Impacts of Motorcycle Demand Management in Yangon, Myanmar

Transcription

1 Workig Paper Impacts of Motorcycle Demad Maagemet i Yago, Myamar Hiroki Iaba ad Hiroori Kato * The Uiversity of Tokyo, Hogo, Bukyo-ku, Tokyo , Japa * kato@civil.t.u-tokyo.ac.jp Abstract. This study aalyzes the potetial impacts of motorcycle demad maagemet ad its cotributio to the trasportatio market i Yago, Myamar, where motorcycles have bee baed sice A vehicle owership model with travel demad models of modal choice, destiatio choice, ad trip frequecy is estimated usig a dataset comprisig 8,289 households ad 24,373 trips i Yago, compiled by the Japa Iteratioal Cooperatio Agecy i Next, a traffic demad forecast system is developed i which a traffic assigmet model is itegrated with a vehicle owership model ad travel demad models to evaluate the impacts of the motorcycle ba. The, the expected impacts of the motorcycle ba are estimated by comparig multiple scearios for 2013 ad The results show that the ba could reduce traffic volume ad vehicle kilometers traveled by approximately 18.0% ad 26.9% i 2013, but oly 4.5% ad 6.0% i I other words, the ba sigificatly cotributes to the mitigatio of the curret urba trasportatio problems; however, it would promote car owership ad the substitutio of motorcycles i lie with icome growth, wipig out the effects of reduced motorcycle trips i the future. These fidigs suggest that developig cities should cosider the log-term dyamics of motorcycle demad maagemet. Keywords: motorcycle ba, urba travel demad, developig city, Yago 1. Itroductio Motorcycles are a key mode of urba trasportatio i may Asia developig cities (Barter, 1998; Tua et al., 2005; Koizumi et al., 2013). I particular, they have bee kow to provide flexible ad iexpesive mobility for low- ad middleicome idividuals. However, the growig umber of motorcycles has led to severe traffic cogestio ad accidets i developig cities (Pha ad Shimizu, 2011; Uy ad Regidor, 2011). To tackle these issues, various related trasportatio policies have bee proposed: motorcycle laes ad parkig as well as improvemets i motorcycle regulatios (Hug, 2006; Istitute for Trasportatio ad Developmet Policy, 2009). A potetial policy measure to regulate the umber of motorcycles is itroducig traffic demad maagemet of motorcycles (Barter, 1998; Lai, 2007), icludig a ba o motorcycles i specific areas, regulatio of motorcycle owership ad parkig, ad cordo pricig for motorcycle users. Although motorcycle demad maagemet could cotribute toward solvig traffic problems i developig cities, its impacts have ot bee sufficietly studied, possibly because few cities have implemeted policies that regulate motorcycle owership ad use (Ye et al., 2009). A exceptio, however, is Yago City, the former capital of Myamar, where motorcycle use has bee prohibited i most urba areas sice 2003 (JICA, 2014; Kojima et al., 2015a). 1

2 Workig Paper Thus, this study aalyzes the potetial impacts of motorcycle demad maagemet ad its cotributio to the trasportatio market i Yago. To do so, a traffic demad forecast system is developed usig large-scale travel data collected by the Japa Iteratioal Cooperatio Agecy (JICA) i The expected impacts of a motorcycle ba are the estimated usig a scearios aalysis, i which traffic demad is simulated with a demad forecast system. A advatage of this study is that it empirically aalyzes the potetial impacts of a motorcycle ba from the viewpoit of ot oly a modal shift but also chages i vehicle owership, trip geeratio, ad destiatio choice usig a cosistet model framework. This approach is expected to provide a uderstadig of the overall impacts of a motorcycle ba i a systematic maer, which could facilitate a holistic discussio o policy implicatios. The remaider of this paper is orgaized as follows. Sectio 2 reviews the extat literature. Sectio 3 describes the curret coditios of the urba trasportatio market i the Yago metropolita area ad presets the data used i the empirical aalysis. Sectio 4 formulates the models of a idividual s vehicle owership ad travel behavior ad presets the estimatio results of the model. Sectio 5 develops the travel demad forecast system usig the estimated models ad presets the simulatio results of the impact aalysis. Sectio 6 summarizes the fidigs ad suggestio for future research. 2. Literature Review Some studies have demostrated the importace of motorcycle owership maagemet ad use i developig cities. For example, Hug (2006) assessed traffic maagemet measures usig multiple criteria ad accordigly, listed strategies for motorcycle-depedet cities. The Istitute for Trasportatio ad Developmet Policy (2009) reviewed existig kowledge o maagemet policies for two- ad three-wheeler use ad summarized recommeded strategies. Ye et al. (2011) estimated the impacts of a motorcycle ba policy usig travel preferece data of local residets i Huizhou, Chia, ad showed that while the policy suppresses motorcycle use, it promotes the use of other trasportatio modes. Xigdog et al. (2009) studied restrictios o motorcycle use across mai urba areas i Guagzhou, Chia, sice 2007 ad reported that the ba decreased the use of motorcycles ad icreased that of public trasportatio, bicycles, ad cars. These studies have highlighted the impact of a motorcycle ba maily o a idividual s modal choice, such as a shift from the use of motorcycles to other trasportatio modes. However, we expect that such a regulatio also iflueces a idividual s trip geeratio ad/or vehicle owership, which could sigificatly affect urba traffic demad. I particular, vehicle owership is strogly coected with its usage i developig cities, where provisios of public trasportatio are lackig (Dissaayake et al., 2001). Thus, this study ivestigates the potetial impacts of a motorcycle ba with a itegrated travel demad model that icorporates trip geeratio, vehicle owership, ad modal choice i Yago City. Although urba trasportatio is oe of the most critical issues i Yago City, it has bee rarely studied maily because of poor data availability. A exceptio is Zhag et al. (2008), who durig Myamar s closed-market era examied the potetial impacts of itroducig a ew public trasit o idividual s behavior usig stated-preferece data i Yago. Similarly, Kato et al. (2010a) aalyzed the route choice behavior of bus commuters usig stated-preferece data for Yago, while Kato et al. (2010b) reported o the city s urba bus system, icludig the regulatory framework ad cost structure of bus operators. Kato et al. (2011) further studied the potetial impacts of itroducig a bus rapid trasit system i Yago by coductig a cost-beefit aalysis usig urba travel demad forecast models. Sice the ecoomic liberalizatio of Myamar i 2012, a icreasig umber of studies have highlighted urba trasportatio problems i Yago. For istace, the Japa Iteratioal Cooperatio Agecy (JICA, 2014) admiistered large-scale travel surveys, icludig the persoal 2

