Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Demand Estimation of a New Light Rail Transit in a Tourist City Considering the Change in Population Distribution –A Case Study for Kanazawa City, Japan–

Tetsuji SATO a*, Eiki TAKASUGI b, Atsuya HANDA c, Keisuke SUGITA d, Kana ISOGAI e a,b,c,d,e Department of Civil and Environmental Engineering, Chiba Institute of Technology, 2-17-1 Tsudanuma, Narashino-shi, Chiba, 275-0016, Japan a E-mail: [email protected] b E-mail: [email protected] c E-mail: [email protected] d E-mail: [email protected] e E-mail: [email protected]

Abstract: Many tourist cities in Asian countries developed or have plans to introduce a new mass transit system such as subway and light rail transit recently to reduce traffic jam and improve convenience of tourists, etc. In tourist cities, the fare of new mass transit lines can be low because tourists increase the demand of the lines compared to other cities. In this paper, we propose an estimation method of the impact of a new mass transit system and its fare level on population distribution in a city and demand of new lines considering the change in population distribution. We also develop an empirical model for Kanazawa city which is one of the famous historical tourist cities in Japan, and estimate the impact of developing a new light rail transit line and the fare level on population distribution and demand of the line.

Keywords: Light rail transit, Demand estimation, Population distribution

1. INTRODUCTION

Several cities in Asia that handle a large number of tourists have recently introduced mass transit systems, such as subway systems, new transportation systems, and light rail transit (LRT), to alleviate traffic congestion, improve the environment, and provide tourists and citizens who commute back and forth to work with convenient forms of transportation. The Okinawa Urban began operation in Okinawa, Japan, in 2003. This monorail system runs for approximately 13 kilometers from Airport to Shuri near the Shuri Castle, which has been named as a World Heritage Site. Several tourist cities in Japan, which have the streetcar system, have introduced low floor type streetcars (light rail vehicles) on the existing streetcar tracks. For example, these were introduced in Kumamoto in 1997, Hiroshima in 1999, Nagasaki in 2011, and Matsuyama in 2017. Each transportation mode has its own capacity. In a width of 3 meters and on an hourly basis, buses can carry 3,000–6,000 people, LRT can carry 3,000–11,000 people, new transportation systems can carry 10,000–20,000 people, and subways can carry 30,000–60,000 people. Generally, the construction and maintenance costs are the highest for subway systems, which is followed by those of new transportation systems, LRT, and buses.

* Corresponding author.

626

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Therefore, estimating the demand (number of users) is essential for determining whether new LRTs, new transportation systems, or subway infrastructures are required and also for determining the most appropriate mode among these. In general, the demand for new transportation lines can be estimated using the four step estimation method. Firstly, the trip generation volume and attracted trip volume is estimated. Subsequently, the trip volume distribution of all transportation modes is determined, which is followed by the trip volume distribution of each mode. Finally, the assigned trip volume (based on the line) can be obtained. The trip generation volume can be estimated in the first step by multiplying the trip generation volume per person (primary unit) with the population that can be expected in the future in each zone. Here, the future population of each zone is estimated using the primary factors cohort. It is important to note that the change in population distribution within a city which is caused by the residential development along the lines and the enhanced convenience achieved using the new lines is not taken into consideration in the ordinary method. Mass transit systems, such as LRT, new transportation systems, and subway systems, significantly impact the population distribution and should, therefore, significantly impact the transportation demand as well. Because of the usage of the newly constructed mass transit systems by tourists, a considerably large demand should be expected in tourist cities when compared with the demand in cities that attract less number of tourists. A high demand generally results in low fares. This may encourage people to relocate to new residential areas along the lines, thereby increasing the usage of those lines and further increasing the demand. There are a lot of previous studies which focus on the modelling of transportation and land use considering the interaction among transport capacity, demand and congestion, location choice of household and population distribution. Recent studies in this field contain Li et al. (2016) and Li et al. (2017). They constructed the integrated co-evolution model of land use and traffic network design based on the reference-dependent theory as for household’s location choice behavior, the multi-criteria stochastic user equilibrium model as for household’s route choice behavior, etc. They also proposed a solution algorithm of the model using the genetic algorithm and conducted numerical experiments. These studies, however, did not consider public transportation system and did not conduct empirical analyses for real cities. Studies analyzing the relationship of newly constructed public transportation system and changes in population distribution within real cities have been recently conducted by Muto et al. (2000), Sato et al. (2017), Takasugi et al. (2018), and Tomioka et al. (2018). Muto et al. constructed a computable urban economic model which assumes an equilibrium in the urban land market and an interaction between household distribution and the trip cost, and evaluated construction of a new transportation system in Gifu city, Japan. Sato et al. and Tomioka et al. have constructed locational equilibrium models of the residential land market for the city of Utsunomiya, Japan; further, they have been used to analyze the long-term impact of the new LRT, which is scheduled for inauguration on March 2022, on the population distribution within the city. Takasugi et al. developed a quasi-dynamic location equilibrium model considering the difference of influences on utility level of households between LRT and BRT, and conducted a case study analysis assuming new construction of LRT or BRT in Maebashi city, Japan. However, these previously conducted researches did not forecast the demand of new public transportation systems based on the population distribution changes; further, new public transportation systems in tourist cities were also not included in these studies. Also, these studies did not consider the impact of the fare of new lines which is likely to be low in tourist cities. Based on the content and focus of the previous studies mentioned above, this paper proposes the estimation method of the impact of a new public transportation development and

