Scholarly article on topic 'Quantifying dimensions of Transportation Diversity: A City-Based Comparative Approach'

Quantifying dimensions of Transportation Diversity: A City-Based Comparative Approach Academic research paper on "Social and economic geography"

CC BY-NC-ND
0
0
Share paper
Academic journal
Transportation Research Procedia
OECD Field of science
Keywords
{"Transportation diversity" / "exploratory factor analysis" / sustainability / "urban mobility" / "non-motorized transport"}

Abstract of research paper on Social and economic geography, author of scientific article — Parag Pareekh, Sudeshna Mitra, Bandhan Bandhu Majumdar

Abstract Urban population growth has been increasing all over the world primarily due to more economic opportunities in cities than in rural areas but generally without a corresponding upgradation of urban transport infrastructure. However, the diverse groups of city dwellers have different transport needs but transport options in the city are often not so diverse. As a result, the transport system in cities around the world is facing challenges such as congestion, pollution, traffic injuries and fatalities, and lack of equity. It is the argument of this paper that the diverse population of cities of 21st century requires diverse forms of transportation to cater the diverse needs for urban mobility. This paper makes an attempt to quantify diversity through a set of measurable variables by performing exploratory factor analysis on data extracted from an international sample of 51 cities (differing in urban and demographic characteristics). The study seeks to identify factors underlying the concept, and is to the best of our knowledge the first such attempt to quantify transportation diversity using the city based comparative approach. EFA was performed to extract factors characterizing transportation diversity from the dataset. The three factors are extracted from the EFA and the association among the variables is examined in light of the theoretical framework of the study. Theoretical expectations such as the nature of association among indicators such as quality transit and non-motorized infrastructure, innovative policies for promoting active modes, and extent of auto use are confirmed in the study's findings.

Academic research paper on topic "Quantifying dimensions of Transportation Diversity: A City-Based Comparative Approach"

Available on line at www.sciencedirect.com

Transportation

ScienceDirect Procedía

Transportation Research Procedía 25C (2017) 3178-3191 * * * **

www.elsevier.com/locate/procedia

World Conference on Transport Research - WCTR 2016 Shanghai. 10-15 July 2016

Quantifying dimensions of Transportation Diversity: A City-Based

Comparative Approach

Parag Pareekha*, Sudeshna Mitrab, Bandhan Bandhu Majumdar0

a Project Associate, Department Of Civil Engineering, IIT Kharagpur, Kharagpur-721302, India bAssistant Professor, IIT Kharagpur, Department of Civil Engineering, Kharagpuru721302, India cSenior Research Fellow, IIT Kharagpur, Department of Civil Engineering, Kharagpur-721302, India

Abstract

Urban populati on growth hits been increasing all over the "world primarily due to more economic opportunities in cities than in rural areas bat generally without a co rresdo nding upgradation of urban transport infra structure. However, the dive rse gronps of city dwell ers have diffe rent transport needs but transpo rt options in the city are often not so diverse. As a result, the transp ort eystem in cities aruund the world is facingchallenges such as concestio^ pollution, traffic injuries and fatalities, and lack of equity. It is the: argument of tht s paper that the diverse population oa c itiete of2 1 st century requires diverse forms of transportation to catee the d^erse needs for urban roobility.This paper makes an attempt to quanrify diversity through a set of measurable vari ables by performing explorato ry factor analysis on data extracted urom an international sample of 51 cities (differi ng in urbae and demograp hic characteristics). The study seek s to identi fy factore unde rlying the concept and is to the best oa our knomied ge ehe first such attempt to quantiay transportation diversity using tha city based comparative apxroach. EbA was performhd to axtract factors characterizing transportation divegsity from the dataset. The three factors are extracted arom the EbA anc the association among the vari ables is examined in lhght oi the theoretical framework or the study. Theoreti cal expectations such ar the nature of aesociation amone indicatore such as qjuimlity traneit and non-motorized infrastructure, innovative policies for promoting active modes, and extent of auto use are confirmed in the study's findings.

© 2017 The Anther. PiMidred by EkCTkr B.V.

P^M-CTTW UMct re^nsbility of WORLD CONFERENCE ON TRANSPORT RESEARCH SOCIETY. Keywords: Transportation diversity; exploratory factor analysis; sastaisabilitd; urban mobility; sos-motorizeh transport

Corresponding author. Tel.: +91-8ddl6e7dei Email address: parag.p areek@gm8il.com

2352-1465 © 2017 The Authors. Published by Elsevier B.V.

Peer-review under responsibility of WORLD CONFERENCE ON TRANSPORT RESEARCH SOCIETY. 10.1016/j.trpro.2017.05.359

1. Introduction and Background

With rapid economic development, urbanization is a natural phenomenon that impacts just about every major sphere of activities in the urban context. According to the recently published United Nations report on world urbanization trends, urban population of the world is projected to grow to an unprecedented high of 66% from the present 54% of the total world population, with the bulk of the urban growth forecasted to take place in India and China (United Nations, 2014). The impact of increased urbanization and economic development is particularly pronounced on urban transportation infrastructure. While demand of motorized transport in general and private motor vehicles in particular tends to increase with economic development and rising urban incomes, capacity augmentation often does not take place at the same rate, resulting in demand supply mismatch. This often leads to a host of commonly experienced transport related problems such as congestion, vehicular pollution, increased travel time and a degraded urban environment. Factors like a rising urban population and constraints in terms of the spatial distribution of land uses make the situation even more critical for the disadvantaged groups in cities with limited mobility options such as the urban poor, women, children and physically challenged road users. As a result, a unified motorized transport option is just not suitable for such diverse population and demands for more diverse and creative solutions (Mitra et al, 2013).