3 Workig Paper trip (PT) survey i Yago metropolita area, ad proposed mid- ad log-term urba trasportatio ivestmet plas. Kojima et al. (2015b) reported o the urba trasportatio market usig data collected by JICA. Tu et al. (2015) aalyzed local car owership usig a ordered logit model with local data for Yago City. Although studies have reported o the uiqueess of motorcycle regulatios i Yago City, to the best of our kowledge, o study aalyzes its impact o urba trasportatio demad i the city. 3. Scope of Study ad Data 3.1. Scope Fig. 1 (a) illustrates the study area, which icludes Yago City (33 towships) ad a part of six adjacet towships: Thalyi, Hmawbi, Helgu, Htatabi, Twatay, ad Kyaukta. The total area is approximately 1,500 km 2 ad the total populatio is approximately 5.7 millio as of The cetral busiess district (CBD) of Yago City is located to the souther part of the city, which is adjacet to where the Yago River ad Bago River meet. As the urba populatio icreased, the urba area was exteded from the CBD to the orth. Accordigly, trasportatio ifrastructure, icludig highways ad rail etworks, were developed to coect the CBD with the orther part of Yago City. The mai residetial areas are located i the orth-easter part of the city, from which may commuters travel to their workplaces i the CBD (Kojima et al., 2015b). Curretly, a vast majority of urba trasporters i Yago rely o the bus system (Zhag, 2008; Kato, 2010a). A factor drivig the icrease i the modal share of buses is the motorcycle ba itroduced i 2003 (31 towships ad areas to the south of the CBD, amely Dala ad Seik Gyi Ka Naug Toe, were ot icluded) to reduce the umber of traffic accidets (Myamar Times, 2014). Fig. 1 (b) shows the motorcycle ba area, that is, where the use motorcycles is ot permitted. Source: JICA, 2014 Source: Kojima et al., 2015 Fig. 1. (a) Study Area i Yago. Fig. 1. (b) Motorcycle ad Bicycle Ba Areas i Yago. 3

4 Workig Paper Note that bicycle use is also prohibited i the CBD. The pealty for violatig the motorcycle ba is 20,000 Myamar Kyats (MMK) (approximately US $20) ad that for ot wearig a helmet is 10,000 MMK (approximately US $10) as of 2014 (Myamar Times, 2014). The motorcycle ba has bee cosidered to cotribute to sigificatly reducig motorcycle owership i Yago (Kojima et al., 2015a). Kojima et al. (2015a) also reported that the mobility of Yago s citizes does ot largely deped o motorcycles ad bicycles. First, this is because the surroudig areas are well coected with the CBD by public trasportatio ad secod, lower icome people who reside to the south of the CBD across the Yago River use ferries to access the district Data This study uses a dataset compiled i the Project for Comprehesive Urba Trasport Pla of Greater Yago (YUTRA) project (JICA, 2014) orgaized by JICA i The YUTRA project team, which icludes oe of the authors, implemeted eleve surveys iteded to capture the travel patters of local people ad the level of trasportatio services such as travel speed, traffic volume by trasportatio mode, ad socioecoomic attributes of trip makers (e.g., age, geder, occupatio, car owership, ad icome). Oe of the eleve surveys is the PT survey iterviewed 44,988 idividuals from 11,286 households i the 39 towships i Yago from February to August The respodets were requested to maitai a daily travel diary, metio the socioecoomic ad demographic characteristics of their households, ad state opiios about trasportatio policies. Details of the PT survey are available i JICA (2014) ad Kojima et al. (2015a). Table 1 presets the descriptive statistics of household vehicle owership by household mothly icome ad locatio of settlemet, that is, iside or outside the motorcycle ba area. Note that this dataset was costructed usig a radom resamplig process from the origial dataset where the sample rate is cotrolled to accout for % of the total households i each towship as of This is because the sample rates of the origial dataset greatly vary amog towships. As a result, 8,289 households comprisig 7,356 households withi ad 933 households outside the motorcycle ba area were obtaied. First, this shows that motorcycle owers ca be see across icome levels, although the motorcycle owership rate margially icreases with icome level, especially outside the ba area. This idicates that motorcycles are popular as a iexpesive trasportatio mode eve i Yago. Secod, the motorcycle owership rate withi the ba area is sigificatly lower tha that outside it. Although this is reasoable, % households withi the ba area may still ow motorcycles because idividuals livig alog the border ca ride their motorcycles outside it. Third, the car owership rate icreases with icome level. Fially, the car owership rate iside the ba area is higher tha that outside it, except amog the lowest icome subgroup (Table 1). This is possibly because car owership is promoted iside the ba area as a substitute for motorcycles. Table 2 shows the descriptive statistics of respodets modal share by vehicle owership ad trip origi, that is, iside ad outside the ba area. Here as well, this dataset was costructed usig a radom re-samplig process of the origial dataset, i which the sample rate is cotrolled to accout for % out of the total households i each towship as of I additio, o-motorized trasportatio modes are excluded from the dataset because they are maily used for itra-zoe short travels. Thus, 24,373 trips cosistig of 22,591 trips from iside ad 1,782 trips from outside the ba area were obtaied. 4

