Submission date:????, 0 Word count words + Tables and Figures 0 0 0 Awareness, Experience, and Potential Choice of Carsharing: Comparative Analysis in Four Cities of Japan Hironori Kato Department of Civil Engineering, University of Tokyo --, Hongo, Bunkyo-ku, Tokyo -, Japan Phone: +---; Fax: +--- E-mail: kato@civil.t.u-tokyo.ac.jp Akihiro Inagi Financial & New Business Unit, Mitsui & Co., Ltd. --, Otemachi, Chiyoda-ku, Tokyo -, Japan Phone: +---0; Fax: +---0 E-mail: A.Inagi@mitsui.com Takahiro Igo Department of Civil Engineering, University of Tokyo --, Hongo, Bunkyo-ku, Tokyo -, Japan Phone: +---; Fax: +--- E-mail:igo-t@ip.civil.t.u-tokyo.ac.jp
0 Abstract. This paper reports the results of the local survey on the carsharing in the four cities of Japan including the people s awareness, experience, expected action regarding the car ownership, and stated choice of carsharing membership. Then, the modal choice and the carsharing membership choice are empirically analyzed. The survey results show that the car owners are more aware of the carsharing than the non-car owners but the car owners have less experience to consider the carsharing-use than the non-car owners; the individuals in the senior subgroups have more positive impressions of carsharing than those in the younger subgroups; 0-0 percent of individuals could stop the car ownership if they would be a member of carsharing service; the individuals seem to make a rational decision on the membership choice under the different hypothetical cases; and the availability of carsharing service and public transportation, trip distance, and household income influences the choice of carsharing membership. Keywords. Carsharing, awareness, experience, modal choice, carsharing membership choice, comparative analysis
0 0 0 INTRODUCTION Carsharing market has been growing in many countries (, ). It is the case in Japan (). Many carsharing services have been provided mainly in urban areas while the number of carsharing members has been increasing rapidly. The carsharing members in Japan were, in 00,, in 00,, in 00,, in 00, and, in 0(). However, it is still unpopular transportation service in Japan. The market potential and the individual s awareness of carsharing have not been well understood by transportation planners. Then although the carsharing could be included in transportation planning, it has not been explicitly taken into an account. For example, the location planning of carsharing station is one of the most critical issues for the business of carshraing operators (, ) while it also influences the public/private transportation demand significantly. However, the carharing market data is not sufficient enough to support their location planning as well as to examine their potential impacts on the transportation demand. Our study team implemented the paper-based questionnaire surveys to understand the awareness, experience, and preference of carsharing service in four cities of Japan. The cities are two cities in Tokyo: Meidai-mae and Hikari-gaoka, a sub-urban city in Kanagawa: Fujisawa, and a prefectural capital in Tochigi: Utsunomiya. This paper aims to report the findings from the comparative analysis and to discuss the implications to transportation policy and carsharing business in Japan. To the best of our knowledge, no study has reported the comparisons of individual s awareness and/or preference on carsharing among multiple cities of Japan. This paper is organized as follows: the background and goal of this paper are presented in the first section. Next the survey method and descriptive statistics of the respondents are presented. The survey results are shown and the findings are discussed. Then the modal choice including the carsharing and the carsharing membership choice are empirically estimated with the collected data. Finally, the paper is concluded with the discussion of implications from the survey results and the further research issues. DATA COLLECTION Survey Areas The study team including the authors made the local survey on local people s daily travel behavior and preference on the choice of hypothetical carsharing service. The paper-based questionnaire surveys were carried out in the four cities in Japan: Meidai-mae, Hikari-gaoka, Fujisawa, and Utsunomiya. The scope of survey area is equal to the circle with 00 meter radius on average of a specific point in each city. TABLE summarizes the characteristics of survey areas. TABLE Characteristics of Survey Areas Survey area Prefecture Population density Public transit availability Land-use characteristics Meidai-mae Tokyo Very high (over,000 persons/km ) Hikari-gaoka Tokyo Very high (nearly,00 persons/km ) Fujisawa Kanagawa Middle (nearly,00 persons/km ) Utsunomiya Tochigi Lower-Middle (nearly,00 persons/km ) Two urban rail services (average distance to the nearest rail station is about 00 m) Bus service Some carsharing services Metro service (average distance to the nearest rail station is about 0 m) Bus service Some carsharing services About.0 km apart from the nearest rail station Poor bus service No carsharing service About. km apart from the nearest rail station Poor bus service No carsharing service High population-density commercial/residential district Private university campus is located in the area 0 km apart from the CBD in Tokyo High-story (over 0 floors) apartment district developed by public corporation km apart from the CBD in Tokyo Typical sub-urban residential district in the Tokyo Metropolitan Area km apart from the CBD in Tokyo Typical car-oriented local city in Japan Residential area National university campus is located next to the area
