The role of information systems in non-routine transit use of university students: Evidence from Brazil and Denmark

(1)

The role of information systems in non-routine transit use of

university students: Evidence from Brazil and Denmark

Sigal Kaplan

a,b,⇑

, Mayara Moraes Monteiro

c

, Marie Karen Anderson

b

, Otto Anker Nielsen

b

,

Enilson Medeiros Dos Santos

d

a

Department of Geography, Hebrew University of Jerusalem, Mount Scopus, 91905 Jerusalem, Israel

b_{Department of Management, Technical University of Denmark, Bygningstorvet 116B, 2800 Kgs. Lyngby, Denmark}

c_{Centro de Tecnologia de Geociências, Universidade Federal de Pernambuco, Rua Acadêmico Hélio Ramos, S/N Campus Universitário, Cidade Universitária, CEP}

50740-530 Recife, PE, Brazil

d

Centro de Tecnologia, Departamento de Engenharia Civil, Setor IV, 1° andar, Sala 078, Universidade Federal do Rio Grande do Norte, Campus Universitário Lagoa Nova, CEP 59078-970 Natal, RN, Brazil

a r t i c l e i n f o

Article history:

Received 22 January 2016

Received in revised form 30 August 2016 Accepted 27 October 2016

Keywords: Information systems Public transport

Technology acceptance model Structural equation models Transit use

a b s t r a c t

In this study we seek to understand the relation between travel information, transit use intentions and night travel. We hypothesize that transit use is related to the perceived use-fulness and the ease-of-use of the system, which are related to information quality and real-time information availability. The hypothesized relations are anchored theoretically in the Technology Acceptance Model and validated empirically in two case-studies: (i) Copenhagen (Denmark), characterized by a highly integrated transit system with an advanced web-based information system; (ii) Recife and Natal (Brazil), characterized by a lower perceived level-of-service and non-integrated information sources. Data from a tailor-made survey of 1123 university students were collected. Structural equation models were employed for explaining the use of transit as a function of the observed respondent characteristics and the latent constructs. The results show that: (i) information search quality and source explain transit use; (ii) information quality underlies level-of-service and familiarity; (iii) the use of real-time information links to information quality and famil-iarity; (iv) general transit use and non-routine use during night and to unfamiliar places are correlated; and (v) the behavioral framework is confirmed with the two case-studies. Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction

The lifestyle of young people is shifting from the traditionally well-rooted and routine lifestyle to a more dynamic one, continuously changing residence place and activity patterns, due to the need for higher education, the competition in the labor market, the need for business travel, higher leisure consumption and globalization. When coupled with growing city complexity and size, this lifestyle, which necessitates people to be constantly on the move in unfamiliar environments, gen-erates high demand for travel information. The more time people spend moving around in complex unfamiliar environ-ments, the more important is getting relevant and reliable information in a clear and efficient format for maintaining the

http://dx.doi.org/10.1016/j.tra.2016.10.029

⇑ Corresponding author at: Department of Geography, Hebrew University of Jerusalem, Mount Scopus, 91905 Jerusalem, Israel and Department of Management, Technical University of Denmark, Bygningstorvet 116B, 2800 Kgs. Lyngby, Denmark.

E-mail addresses:sigal.kaplan@mail.huji.ac.il,siga@dtu.dk(S. Kaplan).

Contents lists available atScienceDirect

Transportation Research Part A

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / t r a

(2)

activity schedule, as well as saving time and money. Nevertheless, the cost effectiveness of transit information systems and their impact on users’ travel patterns and modal shift is unclear. As a result, there are many transit operators that do not provide transit information due to the belief that transit users rely on their past experience and familiarity with the system, word-of-mouth, and operator-specific information (Ibraeva and Figueira de Sousa, 2014). Other transit providers invest in collecting, processing and disseminating real-time information in trip advisor applications (Alves et al., 2012; Farag and Lyons, 2012; Bruglieri et al., 2015). Understanding the relation between information systems and transit use could support transit operators in their decision to invest in advanced information systems.

Recent studies show the importance of seeking transit information for routine travel and long distance trips. In a survey conducted among 200 participants in China, 77% of the respondents sought transit information one to three times in the week prior to the survey, and 77% said that they would like to see transit information before the trip (Hou and Chen, 2013). In a survey conducted in Bristol, respectively 42% and 57% of the respondents sought information always or very often for leisure and business trips over 50 miles within the UK, while 57% obtained transit information for unfamiliar trips (Farag and Lyons, 2012). An experiment in the corridor between Liverpool and Chester showed that better information about price reduction was associated with a significant ridership increase (Ibraeva and Figueira de Sousa, 2014). The mobile application for transit information in Melbourne (Australia) is downloaded more than 4000 times a week (Ibraeva and Figueira de Sousa, 2014). In an experiment conducted in Japan regarding a new information system on-board the train about stop, transfers, information search regarding ‘‘traffic conditions” and ‘‘cabin capacity ratio” were among the leading information searches for both men and women (Matsumoto and Hidaka, 2015).

Ibraeva and Figueira de Sousa (2014)posed the question whether the growing accessibility of information can induce a general modal shift towards transit use. While research is scarce regarding the use of information for planning transit trips and the linkage between seeking transit information and transit use (Farag and Lyons, 2012), studies show a positive relation between transit information and transit ridership.Hou and Chen (2013)reported that a survey revealed that 49.5% and 61.3% of travelers would have been willing to consider to adjust their departure time and mode according to pre-trip information.

Brakewood et al. (2014)conducted a before-after survey for evaluating the impacts of real-time transit information on bus riders in Tampa (Florida), and found a significant change in the waiting time and the feelings associated with the waiting time, but not in the trip frequency or number of transfers.Dyrberg et al. (2015)found a significant relation between the role of information, ease of use of transit terminals and multi-modal route choice in the Copenhagen Region. Two recent studies provided a rigorous statistical analysis to investigate the impact of introducing real-time information on bus ridership, by estimating fixed effect models on longitudinal data, accounting for multiple factors such as level-of-service (LOS), gasoline prices, spatial and temporal effects. In Chicago, the monthly average weekday ridership of bus routes with real-time infor-mation was estimated to be 126 rides a day more than other routes, when controlling for all other factors (Tang and Thakuriah, 2012). In New-York, a similar effect was found, namely an average weekday increase of about 118 trips per route, attributable to providing real-time information (Brakewood et al., 2015).

The aforementioned studies used observed measurable indicators to support a positive relation between information pro-vision as the external stimuli, and transit ridership as the outcome. In the case of transit information propro-vision, observed measures are available in before-after studies when a new information system is introduced. However, research consistently shows that, even when there are no detectable changes in the system, subjective measures such as perceptions, social norms and perceived difficulties are significantly related to transit ridership (e.g.,de Oña et al., 2013; Kaplan et al., 2014). Moreover, adding attitudinal constructs significantly improves the model fit (Spears et al., 2013).Tang and Thakuriah (2011)andFarag and Lyons (2012)focused on the role of travel attitudes.Tang and Thakuriah (2011)linked the willingness to increase transit use to previous experience with transit information, willingness to pay for transit information, and travel attitudes.Farag and Lyons (2012)focused on information use for long-distance trips, and found that information search is negatively related to routine travel and positively associated with less frequent use.

