- Email: [email protected]
A global index for level of service evaluation at airport passenger terminals
Available online at www.sciencedirect.com
Transportation Research Part E 44 (2008) 607–620 www.elsevier.com/locate/tre
A global index for level of service evaluation at airport passenger terminals Anderson Ribeiro Correia a, S.C. Wirasinghe b, Alexandre G. de Barros a
b,*
Aeronautical Institute of Technology, Prac¸a Mal. Eduardo Gomes, 50 Sa˜o Jose´ dos Campos, SP 12.228-901, Brazil b University of Calgary, 2500 University Drive NW, Calgary, Alberta, Canada
Abstract This paper presents a global index for the evaluation of the level of service (LOS) of the operational components at an airport. This index is useful in evaluating the overall LOS on a single scale and according to user perceptions. It is assumed that the overall LOS is a function of the LOS of individual components, such as check-in, departure lounges, etc., as well as socio-economic variables. Regression analysis is used to obtain a mathematical relationship between the global LOS ratings and the LOS of individual operational components. The methodology is illustrated with its application at Sa˜o Paulo/ Guarulhos international airport in Brazil, where 119 enplaning passengers were interviewed. 2007 Elsevier Ltd. All rights reserved. Keywords: Level of service; Airports; Passenger terminal building; Regression analysis
1. Introduction The development of level of service (LOS) measures for airport passenger terminals has been one of the major issues for airport operators in the last several decades. This has motivated a number of LOS studies by air transportation agencies, including the Federal Aviation Administration – FAA (Transportation Research Board, 1987; Airports Council International, 2000; Transport Canada, 1979). Despite the efforts of these agencies, the proposed LOS standards and methods have been the subject of criticism by airport professionals. One of the main concerns is the lack of passenger input. Several studies have also been undertaken to develop methods for LOS evaluation taking into account user perceptions. Most of them have provided results based on poor database, and were not able to provide a high level of significance for testing the hypotheses considered. Additionally, most studies focused on individual components of the airport passenger terminal (check-in counter, departure lounge, etc.), neglecting the overall evaluation. A broad measure reflecting the LOS of the terminal as a whole for a given type of passenger (e.g., departing) would be useful at planning, design and management levels. With this measure, it would be possible to
*
Corresponding author. Tel.: +1 403 220 6713. E-mail address: [email protected] (A.G. de Barros).
1366-5545/$ - see front matter 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.tre.2007.05.009
608
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
identify the level of importance attributed to individual components by passengers, in order to prioritize investments. Additionally, it would provide a measure for comparison between different airports for the purpose of benchmarking. The main challenge when developing an overall measure is the data collection. It is relatively simple to collect characteristics of individual facilities (e.g., waiting time at the check-in counter) as opposed to obtaining overall measures (e.g., total walking distance by an individual). Several issues must be addressed before a research effort is developed to collect overall measures. It is the purpose of this paper to provide a methodology for such an effort, illustrating it with a case study of Sa˜o Paulo/Guarulhos international airport. A complete review of past research on LOS has been presented in Correia and Wirasinghe (2004). Mumayiz and Ashford (1986) proposed a perception–response concept, using graphical displays constructed from passenger responses concerning the LOS provided at airports in England. Omer and Khan (1988) employed the concept of utility theory to develop a relationship between characteristics of facilities (e.g., waiting time, space available) and user responses (0–1) about the LOS offered. Mu¨ller and Gosling (1991) applied a psychometric scaling technique to obtain a quantitative measure of LOS that might be used in a relationship similar to the one developed by Omer and Khan (1988). Seneviratne and Martel (1991) developed LOS standards for several components of the airport passenger terminal. The selection of the most important components and measures was based on a survey of Canadian airports (Martel and Seneviratne, 1999). Ndoh and Ashford (1993) employed theories of perception and scaling to evaluate LOS on airport access, using 12 attributes (e.g., cost, comfort, access to information, etc.). Park (1994) used fuzzy logic to derive LOS measures for specific components of the airport passenger terminal. The methodology was applied to the Seoul Kimpo Airport. Yen (1995) conducted a survey at Austin Municipal Airport in Texas, USA. He applied binary logit models to estimate a ‘‘long’’ model and a ‘‘short’’ model to predict the probabilities that a passenger will rate a service on the basis of perceived time measures. Yen et al. (2001) presented a quantitative model to define the level of service at airport passenger terminals. The model used the fuzzy concept to relate subjective service ratings to time measurements of associated waiting or service processes. Fernandes and Pacheco (2002) utilized data envelopment analysis to evaluate the capacity of 35 Brazilian airports, based on several operational parameters (e.g., number of check-in counters, average space available per passenger, etc.). Magri and Alves (2003) evaluated the LOS offered by six Brazilian airports as a function of 36 subjective parameters suggested by ACI (Airports Council International, 2000). All these studies concentrated on the evaluation of individual components. No study has developed an objective overall LOS measure, reflecting the LOS provided by the airport passenger terminal on a single scale. 2. Airport passenger terminal facilities and characteristics The terminal area is the major interface between the airfield and the rest of the airport. It includes the facilities for passenger and baggage processing, cargo handling, airport maintenance, operations and administration activities. The passenger terminal system has three major functional areas. These functional areas and the activities that occur within them are as follows (Horonjeff and McKelvey, 1994): • The access interface – where the passenger transfers from the access mode of travel to the passenger processing functional area. Circulation, parking, and curbside loading and unloading of passengers are the activities that take place within this functional area. • Processing – where the passenger is processed in preparation for starting, ending, or continuation of an air transportation trip. The primary activities in this functional area are ticketing, baggage check-in, baggage claim, seat assignment, federal inspection, services and security. • The flight interface – where the passenger transfers from the processing functional area to the aircraft. The activities that occur here include assembly, conveyance to and from the aircraft, and aircraft loading and unloading.
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
609
In order to provide a global index for LOS evaluation, it is necessary to specify the type of movement in question. Airport passengers can be divided into three groups according to their movement type: departing, arriving, and transfers. Each of these groups will have a different set of needs and wishes and, in many cases, will even make use of different facilities. For example, departing passengers will not make use of the baggage claim facilities, whereas arriving passengers will not use the enplaning curbside or the check-in lobby. Therefore, each movement type will have a LOS index which is global in the sense that it encompasses the passenger’s full airport experience as opposed to just rating one single facility or service. In this research, only departing domestic passengers are analyzed; however, the methodology presented herein can also be applied to evaluate the LOS for arriving and transfer passengers (de Barros et al., 2006). Additionally, this research concentrates on operational components of the airport passenger terminal: for this reason, parking, public transit or other components located outside of the terminal building are not addressed. In summary, this paper presents an overall LOS evaluation as a function of the following components: • • • • • •
Enplaning curbside. Ticket counter and baggage deposit. Security screening. Departure lounge. Circulation areas (corridors, stairs, elevators, etc.). Concessions.
Four overall LOS measures are included in the analysis, as they have been identified as very important for passengers (Correia et al., 2005): walking distance, orientation, total time and security environment. The last variable differs significantly from security screening. Security environment refers to the users’ perceptions of security throughout the terminal building. This is a subjective variable, which may influence the overall LOS to a certain extent. On the other hand, security screening refers to the quality of the experience of the passenger when being processed at the security screening location. Although some components are not directly managed by the airport operator, they may have considerable influence during the planning and management stages. For instance, check-in counters are usually managed by airlines, but are often planned and built by the airport operator. Moreover, in many countries the airport operator is responsible for assigning check-in areas according to demand priorities. The following are brief descriptions of the components and measures that are analyzed. 2.1. Enplaning curbside The curbside element is the interface between the terminal building and the ground transportation system. High traffic volumes and peaks, plus the complex flows of mixing people and vehicles, may result in extensive traffic congestion at the curbside area. This in turn may cause inconvenience, frustration, and delay to passengers at large airports. Arbitrary standards applied in many airports may lead to oversized or undersized facilities. The airport community is, therefore, interested in a methodology that could lead to rational standards (Siddiqui, 1994). 2.2. Ticket counter and baggage deposit The airline ticket counter is where the airline and passenger make final ticket transactions and check-in baggage for a flight. The ticket transaction takes place at the ticket counter, which is a stand-up desk. To the left and right of the ticket counter position, a low shelf is provided to deposit, check-in, tag and weigh baggage, if necessary. Subsequently, the baggage is passed back by the agent to an outbound baggage conveyance device located near the counter. The check-in counters located at Sa˜o Paulo/Guarulhos international airport operate with a single line for each individual airline. Self check-in and ticketing kiosks are in use at several airports worldwide. However, these devices are not yet used at Sa˜o Paulo/Guarulhos international airport. For this reason, the evaluation of LOS for these components is not addressed in this research.
