Using the Three-Factor Method to Identify Improvement Priorities for Express and Local Bus Services in the Twin Cities

Transcription

1 Using the Three-Factor Method to Identify Improvement Priorities for Express and Local Bus Services in the Twin Cities Xinyi Wu and Jason Cao Humphrey School of Public Affairs, University of Minnesota th Ave. S, Minneapolis, MN, Abstract This paper employs the three-factor theory to identify how express and local bus riders perceive the same service attributes differently, and which service attributes have a critical impact on riders overall satisfaction. It then provides transit agencies recommendations for service improvements. Specifically, we apply ordered logistic models to the 2014 Transit Rider Survey in the Twin Cities to examine how the performance of service attributes affects riders overall satisfaction with express and local buses. By adopting the three-factor method, we classify the attributes into three groups based on their respective contribution to overall satisfaction. We then integrate the classification of the attributes with their average performances to identify improvement priorities. Both the importance of service attributes to overall satisfaction and the improvement priorities differ between express and local buses. Among the tested attributes, vehicles are comfortable, total travel time is reasonable, and reliability should be addressed first for both local and express buses. Key Words: customer satisfaction, importance-performance analysis, quality of service, Kano model, transit Number of words: 5,988 1

2 Introduction Customer satisfaction serves as an important indicator of quality of service in many industries (1). In the realm of public transportation, riders satisfaction can inform transit agencies their service qualities. Identifying the top influential attributes affecting riders overall satisfaction provides targeted strategies to improve the quality of transit service (2). The method of importance-performance analysis (IPA, or called quadrant analysis) has been widely used in examining customer satisfaction surveys (CSS), because of its simplicity and effectiveness. Many scholars and transit agencies have applied the IPA to identify improvement priorities by plotting the importance and performance of transit service attributes in a quadrant. However, the IPA has its limitations. A prominent one is that it is built upon the assumption that the relationship between the performance of attributes and their influence on overall satisfaction with transit is linear and symmetric, which is often not the case. This limitation can be effectively addressed by the three-factor theory we employed in this study. This theory takes the non-linear and asymmetric relationships into account, and can provide more accurate policy implications for transit service improvements. Previous studies in other fields have successfully illustrated the feasibility of using the threefactor theory to determine the relative importance of different attributes from users perspective (3, 4). In this study, we apply the approach to classify the attributes of transit services based upon their relative importance to riders. Using the 2014 Transit Rider Survey in the Minneapolis-St. Paul (Twin Cities) metropolitan area, this study aims to answer the following research questions: (1) Do transit service attributes have varying (linear vs. nonlinear) influences on overall satisfaction? (2) How does the classification of the attributes differ by local and express buses? (3) Do express and local buses have different improvement priorities? The paper is organized as follows. In section 2, we review previous studies related to customer satisfaction of public transit; in section 3, we introduce the data and the method; in section 4, we present our modelling results; and in section 5, we summarize the results. 2. Literature Review 2.1 Satisfaction analysis methodologies As an essential component of the modern transportation system, public transit plays an important role in reducing auto-dependence and traffic congestion (5). Ridership is a key factor that determines the success of a transit system (6). Because it is affected by riders satisfaction with transit service (7), identifying the top influential attributes of riders satisfaction would provide transit agencies targeted strategies to improve the quality of transit services (2). Previous studies have employed a variety of methods to investigate the satisfaction of transit riders using customer satisfaction surveys. de Oña & de Oña (8) classified these methods into two types based on how the service quality was measured: disaggregated models, which measure service attributes individually, and aggregated models, which aggregate individual attributes to 2

