Technical ReportPDF Available

Accessibility and Social Equity in Tel-Aviv Metropolitan Area - examination of the current conditions and development scenarios

Authors:

Abstract and Figures

This research continues the development of a computation tool to examine accessibility in a metropolitan area at high spatial resolution –that of the individual building. We use this tool to examine the issue of equity in transportation infrastructure. Our measure of accessibility is relative – it looks at different modes of mobility, particularly private and public transport, and examines the relative ease of reaching desired destinations using each mode. In this research we extend our discussion into the way public transport serves those who are dependent on it, and who, for lack of financial means, age, or disability are not able to use private cars. To measure the inequity of the transportation system we use the traditional measures defined in economics: the Lorenz Curve, Gini Index and the Poverty Line. Further, we utilize a method to estimate the areas (TAZ) which are more dependent on public transport, beyond low income, car ownership, or actual public transport use. We show that there is a clear pattern of public transport dependency in the metropolitan area of Tel-Aviv: dependency is especially high in the towns of the outer ring of the metropolitan area. Generally speaking, the North-South divide visible in the city of Tel-Aviv – Yafo continues into the metropolitan area along two axes towards the South and Southeast, and the Northeast. Using these concepts we estimate the relative and absolute loss of accessibility, and the accessibility poverty, for the city of Tel-Aviv – Yafo to all jobs throughout the metropolitan area. We compare between the public transport system before and after the bus reform of 2011. Our results show that overall accessibility by public transport, relative to private cars, improved significantly, and thus accessibility poverty was reduced significantly in Tel-Aviv. However, the overall inequity between different zones, particularly with regard to public transport dependency, remained about the same and became even somewhat larger. It will require further analyses to assess the impacts of the recent investments in public transport infrastructure on the equity across the entire metropolitan area.
Content may be subject to copyright.
1
תושיגנ תוינויווישו תיתרובחת ןילופורטמב
לת-ביבא - הניחב לש בצמה םייקה יטירסתו
חותיפ םיידיתע
Accessibility and Social Equity in Tel-
Aviv Metropolitan Area - examination of
the current conditions and
development scenarios
וד" חיפוס
אפור ןדוי
1,4
,ןוסננב קחצי
2
, סנטרמ לרק
3
, ןב ןרעאילא
4
,קינדמ הילטנ
1
1
רבדמב םדאל הקלחמה ,ע םינוכמה"י ש .רבדמה רקחל ןייטשואלב ןוירוג ןב תטיסרבינוא ,
בגנב
2
הםדאה תביבסו היפרגואיגל גוח ,ביבא לת תטיסרבינוא
3
Department of Geography, Planning and Environment, Institute for
Management Research, Radboud University Nijmegen, Netherland
4
יתביבס חותיפו היפרגואגל הקלחמה ,ןב תטיסרבינוא-בגנב ןוירוג
דרשמל שגוהםיכרדב תוחיטבהו הרובחתה
לירפא 201
5
2
ריצקת ירבע םירבדה ירקיע
רקחמ הז ךישממ חותיפב לש ילכ בשחוממ )CityGraph (ססובמ ממ"ג חותינל תושיגנה ןילופורטמב
היצולוזרב תיבחרמ ההובג ,דע תמרל ןיינבה דדובה .תובישחה לש היצולוזר ההובג איה חותינהש
לבוקמה לע יפ ירוזא העונת ,ונניא איבמ ןובשחב תא תונושה ההובגה ךותב ירוזא העונת רקיעב
יחווטב ןמז םירצק ,םהבש ינמז הכילהה לגרב ,הנתמההו תונחתב ,םייתועמשמ רתוי קלחכ ןמזמ
העיסנה ללוכה .תועצמאב ילכ הז ונא םילוכי בשחל תמרב קויד ההובג תא
תושיגנה תיסחיה ןיב
יעצמא העיסנ םינוש .רקחמב הז ונישע שומיש ילכב הז ,ידכ רוקחל תא ירעפ תושיגנה ןיב בכר יטרפ
הרובחתו תירוביצ ןילופורטמב ריעבו לת-ביבא ,ךות האוושה ןיב םירוזא ,םינושה הז הזמ ,תניחבמ
תולתה םהלש הרובחתב תירוביצ.
רויא 1: תושיגנה תיסחיה לתמ-ביבא ופי לכל תומוקמ הדובעה ןילופורטמב ב 7:15 רקובב .תועיסנ
ךשמב 45 תוקד )ןימימ ,(30 תוקד )זכרמב (ו- 15 תוקד )לאמשמ(
רויא 1 םיגדמ תא שומישה ילכב בשחוממה בושיחל תושיגנה תיסחיה לכמ ןינב ריעב לת-ביבא ופי
לכל תומוקמ הדובעה ןילופורטמב ,רשאכ ןמז האיציה אוה 7:15 רקובב ,ךשמו העיסנה 15 ,30 ,ו- 45
תוקד המאתהב .םיעבצה םינושה םיארמ תא סחיה ןיב רפסמ תומוקמ הדובעה םישיגנה הרובחתב
תירוביצ הלאל םישיגנה בכרב )ותואמ םוקמ .(ןתינ תוארל יכ לככ ךשמש ןמז העיסנה הלוע תושיגנה
תיסחיה תרפתשמ ,ןכו בש 45 תוקד טעמכ לכב לת-ביבא ופי הנשי תושיגנ
תיסחי ההובג לש
רתוי מ 0.5.
ידכ ךירעהל תא תדימ תוינויוושה תושיגנב ןיב םירוזא םינוש ונמאתה וניכרצל השולש םידדמ םילבוקמ
ןדמואל תוינויוושה םוחתמ הלכלכה :תמוקע ץנרול ,דדמ ג'יני גשומו וק ינועה .תועצמאב ינש םידדמה
םינושארה ונלוכי תוארהל לשמל יכ תורמל רופישה לודגה לחש תושיגנב תומוקמל הדובע ןילופורטמב
ריעהמ לת-ביבא ופי ,תמר תוינויוושה לש תושיגנה וזה הרתונ תוחפ וא רתוי יפכ התייהש ףאו הדרי
טעמ )דדמ
ג'יני לש 0.26 תשרל םיסובוטואה הנשיה תמועל דדמ ג'יני לש 0.30 תשרל השדחה.(
חותינ ירוזא הז ,ונניא איבמ ןובשחב תא תדימ םתולת לש םיבשותה ירוזאב ריעה םינושה יתורישב
הרובחת תירוביצ ,דעו המכ תכרעמ הרובחתה תירוביצה תמרות תושיגנל לש הלא םיקוקזה הל
רתויב ,ללגב רדעיה תלוכי וא תורשפא דיינתהל בכרב יטרפ .ידכ ןוחבל םיטביה הלא לש תוינויווש ,
ונמאתה יאנתל ץראה עדימלו םייקה תכרעמב ,דדמ תולתל הרובחתב תירוביצ
ססובמה לע םינתשמ
םייפרגומד ויצוסו-םיילכלכ לש הייסולכואה ירוזאב העונתה םינושה )רויא 2 .(ןתינ תוארל יכ תולתל וז
שי םיסופד םייבחרמ םירורב ןה ןילופורטמב ןהו ריעב לת-ביבא ופי .תולת ההובג רתוי ןתינ תוהזל
ירעב הירפירפה )ואלו אקווד קר האצותכ המרמ תילכלכ הכומנ יפכ הארמש תולתה ההובגה חתב"צ
ןיעידומב וא םהושב .(רעפה רורבה ןיב ןופצ םורדו ריעב לת-ביבא ופי ,ךישממ םג םייקתהל םיווקב
3
םייללכ ןילופורטמב ולוכ ינשב םיריצ םיישאר :םורדמ לת-ביבא ופיו ןוויכל םורד םורדו חרזמ ,חרזממו
לת-ביבא ןוויכל ןופצ חרזמ .תמר תוינויוושה לש תולתה הרובחתב תירוביצ ןילופורטמב )דדמ ג'יני
0.235 (הכומנ ףא טעמב וזמ לש תוגלפתה הסנכהה קשמל תיבה ןילופורטמב )דדמ ג'יני - 0.223.(
רויא 2 :דדמ תולתה הרובחתב תירוביצ ןילופורטמב )לאמשמ (ריעבו לת-ביבא ופי )ןימימ(
םתוא םישנא םייולתה הרובחתב תירוביצה םילבוס ןדבואמ לש תויונמדזה )הדובע ,םיתוריש ,וא
םירשק םייתרבח (תיסחי םיעסונל בכרב יטרפ .בוליש לש דדמ תושיגנה תיסחיה דחי םע תמר תולתה
הרובחתב תירוביצ דדמב ןדבוא תושיגנה ,רשפאמ ונל דומאל תא ןדבואה הזה לכב רוזא רוזאו.
רויא 3 הארמ תא ןדבא תושיגנה יסחיה תומוקמל הדובעה לש ישמתשמ הרובחתה תירוביצה העיסנב
הליחתמה ב 7:15 תכשמנהו 45 תוקד )ןהבש עסונה בכרב לתמ-ביבא ופי לוכי
עיגהל טעמכל לכ
תומוקמ הדובעה ןילופורטמב .(רויאה הוושמ ןיב ןדבוא הז ינפל ירחאו המרופרה תשרב םיסובוטואה
לש 2011 ,תאו יונישה לחש האצותכ המרופרהמ .ןתינ תוארל יכ לח רופיש ללוכ לש תושיגנה לכב
יבחר לת-ביבא ופי ,ךא רופיש הז אל היה ינויווש ,אלו חרכהב דקמתה םתואב םירוזא שיש םהל תולת
ההובג רתוי הרובחתב תירוביצ.
4
רויא 3:ןדבא תושיגנה תומוקמל הדובע םישמתשמל ילעב תולת הרובחתב תירוביצ ,ינפל )לאמשמ (ירחאו
)זכרמב (המרופרה םיסובוטואב ב- 2011 .ןימימ יונישה תושיגנב תומוקמל הדובע האצותכ המרופרהמ
דדמ ףסונ חתופש רקחמב ןדמואל העפשהה לש חותיפ תכרעמ הרובחתה לע תוינויוושה אוה דדמ
ינועה יתרובחתה .לע יפ דדמ הז ,םישנא אלל השיג בכרל םילבוס ינועמ יסחי תושיגנב תויוליעפל
ריעהש וא ןילופורטמה העיצמ היבשותל .ןתינ רידגהל תא ינועה יסחיה תומרב
תונוש ,יפכ השענש
רבדה םג םע וק ינועה ילכלכה לבוקמה .לע יפ דדמ הז המרופרה לש תנש 2011 המרג רופישל
יתועמשמ דואמ )תוביבסב 10 (%תמרב ינועה יתרובחתה ריעב לת-ביבא ופי )רויא 4 ,(םא יכ םג ןאכ
םייונישה םה םייסחי ,םירוזאבו םימייוסמ הלח ףא הערה בצמב ינועה יתרובחתה רחאל המרופרה.
רויא 4 :המורתה ינועל תושיגנ לתמ-ביבא ופי תומוקמל הדובע ןילופורטמב )םיזוחאב לש ינוע וק םע50%
יטרפ בכרב תושיגנהמ .(תכרעמ םיסובוטוא ינפל המרופרה )
לאמשמ (רחאל המרופרה )זכרמ (תדימו
יונישה )לאמשמ(
הדובעב וז ,וניארה תא ותובישח לש שומישה בושיחב תושיגנה תמרב היצולוזר ההובג ,דחוימב
תועיסנל תורצק .ונחתיפ םידדמ תכרעהל תוינויוושה לש תכרעמ הרובחתה ,לשו םייוניש תכרעמב
האצותכ תועקשהמ ,וניארהו יכ ךסב-לוכה ,תכרעמ הרובחתה תירוביצה ריעב לת-ביבא ופי הנניא
תנתונ הנעמ דחוימ םתואל םירוזא םהבש תולתה תושיגנב הרובחתל תירוביצ ההובג רתוי .תאז
תורמל תמרש תושיגנה תללוכה הרפתשה דואמ האצותכ המרופרהמ
תכרעמב םיסובוטואה ,יפלו ךכ
הדרי םג תמר ינועה יתרובחתה .ןתינ דיתעב ,ביחרהל תא םידדמה םילכהו ונחתיפש תובשחתהל
םייונישב תושיגנב ךרואל תועש הממיה ,םינפהלו תא יא תואדוה תמרב תורישה הרובחתב תירוביצה
5
תיטרפהו האצותכ שדוגמ .ומכ ןכ ,ןתינ ךירצו םיאתהל תא םידדמה תמרל שוקיבה תויוליעפל תונושה
תולתכ קחרמב ,ףוסבלו ידכ רוציל ילכ םלש רשפאמה תלבק תוטלחה תואיבמה ןובשחב םילוקיש
לש תולע תלעותו ,שי לולכל ותוא ךותב חותינ וטרפ )Pareto (לש תופולח תוגיצמה ינפב ילבקמ
תוטלחהה תא תמר תושיגנה ,תמר תוינויוושה תאו תולעה לש לכ הפולח ואל אקווד תופולחה
תורקיה רתויב ןה ולא תונתונה תואצות תוילמיטפוא ינשל םידדמה םינושה .
ונרעצל ,תוביסמ לש םיישק םיינכט םייבושיחו ,אל רשפאתה ונל תרגסמב רקחמ הז ןוחבל תא
תוכלשהה לש המרופרה םיסובוטואב לע ינפ ןילופורטמה ולוכ ,הדובעו וז ןיידע הכירצ תושעיהל .ומכ
ןכ ררבתה ונל יכ ידכ תושעל שומיש לכשומ ילכב הז ןונכתב תופולח תויתרובחת ,שי רשקל ותוא
ללוחמל םישיחרת ,לכויש רוציל ןפואב יטמוטוא םישיחרת םיטרופמ םישרדנה חותינל לע ידי
תכרעמה .תכרעמה ומכ איהש םויה המיאתמ דחוימב רוביחב ןיב ןונכת הרובחת
לוהינו ןוכנ לש
הרובחתה תירוביצה תמרב ריעה .בושח םירעש ולכוי טולשל לע לוהינ הרובחתה תירוביצה ,םיאתהלו
התוא ןפואב דימתמ םיכרצל םינתשמ לש םישמתשמה ,ילבמ סנכיהל תועקשהל קתע ןונכתב
תויתשתבו ,םימעפל הקיפסמ תזזה המכ תונחת וא יוניש לולסמ לש וק דדוב םייוניש השקש דואמ
ךירעהל תא םתעפשה תועצמאב ילכ ןונכת םייפוקסורקמ וא םילדומ םיטרופמ לש לועפת ינטרפ לש
יליעפמ חת"צ .ומכ ןכ ןתינ תושעל שומיש ילכב הז
ןונכתל תושיגנה םיינפואב ,בולישלו לש תוכרעמ
ףותיש ילכ בכר ריעב.
6
Abstract
This research continues the development of a computation tool to examine
accessibility in a metropolitan area at high spatial resolution that of the individual
building. We use this tool to examine the issue of equity in transportation
infrastructure. Our measure of accessibility is relative it looks at different modes of
mobility, particularly private and public transport, and examines the relative ease of
reaching desired destinations using each mode. In this research we extend our
discussion into the way public transport serves those who are dependent on it, and
who, for lack of financial means, age, or disability are not able to use private cars.
To measure the inequity of the transportation system we use the traditional measures
defined in economics: the Lorenz Curve, Gini Index and the Poverty Line. Further, we
utilize a method to estimate the areas (TAZ) which are more dependent on public
transport, beyond low income, car ownership, or actual public transport use. We
show that there is a clear pattern of public transport dependency in the metropolitan
area of Tel-Aviv: dependency is especially high in the towns of the outer ring of the
metropolitan area. Generally speaking, the North-South divide visible in the city of
Tel-Aviv Yafo continues into the metropolitan area along two axes towards the
South and Southeast, and the Northeast.
Using these concepts we estimate the relative and absolute loss of accessibility, and
the accessibility poverty, for the city of Tel-Aviv Yafo to all jobs throughout the
metropolitan area. We compare between the public transport system before and after
the bus reform of 2011. Our results show that overall accessibility by public transport,
relative to private cars, improved significantly, and thus accessibility poverty was
reduced significantly in Tel-Aviv. However, the overall inequity between different
zones, particularly with regard to public transport dependency, remained about the
same and became even somewhat larger. It will require further analyses to assess
the impacts of the recent investments in public transport infrastructure on the equity
across the entire metropolitan area.
7
Acknowledgments
We would like to acknowledge the Ministry of Transport and Road Safety that
provided funds for this research, and in particular Zeev Shadmi, Head of the
Research and Technological Development Section at the Ministry’s Chief Scientist
Office, for guiding and supporting this research, patiently and with confidence in its
importance and our ability to carry it through to completion. We would also like to
thank the members of the steering committee for their inputs: Yehushua Birotcker,
Dr. Dani Givon, Sarit Levi, and Prof. Yossi Prashker.
We would like to thank the people at Performit Ltd. and particularly Dmitry
Geyzersky, and the people at CityGraph, and particularly Jacob Ben-Arieh and Dr. Eli
Safra, who helped develop the computational tools used in this research. They
worked well beyond the call of duty. Without them this research would have not been
possible.
Finally, this was a complex project, so perhaps it should be made clear who is
responsible for what parts of the research. The major responsibility for developing the
algorithm calculating accessibility at a fine scale and for coordinating the work of the
calculation was Prof. Benenson’s, as was the final mapping of the results. The
development of the ideas on measuring equity in transportation was done by Dr.
Rofè, Prof. Benenson and Dr. Ben-Elia together, but the formulas and the writing of
the relevant chapter, and their development for future extensions were done mainly
by Dr. Ben-Elia. The development of the dependency index was done by Dr. Ro
and Ms. Mednik, and the development of the Transportation Poverty index and the
writing of the relevant chapter were done by Prof. Martens. The mapping of
dependency and poverty were done by Ms. Mednik. The compilation and editing of
the report were principally done by Dr. Rofè with significant help in editing from Dr.
Ben-Elia and Prof. Benenson.
8
Table of Contents
ריצקת ירבע ירקיע םירבדה ............................................................................................. 2
Abstract ..................................................................................................................... 6
Acknowledgments ..................................................................................................... 7
Table of Contents ...................................................................................................... 8
List of Figures .......................................................................................................... 10
