Public transport usage among university students: what to expect based on customer sa"sfac"on survey (CSS) analysis

Recebido: 8 de agosto de 2018 Aceito para publicação: 23 de fevereiro de 2020 Publicado: 31 de agosto de 2020 Editor de área: Sara Ferreira ABSTRACT Customer Sa#sfac#on Surveys (CSS) are valuable tools for both public management and opera#ng companies. It provides useful informa#on to understand public transport users’ needs, to define more effec#ve goals, and to make more asser#ve investments. In the literature, the combina#on of market segmenta#on and the comparison between sa#sfac#on and rela#ve importance data have been rarely assessed. This study inves#gates the tendency for public transport con#nuous adop#on among undergraduate students from Curi#ba (Brazil) through usage habits, perceived sa#sfac#on, rela#ve a:ribute importance and GAP analysis. Overall, the results do not show an inten#on from the studied group to keep using this mode into the future. However, if properly mo#vated, students are more likely to become agents of change and could posi#vely affect behavioral inten#ons toward public transport.


INTRODUCTION
In developing countries, a strong correlation is found between increasing motorization rates, income inequality and income growth (Kutzbach, 2009). This scenario leads to unsustainable congestion volumes, increasing traf#ic accident rates and major environmental and social impacts (Filipović et al., 2009;Gössling, 2013). In addition, bus average speed is usually reduced providing an additional incentive to perform a modal switch (Kutzbach, 2009), thus creating a vicious cycle and aggravating mobility problems.
Increasing road capacity is a frequent, but unfeasible solution to accommodate the escalating trend on car ownership since cities are bound by #inancial and geographical constraints (Ercan et al., 2017). Thereby, adoption of active and public transport on a larger scale is usually linked to a shift in mindset. In this way, urban infrastructure development should be thought as to reduce economic, social and environmental impacts of extreme car dependency (Litman, 1999;Onat et al., 2016).
Sustainable development could be promoted by offering a wider range of travel options and more effective land use (Litman, 1999) or even by creating more compact communities and so reducing trip generation rates and average trip lengths (Ercan et al., 2017). Also, cooperative behaviors should be built through individual involvement on community decisions toward a sustainable environment (Litman, 1999;Van, Choocharukul and Fujii, 2014;Javid et al., 2016). Likewise, the application of soft transport policy measures, such as personalised travel planning, public transport marketing and travel awareness campaigns, have shown a positive effect on reducing car usage (Möser and Bamberg, 2008;Friman, Larhult and Gärling, 2013).
As to increase public transportation ridership, public authorities should also concentrate on adjusting its systems to passengers' needs, once loyalty is strongly associated with overall satisfaction (Lai and Chen, 2011;van Lierop and El-Geneidy, 2016). Consequently, it is essential to have knowledge of which service attributes have a greater impact on perceived quality (dell'Olio, Ibeas and Cecıń, 2010;de Oña et al., 2016;Guirao, Garcıá-Pastor and López-Lambas, 2016), thus developing effective policies (Bordagaray et al., 2014;Abenoza, Cats and Susilo, 2017).
This paper focuses on investigating the tendency for continuous public transport adoption among undergraduate students in the long term. As highlighted by Mouwen (2015), it is more ef#icient to segregate the population and apply resources into a few manageable groups that are likely to have a change in behavior, whilst improvements would still impact positively the overall population (Bordagaray et al., 2014). To this end we applied a customer satisfaction survey (CSS) encompassing habits, satisfaction at attribute and overall levels, relative attribute importance and gap analysis between perceived and expected quality of service. Our work adds to the literature since few studies so far have analysed and compared importance and satisfaction data for a speci#ic market segment.

HIGHLIGHTS OF CUSTOMER SATISFACTION THEORY
Satisfaction is believed to have positive impacts on both behavioral intentions  and loyalty (van Lierop and El-Geneidy, 2016). Therefore, it is a strong indicator of whether a customer will continue to use a service (Zeithaml, Berry and Parasuraman, 1996). Additionally, it is in#luenced by perceived value, related to the discrepancy between bene#its and costs, and service quality, an abstract concept related to the comparison between expected and delivered service (Parasuraman, Zeithaml and Berry, 1985;Widianti et al., 2015).