627

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

its fare on the population distribution within a particular city over the long-term. And we also propose the estimation method of the demand of new lines based on the change of population distribution. Further, an empirical model for Kanazawa, which is one of Japan’s leading tourist cities, is developed and is used to analyze the impact of developing a new LRT line on population distribution within the city and the demand of the line is estimated.

2. DEVELOPMENT OF POPULATION DISTRIBUTION ESTIMATION MODEL

2.1 Outline of the Model

The model in this paper is designed for estimating future population by zone in the long run by assuming the supply and demand of land (or floor) based on the type of residence (including single-family homes and rental apartments) and market equilibrium in each term. While setting the value of the relocation ratio by the current residence type and residence-type ratio after relocation, the model calculates the location volume (the urban social movements) from the supply of land (floor for apartment) by landlord, the land (floor) demand by household based on the relocation destination zone choice probability and market equilibrium through rent adjustment for each residence type in each zone. The model also considers the natural increase (decrease) as well as the moving out of the city and the moving into the city based on the cohort of primary factors in each term. When the population distribution for term t is obtained, the household distribution for each residence type for term t can be determined by converting the population to the number of households in each zone with the average number of household members and multiplying it with the ratio of each residence type. Subsequently, the number of social movements within the city (total number of households intending to move within the city) during terms t to t+1 is determined based on the number of household and relocation ratio for each residence type before relocation. This is followed by the calculation of the total number of households intending to move within the city for each residence type after relocation during terms t to t+1 based on the number of social movements within the city and the ratio of each residence type after relocation. Further, the household distribution after social movement is determined by calculating the location volume in each zone based on the location equilibrium for each residence type. It is further converted to the population distribution that considers only the social movement within the city using the average number of household members for each residence type. Further, the population distribution for term t+1 can be calculated by adding the net natural increase and the net inflow from outside the city from term t to term t+1. With respect to people flowing into the city from outside, we assume that the location determination in each term will be conducted using the same distribution as the population distribution in each term. By repeating this process, the model can estimate the population distribution changes within a particular city over time. In the model, we express residential land (or floor) demand by household, supply by absentee landlord and market equilibrium in each zone by assuming that there are lots of the same lands (or floors) in each zone’s market. This is the main difference as the bid-rent theory which is the general classical method to indicate household’s land choice behavior. Figure 1 shows the flow chart of the population distribution estimation model.

628

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Population distribution in term t Residence type The average ratio household personnel

Household distribution (each residence type) Total number of household intending to move < Household > < Landlord > LRT Fare Demand of land or Supply of land or Social / Natural floor in each zone floor in each zone Increase / Decrease from term t to t+1 Housing Rent land or floor market Rent adjustment Demand = Supply adjustment

Household distribution (each residence type) The average household personnel Population distribution Natural increase / decrease Moving out of / into the city Population distribution in term t+1 Residence type The average ratio household personnel

Household distribution (each residence type) Total number of household intending to move < Household > < Landlord > LRT Fare Demand of land or Supply of land or Social / Natural floor in each zone floor in each zone Increase / Decrease from term t+1 to t+2 Housing Rent land or floor market Rent adjustment Demand = Supply adjustment

Household distribution (each residence type) The average household personnel Population distribution Natural increase / decrease Moving out of / into the city

Figure 1. The flow chart of the population distribution estimation model

2.2 Residential Land (Floor) Demand by Household

We assume that households can be classified into two categories: households with the intention of moving and those without. Households with the intention of moving to each residence type will choose the zone to move to depending on the utility level, which is affected by the convenience of using the transport system, etc.

629

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

The probability of choosing the destination zone by each household can be expressed as a logit model defined in Equation (3). The equation is derived by solving the maximization problem which is expressed by Equations (1) and (2).

1 EPUPPkt=−max ikt ikt ikt ln ikt (1) Pikt i  k s.t. P = 1 (2) i ikt

exp(kU ikt ) P = (3) ikt exp  U i ( k ikt ) where, i : zone, k : residence type, t : term, Pi : choice probability of zone i of household with the intention of moving, and U : utility of household.