List of Tables ........................................................................................................... 11
1 Introduction .................................................................................................... 12
1.1 Goals and Research Questions ................................................................... 12
1.2 Organization of this report ............................................................................ 12
2 Measuring accessibility at the human level ..................................................... 15
2.1 Definition of accessibility .............................................................................. 15
2.2 An operational definition............................................................................... 16
2.3 CityGrapha GIS-based tool for measuring accessibility ............................ 17
2.4 Accessibility in Tel-Aviv Metropolitan Area ................................................... 20
2.5 Justification for high spatial resolution of accessibility analysis .................... 27
3 Measuring accessibility inequity ..................................................................... 29
3.1 Inequity in transportation planning ............................................................... 29
3.2 Measuring horizontal inequity: The Lorenz curve and Gini index .................. 29
3.3 Demonstration: the impact of bus reform on horizontal equity ...................... 31
3.4 Moving from variation in accessibility to measuring magnitude of inequity ... 32
4 Estimating public transport dependence ......................................................... 34
4.1 Transportation need index calculation for Tel-Aviv Metropolitan Area .......... 36
4.2 Public transport dependency results ............................................................ 37
4.3 Conclusions ................................................................................................. 43
5 Evaluating accessibility loss due to public transport dependency ................... 44
6 Evaluating accessibility poverty ...................................................................... 48
6.1 Methodology ................................................................................................ 48
6.2 Results ......................................................................................................... 51
7 Future extensions ........................................................................................... 57
7.1 Accommodating daily variation and travel time unreliability .......................... 57
7.2 Accommodating demand for travel ............................................................... 57
7.3 Integrating equity analysis with cost-benefit analysis ................................... 59
9
8 Summary, conclusions and further research .................................................. 61
8.1 Major findings .............................................................................................. 61
8.2 Difficulties encountered in the research ....................................................... 62
8.3 Future research ........................................................................................... 63
9 References ..................................................................................................... 65
10
List of Figures
Figure 1: Translation of the public transport network to a graph ............................... 19
Figure 2: Illustration of the Neo4J database idea ...................................................... 19
Figure 3: No. of jobs in the Metropolitan Area accessible by car from Tel-Aviv
starting at 7:15am. in 15 minutes (a), 30 minutes (b), and 45 minutes (c) ... 21
Figure 4: Before the bus reform - No. of jobs in the Metropolitan Area accessible
by public transport from Tel-Aviv starting at 7:15 and taking 15 minutes
(a),30 min. (b) and 45 min. (c) .................................................................... 22
Figure 5 : Before the bus reform - Relative accessibility at 7:15 am, for 15 min.
(a), 30 min. (b) and 45 min.(c) long trips. .................................................... 23
Figure 6 : After the bus reform - No. of jobs accessible from Tel-Aviv in the
metropolitan area starting at 7:15 and taking 15 minutes (a), 30 min. (b)
and 45 min. (c). ........................................................................................... 24
Figure 7 : After the bus reform - Relative accessibility to jobs in TAMA from every
building in Tel-Aviv at 7:15 am for trip lengths of 15 (a) 30 (b) and 45
minutes (c) .................................................................................................. 25
Figure 8 : New versus the old bus network - Difference in relative accessibility in
the Tel Aviv Metropolitan area for every building in Tel Aviv Yaffo for a
trip of 15 (a), 30 (b) and 45 (c) minutes ....................................................... 26
Figure 9: New bus network Coefficient of Variation (CV = 100%*STD/MEAN) of
the relative accessibility for buildings within each TAZ. Relative
accessibility is estimated based on trips to the entire Tel Aviv
Metropolitan area, for every building in the city of Tel Aviv. Duration of
trip is 15 (a), 30 (b) and 45 (c) minutes ....................................................... 28
Figure 10 : Representation of the Lorenz curve ........................................................ 30
Figure 11: Lorenz curves and Gini indices for relative job accessibility using the
Old and New bus network in Tel-Aviv. ........................................................ 31
Figure 13: Adults without cars .................................................................................. 38
Figure 12: Unemployed adults .................................................................................. 38
Figure 14: Persons 10-18 years old .......................................................................... 38
Figure 15: Persons above 65 years old .................................................................... 38
Figure 18: Persons with income below median ......................................................... 39
Figure 19: Pubic transportation need index in the Metropolitan area and the city
of Tel-Aviv .................................................................................................. 40
Figure 20: The makeup of weighted Transportation Need Index for the 30 highest
need TAZ .................................................................................................... 41
Figure 21 : Lorenz curves and calculation of Gini index for Public Transportation
Need and Income by TAZ ........................................................................... 42
Figure 22: Relative accessibility loss in the City of Tel Aviv for a 45 min trip to all
jobs in the Metropolitan Area on the old bus network (a), the new bus
network (b) and the changes in accessibility lost RO(OLD) RO(NEW)
caused by reforms to the bus network (c). The relative decrease is
positive when R with the new bus system is lower compared to the the
old system. ................................................................................................. 45
Figure 23: Absolute accessibility loss Li in the city of Tel Aviv for a 45 min trip to
all jobs in the Metropolitan Area , for the old (a) and new (b) bus
11
networks, and the percentage change 100%*[LO(OLD)
LO(NEW)]/LO(OLD) caused by reforms to the bus network (c). .................. 46
Figure 24: Scatterplot of Transportation Need Index and Relative Accessibility to
jobs index from the city of Tel-Aviv-Yaffo. ................................................... 47
Figure 25: The three components for assessing the equity of a transportation
system ........................................................................................................ 49
Figure 26: Share of TAZ with an accessibility level above and below the 20% or
50% poverty line, before and after the reform of the public transport
system. ....................................................................................................... 53
Figure 27: Contribution of TAZ to accessibility poverty in percentage (20%
accessibility poverty line). Old bus network (a), New bus network (b) and
the change in contribution (c). Negative values area TAZ where
accessibility poverty grew after the reform. ................................................. 55
Figure 28: Contribution of TAZ to accessibility poverty in percentage (50%
accessibility poverty line). Old bus network (a), New bus network (b) and
the change in contribution (c). Negative values area TAZ where
accessibility poverty grew after the reform. ................................................. 56
Figure 29: Hypothetical bundles of projects plotted by cost and equity ..................... 59
Figure 30: Possible Pareto Frontier between cost and equity based on Feng and
Zhang (2012) .............................................................................................. 60
List of Tables
Table 1: Statistics of the difference between the values of relative accessibility for
the new and old networks, as dependent on the duration of trip ................. 27
Table 2: Public transportation dependence indicators according to Currie 2010 ....... 35
Table 3: Need indicator weights ............................................................................... 36
Table 4: Number of TAZ and population in each category of need index .................. 43
Table 5: Statistics for the population and zones falling below the accessibility
poverty lines of 50% and 20% of the average car-based accessibility to
employment in Tel Aviv. .............................................................................. 54
12
1 Introduction
This report describes the research project intended to examine the connection
between accessibility and equity in the metropolitan area of Tel-Aviv. In the course of
this research in collaboration with CityGraph Access Ltd, we have completely
revamped the ArcGIS tool Urban.Access, used in our previous research on relative
accessibility (Benenson et al. 2011), and developed a new tool capable of calculating
rapidly the accessibility for every building in a metropolitan area now called
CityGraph.
1.1 Goals and Research Questions
The research questions defined in the proposal were the following:
1. Which accessibility indicator(s) are most suitable to assess and evaluate
transportation and/or land use policies from the perspective of equity?
2. How equitable are the differences in accessibility in the Tel Aviv Metropolitan
Area?
3. What are the impacts of the recent metropolitan bus-reform on accessibility
by public transport and transportation equity?
4. What are the accessibility and equity implications of different scenarios of
transportation and land-use plans and projects?
Most of the research effort in this project went to examine the first question, which
has proven to be more complex and less researched than initially thought. While our
measure of accessibility in itself distinguishes, at a very fine spatial scale, the
difference in accessibility between people living in different areas (called horizontal
inequity), there are different methods to examine the impact of these differences,
taking into account that the distribution of public transport dependency is also
spatially uneven. In the end we have come up with two different ways of measuring
this impact one looking at public transport dependency, and how it leads to
absolute and relative loss of accessibility. The second way is by defining accessibility
poverty and examining how transportation changes affect it spatially. To illustrate
these different measures, we modeled the accessibility from the city of Tel-Aviv
Yafo to all the jobs in the metropolitan area, comparing the situation before and after
the bus reform of 2011.
The last objective proved difficult to attain. This is primarily because of the difficulty of
writing out theoretical scenarios in a sufficiently detailed way to allow analysis by
CityGraph. Starting the research we did not fully recognize this technical difficulty, to
resolve it we would need either close cooperation with a project team in defining the
scenarios, or much more detailed work in describing the scenario changes for the
CityGraph system to work on. We address this problem further in our conclusions.
1.2 Organization of this report
In chapter 2, after a brief recapitulation of the literature on accessibility analysis, we
describe how the CityGraph tool works, and we demonstrate this tool, to examine the
relative accessibility of public transport to private car. In particular we look at
13
accessibility to jobs in the morning, and compare between the accessibility before
and after the bus reform of 2011. For clarity of the presentation we show the results
only for the city of Tel-Aviv-Yafo, but accessibility is measured to all the jobs in the
metropolitan area.
The next step in the research, described in chapter 3, is the theoretical definition of
measures of inequity for transportation accessibility. Here we go back to the historic
work of welfare economics, and use two measures defined in the beginning of the
20
th
century to examine inequality of distribution of goods: the Lorenz Curve, and the
Gini Index. We illustrate the use of these tools in evaluating the equity of accessibility
between areas in Tel-Aviv before and after the bus reform. We conclude this chapter
with theoretical equations for estimating relative and absolute accessibility loss as a
result of traveling by public transport as opposed to a private car.
In order to evaluate the impact of accessibility loss on those people who cannot