As to improve public transport mode-share, transit managers should understand what makes a customer satis#ied. The answers should be applied to increase the system attractiveness and motivate long term usage (van Lierop and El-Geneidy, 2016) and loyalty behaviors, such as spread of positive word-of-mouth and willingly reuse (Lai and Chen, 2011), while discouraging car dependency (Fu and Juan, 2017).
Customer Satisfaction Surveys (CSS) are the most disseminated tool for gathering data concerning perceived performance satisfaction and attribute relative importance (de Oña and de Oña, 2015). Together, attribute-performance and attribute-importance help transit managers visualize a global scenario for passengers' customer evaluation (Wang, Feng and Hsieh, 2010), thus, providing useful information to de#ine more effective goals and more pro#itable and assertive investments (Mouwen, 2015). Additionally, once users' needs and preferences are heterogeneous, it is necessary to understand the variety of behaviors to create more adaptable strategies (Yaya et al., 2015;de Oña et al., 2016).
Service attributes are usually selected based on a combination of methods, such as extensive literature research, focus groups, pilot users survey, and statistical tests (de Oña and de Oña, 2015). Research has proven that model predictive value increases if attributes were chosen accordingly to context (Carrillat, Jaramillo and Mulki, 2007). Given the existence of an extensive number of attributes, they are usually gathered into smaller groups called dimensions. However, there is no consensus on their nature (Morton, Caul#ield and Anable, 2016). Parasuraman, Zeithaml and Berry (1985) initially proposed 10 different dimensions, which vary on a continuum from easy to dif#icult to measure. Through factor and reliability analysis, these dimensions were reduced into 5, leading to the #irst service quality evaluation procedure: the SERVQUAL methodology (Parasuman, Zeithaml and Berry, 1988).
In transportation research, service frequency, reliability, travel time, comfort and cleanliness, information availability, staff behavior, and fare are usually considered (Dell'Olio, Ibeas and Cecin, 2011;Lai and Chen, 2011;Guirao, Garcıá-Pastor and López-Lambas, 2016) and have also been combined into several different dimensions. For example, the Transportation Research Board classify them into availability, easiness to access the public transport system, comfort and convenience factors, likeness of a potential user to become a frequent one (TRB, 2013). Given the complex, fuzzy and abstract nature of service quality, many other attempts to illustrate it can be found in the literature (e.g. Eboli and Mazzulla, 2008;Lai and Chen, 2011;Yaya et al., 2015;Mahmoud and Hine, 2016).
There are two approaches for estimating relative importance of service attributes: Stated and Derived Importance (de Oña and de Oña, 2015). Stated Importance procedures rely on directly asking users for the attribute importance, usually through psychosomatic scales (Dell'Olio, Ibeas and Cecin, 2011). It is a more intuitive method but that has several week points: it increases survey length, which can reduce response rate and accuracy, and it might also fail to differentiate mean importance ratings, once respondents tend to rate attributes at the top levels of the scale, thus yielding factors that have little in#luence in customer satisfaction as relevant (de Oña and de Oña, 2015). Nonetheless, it is most used by operation companies due to its simplicity (Guirao, Garcıá-Pastor and López-Lambas, 2016).

THE CASE STUDY: CURITIBA'S PUBLIC TRANSPORTATION SYSTEM
Curitiba's transit-oriented development was born from a 1965 master plan. In later years, it would become an international reference in sustainable urban development and the Bus Rapid Transit (BRT) model would be exported to over 150 cities around the globe. Currently, the intended and unintended positive co-bene#its from combining mixed and dense land-use and public transport are still noticeable throughout the city (Doll and Oliveira, 2017). Nonetheless, following a national trend, ridership within the system has been decreasing (NTU, 2017) leading to #inancial unbalance and general cuts in investments. On the other hand, car ownership has been increasing and stands as the highest vehicle/habitant ratio among Brazil's State capitals. This study aims to understand whether undergraduate students, which represent about 10% of Curitiba's population, would continue to adopt public transportation on the long term given the current conditions. A questionnaire was applied online through the broadcast mailing systems of three major local universities. It was divided into 4 sections: sociodemographic characteristics, reasons for public transportation adoption, service attribute and, overall satisfaction and stated attribute importance.