610
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
2.3. Security screening The purpose of pre-board screening of passengers is to catch and confiscate objects that could be used as a weapon or otherwise pose a threat to flight security. In order to achieve that goal, it is necessary to inspect both the passenger and his/her carry-on baggage, in a manner that is as little intrusive as possible to avoid causing a high level of stress and discomfort to the passenger. In addition, the screening process must be quick in order to avoid a bottleneck in the passenger flow. The standard model for passenger security checkpoints in most airports is the security channels. Each channel or lane is equipped with an arch-shaped magnetometer and an X-ray machine. Passengers are required to walk through the magnetometer which will sound an alarm if a metal object is detected. Meanwhile, the passenger’s carry-on items – such as purses, laptop computers and small bags – are scanned by the X-ray machine. The channels are laid out at points of access to the boarding gates, creating a ‘‘secure’’ area where all boarding gates are located and accessible only to people who have been screened. 2.4. Departure lounge The departure lounge serves as an assembly area for passengers waiting to board a particular flight and, in many cases, as the exit passageway for deplaning passengers. It is generally sized to accommodate the number of boarding passengers expected to be in the lounge 15 min prior to the scheduled departure time, assuming this is the time when aircraft boarding begins. The space should accommodate seating for these passengers (although not all need to be seated), space for airline processing plus passengers queues, and an exit for deplaning passengers. Advanced processes for designing departure lounges have been discussed in Wirasinghe and Shehata (1988) and de Barros and Wirasinghe (2002). It is customary to allow only ticketed passengers in the departure lounge. This is the procedure adopted at Sa˜o Paulo/Guarulhos international airport. 2.5. Circulation areas In general, the terminal circulation component is considered a pedestrian circulation problem and analyzed by using procedures and standards, such as those suggested in the IATA manuals (International Air Transport Association, 2004). The length of the passenger’s pathway, the passenger’s walking speed, the number of level changes and the degree of interference the passenger encounters along the way are key variables in the assessment of circulation areas. 2.6. Concessions A passenger survey by Seneviratne and Martel (1991) revealed that accessibility to concessions and services is the second most significant characteristic, or indicator, of performance in waiting areas. The concessions in that study included rest rooms, communication facilities (i.e., phones and facsimile), retail outlets and restaurants. 2.7. Walking distance Walking in terminals is one of the most important, most controversial, and least understood activities. Walking distances in some terminals, especially for transfer passengers, have become quite long. The walking distance has been used by many researchers as an important measure of the level of service for an airport passenger terminal: Bandara and Wirasinghe (1992), Seneviratne and Martel (1991), Correia (2000), de Neufville et al. (2002), and de Barros and Wirasinghe (2003). Although its importance as a level of service measure is recognized, there has been no study to evaluate the impact of the walking distance on the LOS according to passenger perceptions. 2.8. Orientation Orientation can be defined as a person’s perception of his/her position relative to the surroundings, while walking, using mechanical systems, or driving a vehicle (Hart, 1985). When an airport provides poor
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
611
orientation systems, this may cause unfamiliar passengers to walk more than necessary, causing inconvenience, frustration, and delays in some extreme cases. Clearly, there is a correlation between orientation and walking distance. Advanced processes for measuring orientation have been proposed by Dada and Wirasinghe (1999). A basic approach to measuring the level of service of orientation has been proposed by Dada and Wirasinghe (2002). 2.9. Total time The main advantage of air transportation over other modes for long distance travel is the reduced travel time between origin and destination. The total travel time is comprised of access time, terminal time and air time. The access time is a matter of major concern. In some cases, this time exceeds the air time. For a journey of 500 km between two large metropolitan areas, the ground time can be as much as twice the air travel time. Similar comparisons can be made between the terminal time and air time. The terminal time can be one of the major components of the total travel time. This fact is especially true when there are many connections along the route. 3. Theoretical framework It is useful to discuss the different assumptions on which the methodology for overall LOS can be based. For this research, we will consider the following three hypotheses: weighted average, maximum value and minimum value. With the weighted average approach, it is assumed that passengers combine their experiences at different terminal components into a weighted average of individual LOS. An important step in this method is determining the weights associated with each component, that is, their relative importance as assigned by passengers. These assigned weights are of high importance for managers and designers, because they will allow them to focus their attention on the most important components. Considering this hypothesis, a bad passenger experience in a given component can be counterbalanced by a good experience in another component. Another approach that can be employed for overall LOS evaluation is based on the maximum LOS value. In this case, it is assumed that passengers evaluate the overall terminal LOS according to the maximum LOS experienced in any of the terminal components. Considering this, a departing passenger experiencing LOS A at check-in, but LOS C for all remaining components, will still assign LOS A to the overall terminal level of service. The opposite of this approach is to assume that passengers evaluate their overall terminal LOS according to the worst experience they face. For instance, if on arrival a passenger experiences LOS A for all components, except for baggage claim, where he/she experiences a LOS E, his/her overall terminal experience will be evaluated as LOS E. Although these two hypotheses are very simplistic in nature, they represent alternative concepts to the weighted average approach, which requires data (component weights) that are difficult to gather. An improvement to the maximum and minimum LOS approaches would be employing statistical measures of LOS, such as mode, median or mean. Suppose for various airport passenger terminal components, we can get a vector representing LOS evaluations for all the individual components, e.g., V : ðA; B; D; A; A; B; B; AÞ:
ð1Þ
This vector could represent LOS evaluations for a departing path: curb (LOS A), check-in (LOS B), departure lounge (LOS D), etc. The most frequent value represented in the above vector is LOS A (four times), which is the modal value. The median is between A and B, and the mean may be determined only if numerical values are assigned to the letters. Coincidentally, LOS A is also the maximum LOS value but that may not be the case for other situations. The minimum value (LOS D) is very far from the mode and occurs for just one component. The unfairness of applying the minimum LOS value approach for this evaluation is clear, especially if the component represented by LOS D is not ‘‘so important’’ according to user perceptions. The mode, median and mean value approach can also be criticized based on a relative weight perspective; it may be the case that the most frequent LOS value actually represents components that do not have high weights according to passengers’ perceptions.
612
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
Although the weighted average approach is more complex, requiring data that are relatively difficult to obtain, it is able to represent a balanced and adequate overall LOS evaluation. This is the approach used in this paper. Thus, we must find a suitable methodology to determine the relative weights of different parameters from the user’s point of view. 3.1. Additive approach The additive approach method is employed for obtaining the composite equations representing the overall level of service for the airport passenger terminal. Using this approach, the composite equation can be developed as follows: X LOS ðOverallÞ ¼ wi LOS ðX i Þ; ð2Þ where wi are positive scaling constants (or weights) among the different terminal components and characteristics, Xi (check-in counter, departure lounge, walking distance, total time, etc.). This function allows us to add the separate contributions of the different attributes to obtain the total level of service measure. It is the best known of the multi-attribute functions, and it is important both because of its relevance to some real problems and its relative simplicity (Keeney and Raiffa, 1976). It should be mentioned that the use of the weighting scheme is possible if certain relations are held. These are known as the concept of worth independence and are defined by the following statements (Rand Corporation, 1969): • The relative importance of satisfying separate attributes does not depend upon the various degrees to which each attribute has itself been satisfied. Rather, their relative importance is conceived as being constant in this respect. • The rate at which increased satisfaction of any given attribute contributes to overall worth is independent of the level of satisfaction already achieved on that and other attributes. Such rates are considered constant. • The rate at which decision makers would be willing to trade off decreased satisfaction on one attribute for increased satisfaction on other attributes, so as to preserve the same overall worth, is independent of the level of satisfaction already achieved by any or all of the attributes. There are some procedures that can be applied to verify whether the explanatory variables are additive independent and if either attribute is independent of the other. We will make use of analysis of correlation between variables to determine the degree of multi-collinearity between them (e.g., Miles and Shelvin, 2001). If it is found that the variables are not independent of each other, the analyst must work on reducing the dimensionality of the problem (Keeney and Raiffa, 1976). 3.2. Weighting values Weighting functions have been used in the past through a variety of available methods, including ranking, rating and pairwise comparisons. The ranking method is useful for obtaining the most important attribute in a given set. Nevertheless, it cannot provide the quantitative preferences for the other attributes. The application of this method to LOS evaluation has been criticized by Mu¨ller and Gosling (1991). An alternative to solve this issue could be the application of the rating method; however, it is not clear if passengers can meaningfully answer questions asking them to assign relative values to widely different metrics. The pairwise comparison scheme – best known as analytical hierarchy process (AHP) – is more complete and can overcome the difficulties associated with the ranking and rating methods (Taylor, 1999; Taha, 1997). The crux of the AHP method is the determination of the relative weights. Assuming that we are dealing with n criteria, the procedure establishes an n · n pairwise comparison matrix that reflects the decision maker’s judgment of the relative importance of the different criteria. These comparisons are made using a preference scale, which assigns numerical values to different levels of preference.
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
613
From the analysis of the three methods, it is clear that the pairwise comparison scheme is also a very good method for obtaining importance relative weights of the components. Unfortunately, however, its data needs are so great that application to an airport passenger terminal may be impractical. We propose an alternative approach, which is capable of obtaining weights without necessarily inquiring passengers directly. This approach can be explained through the following example. In an attitudinal survey, users are asked to declare LOS ratings for each of the ten attributes proposed (check-in counter, departure lounge, walking distance, total time, etc.) along with an overall LOS measure. Then, a regression analysis can be fitted as follows: LOS ðoverallÞ ¼ w1 LOSðA1 Þ þ w2 LOSðA2 Þ þ þ w10 LOSðA10 Þ;
ð3Þ
where LOS (overall) is overall LOS measure; LOS (A1), LOS (A2), . . ., LOS (A10) is LOS ratings for individual components; w1, w2, . . ., w10 is weights. The weights, w1, w2, . . ., w10, are the parameters of the regression equation, which can be obtained with the ordinary least squares (OLS) method. In this case, the weights are ‘‘revealed’’ via the passengers’ perceptions of the relative importance of each component. 4. Data collection A detailed passenger survey was undertaken in order to obtain user opinions about the level of service. Revealed preference technique was employed, meaning that the questions concerned the evaluation of existing and experienced situations. All interviews were made by trained professionals, and the questionnaires were completed in the departure lounge, five minutes before the scheduled boarding time. 4.1. Questionnaires A sequence of logical steps that must be followed to develop a good questionnaire, as suggested by Aaker et al. (1998), was applied. It was utilized in a preliminary survey at three Brazilian airports during the summer of 2003. Some corrections were made, and the improved questionnaire was finally administered at the Sa˜o Paulo/Guarulhos international airport during the summer of 2004. The basic changes were the inclusion of some variables that needed to be present in the LOS evaluation. These variables were suggestions made by airport users to the interviewers: orientation, walking distance, and security environment. Some changes were also made based on the preliminary statistical analysis. Some variables presented a high degree of correlation and were removed from the analysis. The questionnaires were developed with the purpose of getting the following specific information: • Type of flight: international or domestic. • Trip purpose: business or non-business. Passengers’ expectations are very different depending on the trip purpose. • Movement type: arriving or departing. • Gender: male or female. • Airline. • User opinions about LOS: five categories were used for passengers’ ratings: (1) unacceptable, (2) poor, (3) regular, (4) good and (5) excellent. • Facility characteristics. 4.2. Case study of Sa˜o Paulo/Guarulhos international airport Sa˜o Paulo/Guarulhos international airport handled almost 16 million passengers in 2005, making it the busiest airport in South America. The schematic view of the airport terminal buildings is presented in Fig. 1. The terminal buildings are designed as pier fingers. There are some minor commercial stores and services, such as a post office, pharmacy and banks, on the mezzanine floor.