Although socioeconomic profile of transport users varies widely within cities, transport infrastructure and travel options are often not as diverse. For example, dedicated right of way for public transport and non-motorized transport is either missing or is inadequate in urban areas, which is particularly true of cities in the developing countries. As a result, interactions between motorized and non-motorized modes are very frequent, leading to conflicts, accidents and very high traffic related injuries and fatalities. This forces the user to shift to safer and secured modes and demand for motorized personalized transport increases with increased affordability. However, with the help of well-defined policy and planning strategies, it is possible to meet the needs of various classes of road users in an equitable and reconciliatory manner. It is in this context that Transportation Diversity becomes relevant in light of the societal diversity in terms of age, gender, income, physical capacity, habits and inclination and economic characteristics, which is the primary focus of this study.

Litman (2007) defines Transportation Diversity as "the availability of transportation services in a given situation, in terms of their quantity and quality, i.e. at a specific time and at a specific location, considering the users' needs and abilities". Litman (2010) has also postulated that transport diversity is linked to modal split, land use planning, density and distribution of urban population and activities, investment in alternative modes of transport, frequency and quality of transit service, and performance of transport systems in terms of safety, equity and degree of automobile use in the city. Transport diversity is directly linked to the extent and quality of alternative modes i.e. walking, cycling, and public transport (Litman, 2010). Increasing diversity in urban transport systems is therefore directly linked to improvement of the alternative modes of transport. Automobile dependent cities around the world are characterized by low modal shares of public transport and non-motorized modes. A host of factors including functional separation based on land use and urban sprawl, sustained underinvestment in transit and alternative modes, large trip distances, increased investment in road traffic infrastructure in terms of increased road capacity and parking, and lack of transport demand management are responsible for such a state of affairs in such cities. Transportation diversity in such cities is generally low in that quality and adequate options for alternative modes of mobility to different segments of the urban population are often not available (Newman and Kenworthy, 1995; Cervero and Kockelman, 1997; Beuhler and Pucher, 2011).

A detailed literature review reveals a limited amount of research specifically focusing on the transportation diversity as a measure to evaluate urban transportation systems. The following sections review research efforts addressing transport sustainability, travel behavior and land use policy- factors, which in turn affect transport diversity as a function of land use, accessibility, mode split, and urban population density.

Cervero and Kockelman (1997) in their analysis of the impact of the three 'Ds' of the built environment, namely, density, diversity and design on travel behavior, found that compact, diverse and pedestrian orientated neighborhoods can influence urban travel behavior in statistically significant ways by encouraging higher rates of non-auto travel and pedestrian modal share. Holtzclaw (1994) and Ewing et al. (1994) found that high density urban neighborhoods with diverse land uses and greater access to transit averaged significantly less automobile usage per household, relatively less personal transportation costs and higher rates of walking and bicycling than low density

suburban neighborhoods. In a related study, Kenworthy and Laube (1999) investigated an international sample of cities to measure the degree of automobile dependence in cities across a range of parameters such as car ownership rate, transit service and use, population density, land use patterns, mode split and trip distances (for journey-to -work only). They found that cities most dependent on automobiles were generally characterized by low density suburbanized development, scant provision and use of transit, small modal shares of non-motorized modes, significantly higher rates of car use and low investment in transit and non-motorized transport infrastructure.

In this context, it seems pertinent to allude to the long running debate on the efficacy of land use and accessibility based solutions to urban transport problems among two divergent groups of researchers (Klinger, 2013). While some researchers such as Kitamura et al. (1997), Mokhtarian and Salomon (2001) and Schwanen, (2005) question the usefulness and efficacy of implementing pro-density, accessibility and transit improvement strategies, and land use management to curb auto use, spatial sprawl and influence travel behavior in automobile dependent cities, others such as Newman and Kenworthy (2006), Newman et al (1995), Cervero (1995), Cervero and Radisch (1996) and Litman (2006, 2007) argue that these strategies are among the best available tools to tackle growing auto use and its adverse impacts on the urban environment -congestion, emissions and oil dependence (Chapman, 2007). On the other hand, Ewing et al. (1994), while agreeing with the interventionists posit that accessibility oriented solutions such as pedestrianization and high density urban development per se may not be enough to mitigate automobile related problems; urban communities should instead internalize as many facilities and services as possible to cut down trip lengths and vehicle miles travelled.

In their study of trends in bicycle culture, safety, infrastructure, and bicycling rates in the USA and Canada, Pucher and associates (2011) found that cities with high or growing bicycle mode share promoted a range of coordinated policies, plans and programs designed to increase bike mode share, by expanding the network of on street bike lanes, off street bicycle tracts, bicycle parking safety, and integrating bicycle with transit, besides many more such measures. These cities demonstrate ways of increasing and improving transport options for users of alternative modes. Cities such as Freiburg, Stockholm, Copenhagen and Amsterdam in Europe, Hong Kong and Singapore in Asia, and Toronto, Portland and Vancouver in North America are excellent examples of how a coordinated and integrated land use and transport planning, programs and policies provide urban residents with adequate transport options, and an improved urban environment (Newman et al., 1995; Beuhler and Pucher, 2011; Pucher et al., 2011; Cervero, 1995).

It can be seen from the findings of these studies that transport diversity is significantly less in cities with high auto use and low density land use patterns and is high in cities that promote transit and non-motorized modes, curb auto use and contain suburbanization by integrating land use with transport planning and implementing transportation demand management measures. In the context of transport diversity, the studies cited here, concur on a number of issues, namely; the need for cities to move away from low density sprawling land use patterns and high auto reliance towards more balanced transport culture characterized by high density, greater accessibility, greater mode shares for transit and non-motorized modes, more mobility options to vulnerable road users, transportation disadvantaged sections of the urban population, significance of integrating land use planning with transport planning, and very importantly implementing an array of coordinated urban transport policies and programs.