3 construct an overall service quality index (SQI) or a customer satisfaction index (CSI). Among the disaggregated methods, IPA has been widely used to examine transit riders satisfaction. Martilla and James (9) firstly devised this method. They grouped service attributes into four quadrants (A, B, C, and D) based on the performance and importance of each attribute (Figure 1). The classification of service attributes in the grid further determines their hierarchy of improvement priorities. If an attribute has high importance but low performance, it should be prioritized for improvement (Quadrant A). If an attribute has high importance and high performance, it is classified into Quadrant B and service providers should keep up the good work. If an attribute has low importance and low performance, it has a low priority (Quadrant C). Finally, if an attribute has low importance but high performance, it could be a possible overkill (Quadrant D). Using this technique, many studies have analyzed transit service attributes by cross-tabulating the importance and performance of each service attribute (5, 10, 11, 12). Figure 1 Importance-Performance Grid (IPA matrix) The key of IPA is to group all service attributes by their performance and importance. Since the performance is usually measured directly in the CSS, the choice of importance becomes a distinction between different types of IPA studies. Some researchers use derived importance, (i.e. estimated importance based on the coefficients of statistical models), while others used stated importance (i.e. respondents explicit statement on the importance of each attribute) (10). Here we present a few example studies. Some studies chose to estimate the implicit importance by either bivariate or multivariate regressions. Weinstein employed bivariate correlation analysis to calculate the derived importance of each attribute (10). This study used the 1998 CSS data of San Francisco Bay Area Rapid Transit (BART), in which respondents were asked to report the performance of various service attributes on a seven-point scale. Weinstein correlated the performance of each of these attributes with the overall satisfaction and used correlation coefficients to rank the importance of the attributes. Then, Weinstein plotted the importance and performance in an IPA matrix and 3

4 identified the four types of attributes. The targeted attributes for BART s future development include reliability, seats availability, ticket refund and escalators, etc. Weinstein concluded that bivariate correlation is a statistically valid method to rank the importance of different transit attributes. However, this method does not consider the interactions among different service attributes. Figler et al. (11) modified the method used to estimate the importance of each transit attribute. They used the data from a 2008 survey conducted by the Chicago Transit Authority (CTA). Figler et al. introduced an importance index, which was defined as the correlation of attribute with bus loyalty divided by the median correlation of attribute with bus loyalty. A higher importance index represents a greater correlation between attribute performance and bus loyalty. The results of quadrant analysis and regression model showed that two service attribute factor variables ( experience index and perception index ), four trip characteristic variables ( problem experiences, choice rider frequent bus rider and frequent train rider ), and one individual characteristic variable ( age ) have significant impacts on bus loyalty. Among the significant variables, however, age and frequent train rider have counter-intuitive correlations with bus loyalty. Based on the quadrant analysis, this paper identified improvement priorities, including providing services at a reasonable price, reliable services, and responding effectively to problems and issues, etc. Shen et al. (5) built an IPA matrix using a more comprehensive modelling process, with the indicators of performance being satisfaction index, and the indicators of importance being path coefficients of structural equation models (SEM). This study used data from the 2013 passenger survey along Line one of the Suzhou rail transit system in China. It adopted a comprehensive methodology including passenger satisfaction index model, evaluation indicator system and an assisting IPA matrix. Shen et al. proposed a conceptual model as the foundation, and identified six latent variables that are crucial to satisfaction studies. The paper provided detailed improvement implications for transit operation companies based on the IPA matrix. Among the attributes that need to be addressed, equipment and facilities, and information distribution disclosure rank the highest for their relatively higher importance and lower performance. Shen et al. s comprehensive study approach allows more in-depth research of the service attributes as well as their contributions to customer satisfaction, and thus providing more detailed policy implications to transit agencies. Some rider satisfaction surveys asked participants to indicate the importance of transit service attributes. That is, they measured stated importance of service attributes. Stradling et al. (12) summarized the results of previous studies and introduced the concept of disgruntlement into the literature. A respondent would be considered disgruntled with an attribute if he/she considers the attribute important but poorly performed. Then, they used the disgruntlement and importance instead of the performance and importance to construct an IPA matrix. The attributes with high disgruntlement and high importance should be the focus of improvements. Stradling et al. showed the plausibility of using this new concept of disgruntlement to study customer satisfaction, which also provided a new indicator for researchers to compare service qualities across locations, modes, time, and demographics. 4

