Association of built environment with walking distance to public transit stops in hanoi

VIETNAM NATIONAL UNIVERSITY, HANOIVIETNAM JAPAN UNIVERSITY CHU ANH TUAN ASSOCIATION OF BUILT ENVIRONMENT WITH WALKING DISTANCE TO PUBLIC TRANSIT STOPS IN HANOI MASTER'S THESIS MASTER OF

VIETNAM NATIONAL UNIVERSITY, HANOI

VIETNAM JAPAN UNIVERSITY

CHU ANH TUAN

ASSOCIATION OF BUILT ENVIRONMENT WITH WALKING DISTANCE TO PUBLIC

TRANSIT STOPS IN HANOI

MASTER'S THESIS

MASTER OF INFRASTRUCTURE ENGINEERING

Hanoi, 2019

VIETNAM NATIONAL UNIVERSITY, HANOI

VIETNAM JAPAN UNIVERSITY

CHU ANH TUAN

ANNEX 2 LIST OF FORMS FOR MANAGEMENT

Prof HIRONORI KATO

Dr PHAN LE BINH

Hanoi, 2019

I sincerely thank to Prof Nguyen Dinh Duc - the Program Director ofInfrastructure Engineering Program of Vietnam- Japan University (VJU) whoalways has encouraged and deeply care about me.

In addition, I sincerely thank to all members in MIE’s office Specially, Iwould like to give my greatest thanks to Dr Nguyen Tien Dung and Mr Bui HoangTan who always give me strong support and insightful advice

My deep thank to Mrs Nguyen Thi Mai Chi about her precious support for mydata collection

Furthermore, I sincerely thank to the rest of the teachers and staffs of VietnamJapan University for their support and guidance which has inspired me and helped

me to overcome the challenges I faced during the period of study at Vietnam JapanUniversity To all of you, I extend my deepest gratitude

Finally, I would like to send my deepest love and gratitude to my family, whoalways love, support and encourage me every time

Sincerely,

Chu Anh Tuan

Similar to many other developing cities, Hanoi, the capital city of Vietnam, hasalso faced many serious problems caused by spontaneous urban development andmassive private vehicles As a part of solutions, public transportation systems in Hanoiwere paid attention to be invested and developed However, Public transportation meetsonly partially the travelling demand of urban people Even recent years, theproductivity of buses in Hanoi tends to be restrain and even going down in some years.Therefore, how to encourage and attract people to use public transportation modes is amore and more urgent topic This research focuses on a key factor which greatlyinfluences the use of transit services This is Walking distance to transit stop andimpacts of built environment on this walking distance of riders

A Poison Regression model is estimated to analyze the associations of builtenvironment and walking distance to public transit stop using a dataset collectedfrom a questionnaire-based Onboard survey in May 2019 with the geospatial data of

BE variables where the dataset contains 609 respondents of the Hanoi MetropolitanArea The results unveiled that: (1) Walking is the majority mode of both bus tripsfrom home to transit stop and from transit stop to final destination; (2) Theassumptions about walking distance to public stops (threshold of 400 meters) thatplanners have given are consistent with the Hanoi context; (3) BE variables mayhave both positive/ negative impacts on walking distance to public transit stops.Finally, some policy recommendations were produced to support for the use ofthe public transportation modes in Hanoi

Keywords: Walking; Walking distance; Built environment; Transit stop.