While the concept of transportation diversity (Litman, 2007, 2010) and related topics (Cervero and Kockelman, 1996; Klinger et al., 2013) have been introduced and discussed previously in the literature, empirical testing or quantification of the concept has not been done. Ways of quantification and evaluation of transportation diversity suggested by Litman (2010) include evaluation of specific benefits of increased transportation diversity by conducting surveys designed to estimate consumers' willingness to pay for improved transport options. Another way of quantifying transportation diversity according to Litman is use of opinion surveys designed to capture users' mobility concerns and preferences. Litman further recommends comparison of public expenditures by mode as a way to quantify transportation diversity. Litman (2010) also suggests that transportation diversity can be evaluated by examining policies, programs and planning processes in terms of the degree to which they encourage increased automobile use through funding and land use plans such as zoning and tax policies. The premise of this approach is that policies and programs that encourage increased automobile use tend to reduce transport options and thereby transportation diversity, since different modes compete for users and resources (funds and road space). While substantial research has focused on developing qualitative and quantitative approaches for evaluating urban transport systems in terms of automobile dependence (Kenworthy et al., 1999), emissions, non-motorized transport (Pucher et

al., 2011) and public transport quality, studies aimed at evaluating the overall performance of cities in terms of transport options are rather limited. Though a detailed discussion on the specific issues in planning and policy making that impact the measurement of transport options or diversity is beyond the scope in this space, it must be said that increasing transport options particularly for transit and alternative modes have potential positive spillover effects on the entire system as pointed out by Litman (2007). Also, the true preferences of city dwellers in travel behavior analyses are revealed only when they are provided with viable transport options.

With the aim to bridge this gap, this paper empirically tests the concept of transportation diversity by developing an appropriate methodology described below in order to derive a clear set of factors reflecting transportation diversity that can be tested and incorporated in transportation planning. The research is largely exploratory in nature and broadly consists of devising a methodology for the empirical quantification of the concept. The primary objectives of the study are:

1. To identify a set of measurable variables that could be used to explain transportation diversity.

2. To explore through a suitable statistical technique the extent and direction of association among the selected variables such as city characteristics, socioeconomic variables, and transport related variables to uncover the broad latent factors that may be used to explain transportation diversity.

To attain these objectives, this study identifies a set of variables hypothesized to reflect the concept, extracts data on the variables from an international sample of cities and uses the multivariate technique of Exploratory Factor Analysis (EFA) to identify the key factors underlying the concept. In EFA, the investigator has no a-priori hypothesis regarding the number or nature of the variables and it is exploratory in nature. That is, it allows the researcher to explore the main dimensions to generate a theory, or model from a relatively large set of latent constructs often represented by a set of items or indicators (Williams et al., 2012). In statistical modeling, applying knowledge of the underlying data-generating process is a critical step when developing a "starter specification." However, in the absence of well-developed theories, it is often difficult for an analyst to specify a priori which observed variables affect which latent variable. EFA in this context provides a method to discover and define latent variables (Mitra et al. 2005). In this context, it is also important to mention that EFA has been extensively used in research on transportation problems (Klinger et al., 2013; Winters et al., 2011, Heinen et al., 2011); Majumdar and Mitra, 2014; and Mitra et al., 2005).

2. Selection of variables

Identification and selection of the pertinent metrics in the form of observed or measurable variables is essential for empirical testing of concepts such as transportation diversity that cannot be directly measured. Based on the literature review and the conceptual understanding of 'Transportation Diversity', a set of fourteen variables hypothesized to represent the concept of transportation diversity was selected for the purpose of analysis. The selected variables are surrogates or proxies for different aspects of a given urban transportation system and several of these indicators are widely used in urban transport performance analyses. Nine of these variables (first nine variables as mentioned in Table 1) are metric and 5 are (the last five variables as mentioned in Table 1) categorical. The categorical variables are dummy coded in order to incorporate them in the analysis. Table 1 gives a brief description of some of these variables.

Table 1 : A brief description of selected variables

Serial Attribute (Code) Description

1 Public transport modal split (PT) Public transport share is an important indicator of a city's transit performance, transport policy priority to public transport for the given city and general direction of urban mobility patterns of the city. However, it must be mentioned that one should not draw conclusive inferences regarding the city's transport performance based on this measure alone. Use of this variable in any evaluation of transportation systems should be made together with other pertinent measures of performance (Klinger et al., 2013).

2 Walking modal split (WALK) This variable indicates the proportion of trips undertaken on foot in the overall modal split in the city and has been selected to serve as a metric for walking conditions in the sampled cities.

3 Car modal split (CAR) This variable indicates the proportion of urban trips made by the car and is used as a metric for the extent of auto use in the sampled cities.

4 Bike modal split (BIKE) This indicator is used in this research as a metric for the proportion of urban trips made by the bicycle and as a surrogate for bicycle related infrastructure in the sampled cities

5 City population (POP) This demographic measure indicates the spatial distribution of the urban population, and seen in the context of land use, the urban population within municipal limits is closely related to patterns of modal share of various modes, public transport accessibility and performance among others.

6 City area (AREA) It is a measure of the spread of the city and is used to represent land use patterns.

7 Car ownership rate (OWN) This indicates the patterns of auto use and auto-dependence in the city (Kenworthy et al., 1999). Seen together with other measures this variable is an important indicator of urban mobility trends and socio-economic profile of the city. The unit used is number of cars per thousand population of a particular city.

8 GDP per capita (GDP) An economic indicator of the population, it is related to various transportation factors such as car ownership rates, vehicle miles travelled and per capita person trips among others.

9 CO2 emission per capita (EMM) This is an indicator of transport based pollution and extent of auto use in cities (Chapman, 2007).

10 Presence of Dedicated bike lane This indicator as used in this research indicates the priority accorded to bicyclists in the city in

(DED_BIKE) terms of bicycle related infrastructure and bicyclists' right of way.

11 Presence of Shared bike lane This is another indicator of the quality of the urban transport system in a given city and is used

(SHAREDBIKE) a representative surrogate for NMT related facilities in this research.

12 Presence of MRT systems (MRT) Mass Rapid Transit (MRT)is basically a rail-based urban transport system and the variable in the context of this study is represented as either present or absent in the sample of cities. It has been noted to influence travel behavior in the city (Klinger et al., 2013).