0 0 0 0 0 Survey Method The questionnaire sheets in the survey request the respondents to answer the four types of questions. The first is the respondent s daily travel episode on a typical week day and a weekend day including the origin and destination, departure time from home and arrival time to destination, travel purpose, travel mode used, travel time for all trips made. The second is the respondent s stated choices under three hypothetical cases in which different carsharing membership fees, carsharing-use time-based charges, carsharing-use distant-based charges, and accessibility from their home to the nearest carsharing station are presented. The number of levels for each attribute is two or three. The seven types of hypothetical choice cases are prepared in advance on the basis the fractional factorial design. Three cases are presented to the respondents in the survey sheet: the one of them is a base case which is the same for all respondents and the two other cases are randomly selected from the six cases. The respondents are requested to show their choice of participating in carsharing membership and the choice of travel mode for each observed trip made by them when they would participate in the carsharing membership under each hypothetical case. The third is the attributes of respondent s household including personal information such as age, gender, job, marriage status, driving license ownership of all household members, household income; types of car owned by the household and the history of car ownership; the car parking space of the household; the addresses of home; the physical structural type of house; the distance from the home to the nearest bus stop. The final is the respondent s preferences and opinions on the carsharing use, the type of automobiles, and electric vehicle use. The potential respondents are selected randomly from the detail maps of survey areas. The surveyors visit the potential respondents and request them to participate in the survey. When the potential respondent is out of home, the surveyors put the survey sheets with a request letter in its postal mail box. In a case of an apartment, a notice of survey is presented in the apartment in advance of the survey under the support from the apartment managers and then the surveyors visit them. The average time to answer the sheet is about thirty minutes.,000 Japanese Yen (nearly equal to US dollars) are given to all respondents for their contribution. The surveyors visit the respondents twice. The respondents can choose a way to answer from the following options: () they answer the questionnaire sheets immediately during the first visit of surveyors and return the answer sheets to the first surveyors, () they fill in the sheets after the first visit of surveyors and return them during the second visit of surveyors, and () they send their answer sheets by postal mail. The surveys were implemented on February to Feburary, 00 in Meidai-mae; on Feburary to March, 00 in Hikari-gaoka; on March to March, 00 in Fujisawa; and on April to April, 00 in Utsunomiya. 0 potential respondents were selected and the data of 0 individuals were collected in Meidai-mae; 0 potential respondents were selected and the data of individuals were collected in Hikarigaoka; 0 potential respondents were selected and the data of individuals were collected in Fujisawa; and 0 potential respondents were selected and the data of individuals were collected in Utsunomiya. Descriptive Statistics TABLE shows the descriptive statistics of respondents in the four cities. Age is defined as if a respondent is male and 0 otherwise; car ownership is defined as the number of cars owned by the household that the respondent belongs to, transfer is defined as the number of changing the transportation mode from one to another in a trip, and access/egress time is defined as the sum of the travel time from an origin to one public transit station/stop and the travel time from another public transit station/stop to a destination in a trip. First, the average age of respondents is the youngest in Utsunomiya, followed by Hikari-gaoka, Fujisawa, and Meidai-mae. High rate of young respondents in Utsunomiya reflects that the survey area covers the university area. Although Meidai-mae also covers the university area, the average age is the oldest among cities because the survey area includes the apartments for high-income workers. Second, the average gender varies from 0. to 0.. Male respondents are dominant in Hikari-gaoka because many single senior male households reside in the apartment area. Third, the annual household income in Hikari-gaoka is the highest, followed by Meidai-mae, Fujisawa, and Utsunomiya. The reason for higher household income in Hikari-gaoka and Meidai-mae is that they are located in Tokyo, the highest income region in Japan. It should be noted that the standard deviation of household income is higher in Meidai-mae than that in Hikari-gaoka. This is mainly because the survey area in Meidaimae includes both low-income students and high-income workers. The annual household income is the lowest in Utsunomiya because many respondents are the university students. Fourth, the car ownership is the highest in Utsunomiya, followed by Fujisawa, Hikari-gaoka, and Meida-mae. This may reflect that the public transportation network is poor in Utsunomiya and Fujisawa whereas the urban rail network is well-organized in the other two cities in Tokyo. Fifth, the average travel distance is. km in Meidai-mae,.0 km in Hikari-gaoka,. km in Fujisawa, and. km in Utsunomiya, respectively. The travel distance in Hikari-gaoka is the longest probably because many workers commute to the central business district which is located far from their home. Sixth, the average travel time is. minutes in Meidai-mae;. minutes in Hikari-gaoka;. minutes in Fujisawa; and. minutes in Utsunomiya, respectively. This means that the