5 Nathanail (13) also used stated importance. This study used the data of a 2004 national survey conducted along the Hellenic Railway in Greece. Participants indicated the significance and performance of transit service attributes separately on a 10-level scale in the survey. To avoid biased evaluation by less experienced riders, some attributes, such as seat comfort and safety during trip, were measured by railway operators using the condition and amount of facilities provided. This measurement of performance considers both observed and perceived qualities and enables the comparison of service quality across regions and periods. Although these IPA studies have shed lights on the influential service attributes and improvement priorities, they suffer from a few limitations (14). Among them, the IPA assumes that the impact of attribute performance on overall satisfaction is linear. However, the performance of some service attributes often has non-linear and asymmetric relationships with overall satisfaction (15, 16). That is, the impact varies as their performance changes. Many studies have applied alternatives of the Kano Model to provide more accurate policy implications (17, 18). These applications have gradually developed into a factor theory, which classifies service attributes into three groups, as shown in Figure 2 (5): Basic Factors: these attributes have significant impacts on overall satisfaction when they perform poorly, or reduce riders satisfaction if they are not well delivered. However, when these attributes perform well, they do not have significant impacts on overall satisfaction; Performance Factors: these attributes have significant impacts when they perform both well and poorly, or have a linear and symmetric relationship with riders satisfaction; Exciting Factors: these attributes have significant impacts when they perform well; or increase riders satisfaction if they are well delivered. But when these attributes do not perform well, they do not have significant impacts on overall satisfaction. Figure 2 The Three Factor Theory of Satisfaction

6 The three-factor theory can improve the existing IPA approach by 1) addressing the non-linear impact of attribute performance on overall satisfaction, and 2) providing a more accurate policy implications for service improvements. Studies regarding satisfaction in other fields of services have successfully employed the three-factor theory to determine the relative importance of different attributes from users perspective (3, 18). However, the application of the three-factor theory in transit industry is rather limited; as far as we know, only one study employs the theory to examine transit services in Indore, India (19). This paper aims to fill this research gap. It differs from the Indore study in that the latter uses the importance grid to derive the three factors whereas this study applies an alternative method: regressions with dummy variables. The importance grid cross-tabulates implicit and explicit importance to identify different factors (20). This method requires the acquisition of both stated importance and derived importance of each attribute. In the field of transit, Zhang et al. used this approach to study riders' satisfaction with bus, bus rapid transit and van services in Indore, India (19). When stated importance is not available, researchers could adopt regression with dummy variables. This method converts each attribute into two dummy variables (high performance and low performance) and regresses overall satisfaction against these dummy variables. Then, each factor was identified by examining significance level of the coefficients of both dummy variables (18). Because the stated importance was not measured in the survey, this paper uses regression with dummy variables to classify attributes. 2.2 Transit service differentiation Most transit satisfaction studies focus on one type of transit services such as regular bus, light rail, and bus rapid transit (5, 11). Nevertheless, several comparative or qualitative studies have discovered that the importance of service attributes differs from one kind of riders to another. Andreassen found that transit riders with different riding frequencies have different perceptions. This study indicated that train riders have greater concerns over travel time and price level, while bus riders concern the design of stations and platforms more than other attributes (21). Beirão and Cabral suggested that segmentation regarding riders characteristics is necessary in the research of travel behavior and attitudes (22). They found that many factors could have impacts on riders travel behavior. For example, travel purposes can highly affect the importance of different attributes: travel time and reliability are considered important in work- and schoolrelated trips, while available seats, smooth rides, and enough space are considered important in recreational trips. They also showed that car owners and less frequent riders perceive transit service quality lower than frequent riders. Further, Iseki and Smart (23) concluded that lowincome riders care less about daytime riding safety than high-income riders. These studies suggest that it is necessary to differentiate and segment different riders from a policy perspective. Furthermore, Garrett and Taylor suggested that local buses mainly serve low-income and captive riders, while express buses serve high-income and choice riders (24). Based on different characteristics of express and local bus riders and the potential different results they might lead to, we can provide more accurate policy implications by differentiating these two types of riders in our analysis. 6

