found that the majority of EV users is willing to participate in smart-charging for low monthly savings (Rodriguez Jimenez, 2019). In another study, Ashkrof et al. collected data from over 500 EV users in the Netherlands and found that BEV drivers' travel behavior relies on route, vehicle, charging, and socio-economic attributes (Ashkrof et al., 2020).
Consequently, significant research has been conducted to date on EV charging. However, as the literature review reveals a cohesive and consistent image of charging behavior is not visible. This is combined with reliance upon limited data, modeling, simulations, and divergent research questions that are not focused on real-world charging behavior and preferences. In order to address this research gap, we conducted a large-scale, online survey of real-world EV users to identify charging behavior and preferences.
In the following section, we outline the methodology used to conduct the research, the survey type used, sample size, research questions, and general details about the study. Then we present the results of the survey and discuss the findings and their significance. Finally, we provide recommendations and insights from the research. Limitations of the study and areas for future work are outlined afterwards.

Methodology
As outlined above, the research aim of this study is to understand EV users' charging behavior and their perception of current charging infrastructure. In order to capture real-world charging behavior of EV users in Germany, an empirical approach with a quantitative research method was chosen. A two-part survey was developed consisting of a general survey and a stated choice survey. The results of this survey enable a detailed analysis of acceptance, preferences, and opportunities to improve the charging of EVs. These form the basis upon which recommendations for the demand-based design of charging infrastructure in the future can be derived.

Survey design
The survey is designed as an online questionnaire and hosted on the platform Survey Engine. A desired sample of 12,000 private BEV and PHEV users was conceptualized.
The following attributes and their respective quotas for the desired sample were considered: The quotas for the vehicle classes were derived from the general vehicle fleet information according to the KBA's central vehicle registration pool in 2019, which was the most recent database for this matter at that time. For the other two attributes, equal distributions between BEV and PHEV, as well as older (more than two years) and newer (less than or equal to two years) EVs were assumed. The exact distribution of the desired sample is presented in Table 1.
The Institute of Transport Research commissioned the German KBA to draw this desired sample from their central vehicle registration pool. Physical letters were then sent to the respective EV owners according to the desired sample by the KBA, which contained an explanation of the research study aim, privacy regulations, and a URL and QR code to access the online survey. Since the letters were sent to private residences by the KBA, the researchers did not have access to the users' personal information (i.e., name, age, address, state). The authors only had access to the results of the online survey. The survey began in September 2020 and all final results were analyzed in March 2021. The majority of respondents completed the survey within a few weeks of receiving the letter.
The exact survey questions are provided in the results section. Detailed information on the distribution of EV ownership by manufacturer or vehicle model, as well as the usage behavior with such vehicles for the investigated sample is featured in the supplementary material (see Figure  10). The research question for the stated preference section is listed in Sec. 2.3.1.
In choosing a sample size of 12,000 EV users, a return rate of 5% (i.e., 600 respondents) was anticipated. An actual return rate of 44% (5,233 respondents) was achieved. This very high response rate may have been due to the COVID-19 pandemic, which resulted in a national lockdown in Germany with people remaining at home over several months.
Before evaluation, data pre-processing was performed to clean the dataset and filter out faulty iterations that contain inconclusive and contradicting answers or were not answered completely. Incomplete iterations refer to systemrelated errors due to not completing the survey or missing entries (e. g. no declaration of vehicle type as BEV or PHEV). Inconclusive responses refer to faulty declarations of BEV as PHEV to the participants' owned EV (e. g. BMW i3 as PHEV or BMW 2 Series as BEV). After data pre-processing, the dataset contained 4,172 iterations with 2,641 BEV users and 1,531 PHEV users, accounting for a final return rate of about 35%.

General survey
The general survey consisted of 36 questions in total (including a privacy notice question) and were a mix of open-ended and close-ended questions, as found in the supplementary material. The questions can be categorized into three groups: (1) general information about their EV, (2) charging and usage behavior, and (3) socio-demographic, which were asked after the stated preference survey. For evaluation of the results, a descriptive analysis approach by examining percentage frequency distributions for each question was conducted (Babbie, 2016).
