Making the Business Case for Sustainability Related Investments Through a Single Financial Metric

Transcription

1 See discussions, stats, and author profiles for this publication at: Making the Business Case for Sustainability Related Investments Through a Single Financial Metric Chapter June 2012 DOI: / _13 CITATIONS 0 READS 67 1 author: Mark Ferguson University of South Carolina 87 PUBLICATIONS 1,899 CITATIONS SEE PROFILE All content following this page was uploaded by Mark Ferguson on 19 May The user has requested enhancement of the downloaded file.

2 Measuring the Benefit of Offering Auxiliary Services: Do Bag-Checkers Differ in Their Sensitivities to Airline Itinerary Attributes? Mariana Nicolae College of Business, Eastern Michigan University, Ypsilanti, MI Phone: ; Fax: Mark Ferguson Moore School of Business, University of South Carolina, Columbia, SC Phone: ; Fax: Laurie A. Garrow School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA Phone: ; Fax: When firms evaluate their service system design choices, there is typically more uncertainty surrounding the value that a particular auxiliary service provides than there is on the cost of providing that service. To help inform this decision, we propose an approach where we compare the relative value of the segment of passengers who use an auxiliary service to the relative value of the segment that does not use it. We demonstrate this approach for a typical auxiliary service common to the airline industry. In 2008, most U.S. airlines implemented checked baggage fee policies to decrease their costs by reducing the number of customer service agents needed in the check-in and baggage handling processes. The success of this change has led to a current debate at many of these airlines on whether to make further staffing cuts in these areas, essentially making it even less attractive for passengers to check their baggage. Our proposed methodology helps answer whether passengers who continue to check bags in today s baggage-fee era are more or less valuable than passengers who do not check bags. We explore this question empirically by examining, through a stated preference survey, if a history of checking or not checking bags can be used to segment passengers based on how their itinerary choices are influenced by common airline service attributes (on-time performance, itinerary time, number of connections, airfare, and schedule delay). Contrary to the opinions of some top airline executives, we find that the passengers who continue to check bags at airlines that charge baggage fees are generally less sensitive to differences in three of these important service attributes and are less likely to switch airlines when a competing airline improves its offerings along these dimensions. Thus, airlines that charge for checked bags should consider improving the customer experience for their bag-checking passengers, as they represent a potentially more valuable segment class to the airline. Key words : baggage fees; stated preference; airline passenger preferences; itinerary choice; airline price sensitivities History : Received: June 2015; accepted: March 2016 by Nitin Joglekar after 1 revision. 1

3 2 Nicolae, Ferguson and Garrow: Article Short Title 1. Introduction The operations management (OM) literature has a rich history of helping managers make decisions about adding to or subtracting from their product or service portfolios (see Lee 1996 for an example on the product side and Victorino et al for an example on the service side). There has been less work, however, on how to estimate the value that an existing (but operationally expensive) auxiliary service provides that is already included in the main product/service bundle. Examples of operationally expensive auxiliary services include technical support call staff for products, local branches for banking, and room service for higher-end hotels. As Frei (2006) points out, service firms must make trade-offs between what auxiliary services they choose to excel at and which ones they concede a leadership role to their competitors. In the absence of strategically making these trade-off decisions, service firms try to do everything well, which rarely leads to market success. Although operations managers are typically well trained in estimating the cost involved in offering auxiliary services, less is known about how to estimate the benefits an auxiliary service provides. This makes the calculation of cost-benefit trade-offs difficult. Common benefit-focused methodologies, such as customer satisfaction surveys or online customer reviews, often make it difficult to disentangle the value that a particular auxiliary service provides. In this study, we demonstrate one potential way of estimating the benefit of an operationally costly auxiliary service by empirically testing whether the segment of customers who primarily use that auxiliary service is more or less valuable than the segment that does not. For our study, we choose a familiar auxiliary service that has been the subject of much heated debate over the last decade, i.e., the checked bag service provided by airlines. The airline industry is a particularly good choice to demonstrate a methodology that estimates the potential value of passengers who utilize an auxiliary service because airline travel has frequently been identified as a commodity-type service with little brand loyalty among the customer base. In addition, the industry has recently gone through a dramatic restructuring, affecting both the service concept and the service delivery system design choices of the service strategy triad (Roth and Menor 2003). In 2008, oil prices increased to more than $130/barrel (CNN 2010) and the global economic crisis hit. Prior to this crisis, most of the major U.S. airlines had implemented many cost-cutting and revenue-generating measures during the early 2000s as part of their bankruptcy restructuring and merger processes. For competitive reasons, airlines could not raise fares to a level that could overcome the perfect storm that emerged in 2008: soaring fuel costs were followed immediately by plummeting demand. Consequently, 2009 proved to be the worst year on record for U.S. airlines, in terms of year-over-year revenue declines (Southwest Airlines 2009).