614
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
Fig. 1. Layout of Sa˜o Paulo/Guarulhos international airport.
4.3. Summary of responses One hundred and nineteen passengers were interviewed in two surveys carried at Sa˜o Paulo/Guarulhos international airport. The pilot survey was performed on June 16–22, 2003, where 40 passengers were observed and interviewed. The final survey was applied on May 10–16, 2004, where 79 passengers were observed and interviewed. Of the 119 passengers, 47.1% were embarking on a domestic trip, while 52.9% were on an international trip; 44.5% were on a non-business flight, while 55.5% were on a business/combined flight; and finally, 72.3% were male and 27.7% were female. Table 1 and Figs. 2 and 3 present the distribution of responses concerning the level of service ratings provided for individual attributes and for the overall terminal. Table 1 Distribution of responses – Sa˜o Paulo/Guarulhos international airport Category
1 2 3 4 5
Percentage of responses (%) Curb
Check
Sec.
Lounge
Circul.
Conc.
Walking
Orient.
Time
Overall
0.0 9.2 21.8 56.3 12.6
0.0 3.4 15.1 63.0 18.5
1.7 3.4 10.2 54.2 30.5
2.6 6.8 29.1 46.2 15.4
0.8 3.4 16.9 63.6 15.3
0.8 6.8 33.1 44.9 14.4
0.0 11.7 27.3 49.4 11.7
2.6 11.7 18.2 53.2 14.3
0.0 5.1 23.1 55.1 16.7
0.8 3.4 18.6 64.4 12.7
Categories (LOS Ratings): 1-unacceptable; 2-poor; 3-regular; 4-good; 5-excellent.
Percentage of Responses
70 60
Curbside Check-in
50 40
Security Screening Lounge
30 20 10 0 1
2
3 4 Rating Categories
5
Fig. 2. Passenger LOS Ratings – individual components. (1 – unacceptable; 2 – poor; 3 – regular; 4 – good and 5 – execellent.)
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
615
Percentage of Responses
70 60
Circulation Walking Distance
50 40
Orientation Time Concessions
30
Overall Terminal
20 10 1
2
3 4 Rating Categories
5
Fig. 3. Passenger LOS Ratings – overall components. (1 – unacceptable; 2 – poor; 3 – regular; 4 – good and 5 – execellent.)
It can be noted from Figs. 2 and 3 that the LOS ratings are not uniform among attributes. However, the overall LOS ratings are roughly proportional to the LOS of individual attributes. This proportionality indicates that the assumptions proposed in this paper may be reasonable. The next section identifies a function to explain this relationship. 5. Data analysis As previously mentioned, a regression analysis will be performed between the overall LOS user ratings (1–5) and LOS user ratings (1–5) for individual components and attributes. Nevertheless, several a priori steps should be taken. These steps include the application of statistical tests to check for correlation, significance of parameters, and fitness of model. 5.1. Correlation among variables A problem often encountered in multiple regression is multi-collinearity, or the amount of ‘‘overlapping’’ information about the dependent variable that is provided by several independent variables (Taylor, 1999). This problem usually occurs when the independent variables are highly correlated. The correlation factor measures the degree of collinearity between two variables. A correlation of 1 means that there is a perfect (linear) negative relationship. A correlation of +1 means that there is a perfect (linear) positive relationship. A correlation of 0 means that there is no linear relationship. Table 2 presents the correlation among the independent variables calculated by SPSS – Statistical Software.
Table 2 Correlation of variables at the overall level
Curbside Check-in Security scr. Lounge Walking dist. Orientation Total time Circulation Concessions Security env.
Curb
Check-in
Sec. screen.
Dep. lounge
Walk. dist.
Orient.
Total Time
Circ.
Conc.
Sec. environ.
1.0 0.2 0.4 0.3 0.3 0.3 0.3 0.5 0.4 0.4
1.0 0.2 0.2 0.2 0.4 0.4 0.2 0.3 0.1
1.0 0.3 0.3 0.3 0.2 0.3 0.2 0.2
1.0 0.2 0.3 0.3 0.2 0.2 0.3
1.0 0.4 0.6 0.2 0.1 0.1
1.0 0.4 0.3 0.4 0.3
1.0 0.2 0.2 0.1
1.0 0.5 0.4
1.0 0.4
1.0
616
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
The inclusion or removal of variables must be done with care. A model may become non-representative because important variables are not included or improperly removed. Additionally, each airport may have a different model specification, which will be a function of operational, financial and socio-economic characteristics. The correlation between variables is important, but it is not the only criteria for inclusion or removal of variables in the model. The highest correlation value was found between the ratings of total time and walking distance. The reason for this is that passengers spend a considerable time walking in the airport to get to the desired destinations (check-in counter, departure lounge, gate, concessions, services, etc.). There are also fairly strong correlations between the total time and check-in, and between total time and orientation. Passengers spend a long time in check-in counter lines at Sa˜o Paulo/Guarulhos international airport, especially the ones traveling overseas. Some surveyed passengers spent more than 60 min in the check-in counter alone, which may be one of the reasons for the correlation value (0.4) between these two variables. The correlation between total time and orientation may be explained by the fact that users who do not receive good orientation at an airport passenger terminal spend more time than necessary on wayfinding. These same ideas may explain the fairly strong relationship between walking distance and orientation (0.4). It is not convenient to include variables that have a high degree of correlation in a multiple regression model. In this case, it is appropriate to analyze the overall terminal LOS without including the total time variable. Nevertheless, this variable will be indirectly represented by many other variables (curbside, check-in counter, security screening, departure lounge, etc.), because a small share of the total time is included in each of these components. Two other variables require additional attention: circulation and security environment. These variables were originally intended to be part of the overall analysis, but were later removed from the evaluation. The respondent passengers found it very difficult to provide a LOS opinion about these two components, because the definitions of circulation and security environment were not clear to them. Besides, just a few passengers (if any) were able to evaluate the security environment of the airport. Most passengers had no security concerns at the airport. In this case, they had not ‘experienced’ the security environment at the airport at different levels, as they had experienced the waiting time at the check-in counter or the crowdedness at the departure lounge. The correlation values for the independent variables, after excluding total time, circulation, and security environment from the analysis, are presented in Table 3. All these variables will be employed in the regression analysis to obtain a composite measure, because the correlation between these variables is only moderate. 5.2. Composite evaluation The ratings of the variables are combined, according to the following equation: LOS ðoverallÞ ¼ w0 þ w1 LOS ðcurbÞ þ w2 LOSðcheck-inÞ þ w3 LOSðsec: sc:Þ þ w4 LOS ðloungeÞ þ w5 LOS ðwalking dist:Þ þ w6 LOS ðorientationÞ þ w7 LOS ðconcessionsÞ;
ð4Þ
Table 3 Correlation of variables at the overall level
Curbside Check-in Security screening Lounge Walking distance Orientation Concessions
Curb
Check-in
Sec. screen.