To ensure that the actual sample represents the desired sample, the actual sample has to be weighted according to the aforementioned desired sample attributes. Therefore, weighting factors w i (the ratio between the distribution of the actual and model subsamples) are introduced. Hence, a compensation of an underrepresented group is achieved for values over 1 (w i > 1), while overrepresented groups are corrected with a weighting factor value under 1 (w i < 1). An ideal representation would hence result in a weighting factor of 1 (w i ¼ 1). Table 2 lists the weighting factors for the different subsamples.
In the actual sample, EV users of newer vehicles are underrepresented, while there is an overrepresentation of EV users with older vehicles. The underrepresentation is especially visible for PHEV users with newer medium-sized EVs, which show a weighting factor of w i % 9.6. At the same time, the overrepresentation reaches a significant value due to w i % 0.4 for BEV users with older vehicles.

Stated choice experiment
In addition to the general survey, a stated choice survey was conducted. We will first present a short overview of the terms used. Afterwards the structure and content for this part of the survey is presented. Finally, the theoretical background for the analysis of this part is explained. According to J. de Dios Ort uzar and L. G. Willumsen (2011), stated preference describes a collection of different methods in transport research. The most common methods are contingent valuation, conjoint analysis, and stated choice techniques (de Dios Ort uzar & Willumsen, 2011). In this analysis we created a stated choice experiment as part of the survey. To evaluate the data of the stated choice, experiment a discrete choice analysis was conducted. In the following section assumptions and the procedure of the stated choice experiment are described. This is followed by an explanation of the theoretical background used. In 3.2 an evaluation of the results can be found.

Procedure and assumptions
In the stated choice experiment, participants were given a choice between four different charging options out of five possible charging locations and their characteristics (Table  3). Each charging option was based on a different charging location. An example can be found in Figure 2.
In order to show four of five possible alternatives, the choice set was created by using the Modified Federov Algorithm (ChoiceMetrics, 2018). A differentiation was made between the charging locations: home, work, shopping, leisure and fast. Each charging location also differs in the characteristics availability (indicates how often the charging infrastructure is available, i.e., not occupied), power (charging power in kilowatts (kW)), cost (cost per kilowatt hour in Euros (kWh in e)), eco (indicates whether the charged electricity is 100% from renewable source or not), and waiting time (additional waiting time before the charging process (only for fast charging)). Each participant was asked to think about the following situation and select the option which maximizes their personal utility. The following question was asked: Imagine you need to charge your vehicle during the course of the day. For this purpose, all the alternatives for charging given below are available to you. The information on availability, charging power, costs, green electricity, and waiting time differs depending on the location of the charging station.
The stated choice experiment consisted of eight different choice situations in a random order, which were selected randomly from three different choice sets. In each of the three choice sets, all five charging locations were available. In each choice situation, four of the five charging locations were available for choice. The different attributes and their levels (Table 3) were determined by knowledge gained from an expert workshop held on June 2020 and an exchange with researchers from the DLR Institute of Vehicle Concepts (German Aerospace Center, 2021). To further specify the model, other parameters from the base survey were analyzed and were included if they could improve the model.
In order to improve the data quality a Bayesian efficient design was created with the software Ngene (ChoiceMetrics, 2018). The priors gained from an internal pretest with four participants before the field test helped to adjust the choice set and achieve an efficient design. Despite the small sample size of the pretest, the relevant priors were significant and the algebraic sign was as expected. According to the literature a more efficient design can be generated even if it is based on few pieces of information (Bliemer & Rose, 2005).

Theoretical background
The most common approach to address discrete choices in transportation is based on the random utility theory, which assumes that each individual (n) acts rationally and chooses the alternative (i) that maximizes the individual's personal overall utility (U n,i ) (Homo Oeconomicus) (de Dios Ort uzar & Willumsen, 2011;McFadden, 1973).
All model estimations were conducted with the software Python 3.7 and the extension PandasBiogeme (version 3.2.6) (Bierlaire, 2020;Python Software Foundation, 2020). For the model estimation a multinomial logistic regression approach was chosen and only fully completed surveys were considered. Income and gender were not considered in modeling due to their unrepresentative distribution in the sample compared to the German population (see 3.1.1).