5 Workig Paper Table 1. Descriptive statistics of household vehicle owership by settlemet locatio ad household icome. Settlemet area Iside the ba area Outside the ba area Household mothly Car & Motorcycl Car icome Motorcycle e Noe Total kyat/moth Obs. % Obs. % Obs. % Obs. % Obs. 100,000 or lower , , , , , , , , , , , , , ,001 or higher Total , , ,000 or lower , , , , , , , , ,001 or higher Total First, it shows that car use is promoted by car owership. Car owers iside the ba area use their vehicles more tha those outside it. This is probably because the road ifrastructure outside the ba area is so poorly maitaied that car owers have lower icetive to use their vehicles. Secod, motorcycle use is promoted by motorcycle owership (Table 2). Motorcycle owers iside the ba area use motorcycles less tha those outside it simply because it is prohibited. Third, the modal share of car or motorcycle owers i bus use is lower tha the share of those who do ot ow vehicles. I other words, car or motorcycle owership suppresses public trasportatio demad. Fially, the modal share of car owers i taxi use is higher tha the share of those who do ot ow a vehicle both iside ad outside the ba area. This is probably because the household icome of car owers is sufficietly high for them to afford taxi charges. Table 2. Descriptive statistics of modal share by trip origi ad vehicle owership. Trip origi area Vehicle owership Car Motorcycle Bus Taxi Total Obs. % Obs. % Obs. % Obs. % Obs. Car ower 1, , ,531 Iside the ba area Motorcycle ower , ,896 No vehicle ower , , ,467 Total 2, , , ,894 Car ower Outside the ba area Motorcycle ower No vehicle ower Total , , Models 4.1. Vehicle owership model The vehicle owership model is formulated as a discrete choice model (Be-Akiva ad Lerma, 1985) i which a idividual is assumed to choose a optio maximizig his/her utility fuctio from a choice set. The model is specified to be a two-level ested logit model (Louviere et al., 2000): the upper level of the choice set cosists of car or o-car ad the 5

6 Workig Paper lower level comprises Car&Motorcycle or Car uder the car owership optio ad Motorcycle or No Motorized Vehicle uder the o-car optio. The utility fuctios of each optio are assumed to be o V Va, Vb, Vab,, (1) where, is the utility fuctio of household uder the coditio that the household ows a vehicle o (Car&Motorcycle, Car, Motorcycle, or Noe)., is a partial utility fuctio of household if the household chooses a upper level optio a,, is the partial utility fuctio of household if the household chooses a lower-level optio b, ad, is the partial utility fuctio of household determied by the upper ad lower levels of optios a ad b. The probability of choosig a vehicle owership status o for household i the ested logit is expressed as P o b exp V b, exp V V b, ab, V ab, a exp V a, exp V a, a, a,, (2) where Λ, is a log-sum variable defied as, l exp V, V, (3) a b b ab ad is a coefficiet related to the log-sum variable. Ukow coefficiets are estimated usig the full-iformatio maximizatio likelihood method (Louviere et al., 2000) Modal choice model The modal choice model is formulated as the discrete choice model i which a idividual is assumed to choose a optio maximizig his/her utility fuctio from a modal choice set. The idividual s choice set is assumed to cosist of car, motorcycles, buses, ad taxis. It is specified as a stadard logit model (Be-Akiva ad Lerma, 1985) i which the utility fuctio is expressed as V, GC, V, ijm, (4) GC ijm k k mk where, is the utility fuctio of idividual uder the coditio that he/she chooses a trasportatio mode m from origi i to destiatio j., is the geeralized travel cost of trasportatio mode m,, is the kth explaatory variable, ad β is the ukow coefficiet. The, the probability of choosig a trasportatio mode m for idividual who ows o from origi i to destiatio j is 6

7 Workig Paper P ijm, m exp V ijm, exp V ijm,. (5). The ukow coefficiets are estimated by maximizig the likelihood fuctio derived from equatio (5) Destiatio choice model A destiatio choice model is formulated as the discrete choice model i which a idividual is assumed to choose a optio maximizig his/her utility fuctio from a destiatio choice set. It is specified as a stadard logit model icorporatig the log-sum fuctio derived from the modal choice model. The utility fuctio is assumed to be, (6) V ij, lvijl, ij, l where, is the utility fuctio of idividual uder the coditio that he/she travels from origi i to destiatio j,, is the lth explaatory variable, γ is the ukow coefficiet, ad Λ, is a log-sum variable from origi i to destiatio j, which is expressed as, exp ij V ijm, m l. (7). The probability of choosig destiatio j for idividual from origi i is expressed as P ij, ij exp V ij, exp V ij,. (8). The ukow coefficiets are estimated by maximizig the likelihood fuctio derived from equatio (8). The log-sum variable of equatio (7) is formulated usig the utility fuctio of the modal choice model with the estimated coefficiets. Thus, the estimatio process of modal choice ad destiatio choice models ca be regarded the sequetial estimatio of the ested logit model (Hesher, 1986) Trip frequecy model A trip frequecy model is formulated as a aggregated liear regressio model that explais the daily trip frequecy geerated from origi i. Oe of the explaatory variables is the log-sum variable derived from the destiatio choice model. As the log-sum variable represets the expected maximum utility level or expected idirect utility (Williams, 2002), the iduced demad ca be icorporated ito the trip frequecy. This is expressed as 7