0 average travel speeds are. kph in Meidai-mae,. kph in Hikari-gaoka,. kph in Fujisawa, and.0 kph in Utsunomiya, respectively. The average travel speeds in Fujisawa and Utsunomiya are higher than those in the two cities of Tokyo. This is because the individuals in Fujisawa and Utsunomiya use mainly automobile for travel whereas those in Meida-mae and Hikari-gaoka use mainly rail for travel. The rail-use travel time tend to be long because it includes the access/egress travel time and waiting time at stations in addition to the rail-ride travel time. Seventh, the travel cost is the cheapest in Utsunomiya, followed by Fujisawa, Meidai-mae, and Hikari-gaoka. The travel cost is more expensive in the two cities of Tokyo than other cities is because the individuals in Tokyo have to pay the transit fare whereas they individuals in local cities pay the gas cost, which is much cheaper than the transit fare. Eighth, the transfer in Fujisawa and Utsunomiya is nearly equal to zero while that in Meidai-mae and Hikari-gaoka is 0. This is simply because the individuals in Fujisawa and Utsunomiya use only automobile whereas those in Meidai-mae and Hikari-gaoka use rail. As the rail network is TABLE Descriptive Statistics of Respondents in the Four Cities Unit Average S. D. Min. % Median % Max. Meidai-mae (N=0) Age...0.0.0 0.0 0.0 Gender 0. 0. 0.0 0.0.0.0.0 Annual household income Million Yen...0.0.0.0.0 Car ownership 0. 0. 0.0.0.0.0.0 Travel distance Km.. 0..0.. 0. Travel time Minutes...0.0 0.0.0 0.0 Travel cost Yen 0..0 0.0 0.0 00.0.0 0. Transfer 0. 0. 0.0 0.0 0.0 0.0.0 Transfer time Minutes.. 0.0 0.0 0.0 0.0.0 Access/egress travel time Minutes.. 0.0 0.0 0.0.0.0 Hikari-gaoka (N=) Age...0.0.0.0 0.0 Gender 0. 0. 0.0 0.0.0.0.0 Annual household income Million Yen...0.0.0.0.0 Car ownership. 0. 0.0.0.0.0.0 Travel distance Km.0. 0..0... Travel time Minutes...0.0.0.0.0 Travel cost Yen.. 0.0 0.0 0... Transfer 0. 0. 0.0 0.0 0.0 0.0.0 Transfer time Minutes.. 0.0 0.0 0.0 0.0.0 Access/egress travel time Minutes.. 0.0 0.0 0.0.0.0 Fujisawa (N=) Age.. 0.0.0.0.0.0 Gender 0. 0. 0.0 0.0.0.0.0 Annual household income Million Yen.0..0.0.0.0.0 Car ownership. 0. 0.0.0.0.0.0 Travel distance Km..0 0..... Travel time Minutes...0 0.0.0 0.0.0 Travel cost Yen.. 0.0..0. 00.0 Transfer 0. 0. 0.0 0.0 0.0 0.0.0 Transfer time Minutes 0.. 0.0 0.0 0.0 0.0.0 Access/egress travel time Minutes.. 0.0 0.0 0.0 0.0.0 Utsunomiya (N=) Age...0.0.0.0.0 Gender 0. 0. 0.0 0.0.0.0.0 Annual household income Million Yen...0.0.0.0.0 Car ownership. 0. 0.0.0.0.0.0 Travel distance Km.. 0..... Travel time Minutes...0.0.0 0.0 00.0 Travel cost Yen.. 0.0...0. Transfer 0.0 0. 0.0 0.0 0.0 0.0.0 Transfer time Minutes 0.0 0. 0.0 0.0 0.0 0.0 0.0 Access/egress travel time Minutes 0.. 0.0 0.0 0.0 0.0.0
0 0 0 wide and complicated in Tokyo, the rail users are often required to change trains at station. This is also reflected in the transfer time. Finally, the average access/egress travel time is. minutes in Meidai-mae,. minutes in Hikari-gaoka,. minutes in Fujisawa, and 0. minutes in Utsunomiya. This reflects that the nearest stop of public transit is bus stop in Utsunomiya whereas that is rail station in the other cities. SURVEY RESULTS Awareness and Experience of Carsharing TABLE shows the comparison of respondents awareness and experience of carsharing by car ownership subgroup among the four cities. First, no respondent is currently a member of carsharing organizations although the carsharing service is available in Meidai-mae and in Hikari-gaoka. Second, in Meidai-mae and Fujisawa,.0 percent and. percent of total respondents have never heard the carsharing respectively whereas in Hikari-gaoka and Utsunomiya,. percent and 0. has never heard the carsharing respectively. The awareness of carsharing in Hikari-gaoka is lower although the carsharing service is available there. This is probably because the trip distance of respondents are longer than that in other areas; and the long-distance travelers are expected to choose rail rather than carsharing service in the context of Tokyo. The awareness of carsharing in Utsunomiya is lower simply because no carsharing service is available there. Third, in all the cities, more no-car owners have never heard of the carsharing than the car owers. This may mean that the car owners are concerned with the availability of other car-based transportation mode while no-car owners are not interested in the carsharing or they have no car license. Fourth, less than 0 percent of respondents know the names of carsharing operators while very few respondents have considered the carsharing use in Fujisawa and Utsunomiya. This is probably because no carsharing service is available in those two cities. TABLE Awareness and Experience of Carsharing (CS) in the Four Cities (Multiple answers) I have never I am aware of I know the names heard the CS CS by name. of CS operators I have considered to use the CS. I am now a member of CS. Total respondents Meidai-mae No-car owner (.%) (.%) (.%) (.%) 0 (0.0%) Car owner (.%) 0 (.%) 0 (.%) (.%) 0 (0.0%) Total (.0%) (0.%) (.0%) (.%) 0 (0.0%) 0 Hikari-gaoka No-car owner (.