An iterative approach was chosen to obtain the final model. First, a stable model was developed for all participants. In the next step, it was analyzed which findings from the basic survey could be used to provide additional explanatory information and whether these were related to the attributes analyzed. Thereafter, the sample was split into BEV and PHEV users to identify possible differences in decision making. To generate robust results, it was necessary to slightly adjust the model for the different groups.

Analysis of general survey
In total, 30 questions were evaluated to derive insights into the real-world charging behavior of German EV users. These can be clustered to obtain the perception of the charging infrastructure and the EV charging behavior.
3.1.1. Sociodemographic and EV distribution of the sample First, the results in the sociodemographic and EV distribution of the sample are presented. The designated multiple choice questions assess the participants' gender, age, education level, current profession, as well as the living situation (household size, housing type, and monthly net income of household). The corresponding diagrams are shown in Figure S1 in the supplementary material. As a result, the average German EV user of this sample can be identified as a highly educated (e.g., university degree), male, 50 years or older, and who is currently either employed full-time or retired. He lives in a two-person household, in a detached house, and has a monthly net income of at least 3,000 e, while the slight majority of participants declared a monthly net income of 7,000 e. This is rather high in comparison to the average net household income in Germany with 3,580 e in 2019 (Statistisches Bundesamt (Destatis), 2020).
Detailed information on the distribution of EV ownership by manufacturer or vehicle model as well as the usage behavior with such vehicles for the investigated sample is featured in Figure S2 in the supplementary material.

Perception of charging infrastructure
Four questions were assessed for insights into the perception of charging infrastructure by EV users. For this, no differentiation between AC and DC charging was considered. First, EV users were asked to specify the frequency at which charging stations are found at eight different locations in everyday life. These locations include the EV users' home (e.g., privately-owned parking spaces, garages), spaces in close proximity of the EV users' home (e.g., at the roadside), at places of work or training, at places for shopping (e.g., supermarkets, shopping centers), at leisure facilities (e.g., gyms, sport fields, amusement parks, campsites, museums, swimming facilities), at public street spaces within city areas (not the place of residence), along freeways or federal highways, and other locations. Here 'other locations' is mainly specified to be charging stations at hotels or restaurants (20%) although the majority of EV users (36%) did not specify another location or were satisfied with the given options.
As seen in Figure 3, more than 80% of EV users, regardless of the drive technology, find a charging station at home almost every time. In contrast, 40 to 50% of EV users state they never find charging stations at public locations like spaces close to users' residences, at work, or at leisure facilities. During shopping, around one third of all EV users locate charging stations either sometimes or rarely. Almost the same share of BEV users sometimes finds charging stations at public city spaces, along highways or at another location compared to PHEV users. The answers of PHEV users are rather evenly distributed over all answer possibilities. Comparing the answers differentiated by drive technology, BEV users tend to perceive the occurrence of charging stations at public locations more often than PHEV users.
Participants were also asked to specify the available maximum charging power at the described locations, regardless of differences between AC and DC charging. The given charging powers range from 3.7 kW for charging via commercial power outlet to 150 kW for fast charging technologies. As seen in Figure 4, a large share of participants did not answer this question for public locations, especially PHEV user for locations close to the EV users' residences, at work, as well as at leisure facilities (59 to 86%). Despite that, charging powers of 3.7 kW is mainly available at home as stated by 33% of all BEV users and 51% of all PHEV users. The charging power of 7.4 kW has only small shares for both drive technologies at any location, with around 5%. A charging power of 11.1 kW is only mainly noticeable for home charging for both drive technologies, while 22 kW as a maximum charging power is more perceived in public places, especially by BEV users. Only BEV users stated the availability of 50 kW charging power at public city spaces (7%), at shopping facilities, as well as along the highways (20%). Same goes for the availability of fast charging facilities with 150 kW power with one third of all BEV users, namely along highways.