The utility of household can be expressed as the sum of the partial utility and zone specific attractiveness which is not contained in the partial utility (adjustment term). It is expressed by Equation (4).

UViktiktik=+ (4) where, V : partial utility of household, and  : adjustment term.

The partial utility of household is assumed to be expressed by Equation (5). In this paper, considering impact of existence of LRT and its fare level on utility, we introduce the dummy variable of railway (1 if mode of the nearest public transport station or stop from home is railway, 0 otherwise), dummy variable of LRT (1 if mode of the nearest public transport station or stop from home is LRT, 0 otherwise), time required to the nearest public transport station or stop from home on foot, and time required and fare to the downtown area from the nearest transport station or stop by public transportation as explanatory variables of the partial utility function.

T (5) Vikt=+ const a kln( I ikt) + b k ln( r ikt) +++ c k DR ikt d k DL ikt e k T 1 ikt + f k T 2 ikt ++ g k F ikthZ k ikt where, const : constant term, I : household income, r : land rent (floor rent), DR : dummy variable of railway, DL : dummy variable of LRT, T1 : time required to the nearest railway / LRT station or bus stop from home, T2 : time required to travel to the downtown area from the nearest railway / LRT station or bus stop,

630

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

F : fare of public transportation to the downtown area, and Z : vector of other factors which indicate living environment.

Area demand for residential land (floor) per household is determined based on the Roy’s theorem. The residential land (floor) demand in each zone is determined by multiplying the number of households moving into the zone by the area demand for land (floor) per household; These are given by Equations (6)–(8).

hk LIikt= ikt (6) rikt

Diktiktikt L N= (7)

N PiktiktTkt N= (8) where, L : area demand for residential land (floor) per household, D : area demand for residential land (floor), Ni : the number of households moving into zone i, NT : total number of households intending to move, and h : parameter.

2.3 Residential Land (Floor) Supply by Absentee Landlord

The residential land (floor) supply by absentee landlords is given by Equation (11), which is derived by solving the profit maximization problem defined in Equations (9) and (10). Equation (11) indicates the increase in supply as rents increase.

 (9)  iktiktiktikt=−max rSCS ( ) Sikt  S ikt (10) s.t.ln1 -CSS( iktik) =− ikt  Sikt

 ik SSikt =−1 ikt (11) rikt where,  : profit of absentee landlord, S : supply area of residential land or floor, C : supply cost of residential land or floor, S : available area for residential land or floor, and  : parameter.

2.4 Supply and Demand Equilibrium of Residential Land (Floor)

The supply and demand equilibrium of residential land (floor) by residence type in each zone is expressed by Equation (12).

(12) Dikt( r ikt) = S ikt( r ikt )

631

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

3. ANALYSYS OF IMPACT OF DEVELOPING A NEW LRT IN KANAZAWA CITY

3.1 Target City and Setting the Route of LRT

The target of the study is Kanazawa City in Ishikawa Prefecture, which is considered to be one of Japan’s leading tourist cities. Population of Kanazawa City in 2015 is about 466 thousand (National census). The number of tourists from outside the city in 2017 is about 10,221 thousand (Kanazawa City Government). Kanazawa was developed as the castle town of the Kaga Domain (also known as the Kanazawa Domain) during the Edo period, which began in 1603. It was one of the largest cities in Japan along with Edo (present-day Tokyo), Osaka, and Kyoto. The city did not suffer from carpet bombing in World War II; therefore, old streets and old buildings remained as they were in several parts of the city. The historic tourist attractions in Kanazawa include the Kenrokuen Garden, Kanazawa Castle, Higashichaya town, Samurai house site, and Ninjadera Temple. Buses are the only mode of public transportation in the city, and the business district contains a transit mall. Modal shares of car, bus and rail for person trips in Kanazawa metropolitan area are about 67.2%, 4.6% and 1.8% respectively (Kanazawa metropolitan area person trip survey in 2007). By considering the convenience of tourists and road width, the proposed route through which the new LRT has to be built will have the JR (Japan Rail) Kanazawa Station at its center, and it will connect the Ishikawa Prefectural Office, residential areas, and the Kanazawa port on the northwest side of the station with downtown Kanazawa, the Kenrokuen Garden, and Nomachi Station on the south. Figure 2 shows location of Kanazawa city and the route of a new LRT which we assumed.