drive, or do not have access to private transportation, we move in chapter 4 to
describe ways to estimate public transport dependency. After reviewing the literature,
we develop a public transport dependency index for each Travel Analysis Zone
(TAZ). This measure is an estimate of the level of public transport dependency, in
each zone, due to its demographic and socio-economic characteristics. Using the
Lorenz Curve and the Gini index, we show that the spatial distribution of public
transport dependency is somewhat less equal than the distribution of income in the
metropolitan area.
Having estimated accessibility inequity on the basis of the transportation
infrastructure, and on the basis of the demographic and socioeconomic
characteristics of TAZ, we apply the theoretical formulations of chapter 3, to calculate
the relative and absolute accessibility loss due to both of them. This is done in
chapter 5, and is applied to the impact of the bus reform on accessibility to jobs for
trips originating in the city of Tel-Aviv Yafo. The results show, that although there is
an overall improvement in accessibility to jobs from Tel-Aviv Yafo, there is no
relationship between Transportation Dependence and Relative Accessibity the
system is not designed to advantage in some way those that need it most.
In chapter 6, we turn to a different way of measuring the social impact of the
transportation system. This method uses another concept from the socioeconomic
literature to define inequity poverty, and the poverty line. Lack of accessibility may
cause a lack of possibility to engage in life opportunities, i.e. poverty in terms of
activity participation. The related accessibility poverty line can be set at different
levels, depending on policy goals and the standards a society feels committed to
uphold. The concept of accessibility poverty is open in nature and in principle also
allows the assessment of accessibility poverty experienced by persons with access
to a car. In the report, however, we only focus on the accessibility poverty
experienced by the public transport dependent population as defined according to the
index described above. The chapter illustrates the use of this concept to examine the
impact of the bus reform on Tel-Aviv. The chapter shows that the overall increase in
public transport accessibility due to the reform has also led to a substantial reduction
in accessibility poverty experienced by the public transport dependent population.
14
This holds for the ‘compassionate’ as well as the ‘harsh’ accessibility poverty line
applied in the report.
Finally, in chapter 7, we turn to look at future possible extensions of the method, and
its incorporation into practical transportation planning work. We briefly sketch out how
our concepts could be extended to include two aspects of the transportation system
that are crucial to its performance. The first is the accommodation of the variance of
travel times, and the second is the incorporation of varying demand on the
accessibility calculation, which so far were done on all potential destinations, and not
accounting for real demand.
In chapter 8 we summarize and conclude this report, highlighting the major findings
and possibilities for future research.
15
2 Measuring accessibility at the human level
2.1 Definition of accessibility
Accessibility is understood as the ability of people to access necessary or desired
activities by different transportation modes (K.T. Geurs, J.T. Ritsema van Eck, 2001;
Garb and Levine, 2002). It is specified as well by potential destinations and the
easiness of reaching them. The more destinations that could be achieved the higher
the level of accessibility (Handy and Niemeier (1997). Accessibility is identified as a
key criterion to assess the quality of transport policy (Kenyon, 2002). It is an
important characteristic of metropolitan areas, and is considered the reason for
people’s relocation to metropolitan areas (Handy and Niemeier (1997).
Different types of accessibility measures have been developed. (Handy and Niemeier
(1997), Geurs and Ritsema van Eck (2001), Geurs and van Wee (2004). They are
dependent on the number of opportunities and activities in which an individual can
participate, the relative utility that people derive from access, and the cost and travel
time of reaching them (Handy and Niemeier (1997). Geurs and van Wee distinguish
four types of accessibility measures:
a) Infrastructure-based service level of transport infrastructure (e.g. “the
average travel speed on the road network);
b) Location-based accessibility of location (e.g. “the number of jobs within 30
min. travel from origin locations”);
c) Person-based individual accessibility (e.g., “the number of activities in which
an individual can participate at a given time
d) Utility-based the economic benefits that people derive from access to the
destination
The available literature provides excellent examples that apply measures to assess
accessibility (Shen (1998), Blumenberg and Ong (2001), Hess (2005), Kawabata and
Shen (2006) and Kawabata (2009).
The growing interest in sustainable development has emphasized the importance of
public transport accessibility. Accessibility is an essential yardstick for each of the
parts of sustainable development: economic development, environmental quality and
social equity. In case of economic development accessibility enables the exchange of
people and goods and thus to enable the functioning of the economy (Bruinsma,
Nijkamp et al. 1990); social goals relate to access to employment, goods, services
and social contacts (Bertolini, 2005). For the environmental quality dimension,
improvement of the accessibility levels with alternative modes of transport can
reduce car-dependence of metropolitan areas (Shen 1998; Blumenberg and Ong
2001; Hess 2005; Kawabata and Shen 2006; Kawabata 2009; Kwok and Yeh 2004).
This research is focused on the use of an accessibility measure to address the equity
dimension of transportation. The reason for this is that a lack of accessibility has a
serious impact on people's way of living and may prevent them from participation in
the labor market, reaching health care services, as well as having enough contacts
with friends and family (Lucas 2012).
16
The literature distinguishes two forms of accessibility: person-based accessibility and
place-based accessibility (Pirie 1979; Kwan 1999; Miller 2007). Person-based
accessibility refers to a persons' attributes, a person has (or has not) accessibility to
a given set of places. Place-based accessibility is a location (activity) attribute, a
place is accessible (or not) for a given set of people or from a given set of other
locations (Martens, 2012). The perspective taken in this research is directed more to
person accessibility rather than place accessibility as people are the recipients of
social goods.
Accessibility is also related to freedom of choice, and hence to achieving a higher
level of personal fulfillment and satisfaction (Kwan and Hong 1998). In this sense,
accessibility links to mobility 'the ease with which a person can move through space'
(Martens, 2012), or better, to potential mobility which is related to freedom of
movement or the potentiality of movement (Kaufmann 2002, pp. 1314). Personal
accessibility is different from potential mobility; it refers to the ease with which a
destination could be reached within given location and time. (Farrington and
Farrington 2005; Dong et al. 2006; Niemeier 1997). An increase in mobility means
that a person could travel over longer distances or more frequently or both (Sager
2005, pp. 34). Within a higher level of potential mobility more possible opportunities
could be achieved (Kaufmann, 2002).
The social justice perspective of sustainable development uses accessibility as a key
policy indicator of distribution of benefits and burdens over members of society such
as access to employment and critical services (Farrington and Farrington 2005).
Another justice criterion is the principle of need (Sen, 1973). This criterion indicates
that one individual or group may need higher accessibility level than other (Cass et
al. 2005).
2.2 An operational definition
Accessibility is a function of all three components of the land-use transportation
system: a) land-use distribution of jobs and activities, population densities and socio-
economic characteristics of people and their distribution in space and time; b) the
transportation system with its road configuration, modes of travel, and the time, cost
and difficulty of travel from and to any place within the metropolitan area; and c) the
demand and benefits that individuals obtain from traveling to different destinations.
This is a complex system that changes and adapts, reciprocally. However, the speed
of change and adaptation is different for each component in the system. The land-
use system changes most slowly over a period of years, even decades, in
response to demographic and transport infrastructure change thus we can assume
for this research that it is relatively constant in the short run. Individuals adapt quickly
to changes in the other systems and to a great degree are not predictable. We thus
concentrate on accessibility as representing their potential movement choices, and
not necessarily on predicting their actual movements. The transportation system is in
this study the variable most sensitive to policy changes, and therefore we
concentrate on evaluating the impacts of changes in this system on people’s
accessibility potential to land use destinations.
17
Our measure of accessibility is itself relational. We measure the number of
destinations of interest that are available for a person, within a reasonable time frame
(between 30 minutes to an hour) using different transportation modes (usually private
car and public transport, but it could be done also for bicycles and pedestrian
movement).
We define the Mode Access Area (MAA
O
(t)) as: all the destinations that could be
reached using a particular mode (M) from origin (O) within a particular time frame (t).
We define the Mode Service Area (MSA
D
(t)) as: all the area that is served by a
particular destination (D) using a particular mode (M) within a particular time frame
(t).
Given an origin O and destination D, we consider:
Public Transport Access Area PAA
O
(t)
Private Car Access Area CAA
O
(t)
Public Transport Service Area PSA
D
(t)
Private Car Service Area CSA
D
(t)
We measure relative accessibility between private car and public transportation by:
Access areas ratio: AA
O
(t) = PAA
O
(t)/CAA
O
(t)
Service areas ratio: SA
D
(t) = PSA
D
(t)/CSA
D
(t)
The following sections describes how CityGraph, the program developed within a
GIS environment is used to calculate and map these areas in a metropolitan area,
and how these results are further analyzed to calculate the relative accessibility, and
subsequent social inequity.
2.3 CityGrapha GIS-based tool for measuring accessibility
2.3.1 General view
CityGraph is the 3rd generation application based on the previous Urban.Access and
AccessCity applications (Benenson et al, 2010). The main advancement made is the
dramatic increase in the level of spatial resolution on one hand and the computation
speed on the other hand.
CityGraph works at the resolution of individual buildings, bus stops and lines and,
thus, processes very large amounts of GIS data. In a typical metropolitan area we
have 100,000 1,000,000 buildings and between several hundreds and a thousand
PT lines. Tel-Aviv metropolitan, for example, has a 2.5 million population, more than
250,000 buildings and 300 bus lines. This results in 10
9
- 10
10
(1 10 billions)
potential OD pairs. CityGraph uses the metropolitan street and public transport
networks. All available information is fully exploited - traffic directions and speeds,
explicit position of PT lines and stops to estimate accessibility by car and by the PT.
Bus timetable data is a critical non-spatial component necessary to estimate
accessibility by bus. Activities are estimated based on spatially-explicit data on job
distributions, socio economic indicators, car ownership and land use composition.
OD matrix for the Tel Aviv Metropolitan, supplied by NETA at resolution of TAZ for
18
the morning and day hours and for the trips with the public transport and a car is
exploited as an estimate of trip demand.
2.3.2 CityGraph as a graph theory-based application
CityGraph applies graph theory algorithms to estimate the Access and Service areas
for car and bus. First, every building is interpreted as a B-node. The road network
topology is directly transformed into directed graph: junctions into nodes, street
section into links (two-way segments are translated into two links) and travel time into
the impedance.
PT network depends on the time tables and the PT graph is thus constructed for a
given time interval. The PT network is translated into a directed graph based on
quadruples
<PTLINEID, TERMINALDEPARTURFTIME, STOPID, STOPARRIVALTIME>
that are interpreted as a graph PT-nodes. Two PT-nodes
N
1
= <PTLINEID
1
, TERMINALDEPARTURFTIME
1
, STOPID
1
, STOPARRIVALTIME
1
> and
N
2
= <PTLINEID
2
, TERMINALDEPARTURFTIME
2
, STOPID
2
, STOPARRIVALTIME
2
>
in a PT graph are connected by the link L
12
in two cases:
1. PTLINEID
1
= PTLINEID
2
, TERMINALDEPARTURFTIME
1
=
TERMINALDEPARTURFTIME
2
and the stop STOPID
2
is the next to STOPID
1
on
the bus line PTLINEID
1
. The impedance of the link L
12
is equal to
STOPARRIVALTIME
2
STOPARRIVALTIME
1
.
2. The walk time W
12
between N
1
and N
2
is less that the maximal possible walk time
WALK
max
and STOPARRIVALTIME
2
STOPARRIVALTIME
1
< WALK
max
+
WAIT
max
, where WAIT
max
is a maximal waiting time at a stop
In a road graphs, B-node is connected to the road node and a road node is