Furthermore, participants were asked to rate how often they accept additional walking when trying to find a parking spot with charging possibilities. Here a five-level scale starting with 'almost every time' to 'never' was used. Figure 5 reveals that while about one third of all BEV users never have to accept additional walking distances, the remaining two thirds deal with additional walking distances for finding parking spots with charging infrastructure in their daily life. One tenth accepts these inconveniences almost every time, while respectively 20% of all BEV users either accepts this often, sometimes, or rarely. Most BEV users quantified this additional walking distance with a maximum time of 0 to 10 min (61%) and an average of 8.8 minutes. Four out of ten PHEV users never accept additional walking distances so that this inconvenience is accepted by 56% of all PHEV users. While only 7% of PHEV users stated to do so almost every time, a total of 49% of this  subsample accepted this either often (16%), sometimes (13%), or rarely (20%). About 51% of all PHEV users also specified the additional walking distance with a maximum time ranging from 0 to 10 minutes, averaging to about 7.4 minutes. As it is noticeable, PHEV users accept less additional walking distances when parking at locations with charging facilities than BEV users which corresponds to the PHEV users' tendency in finding charging facilities at mainly privately-owned parking spaces.
These results show that public charging infrastructure can be perceived as insufficient in daily life due to the lack of accessibility, particularly at places close to EV users' residence, at work, or leisure facilities. At the same time, EV users usually charge at private properties like home or elsewhere. This is further underlined by the additional walking distances for parking at locations with charging facilities as it was reported by most EV users, which averages to around 8 minutes. Also, a lack of information on available infrastructure is seen, coming from the high abstention rates when asked about available charging powers. BEV users show a slightly higher awareness on available charging powers in public, while also accepting more additional walking distances compared to PHEV users.

EV charging behavior
In connection with the before-assessed locations, the frequency of charging at said locations was examined as well, ranging from (almost) daily to never charging. At home, more than two thirds of all PHEV users charge their EV (almost) daily, while the majority of BEV users charge 1 to 3 days per week (42%) (see Figure 6). Daily home charging is initiated by about 33% of all BEV users. At places close to the EV users' residence, at work, or at leisure facilities, up to one half of all EV users did not specify their charging frequency in daily life (37 to 54%). In addition, these shares of abstention are lower at shopping opportunities, public city spaces, along highways, and other locations, comprising about 10% for BEV users and 20% for PHEV users. At these locations, the majority of EV user charges less than monthly or never. A similar trend is seen at places close to the users' residences, at work, or at leisure facilities for both drive technologies.
Considering the EV users' homes as the most used charging and parking locations, the following question assesses the charging behavior upon arrival. EV users were asked if they would initiate a charging process whenever they arrive at home (see Figure 7). As per the results, PHEV users would start charging their EV upon arrival at home almost every time with a share of 76%; most BEV users (58%) decline this question. A plausible cause can be ascribed to the difference in battery size, which is smaller for PHEVs than for BEVs. In addition, out of all PHEV and BEV users not charging their vehicle almost every time upon arrival, the majority tend to charge their EV at a remaining range of about 0 to 100 km, comprising 27% for BEV and 100% for PHEV users. It should be noted that most of featured PHEVs only have a maximum electric range of up to 100 km. Because the battery size and the range of BEVs varies from vehicle to vehicle, the remaining range is a better indicator compared to the state of charge.
Additionally, the survey assessed the participants' willingness to charge when higher charging power was available. For this, participants were asked to decide whether they consider charging their EV at charging power of 11.1, 50 or 150 kW for parking durations ranging from under 15 minutes to 60 minutes and longer. The results (Figure 8) reveal that about every second BEV user would charge with 150 kW charging power if the parking duration would be less than 15 minutes, as it is with more than one third of all PHEV users. As expected, 40% of all BEV users as well as 29% of all PHEV users consider a parking duration of 60 minutes and longer acceptable when charging with a power of 11.1 kW. Also, charging power of 50 kW is largely expected to result in a parking duration of 15 to 30 minutes for BEV users (37%) as well as PHEV users (25%). At the same time, the abstention in this question rises with higher charging power with about one fifth of all BEV and more than a third of all PHEV users for 150 kW. This is especially apparent for PHEV users that show higher abstention rates than BEV users in comparison. It should be noted that PHEV cannot be charged with such high charging powers, so that this question was meant to evaluate the behavior tendency with higher available charging powers. An elaborated investigation of the remaining range before charging considering the locations examined in Sec. 3.1 is presented in the Figure S3 in the supplementary material.