4 Nicolae, Ferguson and Garrow: Article Short Title 3 U.S. airlines responded to this crisis by introducing new ancillary fees (e.g., charging for previously complementary services such as checking up to two pieces of luggage) and increasing existing ancillary fees (see Garrow et al for a review of product debundling trends). Prior to implementing these new baggage fees, in 2007 the U.S. airlines collected only $467M in baggage fees, mostly from the excessive weight penalties and the checking of more than the two complementary bags. This revenue stream increased to $1.1B in 2008, $2.7B in 2009, and has held steady at $3.3- $3.5B during (Bureau of Transportation Statistics 2015). Thus, the implementation of baggage and other ancillary fees helped the U.S. airline industry return to profitability, e.g., in 2012 the industry collected $3.5B in baggage fees, which accounted for almost half of the industry s total profit of $7B (Bureau of Transportation Statistics 2015). Shortly after carriers implemented baggage fees, they realized several benefits. In addition to the additional revenue generation, these fees altered consumer behavior in a manner that has been beneficial to the airlines costs (e.g., see Allon et al. 2011). A report based on interviews with airline officials conducted by the U.S. Government Accountability Office (2010) has found that after the implementation of baggage fees, the number of checked bags declined by 50% for one airline and 40% for a second airline. The decrease in checked baggage has enabled airlines to decrease fuel costs and/or increase revenue though transporting more cargo. The airlines have also been able to reduce the number of check-in and baggage-handling staff. Passengers who continue to check baggage have also benefited through a reduction in mishandled baggage rates, which declined 40% (from 7 per 1,000 passengers to less than 4 per 1,000 passengers) from 2007 through 2009 (U.S. Government Accountability Office 2010). Airlines are currently debating whether to implement additional reductions in check-in and baggage handling staff. However, a priori, the benefits of implementing additional reductions is not clear. The more passengers who check bags and pay for this service will increase the airline s revenues, but will also increase the airline s costs. Thus, it is unclear from a profitability standpoint whether (after the additional changes observed by successfully implementing baggage fees) an airline will be better or worse off from having more passengers use its baggage checking system. One benefit of implementing further cuts is that staff reductions in this area would make it even less attractive for passengers to check their bags, possibly leading to a further decrease in the number of checked bags and possible additional decreases in the mishandled baggage rates. Conversely, staff reductions could lead to an increase in the number of bag-checking passengers switching to competitor airlines. Complicating this debate is the fact that although airlines are knowledgeable about the cost side of offering a checked baggage service, much less is known about the benefits of offering this service.

5 4 Nicolae, Ferguson and Garrow: Article Short Title In order for those airlines that implemented baggage fees to determine whether they should change their operations policies and spend more (or less) resources to make it easier (or harder) for passengers to check bags, they need to understand more about the benefits provided by this service to their passengers who are currently checking bags. If these passengers are primarily pricesensitive leisure customers who travel infrequently and are not loyal to a particular airline, then any potential loss in this segment of passengers will be more than offset by the cost savings achieved from reductions in the staff supporting the checked baggage process. Stated another way, airlines that charge baggage fees would not mind losing these least valuable passengers to airlines that either do not charge baggage fees or provide better baggage checking services. Conversely, if these passengers represent more loyal and price-inelastic business customers, then reductions in staffing levels could lead to significant revenue losses if passengers choose to switch to other airlines with better baggage checking services. If those passengers who were the most price-sensitive already switched to airlines that did not charge baggage fees when baggage fees were first implemented, then those passengers who remain and continue to check bags may be more valuable than the airline s passengers who do not check bags. To help inform this debate, airlines need to understand how passengers make trade-offs among products across competitors when purchasing an itinerary. Our contribution to the operations literature is that we provide a methodology for helping to explore the benefit that the bag checking process provides. One potential way this benefit can be estimated is by using revealed preference (RP) data of passenger purchase transaction history. The estimation of the benefit through choice models using RP data requires a historical record of the alternatives viewed or considered by passengers when they made their purchase decision. Alternatively, this problem can be addressed using stated preference (SP) data that survey individuals and ask them to select their preferred alternative from a set of two or more alternatives. Both SP and RP data allow the researcher to estimate the airline itinerary choice behavior of the group of passengers who frequently check bags, and assess whether these passengers are statistically different in their itinerary purchasing behavior from those who do not check bags. Capturing RP data, however, has historically been challenging to do in the airline industry because the airlines systems are not designed to capture the other itinerary alternatives that were available when passengers made their purchase decisions. In addition, the airlines typically only have RP data for their own passengers, resulting in a biased sample for drawing market level conclusions. For these reasons, it is most common to use SP data to answer questions such as the one posited in this study. Thus, SP data can help resolve the debate as to how valuable passengers who check bags are, helping inform the decision on how airlines should adjust the staffing and capacity levels of their passenger check-in and baggage handling processes.