13 Parking policy (PARK) This variable has been selected as a metric for this research as parking policy is widely recognized to affect mobility patterns in cities from its relationship with extent of car usage, NMT operating conditions, and land use patterns among others (Cervero and Kockelman, 1997, Kenworthy and Laube., 1999).

14 Presence of smart card or This variable indicates the extent of sophistication of the transport system or presence of

seamless multimodal transport seamless transport in cities - an indicator of smart transport development of an urban section.

system (SMART)

3. Database development

To empirically test the concept of transportation diversity, the present study adopted a city-based comparative approach, whereby a diverse and representative sample of 51 cities from across the world was selected in order to collect secondary data for each of the selected variables. The city based comparative approach has been used in previous research, most notably by Klinger et al., (2013) when they conducted an empirical analysis of a similar latent concept- 'urban mobility cultures'. The comparative approach based on urban transportation data from an international sample of cities provides the desired level of disaggregation not found in comparative analyses based on national or country based data to assess performance of urban transportation systems. The city based comparative approach seems to be promising for the analysis of concepts developed to capture the overall performance of transportation systems in terms of equity, efficiency and transport safety (Kenworthy and Laube, 1999). However, as preferred as this approach is, researchers are thwarted from adopting it mainly on account of the acute dearth of comprehensive international databases on urban transportation, a fact noted by Kenworthy and Laube, 1999), a data limitation also experienced in the course of this study. The authors note that although at the national level, there is a plethora of data available on virtually every area of human concern (for example, from the comparative data publications of the United Nations and World Bank), there is a lack of systematic international databases, especially in the context of urban transportation.

Such limitations of reliable databases severely restrict the scope for comparing the differential performance of cities even within the same region and socioeconomic settings- an exercise that can yield plenty of insightful information and patterns in terms of policies and planning that account for such differential performance. However, a careful review has been done to extract and create a database for cities in this sample from the continents of Asia, Europe, North and South America, and Africa. Data on various parameters were extracted from standard sources such as designated statistical organizations of the sampled cities, national mobility surveys, and published literature available online (Pareekh, 2014). Cities with significant difference in their socio-economic and demographic settings were selected to reveal the influence of city specific characteristics on overall transport performance. Attempt was made to extract data from the most recent years and maintain concurrence with respect to the year of the data among the sampled cities. Table 2 presents the selected cities along with their attribute levels. The cities are presented in ascending order with respect to population. Descriptive statistics of the collected data are presented in Table 3.