Dep. lounge
Walking distance
Orientation
Concessions
1.0 0.2 0.4 0.3 0.3 0.3 0.4
1.0 0.2 0.2 0.2 0.4 0.3
1.0 0.3 0.3 0.3 0.2
1.0 0.2 0.3 0.2
1.0 0.4 0.1
1.0 0.4
1.0
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
617
where LOS (overall) is overall terminal LOS ratings; LOS (curb), LOS (check-in), LOS (sec. screen.), LOS (lounge), LOS (walk. dist.), LOS (orientation) and LOS (concessions) is LOS ratings for each individual attributes; w0 is intercept; and w1, w2, w3, w4, w5, w6, and w7 is parameters of the equation. Substituting the LOS ratings of the above equation by the responses (1–5) of the survey conducted at Sa˜o Paulo/Guarulhos international airport and performing a regression analysis will provide the values w1, w2, w3, w4, w5, w6, and w7 as the parameters of the regression equation, and w0 as the intercept. In this case, the weights are obtained, reflecting the user perceptions of the relative importance of components. The results of the regression analysis are shown in Table 4. It can be noticed from Table 4 that the walking distance parameter has a 94.5% probability of being equal to zero. This motivates us to remove this variable from the analysis. Although this is a very important factor for the overall terminal evaluation, it seems that passengers at Sa˜o Paulo/Guarulhos international airport do not value this factor. This regression analysis includes 78 observations. The 40 observations made in the first survey performed in the summer of 2003 are not included because walking distance, total time, and security environment were not present in those questionnaires. The subsequent calculations in this chapter will now include these 40 observations, because the absent variables are no longer objects of study for the overall LOS. New regression analyses have been performed with the following changes: • Non-inclusion of the walking distance variable. • Addition of the 40 observations made during the Summer 2003 survey. • Removing (stepwise) the variables with high P-values. The results of the best-fit regression analysis are presented in Table 5. The security screening and concessions variables presented high P-values. This means they have a high chance of being equal to zero. In addition, security screening had a negative value. This was expected. The security screening process at Sa˜o Paulo/
Table 4 Parameters, standard error, t Stat, and P-value – original variables Component
Parameters
Standard error
t Stat
P-value
Intercept Curbside Check-in Security screening Lounge Walking distance Orientation Concessions R2 = 0.48 F = 7.86 Observations: 78
0.832 0.359 0.019 0.099 0.117 0.006 0.196 0.027
0.533 0.102 0.108 0.095 0.077 0.083 0.090 0.103
1.562 3.518 0.180 1.042 1.522 0.069 2.191 0.261
0.123 0.001 0.857 0.301 0.133 0.945 0.032 0.795
Table 5 Parameters, standard error, t Stat, and P-value – best-fit regression analysis Component
Parameters
Standard error
t Stat
P-value
Intercept Curbside Check-in Lounge Orientation Purpose R2 = 0.443 F = 17.799 Observations: 118
0.755 0.313 0.114 0.118 0.238 0.243
0.358 0.070 0.075 0.062 0.068 0.103
2.108 4.504 1.513 1.889 3.517 2.365
0.037 0.000 0.133 0.061 0.001 0.020
618
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
Table 6 Parameters, Standard error, t Stat, and P-value – final model Component
Parameters
Standard error
t Stat
P-value
Intercept Curbside Check-in Lounge Orientation Purpose R2 = 0.47 F = 19.538 Observations: 116
0.841 0.246 0.144 0.151 0.229 0.214
0.327 0.065 0.069 0.057 0.063 0.094
2.575 3.809 2.094 2.643 3.656 2.291
0.011 0.000 0.039 0.009 0.001 0.024
Guarulhos international airport is very fast and does not cause any concern to passengers. The concessions component is not mandatory, and some users did not have a strong opinion about its level of service because they had not experienced it. One new variable was included in the analysis, because it fitted well in the regression equation. This was the ‘purpose’ variable; it was included as a dummy variable. It was equal to 0 when the trip purpose was business or combined business/non-business, and equal to 1 when the trip purpose was non-business. The final step in refining the analysis was the detection of possible outliers. The outliers were defined (in this analysis) as responses of passengers that were clearly inconsistent. It might be a passenger that evaluated the overall terminal as excellent, but evaluated all (or most) components as poor or unacceptable. The opposite can also occur: one passenger can evaluate the terminal as poor, but all the components as excellent. Two cases (out of 118 passengers) presented this inconsistency and were removed from the analysis. The results of the regression analysis without these two outliers are presented in Table 6. The parameters of Table 6 can be substituted into Eq. (4) to provide the following relation: LOS ðoverallÞ ¼ 0:841 þ 0:246 LOS ðcurbÞ þ 0:144 LOSðcheck-inÞ þ þ0:151 LOSðloungeÞ þ 0:229 LOS ðorientationÞ þ 0:214ðpurposeÞ:
ð5Þ
According to Eq. (5), the most important component for passengers is the curbside. Although that seems contrary to accepted practice, the reason may be that it was the first component experienced by the 118 surveyed passengers and that curbside is a problematic aspect of Sa˜o Paulo/Guarulhos international airport. The curbside was the first impression they had about the airport and, as such, affected the evaluation of the terminal as a whole. Orientation is the second most important component. The relatively low importance of the check-in and the departure lounge components, when compared to the curbside and orientation components, can be explained by the fact that, at Sa˜o Paulo/Guarulhos international airport, passengers are accustomed to spending a considerable time in these two components. The intercept (0.841) indicates that other variables may be included in this analysis. It signifies that there are one or more components of the overall LOS that are not represented by the explanatory variables included in the model. Future applications of the model may need to test other variables for which data was not collected as part of this research. Although the R2 is relatively moderate (0.47), it is common for estimating this type of model, where several variables are available and subjective assessments are collected for different types of passengers. Nevertheless, despite all these shortcomings, the model was able to explain almost 50% of the variances. 6. Conclusions This research provides a methodology for estimating a global index for LOS evaluations at airport passenger terminals. The main contribution to the field is the identification of the most important airport attributes according to user perceptions. This methodology was applied to Sa˜o Paulo/Guarulhos international airport and helped identify the passenger terminal curbside as one of the most important factors contributing to the passengers’ perception of level of service. The results of the model application also indicate that other
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
619
passenger terminal attributes not tested in this research may have a significant contribution to the passengers’ perceptions of LOS. Airport planners could use this method in practice for forecasting the overall LOS of future facility alternatives. Estimates of LOS for individual components (check-in, lounge, etc.) can be obtained from conventional methods and standards such as those provided by IATA (International Air Transport Association, 2004); however they do not indicate nor provide the overall LOS of an airport terminal. Eq. (5), if calibrated according to the characteristics of its passengers, could provide the overall LOS of each alternative. It also provides a data based method for comparing airports with respect to the level of service provided to passengers. Airport managers can also use the results of this method to obtain the degree of importance that users assign to various components and characteristics of the airport terminal. These weights are useful, as it will be one indication of where airport managers should invest their limited resources. For future research, this methodology can be applied to other airports to provide a more comprehensive understanding of the relationship between overall terminal measures and the level of service associated with them. Other variables not surveyed in this research should also be tested to improve the model’s explanatory power. Acknowledgements This research was supported in part by CAPES and FAPESP – agencies of the government of Brazil providing financial support to the training of highly skilled personnel, by the Natural Sciences and Engineering Research Council of Canada and by the University of Calgary. References Aaker, D., Kumar, V., Day, G., 1998. Marketing Research, 6th ed. John Wiley and Sons, New York. Airports Council International, 2000. Quality of Service at Airports: Standards and Measurements. ACI World Headquarters, Geneva. Bandara, S., Wirasinghe, S.C., 1992. Walking distance minimization for airport terminal configurations. Transportation Research A 22, 59–74. Correia, A.R., 2000. Quantitative Evaluation of Airport Passenger Terminal Configurations. M.Sc. dissertation, Aeronautical Institute of Technology, Brazil (in Portuguese). Correia, A.R., Wirasinghe, S.C., 2004. Evaluation of Level of Service at Airport Passenger Terminals: A Review of Research Approaches. Transportation Research Record 1888, National Research Council, Washington DC. pp. 1–6. Correia, A.R., Wirasinghe, S.C., de Barros, A.G., 2005. A New Procedure for Overall Level of Service Data Collection at Airport Passenger Terminals, in: Proceedings of the 9th ATRS World Conference, CD-ROM, Rio de Janeiro. Dada, E.S., Wirasinghe, S.C., 1999. Development of a New Orientation Index for Airport Terminals. Transportation Research Record 1662, National Research Council, Washington DC. pp. 41–47. Dada, E.S., Wirasinghe, S.C., 2002. Estimating Basic Level of Service for Passenger Orientation in New Airports. In: D. Wang and S.-M. Li (Eds), Proceedings of the 7th Conference of Hong Kong Society for Transportation Studies, Hong Kong. pp. 369–377. de Barros, A.G., Wirasinghe, S.C., 2002. Design and shared use of departure lounge for NLA operations. Journal of Advanced Transportation 36, 187–209. de Barros, A.G., Wirasinghe, S.C., 2003. Optimal terminal configurations for new large aircraft operations. Transportation Research A 37, 315–331. de Barros, A.G., Somasundaraswaran, A.K., Wirasinghe, S.C., 2006. Evaluation of level of service for transfer passengers at airports, in: Proceedings of the 10th ATRS World Conference, CD-ROM, Nagoya. de Neufville, R., de Barros, A.G., Belin, S.C., 2002. Optimal configuration of airport passenger buildings for travelers. Journal of Transportation Engineering 128, 211–217. Fernandes, E., Pacheco, R.R., 2002. Efficient use of airport capacity. Transportation Research Part A 36, 225–238. Hart, W., 1985. The Airport Passenger Terminal, 1st ed. John Wiley and Sons, New York. Horonjeff, R., McKelvey, F.X., 1994. Planning and Design of Airports, 4th ed. McGraw-Hill, New York. International Air Transport Association, 2004. Airport Development Reference Manual, 9th edition. Montreal. Keeney, R., Raiffa, H., 1976. Decisions with Multiples Objectives, 1st ed. John Wiley and Sons, New York. Magri, A.A., Alves, C.J.P., 2003. Convenient Airports: Point of View of the Passengers. ATRS – Air Transport Research Society World Conference, CD-ROM, Toulouse. Martel, N., Seneviratne, P.N.,1990. Analysis of Factors Influencing Quality of Service in Passenger Terminal Buildings. Transportation Research Record 1273, National Research Council, Washington DC. pp. 1–10. Miles, J., Shelvin, M., 2001. Applying Regression and Correlation, 1st ed. Sage, Publications, London.
620
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
Mu¨ller, C., Gosling, G.D., 1991. A framework for evaluating level of service for airport terminals. Transportation Planning and Technology 16, 45–61. Mumayiz, S.A., Ashford, N., 1986. Methodology for Planning and Operations Management of Airport Terminal Facilities. Transportation Research Record 1094, National Research Council, Washington DC. pp. 24–35. Ndoh, N.N., Ashford, N.J., 1993. Evaluation of Airport Access Level of Service. Transportation Research Record 1423, National Research Council, Washington DC. pp. 34–39. Omer, K.F., Khan, A.M., 1988. Airport Landside Level of Service Estimation: Utility Theoretic Approach. Transportation Research Record 1199, National Research Council, Washington DC. pp. 33–40. Park, Y.H., 1994. An evaluation methodology for the level of service at the airport landside system. Ph.D. thesis, Department of Transport Technology, Loughborough University of Technology, UK. Rand Corporation, 1969. Measurement and Evaluation of Transportation System Effectiveness. Memorandum RM-5869-DOT. Seneviratne, P.N., Martel, N., 1991. Variables influencing performance of air terminal buildings. Transportation Planning and Technology 16, 1177–1179. Siddiqui, M.R., 1994. A Statistical Analysis of the Factors Influencing the Level of Service of Airport Terminal Curbsides. M.Sc. thesis, Concordia University, Canada. Taha, H.A., 1997. Operations Research – An Introduction, 6th ed. Prentice Hall, Upper Saddle River, NJ. Taylor III, B.W., 1999. Introduction to Management Science, 6th ed. Prentice Hall, Upper Saddle River, NJ. Transportation Research Board, 1987. Special Report 215: Measuring Airport Landside Capacity. National Research Council, Washington DC. Transport Canada, 1979. A Discussion Paper on Level of Service Definition and Methodology for Calculating Airport Capacity. Report TP 2027, Ottawa. Wirasinghe, S.C., Shehata, M., 1988. Departure lounge sizing and optimal seating capacity for a given aircraft/flight mix – (i) Single Gate; (ii) Several Gates. Transportation Planning and Technology 13, 57–71. Yen, J.R., 1995. A new approach to measure the level of service of procedures in the airport landside. Transportation Planning Journal 24, 323–336. Yen, J.R., Teng, C.R., Chen, P.S., 2001. Measuring the Level of Service at Airport Passenger Terminals: Comparison of Perceived and Observed Time. Transportation Research Record 1744, National Research Council, Washington DC. pp. 17–23.