Considering alternative operation strategies for charging infrastructure, participants were asked if they would charge their EV 'smartly'. This topic was introduced to the participants with an additional explanation paragraph before the actual question stating the following: smart charging refers to an alternative charging strategy, in which charging processes are optimally controlled by grid operators for distributing voltage peaks evenly and more beneficially for the electricity system. Thereby, vehicle parking times can be higher than charging times. Additionally, they were asked to state at which minimum range a smart charging process can commence. The results (Figure 9) show the general approval of smart charging among EV users with a share two thirds, while about one third of BEV as well as PHEV users would consider its use if it is cost-effective. Due to the limited battery capacity and electric range of PHEV, the minimum ranges for smart charging were only evaluated for the different BEV classes.
Looking at Figure 9, most BEV users regardless of vehicle class would prefer minimum ranges between 100 and 150 km before smart charging starts. Also, the wish for higher minimum ranges increases with the vehicle class, so that every second user of a large-sized BEV would choose to charge smartly at a minimum range of 150 km. A similar behavior is seen with users of medium-sized BEV where the majority, comprising of 35%, starts charging smartly at the same range. In comparison, the majority of users of smallsized BEV with around one third prefers smart charging after reaching a minimum range of 100 km. This preference could be linked to the larger battery size in larger BEVs. While being open to smart charging, BEV users, in particular those owning larger-sized BEV, display a wish in flexibility for spontaneous trips, especially when compared to the  average daily trip distance of 39 km in Germany as of 2017 (Nobis & Kuhnimhof, 2018).
In connection to the stated choice survey, participants had to specify which of the charging features was considered important when making decisions for charging. For this, the features of charging location, occupancy of the charging infrastructure, charging power, cost per kWh, sustainability of the charged electricity (100% green electricity), and the additional waiting time at the charging infrastructure were examined. Participants were able to choose multiple answers. As seen in Figure 10, EV users show consideration for all given features with shares of at least 58% regardless of the drive technology. Aspects of comfort through the additional waiting time and the occupancy of the charging infrastructure as well as the cost per kWh were ranked as the three most important features when making decisions with 88 to 91% for either drive technology. While about 75% of BEV users see the remaining features (sustainability of electricity, charging location, and charging power) as important, charging power was considered less important for PHEVs as seen in the share of 58%.
Connected to the perception of charging infrastructure, the results indicate an inhibited charging behavior, particularly in public locations. Here, BEV and PHEV users tend to  never or only monthly use charging facilities at public locations. Regular and daily charging only occurs at home, especially for PHEV users who charge almost every time upon arrival. The majority of BEV users charge 1 to 3 days per week at home. Next to the EV users' logical expectation of shorter parking durations with increasing charging power, the results reveal an increasing uncertainty in parking durations with higher charging powers due to a potential lack of information considering the effects of high charging powers on their EVs, as well as the wide range of energy providers with diverse service systems. Considering smart charging, BEV users show general approval to this charging strategy, while preferring higher remaining ranges beginning at 150 km for larger BEV, after which smart charging can start. For making the decision whether to charge or not, EV users especially refer to comfort, measured by the occupancy of charging infrastructure and additional waiting time, and the charging cost as driving factors. PHEV users tend to put more emphasize on the charging location rather than the charging power in comparison to BEV users.

Analysis of the stated choice experiment
In this chapter, the results of the stated choice experiment are presented and analyzed. In addition to the outcome of the different models, an analysis of the willingness-to-pay (WTP) behavior is conducted. In Table 4 the results of the final MNL models are presented. For the three models, the estimated coefficients (b), the t-test, and the model fit are provided. Overall the estimated coefficients show the expected signs and plausible values.
To obtain a robust model, some coefficients of availability (avail) and green electricity (eco) had to be summarized for certain charging locations. For the public charging locations shopping and leisure and often also fast charging no significant difference was found for the coefficient availability and green electricity. It seems to be very important for the participants to find an available charging place at home (b avail home ). This supports the finding from above that participants mostly charge at home. For BEV users, it is also important to find an available spot at a fast charging station when needed (b avail fast ). For PHEV users, this is not relevant. Green electricity tends to be more important to BEV users than for PHEV users. Especially at home (b eco home ) and at public charging places (b eco shoppingþleisureþfast ) green electricity seems to be more important for EV drivers than at work (b eco work ). Price and charging power were estimated as generic coefficients as there were no significant differences between the considered charging locations. However, significant differences in the evaluation of charging power could be observed for BEV users. First, the charging power was evaluated differently by users with vehicles of higher range than by users with vehicles of lower range (b power low_range ). Second, owners of higherrange vehicles made a difference in their evaluation of normal (b power w/o_fast_high_range ) and fast charging locations (b power fas-t_high_range ). For PHEV users, no significant difference in the evaluation of charging power could be found (b power ). This may be caused by the fact that all PHEV in Germany offer nearly the same range (Patrick Pl€ otz et al., 2021) and charging power is not important for PHEV users (Figure 10).