Table 2: List of selected cities with identified variables

City PT WALK (%) BIKE (%) CAR (%) POP AREA (km2) EMM OWN GDP DED BI KE SHAR E BIK E PARK MR T SM ART

Freiburg 18 24 28 30 229000 153.1 1300 423 35894 Y Y Y N Y

Pittsburgh 18 11 1.5 56.5 306211 147.11 2280 683 52500 Y Y N Y N

Utrecht 18 14 21 47 323617 99 1450 487 53019 Y Y Y Y Y

Zurich 29 35 6 30 380500 87.88 1210 483.7 50600 Y Y Y Y Y

Portland 12 5 8 71 603106 354.51 1450 692 60700 Y Y Y Y N

Vancouver 14 11 1.8 73 603502 114.97 2250 530 41100 Y Y Y Y Y

Baltimore 19.2 6.8 1 58.9 621342 215.04 2700 650 55700 Y Y Y Y N

Oslo 25 34 5 36 626953 454 1400 380 55500 Y Y Y Y Y

Washingto n 38 14 3 45 632323 177 3900 580 71500 Y Y Y Y Y

Seattle 43 6 3 43 634535 223.02 1560 743 65200 Y Y Y Y Y

Boston 33 14 2 45 636479 128.27 2020 634 69300 Y Y Y Y Y

Athens 37 8 2 53 664000 412 1010 650 30500 N Y N Y Y

Frankfurt 20 31 11 43 678691 248.31 1800 549 51600 Y Y N Y Y

Copen hagen 15 25 31 29 772079 167.1 1070 237 39300 Y Y Y Y Y

Amsterdam 20 20 22 38 809244 219 1020 370 46000 Y Y Y Y Y

SFO 34 8 20 35 825863 124.52 1590 620 69000 Y Y Y Y Y

Stockholm 35 17 1 47 870000 188 1510 368 53900 Y Y Y Y Y

Ottawa 22 1.5 8.5 67 883391 2790.2 2050 540 42400 Y Y Y Y Y

Calgary 17 6 1.5 67 1096833 825.29 3000 632 61600 Y Y Y Y Y

Brussels 48 3 2 47 1148100 161 1290 468 45600 N Y N Y Y

Dallas 4.3 2 0.2 76 1241162 904.72 2580 625 55300 N Y N Y N

Munich 21 28 14 37 1353186 310 1390 550 54500 Y Y Y Y Y

Phoenix 3 2 1 88 1488750 1145.4 3900 550 45000 Y Y Y N N

Philadelphia 25 9 2.2 60 1547607 356.29 2140 604 53800 Y Y Y Y N

Barcelona 18 46 1 30 1621000 101.9 620 410 36300 N Y Y N N

Montreal 35 8 3 53 1649519 365.13 1930 446 36200 Y Y Y Y Y

Vienna 37 28 6 29 1741246 414.87 1200 390.2 47800 Y Y Y Y Y

Curitiba 45 21 5 28 1751907 430 750 410 20200 Y Y Y N Y

Hamburg 18 28 12 42 1792600 755.3 997 401 48700 Y Y N Y Y

Houston 4 2.2 0.4 75.5 2160821 1593.0 6200 700 64200 Y Y N Y Y

Paris 33 47 3 12 2234105 105.4 950 480 53900 Y Y Y Y Y

Toronto 35 6.5 2 55 2615060 630.21 2600 490 43900 N Y N Y Y

Chicago 16 19 1 63 2714856 606.1 2910 590 55000 Y Y Y Y Y

Nairobi 10 39 5 17 3138369 696 40 109.6 2080 N N N N N

Madrid 38 38 1 22 3273002 605.7 1050 511 40000 N N Y N Y

Berlin 26 30 13 31 3375200 891.7 774 319 33300 Y Y N Y Y

Los Angeles 3 9 3 83 3857799 1245.2 3800 540 60400 Y Y Y Y N

Kolkata 54 19 11 8 4448679 185 82 24 3100 N N N Y N

Singapore 44 22 1 29 5310000 710 960 105 62500 N Y Y Y Y

Hong Kong 80 1 1 11 7154600 224 378 78 48700 N N Y Y Y

Bogota 62 15 2 15 7451700 1587 510 150 15900 Y N Y Y Y

London 37 20 2 39 8100000 1570 1100 315 52000 Y Y Y Y Y

New York city 32 39 0.6 33 8336697 804.08 3100 220 63200 Y Y Y Y Y

Bangalore 35 26 7 8 8499399 741 204 94.22 4000 N N N Y Y

Seoul 63 4 1 26 10442426 605 850 215 32200 Y Y Y Y Y

Delhi 43 21 12 14 11007835 1113.6 234 197.32 9500 N N Y Y N

Mumbai 45 27 6 8 11700000 546 95 52.78 5900 N N N N N

Lagos 2 35.5 4.5 16 12090000 907 50 23 2600 N N N N N

Tokyo 51 23 14 12 13189000 2188 818 180 41400 N Y Y Y Y

Beijing 23 21 32 20 17311000 3497 275 228 20300 N N Y Y Y

Shanghai 33 27 10 20 20200000 3497 140 54 21400 Y Y Y Y Y

Y-represents presence of that attribute in the city, N- represents of the absence of that attribute in that city, they are binary/dummy coded

Table 3: Descriptive statistics of selected variables

Variable Mean Std. deviation Max. Min.

PT 29.22 16.59 80 2

WALK 18.77 12.49 47 1

BIKE 6.98 8.09 32 0.2

CAR 39.64 21.33 88 8

POP 3845946 4703163 20200000 229000

AREA 718 797.2 3498 87.88

EMM 1538 1215 6200 40

OWN 407 208.2 743 23

GDP 42827 19065 71500 2080

DED_BIKE 0.7 0.46 1 0

SHARED_BIKE 0.8 0.40 1 0

PARK 0.74 0.44 1 0

MRT 0.84 0.37 1 0

SMART 0.74 0.44 1 0

4. Modelling Methodology

For empirically testing the concept of transportation diversity that in terms of measurement is latent in nature, analysis techniques that can measure latent variables are required. Exploratory Factor Analysis, being among the best available tools for analyzing concepts that are latent and exploratory in nature, is used to identify the underlying structure among the fourteen attributes selected for the study. Since the data were not on the same scale, the raw data were standardized in order to convert them to the same scale and to ensure that each variable contributes evenly to the analysis. Subsequently, using this as input, a step wise EFA model was developed based on guidance from the published literature on the subject (Williams et al., 2012; Hair et al., 2010; and Majumdar and Mitra, 2014). The following sections briefly discuss EFA and the step wise procedure adopted in this paper.

4.1. Exploratory Factor Analysis

Factor Analysis is a statistical method used to define the underlying structure among the variables in the analysis (Hair et al., 2010). It is mainly used for the purpose of data reduction and data summarization. Factor analysis is a multivariate statistical procedure that has many uses, most important of them being: a) it reduces a large number of variables into a smaller set of variables (also referred to as factors). b) It establishes underlying dimensions between measured variables and latent constructs, thereby allowing the formation and refinement of theory and c) It provides construct validity evidence of factors (Williams et al., 2012). Factor analysis partitions the variance of each indicator (which is derived from the sample correlation or covariance matrix) into two parts: (1) common variance, or the variance accounted for by the latent variable (s), which is estimated on the basis of variance shared with other indicators in the analysis; and (2) unique variance, which is a combination of reliable variance specific to the indicator (i.e., systematic latent variables that influence only one indicator) and random error variance (i.e., measurement error or unreliability in the indicator) (Stapleton, 1997).

As noted by Washington et al., (2003) and Mitra et al., (2005), in EFA, there is no difference between dependent and independent variables. For a useful EFA model, there should be K < n factors or principal components, with the first factor given as:

— + ^-12^2 + "* + ^1)^) + "* + — l + ^lp^p (1) Subject to:

a'" + a" + - + a')+.. +a'(p—i) + a'+ — 1 (2)

The observed variables are denoted by (p x 1) column vector y, and (q x 1) column vector x, and influence the latent endogenous and exogenous variables, respectively (Mitra et al., 2005). EFA attempts to maximize the VAR [Z1], given the constraints (2) are held. A second factor, Z2, is then sought to maximize the variability across individuals, subject to the constraint (3)

a'" + a'' + - + a')+.. +a'(p—") + — 1 (3)

Additionally, there will be no correlation between Z1 and Z2. Hence a set of (K<n) uncorrelated factors, each containing a number of indicators/attributes defining different dimensions of transportation diversity could be obtained. The step-wise EFA procedure and the various approximations made in this paper are briefed in the following sections.

4.2. Data suitability

The issue of minimum sample size for performing factor analysis (provided that the theoretical framework of the study meets other major criteria stipulated for performing factor analysis) is much debated in the literature. While some researchers such as Hair et al., (2010) stipulate that the researcher should not analyze a sample with fewer than 50 observations and recommend a preferable sample size to be 100 or larger, others such as Sapnas et al., (2004) argue that a sample with even 50 cases is adequate for factor analysis. In this case, the sample consists of an international sample of 51 cities. The study thus satisfies the criterion of the minimum sample size for conducting factor analysis. In the present study, some of the variables were entered in metric form and others were entered in binary form.

4.3. Factor extraction

From among several factor extraction techniques such as principal components analysis (PCA), principal axis factoring (PAF), image factoring and maximum likelihood, the WLSMV (Mean- and Variance-adjusted Weighted Least Square) estimator can be adopted in this paper to cluster observed variables within the dataset such that a given factor accounts for maximum variance of the variables loading on to it (Majumdar and Mitra, 2014). Only the factors having latent roots or Eigen values greater than 1 are considered significant (Hair et al., 2010). In this case, Eigen values for the correlation matrix reveals three factors with an Eigen value greater than 1. Based on this criterion, a 3-factor factor model is selected as the final model from the available models.