%) (.%) (.%) (.%) 0 (0.0%) 0 Car owner (.0%) (.%) (.0%) (.%) 0 (0.0%) Total (.%) (0.%) (.%) (.%) 0 (0.0%) Fujisawa No-car owner (.%) (.%) 0 (0.0%) (.%) 0 (0.0%) Car owner 0 (.%) (.%) (.%) (.0%) 0 (0.0%) Total (.%) (.%) (.%) (.%) 0 (0.0%) Utsunomiya No-car owner (.%) (.%) 0 (0.0%) 0 (0.0%) 0 (0.0%) Car owner 0 (.%) (.0%) (.%) (0.%) 0 (0.0%) Total (0.%) (.%) (.%) (0.%) 0 (0.0%) Note : The percent in parenthesis means the share of responses to corresponding answer out of the total respondents. Note : The respondents showing no answer are removed from the dataset. Note : Total respondents may not be equal to the sum of answered cases because multiple answers are included in each case. FIGURE shows that the respondents impressions of carsharing by age subgroup in the four cities. It reveals the difference of impressions of carsharing among different age subgroups. First, the shares of respondents considering economical in the younger subgroups including 0s or less and 0s and in the senior subgroup including 0s, 0s, 0s and 0s or over are. percent and 0. percent in Meidai-mae,. percent and. percent in Hikari-gaoka,. percent and. percent in Fujisawa, and.0 percent and. percent in Utsunomiya, respectively. This may mean that the individuals in senior subgroups consider more that the carsharing is economical than those in younger subgroups in all the cities. There are the two possible reasons for this. The one is that the younger individuals earn lower income than the senior individuals. The other is that the younger individuals are more aware of carshaing than the senior individuals. The senior individuals have the
0.0 0.0 Economical Environmentally friendly Difficult to use Difficult to understand the charge system To become a member is bothersome 0.0 0.0 0.0 0.0 0.0 0.0 0.00 0s or less 0s or over 0s or less 0s or over 0s or less 0s or over 0s or less 0s or over Meidai mae Hikari gaoka Fujisawa Utsunomiya FIGURE Impressions of Carsharing by Age Subgroup in the Four Cities (Multiple Answers) Note: The respondents showing no answer are removed from the dataset. 0 0 0 hypothetical impression regarding the carsharing cost while the young individuals have the realistic opinion on carsharing cost. Second, the shares of respondents considering environmentally friendly in the younger subgroups and in the senior subgroups are. percent and 0. percent in Meidai-mae,. percent and. percent in Hikari-gaoka,. percent and. percent in Fujisawa, and. percent and. percent in Utsunomiya, respectively. This may mean that the individuals in senior subgroups consider that the carsharing is environmentally friendly than those in younger subgroups in all cities. This probably reflects the difference of concerns with the environmental issues between the younger subgroups and senior subgroups. Third, the shares of respondents considering difficult to understand the charge system in the younger subgroups and in the senior subgroups are.0 percent and. percent in Meidai-mae,. percent and. percent in Hikarigaoka,. percent and. percent in Fujisawa, and. percent and. percent in Utsunomiya, respectively. This means that the individuals in younger subgroups consider that it is difficult to understand the charge system than those in senior subgroups in all cities. Fourth, the share of the individuals considering economical is the highest in Hikari-gaoka, followed by that in Meidai-mae, Fujisawa, and Utsunomiya. This reflects the average wage level in those areas. Fifth, the shares of the individuals considering difficult to use and those considering to become a member is bothersome are the highest in Utsunomiya, followed by those in Fujisawa, Hikari-gaoka, and Meidai-mae. This probably reflects the availability of carsharing service in the cities. FIGURE shows that the expected actions when an individual would be a member of carsharing service. First, -0 percent of the respondents do nothing even if they would be the member while - percent of the respondents stop the car ownership. The share of stop the car ownership is the lowest in Utsunomiya and followed in Fujisawa. This is probably because the local people have the difficulties to use another transportation mode in Fujisawa and Utsunomiya since the public transportation is poor there. Next, - percent of the respondents purchase an additional car in Meidai-mae, Hikari-gaoka, and Utsunomiya whereas 0. percent do it in Fujisawa. -0 percent of the respondents change the car into a new one in Meidai-mae, Hikari-gaoka, and Utsunomiya whereas. percent do it in Fujisawa. In Fujisawa, the individuals do not purchase an additional car but change the car into a new one. This is probably because the household size is larger in Fujisawa than in other areas. Note that Fujisawa is one of the typical suburban areas where the households with a couple and a child or children typically reside in a household. Although they currently have a family car with larger capacity, they may change it into a smaller car in a case of using the carsharing service. TABLE shows the results of stated choice on carsharing membership under different hypothetical cases in the four cities. Case 0 is the base case in which the monthly membership fee is 000 yen, time-based charge is 00 yen per minutes, distance-based charge is yen per kilometer, and the travel time from home to the nearest carsharing station is minutes. Note that US dollaor is equal to 0 yen as of March, 00. The levels of service in Case 0 refer to the typical carsharing service in urban areas in the Tokyo Metropolitan Area.