ii

TABLE OF CONTENTS

ACKNOWLEDGMENT i

TABLE OF CONTENTS ii

LIST OF FIGURES v

LIST OF TABLES vi

LIST OF ABBREVIATIONS vii

CHAPTER 1 INTRODUCTION 1

CHAPTER 2 LITERATURE REVIEW 6

2.1 The relationship between walking and BE 6

2.2 Distance walking and walking to public transit 7

2.3 Studies on walking distance and built environment in the world and Vietnam 10 CHAPTER 3 DATA 13

3.1 Survey Questionnaire 13

3.1.1 Survey area 13

3.1.2 Survey implemention 15

3.2 GIS-based database 20

3.3 Measurement of walking distance to public transit 20

3.4 Measurement of BE variable 22

CHAPTER 4 WALKING DISTANCE MODEL DEVELOPMENT 27

4.1 Overview 27

4.2 Explanatory variables 27

4.3 Methodology 33

CHAPTER 5 ESTIMATION RESULTS 35

5.1 Estimation results of access trip 35

5.2 Estimation results of egress trip 40

CHAPTER 6 DISCUSSION AND CONCLUSION 41

6.1 Walking distance to public transit 41

6.2 Association between BE and walking distance 42

6.3 Limitations 43

APPENDIX 48

A set of survey questionnaire in English 48

A set of survey questionnaire in Vietnamese 55

Correlation between potential explanatory variables 62

iv

LIST OF FIGURES

Page

Figure 1.1: Bus passenger ridership in Hanoi 2

Figure 1.2: Flow of Research 5

Figure 3.1: Survey bus route in Hanoi 14

Figure 3.2: A group of the survey and an interview on a bus route 16

Figure 3.3: An illustration in the survey questionnaire 16

Figure 3.4: Locations of respondents 21

Figure 3.5: Population density by ward 23

Figure 3.6: Job density by ward 24

Figure 3.7: Current bus network and bus stops distribution 26

LIST OF TABLES

Page

Table 2.1: Access modes from home to bus stops in Sydney 8

Table 3.1: The descriptive statistics of attributes of respondents (N=609) 18

Table 4.1: Potential explanatory variables for walking distance model 28

Table 4.2: Descriptive statistics of variables for access trip (N=290) 31

Table 4.3: Descriptive statistics of variables for egress trip (N=299) 32

Table 5.1: Estimation result of Poisson model for walking distance of access trip 38 Table 5.2: Estimation result of Poisson model for walking distance of egress trip 39 Table 6.1: Access modes from home to bus stops and from final stops to destinations in Hanoi 41

Table 6.2: Statistics on walking distances for access trip and egress trip 42

vi

Hanoi urban transport management and operation center

Transit-oriented development

CHAPTER 1 INTRODUCTION

1.1 Introduction

Over the years, Hanoi has witnessed a rapid development of the urbanizationthat leads to the increasing demand in travelling, thus to the rapid increase of thetransportation means, in particular private vehicles In December 2015, it wasreported that the total of private vehicles in Hanoi was 5.8 million units including376,417 cars and 5.4 million motorbikes (Ngoc, 2016), with the exception of thelarge number of vehicles from other provinces going in and out the city during theday Such a massive private vehicles population has caused various transportationissues such as road traffic accidents, traffic jam, and air pollution To deal withthose problems, Public transportation systems in Hanoi were paid attention to beinvested and developed, focusing on buses (only one bus rapid transit (BRT) route

is operating and urban railway are under construction in Hanoi) However, publictransportation meets only partially the travelling demand of urban people,specifically, public transportation system transports only less than 10 percentdemand in Hanoi city (World Bank, 2018) Even recent years, the productivity ofbuses in Hanoi tends to be restrain and even going down in some years (reduced0,35 percent in 2015 compared to that of 2014) (Figure 1.1) In particular, theprogress of construction and operation for the Mass Rapid Transit system (BRT,urban railway) in Hanoi missed the deadline Therefore, how to encourage andattract people to use public transportation modes is a more and more urgent topic

There are various factors that influence the use of public transportation modes,including cost, level of service of the existing system, physical accessibility, spatialaccess, etc all of which contribute to each person’s motivation and ability to decide touse public transportation This research focuses on walking distance to transit stopbecause understanding influences on walking distance to transit stops of riders is animportant indicator of a transit system’s ability to attract people The publictransportation services are widely used when more and more people live and/or work

1

in close proximity to transit stops (Murray et al., 1998) Moreover, walking distance

to transit stops is a very important aspect to define stop catchment areas, which areessential for the planning process of new public transportation lines (Andersen &Landex, 2008), and to evaluate the impacts of transit infrastructure on land-use, and

to design policies for transit-oriented development (TOD) Most studies and transitplanners usually assume distance thresholds of 400 meters for access to bus stops(Queensland Government, 1997; Murray, 2001; Biba, Curtin, & Manca, 2010).However, these distance thresholds have not yet been justified with an empiricalbasis in Hanoi In addition, potential factors affecting the different in walkingdistance to public transport of public transportation users such as demographics, trippurpose, built environment, perceived neighbourhood walkability are still unclear so

particularly pays attention to use measurement objective distance for analysesrelationship with BE In order to compute the BE variables in different geographic

2

units, detailed geospatial data for Hanoi measured in 2010 were installed into aGeographic information system (GIS) database Finally, some implications aremade based on estimation results.

