Scholarly article on topic 'A New Approach to Understand Modal and Pedestrians Route in Portugal'

A New Approach to Understand Modal and Pedestrians Route in Portugal Academic research paper on "Agriculture, forestry, and fisheries"

Share paper
Academic journal
Transportation Research Procedia
OECD Field of science
{"Mode choice" / "walk path choice" / "walking and biking."}

Abstract of research paper on Agriculture, forestry, and fisheries, author of scientific article — Ana Paula Barros, Luis Miguel Martínez, José Manuel Viegas

Abstract The present paper aims at examining which factors interfere on the choices people make of modes of transport or path (when walking), take into account four variables groups: geometrics, syntactic, land use and transportation accessibility. For that purpose, an online questionnaire was applied to formulate a Discrete Choice Model in two parts: mode choice and path choice for walking trips. The findings showed that factors such as safety, comfort and urban form contribute significantly to the choice of path. For the choice of car as transport choice, the most determining factors were: weather (rain, strong sun, cloudy) and periods of the day (night), both of which are examples of factors that are not controllable; and economic elements (the presence of paid parking lot) and the displacement time (total displacement time and time of access to the car). For the choice of bicycle, the main factors were the presence of cycle lanes and bicycle parking. Based on these findings, it was possible to conclude that the car is the transport mode with greater natural preference, followed by walking and finally bicycle.

Academic research paper on topic "A New Approach to Understand Modal and Pedestrians Route in Portugal"

Available online at


Transportation Research



Transportation Research Procedía 10 (2015) 860 - 869

18th Euro Working Group on Transportation, EWGT 2015, 14-16 July 2015,

Delft, The Netherlands

A new approach to understand modal and pedestrians route in


Ana Paula Barros^*, Luis Miguel Martmez6, José Manuel Viegas6

a Instituto Superior Técnico, Universidade de Lisboa, Avenida Rovisco Pais, n°1, 1049-001 Lisbon, Portugal blnternational Transport Forum, Paris, France


The present paper aims at examining which factors interfere on the choices people make of modes of transport or path (when walking), take into account four variables groups: geometrics, syntactic, land use and transportation accessibility. For that purpose, an online questionnaire was applied to formulate a Discrete Choice Model in two parts: mode choice and path choice for walking trips. The findings showed that factors such as safety, comfort and urban form contribute significantly to the choice of path. For the choice of car as transport choice, the most determining factors were: weather (rain, strong sun, cloudy) and periods of the day (night), both of which are examples of factors that are not controllable; and economic elements (the presence of paid parking lot) and the displacement time (total displacement time and time of access to the car).For the choice of bicycle, the main factors were the presence of cycle lanes and bicycle parking. Based on these findings, it was possible to conclude that the car is the transport mode with greater natural preference, followed by walking and finally bicycle. © 2015TheAuthors. Published byElsevier B.V.Thisisanopenaccessarticleunderthe CC BY-NC-NDlicense (http://creativecommons.Org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of Delft University of Technology Keywords: Mode choice; walk path choice; walking and biking.

1. Introduction

Nowadays, interest in pedestrian environments has grown due to the dissemination of the sustainable mobility paradigm in urban areas, which has been encouraging the use of non-motorizes modes. Walking has become an

* Corresponding author. Tel.: +55-6183016116; E-mail address:

2352-1465 © 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license


Peer-review under responsibility of Delft University of Technology

doi: 10.1016/j.trpro.2015.09.039

important transport alternative for short trips in urban areas, thus requiring a deeper analysis of how citizens value the quality of their walking environments.

Several studies have been analyzing walkability in recent years, normally focusing on: (a) physical activity and health(Frank et al. 2006; Giles-Corti 2006; Leslie et al. 2005; Owen et al. 2004)(Frank et al. 2006; Giles-Corti 2006; Leslie et al. 2005; Owen et al. 2004)(Frank et al. 2006; Giles-Corti 2006; Leslie et al. 2005; Owen et al. 2004); (b) creating a pedestrian environment index(Clifton et al. 2007; Krambeck & Shah 2008)(Clifton et al. 2007; Krambeck & Shah 2008)(Clifton et al. 2007; Krambeck & Shah 2008); (c) behavioral models to understand walking as a transport alternative; and (d) urban form impacts in the pedestrian environment(Dieleman et al. 2002; Greenwald & Boarnet 2001; Handy 1996; Lee & Moudon 2006; Schlossberg et al. 2006).