0. 0. 0. Meidai mae Hikari gaoka Fujisawa Utsunomiya 0. 0. 0. 0 Purchase an additional car Change the car into a new one FIGURE Expected Actions in a Case of being a Member of Carsharing Service (Single Answer) Note: The respondents showing no answer are removed from the dataset. TABLE Membership Choice of Carsharing under Different Hypothetical Cases in the Four Cities Case Membership Time charge Distance Join the CS Not available Total fee charge membership Walk time to the nearest CS station Stop the car ownership Not join the CS membership Do nothing Yen/month Yen/ min. Yen/km Minute (%) (%) (%) Meidai-mae 0 000 00... 0 000 00..0. 000 00 0...0 000 00 0... 000 00... 000 00 0... 000 00... Hikari-gaoka 0 000 00. 0.. 000 00 0.0 0.. 000 00 0..0 0. 0 000 00 0...0 0 000 00..0. 000 00 0 0. 0. 0. 000 00. 0. 0. Fujisawa 0 000 00 0... 000 00...0 000 00 0...0 000 00 0.0.0.0 0 000 00... 000 00 0.0 0.0.0 0 000 00... Utsunomiya 0 000 00... 000 00.. 0. 000 00 0... 000 00 0. 0.. 000 00. 0 0. 0. 000 00 0. 0. 0.0 000 00. 0..
0 0 0 0 0 The other cases have the different combinations of fee/charge and accessibility. As stated earlier, three cases including Case 0 and other randomly-selected two cases are presented to a respondent in the survey sheet. The respondents are requested to answer whether to join the carsharing membership for each case. First, in Case 0, the share of respondents joining the carsharing membership is. percent in Meidai- mae,. percent in Hikari-gaoka, 0. percent in Fujisawa, and. percent in Utsunomiya, respectively. The highest share of respondents joining the membership in Fujisawa and the lowest share in Utsunomiya may reflect the share of respondents considering that the carsharing is economical. This means the individual s recognition of carsharing-related cost impacts the carsharing membership significantly. Second, Case is the most preferred in Meidai-mae and Utsunomiya whereas Case is the most preferred in Hikari-gaoka and Fujisawa. There are two possible reasons for that Case is preferred to Case in Meidai-mae and Utsunomiya although the monthly membership fee in Case is three-time more than that in Case. Note that the accessibility in Case is the same as that in Case. The one possible reason is that the trip distance is shorter in Meidai-mae and Utsunomiya than that in Hikari-gaoka and Fujisawa. Note that the distance charge in Case is higher than that in Case. TABLE shows that the average trip distance is. km in Meidaimae and. km in Utsunomiya whereas that is.0 km in Hikari-gaoka and. km in Fujisawa. The other possible reason is the expected monthly carsharing-use time including the travel time by carsharing and the activity time at the destination is longer in Meidai-mae and Utsunomiya than that in Hikari-gaoka and Fujisawa. Note that the time- charge in Case is lower than that in Case.TABLE shows that the average travel time in Meidai-mae and Utsunomiya is shorter than that in Hikari-gaoka and Fujisawa. Additionally, it is difficult to assume that the activity duration in one city is significantly longer than that in another city. Thus the monthly frequency of carsharing-use may be higher in Meida-mae and Utsunomiya than in other cities. As its real reason is unclear from the above analysis, further analysis is required to find the reason. Third, Case is less preferred in all the cities. Particularly in Hikari-gaoka, Fujisawa, and Utsunomiya, the share of the respondents joining the membership is the lowest in Case. Ten-minute walk to the nearest station and the expensive time charge may discourage the individuals to become the membership. Fourth, interestingly over ten percent of respondents prefer Case in Hikari-gaoka whereas about five percent prefer it in other cities. This may mean that the individuals in Hikari-gaoka will expectedly use the carsharing much longer time than those in other cities. Note that the time charge is the cheapest among the cases while the monthly membership fee and the accessibility to the nearest station is the worst among the cases. EMPIRICAL ANALYSIS Modal Choice including Carshraing The data collected in our survey includes the revealed preference (RP) data and the stated preference (SP) data. It is widely known that SP data has more biases than RP data. To analyze the individual s choice behavior with RP data and SP data, an RP/SP combined model is applied here in addition to the RP model and the SP model (Ben-Akiva and Morikawa, ). The RP model means the choice model estimated with only RP dataset while the SP model means the choice model estimated with only SP dataset. The RP/SP model is especially used to correct SP reported biases by introducing the RP information. The empirical analysis assumes the binary mode choice for analytical simplicity. The choice set is assumed as follows. First, if an individual chose the car in the RP question, the choice set of RP model for the trips made by this individual is assumed to be composed of car and an