The contribution of this study to the current knowledge on transit information is four-fold. Firstly, this study is the first to apply the Extended Technology Acceptance Model (ETAM,Venkatesh and Davis, 2000) to investigate the role of information on transit use. The model provides a comprehensive and rigorous behavioral framework that considers the role of perceived information quality, perceived LOS, perceived usefulness of the transit system, perceived difficulties to use the system (i.e., perceived security, perceived cost), and perceived familiarity with the transit system. The hypothesized conceptual framework extends the classical ETAM because the information dimension is not explicitly included in the original ETAM framework.

Secondly, this study is the first to address the role of information use in non-routine travel at night and travel to unfa-miliar places in addition to routine transit use at the intra-metropolitan level. On the one hand, when travel is habitual there is a strong link between travel goal and travel mode with habit reducing the responsiveness to information (Aarts et al., 1997; Farag and Lyons, 2012; Légal et al., 2016). On the other hand, because travelers mainly consider travel time and cost (Nielsen, 2000; Anderson et al., 2014), information is relevant even when transit use is frequent and routine (Sßimsßekog˘lu et al., 2015). The current study is the first to show how information relates to both routine and non-routine transit use.

Thirdly, this study is the first to address students as an important market share of transit users. The attraction and reten-tion of highly educated young people as transit users is important following the understanding that highly skilled young people in their twenties are not transit captives, but rather choose to lead a multi-modal lifestyle (Kuhnimhof et al., 2011, 2012). This study focuses on university students as the new generation of highly-educated young adults, which is rapidly increasing in Brazil and Colombia as well as in other countries and is characterized by high prospects of income

(3)

and motorization rate (Duarte et al., 2016). While the rate of university students in Brazil is still low (24%), the number of young adults attending university has doubled reaching 6.2 million in 2012 (Duarte et al., 2016). Due to the linkage between transport independence and subjective well-being of young adults (e.g.,Delbosc and Vella-Brodrick, 2015), our study joins previous ones (Kaplan et al., 2014; Salva Ramirez et al., 2015; Duarte et al., 2016) in posing the question regarding the moti-vators of highly-educated young people to use sustainable transport modes. Our analysis adds the transit information dimension to previous studies which analyzed other factors such as perceived equity in transit provision (Kaplan et al., 2014), social climate (Salva Ramirez et al., 2015), and system-wide service expectations (Duarte et al., 2016).

Last, this study is the first to show the similarities and differences in information use across regions with different transit networks. The study validates the behavioral framework and the transferability of the structural relations in Brazil and Den-mark. Copenhagen (Denmark) is characterized by a highly integrated and well-structured multi-modal transit system with an integrated web-based information system. In contrast, Recife and Natal (Brazil) are characterized by a transit system that is fragmented, non-coordinated, operated by multiple independent operators, and also by scarce, segmented and informal information sources. The two case-studies enable to shed light regarding the role of information under different conditions of transit provision.

2. Methods

2.1. The theory of planned behavior

This study views transit use from the technology acceptance perspective. The basic building blocks of the framework are inspired by the Extended Technology Acceptance Model (ETAM,Venkatesh and Davis, 2000), seeFig. 1, as a variation of the Technology Acceptance Model (TAM,Davis, 1993) in line with the Theory of Planned Behavior (TPB,Ajzen, 1991). These models stem from the Theory of Reasoned Action (TRA,Ajzen and Fishbein, 1970), stating that behavior is preceded by inten-tions, which are related to attitudes and normative beliefs. The TPB (Ajzen, 1991) postulates that behavioral attitudes, sub-jective norms associated with the behavior, and perceived volitional control, lead to the formation of behavioral intentions. Namely, individual decision-makers express stronger intentions to perform a behavior if they (i) express favorable attitudes towards the behavior, (ii) perceive greater perceived ease of conducting the behavior, and (iii) perceive favorable subjective norms towards the behavior in their social circle. These intentions will eventually transform into observed behavior, pro-vided the availability of resources and volitional control as the ability to choose one’s own behavior (Sigurdardottir et al., 2013).

Davis’ (1993)adaptation of the TRA explains the acceptance of a new technology with two attitudinal measures that could also serve as normative beliefs, namely perceived usefulness and perceived easy-of-use that are related to Bandura’s Self-Efficacy Theory (Bandura, 1977). The perceived usefulness could be viewed as related to performance expectations, while the ease-of-use is related to the perceived locus-of-control as a function of the perceived individual skills. Technology acceptance is described as use behavior preceded by use intentions. The perceived usefulness is associated with the individ-ual perception of utility related to a particular system or a product. The perceived ease-of-use expresses the perceived effort spent in performing the behavior, and thus coincides with the perceived behavioral control (PBC) in the TPB, when the latter

(4)

is interpreted as effort rather than volitional control. In the original TAM, both the ease-of-use and the system usefulness are related to the system features.

Venkatesh and Davis (2000)improved the alignment of the TAM with the TPB and the Theory of Self-Efficacy by formu-lating the ETAM. The alignment with the TPB is achieved by interpreting the notion of volitional control as both ease-of-use and individual choice (i.e., mandatory/voluntary use of the technology) and adding inter-personal motivation to the individ-ual motivation (i.e., subjective norms, social image). The alignment with the Theory of Self-Efficacy is attained by correlating the notion of expected usefulness and expected outcome performance (i.e., results demonstrability, output quality and expe-rience). Notably, the ETAM lacks the relation between the system features as determinants of the perceived usefulness and the perceived ease-of-use.

2.2. The proposed behavioral framework

We extend the ETAM in order to explore the role of information quality on user acceptance of urban transit services. Nota-bly, the TPB is a general theory and the TAM and the ETAM were developed for the context of the use of information and communication technologies at the work place. Therefore, the model framework needs to be adapted to the current context of transit as a complex and dynamic socio-technological product that requires continuous learning by its users. Moreover, information is provided as a system feature, namely an integral part of the transit service and is essential for an efficient use of the system. The model framework is illustrated inFig. 2.

The behavioral framework refers to both general transit use and non-routine transit use, which are hypothesized to be correlated in accordance withAarts et al. (1997)who postulated that temporary mode choice decisions depend on travel habits and information processing. Non-routine travel is hypothesized to be related to greater information search needs in accordance withFarag and Lyons (2012). General transit use is measured by the perceived frequency of transit regardless of trip purpose and time-of-day. Routine travel is a steady behavior with spatial and temporal regularities which affects the decision making process on a recurrent basis (Schlich and Axhausen, 2003; Aarts et al., 1997). Routine/non-routine travel is a latent construct that is not easily measured. While routine travel is defined as recurrent travel to routine activities as a set of customary activities with similarity patterns across days, time-of-day, and location, non-routine travel is defined as infre-quent travel to non-customary activities that occur with high variability over days, time-of-day or location. Nevertheless, both routine and non-routine trips are vague concepts that vary across travelers and cannot be perfectly measured with a single indicator. We used travel situations as measurement indicators in accordance withSchlich and Axhausen (2003)

andAarts et al. (1997). Specifically, the survey respondents were requested to state their perceived transit use frequency for four routine activities (i.e., going to the university, work, city center, regular leisure activities) and two non-routine activ-ities (i.e., going out at night, and going to new/unfamiliar places) as indicators.