12 reasons were selected from the literature (Duarte et al., 2016;Machado-León, de Oña and de Oña, 2016;Gurrutxaga et al., 2017;Xia et al., 2017) to investigate the nature of the factors leading this group to use the public transport. Principal Component Analysis and Pearson Product-Moment Matrix were applied to better understand data covariance structure.
Satisfaction and Stated Importance were evaluated through a 5-point Likert Scale for 17 different positive statements. In both cases, Factor, Item Reliability and Discriminant Analysis were employed to recognize common trends among the sample. In the Discriminant Analysis, heterogeneity was assessed through several sociodemographic and usage descriptive variables, including the effects of Negative Social Safety Experiences (NSSEs), related to encounters with theft and robbery, on satisfaction and relative importance. Finally, a GAP Analysis was assessed to evaluate the difference between mean satisfaction and importance results.

RESULTS AND DISCUSSIONS 4.1. Descrip"ve analysis and public transporta"on usage
The survey received 562 responses from which 544 were drawn as valid. It ful#illed the sampling guidelines set by Bartlett, Kotrlik and Higgins (2001) to ensure generalized results at a 5% sample error level, since 384 answers were required for a population of about 178 thousand people. On average, the sample age mean was of 22 (s.d. 3.4) years old and followed the gender distribution published in the 2010 Brazilian census -52.9% female and 47.1% male.
Our analysis of the undergraduate students' habits shows that 93% of the sample use public transport for educational purposes, followed by leisure activities (60%) and commute trips (41.7%). Additionally, we found a tendency for leisure and shopping trips to be more pronounced among people with high income, while for the ones with lower income, commute and health-related trips are more common. While, in general, respondents use the public transport for up to 2 purposes (52%), the lower income group shows a higher dependency on this mode: 34.5% use public transportation as the main mode for four or more trip purposes.
The system is used mostly during mornings (39%) and afternoons (39.8%). It also has more people riding it twice a day (61.8%), than once a day (25.7%). Predominantly, the sample is composed by frequent users, since 70.2% use it from 5 to 7 times a week. Only 13.1% use it less than twice a week.
Curitiba's public transportation system is composed by 7 distinct bus-line categories, which are easily identi#ied by speci#ic colors re#lecting its planning based on the nexus land use, transportation, and road system. For example, among the most frequent choices in the sample it is possible to identify BRT lines (79%) by the red color, feeder lines (48.7%) by yellow, and direct lines (41.2%) by grey. To make it even more coherent different lines run on different road hierarchies associated with speci#ic population densities. As a result, BRT lines, which serve high-density axis, represent from 44% to 84% of ridership in each corridor.
Furthermore, 72.2% use 2 or more bus lines to complete their usual routes, which might imply in lower travel time satisfaction due to relatively longer transfer time. 75.6% reported #inding overcrowded conditions at least three times a week from which 45.6% face it daily. As expected, this situation is more frequent during peak hours (early mornings and late afternoons) and in the lines with more ridership: BRT, feeder, and direct lines.
PCA showed that the #irst 4 linear combinations held 50.2% of the total variance and yielded 0.686 for the Kaiser-Meyer-Olkin (KMO) criteria, which alone re#lect a low level of sample adequacy. However, Bartlett's Test of Sphericity refused the hypothesis of no common variance at a 95% con#idence level. The scatter plot derived from the 2 #irst components revealed a greater proximity (1) among reasons that do not show an intention to keep using this transportation mode, such as "I do not have a car", "It's my only alternative" and "I do not have a driver's license", and (2) between the ones that favour system attributes (e.g. comfort, frequency, price and speed), traf#ic conditions and environmental awareness.