7 Methodology 3.1 Data and variables This study used the data from the 2014 Metro Transit Rider Survey, which was administered by Metro Transit to users of bus, light rail and Northstar commuter rail services in the Minneapolis- St. Paul (Twin Cities) metropolitan area. Both online and written versions of the surveys were available to respondents. Surveys were distributed on Wednesday, November 5, Thursday, November 6 and Sunday, November 9, and then were collected by November 30, This study focused on riders of express and local buses. The sample includes 5,297 valid bus surveys (25). Metro transit are interested in 24 attributes of bus services (see Table 2 later). To alleviate the burden of participants, Metro Transit divided the 24 attributes into two groups and put them in two versions of surveys (bus survey A and B), respectively. Survey A and B were randomly distributed to bus users. In the surveys, respondents assessed the performance of these service attributes and the overall rating of Metro Transit service on a five-point ordinal scale: Excellent (5), Good (4), Fair (3), Poor (2), and Unacceptable (1). 3.2 Modeling Approach This study used the three-factor theory to identify the service attributes that have significant impacts on the overall satisfaction of express and local bus riders. Regression with dummy variables was adopted to classify the 24 service attributes by examining the impacts of the attributes on the overall satisfaction when they performed well and poorly. In particular, for each of the attributes, we recoded its performance into two dummy variables indicating highperformance and low-performance, respectively. Since survey participants gave high scores to most service attributes, we chose to recode excellent as 1 for the high-performance dummy variables, and recode fair, poor and unacceptable as 1 for the low-performance dummy variables (Table 1). Further, for each attribute, all the missing values (including Don t know/don t use ) were recoded 0 to minimize the impact of missing values on modelling results. We created a dummy variable to indicate whether 0 is a missing value or the average (good) performance. For each version of the surveys, this recoding practice generated 24 dummy variables indicating high performance or low performance of service attributes and 12 dummy variables indicating missing values of the attributes. Then, we developed two ordered logistic models to regress the overall satisfaction against the 36 dummy variables for express and local bus users, respectively. In these two models, the variables of interest are high-performance and low-performance dummy variables. If the low-performance dummy of an attribute has a significant coefficient while its high-performance dummy does not, it is a basic factor. If the high-performance dummy of an attribute has a significant coefficient while its low-performance dummy does not, it is an exciting factor. If both dummy variables of an attribute have significant coefficients, it is a key performance factor. If neither of the dummy variables of an attribute are significant, it is an unimportant performance factor. In this study, we use the p-value of 0.05 as the critical significance level. 7

8 Table 1 The coding methodology of independent variables Categories High Performance Dummy Variable Low Performance Dummy Variable Average Performances Excellent Good Fair Poor Unacceptable Don t know/don t use After classifying attributes, we computed and ranked the average performance of each attribute. Using both the classification and rankings, we identified bus service attributes critical to overall satisfaction. 4. Results and Discussion 4.1 Factor structure Table 2 presents the factor structure of the attributes and the ranking of their average performances. The factor structure is derived from the two ordered logistic models (see Appendix). There are two accessibility attributes in the surveys. Since we did not see any significant differences between these two attributes, we eliminated the statistically insignificant one from survey A. We also eliminated an insignificant attribute, vehicles are environmentally friendly, from survey B due to multicollinearity. As shown in table 2, six of the 22 attributes are classified consistently between express and local bus riders, while sixteen attributes are grouped differently. The possible reasons for the difference are explained below. 8

9 Table 2 Factor structures and attribute rankings Attributes Local Bus Ranking Express Bus Ranking Accessibility Basic 7 Exciting 11 Vehicles are comfortable Basic 12 Basic 16 Driver s manner Exciting 3 Basic 2 Information at bus stops Exciting 19 Basic 19 Drivers calling out street names Exciting 21 Basic 22 Value for the fare paid Exciting 8 Performance 8 Courteous drivers Performance 9 Basic 6 Route going where people need to go Performance 11 Basic 15 Hours of operation for transit service meet my needs Performance 14 Basic 20 Transferring is easy Performance 6 Exciting 9 Personal safety while waiting Performance 16 Exciting 13 Total travel time is reasonable Performance 15 Performance 14 Reliability Performance 20 Performance 18 Safety while riding Performance 13 Performance 5 Paying fare is easy Performance 1 Unimportant 1 Shelter condition/cleanness Performance 22 Unimportant 21 Vehicles are clean Unimportant 18 Basic 12 Fares are easy to understand Unimportant 2 Exciting 4 Easy to identify the right bus Unimportant 4 Exciting 3 Availability of seats Unimportant 17 Exciting 17 Routes and schedules are easy to understand Unimportant 5 Unimportant 7 Availability of the route map and schedule Unimportant 10 Unimportant 10 1) Accessibility is a basic factor among local bus riders but an exciting factor among express bus riders. Local bus riders usually take transit more often and are more likely to depend on transit for their daily travel than express bus users, so the former tend to have a higher demand for accessibility than the latter. 2) Driver s manner, value for the fare paid, drivers calling out street names and information at bus stops are all considered exciting factors among local bus riders, but are basic or performance factors among express bus riders. A higher proportion of local bus riders are captive riders. As a result, they are more dependent on bus services and are less affected by drivers manner and the value for the fare paid than express bus riders. On the other hand, express bus riders are mostly choice riders, so their demand for drivers manner and value for the fare paid need to be delivered properly for them to continue using the services. Likewise, local bus riders use bus services more frequently and thus are more familiar with the routes and schedules, so they have a lower demand for drivers calling out street names and information at bus stops than express 9