KanazawaKanazawa Port Port Kanazawa city

Tokyo Ishikawa Prefectural Office 0 500 1000km JR Kanazawa Station

LRT stations

JR, Semi-public sector railway

Nomachi Private railway Station LRT 0 2 4km

Figure 2. Location of Kanazawa city and the route of a new LRT

632

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

3.2 Target Area and Zone Division for Analysis

The target area of the analysis is the city planning areas that are located inside Kanazawa City. The area designated for urbanization is divided into 500-m GPS meshes, and the urbanization control area is divided into 1-km meshes. The number of analysis zones inside the target area is 449, excluding the meshes where the population was zero in 2015. Figure 3 shows the target area and zones for the empirical analysis.

0 4 8km

Figure 3. The target area and zones

3.3 Parameter Estimation of the Partial Utility Function and the Questionnaire Survey

By considering Endo (2008) as reference, a conjoint analysis method and a questionnaire survey based on an L18 (21 × 37)-type orthogonal table are employed to estimate the parameters of Equation (5) which is the partial utility function for each residence type. This method consists of having the respondents perform a five-stage assessment for 18 hypothetical areas which assume levels of 8 factors in the questionnaire survey and estimating the parameters using individual data of the survey. Here, two levels have to be established for 1 among the 8 factors and three levels have to be established for the remaining 7 factors. Further, the questionnaire is prepared by setting various combinations of the established levels of 8 factors based on the orthogonal table. Based on the explanatory variables of Equation (5), 5 among the 8 factors correspond to a) “land rent (floor rent)”; b) “nearest mode of three public transports (railway, LRT, and bus) from home” which is compatible with dummy variables of railway and LRT; c) “time required to the nearest public transport station or stop from home on foot”; d) “time required to travel from the nearest public transport station or stop to the downtown area”; and e) “the fare from the nearest public transport station or stop to the downtown area”. Based on the variables of the partial utility function in the previous studies, the remaining 3 factors are set to be “the time required to walk from home to the nearest elementary school,” “the maximum flood depth in case of river overflow,” and “the risk of landslide”. These 3 factors are the variables of Vector Z in Equation (5). Only “the risk of landslide” among these 8 factors have two established levels (“no risk” and “having a risk”). Three levels or two levels for each factor are established based on the actual circumstances in Kanazawa City.

633

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Table 1. A part of the questionnaire for single-family homes

The time The nearest The time The fare from required to The time mode required the nearest travel from the required of three public to the nearest public The maximum nearest public to the nearest Land price transports public transport flood depth The risk of Five-stage transport elementary per a square to the transport station or in case of landslide assessment station or school meter downtown station or stop to the river overflow stop to the from home on area stop from downtown downtown foot from home home on foot area area e.g. LRT 5 min 15 min ¥200 14 min 0.5m no risk ¥77,000 4 /5

1 Railway 5 min 10 min ¥150 6 min 0m no risk ¥48,000 /5 2 LRT 5 min 15 min ¥200 14 min 0.5m no risk ¥48,000 /5 3 Bus 5 min 20 min ¥250 22 min 2m no risk ¥48,000 /5 4 Railway 15 min 10 min ¥200 14 min 2m no risk ¥48,000 /5 5 LRT 15 min 15 min ¥250 22 min 0m no risk ¥48,000 /5 . . . 11 Bus 5 min 10 min ¥200 22 min 0.5m having a risk ¥77,000 /5 12 Railway 5 min 15 min ¥250 6 min 2m having a risk ¥77,000 /5 13 LRT 5 min 20 min ¥150 14 min 0m having a risk ¥106,000 /5 14 LRT 15 min 10 min ¥150 22 min 2m having a risk ¥106,000 /5 15 Bus 15 min 15 min ¥200 6 min 0m having a risk ¥106,000 /5 16 Railway 15 min 20 min ¥250 14 min 0.5m having a risk ¥106,000 /5 17 Bus 25 min 10 min ¥250 14 min 0m having a risk ¥106,000 /5 18 Railway 25 min 15 min ¥150 22 min 0.5m having a risk ¥106,000 /5

The questionnaire (for example, single-family homes) based on the orthogonal table is presented in Table 1. Here, land prices are used as proxy variables for land rents in the questionnaire and in Equation (5), because it is difficult to collect data of land rent and land price is defined to be equal to the rent divided by the interest rate. The questionnaire also inquired about certain relevant personal attributes, including age and income level, whether the respondent intends to relocate inside or outside Kanazawa City within five years, and expected frequency of use of LRT. The survey sheets were distributed in the 15 areas centering around the city government branches on November 3 and 4, 2018 by the postal survey. The number of distributions in each area is set according to the population of the area. For single-family homes, 1000 sheets were distributed, out of which 195 were completed. For rental apartments, a total of 1000 sheets were distributed, out of which 39 for apartments with a lot of space (more than 40 m2) and 35 for others (less than 40 m2) were completed. The parameter estimations of the partial utility functions were conducted with the ordinary least squares (OLS) method and the reduction method, which removes insignificant variables at the 5% significance level, and repeat estimations, considering sign conditions and using individual data of the survey. The results of the estimations are shown in Table 2. From the survey, it is also indicated that approximately 1.1% of the respondents living in single-family homes were considering to move within Kanazawa in five years; this figure was about 23.7% for residents of apartments (more than 40 m2) and about 17.1% for residents of apartments (less than 40 m2).