8 Workig Paper F i, qviq, i, q, (9) where, is the daily trip frequecy of idividual travelig from origi i,, is the qth explaatory, is the ukow coefficiet, ad Λ, is the log-sum variable of origi i, which is expressed as i, l exp V ij,. (10). j The ukow coefficiets are estimated usig the ordiary least squared method Model estimatio First, the study area of 39 towships was divided ito 156 traffic aalysis zoes (TAZs) ad a represetative cetroid is assumed i each TAZ. Next, a trasportatio etwork i Yago is assumed, icludig major roads, secodary roads, selected local roads, ad rail etworks, with 1,862 odes ad 2,314 liks. The, sociodemographic ad socioecoomic data are prepared for each TAZ ad travel time ad travel cost data are prepared o the basis of the trasportatio etwork. The ukow coefficiets i the vehicle owership, modal choice, destiatio choice, ad trip frequecy models are estimated usig the dataset. The explaatory variables i each model are selected usig a trial-ad-error process to achieve the best fitted results from a statistical viewpoit. Table 3 shows the estimatio results of the vehicle owership model. HH mothly icome deotes household mothly icome (kyat/moth). MC ba area equals 1 if a household resides withi the ba area ad 0 otherwise. Over five-year stay i MC ba area takes the value of 1 if a household has resided i the motorcycle ba area for more tha five years ad 0 otherwise. Well-maitaied road ifrastructure area equals 1 if a household resides i a towship where road ifrastructure is well maitaied ad 0 otherwise. Poor bus service area equals 1 if a household resides i a towship where urba bus services are poorly provided ad 0 otherwise. The estimatio results show that all coefficiets are statistically sigificat. The model has theoretical cosistecy ad high goodess of fit: the coefficiet of a log-sum variable is betwee 0 ad 1, while McFadde is sufficietly high. HH mothly icome positively iflueces the utility of Car&Motorcycle, Car, ad Motorcycle, give that higher icome households ca afford to purchase vehicles. The coefficiet of l (HH mothly icome/100,000) is the highest for Car&Motorcycle, followed by Car ad Motorcycle because a car is more expesive tha a motorcycle. MC ba area x l (HH mothly icome/100,000) is sigificatly egative for Car&Motorcycle. This possibly meas that a household is less motivated to ow both a car ad motorcycle if it is iside the ba area, which is reasoable give the ba o motorcycle use i the area. I the case of Motorcycle, over five-year stay i MC ba area x l (HH mothly icome/100,000) is sigificatly egative, while over five-year stay i MC ba area is sigificatly positive. Sice more tha 85% of households have mothly icome of over 100,000 kyats per moth (Table 1), over five-year stay i MC ba area egatively iflueces mid- ad high-icome households motorcycle utility. This suggests that motorcycle owership is suppressed iside the ba area. Poor bus service area is sigificatly egative for Car&Motorcycle ad Motorcycle, while wellmaitaied road ifrastructure is sigificatly positive for Car, which is reasoable. 8

9 Workig Paper Table 3. Estimatio results of vehicle owership model. Variable Optio Coefficiet t-statistic l (HH mothly icome/100,000) Car&Motorcycle 4.458** 5.41 l (HH mothly icome/100,000) Car 3.725** 4.38 l (HH mothly icome/100,000) Motorcycle ** 6.38 MC ba area x l (HH mothly icome/100,000) Car& Motorcycle * 2.05 Over five-year stay i the MC ba area x l (HH mothly icome/100,000) Motorcycle 0.438** 2.98 Poor bus service area Car& Motorcycle 0.721* 2.55 Poor bus service area Motorcycle 1.525** Over five-year stay i the MC ba area Car 0.829* 2.34 Over five-year stay i the MC ba area Motorcycle 0.471** 8.98 Well-maitaied road ifrastructure area Car 2.493** 4.80 Log-sum All 0.372** 4.49 Costat Car& Motorcycle ** 4.42 Costat Car ** 3.54 Costat Motorcycle 3.946** 5.61 Number of observatios 8289 Iitial log-likelihood Fial log-likelihood Adjusted ρ Note: ** ad * mea p < 0.01 ad p < Table 4 presets the estimatio results of the modal choice model. Car owership equals 1 if a idividual s household ows a private car ad 0 otherwise. Motorcycle owership equals 1 if a household ows a motorcycle ad 0 otherwise. MC ba area equals 1 if a trip starts from a TAZ, which is located iside the ba area ad 0 otherwise. CBD takes the value of 1 if a trip starts from the CBD ad 0 otherwise. Periphery area takes the value of 1 if a trip starts from a periphery area ad 0 otherwise. Geeralized cost is defied as moetary cost plus time cost estimated usig the value of time ad travel time. The value of time is estimated from the average wage rate. The results show that all coefficiets are statistically sigificat ad McFadde s is sufficietly high. Car owership is sigificatly positive for car use, while MC owership is positive for motorcycle use. I other words, car ad motorcycle owers prefer usig their vehicles to travel. MC ba area is sigificatly egative for motorcycle use, which meas idividuals startig their trip from withi the ba area are less likely to use a motorcycle. CBD is sigificatly positive for taxi use, whereas Periphery area is sigificatly egative for taxi use. This is because a trip startig from the CBD is maily a busiess oe i which a idividual has higher willigess to pay for travel, whereas a trip travelig from peripheral areas is maily a o-busiess trip. Geeralized cost is sigificatly egative, which is reasoable. Table 5 shows the estimatio results of the destiatio choice model. Daytime populatio is the populatio durig the daytime. CBD takes the value of 1 if the destiatio TAZ is the CBD district ad 0 otherwise. Major road equals 1 if the destiatio TAZ is located alog major roads ad 0 otherwise. Riverside takes equals 1 if the destiatio TAZ is located ext to major rivers ad 0 otherwise. Outside rivers equals 1 if the destiatio TAZ is located beyod the major rivers from 9

10 Workig Paper the CBD ad 0 otherwise. Periphery area takes the value of 1 if the destiatio TAZ is categorized as a periphery area ad 0 otherwise. The results show that all coefficiets are sigificat i t-value. The coefficiet of the log-sum variable is betwee 0 ad 1, while McFadde s is large eough. l (Daytime populatio) is sigificatly positive because the daytime populatio maily represets travelers commutig to workplaces. CBD is sigificatly positive because most offices ad commercial facilities are located i the CBD. Major road is sigificatly positive because the areas alog major roads have better accessibility while outside rivers is sigificatly egative owig to poorer accessibility. Riverside ad Periphery area are sigificatly egative because these areas have fewer offices or commercial facilities. Table 4. Estimatio results of modal choice model. Variable Optio Coefficiet t-statistic Car owership Car 2.880** Motorcycle owership Motorcycle ** MC ba area Motorcycle 1.950** CBD Taxi 0.403** 8.64 Periphery area Taxi 0.918** Geeralized cost All ** Costat Car 1.743** Costat Motorcycle 2.578** Costat Taxi 0.364** 7.37 Number of observatios Iitial log-likelihood Fial log-likelihood Adjusted ρ Note: ** ad * mea p < 0.01 ad p < Table 5. Estimatio results of destiatio choice model. Variable Coefficiet t-statistics l (Daytime populatio) 0.515** CBD 0.510** 9.30 Major road 0.823** Riverside 0.815** 9.09 Outside rivers 0.721** 3.24 Periphery area 0.759** 9.77 Log-sum 0.801** Number of observatios Iitial log-likelihood Fial log-likelihood Adjusted ρ Note: ** ad * mea p < 0.01 ad p <