Based on the background presented before, the research questions this studypursue to answer are:

i What is the actual distance walked to public transit stops in Hanoi?

ii How do the built environment characteristics affect the walking

distance to public transit stops?

1.2 Contributions of this research

This research makes significant contributions to the body of knowledge forseveral reasons This research contributes to enhancing our understanding about therelationship between BE and the walking distance to transit stops in case of adeveloping city like Hanoi and providing implications for land use planning and theproper location of transit stops in a public transportation system More specifically,this research aims to offer planning implications based on the following two aspects.Firstly, with regards to the stop area planning aspect, the BE attributes thatpositively impact on the actual walking distance of public transportation users need

to be identified This will suggest transit planners, policy makers and researchershow to motivate people to walk farther for using public transportation Second, withregards to the new stop location choice, this study wants to determine BE attributesthat tend to shorten public transportation's walking distance Consequently, the newtransit stops for new routes or relocate exist transit stop could be located in theplaces with these BE attributes by transit planners Therefore, this study fulfils theresearch gap in giving more evidences to support for both promoting riders walkinglonger and minimize rider's walking distance In addition, to the best of myknowledge, this study is the first study using GIS-based database to investigate therelationship between BE and walking distance in Hanoi

1.3 Research Framework

This thesis consists of five chapters This chapter focuses on the background,problems statement as well as the main objectives of this research The followingchapters are written as follows:

Chapter two reviews previous researches in the literature to clarify the researchobjectives in accordance with current studies This chapter aims to recognize theresearch gap that exists in the literature and needs to fulfil

Chapter three describes the data source as well as procedures for collectingthem This chapter also shows the procedure to develop GIS-based database inHanoi and establish the measurement of BE variables in GIS-based database

Chapter four aims to develop a model to analyse the relationship of BE onwalking distance to transit stop in Hanoi A Poison Regression model is adopted toanalyse the potential impact of BE on walking distance This chapter includes thedata characteristics, descriptive analysis, methodology for model analysis and itsresults

Chapter five displays the estimation results of the walking distance model andexplanation of these results

Chapter six concludes the discussion and the key findings obtained from theestimation Finally, research limitations and future research directions werediscussed The research flow diagram is presented in Figure 1.2 below

4

Figure 1.2: Flow of Research

CHAPTER 2 LITERATURE REVIEW

2.1 The relationship between walking and BE

There were a large number of papers which have studied on the relationshipbetween walking and built environment To investigate this connection betweenwalking and the BE, we need to have a clear understanding of the meaning of walkingand built environment In general, walking recognized as a movement by foot which isone types of the man’s transportation Broadly speaking, in urban context, walking isexplained as short distance moving from one point to the other point (Rafiemanzelat etal., 2017) As defined by Davison and Lawson (2006), The built or physicalenvironment is objective and perceived characteristics of the physical context in whichpeople spend their time (e.g., home, school) including aspects of urban design (e.g.,presence and structure of sidewalks), traffic density and speed, distance to and design

of venues for physical activity, crime, safety and weather conditions On the whole, the

BE includes three characteristics, there are: Design, Density, and Diversity Based onthese atributes of BE and the purpose of the trips, many studies focused on the impact

of BE to walking Ferrer et al., (2015) divided five main categories of builtenvironmental factors influencing walking for transportation including: Safety fromcrime (street lighting, other people, cleanliness, etc.); Traffic safety (traffic volume,traffic speed, crossing waiting times, etc.); Walking facilities (sidewalk width,obstacles, etc.): Aesthetics (presence of green elements, buildings, noise, etc.); andConvenience and other perceptions (availability of car parking, hills and pedestrianvolume, open and wide spaces and length perception) According to these BEcharacteristics, the role of BE for short walking trips were investigated (Ferrer et al.,2015) Furthermore, several studies mentioned the correlation between BE and walking

is moderated based on the varied factors of BE and trip characteristic (Lovasi et al.,2009) The spatial links between the built environment and walking are furtherexplored with the purpose of the trip such as walking for errands and leisure (Feuillet,

et al., 2016) or determined the contribution of the BE to explaining age differences inwalking (Ghani et al., 2018) Some of them