634

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Table 2. Parameter estimation results of the utility function Rental apartments Rental apartments Single-family homes (more than 40 m2) (less than 40 m2) Constant term 5.0288 (23.754) 5.9251 (13.020) 5.5826 (11.451) Household income 0.0008 (3.491**) (ten thousand yen) Land price / Floor rent -0.1094 (-5.970**) -0.2025 (-4.614**) -0.4091 (-4.725**) (ten thousand yen) Dummy variable of railway 0.2316 (3.641**) 0.2102 (1.901*) Dummy variable of LRT 0.2217 (3.394**) 0.1614 (1.417*) The time required to the nearest station or stop -0.0371 (-10.894**) -0.0334 (-5.610**) -0.0373 (-2.767**) from home on foot (min) The time required to travel to the downtown area from -0.0179 (-3.277**) -0.0277 (-2.537**) the nearest station or stop (min) The fare to the downtown area from the nearest station -0.0031 (-4.590**) -0.0030 (-2.611**) or stop (min) The time required to the nearest elementary school -0.0111 (-2.934**) -0.0227 (-3.293**) -0.0113 (-1.350*) from home on foot (min) The maximum flood depth -0.2842 (-8.049**) -0.2542 (-4.357**) -0.3782 (-5.480**) in case of river overflow (m) The risk of landslide -0.2821 (-3.528**) -0.6766 (-4.926**) -0.3685 (-2.345**) Dummy variable of the 60's 2.1334 (8.364**) Note: The figures in parentheses indicate the t value. *indicates signficance at 10% level and **indicates signficance at 1%level.

3.4 Parameter Settings of Other Functions

The logit parameter θ in Equation (3) is set to 1.00 according to the previous studies. Other parameters; τ in Equation (4), h in Equation (6), and σ in Equation (11) are calibrated to match the circumstances of 2015. Here, h is established according to the residence type, and τ and σ are established according to the residence type and zone. While calibrating, the data of 2015 for all the exogenous variables in Equations (3)–(8), (11) and (12) based on the residence type and zone are required. Land price in each zone is estimated with the land price function which is estimated using the public land price data within Kanazawa City and levels of explanatory variables of the function for each zone. Room rent of rental apartment for each resident type in each zone is calculated using the ratio of average land price and average rent in Kanazawa City. The estimation result of the land price function is shown in Table 3.

635

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Table 3. Parameter estimation result of the land price function Constant term 12.3421 (24.472) Front road width 0.1236 (1.632*) Road distance to -0.2281 (-3.409**) the nearest railway station Dummy varable of low -0.2044 (-1.631*) building residential zone Dummy varable of 0.5780 (4.685**) commercial zone Dummy varable of zone -0.5281 (-2.912**) outside the zoning area Dummy varable of zone which the nearest railway 0.5337 (3.055**) station is JR Kanazawa station

Note: The objective variable is ln(land price). The figures in parentheses indicate the t value. *indicates signficance at 10% level and **indicates signficance at 1%level.

To determine the time that is required to travel from home to each facility, the position of home in each zone is assumed to be the center of the zone. The linear distance from home to each facility is measured using the distance measurement function in ArcGIS. It is further converted to road distance by multiplying it with the road linear distance ratio that has been calculated by Morita et al. (2014) (1.3508) and divided by the standard walking speed (80 m/min). As for “The time required to travel from the nearest public transport station or stop to the downtown area”, the time required from the nearest public transport station or stop of the center of each zone to the JR Kanazawa Station by public transport is estimated with the NAVITIME. “The maximum flood depth in case of river overflow” and “the risk of landslide” are evaluated based on the Kanazawa City hazard map (Kanazawa City Government). The supply area of residential land (floor) in Equation (11) is calculated by multiplying the residential land (floor) demand per household with the number of households. For single-family homes, the potential supply area is calculated by multiplying the rate of empty homes in the areas at which buildings currently exist with the field; for rental apartments, it is calculated by multiplying it with the floor area ratio designated for each zoning. In the areas that have been designated for urbanization, farm area that could potentially be transformed into residential areas are added to the potential supply area.

3.5 Estimation of Population Distribution in the Future

We can estimate the future time-series population distribution using the parameters of Equations (3) – (8), (11), and (12) and the basic data for each zone. The estimated population distribution in 2040 (rate of change from 2015) is shown in Figure 4.