available public transit. The choice set of SP model for the trips made by this individual is assumed to be composed of car and carsharing. Second, if the individual chose the public transit in the RP question, the choice set of RP model for the trips made by this individual is assumed to be composed of the public transit and car (for car owner) or other available public transit. When no public transit is available, the alternative mode is assumed to be walk. The choice set of SP model for the trips made by this individual is assumed to be composed of the public transit and carsharing. Third, if the individual chose the walk in the RP question, the choice set of RP model for the trips made by this individual is assumed to be composed of the walk and car (for car owner) or available public transit. The choice set of SP model for the trips made by this individual is assumed to be composed of the walk and carsharing. The utility function is assumed to be a linear function of explanatory variables. After a number of trials and errors considering the various combinations of explanatory variables in the utility function, the utility functions are specified with a set of the same variables among the four cities. This is because we intend to compare the estimated results among the cities. The estimated results are summarized in TABLE. In-vehicle travel time is defined as the travel time including riding public transit, driving a car, and traveling as a passenger of a car; out-of-vehicle travel time is defined as the walking time and the waiting time at public transit stops/stations; travel cost is defined as the travel expense including the public transit fare, the fuel cost of car, and carsharing charge for using the carsharing service; dummy of male (car) is defined as a car-specific variable which is equal to if the individual is male and 0 otherwise; dummy of high income (car) is defined as a car-
TABLE Estimation Results of Modal Choice with RP, SP, and RP/SP Models in the Four Cities Meidai-mae Hikari-gaoka RP SP RP/SP RP SP RP/SP Variable Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat. In-vehicle travel time -0.0 -.** -0.0 -.** -0.0 -.** -0.0 -.** -0.0 -.** -0.0 -.** Out-of-vehicle travel time -0. -.** -0.0-0. -0. -.** -0. -.** -0.0-0. -0.0 -.** Travel cost -0.00 -.** -0.00 -.** -0.00 -.** -0.00 -.0** -0.00 -.** -0.00 -.** Dummy of male (Car) 0.. -0.0-0. 0..0 0.0.** 0. 0. 0.0.** Dummy of high income (Car) 0.00 0.0-0. -. 0.00 0.0 0. 0. -0.0 -. 0. 0. Constant (Rail) -0.0 -.* -.0 -. -0. -.** -0. -. -. -.** -0. -.* Constant (Bus). 0. -. -0....0.** -. -.**.0.0* Constant (Car) -. -.** -. -.0** -.0 -.** -0. -.** -.0 -.** -. -.0** Constant (Carsharing) -.0 -.** -.0 -.* -. -.** -. -.** Scale parameter 0.0.* 0..** Number of observation 0 0 0 Initial log-likelihood -. -. -.0-0. -. -. Final log-likelihood -. -0. -. -. -0. -. Adjusted rho-squared 0. 0. 0. 0. 0. 0. Fujisawa Utsunomiya RP SP RP/SP RP SP RP/SP Variable Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat. In-vehicle travel time -0.0 -.** -0.0 -.** -0.0 -.** -0.0 -.** -0.0 -.** -0.0 -.** Out-of-vehicle travel time -0.0 -.** -0.0 -.** -0.0 -.** -0.0 -.** -0.0-0. -0. -.** Travel cost -0.00 -.** -0.00 -.** -0.00 -.** -0.0 -.** -0.00 -.** -0.0 -.** Dummy of male (Car) 0.0 0. 0.. 0..0 0.0 0. 0.0.0 0.0 0. Dummy of high income (Car) -0.0 -..0.** -0. -0. -0. -0. 0. 0. -0. -0. Constant (Rail) 0. 0. -. -. 0.0 0..0.0. 0..0.0 Constant (Bus) 0. 0. -0. -0. 0. 0...** -.0 -.**.0.** Constant (Car) -0. -.** -. -.** -0. -.** 0.0. -. -.** 0.0. Constant (Carsharing) -. -.** -0.0-0. -. -.** -. -0. Scale parameter 0..** 0.0 0. Number of observation 0 0 Initial log-likelihood -. -0. -. -0. -. -. Final log-likelihood -.0 -. -. -. -0. -. Adjusted rho-squared 0. 0.0 0. 0. 0. 0. Note: ** means the coefficient is significant in percent confidence level and * means the coefficient is significant in 0 percent confidence level. 0
0 0 0 0 specific variable which is equal to if the annual income of household the individual belongs to is over seven million yen and 0 otherwise; constant (rail, bus, car, carsharing) is defined as a mode-specific dummy variable; and the scale parameter which is defined as the ratio of standard deviation in error component of SP-based utility function to that of RP-based utility function. First, in Meidai-mae, the three models have high model fitness. The signs of coefficients are all reasonable. T-statistic of carsharing-specific constant shows it is significantly negative. This means that the carsharing has the negative attractiveness in comparison to other travel mode. The scale parameter estimated in RP/SP model is significant but it is almost equal to zero. This means that the variance of error components in SP model is much larger than that in RP model. The estimated value of in-vehicle time and out-of-vehicle time are. yen per minute and. yen per minute respectively in RP/SP model. The values of walking and waiting time are higher than that