Nonetheless, there is still a lack of consensus on travel behavior modeling on how to assess the effect of physical pedestrian environment design, land use and road traffic in the mode choice process of users regarding their nearby pedestrian infrastructure.

The integration, under the same analysis, of different measures of grid design along with physical attributes of pedestrian infrastructure like sidewalk width, slope or presence of tress, road traffic characterization, land use and activity data is fundamental to fully understand the choice of users in favor of walking.

The paper aims at discussing which factors are relevant for choosing transport modes (motorized or non-motorized) and path (of pedestrians) for different types of displacements and context conditions (i.e. weather). In order to carry out this analysis a questionnaire about urban environment was made available online in four languages, enabling participants from all around the world to partake, allowing assessing of the factors that influence walkability. Although the questionnaire was applied online for every country, the data collected had a greater attrition in Portugal, being the most represented country in the sample with 1,600 answers to the stated preference's section (599 respondents), 75% of which were residents of Lisbon. This collected data was then applied into a Discrete Choice Model (DCM), including a simultaneous choice of mode and walking path, introducing a significant innovation in the literature (ref). This analysis enabled us to evaluate which factors influence the choices made by Portuguese people - more specifically the residents of Lisbon - in terms of transport modes and walking path.

The organization of this study is as follows. After this brief introduction we will present a small literature review focusing on the representative works that has been published about discrete choice modelling applied to pedestrian behavior and the relation of different factors with walkability; afterwards, the methodology will be described, covering both the data collection process and the estimation of the DCM and the discussion of the results; the paper ends with some brief conclusions and further developments. Here introduce the paper, and put a nomenclature if necessary, in a box with the same font size as the rest of the paper. The paragraphs continue from here and are only separated by headings, subheadings, images and formulae. The section headings are arranged by numbers, bold and 10 pt. Here follows further instructions for authors.

2. Literature Review

Literature regarding walking behavior and mode choice, focusing in non-motorized modes has been profuse. We can classify the literature in three research streams in the area: first, researchers that have focused on understanding environmental factors influence the path choice of pedestrians (Reckert and Golob, 1976; Williams, 1977; Schwanen and Mokhtarian, 2005; Whalen, Paez and Carrasco, 2013); secondly, authors that tried to assess the impact of urban form in non-motorized modes (Cervero and Duncan, 2003; Guo and Ferreira Jr, 2008); and lastly urban studies that have focused on the effect of morphological and syntactic aspects, using Space Syntax indices (Medeiros and Holanda, 2010).

In order to analyze users' preferences for displacements (motorized and non-motorized transport), initially, it is worth understanding the functioning of market logic for an individual who needs to make use of a service (or, in the case of research, a space). It is necessary to analyze a set of alternatives available and select the one(s) whose attribute(s) provides highest level of satisfaction. According to Ben-Akiva and Lerman (1985) (1989), the choice results from a procedure performed by the individual which includes the following elements:

(a) The decision maker ('who' - the respondents of the questionnaire)

(b) The alternatives ('what' - the alternatives provided by the questionnaire);

(c) The attributes of the alternatives ('how' - the characteristics of the alternatives); and

(d) The rules for the decision ('why'- Binary logit/Nested - hierarchic: one or another, example of the questionnaire - car or bicycle and Multinomial Logit -with multiple criteria, all at the same time. Example from the questionnaire - car, bicycle and walking) - in this research, both types were used.

The choices are usually based on individual preferences (Bottom et al. 1999) that involve everything from personal aspects - age, gender, etc. - ranging from the characteristics of the alternative - comfort, convenience, speed, efficiency, reliability, safety, time, etc. - to economic aspects - cost of automobile maintenance, travel costs, etc. - all based on maximization of utility (Hoogendoorn and Bovy, 2004).

Drawing a parallel with the research at hand, here the choice would be for a specific path to be crossed on foot or by a method that replaces walking. Regarding the choice of path, some people may prefer to walk through areas with more trees, while others will take more into account the presence of slopes, especially when one or other factor is more prevalent in the life of a given individual. Regarding the modal choice, some people will prefer to walk on sunny days, and others who will drive regardless of displacement time and characteristics of the space. Therefore, the choices are very relative and individual, requiring a thorough study based on each case, taking into account all aspects.