636

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Figure 4. Estimation result of population distribution in 2040 (rate of change from 2015)

3.6 Impact of Developing a New LRT Line on Population Distribution in the Future

We estimate population distribution from 2020 to 2040 with the model assuming that the development of LRT will be completed in 2020. As for fare of LRT, we assume flat rate at two levels; 150 yen and 200 yen. Here, time required from each LRT station to the JR Kanazawa station is calculated assuming that each LRV run at 20 kilometers per hour in the north section of the Kanazawa station and at 12 kilometers per hour in the south section of the Kanazawa station. The estimated rates of change of population from the case without LRT ((with - without) / without) in each zone in 2040 are illustrated in Figure 5.

<150-yen fare> <200-yen fare>

Figure 5. Impact of developing LRT on population distribution in 2040 (Rate of change from the case without LRT)

Table 4 shows the aggregate results of estimated population in zones within 400 meters from LRT stations and other zones in each case (without LRT, with LRT with the 150-yen fare and with LRT with the 200-yen fare) in 2025, 2030, 2035 and 2040. These figure and Table confirm that with LRT, the population increases along the LRT line. When the fare is 150 yen, population increase will take place not only in zones along the line but also in many suburban zones, and the total amount of population increase in zones along the line is larger compared to the case of 200-yen fare.

637

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Table 4. Estimated population in zones within 400 meters from LRT stations and other zones With LRT With LRT Without LRT (150-yen fare) (200-yen fare) Zones within 400 meters 62,838 62,725 61,953 from LRT staitons (1.43%) (1.25%) 2025 399,535 399,648 Other zones 400,420 (-0.22%) (-0.19%) Zones within 400 meters 61,012 60,862 60,537 from LRT staitons (0.78%) (0.54%) 2030 386,563 386,713 Other zones 387,038 (-0.12%) (-0.08%) Zones within 400 meters 58,586 58,455 58,316 from LRT staitons (0.46%) (0.24%) 2035 370,462 370,593 Other zones 370,732 (-0.07%) (-0.04%) Zones within 400 meters 55,214 55,100 55,079 from LRT staitons (0.25%) (0.04%) 2040 348,687 348,801 Other zones 348,822 (-0.04%) (-0.01%)

3.7 Estimation of Demand of LRT based on Population Distribution Demand of developed LRT or the number of LRT users per day is estimated for citizens of Kanazawa city, domestic tourists from outside Kanazawa and foreign tourists, respectively. As for citizens of Kanazawa city, the demand is estimated using population within 20 minutes from LRT station by walk, ratio of each age group, expected frequency of use of LRT for each age group based on the results of the questionnaire we conducted. Equation (13) expresses the estimation formula of the number of LRT users who are citizens of Kanazawa city per day. Table 5 and Table 6 show frequency distribution of use of LRT with 150-yen fare and 200-yen fare in the questionnaire results respectively.

POPRNdis,, agefre ,,  f dis agefre 2 Q = dis age (13) Cf, 365 where, dis : distance group from LRT station, age : age group, fre : expected frequency of use of LRT, f : fare level, QC : the number of LRT users who are citizens of Kanazawa city per day, POP : population, Rfre : share of each expected frequency of use of LRT, and Nfre : the number of days per year for each expected frequency.