6

revealed the effectiveness of different BE attributes on improving walking bydetermining the general and specific features of the major built environment attributes

of residential neighborhoods which could help overcome varied barriers and enhancewalking and cycling activity levels (Wang et al., 2016) They comprise Greenery,public leisure space (e.g roof garden, fitness club, public space), specific road, trail andpath design, safety and security provisions, a wider choice of facilities (e.g sidewalks,cycling paths, treadmills, stairs) as well as some specific design provisions insidebuildings can help overcome some barriers that hinder walking and cycling activitieswithin a residential neighbourhood Ariffin and Zahari (2013) found that the proximity

of destinations, good weather condition, safety and well-designed pedestrian facilitiescan significantly contribute to better perceptions of the walking environment Throughthese studies, it showed that the BE factors have a strong relationship with walking.Nevertheless, a number of research gaps still remain, most studies revealing theassociation of BE and walking were originated from the various fields Consequently,there are only a limited number of studies could provide valuable finding on the designand planning of BE For instance, it has been confirmed that scenery made positiveeffects on walking activities; but there is a lack of researches revealing which type oflandscape design could induce activities Similarly, there is still a lack of evidencedemonstrating the complex relationships and interdependent between walkingconditions, facilities, BE atributes and walking within a residential neighborhood.Accordingly, further studies are needed to provide more appropriate understanding ofthe association between BE and walking activities

2.2 Distance walking and walking to public transit

Public transit (e.g buses, BRT, MRT) is generally not a point-to-point mode

of travel; it requires another mode to reach a pick up point and to get from a dropoff point to the passenger’s final destination The majority of transit users walk toreach to transit systems For example, According to Daniels & Mulley (2013), bysynthesizing the data from the Bureau of Transport Statistics, in Sydney, they

revealed that walking is the access mode for nearly 90% of bus trips from home to transit stops (Table 2.1).

Table 2.1: Access modes from home to bus stops in Sydney

(Source: Bureau of Transport Statistics)

Walking distance is especially important for at least two reasons Firstly, thedistance walked by public transport users to transit stops is a major element of atransit system’s capacity to attract pessengers in its service area , hence walkingdistance has a significant influence on public transport use (Daniels et al., 2013).Ewing and Cervero (2010) using a meta-analysis revealing a 10 percent increase inwalking distance to a transit stop would decrease public transport use by roughly 3percent Other papers have found that for every additional 500 m to reach a station,the probability that a rider will walk to transit system decreases by 50%(Loutzenheiser, 1997), and similarly every 10% decline in transit use when 10%increase in walking distance ( Dill, 2003) Durand, et al., (2016) have identified thecorrelation between distance to a transit stop and the probability it will be accessed

by walking In this study, they also examined even at three kilometres from a transitstop, there is still roughly a 50% possibility a rider will walk to the transit stopversus using a motorized mode Consequently, the distance people are willing towalk to stops seem to be much higher than the frequently cited rules of thumb of

400 meters Data for this study was collected from the California Household TravelSurvey 2012 by the survey questionnaires (both the individual and household-levelwere conducted) and the single-day travel diary via computer-assisted telephone

8

interviewing, online survey, or mailed survey Durand, et al., (2016) revealed thelimitations of the study are difficulties in generalizing the dataset, the lacking ofdetailed address data which make the study further unable to explain the effect ofwalking distance factors on active access to transit.

Secondly, it is admitted that planning for public transport system entails finding afeasible alignment that maximizes population accessibility to transit stops Thepopulation with access to transit are an important indicator to estimates of transit use.The ability to precisely measure walking distances to transit facilities has been elusivegiven the large number of possible walking paths for the population and given thequality of network data available for analysis (Biba, Curtin, & Manca, 2010).Transportation planners often assume the distance which people will walk to access topublic transportation or “rules of thumb” to determine stop spacing, particularly forbuses as these are more flexible but also by land-use planners for urban design toachieve walkable cities and plan transit-oriented developments (TODs)

In this section, the literature reviews the influences on walking distance to publictransport to determine possible explanatory variables for use in the analysis, includingtrip characteristics, socio-demographic attributes, the BE factors, and perceivedwalking conditions For example, El-Geneidy et al., (2014) found that there is anoppositional relationship between the number of transfers and walking distance,whereas the total trip length is positively correlated with walking distance O'Sullivan