The proposed behavioral framework is a modified version of the ETAM, adapted to the context of transit use. In the ETAM, result demonstrability and output quality explain the perceived system usefulness. In the proposed framework, in line with

(5)

Bandura’s Theory of Self-Efficacy, the perceived usefulness of the transit system is expressed in terms of result demonstra-bility, for example time and monetary savings. In line with the ETAM and the TPB, the proposed framework includes pro-transit subjective norms. Social image is driven entirely by subjective norms and hence redundant in the current context. In the ETAM, user experience or familiarity are not related to the ease-of-use, but only to the usefulness of the product. In addition, the ETAM does not explicitly consider information provision and quality. In transit use, user experience, famil-iarity, and information quality, cannot be viewed as external factors that lead to the initial product choice, but rather as essential factors for the efficient and easy use of the transit system. Thus, in the transit context they are viewed as various ease-of-use dimensions. In line withSalva Ramirez et al. (2015), user experience is taken from both the transit provider and the social climate perspective, accounting for both LOS and transit security. Familiarity is viewed as a facilitator for transit use because it reduces the burden associated with navigating and information search (Farag and Lyons, 2012; Kamga et al., 2013). Information quality is related to both the ETAM and the Theory of Self-Efficacy because it is an output quality that is a basic feature in transit systems that allows there efficient use. Information quality can be viewed as intermediate output quality that allows informed and improved transit choices, thus encouraging higher transit use. The proposed framework hypothesizes that information has an additional role on increasing the perceived familiarity and LOS, which in turn affects the transit perceived ease-of-use and usefulness. The construct of job relevance in the original ETAM model is a contextual variable and hence irrelevant in the current transit use context.

The proposed model framework enables to explore the following hypotheses:

H1. Real-time information, as an observed variable, explains the latent constructs of higher perceived LOS, information quality, and perceived familiarity.

H2. General transit use and non-routine travel are explained by the perceived transit usefulness, subjective norms and the perceived ease-of-use.

H3. The perceived transit familiarity and perceived information quality are important components of the perceived ease-of-use of the transit system.

H4. Non-routine transit use positively relates to general transit use and to observed real-time information search. 2.3. Mathematical model

The hypothesized behavioral model structure is statistically tested and validated with a structural equation model (SEM) as a well-established and widely-applied method in behavioral, social and natural sciences (Pugesek et al., 2003). SEM is a general statistical approach for validating a multivariate correlation structure across latent and observed variables in a series of equations. SEM is a generalization of regression that has an advantage in accomodating (i) a correlation structure across multiple dependent variables and (ii) conceptual (latent) constructs that cannot be perfectly measured by a single observed indicator, but require multiple indicators. SEM is commonly viewed as a combination of confirmatory factor analysis (CFA) representing the relationship between the latent constructs and the set of observed indicators, and path analysis represent-ing the dependency structure across the latent constructs (Pugesek et al., 2003). A growing number of recent studies applied SEM to explore mode choice and mode switching intentions of adults (e.g.,Chen and Chao, 2011; Kaplan et al., 2014). The methodology is thoroughly presented byPugesek et al. (2003), and its application in travel behavior research has been reviewed byGolob (2003).

In this study the SEM contains three sets of equations: (i) a set of measurement equations (Eq.(1)); (ii) a set of structural equations (Eq.(2)) linking the latent constructs to socio-economic characteristics, routine use of car and bicycle, place of res-idence and real-time transit information; a set of structural equations (Eq.(3)) relating general and non-routine transit use indicators to the latent constructs and information search.

Irn¼ Zln

a

rþ

t

rnand

t

n Nð0;

R

tÞ for r ¼ 1; . . . ; R ð1Þ

Zln¼ Xlnblþ

x

lnand

x

n Nð0;

R

xÞ for l ¼ 1; . . . ; L ð2Þ

Ykn¼ Zkln

c

Zþ Tkln

c

Tnmnandnn Nð0;

r

2_nÞ for k ¼ 1; . . . ; K ð3Þ

where Irnis the value of an indicator r of the latent construct Z⁄lnas perceived by respondent n, Z⁄lnis the value of latent con-struct l for respondent n, Xlnis a vector of the respondents’ individual characteristics (i.e., socio-economic characteristics, travel habits, place of residence, and travel information sources), Y⁄knis a vector of k independent variables, namely the respondents’ perceived general and non-routine transit use for at night and to unfamiliar places, that are related to the latent constructs Z⁄klnfor each variable k, Tklnis a vector of transit information search possibilities (e.g., during night, to unfamiliar

(6)

activities, to arrive on time, etc.). The error terms follow a normal distribution with respective covariance matrices, and the estimated parameter vectors are

a

r,bl,

c

z, and

c

T.

The commercial software M-Plus served for model estimation. The parameters were estimated simultaneously by using Maximum Likelihood with Huber-White covariance adjustment (Yuan and Bentler, 2000). Standard errors were calculated by adopting the White’s sandwich-based method that produces robust statistics in the presence of non-normality of the indi-cators and the categorical variables (White, 1980). Alongside the traditional descriptive measure of chi-square test of abso-lute model fit (CFI), an additional descriptive goodness-of-fit measure is the Root Mean Square of Approximation (RMSEA) (Browne and Cudeck, 1993). The CFI compares the fit of a target model to the fit of an independent model to a model in which the variables are assumed to be uncorrelated. Kim and Bentler (2006), suggest that a CFI around 0.9 indicates good fit between the model and the data. The RMSEA expresses how well the model would fit the populations’ covariance matrix.

Kim and Bentler (2006)suggest a cutoff value around 0.05 for RMSEA with lower values representing a good model.

3. Data collection 3.1. Case-studies

3.1.1. Recife and Natal, Brazil

Recife, the capital city of Pernambuco’s State, is the core of the fourth most populous metropolitan area in Brazil that com-prises 14 municipalities and 4 million inhabitants, with more than 1.5 million residing in the core city. The metropolitan transit system in Recife is multi-modal including buses, metro, Bus Rapid Transit (BRT), and Light Rail Transit (LRT). The tran-sit system is divided into two systems: the Integrated Structural System (SEI) and the Complementary System.

The SEI comprises 185 bus lines and 25 integrated terminals. The bus lines include 123 feeder lines, 3 perimeter lines, 24 radial lines, 18 inter-terminal lines, 6 transversal lines and 11 circular lines. The BRT system forms part of the SEI and con-sists of 37 stations and of eight lines carrying a daily average of 110,000 passengers. The three metro lines, with a total length of 70.4 km connecting 36 stations, cover five cities of Recife’s Metropolitan Area and carry an average of 245,000 passengers on weekdays. The fare system in the Integrated Structural System is based on five rings. The fares vary according to the ring where the travel is performed, with the most expensive trip being around USD 0.86 (on 20 October 2015, the exchange rate is 1.00 USD = 3.88 R$). There are concessionary fares for students and free fare for elderly (over 65 years old). For a single trip by metro/LRT, the fare is USD 0.41, while for trips that include a combination of the buses and the metro, an integrated fare according to a zone system applies. The complementary system covers the rest of the metropolitan area and is not physically or financially integrated, and the Recife metropolitan area has less than 50 km of exclusive transit lanes, resulting in long travel times due to the operation in mixed traffic.