of in-vehicle time. Second, in Hikari-gaoka, the three models have high model fitness. The signs of coefficients are all reasonable. T-statistic of carsharing-specific constant again shows it is significantly negative. In both RP model and RP/SP model, the car-specific dummy of male is significantly positive. This means that the male individual tends to choose the car than the female. The scale parameter estimated in RP/SP model is also strongly significant and it is smaller than as expected. The estimated value of in-vehicle time and out-of-vehicle time are. yen per minute and. yen per minute respectively in RP/SP model. Third, in Fujisawa, the RP and RP/SP models have high model fitness while the SP model has lower model fitness. One of the possible reasons for this result is that the respondents may feel difficulty to imagine the carsharing stations located near their homes. As the survey area is located in the typical sub-urban districts where the population density is quite low, the hypothetical case of carsharing service may not be realistic for the respondents. The values of in-vehicle time and out-of-vehicle time estimated from the RP/SP model are 0. yen per minute and. yen per minute respectively. The values of walking and waiting are lower than that of in-vehicle time. This is also different from other results. Fourth, in Utsunomiya, the three models have high model fitness. However, although the scale parameter is smaller than, it is not significant statistically. The values of in-vehicle time and out-of-vehicle time estimated from the RP model are. yen per minute and 0. yen per minute respectively while those estimated from the SP model are. yen per minute and 0.0 yen per minute respectively. The value of out-ofvehicle travel time in the SP model is much lower than that of in-vehicle travel time. One of the possible reasons for this is that the respondents cannot appropriately imagine the situation in which they walk to the carsharing station. Another reason is that, compared with other areas, it is highly expected that the local people in Utsunomiya have less knowledge or experience of carsharing service. It should be noted that the most of carsharing service in Japan is operated in urban areas. Carsharing Membership Choice The membership choice is analyzed with the nested logit models using the logsum (LS) variables estimated with the modal choice models shown in FIGURE. The results are shown in FIGURE. CS Membership fee is defined as the daily cost to be the carsharing member. LS variable for non-cs member is estimated with the RP modal choice model whereas LS variable for CS member is estimated with the RP/SP modal choice model. The results show that the model fitness is high enough in all cities. The most of coefficients of explanatory variables are also statistically significant and consistent with their hypothesized effects on utility in the four cities. FIGURE Estimation Results of Carsharing Membership Choice Model Meidai-mae Hikari-gaoka Fujisawa Utsunomiya Variable Coef. t-stat. Coef. t-stat. Coef. t-stat. Coef. t-stat. CS membership fee -0. -.** -0.0 -.** -0.0 -.** -0.0 -.0** LS parameter: Non-CS member 0.0.0** 0..** 0.0.** 0..** LS parameter: CS member 0.0 0. 0.0.** 0.0.** 0.0.** Dummy of CS member.. -. -.** -. -.** -. -.** Number of observation 0 0 Initial log-likelihood -. -0.0-0. -. Final log-likelihood -. -0. -. -. Adjusted likelihood ratio 0. 0. 0. 0. Note: ** means the coefficient is significant in percent confidence level. DISCUSSION The implications of the survey and analysis results are summarized as follows. First, the car owners are more aware of the carsharing than the non-car owners but the car owners have less experience to consider the carsharing-use than the non-car owners. This may mean that the individuals have difficulty to change the car-use
0 0 0 0 0 life into the car-free life once they own the car. On the other hand, for the no-car owners, the carsharing may be attractive because it is one of the opportunities to use a car. They could suggest that the carsharing may increase the car-use because the no-car owners may start to use cars through the carsharing service while the car owners may not change their transportation mode. Second, the individuals in the senior subgroups have more positive impressions of carsharing than those in the younger subgroups. The senior respondents consider the carsharing to be economical and environmentally friendly. They do not consider that the charge system of carsharing is difficult. However, the car ownership rate is higher in senior subgroups than that in younger subgroups because the household income of senior individuals is higher than that of young individuals. This may also suggest that the retired individuals in 0s or 0s may be the potential carsharing users. This is because they may stop the car ownership due to less frequency of out-of-home activities. As Japan is now facing the rapid ageing, the retired individuals