Human choice, therefore, presents an underlying rational decision making process and this process has a functional form. Depending on the behavioural context, a specific functional form may be selected to model that behaviour. At the case in point, the preferences that define consumer behaviour can be represented by a function of utility - based on the variables that influence individual choice. A Discrete Choice Model (DCM) was proposed to the case in point, for the number of options (variables) is finite and of easy ranking. In addition, such Model has been widely used in the field of Transports for similar analysis.

Regarding the DCM, it is noteworthy that its calibration can be performed based on aggregate data - average of choices made by the population (census data) - or disaggregated - the individual information (about people or trips). According to Ortúzar and Willumsen (1990) the advantages of disaggregated data are:

(e) Efficiency in the use of individual information;

(f) Variability of individual information;

(g) The individual level can be used to any level of aggregation

(h) The estimation of parameters for each of the variables considered in the model is explicit, which provides a certain flexibility to represent the variables connected to urban planning policies;

(i) The parameters estimated have a direct interpretation in terms of importance of each variable considered in the choice;

(j) Its stability in time and space.

Thus, the disaggregated models that represent discrete choices allow the probabilities of choice to be estimated based on a behavioral approach.

In DCM, choices are independent and based on their individual utilities in relation to the set of possible choices. Thus, the data for estimation of probabilistic discrete choice models are obtained by the techniques of revealed and stated preferences (Ortúzar and Román, 2003).

The authors state that the first (revealed preferences) are based on the choices made by individuals effectively bringing information about the relative importance of different variables that influence their decisions. The later (stated preferences) also seek the same information, except that in this case, hypothetical scenarios are built and presented to the consumer / user so he can choose. The main advantage these methods present is that they can be used to analyze the demand for alternatives that are not present on the market. However, they have the drawback that individuals do not always do what they say they will (Ortúzar and Román, 2003).

The present research was based on stated preference, in which individuals were exposed to a series of hypothetical scenarios, and the decision rules were both hierarchical and of multiple criteria, to enable greater amplitude in the results.

Given that preference for a certain mode of transport or path is different for each person, that is, there is no single answer for all individuals, the Mixed Nested Logit which considers this variability of choice was chosen, and the analysis was carried out using the Biogeme® software for processing.

3. Methodology

The data required for analyzing the modal and path choices was obtained through a web-based questionnaire about pedestrian experience in urban areas. The questionnaire was conducted in 2013 in several countries around the world, available in four languages (Portuguese, Spanish, English and French) as discussed above. This last multi-language feature was introduced in order to reduce the effort of respondents in answering and to expand the universe of responses. It was composed by four parts: a) socio-demographic characterization of the respondent; b) evaluation of pedestrian environment factors; c) selection among path scenarios and d) daily mobility characterization and profile of transport modes usage.

Fig. 1. General characterization of the path and walking conditions of the stated preferences game in the questionnaire

The questionnaire collected 1,525 complete answers around the world, from which 599 were from Portugal and 524 from Brazil. The sample used was limited to Portugal, trying to minimize the heterogamous of urban contexts respondents are exposed to, yet trying to preserve a significant sample. The sample presents a significant bias

towards higher income population (31.75% of the respondents earn more than 3 times the average annual GDP per capita of their country of residence), high levels of education (57.76% of the respondents have a MSc. or PhD. degree) and an average of 35.68 years old, being 75% of the respondents younger than 42 years old. This significant bias derives from the internet dissemination process that used available academic mailing lists to reach respondents in different continents. Even with the sample characteristics available, the analysis obtained from the spatial and cultural realities may allow obtaining interesting insights about preferences in a pedestrian environment.

In the third stage of the questionnaire (Figure 1) designated selection among path scenarios, the respondent was asked to choose between two alternatives under different conditions in three experiments, resulting in the visualization of six situations. These scenarios were fed by the answers in previous sections of the questionnaire (Section 2), considering a context for the displacement (in various modes, however focusing on walking trips). A set of characteristics were taken into account, such as: straight-line distance between the origin and destination (in meters), trip purpose (work, study, leisure, etc.), time when the displacement takes place (morning, afternoon, evening, etc.) and climate conditions (temperature and cloudiness) and specific travelling conditions that may reduce walking ability, as bags, baby trolley, etc. The combination of variables under a common scenario was obtained through a factorial design.