Transportation Research Part E 44 (2008) 607–620 www.elsevier.com/locate/tre
A global index for level of service evaluation at airport passenger terminals Anderson Ribeiro Correia a, S.C. Wirasinghe b, Alexandre G. de Barros a
b,*
Aeronautical Institute of Technology, Prac¸a Mal. Eduardo Gomes, 50 Sa˜o Jose´ dos Campos, SP 12.228-901, Brazil b University of Calgary, 2500 University Drive NW, Calgary, Alberta, Canada
Abstract This paper presents a global index for the evaluation of the level of service (LOS) of the operational components at an airport. This index is useful in evaluating the overall LOS on a single scale and according to user perceptions. It is assumed that the overall LOS is a function of the LOS of individual components, such as check-in, departure lounges, etc., as well as socio-economic variables. Regression analysis is used to obtain a mathematical relationship between the global LOS ratings and the LOS of individual operational components. The methodology is illustrated with its application at Sa˜o Paulo/ Guarulhos international airport in Brazil, where 119 enplaning passengers were interviewed. 2007 Elsevier Ltd. All rights reserved. Keywords: Level of service; Airports; Passenger terminal building; Regression analysis
1. Introduction The development of level of service (LOS) measures for airport passenger terminals has been one of the major issues for airport operators in the last several decades. This has motivated a number of LOS studies by air transportation agencies, including the Federal Aviation Administration – FAA (Transportation Research Board, 1987; Airports Council International, 2000; Transport Canada, 1979). Despite the efforts of these agencies, the proposed LOS standards and methods have been the subject of criticism by airport professionals. One of the main concerns is the lack of passenger input. Several studies have also been undertaken to develop methods for LOS evaluation taking into account user perceptions. Most of them have provided results based on poor database, and were not able to provide a high level of significance for testing the hypotheses considered. Additionally, most studies focused on individual components of the airport passenger terminal (check-in counter, departure lounge, etc.), neglecting the overall evaluation. A broad measure reflecting the LOS of the terminal as a whole for a given type of passenger (e.g., departing) would be useful at planning, design and management levels. With this measure, it would be possible to
*
Corresponding author. Tel.: +1 403 220 6713. E-mail address: [email protected] (A.G. de Barros).
1366-5545/$ - see front matter 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.tre.2007.05.009
608
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
identify the level of importance attributed to individual components by passengers, in order to prioritize investments. Additionally, it would provide a measure for comparison between different airports for the purpose of benchmarking. The main challenge when developing an overall measure is the data collection. It is relatively simple to collect characteristics of individual facilities (e.g., waiting time at the check-in counter) as opposed to obtaining overall measures (e.g., total walking distance by an individual). Several issues must be addressed before a research effort is developed to collect overall measures. It is the purpose of this paper to provide a methodology for such an effort, illustrating it with a case study of Sa˜o Paulo/Guarulhos international airport. A complete review of past research on LOS has been presented in Correia and Wirasinghe (2004). Mumayiz and Ashford (1986) proposed a perception–response concept, using graphical displays constructed from passenger responses concerning the LOS provided at airports in England. Omer and Khan (1988) employed the concept of utility theory to develop a relationship between characteristics of facilities (e.g., waiting time, space available) and user responses (0–1) about the LOS offered. Mu¨ller and Gosling (1991) applied a psychometric scaling technique to obtain a quantitative measure of LOS that might be used in a relationship similar to the one developed by Omer and Khan (1988). Seneviratne and Martel (1991) developed LOS standards for several components of the airport passenger terminal. The selection of the most important components and measures was based on a survey of Canadian airports (Martel and Seneviratne, 1999). Ndoh and Ashford (1993) employed theories of perception and scaling to evaluate LOS on airport access, using 12 attributes (e.g., cost, comfort, access to information, etc.). Park (1994) used fuzzy logic to derive LOS measures for specific components of the airport passenger terminal. The methodology was applied to the Seoul Kimpo Airport. Yen (1995) conducted a survey at Austin Municipal Airport in Texas, USA. He applied binary logit models to estimate a ‘‘long’’ model and a ‘‘short’’ model to predict the probabilities that a passenger will rate a service on the basis of perceived time measures. Yen et al. (2001) presented a quantitative model to define the level of service at airport passenger terminals. The model used the fuzzy concept to relate subjective service ratings to time measurements of associated waiting or service processes. Fernandes and Pacheco (2002) utilized data envelopment analysis to evaluate the capacity of 35 Brazilian airports, based on several operational parameters (e.g., number of check-in counters, average space available per passenger, etc.). Magri and Alves (2003) evaluated the LOS offered by six Brazilian airports as a function of 36 subjective parameters suggested by ACI (Airports Council International, 2000). All these studies concentrated on the evaluation of individual components. No study has developed an objective overall LOS measure, reflecting the LOS provided by the airport passenger terminal on a single scale. 2. Airport passenger terminal facilities and characteristics The terminal area is the major interface between the airfield and the rest of the airport. It includes the facilities for passenger and baggage processing, cargo handling, airport maintenance, operations and administration activities. The passenger terminal system has three major functional areas. These functional areas and the activities that occur within them are as follows (Horonjeff and McKelvey, 1994): • The access interface – where the passenger transfers from the access mode of travel to the passenger processing functional area. Circulation, parking, and curbside loading and unloading of passengers are the activities that take place within this functional area. • Processing – where the passenger is processed in preparation for starting, ending, or continuation of an air transportation trip. The primary activities in this functional area are ticketing, baggage check-in, baggage claim, seat assignment, federal inspection, services and security. • The flight interface – where the passenger transfers from the processing functional area to the aircraft. The activities that occur here include assembly, conveyance to and from the aircraft, and aircraft loading and unloading.
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
609
In order to provide a global index for LOS evaluation, it is necessary to specify the type of movement in question. Airport passengers can be divided into three groups according to their movement type: departing, arriving, and transfers. Each of these groups will have a different set of needs and wishes and, in many cases, will even make use of different facilities. For example, departing passengers will not make use of the baggage claim facilities, whereas arriving passengers will not use the enplaning curbside or the check-in lobby. Therefore, each movement type will have a LOS index which is global in the sense that it encompasses the passenger’s full airport experience as opposed to just rating one single facility or service. In this research, only departing domestic passengers are analyzed; however, the methodology presented herein can also be applied to evaluate the LOS for arriving and transfer passengers (de Barros et al., 2006). Additionally, this research concentrates on operational components of the airport passenger terminal: for this reason, parking, public transit or other components located outside of the terminal building are not addressed. In summary, this paper presents an overall LOS evaluation as a function of the following components: • • • • • •
Enplaning curbside. Ticket counter and baggage deposit. Security screening. Departure lounge. Circulation areas (corridors, stairs, elevators, etc.). Concessions.
Four overall LOS measures are included in the analysis, as they have been identified as very important for passengers (Correia et al., 2005): walking distance, orientation, total time and security environment. The last variable differs significantly from security screening. Security environment refers to the users’ perceptions of security throughout the terminal building. This is a subjective variable, which may influence the overall LOS to a certain extent. On the other hand, security screening refers to the quality of the experience of the passenger when being processed at the security screening location. Although some components are not directly managed by the airport operator, they may have considerable influence during the planning and management stages. For instance, check-in counters are usually managed by airlines, but are often planned and built by the airport operator. Moreover, in many countries the airport operator is responsible for assigning check-in areas according to demand priorities. The following are brief descriptions of the components and measures that are analyzed. 2.1. Enplaning curbside The curbside element is the interface between the terminal building and the ground transportation system. High traffic volumes and peaks, plus the complex flows of mixing people and vehicles, may result in extensive traffic congestion at the curbside area. This in turn may cause inconvenience, frustration, and delay to passengers at large airports. Arbitrary standards applied in many airports may lead to oversized or undersized facilities. The airport community is, therefore, interested in a methodology that could lead to rational standards (Siddiqui, 1994). 2.2. Ticket counter and baggage deposit The airline ticket counter is where the airline and passenger make final ticket transactions and check-in baggage for a flight. The ticket transaction takes place at the ticket counter, which is a stand-up desk. To the left and right of the ticket counter position, a low shelf is provided to deposit, check-in, tag and weigh baggage, if necessary. Subsequently, the baggage is passed back by the agent to an outbound baggage conveyance device located near the counter. The check-in counters located at Sa˜o Paulo/Guarulhos international airport operate with a single line for each individual airline. Self check-in and ticketing kiosks are in use at several airports worldwide. However, these devices are not yet used at Sa˜o Paulo/Guarulhos international airport. For this reason, the evaluation of LOS for these components is not addressed in this research.