4.4. Factor rotation

The main effect of factor rotation is to redistribute the variance of the initial set of factors to a simpler, theoretically more meaningful factor pattern (Hair et al., 2010). Factor rotation by maximizing high item loadings and minimizing low item loadings produces a more interpretable and simplified solution. In this study, VARIMAX, a widely used orthogonal factor rotation technique is adopted.

4.5. Goodness of Fit

The fit statistics test how well the competing models fit the data. A goodness-of-fit test evaluates the model in terms of the fixed parameters used to specify the model, and acceptance or rejection of the model in terms of the over-identifying conditions in the model. To evaluate the overall goodness of fit of the estimated models, TuckerLewis index (TLI), Comparative Fit Index (CFI) and Root Mean Square Error of Approximation (RMSEA) were calculated and used to guide the final model selection. These goodness-of-fit measures are based on comparisons with a baseline model. The details of these statistics can be found in Bollen (1989), Hoyle (1995) and Washington et al., (2003).

• Tucker Lewis index (TLI): TLI is a comparison of the Normed chi square values for the null model and specified model. The TLI unlike the Normed Fit Index (NFI) is not normed and its values can fall below 0 or above 1. Good models have their TLI values approaching towards 1.

• Comparative Fit Index (CFI): CFI is an incremental model fit index and is an improved version of the NFI. The CFI is normed so that its values range between 0 and 1, with values closer to 1 indicating a good model fit (Hair et al., 2010).

• RMSEA: the RMSEA measures error of approximation and is sometimes also referred to as a population-based index. The RMSEA is really a "badness of fit" index in that a value of 0 indicates the best fit and higher values indicate worse fit.

Model fit indices such as Root Mean Square Error of Approximation (estimate: 0.06), Comparative Fit Index (0.936), and Tucker Lewis Index (0.89) satisfy the cut-off values specified by Muthen and Muthen (1998) for the three factor model derived for this study.

5. Results and discussion

Exploratory factor analysis was performed on the dataset with the 14 variables and the 51 city sample described in the previous sections using the M Plus statistical program. A three factor solution was derived. The three factor model meets the theoretical framework of the study and statistical criteria specified for factor models in the literature. The subsequent labeling of each factor, the factor loadings and the interpretation of the model are discussed in the following sections.

5.1. Factor labeling

This is the most important step in EFA, as the extracted factors should be named based on both the factor loadings of the loaded variables and the theoretical framework of the study (Hair et al., 2010). While labeling, variables with higher factor loadings are given higher importance over variables with weaker loadings. The attribute 'walk modal split' (WALK) is found to have a significant loading (with less than ±0.2 difference) on two factors. It is then clustered based on its appropriateness for representing the suitable factor between them.

Based on these guidelines, the three factors obtained from the three factor model are labeled as "auto-oriented", "city-specific" and "multi-modal inclined". The factor loadings of the three factors and the loaded variables obtained from the rotated component matrix are presented in Table 4.

Table 4: Factor labeling

Factors Variable Factor Loading

CAR 0.985

EMM 0.877

OWN 0.742

Auto-oriented GDP 0.620

WALK -0.521

PT -0.371

BIKE -0.312

POP 0.851

City-specific AREA 0.830

DED_BIKE 0.959

SHARED_BIKE 0.879

Multi-modal inclined SMART 0.725

PARK 0.622

MRT 0.616

5.2. Discussion on the extracted factors

The extracted factors are briefly discussed in the following section.

• "Auto-oriented": This factor was loaded with the maximum number of variables with significant factor loadings. The factor loads with automobile related variables in a positive manner and with non-automobile variables with negative manner. For example, 'Car mode split' (CAR) is found to be associated with the highest loading of 0.985 reveals the prominent effect of increased auto use. This is closely followed by 'CO2 emission rate per capita', another indicator of excessive car use. This indicator/attribute is closely followed by 'car ownership' (OWN) with a factor loading 0.742 and GDP per capita with a factor loading of 0.620, indicating the influence of socio-economic profile on overall transport diversity. Variables such as 'walk modal split' (0.521), 'public transport modal split' (-0.371) and 'bike modal split' (-0.312) which are all non-automobile related variables showed negative loadings, indicate a significant negative association of this latent factor with the former observed variables. Although variables such as public transport modal split are correlated with car ownership and car modal split, the negative loading indicates that they are inversely related.

• "City-specific": This is another factor reflecting the influence of city specific demographic characteristics on overall transport system performance or transport diversity. City population (POP) with a significantly high (positive) loading of 0.851 is closely followed by city area (AREA) with a factor loading of 0.830. This factor indicates a significant influence of such variables on overall transport diversity. The loading of the two spatio-demographic variables 'city population' and 'city area' on to the same factor in the three factor model is in tune with the theoretical assumptions of the study that spatial demographic variables have a strong interactional effect on the city's overall mobility structure- which is also in sync with findings from other studies in the literature focusing on land use-transport interactions. Having said that, it should also be pointed out that spatio-demographics in the urban context have complex interactions with other transport variables such as supply and quality of transport services, socioeconomic variables and transport policy parameters in shaping the overall transport character of any given city. A detailed discussion of such effects and the ways in which spatial variables and transport-related indicators of cities affect each other in terms of the extent of transportation diversity in different parts of cities will require a micro-analysis of a single city with appropriately selected variables and using spatial analytic tool/platform such as Geographical Information System (GIS). Such