Each pair of walking path alternatives is created randomly, including variables that were aggregated into four categories: geometry (4 variables), quality of displacement (8 variables), flows (3 variables) and configurational (5 variables). The variables included in each group are presented in Table 1, showing the possible values that can be presented to the respondent. The values presented did also include the evaluation of the respondent in terms of relevance of the variable for favorable pedestrian environments collected in Section 2 of the questionnaire. Thus, the scenarios already present a synthesis/evaluation derived from previous answers.

Table 1. Variables and presented values in the questionnaire





Quality oof displacement

a) Cold


c) Cloudy

d) Rain

a) Street width

Levels: narrow, medium and wide

b) Sidewalk width

Levels: without sidewalk (0,00m), Narrow (up to 1,00m), medium (from 1,00 to 2,00m), large (more than 2,00m) and complete sidewalk

c) Slope

Levels: streets with a soft slope (less than 2% slope), streets with medium slope (from 2 to 5% slope) and streets with a big slope (more than 5% slope)

d) Buildings height

Levels: 1 floor, between 2 and 4 floors, between 5 and 10 floors and more than 10 floors

e) Type of pedestrian crossing

Levels: without crosswalk, crosswalk without traffic light and crosswalk with traffic light

f) Distance between crosswalk

Levels: small (up to 50m), medium (from 51m to 100m) and large (more than 100m or without crosswalks)

g) Quality of sidewalk

Levels: sidewalk almost without holes, sidewalk with a medium quantity and sidewalk with a large number of holes

h) Parking organization

Levels: cars parked on the sidewalk, cars parked in dedicated areas and without parking on the street

i) Presence of trees

Levels: with a lot of trees (both side on the street - one tree each 5 meters), with some trees (both sides on the street - one tree each 10 meters) and without trees (any tree during the path) j) Lighting

Levels: without public lighting (no lighting pole along the path), bad lighting (1 lighting pole each 50 meters) and good lighting (2 lighting poles each 50 meters) k) Presence of urban furniture

Levels: with and without significant obstacles to the circulation l) Presence of stair and ramps

_Levels: presence of stairs, presence of ramps and none_



m) Flow intensity

Levels: reduced, medium and large n) Type of flow

Levels: streets with predominance of car flow, streets with predominance of pedestrian flow and streets without predominance of flows o) Flow segregation

Levels: pedestrian streets, streets shared with motorized modes with physical segregation of flows, and streets shared with motorized modes without physical segregation of flows

p) Diversity of activities on the street

Levels: predominance of retail or services, predominance of residence, presence of public institutions in a whole block and mixture of all types land uses q) Circulation in outdoor spaces

Levels: street with constant width, squares/gardens between buildings and big outdoors spaces (e.g. squares) r) Presence of streets with walls instead of doors

Levels: high walls, without doors and windows on the street, street with a mix of walls and doors and windows and many doors and windows on the street s) Length of blocks

Levels: small (less than 30m), average (from 30m to 50m) and large (more than 50m) t) Road hierarchy

Levels: local street, main street of the neighborhood, and main street of the city


First, based on this list of variables, the respondent was asked to choose one of two possible paths that he/she prefers to take when walking under the set conditions. The selection is based on relative preference of a path over the other. This preference was discretized five levels, using a Likert Scale - (a) I strongly prefer A over B, (b) I slightly prefer A over B, (c) indifferent, (d) I slightly prefer B over A, (e) I strongly prefer B over A - in order to understand the intensity of preference, trying to further explore the trade-offs considered by the respondent.

After this first stage, the respondent was also questioned regarding mode preference to perform the same trip. Initially, the respondents are asked whether they prefer to use the same path from the above walking scenario, or alternatively choose car, being characterized this mode by parking and fuel costs variables. Subsequently, they are asked if they would prefer to use bicycle, including in the characterization of the scenario the presence of bicycle lanes and parking lots. If the respondent answers 'yes' to both questions, preferring both modes over walking, there will be an additional question, in order to verify which would be the first choice between car and bicycle.

There is recent research that concludes that in the first game of stated preferences experiences, respondents are still insecure about the procedures and it is only on the following games should considered Hess et al. (2011). Nevertheless, in the case of this research, such issues were disregarded and all answers, from all games were considered, in order to expand the sample and consolidate the results.

In order to develop a simultaneous choice model of path and modal choice, a nested logit model was used. Since multiple answers from the same respondent were being used, the data was structured in panel form, in order to improve the model calibration and assessing peoples' perception variability of alternatives.