& Morrall (1996) and Alshalalfah & Shalaby (2007) found that there is a positivelyassociated between walking distance and transit services with high level and short waitingtime Transit riders' demographic attributes, such as age, gender, occupation, income, thenumber of vehicles, and driver’s lisence are also important indicators of walking distance(García-Palomares et al., 2013; El-Geneidy et al., 2014) Previous studies have explored theinfluences of BE characteristics on walking distance because they are important for walkingdistance and public transportation use ( Agrawal, Schlossberg, & Irvin, 2008) In addition,El-Geneidy et al., (2014) and Jiang et al., (2012) found that walking distance has a positivelyassociation with population

density, intersection density, and sidewalk density Jiang et al (2012) concluded thatpublic transport users willing to walk further when the walking environment is highlywalkable Aesthetics and amenity are also potential determinant of walking distance.However, in a study evaluating the effect of a range of factors including amenity andaesthetics on deciding a route for walking, Agrawal et al, (2008) revealed that thefundamental consideration for riders walking to transit stations in the study inCalifornia and Oregon was minimizing the walking time and walking distance Safety(from traffic accident, rather than crime) was a secondary determinant in route choice,whereas amenity and aesthetics appearance was less of a concern.

In summary, there have been clear calls in the literature for relationship ofdistance walking and use of public transit The influences on distance walking to publictransit have also been highlighted Previous studies indicated that walking distance topublic transport may be affected by demographic factors (particularly income, age andgender), by the trip characteristics such as purpose of the overall trip, and by thelocation of the access trip in terms of BE, although the influences are variable Overall,

it shows that the ease of walking elements is impacted by built environment attributesbut it might be taken into account more factors in determining how far people walk topublic transport once they decide to walk to public transport

2.3 Studies on walking distance and built environment in the world and Vietnam

There are several studies on walking distance to transit and BE in Context oflarge cities Wang and Cao (2017) explored that BE correlates strongly with walkingdistance of the egress segments (between transit stops and non-home destinations).Study area of this research is the Mineapolis-St.Paul (Twin Cities) metropolitan area,

US By using the results of the 2010 Transit Onboard Survey with capturing followinginformation of respondents: trip purposes, origin and destination addresses, access andegress modes, transit routes, and demographic characteristics, this study developed fourmodels to compare the effects of the built environment around transit stops uponwalking distance of transit egress trip Focusing on walking distance of

10

egress trip (between transit stops and non-home-ends), and using ArcGIS which is ageographic information system (GIS) for working with maps and geographicinformation to measure dependent variable based on shortest path in street network,Wang and Cao (2017) found that, in term of transit egress, (1) the employmentfactor has a stronger effect on the walking distance to transit stops than thepopulation factor, (2) the number of intersections has a negative correlation withwalking distance to stops within downtown areas, (3) the number of transit stopspositively associated with walking distance to stops within downtown whereas isnegatively correlated with walking distance to stops outside of downtown areas, (4)The diversity in land use (land use mix) around transit stops has a significant andpositive relationship with walking distance to stops which outside of downtown andsuburban centers and (5) compared to the transit stops outside of downtown areas,the walking distacne to transit stops within downtown areas is much more affected

by the job accessibility Kamruzzaman et al, (2016) examined the relationshipbetween urban form and time spent (minutes) walking to transit in Brisbane,Australia, using both cross-sectional and longitudinal research design frameworks.This study concluded both cross-sectional and panel assessment methods confirmedthat the built environment influences walking participation, but it might take intoconsideration of perceptual and attitudinal factors is also important forunderstanding these relationships

One of the most recent studies on walking in Vietnam is written by Minh Tu Tran

et al, (2015) In this study, researchers have indicated that perceived neighbourhoodwalkability has a significant influence on mode choices of short-distance trips incontext of Hanoi, the capital of Vietnam They also assess the walkability of differenttypes of residential neighborhood including: downtown, mixed and new urban Data for this study was collected from a household face-to-face interview survey included a set ofquestions regarding residential neighborhood walkability: 1) residential density, 2)distance to various places, 3) street connectivity,

4) access to services, 5) walking facilities, 6) traffic safety and 7) crime safety and 8)

aesthetics and household and individual attributes: gender, age, employment,education level, household income, household size, vehicle ownership and so on.They found that people are uncomfortable to walk if trip distance is more than 500meters; The share of walking in downtown neighborhood was highest, implying thatresidents in this neighborhood are more likely to walk for short-distance trips; Theinfluences of perceived neighborhood walkability on mode choice of short-distancetrips were empirically confirmed Specifically, accessibility-by-foot, the fear ofcrime, walking facilities and traffic conditions were found to have significantinfluences on mode choice In another study, Tran et al, (2016) found that the morediversity of land use at residence and at working place, the more likely peoplelike towalk.