10 bus riders. Because express bus users ride bus less frequently, they need more information of the services due to their unfamiliarity with the routes. 3) Paying fare is easy and shelter condition/cleanness are considered performance factors among local bus riders but unimportant factors among express bus riders. Most express bus riders use smart cards to pay the fare, which makes the convenience of paying fare less important to express bus riders. In addition, express buses usually have well-designed and developed shelters along the routes, while many local bus stops are in demand for basic facilities. 4) Routes go where people need to go, hours of operation for transit meet my needs, and courteous drivers are considered performance factors among local bus riders, but are basic factors among express bus riders. The difference indicates that, although both local and express bus riders consider these attributes important, they tend to have few impacts on express bus riders when their performance meets certain requirements, likely because their demand for bus services is limited. 5) Transferring is easy and personal safety while waiting are considered performance factors among local bus riders but exciting factors among express bus riders. Few express bus riders need to transfer, and they often wait for buses in the shelters with many amenities including safety and security measures. So they view these two attributes less important. 6) Vehicles are clean is considered unimportant among local bus riders but basic factors among express bus riders. Because the latter are mostly choice riders, they are less likely to tolerate dirty vehicles than local bus riders. 7) Availability of seats, fares are easy to understand, easy to identify the right bus are considered unimportant among local bus riders but exciting factors among express bus riders. In most cases, local bus riders are dependent on bus services, which decreases their demands for seats and fares. The number of express buses is usually much less than local buses, so identifying the right bus has greater importance for express bus riders. 4.2 Improvement priorities Based on the factor structure and the average performance ranking of bus service attributes, we can identify improvement priorities. We considered the attributes that rank from 12 to 22 as poorly-performed attributes. Basic factors that perform poorly have the largest negative impacts, and thus should be the top priorities. Performance factors also have significant negative effects when they do not perform well, which makes them the second priorities. Finally, after transit agencies have fulfilled poorly-performed basic and performance factors, poorly-performed exciting factors could be the third priorities if there are available resources (26). As shown in Table 3, improvement priorities differ between local and express buses. Because of the higher demand of express bus riders, express buses have five more basic factors than local buses. Among the poorly-performed basic factors, vehicles are comfortable is a priority for both local and express buses. As for the second priorities, local buses have four more performance factors than express buses. Both express and local buses perform poorly in terms of total travel time is reasonable and reliability. Even we consider the first and second priorities jointly, local and express buses seem to have quite different improvement priorities. Local bus riders care more about safety and shelter conditions whereas express bus riders 10

11 concern more on riding environment, and information and spatial availability of bus services. The poorly-performed exciting factors are also different between local and express buses. Table 3 Improvement Priorities Local Bus Express Bus Top Priorities Basic factors Vehicles are clean Route going where people need to go Vehicles are comfortable Vehicles are comfortable Information at bus stops Hours of operation for transit service meet my needs Drivers calling out street names Second Priorities Performance factors Safety while riding Hours of operation for transit service meet my needs Total travel time is reasonable Personal safety while waiting Reliability Shelter condition/cleanness Total travel time is reasonable Reliability Third Priorities Exciting factors Information at bus stops Personal safety while waiting Drivers calling out street names Availability of seats Note: the underlying attributes are common for local and express buses. 5. Conclusions This paper explores how riders satisfaction with transit is affected by different attributes of transit services, and how express and local buses differ in service improvement priorities. The three-factor approach regards basic factors and important performance factors, especially those that do not perform well, critical and thus they should be priorities of transit agencies strategic development. By adopting the three-factor theory and ranking average performance, we are able to classify service attributes into different categories, and assess the attributes against their performance. This practice provides more fine-tuned policy implications for transit research and development than traditional importance-performance analysis. Income, ethnicity, car-ownership, travel purposes, and other characteristics of riders might moderate how they perceive transit services. Although this study analyzed local and express bus riders separately, it does not differentiate the riders of different characteristics. Introducing these 11