The high number of answers within the former group (1) indicates a concerning future, once current users are more likely to migrate to individual modes given a shift in their personal #inancial status (Kutzbach, 2009). In addition, as system attributes are the least chosen, behavioral intentions are probably low toward the system, since comfort and convenience factors are strong predictors of loyalty to public transportation modes (TRB, 2013).
Pearson Product-Moment Matrix revealed a slight correlation between traf#ic conditions and system speed (r = 0.37, p < 0.01), and not being able to use the car and the lack of parking spaces (r = 0.27, p < 0.01). Regarding the #irst correlation, it supports that Curitiba' extensive exclusive bus lanes, which have a positive effect in decreasing travel time ratio during peak hours, are appreciated by users. The second correlation might indicate that reducing parking spaces within the city will be an effective measure to decrease car ridership within chosen areas.

Sa"sfac"on at A?ribute and Overall Levels
Satisfaction was evaluated through a 5-point Likert scale in a 17-question survey. The descriptive analysis shows fare (1.62), safety and security (2.17), and crowding conditions (2.30) having the worst results. Skewness values re#lected that for the #irst 10 attributes, the left tail is longer in the distribution curve, therefore indicating higher dissatisfaction, as presented in Table 1. On the other hand, for overall satisfaction the distribution is approximatively even. Safety and security, staff behavior, exclusive bus lanes availability, and application of sustainable technologies were excluded from the analysis, due to its loading factors (<0.5). Furthermore, as Integration and Critical Attributes have low values for Cronbach's alpha, they can not be considered as reliable factors.
A Discriminant Analysis was performed between satisfaction variables and several descriptive ones. Women were found to be, on average, 6% less satis#ied with the system than men, which is a re#lex of the lower rates this group gave for fare (-12.1%), bus stops infrastructure (-12%), accessibility (-11.1%), safety and security (-10.4%) and vehicle interior conditions and crowding conditions (-10.3%). This might be because women, in this sample, are more dependent on public transport than men, therefore facing more overcrowding conditions (+11%) and being more likely to be robbed or to see a robbery in the system (+8%). As a result, they have chosen more frequently the reasons that do not show an intention to keep using this mode.
Dependency, in general, was found to be a reducer of satisfaction. In the sample, those who use public transport more frequently showed lower satisfaction rates. For example, the group who use it three times a day is 9% less satis#ied than people who use it twice a day. Overall satisfaction levels also decreased among people using the system more times a week and for more trip purposes. Moreover, these users usually ride feeder lines more often, which serve areas further away from the structural axis and have lower service frequency. As a result, safety and security, transit connectivity, fare, bus stop infrastructure, accessibility and travel time scores were affected.
On the other hand, the group that use public transportation due to traf#ic conditions perceives travel time (+24.9%), fare (+18.5%), service frequency (+12.6%), accessibility (+12.1%) and exclusive bus lanes (+10.5%) much better than other groups. However, the ones who use it because they do not have a car or for being the only alternative granted lower results for fare, service frequency, accessibility, travel time and crowding conditions. These results, therefore, strengthens the idea that the construction of positive behaviors and mindsets toward public transportation could have promising results.

Rela"ve Importance through Descrip"ve Analysis
Evaluating service attributes importance through Descriptive Analysis is the most used procedure to investigate relative importance. In this research, respondents were asked to rate the same 17 variables in a psychometric scale (extremely unimportant to extremely important). As anticipated, mean values were much higher than the ones found for satisfaction. Customers leaned toward scoring nearly all attributes at the top scales, hence making it dif#icult to differentiate them. Nonetheless, item reliability test resulted in 0.894 for the Cronbach's alpha test, thus indicating a consistent set of variables, even though there was only up to 5% difference between two subsequent attribute scores.