So at the end of it, this section comes to the conclusion that there is few researchexploring the built environment correlates of walking distance of both access trip andegress trip In addition, studies on walking in Hanoi, reliance on self-reported walkingdistance measures has been a weakness because of respondents’ abilities to accuratelyestimate walking distance Accordingly, the study have been conducted to fills thesegaps in the literature and extends the stream of these studies by (1) calculating theactual distance walked to public transit stops to define a catchment area for publictransport in Hanoi and (2) determining how built environment characteristics influencewalking distance to transit stops of both access segment from origins to transit stopsand egress segment from transit stops to final destinations

12

Based on data from Hanoi Bus Map and latest Hanoi bus routes database ofHanoibus, 15 bus lines that get good coverage of the entire main streets of Hanoicity were selected, including routes 02, 03A, 08A, 09, 14,16, 19, 22A, 23, 26, 27,

30, 32, 44, 50 (Figure 3.1) According to Hanoi urban transport management andoperation center (TRAMOC) data, these fifteen bus lines account for approximately20% of the total passengers of Hanoi's public transport system Each route has alarge volume of passengers, for example, line 02 has 2,795.419 passengers in thefirst quarter of 2019, line 3A is 1,100,677 passengers, etc By selecting these surveyroutes, this study may mitigate spatial aggregation bias compared to the other kind

of survey such as household survey or bus stop survey which need to have a muchlarger sample Furthermore, thanks for permission and support from TRAMOC, thesurvey was very successful and received cooperation from many passengers

Figure 3.1: Survey bus route in Hanoi

(Source: Created by the author, using source map from Hanoibus)

14

3.1.2 Survey implemention.

This survey was designed and conducted by writer with the support ofVietnam Japan University A questionnaire was designed for an interview-basedOnboard survey Specifically, a set of survey questionnaire consists of three parts.The first part is about socio-demographic information of respondents including 14questions such as gender, age, household income, household size and workers,vehicles availability and driver license These information provide an understanding

of public transportation users and were used to improve the explanatory power ofwalking distance model Part two begins from question 15 to question 28 focusing

on transit trip information of respondents including origin and destination Address,origin and destination activity, boarding stop and alighting stop location, number oftransfers, transit usage history, frequency of transit Use and the reasons for choosingtransit In addition to provide an in-depth understanding of how people use thepublic transportation system, these information is very important to calculateobjectively walking distance to public transit of respondents The last part of thesurvey questionnaire is Perceived neighborhood walkability The Likert scale wasused to measure and evaluate respondent sentiment on: conflict with other mode,cleanliness, level road, cross the street, drainage, step up and down, and walkingamenities Respondents must choose a positive or negative answer with 5-pointLikert scale corresponding to statements “Very Poor – Poor – Fair – Good –Excellent” In particular, each situation has an illustration so that respondents caneasily imagine (Figure 3.2)

Figure 3.2: An illustration in the survey questionnaire

(Source: Created by author)

609 passengers had participated in this survey Twelve surveyors from a localprofessional survey team were divided into 3 groups (4 surveyors/a group)performed 15 different bus routes At each bus line, the survey questionnaire will bedistributed directly to the passengers by surveyors With a desirable total of over

600 random samples and divided by 15 bus routes, each bus route was implementedabout 40 to 50 respondents In which, 20 samples at peak hours (17h - 18h and7h15-1h15) and 20-25 samples at off-peak hours (the remaining hours)

Figure 3.3: A group of the survey and an interview on a bus route.

(Photo taken by author)

16

In the week before the survey begin, surveyors were interviewed, hired andtrained In the survey process, surveyors interviewed passengers face-to-face andinterpreted for respondents to answer all questions correctly In order to ensure thesimplest and least expensive way to complete the necessary assignments, Surveywas distributed and collected as soon as passengers get on the bus because it is moreconvenient when people have more free time to answer questions (limited time on abus) In particular, in this study, the walking distance data will be collectedobjectively Accordingly, the address of an origin, destination, first bus stop locationand last bus stop location of respondents paid special attention to record in thesurvey in detail and clearly whereby the walking distance will be measuredobjectively on the ArcGIS tool.