610
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
2.3. Security screening The purpose of pre-board screening of passengers is to catch and confiscate objects that could be used as a weapon or otherwise pose a threat to flight security. In order to achieve that goal, it is necessary to inspect both the passenger and his/her carry-on baggage, in a manner that is as little intrusive as possible to avoid causing a high level of stress and discomfort to the passenger. In addition, the screening process must be quick in order to avoid a bottleneck in the passenger flow. The standard model for passenger security checkpoints in most airports is the security channels. Each channel or lane is equipped with an arch-shaped magnetometer and an X-ray machine. Passengers are required to walk through the magnetometer which will sound an alarm if a metal object is detected. Meanwhile, the passenger’s carry-on items – such as purses, laptop computers and small bags – are scanned by the X-ray machine. The channels are laid out at points of access to the boarding gates, creating a ‘‘secure’’ area where all boarding gates are located and accessible only to people who have been screened. 2.4. Departure lounge The departure lounge serves as an assembly area for passengers waiting to board a particular flight and, in many cases, as the exit passageway for deplaning passengers. It is generally sized to accommodate the number of boarding passengers expected to be in the lounge 15 min prior to the scheduled departure time, assuming this is the time when aircraft boarding begins. The space should accommodate seating for these passengers (although not all need to be seated), space for airline processing plus passengers queues, and an exit for deplaning passengers. Advanced processes for designing departure lounges have been discussed in Wirasinghe and Shehata (1988) and de Barros and Wirasinghe (2002). It is customary to allow only ticketed passengers in the departure lounge. This is the procedure adopted at Sa˜o Paulo/Guarulhos international airport. 2.5. Circulation areas In general, the terminal circulation component is considered a pedestrian circulation problem and analyzed by using procedures and standards, such as those suggested in the IATA manuals (International Air Transport Association, 2004). The length of the passenger’s pathway, the passenger’s walking speed, the number of level changes and the degree of interference the passenger encounters along the way are key variables in the assessment of circulation areas. 2.6. Concessions A passenger survey by Seneviratne and Martel (1991) revealed that accessibility to concessions and services is the second most significant characteristic, or indicator, of performance in waiting areas. The concessions in that study included rest rooms, communication facilities (i.e., phones and facsimile), retail outlets and restaurants. 2.7. Walking distance Walking in terminals is one of the most important, most controversial, and least understood activities. Walking distances in some terminals, especially for transfer passengers, have become quite long. The walking distance has been used by many researchers as an important measure of the level of service for an airport passenger terminal: Bandara and Wirasinghe (1992), Seneviratne and Martel (1991), Correia (2000), de Neufville et al. (2002), and de Barros and Wirasinghe (2003). Although its importance as a level of service measure is recognized, there has been no study to evaluate the impact of the walking distance on the LOS according to passenger perceptions. 2.8. Orientation Orientation can be defined as a person’s perception of his/her position relative to the surroundings, while walking, using mechanical systems, or driving a vehicle (Hart, 1985). When an airport provides poor
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
611
orientation systems, this may cause unfamiliar passengers to walk more than necessary, causing inconvenience, frustration, and delays in some extreme cases. Clearly, there is a correlation between orientation and walking distance. Advanced processes for measuring orientation have been proposed by Dada and Wirasinghe (1999). A basic approach to measuring the level of service of orientation has been proposed by Dada and Wirasinghe (2002). 2.9. Total time The main advantage of air transportation over other modes for long distance travel is the reduced travel time between origin and destination. The total travel time is comprised of access time, terminal time and air time. The access time is a matter of major concern. In some cases, this time exceeds the air time. For a journey of 500 km between two large metropolitan areas, the ground time can be as much as twice the air travel time. Similar comparisons can be made between the terminal time and air time. The terminal time can be one of the major components of the total travel time. This fact is especially true when there are many connections along the route. 3. Theoretical framework It is useful to discuss the different assumptions on which the methodology for overall LOS can be based. For this research, we will consider the following three hypotheses: weighted average, maximum value and minimum value. With the weighted average approach, it is assumed that passengers combine their experiences at different terminal components into a weighted average of individual LOS. An important step in this method is determining the weights associated with each component, that is, their relative importance as assigned by passengers. These assigned weights are of high importance for managers and designers, because they will allow them to focus their attention on the most important components. Considering this hypothesis, a bad passenger experience in a given component can be counterbalanced by a good experience in another component. Another approach that can be employed for overall LOS evaluation is based on the maximum LOS value. In this case, it is assumed that passengers evaluate the overall terminal LOS according to the maximum LOS experienced in any of the terminal components. Considering this, a departing passenger experiencing LOS A at check-in, but LOS C for all remaining components, will still assign LOS A to the overall terminal level of service. The opposite of this approach is to assume that passengers evaluate their overall terminal LOS according to the worst experience they face. For instance, if on arrival a passenger experiences LOS A for all components, except for baggage claim, where he/she experiences a LOS E, his/her overall terminal experience will be evaluated as LOS E. Although these two hypotheses are very simplistic in nature, they represent alternative concepts to the weighted average approach, which requires data (component weights) that are difficult to gather. An improvement to the maximum and minimum LOS approaches would be employing statistical measures of LOS, such as mode, median or mean. Suppose for various airport passenger terminal components, we can get a vector representing LOS evaluations for all the individual components, e.g., V : ðA; B; D; A; A; B; B; AÞ:
ð1Þ
This vector could represent LOS evaluations for a departing path: curb (LOS A), check-in (LOS B), departure lounge (LOS D), etc. The most frequent value represented in the above vector is LOS A (four times), which is the modal value. The median is between A and B, and the mean may be determined only if numerical values are assigned to the letters. Coincidentally, LOS A is also the maximum LOS value but that may not be the case for other situations. The minimum value (LOS D) is very far from the mode and occurs for just one component. The unfairness of applying the minimum LOS value approach for this evaluation is clear, especially if the component represented by LOS D is not ‘‘so important’’ according to user perceptions. The mode, median and mean value approach can also be criticized based on a relative weight perspective; it may be the case that the most frequent LOS value actually represents components that do not have high weights according to passengers’ perceptions.
612
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
Although the weighted average approach is more complex, requiring data that are relatively difficult to obtain, it is able to represent a balanced and adequate overall LOS evaluation. This is the approach used in this paper. Thus, we must find a suitable methodology to determine the relative weights of different parameters from the user’s point of view. 3.1. Additive approach The additive approach method is employed for obtaining the composite equations representing the overall level of service for the airport passenger terminal. Using this approach, the composite equation can be developed as follows: X LOS ðOverallÞ ¼ wi LOS ðX i Þ; ð2Þ where wi are positive scaling constants (or weights) among the different terminal components and characteristics, Xi (check-in counter, departure lounge, walking distance, total time, etc.). This function allows us to add the separate contributions of the different attributes to obtain the total level of service measure. It is the best known of the multi-attribute functions, and it is important both because of its relevance to some real problems and its relative simplicity (Keeney and Raiffa, 1976). It should be mentioned that the use of the weighting scheme is possible if certain relations are held. These are known as the concept of worth independence and are defined by the following statements (Rand Corporation, 1969): • The relative importance of satisfying separate attributes does not depend upon the various degrees to which each attribute has itself been satisfied. Rather, their relative importance is conceived as being constant in this respect. • The rate at which increased satisfaction of any given attribute contributes to overall worth is independent of the level of satisfaction already achieved on that and other attributes. Such rates are considered constant. • The rate at which decision makers would be willing to trade off decreased satisfaction on one attribute for increased satisfaction on other attributes, so as to preserve the same overall worth, is independent of the level of satisfaction already achieved by any or all of the attributes. There are some procedures that can be applied to verify whether the explanatory variables are additive independent and if either attribute is independent of the other. We will make use of analysis of correlation between variables to determine the degree of multi-collinearity between them (e.g., Miles and Shelvin, 2001). If it is found that the variables are not independent of each other, the analyst must work on reducing the dimensionality of the problem (Keeney and Raiffa, 1976). 3.2. Weighting values Weighting functions have been used in the past through a variety of available methods, including ranking, rating and pairwise comparisons. The ranking method is useful for obtaining the most important attribute in a given set. Nevertheless, it cannot provide the quantitative preferences for the other attributes. The application of this method to LOS evaluation has been criticized by Mu¨ller and Gosling (1991). An alternative to solve this issue could be the application of the rating method; however, it is not clear if passengers can meaningfully answer questions asking them to assign relative values to widely different metrics. The pairwise comparison scheme – best known as analytical hierarchy process (AHP) – is more complete and can overcome the difficulties associated with the ranking and rating methods (Taylor, 1999; Taha, 1997). The crux of the AHP method is the determination of the relative weights. Assuming that we are dealing with n criteria, the procedure establishes an n · n pairwise comparison matrix that reflects the decision maker’s judgment of the relative importance of the different criteria. These comparisons are made using a preference scale, which assigns numerical values to different levels of preference.
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
613
From the analysis of the three methods, it is clear that the pairwise comparison scheme is also a very good method for obtaining importance relative weights of the components. Unfortunately, however, its data needs are so great that application to an airport passenger terminal may be impractical. We propose an alternative approach, which is capable of obtaining weights without necessarily inquiring passengers directly. This approach can be explained through the following example. In an attitudinal survey, users are asked to declare LOS ratings for each of the ten attributes proposed (check-in counter, departure lounge, walking distance, total time, etc.) along with an overall LOS measure. Then, a regression analysis can be fitted as follows: LOS ðoverallÞ ¼ w1 LOSðA1 Þ þ w2 LOSðA2 Þ þ þ w10 LOSðA10 Þ;
ð3Þ
where LOS (overall) is overall LOS measure; LOS (A1), LOS (A2), . . ., LOS (A10) is LOS ratings for individual components; w1, w2, . . ., w10 is weights. The weights, w1, w2, . . ., w10, are the parameters of the regression equation, which can be obtained with the ordinary least squares (OLS) method. In this case, the weights are ‘‘revealed’’ via the passengers’ perceptions of the relative importance of each component. 4. Data collection A detailed passenger survey was undertaken in order to obtain user opinions about the level of service. Revealed preference technique was employed, meaning that the questions concerned the evaluation of existing and experienced situations. All interviews were made by trained professionals, and the questionnaires were completed in the departure lounge, five minutes before the scheduled boarding time. 4.1. Questionnaires A sequence of logical steps that must be followed to develop a good questionnaire, as suggested by Aaker et al. (1998), was applied. It was utilized in a preliminary survey at three Brazilian airports during the summer of 2003. Some corrections were made, and the improved questionnaire was finally administered at the Sa˜o Paulo/Guarulhos international airport during the summer of 2004. The basic changes were the inclusion of some variables that needed to be present in the LOS evaluation. These variables were suggestions made by airport users to the interviewers: orientation, walking distance, and security environment. Some changes were also made based on the preliminary statistical analysis. Some variables presented a high degree of correlation and were removed from the analysis. The questionnaires were developed with the purpose of getting the following specific information: • Type of flight: international or domestic. • Trip purpose: business or non-business. Passengers’ expectations are very different depending on the trip purpose. • Movement type: arriving or departing. • Gender: male or female. • Airline. • User opinions about LOS: five categories were used for passengers’ ratings: (1) unacceptable, (2) poor, (3) regular, (4) good and (5) excellent. • Facility characteristics. 4.2. Case study of Sa˜o Paulo/Guarulhos international airport Sa˜o Paulo/Guarulhos international airport handled almost 16 million passengers in 2005, making it the busiest airport in South America. The schematic view of the airport terminal buildings is presented in Fig. 1. The terminal buildings are designed as pier fingers. There are some minor commercial stores and services, such as a post office, pharmacy and banks, on the mezzanine floor.