6 Nicolae, Ferguson and Garrow: Article Short Title 5 To answer this question empirically, we use SP data from an internet-based survey of 878 domestic U.S. passengers that was conducted by Resource Systems Group, Inc. (RSG) in Although the survey was not specifically designed to answer our research question, it is part of a series of periodical surveys conducted at considerable expense (over $20,000 per data collection effort) and has been extensively vetted by other researchers (as described in the Literature Review section). The data that were used in this study came from a nationally-representative sample of air passengers using a panel maintained by an independent survey sampling provider. The itinerary information is self-reported based on the respondent s most recent paid domestic air trip. Respondents answer a series of SP trade-off questions (price, number of connections, etc.) which are tailored to their most recent paid domestic trip (see Section 4). The survey also asks respondents whether they checked bags on their last paid domestic U.S. flight and, if so, whether they paid extra for this service. Accordingly, we group passengers into baggage checkers and non-baggage checkers segments and differentiate between baggage checkers who pay bag fees and those who do not pay bag fees. Our findings indicate that the segment of passengers who continue to check bags and pay for this service is not the least valuable passenger segment that was conjectured by one airline executive we talked to before conducting the study. In contrast, this bag-checking passenger segment appears to be less price sensitive as well as less sensitive to itinerary time and the number of connections than the segment of passengers who check bags and do not pay fees and the segment of passengers who do not check bags. Thus, the bag-checking passenger segment who pays for this service appears to be more loyal to a given airline than the non bag-checking segment or the bag-checking segment that does not pay fees when a competitor improves its attractiveness along one or more of these dimensions. This finding may have already influenced one airline to improve its performance in the baggage-handling process, as discussed in the Conclusions section. One of the theoretical contributions of this paper is that it is among the first studies in the OM literature that explores customer choice using discrete choice models that control for noncompensatory behavior (e.g., in our setting we account for passengers who do not make tradeoffs among alternatives but rather always select the lowest-priced itinerary). Accounting for this behavior, we investigate how passenger preferences for on-time performance, itinerary time, number of connections, airfare, and schedule delay differ among those passengers who check bags and do not pay fees, those passengers who check bags and pay fees, and those passengers who do not check bags. Our paper contributes to the literature by providing evidence that those passengers who check bags and pay bag fees represent a potentially valuable passenger segment, and that airlines should carefully consider making any additional operational changes that could impact the service experience of this group. Thus, our findings provide guidance for key airline staffing level decisions

7 6 Nicolae, Ferguson and Garrow: Article Short Title and capacity adjustments to the baggage checking process. We also suggest a potential new way that airlines can segment passengers, i.e., based on their baggage checking behavior. The remainder of this study is organized as follows. In Section 2 we review the relevant literature on air itinerary choice and motivate the factors we include in the hypotheses tested in this study, described in Section 3. In Sections 4 and 5 we describe our data and methodology, respectively. In Section 6 we present our results and discuss the implications of these results for practice. We conclude in Section 7 with a discussion of limitations of the study and directions for future research. 2. Literature Review The identification of higher value customer segments has received extensive research effort (e.g., see Bodea and Ferguson 2014 for a recent review). For the most part, however, the research has focused on determining the demographic and/or psychographic attributes of a customer segment so that higher valued products or services can be marketed to the segments with the corresponding highest utilities and ability to pay. For auxiliary services such as the airline baggage checking service, although the airlines can tailor their pricing for this service to different segments, the actual service process requires that all passenger segments use the same resources (once bags are checked, they are all treated the same). Thus, the identification of passenger segments is different in our problem versus in the traditional segmentation research in that passenger attribute data is not very useful since an airline is unlikely to provide separate bag-checking processes for different segments (or chooses not to provide different bag-checking processes). Instead, our aim is to estimate the relative value of passengers who self-select to use the bag-checking process. In some ways, our work is similar to previous empirical work exploring the customer efficiency concept by Xue and Harker (2002), where a more efficient customer is defined as one who consumes less of a resource for the same amount of output. An example from this literature is Xue et al. (2007), who use econometric models to determine if bank customers who self-select to use less labor intensive channels (e.g., customers who primarily use ATMs) are more or less valuable to the bank than customers who do not. Such studies require some heterogeneous measure of customer financial value, such as the amount of deposits in non-interest bearing accounts from which some dependent variable of interest, such as choice of channel, can be regressed against. This type of historical data may be difficult and time consuming to obtain by the firm or researcher, if it is available at all. In contrast, SP survey data can be collected over a short period of time through well-established instruments. Despite its popularity in other disciplines, the analysis of SP survey data using customer choice models (CCM) is still rare in the OM literature. Some of the first OM papers to use this methodology include Pullman et al. (2001) and Verma et al. (2001). A summary of papers that use SP/CCM