Safety and security (4.67), fare (4.61), fare integration (4.54), reliability (4.54), service frequency (4.53) and accessibility (4.30) appeared among the most important attributes ( Table  2). Application of sustainable technologies (3.87) and staff behavior (3.77) were among the lowlevel importance attributes. Surprisingly, transit connectivity (3.64) and crowding conditions (3.60) also appeared among this category. The #irst six factors were correlated to Overall Satisfaction through Multiple Linear Regression Analysis (Ordinary Least Squares). The R² was of 76.83%, which indicates how much variance these attributes can explain at a 95% con#idence level. A Factor Analysis was performed following the same principles as the previous one. This was possible since all collected data was at the same scale and because the Cronbach's alpha results were extremely positive (0.894). The Factor Analysis aimed at explaining underlying patterns of data correlation through the reduction in the number of components. The data set produced 0.926 for the KMO test (great sample adequacy) refusing the non-correlation hypothesis through Bartlett's Test of Sphericity at 95% con#idence level. The two derived factors explained 47.1% of variance and were related to Comfort and Accessibility and System Performance, as presented below: • Comfort and Accessibility: This set of factors is much more reliable than the one found before, due to its higher level of consistency. System Performance is related in a wider level to service characteristics and it re#lects the most relevant factors found in this Stated Importance procedure. Comfort and Accessibility is a combination of supplementary and convenience features. Crowding conditions and fare connectivity were excluded from the analysis due to low factor loadings (<0.5).
The correlations between each variable and the Overall Satisfaction were further investigated through Model Regression Analysis. Service frequency (43.5%), bus stop infrastructure (41.2%), accessibility (37.9%), *leet conservation (35.5%), vehicle interior conditions (31.1%) showed the greater R² values, thus indicating the importance of investments on Comfort and Accessibility attributes as to retain and increase ridership among the studied group.
Discriminant Analysis showed that Service Performance attributes usually do not vary in importance among the analysed descriptive variables. Nevertheless, results from Comfort and Accessibility features differ given variations on perceived satisfaction. For example, women are more sensitive to crowding conditions (+14%) and accessibility (+7%); the lower income group values transit connectivity (+10%), bus stop infrastructure, road conditions and exclusive bus lane availability (+8%); the ones who use the bus for several trip purposes value accessibility (+10%), transit connectivity (+9%), travel time and *leet conservation (+9%); the ones who use it several times a day highly regard transit connectivity, crowding conditions (+8%), accessibility and *leet conservation (+6%). All these attributes were the ones which had lower satisfaction rates for each group, therefore that is probably what makes them more noticeable and, by consequence, more important.
Also, users who face more overcrowding conditions value better crowding conditions (+16%), vehicle interior conditions (+9%) and #leet conservation (+7%). Users who have seen a robbery or have been robbed in the system attribute a relatively higher value to safety and security (+3%). This value may seem a low increase, but it is an attribute that have not shown any variation for any other descriptive variable. Furthermore, the group that use public transportation due to traf#ic conditions were less troubled with crowding (-6%) and more concerned with exclusive bus lane availability (+7%) and transit connectivity (+5%). On the other hand, the ones that use the bus for the reasons of not having a car or for being the only alternative give more weight to crowding conditions.
Given the results, once Service Performance attributes are well stablished, public authorities and operation companies might want to focus on developing better crowding conditions and improving transit connectivity and *leet infrastructure as to retain and increase ridership among undergraduate students.

GAP Analysis: Expected vs. Perceived Service
Service quality is believed to be perceived along a spectrum ranging from ideal to unacceptable quality, in which a customer position depends on the nature of the existing gap between expected service and how the delivered service is perceived. Consequently, when expectations are higher than what is delivered, perceived quality tend to the lower side of the satisfaction range (Parasuraman, Zeithaml and Berry, 1985). Therefore, it is important for public authorities and service providers to understand the needs of public transportation users. By understanding these needs, services perceived as of high quality can be delivered, in#luencing positive wordof-mouth and increasing ridership. This knowledge is also an asset for developing effective policies and marketing campaigns.