connected to a B-node if the walk time between them is less than W
max
.
In a PT graph a B-node is connected to a PT node N
1
, given the trip start time T
start
,
the walk time between the B-node and a PT-node is less than WALK
max
, and
STOPARRIVALTIME
1
> T
start
+ WALK
max
and STOPARRIVALTIME
1
< T
start
+
WALK
max
+ WAIT
max
.
The link impedance between the B-node and road node is equal to a walk time
between them. For a PT-graph it is a travel time between stops or walk time between
building and stop plus waiting time to the arriving bus
Figure 1 presents a description of the process of translation of a typical bus journey
from origin to destination into a sequence of connected links
CityGraph is developed with the help of the Neo4J graph database
(http://www.neo4j.org/
) as depicted in Figure 2.
19
PT-node:
<PTLINEID = 1,
TERMNALDEPARTURETIME = 6:57,
STOPID = 4,
STOPARRIVALTIME =
Figure 1: Translation of the public transport network to a graph
20
2.4 Accessibility in Tel-Aviv Metropolitan Area
The following figures show the various measures of relative accessibility in the Tel-
Aviv Metropolitan area. First we calculate accessibility by car to jobs. Second, we
calculate the accessibility by bus before the bus reform and calculate the relative
accessibility. Then we compare the pre- and post bus reform accessibility levels. All
calculations are made for the morning hours, for the travelers who are willing to start
their travel Sunday morning, between 7:00 and 7:15. Maximal aerial distance
between the traveler's origin and bus stop or between bus stop and traveler's
destination is set 400 m and maximal waiting time at transfer is 10 minutes.
The accessibility maps below are built for all origins in the City of Tel-Aviv, traveling
to all of the jobs in the metropolitan area. Due to computational limitations, we prefer
to perform these calculations for the separate city or group of cities that contain up to
100,000 buildings but not for the entire metropolitan area with its more than 250,000
buildings..
Figure 3 shows the number of jobs throughout the metropolitan area
accessible by car from the city of Tel-Aviv departing at 7:15 am, and taking 15
to 45 minutes.
Figure 4 shows the number of jobs accessible by bus throughout the
metropolitan area and originating in Tel-Aviv starting out at 7:15 am and
taking 15 to 45 minutes, before the bus reform of 2011.
Figure 5 shows the relative accessibility to jobs between public transport and
car in the same 15-minute intervals.
Almost all the jobs in the metropolitan area are reachable by car in 45 minutes from
all of the city of Tel-Aviv-Yaffo (Figure 3). Accessibility by bus is lower, particularly for
shorter trips (Figure 4). The clear advantages of central locations, where relative
accessibility by bus reaches high values approaching 1 can be seen in Figure 5.
Figure 6 shows the number of jobs accessible by PT throughout the
metropolitan area, originating in Tel-Aviv at 8am and taking 45 minutes - after
the 2011 bus reform.
Figure 7 shows the relative accessibility to jobs in the metropolitan area
originating in Tel-Aviv at 7:15 am on the new bus network.
Finally, Figure 8 shows the gains and losses in accessibility to jobs as a result
of the bus reform.
We can see from the map in Figure 8 that for trips originating in the city of Tel Aviv
the bus system reform is generally positive and positive effects essentially increase
with the increase in a trip duration (Table 1). However, the results are not clear cut
particularly for short trip durations. Some areas, e.g. Shderot Yerushalaim in Jaffa,
and the Old North area around the municipality building particularly stand out.
21
a b c
Figure 3: No. of jobs in the Metropolitan Area accessible by car from Tel-Aviv starting at 7:15am. in 15 minutes (a), 30 minutes (b), and 45 minutes
(c)
22
a b c
Figure 4: Before the bus reform - No. of jobs in the Metropolitan Area accessible by public transport from Tel-Aviv starting at 7:15 and taking 15
minutes (a),30 min. (b) and 45 min. (c)
23
a b c
Figure 5 : Before the bus reform - Relative accessibility at 7:15 am, for 15 min. (a), 30 min. (b) and 45 min.(c) long trips.
24
a b c
Figure 6 : After the bus reform - No. of jobs accessible from Tel-Aviv in the metropolitan area starting at 7:15 and taking 15 minutes (a), 30
min. (b) and 45 min. (c).
25
a b c
Figure 7 : After the bus reform - Relative accessibility to jobs in TAMA from every building in Tel-Aviv at 7:15 am for trip lengths of 15 (a)
30 (b) and 45 minutes (c)
26
a b c
Figure 8 : New versus the old bus network - Difference in relative accessibility in the Tel Aviv Metropolitan area for every building in Tel
Aviv Yaffo for a trip of 15 (a), 30 (b) and 45 (c) minutes
27
2.5 Justification for high spatial resolution of accessibility analysis
We are well aware that most transportation planners use TAZ as the most basic level
of analysis mainly due to availability of the data and lack of high-end computation
facilities. We argue that for evaluating accessibility by public transport a high
resolution view of the accessibility pattern is essential. The justification comes from
the high intra-variability in the TAZ themselves. We demonstrate this in Figure 9,
which shows substantial intra-zonal variation. The reason is that the walking and
waiting time at the bus stops takes up a significant share of the total door-to-door
travel time by public transport. Variability of walking speed enables background
estimate of the travel time coefficient of variation (CV). The variation is relatively high,
~ 15%, for the short trip of 15 minutes and decreases to 5-6% for 45 minute trips.
The variation of the travel time can be translated into variation of accessibility
measures and results in the similar background estimates of the measures' "inherent"
variability. Comparing these background values to the empirical estimates we can
see from Figure 9 that the value of the CV for many Tel Aviv TAZ areas is far above
the threshold value. This implies that analysis at the zonal level will ignore that
different people in the same TAZ will experience different levels of accessibility. The
outcome is that analysis at zonal resolution will overestimate the patronage of public
transport especially for shorter trips.
Thus, as described above, zonal analysis will not capture well the penalizing effects
of walking and waiting time on public transport ridership accurately enough. It is likely
that this overestimates the share of public transport in the mode choice model
applied at the TAZ level.
Table 1: Statistics of the difference between the values of relative accessibility for the
new and old networks, as dependent on the duration of trip
Duration of trip, minutes 15 30 45
Average 0.0033 0.047 0.078
STD 0.096 0.102 0.086
Minimum -0.73 -0.40 -0.49
Maximum 1.11 0.74 0.75
28
a b c
Figure 9: New bus network Coefficient of Variation (CV = 100%*STD/MEAN) of the relative accessibility for buildings within each TAZ.
Relative accessibility is estimated based on trips to the entire Tel Aviv Metropolitan area, for every buil
ding in the city of Tel Aviv. Duration
of trip is 15 (a), 30 (b) and 45 (c) minutes
29
3 Measuring accessibility inequity
3.1 Inequity in transportation planning
There are two general categories of transport equity: horizontal equity and vertical
equity (Currie, 2011; Litman, 2007). Horizontal equity (fairness or egalitarianism) is
concerned with providing equal resources to individuals or groups assumed as equal
in ability. It avoids favoring one individual or group over another and services are
provided equally regardless of need or ability. Vertical equity (social justice,
environmental justice or social inclusion) is concerned with distributing resources
between individuals of different abilities and needs. Vertical equity favors groups
based on socioeconomics or specific needs in order to make up for overall societal
inequalities. These two frameworks reflect two contrasting perspectives in public
transport planning. The horizontal equity framework focuses on distributing services
to the maximum number of users; it encapsulates the ‘‘mass transit’’ perspective
(Betts, 2007). The vertical equity perspective is also called the ‘‘social transit’’
perspective (Betts, 2007) as it prioritizes specific user groups. Its goal is to provide
transit access to those with greatest ‘‘need,’’ e.g. people without private transport or
specific demographic groups such as the low-income, youth or ethnic minorities
(Murray and Davis, 2001; Ward, 2005; Garrett and Taylor, 1999; Deakin, 2007;
Taylor et al,.
2009; Sanchez et al., 2007; Litman, 2007). Recently there has been
renewed interest in equity in transport contexts in particular for public transport
provision (Welch and Mishra, 2013; Feng and Zhang, 2012; Pangbourne et al. 2011;
Levinson, 2010).
3.2 Measuring horizontal inequity: The Lorenz curve and Gini index
In economics, the Lorenz Curve is a graphical representation of the cumulative
distribution function of wealth across the population (Lorenz, 1905). It can be applied
not just to income but to any quantity that can be cumulated across a population.
They have been applied in a range of disciplines, from studies of biodiversity to
business modelling and even within transport (Fridstrom et al., 2001). Fig. 6 is a
hypothetical example of a Lorenz curve. The blue line represents a population of
perfectly equitable income distribution; the red line represents an inequitable
distribution of (e.g., 70% of the population shares about 25% of the population’s
income).
30
It is important to note that the Lorenz Curve does not imply that perfect equity is
possible or even desirable.
The Gini index (Gini, 1912) is a statistical measure of the distribution of an attribute
(e.g. income). Whereas the Lorenz curve is a visual representation of inequality, the
Gini index is a single simple mathematical metric representing the overall degree of
inequality. Graphically, it is a ratio of the area between the line of equality and the
Lorenz curve (marked ‘‘A’’ in Figure 10) divided by the total area under the line of
equality (A + B in Figure 10). Using Gini indexes the distribution of two different
Lorenz curves can be mathematically compared. Statistically, Gini is a measure of
equity variance, computed as half of the Relative Mean Differenceof the value of
an attribute between two randomly chosen objects. That is, Gini as such is one of the
measures of variation (as the STD is) and is independent of the average value of the
distribution. For a rank-ordered sample S of the population i.e. y
i
< y
i+1
, the Gini index
is computed as:
The Gini index is always between 0 and 1. A value of 0 implies complete equality
whereas a value of 1 suggests complete inequality. The lower the value the more
equal is the distribution in question.
The Lorenz curve and Gini index have been known for many years as a
representation of inequality and its aggregate measure. Gini type indexes have been
extensively applied not only for distribution of income but also for education and
health as well as biodiversity. For any attribute Y the Lorenz curve and Gini index are
simple to compute based on the distribution of Y in population and they are scale
independent i.e. they can be based on the value of Y for individual objects (e.g.
Figure 10 : Representation of the Lorenz curve
31
buildings) and for the aggregates of objects (say, average values of Y over statistical
areas)
.
We suggest computing the Lorenz Curve and calculate the Gini index for two
populations: (a) G
T
- for entire population of interest (e.g. residents of the area,
elderly people, etc); (b) G
B
- for public transport users only. In both cases, Lorenz
curves and Gini index will be constructed for the same attribute - bus/car access and
service area ratios calculated for every object. The Gini index and Lorenz curve allow
easy comparisons between two alternative transport systems e.g. before and after
the Tel-Aviv bus reform. G
T
, constructed for the entire city or city quarter, allows
comparing PT accessibility inequality between different cities and or between regions
within the same city. G
B
may perform better if we can estimate the bus users'
population before and after the reform.
3.3 Demonstration: the impact of bus reform on horizontal equity
We can illustrate the use of the Lorenz Curve and Gini Index by examining them for
the old and new bus systems and their impact on the trips originating in the city of
Tel-Aviv (Figure 11).
Figure 11: Lorenz curves and Gini indices for relative job accessibility using the Old
and New bus network in Tel-Aviv.
32
It’s clear from the figure, that despite the overall improvement in accessibility, the
inequity in distribution of accessibility grew somewhat within the city of Tel-Aviv as a
result of the reform. According to Figure 11, the improvement of accessibility for the
low accessible areas was insufficient to equalize them with the high accessible ones
3.4 Moving from variation in accessibility to measuring magnitude
of inequity
Although the Lorenz curve and the Gini index provide a visual and quantitative
representation of inequity, scale independence remains an issue. This problem
arises especially when two accessibility distributions are similar but the averages are
very different. In this case neither the Lorenz curve nor Gini index will show any
difference between the two evaluated systems, although clearly in terms of scale
they are so. As such we need also to examine the average accessibility. Moreover,
we need also to generate new measures which capture the magnitude of difference.
For this end, two new measures are introduced: The relative loss of accessibility and
its total loss.
3.4.1 Relative accessibility loss
This measure is computed as:

=
1




for access
or

=
1




for service areas for any origin or
destination object for a trip duration of t minutes and
P
B
/P
T
is the public transport use
share (e.g. estimated from the mode choice model). R allows ranking regions and
population groups by their accessibility. If R = 0 (i.e. no loss), there are either no bus
users, or bus and car service areas are equal. If the entire population uses the bus
then R = 1. R can be estimated at any spatial level from individual buildings up to the
entire metropolitan area.
3.4.2 Absolute accessibility loss
This measure is computed as

=
(




)
for access areas,
or

=
(




)
for service areas.
L measures the total population loss of accessibility caused by the use of a public
transport instead of car. L is calculated for a certain population group or population of
a region and allows assessment of a total gained/lost accessibility due to
improvement of the PT. If all group members use cars then L = 0. If service areas by
bus and car are equal, there is no loss of accessibility and L = 0. , if all group
members use bus, L is very high and equal to the number of travelers multiplied by
the job losses for each of them (i.e. the difference in number of jobs accessible by
car and public transport). Different from R, L depends on the size of the public
transport users' population P
B
and, thus, allows comparison between the outcomes of
essential PT improvement that is available to a few passengers and the minor PT
improvement that is available to many passengers.
An example of the use of these indices to calculate the relative and absolute
accessibility losses in Tel-Aviv due to public transportation dependence is shown in
chapter 5.
33
3.4.3 Advantages and disadvantages of the proposed indices
Advantages of the proposed indexes:
Horizontal comparisons across all or some of the users
Vertical comparisons along spatial and non-spatial categories e.g. different
user groups / neighborhoods
Parsimony: Reflects only one component of accessibility
Flexibility: Can accommodate any change in the PT system independent of
demand data availability.
Universality: Can be applied to any parameter-free accessibility measure:
e.g. access area, travel time etc.
Potentially highly sensitive to PT level of service changes
Disadvantages of the proposed measures
Context dependence: indexes are computed for typical trip purpose or
destination type, e.g. jobs or hospital.
Oversimplicity: Reflects only one component of accessibility
Some comparisons are dependent on availability of mode choice model
Time-of-day dependency: PT accessibility during morning peak can be very
different than off peak or evening, with different frequencies and travel times.
Can be resolved by averaging, in different way (weighted inequality, minimal
inequality, maximal inequality over time.
34
4 Estimating public transport dependence
So far we have been discussing horizontal equity measures of accessibility. The
following section discusses measures for estimating vertical equity of accessibility
based on needs or dependence on public transport.
The growing interest in the issue of transport disadvantage and how it relates to
social exclusion stimulated studies that began to show the interrelationship between
poverty, transport disadvantage, access to life essential activities and services, and
transport related social exclusion (Lucas et al., 2001, Kenyon, 2003, Kenyon et al.,
2003, Lucas 2004, Currie et al., 2007, Delbosc and Currie, 2011). The concept of a
transport disadvantaged population has grown from the fact that traditional
transportation planning methods usually aimed to satisfy travel demand and do not
take into consideration social-economic aspects (Hine and Mitchell, 2001). However
analysing public transport disadvantage and social needs together with transport
provision are essential in order to see where the system can be improved (Lucas,
2004; Fainstein, 2005; Currie, 2009).
Public transport disadvantage measure is recognized as a complex multi-dimensional
characteristic of location and individuals. The concept includes a set of individual
characteristics related to location access, access to mobility, and personal
circumstances, such as physical and social characteristics that could limit personal
access (Delbosc, et al, 2011, Lucas 2004).
First, transport disadvantage measure is related to location analysis. This measure
requires a determination of travel time, cost and distance to key life opportunities
such as employment, medical centres, shops, education centers and social networks.
(Church et al., 2000, Schonfelder and Axhausen, 2003, Dodson et al., 2006).
However, taken alone this system-based approach to travel provision misses the
population needs and different circumstances of participants that could influence to
the quality of access (Delbosc, 2011).
Another parameter of transport disadvantage is a mobility based measure that
considers a set of indicators including level of car ownership and public transport
service (Hine, 2004, Hurni, 2005 and Currie, 2009). Thus individuals without access
to a car (Clifton and Lucas, 2004) and poor public transport provision are considered
as transport disadvantaged (Currie, 2009).
A third aspect is related to socio-economic aspects and specific groups of people and
their needs. Those with physical or mental impairments (Casas 2007), the elderly
(Rosenbloom and Morris, 1998), (Siren, 2007), unemployed youth (Currie,
2007, Hurni, 2007), single parents (Fritze, 2007, Hurni, 2007, Titheridge and
Solomon 2008), people with low-income, or with cultural and language barriers, etc
(Litman, 2010).
The literature provides examples that consider these aspects separately and
together. Lucas, 2004 proposed an approach to combining transport disadvantage
35
and accessibility assessment. She argues that policy makers should take into
account individual circumstances, interactions between people, activities that they
need and would like to attend and the transport options that they have. She claims
that transport and social disadvantages interact and they cause transportation
poverty that leads to inaccessibility to essential activities and thus to social exclusion.
Therefore, both a social disadvantage index (income, employment status, skills level,
health problems, poor housing) and transport disadvantage indices (car ownership,
poor public transport services, high cost of fares, no information, fear of crime, etc)
should be taken in account together (Currie, 2010, Lucas, 2004).
For a broader analysis of public transport supply, Currie (2004, 2009, 2010)
proposed an aggregate indicator based on social needs with an analysis of
disadvantaged groups using recent census data. He elaborates a 'needs gap' index
that spatially compares locations of public transport supply and service quality with
demand, that was estimated from socio-economic factors. The measure consists of
weighted indexes of social and transport needs that combines characteristics of
income, employment status, car ownership and health together (Casas, et al.,
2007, Currie, 2009, Currie, 2010, Delbosc and Currie, 2011). The weights of the
parameters were based on the results of transportation preference studies.
Table 2: Public transportation dependence indicators according to Currie 2010
Need indicator
Weight
Adults without cars
0.19
Accessibility
0.15
Persons aged over 60 years
0.14
Persons on a disability pension
0.12
Low income housholds
0.10
Adults not in the labour force
0.09
Students
0.09
Persons 5-9 years
0.12
One parameter that is not available from Census is accessibility. Accessibility is a
measure of locational opportunity or disadvantage. The accessibility measure used
by Currie was the straight line distance to Melbourne central business distric (GPO)
from the CCD centroid (Currie, 2007). All the rest of the indicators are taken from the
availabe census. Applied weightings are derived from an analysis of travel survey
accounts since they relate to these census indicator groups. Weights that applied to
indicators are based on the degree of low trip making (Currie, 2007). A score for
each need is identified by standardisation of each value of applied need indicators.
Standardisation is based on the relationship between the score and highest value of
this indicator. Each standardised value is weighted and resets the scores between
the values 0 and 100. The final need index is produced by adding together the
weighted standardised values.
Then measures are aggregated and a comparison is made between transport needs
and public transport provision in order to identify spatial gaps between needs and
supply, in order to show where transport system improvements should take place.
36
The structure of the index proposed by Currie was used for Australian cities and later
applied for Palermo (Italy) (Amoroso et al., 2010). The methodology was also used
for Santiago de Cali, Colombia (Jaramillo, 2012), however with modified weightings
more in accordance with its different socio-economic conditions. The basic structure
taken to calculate the index for transport social needs is the index proposed by
Currie. For Santiago de Cali the index has been calculated for the districts of the city
due to the fact that there is no available data on transport analysis areas. The index
itself is a weighted summary of transport disadvantage indicators of different factors
within the district. These factors are showing the most social need for better public
transportation. Indicators that represent transport disadvantage of each factor of
public transport need are standardized between values 0 and 1 (Jaramillo, 2012).
4.1 Transportation need index calculation for Tel-Aviv Metropolitan
Area
Transportation need is defined as the number of people in a given geographic area
who are likely to require a public transport service. The measure is the index of
transportation social needs related to transport disadvantages for each of the
transportation analysis zones (TAZ) of Tel-Aviv Metropolitan area. For the calculation
of transportation needs index, the method proposed by Currie, was taken as a
baseline, modified to the available data and socioeconomic conditions relevant to
Israel. The index itself is a weighted summary of transport and social disadvantage
indicators within the area of analysis. The final index consists of the following need
indicators: adults without cars (expressed by the difference between the number of
adults and the number of cars in the TAZ), persons aged over 65, persons with
disabilities, low income households (below median income), unemployed persons (as
defined by the CBS - aged over 15 without a job), students, and persons aged 10-18.
These statistics are used by NTA in their model to determine mode choice within
TAZ. However, lacking a detailed study of the likelihood of taking public
transportation based on socio-economic characteristics, we assume that the most
important factor among all the factors involved is the lack of access to a car by
persons of driving age, therefore we have given this factor and income factor highest
weight, and allowing all the other factors an equal weight.
Weighting was applied in the following way: 25% for adults without cars, with the
balance divided equally among the other groups (13%), Table 2.
Table 3: Need indicator weights
Need indicator
Weight
1
Adults over 18 without cars (Census, 2010)
0.25
2
Persons aged over 65 (Census, 2010)
0.13
3
Persons with disabilities (Census, 2008)
0.13
4
Low income households (below median income) (Census,
2008)
0.13
5
Persons over 15 without a job (Census, 2010)
0.13
6
Students (Census, 2010)
0.13
7
Persons 10-18 (Census, 2010)
0.13
37
The overall index is calculated as the weighted sum of the indicators for all chosen
factors within each TAZ. Relatively disadvantaged areas are those with higher index.
The index is calculated as the following:
NS
TAZ
= SI
1
*W
1
+ SI
2
*W
2
+ …+ SI
n
*W
n
Where:


Transportation Need Index for the TAZ under analysis;
SI1 SIn standardized indicator
W1-Wn are the weights for the indicators
All indicators of transport disadvantage are standardized so they receive values
between 0-100. Relatively disadvantaged areas are those with higher index score
and advantaged are those with lower index.
Equation for standardization: SI = 100*(I - I
min
) / (I
max
- I
min
)
SI Standardized Indicator
I Indicator of the disadvantage factor for the TAZ
I
min
Minimum value for the indicator for the TAZ
I
max
Maximum value for the indicator for the TAZ
The method requires demographic data. For this project, Israeli Bureau of Statistics
was chosen as a source of data (ICBS). Parameters that indicate persons with
disabilities and income of households are taken from Census, 2008 and only
available for TAZ numbering 2,000 residents and over. Parameters for localities with
the population less than 2000 residents were taken from the subquarter level. All
other indicators are taken from Census, 2010 and are available for all TAZ.
4.2 Public transport dependency results
The computed need indicators are shown in Figures 12-18. The data exists only for
urban localities in the metropolitan area (cities and local authorities). The grey areas
are areas where no people live,
One can clearly see that the different variables are distributed differently throughout
the metropolitan area, and do not necessarily coincide.
38
Figure 13: Adults without cars
Figure 12: Unemployed adults
Figure 14: Persons 10-18 years old
Figure 15: Persons above 65 years old
39
Figure 16: Persons with disabilities
Figure 17: Number of students
Figure 16: Persons with income below
median
40
Figure 19 shows the spatial distribution (in deciles) of the transportation need index
for the Tel Aviv Metropolitan area. and the city of Tel-Aviv (enlarged on the left). The
relative disadvantage of outer ring towns and suburbs, as well as the North-South
divide in the metropolitan area and within the city are well observable.
Figure 17: Pubic transportation need index in the Metropolitan area and the city of Tel-
Aviv
Figure 20 illustrates the share of values of each need indicator factor. The result is
presented for the 20 highest need TAZ. Usually the two most significant factors in
determining dependency on public transportation are low income and not owning a
private vehicle. Although in the case of the two most dependent TAZ: a TAZ in
Ashdod and Elad, these are compounded by pensioners above 65 years and a high
population between the ages of 10-18 of age that do not drive.
41
Figure 18: The makeup of weighted Transportation Need Index for the 30 highest need
TAZ
Figure 21 below shows the Lorenz curves and Gini Indices for the transportation
need index and income index. It shows that the inequality of distribution of public
transport need is almost the same as the inequality of spatial distribution of income in
the Tel-Aviv Metropolitan Area.
42
Figure 19 : Lorenz curves and calculation of Gini index for Public Transportation Need
and Income by TAZ
Table 4 present a summary of the individual need indicator of transportation need
index using the methodology described above. The need index was assembled into
deciles; highest score represents the highest need. The population for each TAZ is
also identified. An interesting phenomenon is that the higher the need index, the
larger are the share of the population of the TAZ. This is in itself worthy of further
investigation.
0
10
20
30
40
50
60
70
80
90
100
0 10 20 30 40 50 60 70 80 90 100
INDEX
TAZ
Lorenz curves and Gini Index for Tel-Aviv Metropolitan
Equality line
Need Index
Income index
Need index
Gini = 0.36
Income index
Gini = 0.40
43
Table 4: Number of TAZ and population in each category of need index
Index decile Number of TAZ Population
Population
percentage
0 no people 260 0 0
1 (0.0 - 10.1) 96 132,923 3.97
2 (10.2 - 13.6) 96 159,404 4.76
3 (13.7 - 16.9) 96 212,373 6.34
4 (17.0 - 19.8) 96 247,553 7.40
5 (19.9 - 22.6) 96 290,231 8.67
6 (22.7 - 25.5) 96 332,801 9.94
7 (25.6 - 28.8) 96 341,358 10.20
8 (28.9 - 32.3) 96 393,929 11.77
9 (32.4 - 37.9) 96 450,893 13.47
10 (38.0 - 100) 95 786,051 23.48
Total: 1219 3,347,516 100.00
4.3 Conclusions
Overall more than 23% of the population or 786,051 residents have relatively high
public transportation dependency score, which is above the mean score in the
Metropolitan Area.
The highest transportation need score zones are:
In the outer ring of the metropolitan area.
In an axis starting from Tel-Aviv Yaffo’s southern neighborhoods and
extending to Bat-Yam and towards the Southeast
In another axis starting from Tel-Aviv Yaffo’s eastern neighborhoods and
towards the Northeast.
The clear North-South divide in the central city is continued in the
Metropolitan Area as a whole with the Southern part showing greater
dependency.
44
5 Evaluating accessibility loss due to public transport
dependency
Having calculated the relative accessibility to jobs in the Metropolitan Area from Tel-
Aviv, and the Transportation Need Index for the entire Metropolitan Area, we can
now proceed to evaluate the accessibility loss due to the reliance on public transport
for travel to work. We will use the equations presented in sections 3.4.1 and 3.4.2
above for relative and absolute accessibility loss, but adapting it to the need index of
each TAZ, and the size of the dependent population.
Thus, in the expression for relative accessibility we use the unstandardized need
index in place of P
B
/P
T
this reflects the ratio of public transport dependent
population to the overall travelling population of the TAZ. In the expression for
absolute accessibility loss we use the dependent population as an estimate of P
B
.
Figures 22 and 23 show the relative (R
i
) and absolute (L
i
) accessibility loss, for the
old and new bus networks of Tel-Aviv. They also show the difference between the
two systems and where these losses were reduced or became larger. The
importance of these maps is that they combine both the horizontal inequality due to
the layout of the public transport system, as well as the vertical inequality due to the
spatial distribution of public transportation dependency throughout the city.
45
a b c
Figure 20: Relative accessibility loss in the City of Tel Aviv for a 45 min trip to all jobs in the Metropolitan Area on the old bus network (a), the new
bus network (b) and the changes in accessibility lost R
O
(OLD) – R
O
(NEW) caused by reforms to the bus network (c). The relative decrease is
positive when R with the new bus system is lower compared to the the old system
.
46
a b c
Figure 21: Absolute accessibility loss L
i
in the city of Tel Aviv for a 45 min trip to all jobs in the Metropolitan Area , for the old (a) and new
(b) bus networks, and the percentage change 100%*[L
O
(OLD) – L
O
(NEW)]/L
O
(OLD) caused by reforms to the bus network (c).
The relative decrease is positive when L with the new bus system is lower compared to the loss with the old system.
47
To examine further the relationship between the public transportation system and
socio-economic dependence on public transport we look at the scatterplot relating
between the Transportation Need Index (as calculated for the entire metropolitan
area in chapter 6), and the accessibility to jobs from the city of Tel-Aviv index for a 15
minute trip at 7:15 am (PAA
O15
). For a system that serves well the dependent
population we would expect a strong positive correlation between the two variables.
We see that there is no correlation between the two variables. Rather there are two
groups of TAZ: one with low accessibility index (below 0.3), which spans the whole
spectrum of public transport need, and the second, with higher levels of accessibility
(above 0.3), which concentrates in the middle to low public transport need index
levels. This means that there is no connection between PT service and the need for
it. Thus, the public transport system in Tel-, which is the city with the highest level of
service in the metropolitan area, does not serve particularly well the people most
dependent on it.
Figure 22: Scatterplot of Transportation Need Index and Relative Accessibility to jobs
index from the city of Tel-Aviv-Yaffo.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.00 10.00 20.00 30.00 40.00 50.00 60.00
Relative Accessibility index
Public Transportation Need index
Accessibility index versus need index
New bus network system, 15 min trips
48
6 Evaluating accessibility poverty
6.1 Methodology
From the perspective of equity, the main purpose of the transportation system is to
enable persons to travel from one place to another in order to participate in out-of-
home activities (cf. Martens 2011). A fair transportation system would enable all
citizens to participate in the activities deemed normal in a particular society, such as
employment, education, or health care. Accessibility poverty occurs if systematic
problems of accessibility to activities lead to significant impacts on a person’s life,
such as unemployment, deterioration of health, or social isolation (e.g., Kenyon,
Lyons et al. 2002; Farrington and Farrington 2005; Lucas 2012; Martens 2013).
Thus, equity does not require equality in accessibility, but it does require a certain
minimum level of accessibility.
A fair transport system, then, is a system that provides every citizen with a sufficient
level of accessibility to participate in activities deemed normal to society (cf. the
social exclusion literature). From a fairness perspective, transport policies should first
and foremost address the accessibility needs of people who do not receive sufficient
accessibility (see for a more elaborate discussion, Martens, forthcoming).
This general formulation encompasses three important elements for the assessment
of the equity of a particular transportation system. First, this assessment requires the
delineation of a level of sufficient accessibility. We term this level the accessibility
poverty line, in analogy with the income poverty line. Second, it requires an
assessment of the intensity of the accessibility poverty experienced by a person. This
intensity is determined by the distance between a person’s accessibility level and the
accessibility poverty line. The larger the distance between a person’s accessibility
level and the poverty line, the stronger the intensity of the accessibility poverty
experienced by that person. Third, the assessment of the equity of a particular
transportation system requires the identification of the number of people that are
experiencing accessibility poverty. Clearly, the higher the number of persons facing a
sub-standard level of accessibility, the less fair the transportation system.
Figure 25 depicts the three components of the assessment of the equity of a
transportation system. The figures (a) and (b) both depict a continuum of accessibility
levels experienced by different persons and groups of a population in a study area,
and each includes an accessibility poverty line. Figure 25 (a) represents the intensity
or depth of accessibility poverty: person P
2
is farther removed from the accessibility
poverty line than person P
1
and thus faces a more severe level of accessibility
poverty than person P
1
. Person P
3
has an accessibility level above the poverty line
and thus faces no accessibility poverty at all. Her situation will therefore not be taken
into account in the assessment of the equity of a transportation system.
49
Figure 23: The three components for assessing the equity of a transportation system
Figure 25 depicts the three components of the assessment of the equity of a
transportation system. Figures 25 (a) and (b) both depict a continuum of accessibility
levels experienced by different persons and groups of a population in a study area
and each includes an accessibility poverty line. Figure 25 (a) represents the intensity
or depth of accessibility poverty: person P
2
is farther removed from the accessibility
poverty line than person P
1
and thus faces a more severe level of accessibility
poverty than person P
1
. Person P
3
has an accessibility level above the poverty line
and thus faces no accessibility poverty at all. Her situation will therefore not be taken
into account in the assessment of the equity of a transportation system.
Figure 25 (b) represents the size of the accessibility poverty. Clearly, the more
people experience accessibility poverty, the less equitable a transportation system is.
In the figure, groups G
1
and G
2
experience an identical level of accessibility. Both are
positioned below the accessibility poverty line, but the number of people belonging to
group G
2
is larger than the number belonging to group G
1
, as depicted by the size of
the circles. All else being equal, a study area with a larger group of persons below
the accessibility poverty line is deemed to provide a less equitable transportation
system. The size of group G
3
does not directly influence the size of the accessibility
poverty, but it does so indirectly: the larger the share of the total population that is
located above the poverty line, the fairer a transportation system. Thus, a society in
which G
1
represents the number of people below the accessibility poverty line may
well be less fair than a society with G
2
number of people below the poverty line, in
case G
1
represents a larger share of the total population than G
2
.
50
How can we quantify the equity of a transportation system? The measure we are
seeking should be sensitive to at least three components of accessibility poverty: the
position of the poverty line, the share of the population that falls below the poverty
line, and the exact level of accessibility experienced by people below the poverty line
(i.e., the distribution of people below the poverty line; see Figure 25 (a)). In order to
define an indicator, we turn to a poverty measure that has been developed in the
domain of income poverty.
In the domain of income poverty, an indicator has been developed in the 1980s
which can capture all these properties. The indicator is called the P
2
measure and
was proposed first by Foster et al. (1984). The measure takes both the intensity and
the size of poverty into account in assessing the level of income poverty in a society
(Ravallion 1992). With q number of subgroups with incomes no greater than the
poverty line z in a total population N, and y
i
representing the income of the ith
subgroup, the P
2
measure is expressed by:
=
1