8 Nicolae, Ferguson and Garrow: Article Short Title 7 to access customer choice behavior on trade-offs along the product-service interface in the hospitality industry can be found in Verma (2010). To our knowledge, however, there has been no SP/CCM paper in the OM literature that addresses how customers choose among air travel options. Thus, we turn to the related research from the transportation discipline. Table 1 Review of Selected SP/CCM Studies from the Transportation Literature Study Market(s) Variables and Controls Segments Airfare, schedule delay with respect to departure time Proussaloglou and Koppelman (1999) Wen and Lai (2010) Collins et al. (2012) Espino et al. (2008) Pereira et al. (2007) Brey and Walker (2011) Garrow et al. (2007) Chicago- Denver and Dallas-Denver Taipei-Tokyo and Taipei-Hong Kong Sydney- London and Sydney-Paris Gran Canaria - Madrid Madeira and Oporto Continental U.S. markets Continental U.S. markets FF membership, carrier market presence, respondent s perception of carriers quality of service, airline, fare class (captures cabin and ticket restrictions) Airfare, schedule delay with respect to departure time, flight frequency, on-time performance Airline, check-in service, cabin crew service, leg room Airfare, arrival time, itinerary time, number of connections FF membership, carbon tax, airline, legroom, ability to select seat at time of booking, entertainment system (shared/personal), exchange fee Airfare, flight frequency, on-time performance Exchange fee, onboard food offerings, leg room Airfare, schedule delay with respect to departure time, itinerary time, connection indicator Multiple carriers on itinerary, legroom, departure time of day (TOD) Airfare, schedule delay with respect to departure time, itinerary time, connection on same or different airline Distance, household income, length of stay, booking days from departure Segmentation of schedule delay and fare by business/leisure Segmentation includes income, gender, age, party size (1,2+), trip purpose (business/leisure) None Segmentation of reliability by business/leisure trip purpose; segmentation of food offerings by cabin (first/business vs. coach) TOD segmented by party size (1,2+), direction and number of time zones traveled; only estimated leisure travel segment Price segmented by leisure and self-pay business vs. reimbursed business Table 1 provides an overview of selected SP/CCM studies from the transportation literature and

9 8 Nicolae, Ferguson and Garrow: Article Short Title Table 2 reviews studies that have explicitly used RSG s SP air travel survey (the same survey data we use in our study). The tables summarize the variables and controls included in the reported choice model along with which markets were included in the database and segmentations included in the model. The studies based on the RSG surveys included U.S. markets; thus the table shows the year the data was collected instead of the markets. Controls included in the model for purely methodological reasons (e.g., a scale parameter to account for combining revealed preference and stated preference data) are omitted from the tables. Several factors are common across the studies. Airfare is included in all studies. Itinerary time (defined as the time to travel from the time the first leg departs at the origin to the time the last leg arrives at the destination) and the number of connections (or indicator for one or more connections) is included in all studies that include a large number of markets. That is, for those studies in which only one or two markets were included, it may not have been possible to include itinerary times or the number of connections due to a lack of variation in these variables across the markets. For example, the Proussaloglou and Koppelman (1999) study focused only on nonstop flights between the Chicago-Denver and Dallas-Denver markets; the itinerary times for the first market (which included seven nonstop flights) differed by at most 28 minutes. The itinerary times for the second market (which included 12 nonstop flights) differed by at most 17 minutes. Thus, itinerary times and the number of connections could not be included in the Proussaloglou and Koppelman study. Ten of the 13 studies shown in Tables 1 and 2 contain a measure of departure and/or arrival time preference. Two of the studies that do not include this measure (Hess 2007 and Hess 2008) were more focused on methodological concepts; for these papers the author notes the specification is by no means complete in terms of the attributes included (Hess 2007). During a SP survey, it is possible (and common) to ask respondents what their preferred departure and/or arrival times were for a recent air trip and/or what these preferred times would be for a future hypothetical trip. Schedule delay functions are then used to model the disutility associated with the difference between the passengers preferred departure (arrival) time and the scheduled departure (arrival) time of the itinerary. Brey and Walker (2011) provide an excellent review of how schedule delay functions have been used in the transportation literature, and the aviation literature in particular. Small (1982) was the first to obtain empirical estimates of schedule delay (his focus was on commuters decisions of when to travel to work). His methodology incorporated different per-minute penalties for arriving early versus late to work and a one-time penalty for arriving late. Within the aviation studies shown in Tables 1 and 2, the majority use a schedule delay function that incorporates different slopes for itineraries that depart earlier or later than the respondent s preferred departure (or arrival) time. The majority of the studies shown in the tables that do not incorporate a schedule

