Passenger Traffic Implications of Alliance
Constellations in the Airline Industry
Passenger Traffic Implications of Alliance
Constellations in the Airline Industry
Babke Schepers
s1336266@student.rug.nl
January 2008
1
stSupervisor: Drs. H.C. Stek
2
ndSupervisor: Dr. Drs. H.A. Ritsema
University of Groningen Landleven 5 9747 AD Groningen
ABSTRACT
The formation of alliance constellations (multi-partner alliances) in the airline industry has enabled airlines to offer more destinations at lower cost with smooth and frequent connections. This research examines the impact of constellation membership on passenger traffic growth and the factors that drive this impact. Building on existing literature, it hypothesizes that airlines benefit from entering alliance constellations in terms of increased passenger traffic growth. The analyses of airlines in constellations show that constellation membership is negatively related to annual passenger traffic growth, thereby suggesting that airlines do not benefit from joining a constellation. Of the factors investigating group attributes, the analysis revealed a possible advantage for airlines controlling valuable resources leading to higher growth in passenger traffic. Previous research findings were confirmed with constellation size and network diversity being significant drivers of annual passenger traffic growth. The results extend research in understanding the relationship between alliance constellation membership and passenger traffic growth.
Key words: alliance constellations, airline industry, passenger traffic
TABLE OF CONTENTS Executive Summary………v 1 Introduction………..………...1 1.1 Problem Statement……..………..………4 1.2 Research Question……….…………...6 1.3 Definitions……….………...7 1.4 Thesis Outline……….……….…….8
2 Theory and Hypotheses………..9
2.1 Alliances...………..…9
2.2 Airline Alliances ………..12
2.3 Airline Alliances Benefits……….………..13
2.4 Airline Alliances and Passenger Traffic Growth………16
2.5 Constellation Specific Attributes………....20
2.6 Member Specific Attributes………....23
3 Research Methodology………..………...…………28
3.1 Analysis Method……….………28
3.2 Measurement of Dependent Variable………….………...29
3.3 Measurement of Independent Variables………...……...30
3.4 Control Variables………31
3.5 Data Collection ………..32
4 Results………..………..34
4.1 Descriptive Statistics………34
4.2 Constellation Membership (H1a)……….37
4.3 Members vs. Non-Members (H1b)….…………..………...39
4.4 First Year Impact (H1c)………..41
4.5 Constellation Size (H2)...……….………...43
4.6 Network Diversity (H3)………..……45
4.7 Control of Valuable Resources (H4)……….……47
5 Discussion and Conclusion.………52
5.1 Limitations and Future Research...……….52
5.2 Concluding Remarks………...54
References………..………56
Appendices Appendix A: Constellations and Member Airlines……….62
EXECUTIVE SUMMARY
Following, attributes of alliance constellations were investigated to examine whether constellation characteristics play a role in aggregated performance of members. Two attributes were investigated; size of the constellation and the diversity in their aggregated route network (number of destination countries). A larger constellation creates more opportunities for the bundling of resources (e.g. through improving service frequency) and could therefore attract more passengers. The number of countries the members of a constellation serve enables the constellation to provide the traveler with services to many destinations, while offering smooth travel connections and convenient travel times. Constellation size (in terms of passengers carried and distances flown) and the number of countries a constellation serves were both put forward as drivers raising passenger traffic growth for their constellation members. However, their predicting power was not strong compared to the power of overall industry passenger traffic growth which was included as a control variable. Furthermore, these results were based on a relatively low number of observations, which weakens the strength of regression analysis, and therefore these results should be interpreted with care.
The analysis continued by investigating member specific attributes. Referring to power dependency theory, the control of valuable resources can be viewed as a source of power for airlines within a constellation. Based on the analysis of control of valuable resources, where all member airlines that are important for the constellation considering their access to important airports and/or destinations were identified, a pattern that validates the prediction that airlines with control over these resources could achieve higher growth rates of passenger traffic is found, however the impact was not as strong as expected.
using a measure of passenger traffic growth industry wide) was much stronger in predicting the passenger traffic growth of airlines than the suggested constellation and member specific attributes. A possible explanation for the contradictory findings as compared to previous studies is the limitations on data available, making it hard to sufficiently measure the impact of constellation membership on passenger traffic growth. Furthermore, it remains methodologically challenging to isolate the impact of alliance membership from other factors that possibly affect passenger traffic growth of airlines.
1 INTRODUCTION
The airline industry has been under several pressures in the last decade, with increasing fuel prices, terrorist threats, the SARS epidemic, and the rise of low cost carriers (Wagner et al., 2005). Even more accelerated by the deregulation in the US and the economic integration in Europe, the previous relatively predictable and stable environment for airlines has shifted to a volatile and chaotic one (Klint and Sjöberg, 2003). The industry has structurally changed, being bipolarised with on the one hand low-cost carriers offering simplified services from A to B, and on the other hand the network carriers providing a seamless travel experience by investing in network size and long-haul routes (Tieman, 2006; PATA, 2006).