Natal, the capital of Rio Grande do Norte’s State and the nineteenth most populated city in Brazil, hosts about 800.000 inhabitants, and is the core of a metropolitan area comprising twelve municipalities and 1.5 million inhabitants. The transit system managed by the municipality of Natal is composed by buses and microbuses: there are 646 buses in operation trans-porting 530,000 passengers daily and the fare is USD 0.68. The fleet of 177 microbuses run by individual operators is respon-sible for about 50,000 daily passengers. Two light rail transit lines, with a total length of 56.2 km connecting 22 stations, cover four municipalities in the Metropolitan Area (Natal, Parnamirim, Ceará Mirim e Extremoz). On weekdays, the light rail services transport 8200 passengers daily, and the fare is USD 0.13. There are concessionary fares for students and free fare for elderly (over 65 years). Bus services between metropolitan urban cores and Natal are managed by the State of Rio Grande do Norte and comprises around 250 buses and a hundred microbuses with a global daily patronage around 175,000 passengers. There are several information sources offered for the transit system in Recife and Natal. In the two cities, information is provided on buses regarding line name/number, fare price, direction (suburbs or city center), a short list of main destinations covered by the line, operating company, and customer service number. The information at bus stops varies: some of the bus stops provide written information regarding the bus lines that pass at the stop, while other bus stops are not clearly marked. As of March 2015, only 887 stops out of 5816 bus stops in Recife (15.3%) provide information regarding the list of lines that service the stop. In both cities, authorities’ official web-sites offer information about line itineraries, stops, timetables and fares, and there are also free central call services, which are heavily used. each month the customer central call service receives about 8000 calls and 70% of them are related to information on issues such as itinerary of buses and student cards. To cover the need for real-time information, private initiatives of information services have emerged such as the Citta Mobi app (Urbana-PE - Union of Passenger Transport Companies in the state of Pernambuco), the Moovit app and Google Maps. The Citta Mobi app offers real-time information about bus arrival time, showing the route and stops and whether the vehicle is adapted to people with disabilities. The Moovit app offers static integrated information combining bus/BRT and metro/RLT and allows planning trips based in origin-destination points. Google Maps provides static integrated information combining bus/BRT and metro/RLT.

3.1.2. Copenhagen, Denmark

The Greater Copenhagen Area (CGA) comprises 18 municipalities extending for about 3000 sq. km with a population of about 2 million inhabitants. The planning and the development of the transit system in the GCA follows the ‘‘finger-plan” directive, which indicates five cities (Køge, Roskilde, Frederikssund, Hillerød and Helsingør) as the direction from

(7)

Copen-hagen for the ‘‘fingers” (corridors) to be served by backbone transit and road connections. The transit network of the GCA consists of seven major modes: (i) metro, (ii) local trains, (iii), suburban trains (S-trains), (iv) regional and intercity trains, (v) regular buses, (vi) high-frequency buses (A-buses), and (vii) suburban and express buses (S-buses and E-buses). The metro serves central Copenhagen and the airport. The Reg- and IC-trains lead to north and west of Copenhagen as well as to the airport and further by a bridge to the urban Malmö Region in Southern Sweden with about 1 million inhabitants. The S-trains follow the radial finger lines from central Copenhagen to the mentioned five cities. E-buses and S-buses serve the S-train stations (primarily in rings), A-buses operate in central Copenhagen, and the remaining buses run in Copenhagen, the suburbs and the rural areas at a lower frequency. The system carries 220 million passengers per year and comprises 448 bus lines and 13,500 stops. The fare system is a zone-based system and different fares apply according to the payment method (i.e., smart travel card, mobile pay, periodic cards, single tickets). The average trip fare is 4.7$ (on 20 October 2015, the exchange rate is 1.00 USD = 6.57 DKK).

The transit system has an integrated computerized on-line information system (Rejseplanen), which enables the user to search for transit options by departure or arrival time, and offers real-time information. The provided information includes walk times, departure times, transfers, maps, fare price, delays and changes in the system. The information can be accessed via a mobile app and the system allows purchasing tickets. The app is free of charge, and the information is open access and hence used in third party products as well (e.g., Google Maps). In addition, a paid app (‘‘n

æ

ste bus”) provides information about the whereabouts of the buses and their estimated arrival time to the nearest bus stop. Real-time information is pro-vided at train stations and at bus stops in the city center and major locations. While there is a phone service, the commu-nication with the transit service provider regarding information/LOS/service changes/LOSt and found and other issues is via the automatic system.

3.2. Survey design and administration

A web-based survey was designed to investigate the research hypotheses for the case-studies. The survey elicited transit use, respondents’ socioeconomic characteristics, transit use patterns, perceived experience with transit, social norms, search and use of transit information, and the importance attributed to various information sources.

The frequency of transit use included frequency to mandatory and non-mandatory activities. The socio-economic char-acteristics comprised gender, age, residential arrangements, access to mobile phone with internet, employment/study status, access to travel options, form of payment for transit use, residential location and monthly expenses. Questions related to the transit use patterns elicited the frequency of each travel mode and the preferred travel mode to mandatory and non-mandatory activities. The perceived LOS was elicited in accordance with the SERVQUAL approach (Parasuraman et al., 1985), in line with previous studies regarding perceived transit LOS (e.g.,de Oña et al., 2013; Machado-León et al., 2016). Nevertheless, the included LOS items are those that are relevant to the role of information use and quality, which is the focus of the current analysis. Users’ perceived experience with transit enquired about the perceived LOS in terms of travel time and waiting time (speed/punctuality), information services (customer service/responsiveness/information/empathy), comfort (tangibles/individual space), coverage (accessibility/proximity), service reliability (reliability), cost (accessibility/availabil-ity), usefulness in terms of going to activities (accessibility/proxim(accessibility/availabil-ity), saving time and money, and avoiding driving (per-ceived quality with respect to other modes). The dimensions of information and security are related to the LOS but were investigated independently due to their relative importance in the current study context. In information provision, the emphasis was placed on the information content rather than courtesy due to short and informative communications and due to irrelevance when the information provision is via an on-line system and does not require personal communication. The sample statistics in the current study, reported inTable 3in the results section, support this approach as only 7% of the sample in Recife and Natal, and only 2% in Copenhagen stated that they call the information center of the transit provider. The familiarity with the transit system was elicited in terms of wayfinding, line combinations and frequency. Information quality was elicited in terms of reliability, clarity, completeness and efficiency, and personal security while travelling alone or at night. Questions about the social norms involved attitudes of family and friends towards the transit LOS and transit use. The questions were rated on a 5-point Likert scale from ‘‘strongly disagree” to ‘‘strongly agree”. A balanced questionnaire design including both positively and negatively phrased items was chosen as a well-established and widely-used coping strategy with acquiescent responding, namely the tendency to agree with the questionnaire items (Ferrando and Lorenzo-Seva, 2010). Reverse scoring was employed as the standard procedure in the data analysis in order to combine the negatively and positively phrased items.

The survey was administered in English, Danish, and Portuguese, to students from the Federal University of Pernambuco, Federal University of Rio Grande do Norte, the Technical University of Denmark and Copenhagen University during October 2015 via campus networks. Notably, the survey respondents are students, and therefore the results are not generalizable to all transit riders or other population groups which are significantly different in their socio-economic characteristics and accessibility to technology and information sources. Nevertheless, accessibility to technology should not impose a barrier because, according to an estimate of the Brazilian Internet Steering Committee, 85% of the Brazilians above 10 years of age own a mobile phone and 76% of the phones are estimated to be ‘‘smartphones”

The questionnaire took about 10 min and respondents were offered to participate in a raffle of 32 gift cards that could be redeemed at large stores as an incentive to complete the survey. In Denmark, the raffle comprised of 15 gift cards, varying between DKK100 and 500 (USD 15–75). In Brazil, the raffle consisted of 17 gift cards varying from R$50 to 400 (USD 13–100).