The analysis of the charging price showed a significant correlation with the annual mileage. Price plays a more important role for persons with high annual mileage (b price high_mileage ) than for persons with low annual mileage (b price low_mileage ). BEV users in particular react much more sensitively here. Thereby, the difference between the two groups is twice as high for BEV users as it is for PHEV users. This can be explained by the fact that a higher price per kWh has a greater overall effect due to the battery size. As shown in Table 4, charging power plays an important role for BEV users. Therefore, Figure 11 is only focused on this user group. It can be seen that charging power becomes much more important for BEV users with higher ranges. Contrary to the assumption that charging power is more important for fast charging, it turns out that the coefficient for charging power is in fact lower for fast charging. This is caused by the significantly higher power of fast charging processes (Table 3). Charging power is important for fast charging processes, but the coefficient appears smaller due to the strong increase in charging power compared to normal charging. Compared to the normalized differences in charging power, the normalized differences in charging price depending on the annual mileage are much smaller.
The results of the models from Table 4 can be used to evaluate and compare the WTP for additional charging power. In Figure 12, the results of this analysis for users of PHEVs (in yellow) and BEVs (in blue) are presented. The willingness to pay for additional charging power is plotted on the X-axis. The unit to measure this is the additional price per kWh of charged energy that a vehicle user is willing to pay more for each kW of additional charging power on top of a fully charged battery. The figure shows that BEV users are willing to pay more for additional charging power. Furthermore, the WTP decreases with higher annual mileage (light vs. dark). In the case of BEVs, it is also noticeable that users of EV with a higher range are willing to pay more for additional charging power. It can be assumed that a higher range is essentially made possible by a higher battery capacity. However, the longer charging time required due to a larger battery capacity can be compensated by a higher charging power. To save time, public charging options with higher charging power become particularly interesting for users of vehicles with a longer range. As can be seen in Figure 12, there is a difference between normal and fast charging for these EV users caused by the reasons explained above.

Discussion
Looking at previous studies on early adopters of EVs in Germany from 2015, we asked similar questions in order to compare the development of EV ownership and usage in Germany. The results were found to be similar. As shown in Trommer et al. (2015), we also identified EV users as a rather homogenous group, which is mostly male, highly educated, and with a high income. The presented results only apply to private EV owners of the generated sample by KBA. Since the final respondents were from users reached via KBA's physical letters, we were not able to entirely control the final sample make-up. Regarding the observed increase of EV sales as shown in Figure  1, one can expect higher shares for these user groups in possible future surveys as EVs become a mass market. Despite the  high return rate with 4,172 verifiable iterations, the sample also featured a considerably small amount of PHEV users, as well as users of newer EV so that representativeness of the answers coming from these user groups should be viewed critically even with weighting. This is particularly true for the group of PHEV users with newer smaller-or medium-sized EVs. As the respective subsamples with 66 and 25 respondents are relatively small, the high weighing factors could increase the errors of a small sample size.
As of May 2021, new EV registrations Germany accounted for over 23% (over 11% for both BEVs and PHEVs) of all vehicle registrations in this month (Kraftfahrtbundesamt, 2021h), so that EV adoption has reached the early majority stage already (Rogers, 1971). Rogers lists innovators as the first 2.5% of users, followed by early adopters (the following 13.5% of users). While the survey was conducted at the end of 2020 and the respondents had purchased their vehicles prior to this, the findings are extremely relevant for the early majority. The challenges and situation faced by our respondents (e.g., access to public infrastructure, public subsides, electricity prices) will be faced by the early majority as well. Thus, the insights from this work are of critical and time importance as the EV market grows to a mass market.