12 variables into transit satisfaction studies could be a direction for future research if the interest is in enhancing transit service satisfaction for different rider segments. Further, since most respondents in this sample chose to give high scores to service attributes, the percentages of low performance ratings were extremely small. The regression results change greatly when using different coding thresholds to generate dummy variables. Future studies should examine the distribution of the performance of service attributes since the results may be sensitive to the small percentages. Further, the three-factor theory, as well as the IPA, does not consider the costs associated with improvement priorities. When budget is constrained, transit agencies should consider cost-effective improvement priorities. Nevertheless, this study shows the feasibility of using the three-factor theory to classify different attributes of transit services. It also demonstrates how riders of different transit services perceive service attributes differently. This attempt could help transit agencies to design more specific development plans. The results show that local and express buses have different improvement priorities. For local buses, vehicles are comfortable is the top priority. For express buses, many attributes are considered as top priorities, including vehicle cleanness, routes, operating hours, information at bus stops, vehicle comfort, and drivers calling out street names. As for the second priorities, express bus riders concern more about travel time and bus reliability. While local bus riders pay extra attention to travelling safety, waiting safety, and shelter conditions. Finally, if the first and second priorities have already been addressed, transit agencies could distribute resources to the third priorities, which include information at bus stop and drivers calling out street names for local buses, and personal safety while waiting and availability of seats for express buses. Among the improvement priorities, vehicles are comfortable, total travel time is reasonable, and reliability should be the focus for both local and express buses. 12

13 References 1. Aniley, B., & Negi, R. Service quality attributes affecting passengers satisfaction with HIGER city bus. Ethiopian Journal of Business and Economics (The), 2015, Vol.1, No. 2, pp Cao, J., Cao, X., Zhang, C., & Huang, X. The gaps in satisfaction with transit services among BRT, metro, and bus riders: Evidence from Guangzhou. Journal of Transport and Land Use, Füller, J., Matzler, K. Customer delight and market segmentation: An application of the three-factor theory of customer satisfaction on life style groups. Tourism management, 2008, Vol. 29, No. 1, pp Busacca, B., Padula, G. Understanding the relationship between attribute performance and overall satisfaction: Theory, measurement and implications. Marketing Intelligence & Planning, 2005, Vol. 23, No. 6, pp Shen, W., Xiao, W., & Wang, X. Passenger satisfaction evaluation model for Urban rail transit: A structural equation modeling based on partial least squares. Transport Policy, 2016, pp de Oña, J., de Oña, R., Eboli, L., & Mazzulla, G. Perceived service quality in bus transit service: a structural equation approach. Transport Policy, 2013, pp Lai, W. T., & Chen, C. F. Behavioral intentions of public transit passengers the roles of service quality, perceived value, satisfaction and involvement. Transport Policy, 2011, Vol. 18, No.2, pp de Oña, J., & de Oña, R. Quality of service in public transport based on customer satisfaction surveys: A review and assessment of methodological approaches. Transportation Science, 2014, Vol. 49, No.3, pp Martilla, J. A., & James, J. C. Importance-performance analysis. The journal of marketing, 1977, pp Weinstein, A. Customer satisfaction among transit riders: how customers rank the relative importance of various service attributes. Transportation Research Record: Journal of the Transportation Research Board, 2000, No. 1735, pp Figler, S., Sriraj, P., Welch, E., & Yavuz, N. Customer Loyalty and Chicago, Illinois, Transit Authority Buses: Results from 2008 Customer Satisfaction Survey. Transportation Research Record: Journal of the Transportation Research Board, 2011, No. 2216, pp Stradling, S. G., Anable, J., & Carreno, M. Performance, importance and user disgruntlement: A six-step method for measuring satisfaction with travel modes. Transportation Research Part A: Policy and Practice, 2007, Vol. 41, No. 1, pp Nathanail, E. Measuring the quality of service for passengers on the Hellenic railways. Transportation Research Part A: Policy and Practice, 2008, Vol. 42, No. 1, pp