(8)

4. Results

4.1. Sample characteristics

In total, the survey yielded 1123 complete responses (63.2% from Brazil). The survey dropout rate was low as in Copen-hagen the rate of complete responses from the survey entries is 95%, and in Recife and Natal is 93%. The sample corresponds to 5% of the targeted university students in Copenhagen. In Recife/Natal, the sample corresponds to 1% of the university attendees and 5% from the students who read the e-mail.

Table 1presents the sample characteristics.Table 2presents the statistics of the perceived LOS and information quality.

Table 3presents the summary statistics of the use and importance of the various information sources.

The difference between the respondents in Copenhagen and the respondents in Recife and Natal in terms of LOS and infor-mation quality and availability is evident when combining the categories ‘‘strongly agree” and ‘‘agree” into a single category and the same with the ‘‘strongly disagree” and ‘‘disagree” categories from the 5-point Likert scale. The respondents in Recife and Natal perceived a lower LOS and information quality compared to Copenhagen. In Recife and Natal, more than 85% per-ceived long travel times and overcrowding, compared to 30% in Copenhagen. Less than 20% were satisfied with the transit cov-erage in Recife and Natal, compared to more than 60% in Copenhagen. Over 60% in Recife and Natal perceived frequent delays, compared to 30% in Copenhagen. In both case-studies, over 75% of the respondents perceived the transit system as expensive. In Recife and Natal less than 15% thought that there are good information services, compared to almost 60% in Copenhagen.

Table 2

LOS and information quality perceptions in Copenhagen versus Recife and Natal.

Item Categories (%)

Strongly disagree Disagree Neither disagree nor agree Agree Strongly agree

The travel and waiting time is too long Brazil 0.28 3.24 9.30 42.11 45.07

Denmark 6.05 34.38 27.85 27.60 4.12

There are good information services Brazil 22.82 44.79 19.01 11.55 1.83

Denmark 1.69 17.68 21.79 48.18 10.65

The vehicles and stations are too crowded Brazil 0.28 1.27 8.03 34.51 55.92

Denmark 3.39 32.93 33.90 23.73 6.05

The public transport coverage is good Brazil 22.11 42.68 18.73 13.24 3.24

Denmark 3.15 13.32 20.82 54.72 7.99

There often are service cancellations/delays Brazil 1.69 10.85 24.08 37.32 26.06

Denmark 4.36 33.66 31.72 21.07 9.20

Public transport is expensive Brazil 1.41 6.34 17.18 34.93 40.14

Denmark 0.97 2.66 12.59 31.96 51.82

Table 1

Sample characteristics in Copenhagen versus Recife and Natal.

Variable Country Categories (%)

Gender Male Female

Brazil 51.41 48.59

Denmark 62.71 37.29

Age 20 or less 21–25 26–30 31–35 Over 35

Brazil 25.07 55.49 12.11 3.24 3.24

Denmark 15.74 59.08 17.43 5.33 2.42

Residence Alone Roommates Parents Spouse

Brazil 6.90 10.70 71.97 10.42

Denmark 40.44 21.07 14.53 23.97

Employment status Only study Work/study Only work

Brazil 61.83 34.65 3.52

Denmark 54.72 41.40 3.87

Mobile/internet availability Paid internet Free internet No internet

Brazil 73.38 21.55 5.07

Denmark 87.41 7.75 4.84

Transit availability Available

Brazil 93.66

Denmark 94.92

Transit use Always Often Occasionally Rarely Never

Brazil 48.31 18.45 14.93 14.23 4.08

(9)

In Recife and Natal, the three most commonly used information sources are other passengers (89.01%), Google Maps (75.35%), and on-line information (72.11%). In Copenhagen, these are on-line information (93.95%), real-time information in vehicles or stops (85.47%), and Google Maps (81.60%). In both case-studies, less than 10% considered calling the public transport company as important. In Recife and Natal, asking passengers or the drivers was considered important by 88.71% compared to 21.07% in Copenhagen. In Recife and Natal, the four main reasons for information search are the need to arrive on time (48.17%), and going to unfamiliar places (42.96%), the university (37.61%), and the city center (37.46%). In Copenhagen, the reasons are the need to arrive on time (81.84%), going to unfamiliar places (64.89%), going out at night (44.31%), and going to the city center (38.98%). In both case-studies, over 90% of the respondents indicated that information on departure/arrival times, fastest route and changes/delays is important. Of the respondents in Recife and Natal, 60% indi-cated that they would use transit more frequently with better information (58.87%).

4.2. Exploratory factor analysis

Exploratory factor analysis served to extract the latent constructs for each case-study. Tests of internal consistency and sample adequacy constituted the necessary preliminary conditions for conducting factor analysis and obtaining meaningful results. The Spearman correlation matrix among the indicators provided the input for both the tests and the factor analysis. The activity-pattern items obtained in the survey demonstrate good internal consistency (Cronbach’s alpha = 0.791) and Table 3

The use of information sources and their importance in Copenhagen versus Recife and Natal.

Item Categories (%)

Use = yes Very important/important Use = yes Very important/important

Asking other passengers/transit personnel 89.01 88.17 35.35 21.07

Using real-time information in vehicles or stops 56.90 79.86 85.47 73.61

Calling the transit company 7.18 26.62 2.18 6.54

Using on-line information apps/internet 72.11 88.31 93.95 90.31

Using google maps 75.35 80.99 81.60 67.80

Table 4

Rotated factor analysis solution for Recife and Natal.

Factor

F1 F2 F3 F4 F5 F6

The travel and waiting time is too long 0.109 0.071 0.453 0.053 0.054 0.216

There are good information services 0.447 0.027 0.074 0.043 0.029 0.050

The vehicles and stations are too crowded 0.017 0.104 0.382 0.130 0.161 0.187

The public transport coverage is good 0.174 0.074 0.190 0.018 0.077 0.254

There often are service cancellations/delays 0.203 0.129 0.295 0.098 0.031 0.032

Public transport is expensive 0.090 0.147 0.189 0.282 0.141 0.321

Public transport allows me to go to my activities 0.087 0.325 0.028 0.632 0.015 0.099 Public transport allows me to go to new places 0.148 0.205 0.004 0.643 0.080 0.250

Public transport allows me to save time 0.197 0.028 0.326 0.142 0.035 0.276

Public transport allows me to save money 0.103 0.045 0.099 0.006 0.061 0.502

Public transport allows me to avoid driving 0.032 0.015 0.011 0.048 0.059 0.541

Public transport allows me to use the travel time to do things 0.119 0.030 0.159 0.091 0.055 0.366 I easily remember line combinations when I am asked 0.042 0.616 0.097 0.303 0.101 0.029 I usually remember the best route to arrive to my destination 0.003 0.721 0.090 0.273 0.047 0.027 I usually remember the frequency of the transit lines I need 0.056 0.690 0.021 0.214 0.053 0.039 I usually remember which stop is the closest to my destination 0.109 0.768 0.112 0.188 0.015 0.059 I usually remember the travel time to arrive to my destination 0.119 0.659 0.021 0.129 0.111 0.102 The travel time/waiting time information is reliable 0.602 0.077 0.254 0.032 0.009 0.081 The information about delays/changes is reliable 0.594 0.043 0.146 0.036 0.059 0.109 The information system provides efficient routes 0.763 0.095 0.079 0.031 0.008 0.188