On the topic of home charging, our finding about the dominance of home charging as well as the lack of charging in public also corresponds to the study by Trommer et al. (2015) and Morrissey et al. (Morrissey et al., 2016). A similarly low availability of public charging infrastructure in Germany was published by LeasePlan with around 0.53 public chargers per 1,000 inhabitants (LeasePlan, 2021). This value is around six times lower than in the Netherlands (3.54) or Norway (3.4). The same goes for fast charging facilities relative to the highway availability which is, with a value of 52 locations per 100 km, four times and 15 times smaller than in the UK or Norway (LeasePlan, 2021). As a counter measure, the new German law for fast charging promises the nationwide construction of 1,000 fast charging stations by 2023 so that research on this topic should be conducted in the near future to account for this build-up (Bundesministerium f€ ur Verkehr und digitale Infrastruktur, 2021). A successful build-up of fast charging infrastructure might change the perception of charging infrastructure. Due to the high usage of home charging in daily life, funding or other incentives supporting the acquisition of private charging infrastructure by EV users could help in realizing a rapid transformation. While most participants prefer home charging where parking time is usually higher than charging time, we also found that smart charging is widely accepted, especially when charging prices are lower as found by Rodriguez Jimenez (Rodriguez Jimenez, 2019). This acceptance is very important for grid operators who need the additional flexibility to increase the amount of renewable energy in the grid. To take advantage of this high acceptance, it is important to make sure connected vehicles have enough range for spontaneous trips.
Other factors also inhibiting the frequent usage of public charging infrastructure need to be addressed. The identified high abstention rates from several survey questions show a potential lack of knowledge about public charging infrastructure, which would be in accord with EV users only charging at home. Considering the age of the representative EV user, attention should be directed to simplifying the charging process when it comes to ease-of-use and the payment process. In-depth analysis on these topics could form the basis for recommendations for energy providers and others operating public charging infrastructure. It would also help in identifying where EV users lack in knowledge on new charging technologies.
Analysis of the stated choice data showed that decision making differs from BEV to PHEV users. The availability of electricity from renewable sources, charging power, and charging price play a larger role for BEV users. The interaction of charging power and charging price could be observed by a willingness-to-pay analysis. It is clear that BEV users and less frequent drivers are willing to pay more for additional charging power, while a BEV user's willingness to pay increases with the range of their EVs.

Conclusion
In this research, the quantitative method of conducting a survey was chosen for acquiring insights into the charging behavior, preferences, and perception of German EV users toward current charging infrastructure. This two-part survey, consisting of a general survey and a stated choice experiment, was conducted over a two-month period.
The results showed the dominance of at home charging. Public charging infrastructure was viewed to be insufficient. PHEV users in particular tend to charge every time upon arriving at home, while BEV users more strongly perceive the wide range of charging infrastructure and wish for more flexibility when making spontaneous trips. Comfort in using charging infrastructure, through the aspects of occupancy as well as additional waiting time before charging, plays an important role next to charging prices when a decision for charging is made. Compared to PHEV users, BEV users with high mileage are more sensitive to charging prices as well as available charging power whereas the latter factor exhibits a dependency on the remaining BEV range. At fast charging locations, charging power has been identified to be less important of a factor than at all other locations. Additionally, higher abstention rates were visible with higher charging powers, indicating a certain lack of knowledge with this charging technology.
The scope of this research is limited to private EV owners in Germany. Charging behavior and preferences of commercial vehicle owners and international users were not covered in this study. Additionally, this research was conducted during the COVID-19 pandemic which affected the transportation sector on a great scale. Due to its impact on the perception of individual mobility, it has a considerable potential of distorting the perception of EV users on the charging infrastructure (Eisenmann et al., 2021). A postpandemic follow-up survey can therefore provide further insights on this topic.
Upcoming research topics that arise from the acquired results include further investigations on the real-world utility factor of PHEVs and in-depth looks into other reasons for the inhibited charging behavior at public charging facilities (e.g., the examination of the ease-of-use in daily life or potential problems during payment). One topic that became apparent during the survey period was the relationship between owning a photovoltaic energy system at home and an EV. Future research should examine the connection between the usage and perception of renewable electricity at home and charging behavior. Other potential research topics include charging behavior at work and the role of vehicleto-grid charging in context to the EV user.