analysis in the context of the concept of transportation diversity can yield meaningful insights and provide a much deeper understanding of how diversity varies with key parameters impacting the city's mobility patterns. The methodology adopted for this study in terms of selection of the international sample of 51 cities and the 14 variables described above, primarily sought to provide an empirical justification of the concept using the city based comparative approach as discussed in previous sections. Therefore the inclusion of the two spatio-demographic variables is meant to capture to an extent their interactions with other variables. It should also be pointed out that an added advantage of analyzing the international sample of cities from different parts of the globe with different transport related, urban form/land use and socio economic characteristics is that this sample itself will also likely reflect, among other things, the interactive effect among spatio-demographic and transport related characteristic of cities. • "Multi-modal inclined": This factor represents the influence of innovative or smart infrastructure on overall transport performance. The highest significant loading associated with "presence of dedicated bicycle lane" reveals the importance of the provision of adequate bicycle infrastructure for increasing overall transport efficiency. Significant loading of variables such as presence of MRT systems, presence of smart card/seamless transport, improved parking policy and shared bike lane are found to be correlated with this factor indicating the importance of innovative infrastructures, which may also increase overall transport diversity. The pattern of factor loadings on this factor also points to an associative effect on urban transport system arising from the interaction among the loaded variables.

The identified variables representing different facets of the urban transport system and the factors derived from the factor model in this study may be used as a tool for rating the city's transport system at different scales (corresponding to the underlying variables for each broad factor). As it is widely recognized that the quality of the city's transport system has substantial impact on overall urban livability and urban residents' perception of the city, these variables could be further tested through an appropriate scale to estimate how transportation diversity affects other 'Quality of Life' parameters such as prosperity, equity, and inclusiveness. The factors as derived and interpreted in this study are envisaged to provide guidance to urban economists and planners in terms of the areas to focus on for improving transport options in the urban context.

6. Conclusions and recommendations

This paper seeks to quantify the concept of "transport diversity" by identifying and developing a set of variables reflecting the concept, extracting data on such variables from an international sample of cities, and analyzing the association among the variables using exploratory factor analysis. At the outset, a set of fourteen variables theorized to reflect different dimensions of transport diversity were identified from the literature. Based on the results from the exploratory factor analysis, three latent structures are identified, which load on to the observed city specific transport related variables. The three factors labeled "auto-oriented", "city- specific" and "multi-modal inclined" can be used to explain transport diversity. Among the three factors, "auto-oriented" and "multi-modal inclined" point to the city's transport orientation, whereas "city- specific" purely captures the city's geographic and demographic characteristics. Such findings suggest that reduction in the modal share of the car or a relatively less car ownership rate and a corresponding increase in public transport and NMT (both bike and walk) modal shares may subsequently lead to an increase in efficiency of the transport system and are probably indicative of transport diversity in cities, other necessary conditions also being present. Furthermore, the factor labeled "multi modal inclined" indicates a positive interactional effect of the association among variables such as bicycle facilities and the sophistication of the urban transport system characterized by mass rapid transit, parking policies and smart cards (indicating use of technology to promote use of transit). The factor "city specific" points to the significance of demographic and spatial variables as important influences on the urban transport system which concurs with findings from a number of previous studies as discussed in the previous section. Transportation diversity has a strong spatio-demographic dimension to it, to incorporate which we have used the two spatio-demographic variables. The loading of the two variables on to the same factor in the three factor model extracted from factor analysis in this study accords with the study's theoretical expectations. However, the authors recognize that interactions among spatio-demographic/land use variables and key transport related parameters are complex and the resulting effects on transportation diversity

may not have been adequately captured within the analytic framework adopted in this study which primarily sought to provide an empirical justification for the concept using the city based comparative approach and extracting broad latent factors underlying the concept. While this is a limitation of the study, it is felt that a micro analysis with a single city as unit of analysis incorporating spatial/urban form variables using spatial analysis tool/platform such as geographical information system (GIS) will provide a much more deeper and comprehensive understanding of transport diversity in terms of its spatio-demographic dimension. Future research should therefore focus along this direction. However, it should be pointed out that increases in modal share of transit and the active modes per se do not necessarily mean an improvement in urban mobility or increase in transport diversity in cities.

The proposed methodology and broad outcome of the study could have three different applications. Firstly, from a city planning perspective, based on the factor loadings of the selected variables, variables that influence diversity in significant manner can be prioritized. For example, from the latent factor labeled "multi-modal inclined", the variable with the highest loading, namely, "DED_BIKE" (though it is correlated with the other variables loading on to that factor, a direct interpretation may not be appropriate) can be tested to check how the inclusion or omission of this variable impacts the diversity of cities. Secondly, another application of the study's outcome could be devising an indicator of transportation diversity. This may be accomplished by constructing a transportation diversity index based on the three broad factors and the variables associated with them. Such an index could be instrumental in evaluation of diversity of the transport system for a given set of cities. The index can also be used to test the influence of policy measures such as introduction of smart card to integrate different transit modes and reduce transaction time, and parking innovations such as park-and-ride options on overall transportation diversity of the city. Development of such an index for different cities and a confirmatory analysis on the interrelationships among the various factors and the associated variables using latent variable estimation procedures such as Structural Equation Modeling (SEM) are envisaged as future extension to the present study.

Before closing the discussion on the broad findings of this study, it is important to mention that while the methodology of identifying representative attributes/indicators and selecting the city as the unit of comparison for analysis in this paper is aimed to uncover the latent structure underlying the concept, the factors obtained from the analysis are meant to be only indicative of the extent and direction of the diversity of the transport system in the city and are by no means exhaustive; as multiple quantitative as well as qualitative approaches can be adopted to evaluate a measure as complex as transportation diversity, as previously alluded to in the discussion of the literature on the subject. While this is true, it is certain that any approach for evaluation of the concept must be at the network level as transportation is provided by an integrated system; hence the use of the city based approach for the purpose of analysis in this study. This study, to the best of our knowledge, is the first such attempt to empirically test the concept of transportation diversity and seeks to contribute to the pool of research on sustainable transport in cities.

References

Bollen, K.A. 1989. Structural Equations with Latent Variables, New York: John Wiley & Sons, Inc.