638

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Table 5. Frequency distribution of use of LRT (150-yen fare) in the questionnaire results for each age group (%) 20's and 70's and 30's 40's 50's 60's under over Almost every day 18.52 9.09 18.75 9.76 10.81 15.63 About 3-5 times a week 7.41 13.64 12.50 19.51 5.41 15.63 Within 5 About 1-2 times a week 37.04 13.64 21.88 21.95 18.92 37.50 min to the About 1-2 times a month 18.52 31.82 12.50 26.83 27.03 12.50 LRT station Several times a year 14.81 22.73 25.00 12.20 24.32 9.38 from home on foot Les than 1 time a year 3.70 0.00 0.00 0.00 2.70 3.13 No use 0.00 9.09 9.38 9.76 10.81 6.25 Total 100.00 100.00 100.00 100.00 100.00 100.00 Almost every day 16.00 9.09 11.11 8.82 11.11 3.13 About 3-5 times a week 8.00 9.09 7.41 11.76 0.00 25.00 Within 5-10 About 1-2 times a week 28.00 4.55 11.11 14.71 7.41 34.38 min to the About 1-2 times a month 28.00 13.64 18.52 26.47 25.93 9.38 LRT station Several times a year 16.00 36.36 37.04 17.65 29.63 18.75 from home on foot Les than 1 time a year 4.00 0.00 0.00 2.94 3.70 0.00 No use 0.00 27.27 14.81 17.65 22.22 9.38 Total 100.00 100.00 100.00 100.00 100.00 100.00 Almost every day 8.00 4.35 4.00 9.38 0.00 9.09 About 3-5 times a week 8.00 8.70 8.00 3.13 4.35 4.55 Within 10-15 About 1-2 times a week 4.00 0.00 4.00 3.13 0.00 36.36 min to the About 1-2 times a month 24.00 13.04 4.00 9.38 26.09 18.18 LRT station Several times a year 28.00 21.74 36.00 34.38 8.70 4.55 from home on foot Les than 1 time a year 8.00 4.35 0.00 6.25 13.04 0.00 No use 20.00 47.83 44.00 34.38 47.83 27.27 Total 100.00 100.00 100.00 100.00 100.00 100.00 Almost every day 4.00 4.55 4.00 8.82 0.00 5.00 About 3-5 times a week 0.00 9.09 4.00 2.94 0.00 5.00 Within 15-20 About 1-2 times a week 12.00 0.00 4.00 2.94 4.55 10.00 min to the About 1-2 times a month 8.00 4.55 4.00 5.88 18.18 30.00 LRT station Several times a year 24.00 18.18 20.00 23.53 9.09 10.00 from home on foot Les than 1 time a year 8.00 0.00 0.00 8.82 13.64 0.00 No use 44.00 63.64 64.00 47.06 54.55 40.00 Total 100.00 100.00 100.00 100.00 100.00 100.00

639

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

Table 6. Frequency distribution of use of LRT (200-yen fare) in the questionnaire results for each age group (%) 20's and 70's and 30's 40's 50's 60's under over Almost every day 19.23 9.09 19.23 12.82 0.00 12.00 About 3-5 times a week 3.85 13.64 11.54 5.13 15.15 4.00 Within 5 About 1-2 times a week 30.77 13.64 19.23 23.08 15.15 32.00 min to the About 1-2 times a month 23.08 27.27 3.85 30.77 33.33 28.00 LRT station Several times a year 15.38 27.27 30.77 7.69 18.18 4.00 from home on foot Les than 1 time a year 0.00 0.00 0.00 0.00 3.03 0.00 No use 7.69 9.09 15.38 20.51 15.15 20.00 Total 100.00 100.00 100.00 100.00 100.00 100.00 Almost every day 15.38 9.09 8.00 8.82 4.00 3.23 About 3-5 times a week 11.54 9.09 8.00 8.82 8.00 9.68 Within 5-10 About 1-2 times a week 19.23 4.55 16.00 8.82 0.00 35.48 min to the About 1-2 times a month 23.08 13.64 12.00 26.47 32.00 19.35 LRT station Several times a year 23.08 36.36 32.00 20.59 32.00 19.35 from home on foot Les than 1 time a year 0.00 0.00 0.00 0.00 4.00 0.00 No use 7.69 27.27 24.00 26.47 20.00 12.90 Total 100.00 100.00 100.00 100.00 100.00 100.00 Almost every day 4.00 9.09 0.00 11.43 0.00 0.00 About 3-5 times a week 8.00 4.55 8.33 0.00 3.85 4.76 Within 10-15 About 1-2 times a week 8.00 0.00 4.17 5.71 3.85 33.33 min to the About 1-2 times a month 20.00 13.64 8.33 8.57 19.23 19.05 LRT station Several times a year 32.00 18.18 16.67 31.43 19.23 4.76 from home on foot Les than 1 time a year 8.00 0.00 4.17 2.86 11.54 0.00 No use 20.00 54.55 58.33 40.00 42.31 38.10 Total 100.00 100.00 100.00 100.00 100.00 100.00 Almost every day 0.00 4.55 0.00 6.06 0.00 0.00 About 3-5 times a week 4.00 9.09 4.17 0.00 0.00 0.00 Within 15-20 About 1-2 times a week 12.00 0.00 8.33 6.06 4.35 15.00 min to the About 1-2 times a month 8.00 4.55 0.00 6.06 13.04 15.00 LRT station Several times a year 24.00 18.18 16.67 24.24 17.39 15.00 from home on foot Les than 1 time a year 4.00 0.00 0.00 9.09 13.04 0.00 No use 48.00 63.64 70.83 48.48 52.17 55.00 Total 100.00 100.00 100.00 100.00 100.00 100.00

The demand of domestic tourists is estimated using the actual average number of tourist visitors in 2015 and 2016, the actual modal share of public transport (bus) on the assumed LRT route and the actual average number of nights stayed based on the questionnaire survey conducted by Ishikawa Prefecture. The demand of foreign tourists is estimated using the expected number of tourist visitors to Japan in the future (Japan National Tourism Organization), the actual rate of visiting to Ishikawa Prefecture in 2017, the actual modal share of public transport (bus) on the assumed LRT route and the actual average number of nights stayed based on the questionnaire survey conducted by Ishikawa Prefecture. Although

640

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

it is considered that the fare level of LRT has an impact on demands of domestic and foreign tourists, in this analysis we assume that demands of domestic and foreign tourists are constant with the fare level. To estimate the change in demands of tourists in response to the fare level, it is necessary to develop the demand estimation model for tourists considering fare level and conduct a questionnaire survey for tourists in Kanazawa City. Table 7 shows the demand.