The available data that characterizes each transport alternative (modal or path choice) are defined by a utility function. Due to the hierarchical nature of the decision process, a nested formulation was considered. The used nested structure is presented in Figure 2. The used structure considers to types modes: motorized (private car) and non-motorized (the two walking paths and bicycle).

Fig 2 Modal choices present in the nested logit model

4. Model Results

After testing various configurations of model specification, the final model is presented in Table 2. The obtained model calibration present an average quality fit with a p2 of 0.163. Both the nested and panel structure of the model showed to be significant, being the estimated parameters presented in Table 2.

Table 2 shows the obtained coefficients of the walking path or mode attributes considered in utility function of the three alternatives. For easier interpretation these attributes are grouped into six categories, depending their role in route and mode choice.

Regarding the alternative specific constant or independent terms (ASC), the obtained values represent the people's predisposition to choose each mode, and the values of sigma (Sigma) (how variance affects this predisposition). The values show a strong natural preference towards the car, being the bicycle and walking terms negative. However, these last two modes present high Sigma values, showing that some persons have a deviant behavior from the average, or either preferring these modes strongly or having a very negative perception about these modes.

The car utility function included three of the weather conditions tested in the experience, leaving the Rain condition as reference and equal to 0. The obtained results show that this weather condition is the one that mostly favors car choice, being the other weather coefficients negative. The results do also indicate that the weather condition order that enhances car use is: Rain (0.00), Cloudy (-0.36), Sunny (-0.80) and Cold (-0.95).

The other attributes of the model can be evaluated in terms of their intensity (regardless of the positive or negative sign), which drive the modal and walking route choice in Portuguese cities and the statistical significance (based on the robust p value). As all the attributes in the model are binary, with the exception of travel time, the relevance of each attribute in the decision process can be directly compared by the coefficient values.

Next we discuss the results for the other model variables sorted by the magnitude of the impact on route and mode choice. These variables are:

1) Paid car parking (-0,69), that shows that parking is a fundamental policy the divert users to more sustainable transport modes;

2) Presence of bicycle lanes (0,67), that promotes significantly bicycle use;

3) Night (0,61), that increases the willingness to use car to travel;

4) Streets shared with motorized modes (no segregation) (-0,43), that evidences that pedestrian prefer to walk in pedestrian streets;

5) Slope bicycle (-0,16), that shows how this element may penalize significantly riding a bicycle;

6) Absence of lighting (-0,29), that shows the safety and security concerns of pedestrians;

7) Pedestrian streets (0,23), that potentiates walking activity;

8) Absence of crosswalk (-0,18), that demonstrates once again the pedestrian safety concerns;

9) Streets with declivity lower than 2% (0,19), that evidences the greater willingness to walk in flat areas;

10) Travel time (-0,07), that presented the same coefficient for the three alternatives, showing the omnipresence of the value of time even for very short trips;

11) Access time to car (-0,06), that proved to have a smaller coefficient that travel time regarding car use;

12) Streets with high walls (-0,17), that penalizes walking mode choice due to security issues created by lack of activity and doors in the street;

13) Presence of bicycle parking (0,45), that proves the relevance of parking facilities for bicycles in urban contexts to promote this mode;

14) Short distances between crosswalks (up to 50m) (0,14), that evidences that pedestrians tend to choose paths where crosswalks are frequent;

15) Slope conditioned (-0,29), that demonstrates persons carrying heavy bags or backpacks, with baby strollers, etc., tend to avoid the streets with slope;

16) Local streets (0,12), that indicates that people prefer to walk on calm streets with less traffic;

17) Presence of urban furniture with significant obstacles to the circulation (-0,14), that shows that badly placed urban furniture may be a significant barrier to walking, especially for impaired pedestrians;

18) Sidewalk with bad pavement (-0,15), that demonstrates how this quality component may affect route and mode choice;

19) Good lighting (0,09), that once gain prove safety and security concerns of pedestrians (especially during night), being less relevant than having a bad lighting system;

20) Sidewalk with good pavement (0,13), that shows less relevance that its symmetrical attribute (bad pavement), showing once again that people value strongly penalties than positive aspects;

21) Presence of stairs (-0,12) that shows the relevance of the physical effort when choosing walking routes;

22) Streets without parking on the sidewalk (0.12), that indicates the preference for walking in street with greater visibility and more spatial degrees of freedom;

23) Streets with illegal parking (-0.12) that reinforces the penalty introduced by barriers to walking as the bad placed urban furniture;

24) Streets without trees (-0,11) that reveals the decrease in utility of walking routes not protected from sun exposition, especially in sunny days;

25) Width conditioned (-0.29) that shows that persons carrying heavy bags or backpacks, baby strollers, etc., present a greater concern regarding the sidewalk width;

26) Prevalence of retail and services (0.09) that evidences the preferences of pedestrians towards streets with activity, due to security but also for human interaction.