11 Workig Paper Fially, Table 6 presets the estimatio results for the trip frequecy model. Daytime populatio desity is defied as the populatio per hector durig the daytime. CBD takes the value of 1 if the TAZ is located i the CBD ad 0 otherwise. Major road equals 1 if the TAZ is located alog major roads ad 0 otherwise. Riverside equals 1 if the TAZ is located ext to major rivers ad 0 otherwise. Outside rivers equals 1 if the TAZ is located beyod the major rivers from the CBD ad 0 otherwise. Periphery area takes the value of 1 if the TAZ is categorized as a periphery area ad 0 otherwise. All coefficiets are sigificat i t-value ad is high eough. All explaatory variables sigs are reasoable. Table 6. Estimatio results of trip frequecy model. Variable Coefficiet t-statistics l (Daytime populatio desity) 0.091** 6.79 CBD 0.571** 6.44 Major road 0.235** 2.89 Riverside 0.240** 2.64 Outside rivers 0.364** 4.32 Periphery area * 2.15 Log-sum 0.174** 3.35 Costat * 2.41 Number of observatios 155 Adjusted R Note: ** ad * mea p < 0.01 ad p < Sceario Aalysis 5.1. Travel Demad Forecast System A travel demad forecast system is developed usig the estimated models with a traffic assigmet i the road ad public trasportatio etwork to estimate the travel demad uder give scearios. Fig. 2 illustrates the structure of the travel demad forecast system. The vehicle owership i a zoe is computed usig the vehicle owership model uder give socioecoomic ad sociodemographic variables, while a O-D matrix by travel mode is calculated usig the trip frequecy, destiatio, ad modal choice models with socioecoomic ad sociodemographic variables, which are give by sceario. The O-D matrix by travel mode ad vehicle owership is estimated as o Pi Fi Pij Pijm,, o T ijm, I i,, (11) where is a set of idividuals who resides i zoe i. The geeralized cost iserted i the modal choice model is computed through the traffic assigmet i the road etwork because the traffic cogestio of a road lik iflueces travel time. The road traffic volume is computed o the basis of the user equilibrium priciple whe the O-D matrix by mode is give. The followig BPR fuctio is assumed for the lik performace fuctio i the user equilibrium assigmet: 11

12 Workig Paper Scearios SE variables Vehicle Owership Model Vehicle owership i zoe i: Populatio of zoe i: Trip Frequecy Model Trip geeratio from zoe i: Log-sum: Destiatio Choice Model Log-sum: OD matrix: OD matrix by vehicle owership: Modal Choice Model Geeralized cost: OD matrix by mode by vehicle owership: OD matrix by mode: Network data Travel time: Traffic Assigmet (UE) Lik flow: Fig. 2. Structure of Travel Demad Forecast System x t a xa ta 1 C a a 0, (12) where,, a,, ad, respectively, are the travel time, free flow time, traffic volume, ad capacity of lik a, ad ad are parameters. α ad are assumed to be 3.0 ad 4.0 followig JICA (2014). As show i Fig. 2, our travel demad forecast system requires a iterative computatio process. The covergece criterio is assumed as follows: a x g a x x g a g1 a 0.05, (13) where is the traffic volume of lik a of the gth iteratio. Table 7 shows the observed modal share versus coverged modal share estimated usig the travel demad forecast system, while Fig. 3 shows the observed O-D traffic volume versus coverged O-D traffic volume estimated with the travel demad forecast system. Both show that the estimated traffic demad through the travel demad forecast system is well reproduced. Table 7. Observed modal share versus estimated modal share. Car Motorcycle Bus Taxi Observed modal share (%) Estimated modal share (%)

13 Workig Paper 5.2. Scearios Aalysis The impact of the motorcycle ba o Yago s trasportatio market is simulated usig the developed travel demad forecast model. Four scearios are set up for the aalysis. I Sceario 1, the ba cotiues to exist as of 2013, which represets the curret coditio. I Sceario 2, the ba is lifted i I Sceario 3, the ba cotiues to exist i 2035, Fig. 3. Observed OD versus estimated OD traffic volume the target year i JICA s (2014) log-term urba trasportatio master pla for Yago City. Fially, i Sceario 4, the ba is lifted i The expected chages i the sociodemographic ad socioecoomic coditios ad trasportatio etwork are icorporated ito Scearios 3 ad 4 followig the log-term master pla (JICA, 2014). First, future populatio is projected i each muicipality ad the divided ito TAZs usig the curret share of populatio i each muicipality. Secod, the trasportatio etwork, which is exteded from the curret etwork as of 2035 ad icludes 2,197 odes ad 3,056 liks, is assumed. This extesio of trasportatio etwork icludes the itroductio of two urba mass trasit (UMRT) lies, te bus rapid trasit (BRT) lies, urba expressways, ad tuels or bridges crossig the rivers. Third, the household icome i 2035 is assumed to be 3.96 times higher tha that i 2013, reflectig the projected average icome growth i Yago. Fourth, the value of time i 2035 is also assumed to be 3.96 times higher tha that i Fifth, the bus service is assumed to improve such that the poor bus service area i the vehicle owership model is zero i Fially, for aalytical simplicity, the fare of public trasport icludig bus, UMRT, ad BRT is assumed to follow the fare table of the curret bus system, while the travel time of public trasport is assumed to follow the travel time of road traffic. To simulate the traffic demad without the ba i Scearios 2 ad 4, MC ba area ad over five-year stay i the MC ba area i the vehicle owership model are assumed to be zero. The MC ba area i the modal choice model is also assumed to be zero for all households ad idividuals. Note that the traffic volume of vehicles other tha cars, motorcycles, buses, ad taxis (e.g., trucks) is also cosidered i the traffic assigmet give the possible ifluece o traffic cogestio ad travel time. Although the traffic volume of other vehicles is assumed to grow i lie with ecoomic growth i Scearios 2 ad 4, for aalytical simplicity, it is assumed that whether the motorcycle ba area is itroduced does ot affect the traffic volume of other vehicles. Table 8 shows the results of vehicle owership by type of vehicle estimated i the four scearios. Table 9 shows the traffic volume ad vehicle kilometers traveled usig the said travel modes for the four scearios. Figs. 4 ad 5 show traffic volume ad traffic cogestio i the road etwork with ad without the ba area for 2013 ad Table 8 shows that, i 13