Table 7. Estimated demand of LRT (people per day) <150-yen fare> Domestic tourists Citizons of from outside Foreign tourists Total Kanazawa city Kanazawa 2020 48,850 52,045 5,160 106,055 2025 47,728 52,045 5,429 105,203 2030 46,123 52,045 6,569 104,737 2035 43,873 52,045 7,969 103,888 2040 41,358 52,045 9,369 102,773

<200-yen fare> Domestic tourists Citizons of from outside Foreign tourists Total Kanazawa city Kanazawa 2020 39,758 52,045 5,160 96,963 2025 38,565 52,045 5,429 96,039 2030 37,118 52,045 6,569 95,732 2035 35,149 52,045 7,969 95,163 2040 32,946 52,045 9,369 94,361

4. CONCLUSION

In this study, a model for estimating the impact of developing LRT in a tourist city and its fare on the future population distribution is developed and an estimation method of demand of LRT based on the estimated population distribution and a questionnaire survey for citizens is indicated. Further, an empirical model for Kanazawa, which is one of Japan’s leading tourist cities, is developed and is used to analyze the impact of developing a new LRT line on population distribution within the city and demand of the line is estimated. The results of the analysis show that the installation of LRT in Kanazawa City increases the population along the LRT, the number of users of LRT per day is expected to be about 103 thousand with 150-yen fare and about 94 thousand with 200-yen fare in 2040, etc. However, the model for estimating the impact of developing LRT we developed does not consider the modal shift from car to LRT, the interaction between road congestion and household’s location choice and change in the number of households intending to move by developing the new LRT. As for demand estimation of the new LRT in Kanazawa city, it is necessary to develop the demand estimation model for domestic and foreign tourists considering fare level and investigate the demand elasticities of fare level by conducting a questionnaire survey, etc. Developing new models and conducting the empirical analysis considering these are our future works. The future challenge also includes the cost benefit analysis of LRT using the estimated demand and the effect analysis of introduction of a feeder bus system and a park and ride system in addition to LRT development.

641

Journal of the Eastern Asia Society for Transportation Studies, Vol.13, 2019

REFERENCES

1. Endo, K. (2008). Factor analysis of employment decision of women with college degrees using the conjoint method. The public economy program of School of Public Policy, Hitotsubashi University. (in Japanese) 2. Li, T., Wu, J., Sun, H., Gao, Z. (2016). Integrated co-evolution model of land use and traffic network design. Networks and Spatial Economics, 16(2), 579-603. 3. Li, T., Sun, H., Wu, J., Ge, Y. E. (2017). Optimal toll of new highway in the equilibrium framework of heterogeneous households' residential location choice. Transportation Research Part A: Policy and Practice, 105, 123-137. 4. Morita, M., Suzuki, K., Okunuki, K. (2014). An empirical study of the ratio of the road distance to the straight line distance in major cities in Japan. Theory and Applications of GIS, 22(1), 1–7. (in Japanese) 5. Muto, S., Ueda, T., Takagi, A., Tomita, T. (2010). The benefit evaluation considering the relocating sectors with computable urban economic model, Infrastructure Planning Review, 17, 257-266. (in Japanese) 6. Railway Bureau, Ministry of land, Infrastructure, Transport and tourism., Institute for transport policy studies. (2012). Cost-effective analysis manual of rail projects 2012. (in Japanese) 7. Sato, T., Sasaki, T., Chikuma, M. (2018). Cost-benefit analysis of developing a light rail transit and feeder bus system in Utsunomiya city considering the change in population distribution, Asian Transport Studies, 5(1), 151–164. 8. Takasugi, E., Sato, T., Chikuma, M. (2018). Development of the method of measuring the benefits of developing light rail transit and bus considering differences of these systems and influence on urban population distribution and a case study for Maebashi city, Japan, Journal of the City Planning Institute of Japan, 53(3), 1341-1347. (in Japanese) 9. Tomioka, H., Morimoto, A. (2018). A Study on the population increase effect by introducing LRT using CUE model, Journal of the City Planning Institute of Japan, 53(3), 1348-1354. (in Japanese)

642