Table 2. Results of the calibrated nested model

Nest brunches




ASC Sigma

Day period , weather and time of displacement Night Cold Sunny Cloudy

Time and costs attribute

Access time to car Paid car parking Time

Geometry and dedicated infrastructure

Bicycle lane Slope bike Parking bike Soft slope Slope conditioned Width conditioned

Quality of displacement

Illegal parking Without parking on the street Without trees Presence of stairs Short distance between crosswalks Bad lighting Goof lighting Presence of urban furniture Without crosswalk Sidewalk with good pavement Sidewalk with bad pavement

0.00 (fixed)

1.28 ***

0.61*** -0.95*** -0.80*** -0.36*

-0.06** -0.69*** -0.07***

Pedestrian street Streets shared with motorized modes (no segregation)


Retail and services oriented Presence of high walls on the street Local street

Nested scale (n)

0- test significance

1- test significance

-0.58 * -1.06 **

-0.07 *

0.19 ** -0.29* -0.17 -*

-0.12 -* 0.12 -* -0.12 -* -0.12 -* 0.14 *

-0.29 **

0.14 -* -0.14 -* -0.18 ** 0.13 -* -0.15 -*

0.23 ** -0.43 **

0.09 -*

-0.17 * 0.12 -*

-0.50 * -1.14 **

-0.07 *

0.67 * -0.16 * 0.45

3.46*** 1.51 -*

"significant at the 99% level; "significant at the 95% level; *significant at the 90% level, -

significant at the 80% level.

Out of the variables that can be controlled by urban planners, 'Paid car parking' presents the greatest relevance in terms of mode choice. This fact ratifies that measures of parking charging in urban centers, if combined with other policies and interventions, may be a key aspect to promote sustainable mobility.

Regarding the use of the bicycle, it is worth noting the fact that bicycle lanes can strongly induce people to use it, especially when combined with bicycle parking. However, a proper bicycle lanes location should be carefully studied to avoid routes with very steep slopes (Slope bike). The effect of these variables should be combined with the most negative ASC and larger Sigma for this mode, indicates that people have more resistance to use their bicycles (even than to walk), in particular with the lack of cycling infrastructure, turning people skeptical about the use of this mode.

It should be also highlighted that the variables found as more relevant for walking route choice are infrastructure related, while time, weather, cost, comfort, safety and security are more relevant for mode choice.

It is worth clarifying that the variables presented in Table 1 and not introduced in the model did not show a significant impact in mode and walking path choice.

The importance of trade-offs must be highlighted. In this particular case, some trade-offs of the model ware computed against travel time, as cost is not present in walking route choices is represented by how much people are willing to pay in order to receive a certain benefit.

Two examples can be used to exemplify their role in walking route choice, the first related to urban form and the last to walking impedances:

(a) Presence of high walls on the street (2.43), which means that people are willing to walk and extra 2.43 minutes in order to avoid streets with high walls in their displacements;

(b) Soft slope (2.71), which shows that people are willing to walk an extra 2.71 seconds in order to avoid streets with slope greater than 5%.

Therefore, depending on the importance given to the other variable in the trade-off, people are willing to spend more time in their overall walking displacements.

4. Conclusion

This article presents a new modal and path choice model, formulated under a discrete choice model approach. The model is based on behavioral data collected through a web based questionnaire that presented a large set of factors relevant for the definition of a pedestrian environment. These factors encompassed infrastructure attributes, traffic characterization, built environment typology, morphologic and syntactic aspects.

The model obtained presented a good fit and ability to predict the modal and path choice of respondents in the pedestrian environments.

The results show that the factors considered very important to people to affect their mode choice are related to weather, travel costs and travel time. In terms of path choices, the most important factors are those related to comfort (i.e. the presence of trees, slopes, width of sidewalks, etc.), safety (i.e. Presence of crosswalks, good lighting, hindrances on the way, etc.) and urban form (presence of high walls, street hierarchy and land use). For cyclist, the presence of cycle lanes and bicycle parking are the key elements that drive their choice. Based on these findings, the car was confirmed to be the modal most widely chosen by people, followed by walking and then the bicycle.