Buehler, R., Pucher, J. 2011. Sustainable Transport in Freiburg: Lessons from Germany's Environmental Capital. International Journal of

Sustainable Transportation, 5, 43-70. Cervero, R. 1995. Sustainable New Towns: Stockholm's Rail-Served Satellites. Cities, 12 (1), 41-51.

Cervero, R., Kockelman, K. 1997. Travel Demand and the 3ds: Density, Diversity, and Design. Transportation Research, Part D, 2 (3), 199-219. Cervero, R., Radisch, C. 1996. Travel choices in pedestrian versus automobile oriented neighborhoods. Travel choices in pedestrian versus

automobile oriented neighborhoods, 3 (3), 127-141. Chapman, L. 2007. Transport and climate change: a review. Journal of Transport Geography 15, 354-367.

Ewing, R., Haliyur, P., Page, G.W. 1994. Getting around a traditional city, a suburban planned unit development pud), and everything in between.

Transportation Research Record 1466. Journal of the Transportation Research Board. Fitzroy, F., Smith, I. 1998. Public transport in Freiburg : why did patronage double in a decade ? Transport Policy 5,163-173. Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E. 2010. Multivariate Data Analysis. Seventh ed. 2013. Pearson education Inc. Heinen, E., K. Maat, and B. Van Wee. Day-to-Day Choice to Commute or Not by Bicycle. In Transportation Research Record: Journal of the

Transportation Research Board, No. 2230, Transportation Research Board of the National Academies, Washington, D.C., 2011, pp. 9-18 Holtzclaw, J. 1994. Using residential patterns and transit to decrease auto dependence and costs. "http://docs.nrdc.org/smartGrowth/files/sma_09121401a.pdf' Accessed on August 15th, 2015

Hoyle, R. Ed. 1995. Structural Equation Modeling: Concepts, Issues, and Applications. Thousand Oaks, CA: Sage Publications, Inc.

Kenworthy, J. R., Laube, F. B. 1999. Patterns of automobile dependence in cities: an international overview of key physical and economic dimensions with some implications for urban policy. Transportation Research Part A: Policy and Practice, 33(7), 691-723.

Kitamura, R., Mokhtarian, P.L., Laidet, L. 1997. A micro-analysis of land use and travel in five neighborhoods in the San Francisco Bay Area. Transportation Volume 24 (2), 125-158.

Klinger, T., Kenworthy, J.R., Lanzendorf, M. 2013. Dimensions of urban mobility cultures - a comparison of German cities. Journal of Transport Geography, 31, 18-29.

Litman, T. 2006. Community Cohesion as a Transport Planning Objective. Victoria Transport Policy Institute.

Litman, T. 2007. You Can Get There From Here: Evaluating Transportation Choice," Transportation Research Record 1756. Journal of the Transportation Research Board. pp. 32-41.

Litman, T. 2010. Transportation Cost and Benefit Analysis II - Transportation Diversity. Victoria Transport Policy Institute, pp. 5.9-1-10.

Majumdar, B.B., Mitra, S. 2014. Identification of factors influencing bicycling in small sized cities: A case study of Kharagpur, India. Case Studies on Transport Policy.

Mitra, S., Pareekh, P., Majumdar, B.B. 2013. Urbanization and Transportation Diversity. Shelter, Vol. 14 No.2 October 2013. HUDCO -HSMI Publication.

Mitra, S., Washington, S., Dumbaugh, E., Meyer, M.D. 2005. Governors Highway Safety Associations and Transportation Planning: Exploratory Factor Analysis and Structural Equation Modeling. Journal of Transportation and Statistics, 8, 01.

Mokhtarian, P.L., Salomon, I. 2001. How derived is the demand for travel? Some conceptual and measurement considerations. Transportation research part A: Policy and practice.

Muthen, L.K., Muthen, B.O. 1998-2010. Mplus User's Guide. Sixth Edition. Los Angeles, CA.

Newman, P., Kenworthy, J. 2006. Urban design to reduce automobile dependence. Opolis, 2(1)

Newman, P., Kenworthy, J., Vintila, P. 1995. Physical planning in an age of urban cynicism. Cities, Vol. 12, No. 1, pp. 53—65.

Pareekh, P. 2014. A Critical Analysis of Transportation Diversity in Different Cities in the World (unpublished Master's thesis). Indian Institute of Technology Kharagpur, Kharagpur, India.

Pucher, J., Buehler, R., and Seinen, M. 2011. Bicycling renaissance in North America? An update and re-appraisal of cycling trends and policies. Transportation research part A: policy and practice 45 (6), 451-475.

Sapnas, K.G., Zeller, R.A. 2004. Minimizing sample size when using exploratory factor analysis for measurement. J Nurs Meas. 2002 Fall, 10(2), 135-54.

Schwanen, T., Mokhtarian, P L. 2005. What affects commute mode choice: neighborhood physical structure or preferences toward neighborhoods? Journal of Transport Geography, 13 (1), 83-99.

Stapleton, C.D. 1997. Basic Concepts and Procedures of Confirmatory Factor Analysis. Texas A&M University.

Tang, S., Lo, H.K. 2008. Hong Kong: The Impact of Public Transit on the Viability and Sustainability of Mass Railway Transit -The Hong Kong Experience. Transportation Research Part A: Policy and Practice, 42 (4), 563-576.

Washington, S., Karlaftis, M.., Mannering, F. 2003. Statistical and Econometric Methods for Transportation Data Analysis. Boca Raton, FL: Chapman & Hall/CRC.

Williams, B., Brown, T., Onsman, A. 2012. Exploratory factor analysis: A five-step guide for novices. Australasian Journal of Paramedicine, 8, Issue 1-10.

Winters, M., G. Davidson, D. Kao, and K. Teschke. Motivators and Deterrents of Bicycling: Comparing Influences on Decisions to Ride. Transportation, Vol. 38, 2011, pp. 153-168

United Nations, Department of Economic and Social Affairs, Population Division. 2014. World Urbanization Prospects: The 2014 Revision, Highlights (ST/ESA/SER.A/352).