The value of P
2
ranges from 0 to 1, with a score of 0 indicating the case of an entire
population with an income level above the poverty line, and a score of 1 the case of
an entire population below the poverty line. The larger the P2 score, the larger the
income poverty and the less fair the income distribution in a society.
The P
2
measure has an important benefit: it is decomposable. This means that the
measure makes it possible to identify the contribution of each subgroup i to overall
poverty. In order to determine the contribution of a subgroup to the overall poverty
level, the subgroup poverty level is weighed by its population share and then
expressed as a percentage of overall poverty, which, when contributions of all
subgroups are summed, add up to exactly 100% (Foster et al. 1984; Foster and Sen
1997). By eliminating poverty in one subgroup, the overall poverty level will decrease
exactly by that subgroup’s rate.
The P
2
measure can be used to develop a measure to assess and compare
accessibility poverty within and between areas. Because of its decomposability, the
measure can also be employed to assess the extent of accessibility poverty
experienced by various groups in the population and the extent to which each sub-
group is contributing to overall accessibility poverty (allowing a comparison of, for
instance, the contribution of the urban population versus the suburban or rural
population, or a comparison of different socio-economic groups, such as the Jewish
versus the Arab population). Hence, in the following, we develop our index of
accessibility poverty based on the P
2
measure.
The translation of the P
2
measure to accessibility is a rather straightforward exercise.
It requires the delineation of a poverty line and the distinguishing of population
groups. Clearly, the accessibility sufficiency threshold is identical to the income
poverty line. The population groups have been discussed before. In line with Foster
et al. (1984) it is then possible to define the accessibility poverty in region r (AP
r
) as:
51