14 Workig Paper 2013, car (Car&Motorcycle ad Car) ad motorcycle owers (Car&Motorcycle ad Motorcycle) accout for 11.5% ad 11.1% i Sceario 1 but 9.3% ad 25.2% i Sceario 2. It also shows that, i 2035, car (Car&Motorcycle ad Car) ad motorcycle owers (Car&Motorcycle ad Motorcycle) accout for 34.4% ad 12.5% i Sceario 3 ad 31.2% ad 33.3% i Sceario 4. This meas that the ba promotes car owership but suppresses motorcycle owership. Table 9 shows that, i 2013, the total traffic volume based o the persoal car uit i the ba area is 657,483 i Sceario 1, while that outside the ba area is 802,291 i Sceario 2, which meas the ba suppresses total traffic volume by 18.0%. O the other had, i 2035, the total traffic volume i the ba area is 2,117,428 i Sceario 3, while that without the ba is 2,218,215 i Sceario 4. I other words, the motorcycle ba suppresses total traffic volume by 4.5%, despite the sharp reductio i motorcycle traffic volume by 74.7%. The impact of the ba is smaller i 2035 because by the, car traffic volume has sigificatly icreased compared to that i Table 9 also shows that, i 2013, the total vehicle kilometers traveled with the ba is 5,022,697 km i Sceario 1, while that without the ba is 6,873,811 i Sceario 2. I 2035, the total vehicle kilometers traveled with the ba is 21,186,150 km i Sceario 3, while that without it is 22,530,332 km i Sceario 4. This meas the motorcycle ba reduces the vehicle kilometers traveled by approximately 26.9% i 2013 ad 6.0% i 2035, which suggests that the ba reduces vehicle kilometers traveled i 2013 but loses its effectiveess i Figs. 4 ad 5 show chages i traffic volume from 2013 to Lik traffic flows sigificatly icrease i wider areas with cogested traffic by 2035 give the growth of populatio ad icome. They also show that the rig road liks have greater traffic volume i 2035 tha 2013 maily because of idustrial developmet ad rapid populatio growth i suburba areas. Table 8. Vehicle owership estimated i the four scearios. Sceario (%) Car & Motorcycle Car Motorcycle Noe Sceario 1 (with the ba area i 2013) Sceario 2 (without the ba area i 2013) Sceario 3 (with the ba area i 2035) Sceario 4 (without the ba area i 2035) Table 9. Traffic volume ad vehicle kilometers traveled by trasportatio mode i the four scearios. Sceario Car Motorcycle Bus Taxi Total Sceario 1 Traffic volume (PCU) 224,164 40, , , ,483 Vehicle km 1,798, ,474 1,755, ,972 5,022,697 Sceario 2 Traffic volume (PCU) 202, , , , ,291 Vehicle km 1,696,048 2,428,442 1,849, ,553 6,873,811 Sceario 3 Traffic volume (PCU) 1,210, , , ,689 2,117,428 Vehicle km 12,126,862 1,687,358 3,096,052 4,275,878 21,186,150 Sceario 4 Traffic volume (PCU) 1,035, , , ,372 2,218,215 Vehicle km 10,275,523 5,265,165 2,976,146 4,013,498 22,530,332 Note 1: PCU is passeger car uit. Note 2: Bus i Scearios 3 ad 4 iclude covetioal bus ad bus rapid trasit. 14

15 Workig Paper Fig. 4 shows that the average volume or capacity with the ba is 0.31 i Sceario 1, while that without the ba is 0.38 i Sceario 2. Fig. 5 also shows that the average volume or capacity with the ba is 0.61, while that without it is This meas that the ba cotributes toward reducig traffic cogestio. However, the distributio of volume or capacity i Sceario 2 seems to largely differ from that i Sceario 1, while that i Sceario 4 appears similar to that i Sceario 3. Fig. 4 (a) ad (b) show that the traffic cogestio i Sceario 1 sigificatly varies from that i Sceario 2, particularly i the road etwork of the CBD. Thus, although the traffic cogestio i the CBD is mitigated by the ba regulatio i 2013, it may ot be as effective i doig so i Discussio The results showed that the impacts of the ba o the reductio of total traffic volume ad vehicle kilometers traveled are sigificat i 2013, although they are less sigificat i This suggests that the effects of the motorcycle ba could weake i the future, possibly because household icome gradually could icrease with ecoomic growth i the city, eablig the local people to purchase cars rather tha motorcycles. Thus, potetial motorcycle users shift to car users oce they ca afford to do so. Fig. 4. (a) Traffic volume ad traffic cogestio i the road etwork with the motorcycle ba i 2013 (Sceario 1) Fig. 5. (a) Traffic volume ad traffic cogestio i the road etwork with the motorcycle ba i 2035 (Sceario 3) 15