14 Arbore, A., & Busacca, B. Rejuvenating importance-performance analysis. Journal of Service Management, 2011, Vol. 22, No. 3, pp Kano, N., Seraku, N., Takahashi, F., Tsuji S. Attractive quality and must-be quality. Journal of the Japanese Society for Quality Control, 1984, Vol.14, No. 2, pp Berger, C., Blauth, R., Boger, D., Bolster, C., Burchill, G., DuMouchel, W., & Timko, M. Kano s methods for understanding customer-defined quality. Center for Quality Management Journal, 1993, Vol. 2, No. 4, pp Anderson, E. W., & Mittal, V. Strengthening the satisfaction-profit chain. Journal of Service research, 2000, Vol. 3, No. 2, pp Busacca, B., Padula, G. Understanding the relationship between attribute performance and overall satisfaction: Theory, measurement and implications. Marketing Intelligence & Planning, 2005, Vol. 23, No. 6, pp Zhang, C., Cao, J., Nagpure, A.S. Applying Three-Factor Approach to Customer Satisfaction of Transit Services in Indore, India. Presented at the 56th ACSP Conference, Houston, TX, Matzler, K., & Sauerwein, E. The factor structure of customer satisfaction: An empirical test of the importance grid and the penalty-reward-contrast analysis. International Journal of service industry Management, 2002, Vol. 13, No. 4, pp Wallin Andreassen, T. (Dis) satisfaction with public services: the case of public transportation. Journal of Services Marketing, 1995, Vol. 9, No. 5, pp Beirão, G., Cabral, J. S. Understanding attitudes towards public transport and private car: A qualitative study. Transport policy, 2007, Vol. 14, No. 6, pp Iseki, H., & Smart, M. How Do People Perceive Service Attributes at Transit Facilities? Examination of Perceptions of Transit Service by Transit User Demographics and Trip Characteristics. Transportation Research Record: Journal of the Transportation Research Board, 2012, No.2274, pp Garrett, M., Taylor, B. Reconsidering social equity in public transit. Berkeley Planning Journal, 1999, Vol. 13, No Metro Transit. Metro Transit Rider Survey, Matzler, K., Sauerwein, E., & Heischmidt, K. Importance-performance analysis revisited: the role of the factor structure of customer satisfaction. The Service Industries Journal, 2003, Vol. 23, No. 2, pp

15 Appendix Survey A regression results Local Buses Express Buses Coefficients P-value Coefficients P-value Routes go where people need to go (Low) Routes go where people need to go (High) Paying fare is easy (Low) Paying fare is easy (High) Safety while riding (Low) Safety while riding (High) Driver s manner (Low) Driver s manner (High) Vehicles are comfortable (Low) Vehicles are comfortable (High) Value for the fare paid (Low) Value for the fare paid (High) Availability of seats (Low) Availability of seats (High) Easy to identify the right bus (Low) Easy to identify the right bus (High) Drivers calling out street names (Low) Drivers calling out street names (High) Transferring is easy (Low) Transferring is easy (High) Shelter condition/cleanness (Low) Shelter condition/cleanness (High) Shelter condition/cleanness (Missing) Transferring is easy (Missing) Routes go where people need to go (Missing) Paying fare is easy (Missing) Safety while riding (Missing) Driver s manner (Missing) Vehicles are comfortable (Missing) Value for the fare paid (Missing) Availability of seats (Missing) Easy to identify the right bus (Missing) Drivers calling out street names (Missing) N = 1,497; R 2 = 32.74% N = 1,225; R 2 = 35.28%

16 457 Survey B regression results Local Buses Express Buses Coefficients P - value Coefficients P - value Personal safety while waiting (Low) Personal safety while waiting (High) Hours of operation for transit service meet my needs (Low) Hours of operation for transit service meet my needs (High) Total travel time is reasonable (Low) Total travel time is reasonable (High) Reliability (Low) Reliability (High) Vehicles are clean (Low) Vehicles are clean (High) Routes and schedules are easy to understand (Low) Routes and schedules are easy to understand (High) Fares are easy to understand (Low) Fares are easy to understand (High) Information at bus stops (Low) Information at bus stops (high) Availability of the route map and schedule (Low) Availability of the route map and schedule (High) Courteous drivers (Low) Courteous drivers (High) Accessibility (Low) Accessbility (High) Personal safety while waiting (Missing) Hours of operation for transit service meet my needs (Missing) Total travel time is reasonable (Missing) Reliability (Missing) Vehicles are clean (Missing) Routes and schedules are easy to understand (Missing) Fares are easy to understand (Missing) Information at bus stops (Missing) Availability of the route map and schedule (Missing) Courteous drivers (Missing) Accessbility (Missing) N = 1,391; R 2 = 35.34% N = 1,184; R 2 = 42.02% 16