10 Nicolae, Ferguson and Garrow: Article Short Title 9 Study Hess et al. (2007) Warburg et al. (2006) Adler et al. (2005) Hess (2007); Hess (2008) Theis et al. (2006) This study Table 2 Survey Variables and Controls Year Airfare, schedule delay with respect to arrival times, itinerary time, connection indicator, on-time performance Representative Studies based on RSG s SP Air Travel Surveys Preferred airport, equipment type, FF membership and level, indicator airport is closest to home; various interaction terms Airfare, schedule delay with respect to arrival time, itinerary time, number of connections, on-time performance Preferred airlines, preferred airports, equipment type, FF membership and level, access time Airfare, schedule delay with respect to arrival time, itinerary time, number of connections, on-time performance Preferred airlines, preferred airports, equipment type, access/egress time Airfare, itinerary time, on-time performance, number of connections FF membership and level, access time, indicator airport is nearest respondents home Airfare, itinerary time, number of connections, on-time performance Preferred airlines, preferred airports, FF membership and level, access time, night departure indicator, connection times Airfare, schedule delay with respect to departure time, itinerary time, number of connections, on-time performance Segments Business, holiday travel, visiting family and friends Extensive demographic/trip interaction terms with multiple factors including gender, party size (1,2+), income, employment type, vehicle ownership, day of travel, length of stay, whether passenger checked bag, whether passenger paid for trip Business vs. leisure None None Bag-checkers who pay bag fees, bag-checkers who do not pay bag fees, non-bag checkers Preferred airlines, FF membership and level

11 10 Nicolae, Ferguson and Garrow: Article Short Title delay function include other measures of departure and/or arrival time preferences. For example, Collins et al. (2012) include arrival time preferences and Espino et al. (2008) and Pereira et al. (2007) include flight frequency. Flight frequency can be viewed as a (weak) proxy for schedule delay, i.e., as the number of flights increases, the expected difference between a passenger s desired and scheduled flights times decreases. The majority of studies shown in Tables 1 and 2 also include a measure of schedule reliability (referred to as on-time performance in the tables). Information about on-time performance was available and included in online screen displays from at least For example, an article in Computerworld discusses flight delays and notes that both Expedia and Travelocity had tools that tracked on-time performance of selected flights (Meehan 2000) and an article in Travel Weekly notes that Expedia was listing on-time performance in its display screens (Cogswell 2000). All of the studies based on RSG SP survey data represent on-time performance as a percentage of flights that arrive on-time at the final destination. Other measures of reliability have been used in the literature. For example, Espino et al. (2008) and Pereira et al. (2007) include performance guarantees in their survey (these guarantees include no compensation, a free ticket for the same trip, and full reimbursement of the original ticket). The first line under the variables and controls headings in the tables show those variables that are most common across the studies: airfare, itinerary time, connections, schedule delay and on-time performance. The second line shows the range of other factors (and controls) included in these studies. Many of these factors are specific to the underlying goals and/or scope of the survey. The majority of these additional factors are for onboard amenities and/or multi-airport choices (neither of which is the focus of our current study). The respondent s frequent flyer (FF) affiliation and/or membership level (e.g., basic, elite) is most often included as a control in the utility function. A wide range of passenger segmentations have been used or proposed in these studies. The most common segmentation is based on trip purpose (e.g., business versus leisure). Other segmentations are typically based on the underlying goals of the study. For example, Brey and Walker (2011) were focused on estimating time-of-day preferences, and thus interacted socio-demographic and trip-characteristics with time-of-day factors. Similarly, the goal of Warburg et al. (2006) was to understand how itinerary preferences varied along a wide range of socio-economic and trip characteristics so it should not be surprising that a large number of interaction terms were included in their model. The studies shown in Tables 1 and 2 informed our methodology, specifically the factors we included in our utility function, the segments we used, and our research hypotheses. The previous papers identify the key drivers behind passenger choice of air travel itineraries. We use five of the