The information is clear and complete 0.811 0.084 0.064 0.033 0.078 0.088

The information is easy to find 0.735 0.105 0.015 0.062 0.057 0.027

I am concerned about being pickpocketed/robbed 0.055 0.001 0.298 0.063 0.623 0.061

I am concerned about being harassed 0.014 0.019 0.064 0.063 0.687 0.182

I am concerned about walking/waiting at night 0.008 0.013 0.220 0.057 0.578 0.024

I am concerned about drunk passengers 0.031 0.020 0.009 0.024 0.629 0.053

Most of my friends use public transport 0.005 0.149 0.070 0.506 0.141 0.001

Most of my family members use public transport 0.039 0.167 0.099 0.521 0.137 0.099 My parents prefer that I do not take public transport 0.025 0.186 0.075 0.463 0.117 0.049 Most of my friends think that public transport is uncomfortable 0.084 0.034 0.683 0.009 0.217 0.041 Most of my friends think that public transport is unsafe 0.079 0.058 0.637 0.016 0.376 0.017 Most of my friends think that public transport is inefficient 0.152 0.010 0.613 0.002 0.056 0.106

(10)

good sampling adequacy according to Kaiser-Meyer-Olkin measure (KMO = 0.837). The Spearman correlation matrix shows the existence of correlations without multi-collinearity. The result of the Bartlett’s test of sphericity rejects the null hypoth-esis that the correlation matrix is an identity matrix (p = 0.000).

Exploratory principal axis factor analysis with Varimax rotation (Kaiser normalization) produced six factors according to the Scree plot. The factor loadings are presented inTables 4 and 5. In order to facilitate factor labeling, the dominant items, marked in bold, were defined as those with an absolute value of the loading greater than 0.30.

The factor structure for both case-studies is similar, with the main difference that in Recife and Natal the perceived factors are thematic, mixing personal attitudes and subjective norms, while in Copenhagen there is a differentiation between atti-tudes and subjective norms. Possibly, the result is related to information sharing among passengers in Recife and Natal. As reported in Section4.1, of the survey respondents in Recife and Natal 89.01% viewed other passengers as one of the three main information sources. The first factor (F1) for the Copenhagen case-study is labeled ‘‘Perceived level of service”, while for Recife and Natal the factor has a slightly different composition combining attitudes and subjective norms and is labeled ‘‘Perceived level of service to self and others”. The next three factors are labeled ‘‘Perceived information quality” (F2), ‘‘Per-ceived transit security” (F3), and ‘‘Per‘‘Per-ceived transit familiarity (F4)”. The factors F2, F3 and F4 are applicable to both case-studies. The fifth factor (F5) is labeled ‘‘Perceived transit usefulness” in Copenhagen and ‘‘Perceived transit usefulness to self and others” in Recife and Natal. The sixth factor (F6) in Recife and Natal is labeled ‘‘Perceived transit convenience”, while the last factor (F7) in Copenhagen is labeled ‘‘Pro-transit subjective norms”.

4.3. Model estimation results

The estimation results are presented inTables 6–8for Recife and Natal, andTables 9–11for Copenhagen.

For the two case-studies,Tables 6 and 9of the model present respectively the estimates of the measurement equations of the confirmatory factor analysis that corresponds to the exploratory factor analysis presented inTables 4 and 5. The objective of the exploratory factor analysis was to identify a set of latent constructs underlying a set of measured variables when with-out a priori hypotheses, while the objective of the confirmatory factor analysis was to test whether the data fit a hypothe-sized measurement model. In this study, the factor structure is revealed by exploratory factor analysis and is employed as part of the model structure with confirmatory factor analysis, which enhances the structural validity of the proposed model. The factor structure for the two case-studies is similar. The only difference is that in Copenhagen, the ETAM shows a clear Table 5

Rotated factor analysis solution for Copenhagen.

Factor

F1 F2 F3 F4 F5 F6

The travel and waiting time is too long 0.218 0.039 0.046 0.204 0.100 0.526

There are good information services 0.507 0.060 0.022 0.233 0.020 0.127

The vehicles and stations are too crowded 0.285 0.132 0.068 0.116 0.018 0.181

The public transport coverage is good 0.242 0.050 0.040 0.251 0.139 0.350

There often are service cancellations/delays 0.232 0.050 0.083 0.044 0.098 0.531

Public transport is expensive 0.017 0.033 0.093 0.075 0.202 0.320

Public transport allows me to go to my activities 0.145 0.237 0.062 0.165 0.603 0.096 Public transport allows me to go to new places 0.178 0.184 0.012 0.188 0.570 0.086

Public transport allows me to save time 0.111 0.003 0.003 0.295 0.466 0.383

Public transport allows me to save money 0.089 0.017 0.037 0.046 0.520 0.173

Public transport allows me to avoid driving 0.061 0.059 0.044 0.043 0.331 0.064

Public transport allows me to use the travel time to do things 0.064 0.011 0.036 0.003 0.223 0.223 I easily remember line combinations when I am asked 0.017 0.670 0.028 0.051 0.032 0.051 I usually remember the best route to arrive to my destination 0.029 0.850 0.010 0.014 0.022 0.011 I usually remember the frequency of the transit lines I need 0.050 0.571 0.040 0.002 0.176 0.023 I usually remember which stop is the closest to my destination 0.065 0.706 0.027 0.024 0.039 0.007 I usually remember the travel time to arrive to my destination 0.060 0.653 0.004 0.013 0.148 0.003 The travel time/waiting time information is reliable 0.503 0.095 0.039 0.153 0.013 0.394 The information about delays/changes is reliable 0.526 0.015 0.013 0.122 0.047 0.320 The information system provides efficient routes 0.586 0.014 0.080 0.027 0.153 0.116

The information is clear and complete 0.798 0.011 0.093 0.063 0.082 0.100

The information is easy to find 0.655 0.092 0.160 0.107 0.001 0.089

I am concerned about being pickpocketed/robbed 0.068 0.013 0.698 0.038 0.024 0.085

I am concerned about being harassed 0.148 0.071 0.819 0.040 0.005 0.054

I am concerned about walking/waiting at night 0.059 0.026 0.724 0.006 0.023 0.219

I am concerned about drunk passengers 0.014 0.040 0.705 0.033 0.130 0.009

Most of my friends use public transport 0.176 0.167 0.019 0.314 0.265 0.085

Most of my family members use public transport 0.115 0.054 0.093 0.344 0.324 0.051 My parents prefer that I do not take public transport 0.071 0.133 0.353 0.325 0.134 0.052 Most of my friends think that public transport is uncomfortable 0.161 0.004 0.137 0.671 0.037 0.162 Most of my friends think that public transport is unsafe 0.138 0.011 0.451 0.497 0.006 0.021 Most of my friends think that public transport is inefficient 0.221 0.107 0.015 0.620 0.190 0.252

(11)

distinction between the usefulness and the LOS of the system to oneself and the subjective norms. In Recife and Natal, respondents perceived the usefulness and the LOS to self and others without a clear distinction. The respective CFI, indicating the goodness-of-fit of the confirmatory factor analysis, are 0.956 and 0.909 for Copenhagen and Recife/Natal respectively, showing excellent data fit.