While this agreement applies to the most travelled region in the world, the options for network carriers to expand their networks globally remain limited under the complex restrictions on ownership and access. The only way for airlines to circumvent the regulatory constraints is by forming international alliances which allows them to globalize their networks while maintaining relatively autonomous (Koza and Lewin, 1999). As a result of this fertile context for alliances, the airline industry observed a virtual explosion of cooperative agreements. This has moved beyond the conventional two-company alliances; the industry has witnessed the formation of global alliance groups, counting up to 17 airlines participating in them. Multilateral in nature, with an agreement that is applicable to all partners, they include joint decision making and common investments, resembling real-life companies with their own CEOs and corporate identities. Originally, they were set up as marketing tools to pool resources and attract more passengers, focusing on combined frequent flyer programs and low integration between partners. However, the increasing demands of travellers for lower fares and frequent services to virtually every destination in the world made airlines realize that this needed a deeper integration between airlines.
The first global alliance groups, or constellations1, were formed in the 1990s, but many were abandoned in their first years of operation due to management difficulties and/or conflicting interests (see figure 1.1 for a brief overview). In the beginning of the alliance race, airlines joined constellations without a master plan or simply because they feared to be left behind. These first alliance constellations can be seen as the ‘trial phase’ of global airline alliances after which the situation stabilized. Three alliance groups remain after the frenzy of alliance formation: Star Alliance, oneworld and SkyTeam. What differentiates these constellations from previous ones is their cooperation in a broad range of activities and the
large number of airlines involved. They aim to increase the level of commitment in order to reap all the benefits possible. Although the circumvention of regulatory boundaries and the creation of synergies a constellation enables are important for airlines, the core driver of alliance formation in the airline industry is the combination of route networks (Iatrou, 2004). As noted by Baker and Field in 2003: the winner is that alliance that offers the most destinations in the most ways.
In a highly fragmented industry with 900 airlines flying the skies, the 37 airlines allied might not seem powerful. But if examined on the alliance level, with 17 of the 20 biggest airlines signed up, it is observed that the three alliance constellations cover around 60% of airline traffic. As can been seen from figure 1.2, they continue to grow and have turned into powerful blocks making them clearly a force to be reckoned with and a potential threat to non-aligned airlines. It enables them to take coordinated actions and present a unified front against outside pressures (Khanna and Rivkin, 2001).
Star Alliance 1989 ’90 ‘91 ‘92 ’93 ‘94 ‘95 ’96 ‘97 ’98 ’99 ’00 ‘01 ’02 ’03 ’04 ’05 ’06 2007 “Wings” Global Excellence EQA Qualifyer KLM/Alitalia Atlantic Excellence oneworld SkyTeam (source: Iatrou, 2004)
With this growing importance of these constellations, the airline industry becomes an interesting field of analysis. To what extent are they successful? According to Klint and Sjöberg (2003), the most obvious sign of alliance success are the benefits that are created for partner firms. After all, firms enter them in order to gain from it. Although airlines conceive alliance success quite differently, increasing the level of passenger traffic is one of the main goals of alliances and commonly used by airlines for measuring their operational performance, with standardized financial information being unavailable internationally. Consequently, this thesis focuses on annual increases in passenger traffic to measure the benefits an airline gains as result of participation in an airline constellation.
1.1 Problem Statement
How can airlines benefit from entering constellations and where do these benefits originate from? With three large airline constellations being active in the industry, competition can no longer be viewed from an individual firm level when firms also compete and operate as part of a larger network. Traditional models of strategy and organization do not apply here since it lacks to take group based advantages and/or competition between groups into account. Yet, it
Figure 1.2 Growing constellations 1997-2007 (number of airlines allied)
is important to understand the issues related to the formation of these constellations. It can have substantial implications for airlines, when its performance can depend on which group it chooses to join (Gomes-Casseres, 1994).
1.2 Research Question
Through this study, a contribution to existing literature is made by examining the impact of constellation membership for member airlines in terms of their ability to improve passenger traffic growth. The main research question is stated as follows: Does membership in alliance constellations has an impact on passenger traffic growth, and how is this impact driven? At the basis of this question is the assumption that the structure of the constellation and that the position of a firm within it can influence the gains created by the members. In order to answer the main research question, it is of importance to investigate constellations and in special how they work in the airline industry. Existing literature on alliances and in special airline constellations can help to explain how constellations can create benefits for partner firms. The following questions need to be answered:
1. What theories are important when discussing alliances?
2. What does existing literature reveal about alliance benefits to firms?
3. How do these theories and literature apply to constellations in the airline industry? 4. How can constellation membership affect passenger traffic growth?