16 Workig Paper I other words, policies must accout for log-term traffic demad maagemet i developig cities. Dyamic policy implemetatio is a key issue, particularly i developig coutries, as poited out by Acharya (2005); however, the dyamics of a traffic maagemet policy have bee rarely discussed. A possible reaso is that studies discussig traffic maagemet policies ted to focus o developed cities, where ecoomic developmet is rather stable. I additio, JICA (2014) preseted the short-, mid-, ad log-term urba trasportatio plas, i which differet traffic maagemet policies were proposed for the Yago metropolita area. However, the short- ad mid-term trasportatio plas highlighted car parkig maagemet rather tha the curret motorcycle ba regulatio, reflectig the traffic capacity costraits caused by o-street parkig i the dowtow area of Yago. Our aalysis reveals that, i the ear future, the motorcycle ba policy should be replaced by other trasportatio policy measures such as car bas or cogestio charges for both motorcycles ad cars. I additio, a expaded trasportatio etwork ad improved public trasit system should be highlighted, as poited out by Morichi (2005). Oe of the reasos uderlyig this shift from motorcycles to cars is the difficulty public trasportatio faces i attaiig sufficiet ridership owig to its poor service quality. Although idividuals may have to opt for the bus or railways i the early stage of developmet, they are expected to gradually switch their travel mode from public to persoal trasportatios (e.g., cars). Additioally cosistecy of urba trasportatio policy with other related strategies may be also required. Particularly, the car import strategy should be harmoized with the motorcycle maagemet policy. The car owership would icrease sigificatly i accordace with the icome growth if the car import would be iappropriately deregulated. Oe of the challeges i Yago towards the cosistet policy developmet is a fragmeted orgaizatioal system relatig the urba trasportatio issues. For istace, curretly, the Yago Regioal Govermet is resposible for the motorcycle ba while the Yago City Developmet Committee is resposible for other urba trasportatio policies. To pursue the itegrated trasportatio policy, poor istitutioal structure should be properly improved, such as the establishmet of a ew orgaizatio that hadles comprehesively the Yago s urba trasportatio. Aother issue i the dyamic urba trasportatio strategy is cosesus buildig amog stakeholders to chage the existig policy. It is widely kow that a ewly itroduced traffic demad maagemet ofte faces strog oppositios from locals, tough debates amog stakeholders, ad varyig attitudes from differet groups (Schlag ad Teubel, 1997; Sdhlag ad Schade, 2003; Whittles, 2003; Wislott-Hiselius et al., 2009; Xiao ad Lu, 2013). This is also the case i developig cities (Mahedra, 2011). Similar challeges are expected i the removal or replacemet of the curret motorcycle ba. I the PT survey, respodets were asked if they thought the curret limitatio o the use of motorcycles i Yago should cotiue, to which 58% respodets supported the regulatio that could possibly impair citize s coveiece; i particular, the residets i the CBD ad surroudig areas favored the ba (Kojima et al., 2015b). This implies that people recogize that the motorcycle ad bicycle ba improves traffic safety i the CBD. Uder this coditio, it may be difficult for policy makers to chage the existig regulatio. Thus, a careful moitorig of vehicle owership ad urba traffic demad is required to make a decisio regardig the removal of the motorcycle ba ad to do as at the most appropriate time so that the decisio is accepted by the public. 6. Coclusio This study aalyzes the potetial impacts of motorcycle demad maagemet ad its cotributio to the trasportatio market i Yago, where motorcycle use has bee prohibited sice To do so, a traffic demad forecast system is 16

17 Workig Paper developed, where the vehicle owership model ad travel demad models are estimated usig large-scale data collected i Yago. The expected impacts of the ba are aalyzed for both 2013 ad 2035 usig a scearios aalysis, i which the traffic demad is simulated usig a demad forecast system. The results imply that the motorcycle ba policy cotributes to the mitigatio of urba trasportatio problems i the early stage of ecoomic developmet; however, i the mid ad log term, it would promote car owership or use, reducig or eve elimiatig the positive impacts of reduced motorcycle traffic. This suggests a strog requiremet to develop log-term dyamic trasportatio strategies, particularly i developig cities. As show i this study, the effects of motorcycle demad maagemet is expected to dyamically chage i lie with icome growth. Thus, the motorcycle ba regulatio should be well maaged i cojuctio with additioal trasportatio policies or should be replaced by other trasportatio policies such as a additioal traffic maagemet of cars ad/or a improved public trasportatio Numerous issues eed to be addressed i future research. First, this study uses cross-sectioal data, makig it difficult to cofirm the reliability of future trasportatio demad forecast. Studies could examie whether idividuals or households really chage their decisios regardig their travel behavior i a static maer. For istace, i the cotext of our study, those who resided i the area with the ba for several years are likely to have become accustomed to daily travel patters i which o motorcycles are used. I this case, eve if the ba was lifted, idividuals attitudes or prefereces may ot sigificatly icrease motorcycle traffic volume, as show i our sceario aalysis. O the other had, motorcycle users may cotiue to use motorcycles, despite potetially aggressive improvemets i public trasport (Tua, 2014). Next, a combiatio of multiple policies should be discussed, such as the motorcycle ba as well as public trasit ivestmet. May studies have highlighted itegrated trasportatio strategies, such as May ad Roberts (1995) ad Hull (2005, 2008). As discussed by Morichi ad Acharya (2012), such itegrated trasportatio policies are critical to overcome traffic problems i developig cities. I additio, sice the timig of itroducig a mass trasit system should be harmoized with traffic demad maagemet, the time schedule of log-term urba trasportatio strategies should be carefully aalyzed. Fially, from the log-term perspective, the iteractio of urba trasportatio system with lad-use patters ad/or idustrial structure should be examied. Ackowledgemets This study is supported by Dr. Pha Le Bih (Japa Iteratioal Cooperatio Agecy), Mr. Takashi Shoyama (Almec Co.), Mr. Makoto Okamura (Almec Co.), ad Ms. Momoko Kojima (Almec Co.), who collected data ad arraged local iterviews. The Miistry of Trasportatio ad Commuicatio Departmet, Myamar; Yago City Developmet Committee; Yago Road Trasport Admiistratio Departmet; ad Yago City Traffic Policy provided us with useful iformatio. Mr. Ryosuke Abe (The Uiversity of Tokyo) supported the techical aspects of model developmet. We deeply thak them for their kid support. Refereces Acharya, S. R., Motorizatio ad Urba Mobility i Developig Coutries Explorig Policy Optios through Dyamic Simulatio. Joural of the Easter Asia Society for Trasportatio Studies 6, Barter, P. A., A Iteratioal Comparative Perspective o Urba Trasport ad Urba Form i Pacific Asia: the Challege of Rapid Motorizatio i Dese Cities. Doctoral Dissertatio, Murdoch Uiversity. 17