12 Nicolae, Ferguson and Garrow: Article Short Title 11 most commonly identified drivers (on-time performance, itinerary time, number of connections, airfare, and schedule delay) to test whether passengers can be divided into segments with different substitution behaviors based on whether they commonly check bags or not (and whether they pay baggage fees). Next, we explain why new segmentation classes are needed and describe the previous literature that comes the closest to answering this question. Recognizing that airlines are facing domestic changes in their business environment and in passenger behavior, Teichert et al. (2008) assert that class flown and trip purpose (i.e., business vs. leisure) have become obsolete ways of segmenting airline passengers since they cannot be determined in advance and can rarely be determined based on passengers historical buying behavior. In our attempt to find a new and more robust segmentation method, we use a utility function that includes key factors that influence itinerary choice and add to this literature by including the checking or non-checking of bags as a possible passenger segmentation scheme. Our segmentation further accounts for whether or not those who check bags also pay bag fees. More specifically, we explore whether the fact that passengers check bags but do not pay fees, check bags and do pay fees, or do not check bags identifies distinct passenger segments, similar to how trip purpose has been explored in previous studies. Note that although the SP data identify respondents as either business or leisure passengers, this classification is not useful for the purposes of our study, as it is not economical for an airline to create separate baggage checking processes for business and leisure passengers. Thus, our interest is in determining whether baggage checkers in general (either business or leisure travelers) exhibit different airline choice behavior than non-baggage checkers. Further, although the SP data provide information about the respondents frequent flyer status across multiple airlines, we do not use this information for segmentation purposes as airlines cannot (or choose not to) design different baggage checking system for their frequent versus non-frequent flyers. For these reasons, we only include frequent flyer status as a direct control in our models. 3. Hypotheses Development Multiple product and passenger attributes influence itinerary choice. The objective of our study is to investigate how passenger valuations of product attributes differ among baggage checkers and non-baggage checkers. We further differentiate between baggage checkers who pay bag fees and baggage checkers who do not pay fees. The conceptual framework shown in Figure 1 presents our hypotheses regarding how these three segments value five important product attributes: on-time performance, itinerary time, number of connections, airfare and schedule delay. The signs included in each attribute box indicate the expected influence of the attribute on itinerary choice, e.g., an increase in itinerary time is hypothesized to lead to a decrease in utility. The dotted lines represent

13 12 Nicolae, Ferguson and Garrow: Article Short Title interaction terms that capture differences in passenger segments, namely those passengers who check bags and do not pay bag fees, those who check bags and pay bag fees, and those that do not check bags. This section provides the motivation for each of our hypotheses. Figure 1 Conceptual Model Within the U.S., carriers that have at least 1% of total domestic scheduled-service passenger revenues are required to report on-time flight departure and arrival information to the U.S. Department of Transportation (DOT), that then publishes a rank of these carriers on-time performance which is widely reported by the media as the official metric of flights on-time performance. In ranking flight delays among carriers, the DOT uses the percentage of flights with delayed arrivals where any flight with an arrival delay of more than 15 minutes is classified as delayed and any flight with an arrival delay of 15 minutes or less is classified as on-time. Although the duration of the delay (above 15 minutes) is not factored in the on-time performance rankings, this metric can impact individuals perceptions and itinerary valuations. Suzuki (2000) suggests that passengers choice of airlines may be affected by the on-time arrival experiences of passengers. According to Mazzeo (2003, p. 277), the expected on-time performance is a key non-monetary component of an air traveler s utility function. Such a consumer would compare prices and expected on-time performance of the competing carriers on the route for which he or she was buying a ticket. To the extent that the consumers expectation of future delays are based on a carrier s past on-time performance on that route, one potential cost of flight delays for airlines is reduction in future

14 Nicolae, Ferguson and Garrow: Article Short Title 13 demand. Adler et al. (2005) confirm the on-time performance as an important selection criterion for air passengers based on a SP survey conducted in In more recent years, the historical on-time performance of flights has been made more transparent to passengers when they make their flight decisions through its availability on airlines and travel agencies websites. There is also some prior evidence that passengers who check bags value on-time performance differently than passengers who do not check bags. Based on a 2001 survey of business passengers, Warburg et al. (2006) find that business passengers who check bags are less time-sensitive and subsequently less impacted by on-time performance than business passengers who do no check bags. We expect this relationship to also hold for leisure passengers (our survey respondents include both business and leisure passengers although the survey does not explicitly distinguish between these two groups). In addition, we expect bag fees to make a difference between passengers who check bags and pay versus those who check bags but do not pay. Thus, we suggest the following hypotheses: Hypothesis 1A. The historical on-time performance of an itinerary has a positive impact on the utility of that itinerary. Hypothesis 1B. The historical on-time performance of an itinerary has a different impact on the utility of that itinerary for passengers who check bags and do not pay bag fees, versus passengers who check bags and pay bag fees, versus passengers who do not check bags. In addition to arriving on time, passengers prefer short-duration flights over long-duration flights on the same route. As previously mentioned, Warburg et al. (2006) find that business passengers who check bags are less time-sensitive than business passengers who do not check bags. They explain their finding through the additional time those business passengers are willing to spend at the origin airport checking bags and then retrieving them at the destination airport. We expect this finding to be true for the general population of air passengers as well. We also expect bag fees to make a difference between passengers who check bags and pay versus those who check bags but do not pay. Thus, the following hypotheses are suggested: Hypothesis 2A. The itinerary time has a negative impact on the utility of that itinerary. Hypothesis 2B. The itinerary time has a different impact on the utility of that itinerary for passengers who check bags and do not pay bag fees, versus passengers who check bags and pay bag fees, versus passengers who do not check bags. A priori, it is unclear whether bag checkers or non-bag checkers will exhibit a stronger preference for nonstop itineraries. With a few exceptions, most studies find that passengers prefer itineraries with fewer connections (e.g., Koppelman et al. 2008). As noted earlier, passengers who check bags