Table 3.1 summarizes the descriptive statistics of the sample dataset with thesocio-economic profiles of the survey areas Female respondents account for 54.20% ofthe total, reflecting the proportion of riders using buses by gender distribution in Hanoi.The average age of respondents is only approximately 33 years of age, which reflectsthe fact that the young people dominate the use of public transportation in Hanoi Themost common job type occupies the highest proportion in the use of bus areknowledge-intensive labour (54.98%), including government officer (14.3%), private-company officer (31.5%), university researcher (2.1%), doctor (1.57%), and schoolteacher (5.51%) Service workers, unskilled workers, and street vendors/shopkeepersaccount for 9.45%, 6.56%, and 6.69%, respectively whilehousewives/unemployed/retired people and pupils/students account for 4.86% and11.55%, respectively In relation to motorbikes, 92.52% of respondents are owners, and95.54% have a motorbike license These results reflect the fact that most Hanoi peopleuse motorbikes for commuting In addition, 11.02 % of respondents own cars, and22.18% of respondents have a car license, which may mean that the number of cars willgrow in the future The results also show that the average number of householdmembers is 3.23 while the average number of working members is 2.09 This meansthat most respondents belong to quite small families, typically nuclear families Theaverage monthly household income is around 20 million VND, which

is slightly higher than the average household income in Hanoi The average

numbers of motorbikes and bicycles owned by respondents’ households are 2.19

and 1.26, respectively The average width of the road accessing their place of

residence is 3.78m (one-lane road) This reflects the poor condition of streets in

Hanoi, with narrow lanes and substandard road space, which could be one of major

factors leading to the dominant share of motorbikes in the city

Table 3.1: The descriptive statistics of attributes of respondents (N=609)

equivalent

Usable Minimum Maximum Mean Median Std.

Deviation motorcycles

There are still barriers to collecting the necessary information Firstly, there

are many participants who do not want to provide details of special housing-related

addresses In the data entry process, we need to consider and filter out the most

relevant and detailed answers Secondly, the survey has divided the time frame

including peak hours and off-peak hours During rush hour, there were many

difficulties because the number of people getting into the bus was very high, therewas a lack of space for convenient interviews

19

3.2 GIS-based database

In order to compute objectively walking distance and BE variables, it is essential

to have a complete GIS-base database of Hanoi city This research used GIS-basedatabase developed by Nguyen et al., (2018) This is a very valuable database becausethere was no GIS based system for Hanoi before In terms of creating this GIS-basedatabase, Nguyen et al., (2018) revised the primary data source from current mapsformatting in DWG files which developed by Hanoi Urban Planning Institute andconverted to GIS by using AutoCAD software After that, these maps were converted

to ArcGIS by using DWG-to-GIS convert tool After converting these maps to GIS,they transformed these maps into a global geographical coordinate system because theprimary maps are developed in VN2000, which is a local projected coordinate system.All the maps were transformed to global geographical coordinate system by usingArcGIS 10.4 software provided by ERSI

To complete this database, the data of spatial land use and transportationsystem attributes were collected and coded into a GIS-based database Once thedatabase was available, the respondent's addresses were inserted into Google EarthPro software via “Add Placemark” tool and transferred to GIS-based databased forcomputing variables

3.3 Measurement of walking distance to public transit

The On-board survey in Hanoi recorded the detailed address of origin, the first busstop, last bus stops and destination which were then geocoded to GIS-baseddatabase On a bus, the surveyor asked respondents for these location and recordedthem into questionnaire sheet Of the total respondents who participated the survey,there were 290 samples with detailed addresses for access trips and 299 sampleswith detailed addresses for egress trip (Figure 3.4) Because of privacy andconfidentiality reasons, these addresses and bus stops location were not provided inthe dataset, only the estimated walking distance was showed

20

Figure 3.2: Locations of respondents

(Source: created by writer, using source map from open street map)

Calculation of walking distance to public transit stop is a significant issue for thisresearch Objective distance can be measured in two ways without using trackingdevices: the shortest distance walking route along the street network (calculated using anetwork analysis function) and the straight-line or Euclidean distance (calculated using

a distance measuring tool) Both are measured objectively The first one is more

exactly than the other one The origin address, used bus stop and destination addresswere located clearly All these locations will be geo-coded as x, y coordinates into aGeographic Information System (GIS), after that, a network analysis tool in ArcGIS10.6 provided by ERSI was used It will analyse to calculate the shortest route andthe objective distance to public transit of each respondent According to Daniels &Mulley (2013), this is an approximation of the actually walking distance for severalreasons Riders might walk through open space and park rather than the main roadnetwork, which reduce their walking distance Otherwise, people might also choose

a longer path than the shortest road distance because the longer path is moreattractive, facilities or avoids negative elements However, the walking distance tothe bus stop is not too long Therefore, the walking distance which was measuredbased on the shortest path in street network using ArcGIS is accurate and close to themost actual walking distance