=
1

where N represents the total population in region r; q the number groups living in
region r experiencing accessibility levels below the accessibility poverty line z; and y
i
the accessibility level experienced by the i-th group below the accessibility poverty
line z.
6.2 Results
The aim of this analysis is to analyze the extent of accessibility poverty across the
Tel-Aviv region. This analysis takes into account the intensity of the accessibility
poverty (the ‘depth’ of the accessibility poverty) and the number of persons that are
affected by accessibility poverty (the size of the accessibility poverty), as discussed
above.
The first step in the assessment of accessibility poverty in the Tel-Aviv region
consists of the demarcation of the accessibility poverty line (z in the equation above).
As there is virtually no empirical research that systematically investigates the
relationship between (low) levels of accessibility and the ability of persons to engage
in out-of-home activities, we determine the accessibility poverty line in a pragmatic
way and carry out the analyses for two different poverty lines. By doing the latter, we
avoid drawing conclusions based on a single, inevitably arbitrary, accessibility
poverty line.
The pragmatic approach to establishing the poverty lines is relatively straightforward.
The approach starts from the average accessibility level experienced by persons with
access to a car as observed in a study area. Since the car is the dominant mode of
transportation in the Tel-Aviv metropolitan area, it may be assumed that land uses
have organized around the mobility and accessibility afforded by the car. Persons
with an accessibility level at or above the average car-based accessibility level may
thus be assumed to have no problems to participate in out-of-home activities. This
may also hold for persons with an accessibility level somewhat below the average.
Yet, it may also be clear that persons experiencing a substantial negative deviation
from the average accessibility level, i.e. persons with a relatively low accessibility
level, are more prone to face accessibility poverty. Following this line of reasoning,
the poverty line is a particular accessibility value that lies below the average car-
based accessibility level as observed in a particular study area. Lacking empirical
research on the relationship between accessibility and activity participation, we
define the poverty line in a pragmatic and inevitably arbitrary way as a percentage of
the average car-based accessibility (e.g., 20%, 30% or 60% of the average car-
based accessibility level).
In the analyses that follow, we only relate to the accessibility levels as calculated for
a 30 minute travel time threshold, using a cumulative opportunity measure of
accessibility. Based on this measure, and weighing accessibility for the population
size of each TAZ, an ‘average’ person in the city of Tel-Aviv had access to 561,961
jobs within 30 minutes travel by car, before the public transport reform. The average
is somewhat higher after the reform, namely 562,267 jobs, due in part to a general
52
increase in employment in the Tel-Aviv metropolitan area. We have subsequently
derived two poverty lines from these average car-based accessibility levels. We
distinguish between a ‘compassionate’ poverty line of 50% of average car-based
accessibility and a ‘harsh’ threshold of 20% of the average. The 50% threshold
translates into 280,980 jobs before the reform of the public transport system and
281,133 jobs after the reform. For the 20% threshold it translates to 112,392
respectively 112,453 jobs. All zones (TAZs) that provide a lower level of accessibility
by public transport than these values are considered to fall below the poverty line and
are included in the measurement of accessibility poverty for Tel-Aviv. Zones with a
transit-based accessibility level above the poverty lines are assumed to provide
sufficient accessibility to the transit-dependent population and thus are not included
in the assessment of accessibility poverty. Based on the four poverty lines, it is now
possible to assess the changes in accessibility poverty as a result of the reform in the
public transport system in the Tel-Aviv metropolitan area.
Let us first present the key results of the analyses. The most important conclusion is
that the reform of the public transport system in the Tel-Aviv metropolitan area has
led to a reduction in the accessibility poverty experienced by persons dependent on
the public transport system. This holds for both poverty lines.
This reduction is the direct result of the fact that the average accessibility level in Tel-
Aviv (as defined in terms of the number of jobs that can be reached within a
particular time window) has increased much more for public transport than for the
private car. For instance, for the 30 minutes travel time threshold, the former has
improved by 8.6%, while the latter has remained roughly stable. The former is
obviously in large part the result of the public transport reform, while the latter is
merely an outcome of the increase in overall employment in the Tel-Aviv metropolitan
area in combination with small changes in the performance of the road system.
Weighted by population size, the general population has benefitted slightly more from
the public transport reform than the transit-dependent population: 9.6% versus 8.2%.
The fact that only the general population has seen a stronger increase than the
improvement in the average, unweighted, accessibility level (9.6% versus 8.6%),
underscores that the reform has not been directed specifically to the transit-
dependent population, an observation made before in this report and supported by
the analyses in the previous chapter.
How large are the reductions in accessibility poverty due to the public transport
reform? Let us briefly present the results separately for the two poverty lines defined
above.
For the 50% poverty line, the AP
r
score drops by 25%. This substantial reduction is
the result of the fact that the number of zones with an accessibility level below the
poverty line decreases substantially due to the reform of the public transport system.
Out of total of 214 zones in the city of Tel-Aviv, the number of zones drops from 99
before the reform to 77 after the reform, a decrease of 22% (see Figure 26). The
share of the transit-dependent population below the accessibility poverty decreases
even stronger, with 32%. These positive results are somewhat overshadowed by the
fact that the average accessibility shortfall (i.e., the negative deviation from the 50%
poverty line) as experienced by the groups below the poverty line shows an increase
53
of 10% (from 102,557 jobs before the reform to 113,585 jobs after the reform). Thus,
the number of persons falling below the accessibility poverty line decreases, but the
‘average’ person below the threshold is worse-off after the reform than before the
reform of the public transport system.
For the 20% poverty line the results are even more spectacular. In this case, the AP
r
score drops by 68%. The reduction can be largely ascribed to a sharp decrease in
the number of zones falling below the accessibility poverty line (58%), from 33 before
the reform to 14 after the reform. The number of transit-dependent persons
experiencing accessibility poverty shows a comparable reduction (57%). Like for the
50% poverty line, the average accessibility shortfall experienced by the population
groups falling below the poverty line increases, from 33,141 jobs before the reform to
38,203 jobs after the reform, an increase of 15.3%. This result underscores, again,
that a reform which is not directed explicitly to the transit-dependent population may
also have detrimental effects for this population group.
Note that the absolute values on the AP
r
-score have little meaning without a
comparison to other cities or regions. The changes in the score, however, do reflect
the changes in the extent of the accessibility poverty experienced by the transit-
dependent population in the city of Tel-Aviv. Note that substantially higher levels of
accessibility poverty may be expected for the entire Tel-Aviv metropolitan region,
because of the relatively poorer public transport service in the suburban parts of the
region.
Figure 24: Share of TAZ with an accessibility level above and below the 20% or 50%
poverty line, before and after the reform of the public transport system.
The reduction in the level of accessibility poverty obviously is reflected in the
decrease of the share of the total population in Tel-Aviv that experiences accessibility
poverty. In large part due to the reform of the public transport system, that share has
dropped from 11.6% before the reform to 7.8% after the reform, for the 50% poverty
line (see Table 5.1). In other words, based on our analyses, about 1 in 12 persons in
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Before 20%
After 20%
Before 50%
After 50%
Above
Below
54
the city of Tel-Aviv experiences accessibility poverty in the current situation if the 50
poverty threshold is used.
The situation is obviously more ‘rosy’ if the 20% poverty line is uphold as the
standard. In that case, the share of total population experiencing accessibility poverty
has dropped from 2.1% before the reform to 0.9% after the reform. So if it is
assumed that an accessibility level of 20% of the average car-based accessibility
level is sufficient for persons to adequately participate in out-of-home activities, than
less than 1 in 100 persons experience accessibility poverty in the city of Tel-Aviv.
Table 5: Statistics for the population and zones falling below the accessibility poverty
lines of 50% and 20% of the average car-based accessibility to employment in Tel Aviv.
50% poverty line
20% poverty line
Before
After
Change
Before
After
Change
TAZs
99
77
-22.2%
33
14
-57.6%
Dependent population
(# persons)
46,767 31,707 -32.2% 8,338 3,622 -56.6%
Average shortfall
(# jobs)
102,557 113,585 10.8% 33,141 38,203 15.3%
Share of total
population
11.6% 7.8% -32.2% 2.1% 0.9% -56.6%
55
a b c
Figure 25: Contribution of TAZ to accessibility poverty in percentage (20% accessibility poverty line). Old bus network (a), New bus
network (b) and the change in contribution (c). Negative values area TAZ where accessibility poverty grew after the reform.
56
a b c
Figure 26: Contribution of TAZ to accessibility poverty in percentage (50% accessibility poverty line). Old bus network (a), New bus
network (b) and the change in contribution (c). Negative values area TAZ where accessibility poverty grew after the reform.
57
7 Future extensions
The following are some of the extensions of the model that we believe will make it an
even better tool for analysis of real-world accessibility:
7.1 Accommodating daily variation and travel time unreliability
Daily variation: Each of the above indices varies by hours of a day as frequencies
and timetables change between the morning and afternoon. This will result in
different Lorenz curves and Gini Index between different periods of the day. To
incorporate daily variation of the Lorenz curves and Gini into the final estimates of
inequality we can use weighted average of the Lorenz(t) and Gini(t) for different
hours h of the day:
Lorenz =
Σ
h
Lorenz(h)*w(h),
where w(h)=TravelDemand(h)/
Σ
h
TravelDemand(h)
and
Gini =
Σ
h
Gini(h)*w(h),
Travel time Unreliability: Different traffic conditions may result in different travel
times between the same Origin (O) and Destination (D). In addition, for the buses,
irrespective of the traffic conditions, passengers are unsure about bus arrival time or
unaware of a timetable and, thus, arrive to the stop in advance, or be late for the bus.
None of these uncertainties is accounted for in the indices above which by default
apply shortest Bus and Car travel times to estimate accessibility.
High-performance computing allows introducing unreliability in travel time (or, more
general, cost). We can simulate travel time/cost uncertainty (introducing traffic jams,
non-recurring delays, etc.) using Monte Carlo procedures for estimating the
distribution of the OD travel time/cost for each OD-pair. We can assume congestion-
dependent variation in arrival time if buses do not use a special bus lane, and cancel
this variation when special lanes are operational. Every characteristic
of the OD trip
can be represented by its distribution and distribution’s parameters.
7.2 Accommodating demand for travel
We measure relative accessibility between private car and public transport by:
Access areas ratio: AA
O
(t) = PAA
O
(t)/CAA
O
(t)
Service areas ratio: SA
D
(t) = PSA
D
(t)/CSA
D
(t)
These measures relate only to the number of destinations (e.g. jobs) but not to their
importance in terms of travel flows. Instead of number of jobs (i.e. each job area has
an equal weight) we need to weight each job area by the number of trips to the job
zone and account for the travel time between zones i.e. hours travelled between
zones.
e.g. from O
1
there are trips to two destinations: 90 trips to D
1
and 10 trips to D
2
during
one hour by mode M then we compute: 90xt
11
+10xt
12
= hours travelled from O
1
by
58
mode M. If we consider fraction of OD trips among all trips, as 0.9 and 0.1 above we
can use it as: 0.9xt
11
+0.1xt
12
= average travel time from O
1
. We define:
MTT
O
= the average travel time to those job destinations that could be reached using
a particular mode (M) from origin (O)
MTT
D
= the average travel time from origins that could be reached using a particular
mode (M) from destination (D)
Given an origin O and destination D, we consider:
Average Public Transport Access travel time PATT
O
Average Private Car Access travel time CAT
O
Average Public Transport Service travel time PSTT
D
Average Private Car Service travel time CSTT
D
Note: the numbers of trips can vary for each mode and, more complicated, be
dependent on the level of service of the mode. That is, the values of 90 and 10 above
can be characteristic of buses but not of cars and improvement of the service in
regards to the destination 1 can influence one of this number and not the other.
We now measure relative accessibility between private car and public transport trip-
hours by:
Access travel time ratio: ATTR
O
= PATT
O
/CATT
O
Service travel time ratio: STTR
D
= PSTT
D
/CSTT
D
The equity indexes we have developed will change only in terms of the bus/car
average travel time ratio (i.e. we divide by travelled-hours not by number of zones)
7.2.1 Relative accessibility loss
This measure is transformed to:
=
(
1 
)
Where: P
B
/P
T
is the bus use share for origin/destination i (estimated from the mode
choice model) in the population and ATTR
O
is the average access travel time ratio for
origin O. For service areas the equivalent measure is:
=
(
1 
)
, where
STTR
D
is the average service travel time ratio for destination D.
7.2.2 Absolute accessibility loss
This measure is computed as
=
(


)
for access areas and
=
(


)
for service areas. The measure takes the average travel
time difference and multiples it by the number of PT users thus it is the average
number of passenger hours lost due to travelling by public transport. In the rare case
that the difference is negative (i.e PT is more advantageous than car) the measure
can be normalized to zero.
Generalizations:
1. Consider not all OD flows but the flows above some threshold number
2. Consider flows to areas at a distance/travel time less than a threshold value
59
In sum, the method proposed to measure horizontal equity is very flexible and can
accommodate various forms of measurements that relate to the intensity of travel
between the spatial units (buildings, zones etc.
7.3 Integrating equity analysis with cost-benefit analysis
We suggest that equity analysis should be integrated in a wider cost-benefit analysis
of transportation investments. The idea adopts the notion of a Pareto frontier. The
Pareto frontier maps the different projects as bundles on the plain of cost and equity.
The projects that have the lowest costs for obtaining similar levels of equity can be
regarded as lying on the Pareto frontier. This idea is shown in Figure 27.
Figure 27: Hypothetical bundles of projects plotted by cost and equity
The boxed points represent feasible choices. Point Z is not on the Pareto Frontier
because it is dominated by both point X and point Y. Points X and Y are not strictly
dominated by any other, and hence lie on the frontier.
This idea has been demonstrated for equity analysis (Gini index) at the TAZ level in a
paper by Feng and Zhang (2012). In Figure 28, we show the plotted results obtained
by Feng and Zhang for various projects by construction cost and Gini. The red line
shows the possible Pareto Frontier. Clearly, higher equity levels do not necessarily
imply higher costs.
Equity
Cost
Z
X
Y
60
Figure 28: Possible Pareto Frontier between cost and equity based on Feng and Zhang
(2012)
Lastly, we can add to the Pareto frontier, one more dimension the average
accessibility level provided by given projects. Thus we would aim to get higher
accessibility and higher equity at lowest construction costs.
61
8 Summary, conclusions and further research
This project set out to continue our work on measuring accessibility of transportation
systems, and comparing between private car and public transport accessibility. We
continued to develop CityGraph, as a GIS platform that calculates accessibility at
high spatial resolution, down to the level of the individual building. Many technical
problems were encountered by the sheer amount of data that has to be processed,
and some of them still need resolution and perfection.
We defined measures for relative and absolute accessibility loss due to public
transport use. This loss could be calculated on the basis of low income, car
ownership, or actual mode split per area. Feeling that none of these measures
captures well the loss of accessibility due to dependence on the public transport
system, we developed our own estimate of public transport dependency. By applying
this estimate to the formulas of accessibility loss, we arrived at an estimate of the
loss of accessibility to jobs suffered by people who are dependent on public transport
for their mobility needs. This calculation was carried out for origins in the city of Tel-
Aviv Yaffo only.
Finally, we defined and calculated accessibility poverty, which is yet another measure
of inequity in transportation. This measure was calculated for the same public
transport dependent groups but highlights those areas and population groups that
have access to jobs below a certain predetermined threshold.
8.1 Major findings
8.1.1 Calculation of accessibility at high resolution is important and possible
We have shown that it is possible to calculate accessibility at high resolution, and
that it is important as the results at a fine scale resolution are significantly different
than at the level of zones (TAZ). There is much internal variance inside TAZs,
particularly for those of large size, and those in peripheral areas where public
transport lines are not dense. Due to a lack of time and resources, we examined job
accessibility only, and have not looked at accessibility to public services or
commerce this still needs to be examined further.
Accessibility by Public Transport in Tel Aviv
There is a clear pattern of public transport accessibility in Tel-Aviv. Central areas are
highly accessible and some areas achieve accessibility levels that are close to that of
the car. At 45 minutes, most of the city achieves relative accessibility levels that are
higher than 0.5.
The New bus network improves global accessibility levels in Tel-Aviv.
The reform in the bus network carried out in 2011 generally improved the relative
accessibility by public transport in the city of Tel-Aviv - Yafo. However the effect is
not spread equally across all the city, and some areas have reduced accessibility
indeed the Lorenz curve and Gini index of the pre and post reform bus networks
show, that at least in Tel-Aviv, the inequality of accessibility between areas grew
somewhat.