15 14 Nicolae, Ferguson and Garrow: Article Short Title are less time-sensitive, as they need to spend additional time at the origin airport checking their bags and at the destination airport waiting for and retrieving their bags. Compared to a nonstop itinerary, a connecting itinerary has a longer itinerary time. Thus, it can be argued that those passengers who check bags will have weaker preferences for nonstop flights than passengers who do not check bags. However, it is important to note that for passengers who check bags, connecting itineraries are associated with an increased risk of lost or delayed bags. For example, Graham et al. (2010) find evidence consistent with the belief that business passengers are more risk adverse and/or time sensitive on the outbound portion of their trips, and thus are less willing to select outbound connections due to the increased itinerary time and additional risk that checked baggage may be delayed or lost. Conversely, it can be argued that those passengers who check bags will have a stronger preference for nonstop itineraries than passengers who do not check bags. Based on these conflicting arguments no direction can be offered on how, at the aggregate level, passengers who check bags differ from those who do not check bags with regard to their connection preferences. In addition, we expect bag fees to make a difference between passengers who check bags and pay versus those who check bags but do not pay. Thus, we propose the following hypotheses: Hypothesis 3A. The number of connections of an itinerary has a negative impact on the utility of that itinerary. Hypothesis 3B. The number of connections of an itinerary has a different impact on the utility of that itinerary for passengers who check bags and do not pay bag fees, versus passengers who check bags and pay bag fees, versus passengers who do not check bags. Multiple studies have found that price plays a large role in passengers itinerary choices. Further, some studies have estimated price as a function of trip purpose (e.g., business versus leisure) and found some segments to be more price sensitive than others. One problem with these identified segments is that they are not predictive. That is, an airline can rarely determine whether a potential passenger is a business or leisure passenger. In contrast, a segmentation based on whether a passenger has historically checked bags is easily implementable. One of the questions that we attempt to answer is whether this segment of passengers is more or less sensitive to price differences among competing itineraries offered by different airlines. The previous literature provides little guidance on this question. Passengers who check bags may predominately represent infrequent leisure passengers who are less loyal to a particular airline and select an itinerary based on a small difference in price or other product attribute. In this context, a history of checking bags would be a distinguishing characteristic of a (more price sensitive) leisure passenger versus a (less price sensitive) business passenger. Conversely, it can be argued that the leisure passengers who continue to check bags in

16 Nicolae, Ferguson and Garrow: Article Short Title 15 today s baggage fee era (after 2008) represent a more loyal and price inelastic customers than the leisure passengers who no longer check bags. These opposite arguments cannot offer a direction on how, at the aggregate level, passengers who check bags differ from those who do not check bags with regard to their price sensitivities, and further, how passengers who check bags and pay fees differ from those who do not pay fees when checking bags. Thus, we suggest the following hypotheses: Hypothesis 4A. The airfare of an itinerary has a negative impact on the utility of that itinerary. Hypothesis 4B. The airfare of an itinerary has a different impact on the utility of that itinerary for passengers who check bags and do not pay bag fees, versus passengers who check bags and pay bag fees, versus passengers who do not check bags. Our final hypothesis involves passengers sensitivities to departure delay, a factor that was included in half of the previous studies discussed in the Literature Review section. We expect passengers sensitivities to departure delay to be similar to their sensitivities to on-time departure performance and itinerary time, as the latter also affect the probabilities that passengers will arrive at their destinations before some trip-specific preferred time. For example, passengers traveling to a meeting or a social event will typically want to arrive at their destinations before the event begins. Such passengers will have stronger preferences for offered departure times that are closest to their (stated) preferred departure times than passengers who are not traveling for a specific event. We also expect this effect may differ for offered itineraries that depart earlier/later than the passengers stated preferred times. The reasons for expecting a difference between earlier versus later departures is that some passengers may be more constrained on the earliest they are available for a departure (dropping kids off at school or other morning responsibilities) whereas passengers traveling for an event may prefer to arrive for the event early rather than late. A priori, we do not know which of these effects will be stronger. Thus, we test for both effects by separately including variables for differences when the departure times were earlier from the preferred departure time and differences when the departure times were later than the preferred time. To be consistent with previous studies, we use the terminology schedule delay with early departure and late departure to represent these effects. There are also reasons for expecting these effects to differ for passengers who check bags. For early departures, passengers who prefer to depart at a later time may be timepressed to arrive at the airport in time to check-in their bags, thus making them more sensitive to itineraries with earlier departure times. For late departures, passengers who check bags face the additional risk of being delayed for a potential event because of delays (or even lost luggage) in the baggage claim process so we expect passengers who check bags to be less sensitive to schedule