614
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
Fig. 1. Layout of Sa˜o Paulo/Guarulhos international airport.
4.3. Summary of responses One hundred and nineteen passengers were interviewed in two surveys carried at Sa˜o Paulo/Guarulhos international airport. The pilot survey was performed on June 16–22, 2003, where 40 passengers were observed and interviewed. The final survey was applied on May 10–16, 2004, where 79 passengers were observed and interviewed. Of the 119 passengers, 47.1% were embarking on a domestic trip, while 52.9% were on an international trip; 44.5% were on a non-business flight, while 55.5% were on a business/combined flight; and finally, 72.3% were male and 27.7% were female. Table 1 and Figs. 2 and 3 present the distribution of responses concerning the level of service ratings provided for individual attributes and for the overall terminal. Table 1 Distribution of responses – Sa˜o Paulo/Guarulhos international airport Category
1 2 3 4 5
Percentage of responses (%) Curb
Check
Sec.
Lounge
Circul.
Conc.
Walking
Orient.
Time
Overall
0.0 9.2 21.8 56.3 12.6
0.0 3.4 15.1 63.0 18.5
1.7 3.4 10.2 54.2 30.5
2.6 6.8 29.1 46.2 15.4
0.8 3.4 16.9 63.6 15.3
0.8 6.8 33.1 44.9 14.4
0.0 11.7 27.3 49.4 11.7
2.6 11.7 18.2 53.2 14.3
0.0 5.1 23.1 55.1 16.7
0.8 3.4 18.6 64.4 12.7
Categories (LOS Ratings): 1-unacceptable; 2-poor; 3-regular; 4-good; 5-excellent.
Percentage of Responses
70 60
Curbside Check-in
50 40
Security Screening Lounge
30 20 10 0 1
2
3 4 Rating Categories
5
Fig. 2. Passenger LOS Ratings – individual components. (1 – unacceptable; 2 – poor; 3 – regular; 4 – good and 5 – execellent.)
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
615
Percentage of Responses
70 60
Circulation Walking Distance
50 40
Orientation Time Concessions
30
Overall Terminal
20 10 1
2
3 4 Rating Categories
5
Fig. 3. Passenger LOS Ratings – overall components. (1 – unacceptable; 2 – poor; 3 – regular; 4 – good and 5 – execellent.)
It can be noted from Figs. 2 and 3 that the LOS ratings are not uniform among attributes. However, the overall LOS ratings are roughly proportional to the LOS of individual attributes. This proportionality indicates that the assumptions proposed in this paper may be reasonable. The next section identifies a function to explain this relationship. 5. Data analysis As previously mentioned, a regression analysis will be performed between the overall LOS user ratings (1–5) and LOS user ratings (1–5) for individual components and attributes. Nevertheless, several a priori steps should be taken. These steps include the application of statistical tests to check for correlation, significance of parameters, and fitness of model. 5.1. Correlation among variables A problem often encountered in multiple regression is multi-collinearity, or the amount of ‘‘overlapping’’ information about the dependent variable that is provided by several independent variables (Taylor, 1999). This problem usually occurs when the independent variables are highly correlated. The correlation factor measures the degree of collinearity between two variables. A correlation of 1 means that there is a perfect (linear) negative relationship. A correlation of +1 means that there is a perfect (linear) positive relationship. A correlation of 0 means that there is no linear relationship. Table 2 presents the correlation among the independent variables calculated by SPSS – Statistical Software.
Table 2 Correlation of variables at the overall level
Curbside Check-in Security scr. Lounge Walking dist. Orientation Total time Circulation Concessions Security env.
Curb
Check-in
Sec. screen.
Dep. lounge
Walk. dist.
Orient.
Total Time
Circ.
Conc.
Sec. environ.
1.0 0.2 0.4 0.3 0.3 0.3 0.3 0.5 0.4 0.4
1.0 0.2 0.2 0.2 0.4 0.4 0.2 0.3 0.1
1.0 0.3 0.3 0.3 0.2 0.3 0.2 0.2
1.0 0.2 0.3 0.3 0.2 0.2 0.3
1.0 0.4 0.6 0.2 0.1 0.1
1.0 0.4 0.3 0.4 0.3
1.0 0.2 0.2 0.1
1.0 0.5 0.4
1.0 0.4
1.0
616
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
The inclusion or removal of variables must be done with care. A model may become non-representative because important variables are not included or improperly removed. Additionally, each airport may have a different model specification, which will be a function of operational, financial and socio-economic characteristics. The correlation between variables is important, but it is not the only criteria for inclusion or removal of variables in the model. The highest correlation value was found between the ratings of total time and walking distance. The reason for this is that passengers spend a considerable time walking in the airport to get to the desired destinations (check-in counter, departure lounge, gate, concessions, services, etc.). There are also fairly strong correlations between the total time and check-in, and between total time and orientation. Passengers spend a long time in check-in counter lines at Sa˜o Paulo/Guarulhos international airport, especially the ones traveling overseas. Some surveyed passengers spent more than 60 min in the check-in counter alone, which may be one of the reasons for the correlation value (0.4) between these two variables. The correlation between total time and orientation may be explained by the fact that users who do not receive good orientation at an airport passenger terminal spend more time than necessary on wayfinding. These same ideas may explain the fairly strong relationship between walking distance and orientation (0.4). It is not convenient to include variables that have a high degree of correlation in a multiple regression model. In this case, it is appropriate to analyze the overall terminal LOS without including the total time variable. Nevertheless, this variable will be indirectly represented by many other variables (curbside, check-in counter, security screening, departure lounge, etc.), because a small share of the total time is included in each of these components. Two other variables require additional attention: circulation and security environment. These variables were originally intended to be part of the overall analysis, but were later removed from the evaluation. The respondent passengers found it very difficult to provide a LOS opinion about these two components, because the definitions of circulation and security environment were not clear to them. Besides, just a few passengers (if any) were able to evaluate the security environment of the airport. Most passengers had no security concerns at the airport. In this case, they had not ‘experienced’ the security environment at the airport at different levels, as they had experienced the waiting time at the check-in counter or the crowdedness at the departure lounge. The correlation values for the independent variables, after excluding total time, circulation, and security environment from the analysis, are presented in Table 3. All these variables will be employed in the regression analysis to obtain a composite measure, because the correlation between these variables is only moderate. 5.2. Composite evaluation The ratings of the variables are combined, according to the following equation: LOS ðoverallÞ ¼ w0 þ w1 LOS ðcurbÞ þ w2 LOSðcheck-inÞ þ w3 LOSðsec: sc:Þ þ w4 LOS ðloungeÞ þ w5 LOS ðwalking dist:Þ þ w6 LOS ðorientationÞ þ w7 LOS ðconcessionsÞ;
ð4Þ
Table 3 Correlation of variables at the overall level
Curbside Check-in Security screening Lounge Walking distance Orientation Concessions
Curb
Check-in
Sec. screen.
Dep. lounge
Walking distance
Orientation
Concessions
1.0 0.2 0.4 0.3 0.3 0.3 0.4
1.0 0.2 0.2 0.2 0.4 0.3
1.0 0.3 0.3 0.3 0.2
1.0 0.2 0.3 0.2
1.0 0.4 0.1
1.0 0.4
1.0
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
617
where LOS (overall) is overall terminal LOS ratings; LOS (curb), LOS (check-in), LOS (sec. screen.), LOS (lounge), LOS (walk. dist.), LOS (orientation) and LOS (concessions) is LOS ratings for each individual attributes; w0 is intercept; and w1, w2, w3, w4, w5, w6, and w7 is parameters of the equation. Substituting the LOS ratings of the above equation by the responses (1–5) of the survey conducted at Sa˜o Paulo/Guarulhos international airport and performing a regression analysis will provide the values w1, w2, w3, w4, w5, w6, and w7 as the parameters of the regression equation, and w0 as the intercept. In this case, the weights are obtained, reflecting the user perceptions of the relative importance of components. The results of the regression analysis are shown in Table 4. It can be noticed from Table 4 that the walking distance parameter has a 94.5% probability of being equal to zero. This motivates us to remove this variable from the analysis. Although this is a very important factor for the overall terminal evaluation, it seems that passengers at Sa˜o Paulo/Guarulhos international airport do not value this factor. This regression analysis includes 78 observations. The 40 observations made in the first survey performed in the summer of 2003 are not included because walking distance, total time, and security environment were not present in those questionnaires. The subsequent calculations in this chapter will now include these 40 observations, because the absent variables are no longer objects of study for the overall LOS. New regression analyses have been performed with the following changes: • Non-inclusion of the walking distance variable. • Addition of the 40 observations made during the Summer 2003 survey. • Removing (stepwise) the variables with high P-values. The results of the best-fit regression analysis are presented in Table 5. The security screening and concessions variables presented high P-values. This means they have a high chance of being equal to zero. In addition, security screening had a negative value. This was expected. The security screening process at Sa˜o Paulo/
Table 4 Parameters, standard error, t Stat, and P-value – original variables Component
Parameters
Standard error
t Stat
P-value
Intercept Curbside Check-in Security screening Lounge Walking distance Orientation Concessions R2 = 0.48 F = 7.86 Observations: 78
0.832 0.359 0.019 0.099 0.117 0.006 0.196 0.027
0.533 0.102 0.108 0.095 0.077 0.083 0.090 0.103
1.562 3.518 0.180 1.042 1.522 0.069 2.191 0.261
0.123 0.001 0.857 0.301 0.133 0.945 0.032 0.795
Table 5 Parameters, standard error, t Stat, and P-value – best-fit regression analysis Component
Parameters
Standard error
t Stat
P-value
Intercept Curbside Check-in Lounge Orientation Purpose R2 = 0.443 F = 17.799 Observations: 118
0.755 0.313 0.114 0.118 0.238 0.243
0.358 0.070 0.075 0.062 0.068 0.103
2.108 4.504 1.513 1.889 3.517 2.365
0.037 0.000 0.133 0.061 0.001 0.020
618
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
Table 6 Parameters, Standard error, t Stat, and P-value – final model Component
Parameters
Standard error
t Stat
P-value
Intercept Curbside Check-in Lounge Orientation Purpose R2 = 0.47 F = 19.538 Observations: 116