In both case-studies, the model including the set of measurement and structural equations fits reasonably well. The ratio between chi-square and degrees of freedom are 2.54 for Recife and Natal and 2.38 for Copenhagen, which are below the max-imum value (Ullman, 1996). The respective RMSEA are 0.047 and 0.058, below the maximum accepted values of 0.08 and Table 6

Measurement equations relating the indicators to the factors: Recife and Natal.

Variable Est. C.R. Variable Est. C.R.

Transit level-of-service (F1) Information quality (F2)

The travel and waiting time is too long (R) 1.000 – There are good information services 1.000 – The vehicles and stations are too crowded (R) 0.973 10.419 The travel time/waiting time information is

reliable

1.625 14.977 The public transport coverage is good (R) 1.536 13.880 The information about delays/changes is

reliable

1.507 14.640 Most of my friends think that public transport is unsafe (R) 1.679 14.179 The information system provides efficient

routes

1.755 15.842 Most of my friends think that public transport is inefficient (R) 1.359 14.054 The information is clear and complete 1.894 15.605 The information is easy to find 1.675 15.390

Transit security (F3) Transit Familiarity (F4)

I am concerned about being pickpocketed/robbed (R) 1.000 – I easily remember line combinations when I am asked

1.000 – I am concerned about being harassed (R) 0.822 18.651 I usually remember the best route to arrive to

my destination

1.128 31.270 I am concerned about walking/waiting at night (R) 0.949 19.737 I usually remember the public transport

frequency of the lines I need

1.029 30.649 I am concerned about drunk passengers (R) 0.653 15.505 I usually remember which stop is the closest to

my destination

1.126 32.459 I usually remember the travel time to arrive to

my destination

0.997 26.979

Transit usefulness (F5) Transit Convenience (F6)

Public transport allows me to go to my activities 1.000 – Public transport allows me to save money 1.000 – Public transport allows me to go to new places 0.906 16.960 Public transport allows me to avoid driving 1.219 6.924 Most of my friends use public transport 0.563 11.337 Public transport allows me to use the travelling

time to do things

0.837 7.013 Most of my family members use public transport 0.586 11.649

My parents prefer that I do not take public transport (R) 0.543 11.309 Note: (R) – Reversed coding in the case of negatively-phrased items.

Table 7

Structural equations relating the factors to respondents’ characteristics: Recife and Natal.

Level-of-service (F1) Information quality (F2)

Male 0.346 6.044 Less than 18 0.286 2.624

Pay with cash 0.109 1.795 High expense with transport 0.317 2.123

Study and work 0.090 1.606 Ask people about PT info 0.119 2.035

Lives in Recife 0.168 2.375 Use real-time info 0.099 2.390

Lives in Natal 0.198 2.588 Lives in Natal 0.134 2.178

Lives in Natal Metropolitan Region 0.210 2.516 Access to bicycle 0.058 1.689

Male 1.035 10.887 Less than 18 0.437 3.520

Lives in Natal Metropolitan Region 0.255 1.674 Access to car/car sharing 0.263 3.715

Access to car/car sharing 0.232 2.400 Pay with cash 0.563 7.439

Use real-time info 0.236 2.863 High expense with transport 0.410 1.831

Lives in Recife 0.226 2.612

Lives in Natal 0.285 2.877

Transit usefulness (F5) Transit convenience (F6)

Use GPS/paid internet 0.281 3.680 Male 0.185 2.912

Access to PT system 0.402 2.683 Live with parents 0.140 1.803

Pay with cash 0.520 5.669 Live with partner 0.222 1.796

Low expense with transport 0.323 2.531 Pay with cash 0.162 2.157

Lives in Recife 0.230 2.309 Access to car 0.265 3.870

(12)

0.10. The respective CFI are 0.880 and 0.777, which are reasonable according toLoehlin (1998). Therefore, the statistical model confirms the general behavioral framework and the research hypotheses for the two case-studies.

Tables 7 and 10of the model show the relationship between the latent constructs and the respondents characteristics. Most importantly, we focus on the effect of the use of real-time information on the various latent constructs. The model con-firmsH1only for the Recife and Natal case-studies. Real-time information, as an observed variable, explains the latent con-structs of higher perceived LOS, information quality, and perceived familiarity. In Recife/Natal, as shown inTable 7, the use of Table 8

Structural equations relating the latent factors to transit use: Recife and Natal.

Explanatory/mediator variable Dependent variable Est. C.R.

Information quality (F2) Transit Familiarity (F4) 0.124 2.059

Transit security (F3) Level-of-service (F1) 0.398 11.171

Information quality (F2) 0.432 7.797

Level-of-service (F1) Transit Usefulness (F5) 0.069 1.036

Transit convenience (F6) 0.366 4.746

Transit Familiarity (F4) General transit use frequency 0.498 8.866

Transit Usefulness (F5) 0.746 13.534

Transit Familiarity (F4) Non-routine transit travel at night 0.299 5.333

Information search for work trips 0.232 2.031

Information search for night trips 0.937 7.062

Information search for trips to unfamiliar places 0.359 3.220

Transit Familiarity (F4) Non-routine transit travel to unfamiliar/new places 0.269 4.750

Information search for work trips 0.206 1.810

Correlation patterns

Non-routine transit travel at night General transit use frequency 0.123 2.566

Non-routine transit travel to unfamiliar/new places General transit use frequency 0.050 1.062 Non-routine transit travel to unfamiliar/new places Non-routine transit travel at night 0.488 13.456

Table 9

Measurement equations relating the indicators to the factors: Copenhagen.

Transit level-of-service (F1) Information quality (F2)

The travel and waiting time is too long (R) 1.000 – There are good information services 1.000 – The public transport coverage is good 0.987 9.355 The travel time/waiting time information is reliable 1.166 13.840 There often are service cancellations/delays (R) 0.733 7.560 The information about delays/changes is reliable 1.081 14.404 Public transport is expensive (R) 0.481 4.882 The information system provides efficient routes 1.112 14.048 The information is clear and complete 1.377 16.018 The information is easy to find 1.057 14.546

Transit Security (F3) Transit Familiarity (F4)

I am concerned about being pickpocketed/robbed (R) 1.000 – I easily remember line combinations when I am asked 1.000 – I am concerned about being harassed (R) 1.171 26.857 I usually remember the best route to arrive to my

destination

1.220 20.777 I am concerned about walking/waiting at night (R) 1.030 28.057 I usually remember the public transport frequency of

the lines I need

0.815 15.793 I am concerned about drunk passengers (R) 0.931 24.915 I usually remember which stop is the closest to my

destination

1.052 21.225 My parents prefer that I do not take public transport (R) 0.513 9.005 I usually remember the travel time to arrive to my

destination

0.933 17.474

Transit Usefulness (F5) Pro-transit subjective norms (F7)

Public transport allows me to go to my activities 1.000 – Most of my friends use public transport 1.000 – Public transport allows me to go to new places 0.962 13.29 Most of my family members use public transport 1.069 8.032 Public transport allows me to save time 0.911 12.49 Most of my friends think that public transport is

uncomfortable (R)

1.556 9.192 Public transport allows me to save money 0.525 6.87 Most of my friends think that public transport is

unsafe (R)

1.037 7.694 Most of my friends think that public transport is

inefficient (R)

1.710 9.661

(13)

real-time information is positively and significantly relates to the perceived information quality and transit convenience, while it negatively and significantly relates to transit security. For the Copenhagen case-study, the use of real-time informa-tion is not statistically significant. Possibly, since in Copenhagen real-time informainforma-tion is widely used as the dominant infor-mation source, it does not affect the perception of the system among users in Copenhagen.