In terms of methodology, the Discrete Choice Model in a combined and hierarchical formulation was an innovation in the literature, since there are no models with the same types of specification that allow to identify the impact of attributes in both the modal and path choices simultaneously. Moreover, the integration of elements related to the urban configuration with the stated preferences of the respondents is also considered a relevant contribution.

Nevertheless, creating new ways of collecting information about the trip dairies of some of the participants is paramount for obtaining revealed preferences, those based on individual's travel habits. This would in turn help gather information necessary to calibrating a combined stated-revealed model.

Although the results obtained were quite satisfactory, we should acknowledge a significant bias of the sample towards more educated people and higher income strata. This fact was derived from the dissemination process through academic mailing lists.

Further developments of this study may consider an expansion in the sample, both in geographical and socioeconomic terms. In order to understand the travel behavior of a certain location and obtain data that could potentially influence public policy decision making, it is paramount to broaden the database to include all sectors of society. This would correct the sample bias by surveying a greater diversity of respondents (especially regarding education income and age). Moreover, the understanding of human behavior towards travelling could benefit from a wider geographical sample, since the number of collected responses was insufficient to calibrate the discrete choice model in other world regions.

Finally, the proposed model may result in a very useful and comprehensive instrument to assess pedestrian choices, supporting land use and mobility planning at local scale.


We would like to acknowledge all the institutions or organizations that helped on the dissemination of the web-based questionnaire that was developed for this study ( We should especially mention (WCTRS, MIT-Portugal, Instituto Superior Técnico (1ST), Rede de Polos Geradores de Viagens (Rede PGV) and Grupo de Estudos em Transportes (GET).


Ben-Akiva, M & Lerman, SR (1985), Discrete Choice Analysis, MIT Press, Cambridge, Massachusetts.

Bottom, J, et al. (1999), 'Investigation of route guidance generation issues by simulation with DynaMIT', Transportation and Traffic Theory, pp. 577-600.

Clifton, KJ, et al. (2007), 'The development and testing of an audit for the pedestrian environment', Landscape and Urban Planning, vol. 80, no. 1-2, pp. 95-110.

Dieleman, FM, et al. (2002), 'Urban form and travel behaviour: Micro-level household attributes and residential context', Urban Studies, vol. 39, no. 3, pp. 507-27.

Frank, LD, et al. (2006), 'Many pathways from land use to health - Associations between neighborhood walkability and active transportation,

body mass index, and air quality', Journal of the American Planning Association, vol. 72, no. 1, pp. 75-87. Giles-Corti, B (2006), 'People or places: What should be the target?', Journal of Science and Medicine in Sport, vol. 9, no. 5, pp. 357-66. Greenwald, M & Boarnet, M (2001), 'Built Environment as Determinant of Walking Behavior: Analyzing Nonwork Pedestrian Travel in Portland,

Oregon', Transportation Research Record: Journal of the Transportation Research Board, vol. 1780, no. -1, pp. 33-41. Handy, S (1996), 'Urban Form and Pedestrian Choices: Study of Austin Neighborhoods', Transportation Research Record: Journal of the

Transportation Research Board, vol. 1552, no. -1, pp. 135-44. Krambeck, H & Shah, J (2008), 'Evaluating and Comparing Quality of Pedestrian Infrastructure and Services Across Developing Cities: Global

Walkability Index', in 8fh Annual Meeting of the Transportation Research Board, Washington D.C. Lee, C & Moudon, AV (2006), 'The 3Ds+R: Quantifying land use and urban form correlates of walking', Transportation Research Part DTransport and Environment, vol. 11, no. 3, pp. 204-15. Leslie, E, et al. (2005), 'Residents' perceptions of walkability attributes in objectively different neighbourhoods: a pilot study', Health & Place, vol. 11, no. 3, pp. 227-36.

Owen, N, et al. (2004), 'Understanding environmental influences on walking: Review and research agenda', American Journal of Preventive Medicine, vol. 27, no. 1, pp. 67-76.

Schlossberg, M, et al. (2006), 'School trips - Effects of urban form and distance on travel mode', Journal of the American Planning Association, vol. 72, no. 3, pp. 337-46.