17 16 Nicolae, Ferguson and Garrow: Article Short Title delays than passengers who do not check bags. In addition, we expect bag fees to make a difference between passengers who check bags and pay versus passengers who check bags but do not pay. Thus, we propose the following hypotheses: Hypothesis 5A. The schedule delay has a negative impact on the utility of that itinerary. Hypothesis 5B. The schedule delay has a different impact on the utility of that itinerary for passengers who check bags and do not pay bag fees, versus passengers who check bags and pay bag fees, versus passengers who do not check bags. 4. Data The data for our study were provided by RSG. In Spring of 2012, RSG conducted an internetbased survey of passengers who in the previous six months had taken an airline trip that originated and/or terminated in the U.S. To be eligible for the survey, the trip must have been paid for by the individual or the company the individual worked for. Information about the respondent s most recent air trip and airline preferences was obtained directly from the respondent. This information includes the respondent s outbound origin and destination airports, number of connections, flight departure and arrival times, flight on-time performance, actual flight departure and arrival times, the respondent s preferred departure and arrival times, itinerary time defined as the total travel time for the air portion of the trip, operating airline(s), equipment type(s), class(es) of service (e.g., coach or first class), paid fare, the number of checked bags, the number of carry-on bags, and the total amount the passenger paid in baggage fees. Respondents were shown a list of 24 airlines and were asked to: (1) rank their top three preferred airlines and their least preferred airline; and, (2) indicate whether they were a member of each airline s frequent flyer program (and their status level if they were a member). Finally, respondents provided socio-economic information, e.g., age, gender, household income, and employment status. Eight choice scenarios were presented to each respondent. These scenarios were customized to each individual based on her most recent air trip and airline preferences. As shown in Figure 2, for each scenario individuals were asked to choose between two itinerary options. The attributes 1 and their corresponding levels used in the choice scenarios are summarized in Table 3. In the literature, departure time preferences are modeled either using the scheduled flight departure time or, if known, the difference between the passenger s preferred departure time and scheduled flight departure time (e.g., Parker and Walker 2006). The latter is often referred to as a schedule delay function. Schedule delay functions incorporate different preference weights for 1 Equipment type (widebody, standard jet, regional jet and propeller) was also included as an attribute on the survey; however, the equipment type was highly correlated with itinerary time (correlation based on Spearman nonparametric correlation test=0.5834); thus, we only included itinerary time in the analysis (reflected in Figure 2 and Table 3).

18 Nicolae, Ferguson and Garrow: Article Short Title 17 Figure 2 Example Choice Scenario Airline Flight departure time Flight arrival time Number of connections Itinerary time On-time performance One-way fare Table 3 Description of Itinerary Attributes and Levels The airline that markets and operates the outbound itinerary. Only one airline is associated with an itinerary. Passengers rank their top three preferred carriers and their least preferred (out of 24) carriers. These four airlines, which will differ across respondents, are included in the SP choice sets. The departure time of the first leg of the outbound itinerary, expressed in the local time of the departure airport. Departure time is modeled as minutes earlier or later than an individuals preferred departure time (the survey asked respondents at what time they would have most preferred to depart). The arrival time of the last leg of the outbound itinerary, expressed in the local time of the arrival airport. The number of connections associated with an outbound itinerary. A nonstop itinerary has zero connections. One-stop and two-stop connections are also included in the SP survey. A single connection adds 90 minutes to the itinerary time. Double connections are only included where the nonstop flight time to serve an origin and destination pair is greater than four hours. Double connections add either 90 minutes or 150 minutes to the itinerary time. Itinerary time (shown as total travel time in Figure 2 is the total time the passenger travels by air, defined as the arrival time of the last leg of the outbound itinerary minus the departure time of the first leg of the outbound itinerary where the arrival and departure times are expressed in Coordinated Universal Time (UTC) to properly account for time zone changes. The percentage of time the flight legs associated with the itinerary arrive on-time at their destination airports. Four levels are included in the SP survey: 60%, 70%, 80%, and 90%. One-way fare (inclusive of taxes). Four levels are included in the SP survey based on the fare the individual recalled paying for the last itinerary. These levels are 40% lower, 20% lower, 20% higher, and 40% higher than the previous fare. flights that depart earlier versus later than the individual s preferred departure time. Intuitively, we expect itineraries that depart either earlier or later than the individual s preferred departure time