3.4 Measurement of BE variable

Each BE variable were computed for four different buffer including 100, 200,

500, and 1000 meters This ensures that the parameters are calculated on a diverserange of spaces

Figure 5 illustrates the ward-scale population density distribution in Hanoi

22

Figure 3.3: Population density by ward (Created by the writer, using database:

Nguyen et al., (2018); base map: Openstreet map)

Job density

Data about employment density in Hanoi is provided by Hanoi statistical Office(HSO) in 2017 This data consists of number of employment in both state and privatecompanies in ward-scale level Employment density is measured following the sameformulation as population density measurement, as shown below:

, =

∑ ∈

,

Figure 3.4: Job density by ward

(Created by the writer, using database: Nguyen et al., (2018);

base map: Open street map)

24

area in ward k; and AiR is the area of the buffering zone (R) for the individual n.Figure 3.6 illustrates the ward-scale Job density distribution in Hanoi.

Entropy Index of land use mix

Land use mix is measured by using entropy index which measures theheterogeneity of the spatial unit A land use map in 2010 was used to compute theentropy index The map categorizes the land-use patterns into 3 different types:residential land, public-used land (such as offices, market, hospital, departmentstore, etc.) and other purposes (such as industrial land, transportation) The entropyindex is estimated based on following equation (Frank et al., 2005):

( ) = − ∑ =1

, ( , )

( , )

number of land use types in the buffering zone (R)

Number of bus stops and bus frequencies

By using the GIS-base database which all bus stops were pointed on the map,the number of bus stops of each buffer were counted The bus stop map provided byHanoibus – Transerco – the local bus operator in Hanoi Bus frequency wascomputed based on the bus schedule provided by Hanoibus – Transerco, that equal

to the total number of bus passes all bus stations in a day within each buffer Figure3.7 shows the bus network and bus stop distribution in Hanoi

Number of road intersection

Number of way road intersection is computed by counting total number of way or more within a buffer area This variable explains the street connectivity andaccessibility for riders

4-Number of public facilities.

A source maps about current land use in Hanoi was used to show currentpublic spaces The footmark of each public space was pointed by using a Feature-to-Point tool in ArcGIS Afterward, the number of public facilities is computed bycounting total number of public facilities such as market, hospital, department storeand so on within a buffer area

Figure 3.5: Current bus network and bus stops distribution

(Created by the writer, using database: Nguyen et al., (2018); base

map: Open street map)

26

CHAPTER 4 WALKING DISTANCE MODEL DEVELOPMENT

4.1 Overview

This chapter develops a model for walking distance to public transit in Hanoi.Three sections will be displayed in this chapter including: overview, explanation ofthe variables and methodology for walking distance model development Thepurposes for developing walking distance functions in Hanoi is to examine anassociation of built environment with walking distance

4.2 Explanatory variables

Table 4-1 shows the explanatory variables used for the walking distance modeldevelopment The explanatory variable can be categorized into four groups: builtenvironment, individual characteristics, trip characteristics and perceivedneighborhood walkability The built environment consists of seven variables:population density, employment density, land use mix, bus stops, bus frequency andintersections These are continuous variables measured objectively by ArcGISanalysis tool The individual characteristics includes dummy variables named male,young people, senior people, motorbike license, car license Beside, ranks ofhousehold income (low income, middle income and high income) was created.Perceived neighborhood walkability is a series of questions: conflict with othermode, cleanliness, level road, cross the street, drainage, step up and down andwalking amenities which used a Likert scale to assess the current walking condition

of respondents Finally, trip characteristics consist of a continuous variable and twodummy variables These are total buses, main transportation: motorbike and Othervehicle available, respectively

Table 4.1: Potential explanatory variables for walking distance model

Built environment attributes

Land Use

Index

Individual characteristics

male

“between 18 to 25 years old”

“over 55 years old”

has a car driver's license

Perceived neighbourhood walkability