18 Workig Paper Be-Akiva, M., Lerma, S., Discrete Choice Aalysis: Theory ad Applicatio to Travel Demad, MIT Press, Cambridge. Deg, X., Xu, J., Wag, B., Traffic Coutermeasures Research for Guagzhou City i Traffic Mode Trasferrig Period after Motorcycle Forbidde Ba Effect. Joural of Trasportatio Systems Egieerig ad Iformatio Techology 9(4), Dissaayake, D., Morikawa, T., Trasport Policy Aalysis for Developig Coutries usig Nested Logit Model of Vehicle Usage, Mode Choice ad Trip-chai. Joural of the Easter Asia Society for Trasportatio Studies 4, Hesher, D. A., Sequetial ad Full Iformatio Maximum Likelihood Estimatio of a Nested Logit Model. The Review of Ecoomics ad Statistics 68(4), Hull, A., Itegrated Trasport Plaig i the UK: From Cocept to Reality. Joural of Trasport Geography 13(4), Hull, A., Policy Itegratio: What Will It Take to Achieve More Sustaiable Trasport Solutios i Cities? Trasport Policy, 15(2), Hug, K. V., Traffic Maagemet i Motorcycle Depedet Cities. Doctoral Dissertatio, Darmstadt Uiversity of Techology. Istitute for Trasportatio & Developmet Policy (2009) Best Practices o Regulatio ad Desig for Motorized ad No-Motorized Two ad Three Wheelers i Urba Traffic, last accessed o March 31, Japa Iteratioal Corporatio Agecy (JICA)., Project for the Strategic Urba Developmet Pla of the Greater Yago (SUDP). Japa Iteratioal Corporatio Agecy (JICA)., Project for Comprehesive Urba Trasport Pla of the Greater Yago (YUTRA). Kaeko, Y., Fukuda, A., Srisurapao, V., Oda, T., Estimatio of the Impact of Area Licese Scheme with Multi- Class User Equilibrium Model. Joural of the Easter Asia Society for Trasportatio Studies 4, Kato, H., Iagi, A., ad Htu, P. T. T., 2010a. Route Choice Aalysis of Bus-use Commuters i Yago, Myamar, Proceedigs of the 7th Iteratioal Coferece o Traffic ad Trasportatio Studies, Kumig (Chia), Aug Kato, H., Iagi, A., Saito, N., Htu, P. T. T., Feasibility Aalysis for the Itroductio of a Bus Rapid Trasit System i Yago, Myamar. Joural of the Easter Asia Society for Trasportatio Studies 9, Kato, H., Saito, N., Iagi, A., Myit, U. A., 2010b. Regulatory Framework ad Operatioal System of Urba Bus Trasportatio i Yago, Myamar. Preseted at the Trasportatio Research Board 2010 Aual Meetig, Washigto D. C. (U. S.), Ja Kojima, M., Futose, T., Bih, P. L., Kato, H., 2015a. Yago s Urba Trasportatio: Traffic Demad ad its Quality of Service. Preseted at the 11th Easter Asia Society for Trasportatio Studies, Cebu (Philippies), Aug Kojima, M., Futose, T., Bih, P. L., Kato, H., 2015b. Urba Trasport i Yago, Myamar: Evidece from Large-Scale Surveys. Preseted at the Trasportatio Research Board 2015 Aual Meetig, Washigto D. C. (U. S.), Ja

19 Workig Paper Koizumi, Y., Nishimiya, N., Kaeko M., Urba Trasportatio Characteristics ad Urba Mass Trasit Itroductio i the Cities of Developig Coutries. Joural of the Easter Asia Society for Trasportatio Studies 10, Lai, W.T., Lu, J. L., Modelig the Workig Mode Choice, Owership ad Usage of Car ad Motorcycle i Taiwa. Joural of the Easter Asia Society for Trasportatio Studies 7, Louviere, J. J., Hesher, D. A. Swait, J. D., Stated Choice Methods, Cambridge: Cambridge Uiversity Press. Mahedra, A., Optios for Travel Demad Maagemet: Traffic Bas versus Pricig. I: Dimitriou, H. T., Gakeheimer, R., (Eds.) Urba Trasport i the Developig World: A Hadbook of Policy ad Practice, Edward Elgar Publisher, May, A. D., Roberts, M., The Desig of Itegrated Trasport Strategies. Trasport Policy, 2(2), Morichi, S., Log-term Strategy for Trasport System i Asia Megacities. Joural of the Easter Asia Society for Trasportatio Studies 6(1), Morichi, S., Archarya, S. R., Trasport Developmet i Asia Megacities: A New Perspective, Spriger. Myamar Times., Ba o Motorbikes Ligers, 4 July last accessed o July 4, Schlag, B., Teubel, U., Public Acceptability of Trasport Pricig. IATSS Research 21, Sdhlag, B., Schade J., Acceptability of Urba Trasport Pricig Strategies. Trasportatio Research Part F: Traffic Psychology ad Behaviour 6(1), Tu, N., Kaitpog, K., Jiwattaakulpaisar, P., Kato, H., Study of Vehicle Owership i Yago usig Household Data. Preseted at the 8th ATRANS Symposium: Youg Researcher s Forum 2015, Bagkok (Thailad), Aug Tua, V. A., Shimizu, T., Modelig of Household Motorcycle Owership Behavior i Haoi City. Joural of the Easter Asia Society for Trasportatio Studies 6, Tua, V. A., Possibilities of Modal Shift to Public Trasport i Developig Asia Cities A Empirical Examiatio i Haoi City. Preseted at the 7th ATRANS Symposium: Youg Researcher s Forum 2014, Bagkok (Thailad), Aug Whittles, M., Urba Road Pricig: Public ad Political Acceptability, Ashgate Publicatio Ltd. Williams, H. C. W. L., O the Formatio of Travel Demad Models ad Ecoomic Evaluatio Measures of User s Beefit. Eviromet ad Plaig A 9, Wislott-Hiselius, L., Brudell-Freij, K., Vaglad, A., Byström, C., The developmet of public attitudes towards the Stockholm cogestio trial. Trasportatio Research Part A 43, Xiao, H., Lu, J., A comparative study of public ad academic attitudes to TDM measures: Evidece from Najig, Procedia Social ad Behavioral Scieces 96, Ye, L., Wag, Q., Case Study of Motorcycle Use ad Policy Aalysis i Huizhou, Chia. Joural of Trasportatio Egieerig 137, Zhag, J., Fujiwara, A., Thei, S., Capturig Traveler s Stated Mode Choice Prefereces uder Ifluece of Icome i Yago City, Myamar. Joural of Trasportatio Systems Egieerig ad Iform 19