0.841 0.246 0.144 0.151 0.229 0.214
0.327 0.065 0.069 0.057 0.063 0.094
2.575 3.809 2.094 2.643 3.656 2.291
0.011 0.000 0.039 0.009 0.001 0.024
Guarulhos international airport is very fast and does not cause any concern to passengers. The concessions component is not mandatory, and some users did not have a strong opinion about its level of service because they had not experienced it. One new variable was included in the analysis, because it fitted well in the regression equation. This was the ‘purpose’ variable; it was included as a dummy variable. It was equal to 0 when the trip purpose was business or combined business/non-business, and equal to 1 when the trip purpose was non-business. The final step in refining the analysis was the detection of possible outliers. The outliers were defined (in this analysis) as responses of passengers that were clearly inconsistent. It might be a passenger that evaluated the overall terminal as excellent, but evaluated all (or most) components as poor or unacceptable. The opposite can also occur: one passenger can evaluate the terminal as poor, but all the components as excellent. Two cases (out of 118 passengers) presented this inconsistency and were removed from the analysis. The results of the regression analysis without these two outliers are presented in Table 6. The parameters of Table 6 can be substituted into Eq. (4) to provide the following relation: LOS ðoverallÞ ¼ 0:841 þ 0:246 LOS ðcurbÞ þ 0:144 LOSðcheck-inÞ þ þ0:151 LOSðloungeÞ þ 0:229 LOS ðorientationÞ þ 0:214ðpurposeÞ:
ð5Þ
According to Eq. (5), the most important component for passengers is the curbside. Although that seems contrary to accepted practice, the reason may be that it was the first component experienced by the 118 surveyed passengers and that curbside is a problematic aspect of Sa˜o Paulo/Guarulhos international airport. The curbside was the first impression they had about the airport and, as such, affected the evaluation of the terminal as a whole. Orientation is the second most important component. The relatively low importance of the check-in and the departure lounge components, when compared to the curbside and orientation components, can be explained by the fact that, at Sa˜o Paulo/Guarulhos international airport, passengers are accustomed to spending a considerable time in these two components. The intercept (0.841) indicates that other variables may be included in this analysis. It signifies that there are one or more components of the overall LOS that are not represented by the explanatory variables included in the model. Future applications of the model may need to test other variables for which data was not collected as part of this research. Although the R2 is relatively moderate (0.47), it is common for estimating this type of model, where several variables are available and subjective assessments are collected for different types of passengers. Nevertheless, despite all these shortcomings, the model was able to explain almost 50% of the variances. 6. Conclusions This research provides a methodology for estimating a global index for LOS evaluations at airport passenger terminals. The main contribution to the field is the identification of the most important airport attributes according to user perceptions. This methodology was applied to Sa˜o Paulo/Guarulhos international airport and helped identify the passenger terminal curbside as one of the most important factors contributing to the passengers’ perception of level of service. The results of the model application also indicate that other
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
619
passenger terminal attributes not tested in this research may have a significant contribution to the passengers’ perceptions of LOS. Airport planners could use this method in practice for forecasting the overall LOS of future facility alternatives. Estimates of LOS for individual components (check-in, lounge, etc.) can be obtained from conventional methods and standards such as those provided by IATA (International Air Transport Association, 2004); however they do not indicate nor provide the overall LOS of an airport terminal. Eq. (5), if calibrated according to the characteristics of its passengers, could provide the overall LOS of each alternative. It also provides a data based method for comparing airports with respect to the level of service provided to passengers. Airport managers can also use the results of this method to obtain the degree of importance that users assign to various components and characteristics of the airport terminal. These weights are useful, as it will be one indication of where airport managers should invest their limited resources. For future research, this methodology can be applied to other airports to provide a more comprehensive understanding of the relationship between overall terminal measures and the level of service associated with them. Other variables not surveyed in this research should also be tested to improve the model’s explanatory power. Acknowledgements This research was supported in part by CAPES and FAPESP – agencies of the government of Brazil providing financial support to the training of highly skilled personnel, by the Natural Sciences and Engineering Research Council of Canada and by the University of Calgary. References Aaker, D., Kumar, V., Day, G., 1998. Marketing Research, 6th ed. John Wiley and Sons, New York. Airports Council International, 2000. Quality of Service at Airports: Standards and Measurements. ACI World Headquarters, Geneva. Bandara, S., Wirasinghe, S.C., 1992. Walking distance minimization for airport terminal configurations. Transportation Research A 22, 59–74. Correia, A.R., 2000. Quantitative Evaluation of Airport Passenger Terminal Configurations. M.Sc. dissertation, Aeronautical Institute of Technology, Brazil (in Portuguese). Correia, A.R., Wirasinghe, S.C., 2004. Evaluation of Level of Service at Airport Passenger Terminals: A Review of Research Approaches. Transportation Research Record 1888, National Research Council, Washington DC. pp. 1–6. Correia, A.R., Wirasinghe, S.C., de Barros, A.G., 2005. A New Procedure for Overall Level of Service Data Collection at Airport Passenger Terminals, in: Proceedings of the 9th ATRS World Conference, CD-ROM, Rio de Janeiro. Dada, E.S., Wirasinghe, S.C., 1999. Development of a New Orientation Index for Airport Terminals. Transportation Research Record 1662, National Research Council, Washington DC. pp. 41–47. Dada, E.S., Wirasinghe, S.C., 2002. Estimating Basic Level of Service for Passenger Orientation in New Airports. In: D. Wang and S.-M. Li (Eds), Proceedings of the 7th Conference of Hong Kong Society for Transportation Studies, Hong Kong. pp. 369–377. de Barros, A.G., Wirasinghe, S.C., 2002. Design and shared use of departure lounge for NLA operations. Journal of Advanced Transportation 36, 187–209. de Barros, A.G., Wirasinghe, S.C., 2003. Optimal terminal configurations for new large aircraft operations. Transportation Research A 37, 315–331. de Barros, A.G., Somasundaraswaran, A.K., Wirasinghe, S.C., 2006. Evaluation of level of service for transfer passengers at airports, in: Proceedings of the 10th ATRS World Conference, CD-ROM, Nagoya. de Neufville, R., de Barros, A.G., Belin, S.C., 2002. Optimal configuration of airport passenger buildings for travelers. Journal of Transportation Engineering 128, 211–217. Fernandes, E., Pacheco, R.R., 2002. Efficient use of airport capacity. Transportation Research Part A 36, 225–238. Hart, W., 1985. The Airport Passenger Terminal, 1st ed. John Wiley and Sons, New York. Horonjeff, R., McKelvey, F.X., 1994. Planning and Design of Airports, 4th ed. McGraw-Hill, New York. International Air Transport Association, 2004. Airport Development Reference Manual, 9th edition. Montreal. Keeney, R., Raiffa, H., 1976. Decisions with Multiples Objectives, 1st ed. John Wiley and Sons, New York. Magri, A.A., Alves, C.J.P., 2003. Convenient Airports: Point of View of the Passengers. ATRS – Air Transport Research Society World Conference, CD-ROM, Toulouse. Martel, N., Seneviratne, P.N.,1990. Analysis of Factors Influencing Quality of Service in Passenger Terminal Buildings. Transportation Research Record 1273, National Research Council, Washington DC. pp. 1–10. Miles, J., Shelvin, M., 2001. Applying Regression and Correlation, 1st ed. Sage, Publications, London.
620
A.R. Correia et al. / Transportation Research Part E 44 (2008) 607–620
Mu¨ller, C., Gosling, G.D., 1991. A framework for evaluating level of service for airport terminals. Transportation Planning and Technology 16, 45–61. Mumayiz, S.A., Ashford, N., 1986. Methodology for Planning and Operations Management of Airport Terminal Facilities. Transportation Research Record 1094, National Research Council, Washington DC. pp. 24–35. Ndoh, N.N., Ashford, N.J., 1993. Evaluation of Airport Access Level of Service. Transportation Research Record 1423, National Research Council, Washington DC. pp. 34–39. Omer, K.F., Khan, A.M., 1988. Airport Landside Level of Service Estimation: Utility Theoretic Approach. Transportation Research Record 1199, National Research Council, Washington DC. pp. 33–40. Park, Y.H., 1994. An evaluation methodology for the level of service at the airport landside system. Ph.D. thesis, Department of Transport Technology, Loughborough University of Technology, UK. Rand Corporation, 1969. Measurement and Evaluation of Transportation System Effectiveness. Memorandum RM-5869-DOT. Seneviratne, P.N., Martel, N., 1991. Variables influencing performance of air terminal buildings. Transportation Planning and Technology 16, 1177–1179. Siddiqui, M.R., 1994. A Statistical Analysis of the Factors Influencing the Level of Service of Airport Terminal Curbsides. M.Sc. thesis, Concordia University, Canada. Taha, H.A., 1997. Operations Research – An Introduction, 6th ed. Prentice Hall, Upper Saddle River, NJ. Taylor III, B.W., 1999. Introduction to Management Science, 6th ed. Prentice Hall, Upper Saddle River, NJ. Transportation Research Board, 1987. Special Report 215: Measuring Airport Landside Capacity. National Research Council, Washington DC. Transport Canada, 1979. A Discussion Paper on Level of Service Definition and Methodology for Calculating Airport Capacity. Report TP 2027, Ottawa. Wirasinghe, S.C., Shehata, M., 1988. Departure lounge sizing and optimal seating capacity for a given aircraft/flight mix – (i) Single Gate; (ii) Several Gates. Transportation Planning and Technology 13, 57–71. Yen, J.R., 1995. A new approach to measure the level of service of procedures in the airport landside. Transportation Planning Journal 24, 323–336. Yen, J.R., Teng, C.R., Chen, P.S., 2001. Measuring the Level of Service at Airport Passenger Terminals: Comparison of Perceived and Observed Time. Transportation Research Record 1744, National Research Council, Washington DC. pp. 17–23.