It was found a 52.4% mean difference between mean Overall Satisfaction and mean Overall Importance (Table 3). The 17-variable data set for both criteria was also put through a Canonical Correlation Analysis, which yielded 45.3% for the #irst eigenvalue, thus indicating the existence of a GAP between expected and perceived services among the studied group. Furthermore, since importance values are higher, satisfaction tends to be perceived on the lower side of the spectrum. Therefore, once the most chosen reasons for usage does not favour loyalty toward public transportation and 58.8% of the attributes have a broader dissatis#ied distribution, undergraduate students are not likely to keep adopting this modal. At the attribute level, the expected -perceived gap was assessed by the difference between mean satisfaction and importance results. By comparing, satisfaction and relative importance scores, the attributes are put into perspective. It enables the transit analyst to perceive easily TRANSPORTES | ISSN: 2237-1346 42 which attributes are more critical and will have more impact on the public perception if improved. For this sample, the most pronounced difference was found for fare (3.00) and safety and security (2.50). These attributes are characterised by low satisfaction and high importance, therefore more effort should be put into improving them, as depicted in the Quadrant Chart Matrix from Figure 1. Through Multiple Linear Regression Analysis, it was found that the #irst 7 attributes could hold 81.1% of sample variance. In comparison, through the relative importance results alone it was needed 8 factors to explain the same amount. Thus, this analysis aids to better select where to invest and market improvements as to create a positive sphere around public transport according to local needs.

CONCLUSION
Cities are bound by #inancial and geographical constraints and, therefore, continuous increase in road capacity is an unfeasible solution to accommodate current trends on car ownership. In this way, promotion of public transport is necessary as to reduce economic, social, and environmental impacts of extreme car dependency, such as unsustainable congestion volumes, increasing traf#ic accident rates, health-related issues, and overall quality of life.
We investigated the tendency for continuous public transportation adoption among undergraduate students from Curitiba. In this sense, the results indicate that under the current conditions the analysed group is likely to perform a mode switch towards individual modes. For instance, "I do not have a car", "It's my only alternative", and "I do not have a driver's license" were among the most common reasons for using public transport, while service and comfort aspects of the service were signi#icantly less favored. Thus, suggesting negative behavioral intentions and attitudes towards public transport.
The analysis of satisfaction provides further evidence to this claim. 58.8% of the attributes have a broader dissatis#ied distribution from which Fare, Safety and Security, and Crowding Conditions exhibited the worst results. Additionally, those who are more dependent on public transport, thus using the system more times a day, more times a week or for more purposes, were found to display worst overall satisfaction results, which can be attributed to reduced satisfaction with Safety and Security, Transit Connectivity, Fare, Bus Stop Infrastructure, Accessibility, and Travel Time. On the same note, Negative Social Safety Experiences and facing overcrowding conditions often were also found to have a negative effect on overall satisfaction, while using the system due to traf#ic conditions was found to have positive in#luence on it. Therefore, creating an environment that reduces negative experiences are paramount to creating positive attitudes and, thus, retaining and increasing ridership.
On the same note, the analysis of relative importance revealed that respondents were likely to place more importance on the attributes that they perceived less satisfaction. Thus, highlighting the importance of creating an environment with reduced negative experiences as they are likely to become salient. For example, among those more dependent on public transport, Transit Connectivity, Crowding Conditions, Fleet Conservation, and Accessibility were shown to have more importance than for those less dependent on this travel mode. Overall, Safety and Security, Fare, Fare Connectivity, Fare Integration, Reliability, Service Frequency, and Accessibility were found to be the most important.
Finally, by comparing satisfaction and relative importance data through GAP analysis we yielded interesting results, such as evidencing a smaller number of attributes that are more likely to improve user intention towards public transport given their current levels of satisfaction and importance. Fare and Safety and Security are found to be the most critical attributes, suggesting that more effort should be put into their improvement as to retain and prospect more users. Additionally, Comfort and Accessibility attributes should be improved, once they have a strong in#luence on satisfaction as shown by the Model Regression Analysis. Consequently, public policies targeting reducing fares and overcrowding conditions and improving safety and security and the bus #leet infrastructure are necessary.
Moreover, a mindset shift in planning for urban infrastructure is needed as to endorse more sustainable development. The literature provides a range of acknowledged options towards decrease car ownership and increase the use of public transportation. However, the success of any measure depends on understanding the trends, the context and the expectations the population, in question has for a better quality of life.