may be the critical potential target in the future. Third, 0-0 percent of individuals could stop the car ownership if they would be a member of carsharing service. This may mean that a considerable number of individuals may give up using cars once they become the carsharing members. This may suggest that the availability of reasonable carsharing service is the most important to reduce the car ownership. Fourth, the individuals seem to make a rational decision on the membership choice. Lower membership fee, cheaper charge of carssharing-use, and better accessibility to the nearest carsharing station are preferred. This suggests that the balanced combination of the membership fee, time-charge, distance-charge, and accessibility is critical to achieve more carsharing members. Fifth, the comparative analysis among the four cities showed that the availability of carsharing service, service quality of public transportation, trip distance, and household income influence the choice of carsharing membership. More individuals have the chance to become the member in the area where the carsharing service is available. Individuals have less motivation to use the carsharing in the area where the car is the dominant transportation mode mainly due to the poor public transportation. The market potential of carsharing may be weaker in the area where the average trip distance is longer. The individuals in the lower-income area may not consider the carsharing is economical. These may suggest that the potential market of carsharing is located in the urban areas where middle- to high-income people reside and where rich public transportation service is provided. Finally, the modal choice models and the carsharing membership choice models are successfully estimated although the model fitness of modal choice models in Fujisawa and Utsunomiya is not high enough. They could be useful for analyzing the potential demand of carsharing. CONCLUSIONS This paper reported the results of the questionnaire survey on the carsharing in the four cities of Japan. The awareness, experience, expected action regarding the car ownership, stated choice of carsharing membership, and modal choice were analyzed with the collected data. Then, the modal choice and the carsharing membership choice were analyzed. The further research issues are summarized as follows: First, the modal choice models in Fujisawa and Utsunomiya should be further explored. The model with non-linear utility function may be tried for better estimation results. Next, the demand models for carsharing should be verified with the observed dataset of modal choice of existing carsharing users. One of the difficulties of this analysis is to collect the data of existing carsharing users. Actually no carsharing user is included in the respondents of our household surveys. The customer information of carsharing service is usually collected by the private carsharing operators and it is not open to the public. Thus it may be necessary to carry out the joint research with them to verify it. ACKNOWLEDGEMENTS The carsharing survey was carried out in a joint research project with Showa Shell Sekiyu K.K. funded by the Ministry of Economy, Trade and Industry, Japan. We thank Mr. Masahiro Kakuwa, Ms. Nami Kitamura, Yuichi Shinada, Mr. Naoki Kawasaki, and Ms. Yumi Tokuda (Showa Shell Sekiyu K.K.) for their support to the survey. REFERENCES. Shaheen, S., A.P. Cohen, and M.S. Chung. North American carsharing: 0-year retrospective, In: Transportation Research Record: Journal of the Transportation Research Board, No. 0, Transportation Research Board of the National Academies, Washington, D.C., 00, pp..
0. Shaheen, S.A. and A.P. Cohen. Growth in worldwide carsharing, In: Transportation Research Record: Journal of the Transportation Research Board, No., Transportation Research Board of the National Academies, Washington, D.C., 00, pp... Barth, M., S. A. Shaheen, T. Fukuda, and A. Fukuda. Carsharing and station cars in Asia: Overview of Japan and Singapore. In Transportation Research Record: Journal of Transportation Research Board, No., Transportation Research Board of the National Academies, Washington D.C., 00, pp.0.. Foundation for Promoting Personal Mobility and Ecological Transportation. http://www.ecomo.or.jp/ environment/carshare/carshare_graph0..html (last accessed July 0, 0, in Japanese).. Barth, M., and M. Todd. Simulation model performance analysis of a multiple station shared vehicle system, Transportation Research Part C, Vol., No.,, pp... Kek, A.G.H., R.L. Cheu, and M.L. Chor. Relocation simulation model for multiple-station shared-use vehicle systems, In: Transportation Research Record: Journal of the Transportation Research Board, No., Transportation Research Board of the National Academies, Washington, D.C., 00, pp... Ben-Akiva M, and T. Morikawa. Estimation of mode switching models from revealed preferences and stated intentions, Transportation Research Part A, Vol., No., 0, pp..