Tables 8 and 11of the model presents the estimates of the structural equations associated with the relationship across the latent constructs. The statistical model confirmsH2. In Recife/Natal (Table 8) general transit use and non-routine travel are explained by the perceived transit familiarity and usefulness to self and others. In Copenhagen (Table 11), general transit use and non-routine travel are explained by the perceived transit usefulness, transit familiarity and subjective norms.

The statistical model in both case-studies confirmsH3. The perceived transit familiarity and perceived information qual-ity are important components of the perceived ease-of-use of the transit system. In both case-studies (Tables 8 and 11), information quality (F2) is significantly associated with LOS and transit familiarity, and has higher weight than transit secu-rity (F3) in its effect on the perceived LOS.

The correlation patterns inTables 8 and 11confirmH4. The error terms of the general and non-routine transit use for night trips and unfamiliar trips are positively correlated.

Table 10

Structural equations relating the factors to respondents’ characteristics: Copenhagen.

Level-of-service (F1) Information quality (F2)

Male 0.146 1.604 Lives with roommate/friend 0.201 2.159

Live with child/children 0.862 1.976 Access to bike sharing 0.279 1.563

Access to PT system 0.217 1.375

Pay with monthly card 0.218 1.534

Lives in Copenhagen 0.162 2.026

Male 0.226 2.477 Live with parents 0.218 1.717

Lives in Copenhagen 0.201 2.010 Live with partner and child 1.078 1.794

Access to bicycle 0.543 3.349 Pay with phone app 0.478 2.668

Pay with phone app 0.349 1.598 Pay with monthly card 0.643 4.127

Pay with student card 0.442 3.219

Transit usefulness (F5) Pro-transit subjective norms (F7)

Car use 0.242 1.830 Pay with monthly card 0.431 3.248

Access to bicycle 0.447 2.767 Low expense with transport 0.218 1.993

Average expense with transport 0.207 1.704

Table 11

Structural equations relating the latent factors to transit use: Copenhagen.

Explanatory/mediator variable Dependent variable Est. C.R.

Transit security (F3) Level-of-service (F1) 0.132 2.817

Information quality (F2) 0.646 9.937

Information quality (F2) Transit Familiarity (F4) 0.185 2.958

Level-of-service (F1) Transit Usefulness (F5) 0.593 7.374

Pro-transit subjective norms (F7) 0.863 7.732

Transit Familiarity (F4) General transit use frequency 0.259 3.512

Transit Familiarity (F4) Non-routine transit travel at night 0.252 3.781

Information search for arriving on time 0.304 1.851

Transit Familiarity (F4) Non-routine transit travel to unfamiliar/new places 0.148 2.195

Information search for arriving on time 0.500 2.974

Correlation patterns

Non-routine transit travel at night General transit use frequency 0.145 2.937

Non-routine transit travel to unfamiliar/new places General transit use frequency 0.317 6.998 Non-routine transit travel to unfamiliar/new places Non-routine transit travel at night 0.413 9.456

(14)

Tables 8 and 11confirmH4. In Recife/Natal, non-routine transit use is associated with real-time information search for work trips, night trips and trips to unfamiliar places. In Copenhagen, non-routine transit use associates with information search for night trips, unfamiliar places and for information search in order to arrive on time.

5. Conclusions

This study addresses the behavioral mechanism underlying the effect of information provision and quality on transit use of students as highly skilled young adults, which is of interest to transport policy-makers and transit operators. The findings show that information quality and real-time information are important factors influencing transit use by young adults with high prospects of income and motorization rate, and that information has a system-wide influence on the perceived quality of the transit system, rather than merely being an additional service. The findings show also that information quality is important both for general as for non-routine transit use, with real-time information search associated with trips at night and to unfamiliar places. The information search is associated with the need for punctuality in Copenhagen, and work-trips in Recife/Natal, which may be attributed to differences in the current information quality. Recent studies show that transit information is an integral part of service quality (e.g.,de Oña et al., 2013). Our results show the complex relations between transit information and transit use. Our findings show that information quality explains the perceived transit LOS and perceived transit familiarity, which influences the perceived usefulness of the transit system, and eventually leads to higher frequency of transit use.

Two case-studies, one from Brazil and the other from Denmark, show the validity of the proposed behavioral framework and the transferability of the behavioral mechanism across regions with differing cultures and transit level-of-service. The two case-studies differ geographically, culturally, and in terms of transit and information provision. Yet, many similarities were found regarding the information sources used, the reasons for searching information, the type of information sought, and the importance attributed to various information sources. Similar structural relations show the role of information as a motivator for transit use intentions in both case-studies.

While the findings seem common knowledge at first glance, the staggering fact is that integrated information systems are relatively new and Copenhagen is a unique example of an established, integrated on-line and real-time information system, provided by the authorities. Formal integrated transit information systems are scarce and information provision is left to pri-vate initiatives because transit operators do not provide high-quality transit information (Ibraeva and Figueira de Sousa, 2014). Interestingly, while every consumption product is coupled with information leaflet, information on public transport as the backbone of urban systems is lacking.Ibraeva and Figueira de Sousa (2014)state that ‘‘Some companies to believe that residents already know about their products anyway, others may think that detailed information is not necessary because a client can always ask a driver or company’s representative for it if needed”. The findings of the current study show that, when on-line information resources are unavailable, passengers regularly consult other passengers and the transit personnel. Neverthe-less, regardless of the availability of formal information, most passengers use and consider important on-line information resources and applications. The unavailability of transit information is not unique to the current case-study. In many cities, both in developed, emerging economies and developing countries throughout South-America, Asia, the Middle-East, South and Eastern Europe transit passengers transit passengers are relying on informal on-line applications, developed by private entrepreneurs, for obtaining real-time transit information. While some of these applications enjoy the support of the formal authorities and rely on formal data, there is merit in an integrated real-time information system, where the system notifies the users about service interruptions, and where there is continuous monitoring of data accuracy and reliability. In fact, the Danish case-study shows that in the existence of such a system, the vast majority of the passengers will use it and the con-sultations with transit personal can reduce from 90% to 30% compared to the case-study in Brazil, where most passengers consult the transit personnel. While many transport operators view information as an additional tool independent from the overall LOS, the findings of this study show that on-line and real-time information are essential building blocks of the transit system. As such, while it is important to encourage spontaneous private initiatives, public sector authorities should provide formal integrated, on-line and real-time information systems as an integral part of transit provision. The policy implication is that, compared to the high costs of transit infrastructure and LOS improvements in terms of frequency, avail-ability and reliavail-ability, improving information quality could be attractive for increasing transit ridership with relatively mod-est invmod-estments.

Acknowledgements

The authors gratefully acknowledge that the study forms part of the project ‘‘Integrated Public Transport Optimization and Planning” (IPTOP), funded by the Innovation Fund Denmark. The work was conducted during the stay of Mayara Moraes Monteiro at DTU, supported by CNPq Brazil - National Council for Scientific and Technological Development. The authors are thankful to three anonymous reviewers for their helpful comments.

References

Aarts, H., Verplanken, B., van Knippenberg, A., 1997. Habit and information use in travel mode choices. Acta Psychol. 96, l–14.