5. How is this impact driven?
6. What does this impact imply for airlines?
1.3 Definitions
1.4 Thesis Outline
This chapter has provided an introduction to this thesis by discussing the problem statement and the questions that have to deal with this problem. Following a deductive research approach, this study starts with theory that will be narrowed down into testable hypotheses (Gill and Johnson, 2002). In the following section, the theoretical framework based on existing theories and previous research findings concerning the topic of interest is build which provides the basis for the formulation of hypotheses. In order to address the hypotheses, data needs to be collected which is the topic in the subsequent section, dealing with the methods and data used. This leads to the testing of hypothesis with specific data in order to confirm or refute the originally proposed theories which empirical findings are discussed in the results section. Finally, the last section embodies the conclusion and discussion part of the research addressing limitations and suggesting directions for further research. The stages of the research process are depicted in figure 1.3 below.
(source: Bryman and Cramer, 2001)
Figure 1.3 Research process of deductive research
Theory Hypotheses Operationalization of concepts
2 THEORY AND HYPOTHESES
In the previous section the problem statement has been described including the questions that have to deal with this problem. In this section the theoretical framework will be build in order to answer the main research question. Previous literature and theories are used to explain why alliances are formed and how they operate. This discussion leads to the formulation of hypotheses about the effects of airline constellation membership on passenger traffic growth that will be tested during this research.
2.1 Alliances
First, it is important to find out why alliances are formed in the first place. In a rapidly changing environment led by globalization and liberalization, firms are facing hyper competition, high levels of uncertainty and high speed technological innovations. In order to respond to these threats, teaming up with other firms became a necessity for survival (Stuart, 2000). Interfirm cooperation is enabling firms to gain access to capabilities and resources that they could not have obtained when operating alone. Alliances are the perfect way for establishing market reach (or market entry), growth and knowledge (Inkpen, 2001). Through an alliance, competencies are shared among firms and this automatically means that a firm loses complete autonomy when certain decisions or activities involve the cooperation of their alliance partners. Nonetheless, in comparison with more extreme types of interfirm cooperation, such as mergers or acquisitions, alliances offer a relatively easy exit option since they have a lower level of involvement.
achieve economies of scale or specialization (1), creating opportunities for increased profitability when they benefit from improved efficiency and effectiveness (Inkpen, 2001). Alliances also offer a solid solution when market access (2) is needed for further growth. Cooperation between firms can also involve the sharing of financial risks (3) and/or knowledge sharing (4) which can create enhancement of innovative capacity and can lead to first-mover advantages (Kale et al., 2002). Taken together, alliance formation enable firms to create a competitive advantage over other firms when they can achieve cost savings, develop faster or better, and form a barrier for entry to new firms. Synergies can be created, which are the result of joint activities that are more value together than they are apart (Lavie, 2006).
The essential element of an alliance is that it creates a win-win situation for all partners involved which is achieved when the partners have complementary competencies. This is not always the case, for example when alliances are formed between competitors to reduce competition or create barriers to entry for new players (Silverman and Baum, 2002). Complementary products, customers, and/or knowledge that are combined present the easiest way for achieving synergies through the strengthening of weaknesses and improvement of capabilities of partner firms. Technological innovation has also boosted interfirm partnerships, where industries started to converge. Take for example the combination of a computer, cell phone and photo camera in one product which can be achieved the fastest and the easiest when resources are shared among different specialized firms.
also decide to leave the alliance, while being it at a certain cost. The dynamic nature of an alliance is a challenge for the management and an alliance can only succeed when it is benefiting all partners involved.
The dynamic nature and problematic alignment of different partners within alliances already reveals its fragility. Failure rates of alliances are extremely high, where most alliances fail due to lack of trust, unsolvable differences between firms and/or expectations that were not fulfilled (Hamel et al., 1989). On the other hand, alliances are highly recognized as beneficial for firms when only the announcement of new partnerships often boosts firm value.
2.2 Airline Alliances
In light of the discussion above, this section will continue by focusing on alliances in the airline industry. The airline industry is no different from other industries when discussing its ultimate aim; profit maximization. In the rapidly changing environment as explained above, airlines started to adapt new strategies for achieving growth, where access and expansion to new areas became the means through which that could be accomplished. Consequently, focus on international expansion became the norm, and alliances became the most convenient way to realize it (Nunes et al., 1997). Airlines that opt to do it alone face high investments with a high level of risk involved, and have to deal with the restrictions resulting from bilateral agreements. Instead of investing high amounts in landing fees and advertising, an airline that decides to form an alliance can connect with a partner’s route network and establish hubs.
Airline alliances occur in many different forms, depending on their specific goals. If the focus within an alliance lays on the joint use of resources, it is usually aimed to reduce costs and/or improve productivity. This occurs for example through joint use of facilities, joint development of systems, and/or joint maintenance. When it is opted to improve sales and increase revenues, codesharing, coordination of flight schedules and joint advertising are common fields of cooperation (Wang and Evans, 2002). While these fields all involve a different level of commitment, they are all likely to have positive impacts on airline performance.
(Iatrou, 2004). They can also be used as trials to test fit between partners, and can be followed by further involvement between airlines. High involvement is needed to establish truly global alliances, including the jointly creation of flight schedules and setting of fares. Global alliances are by definition more strategic as they involve the mingling of assets and are entered for the long term. Cooperation is established in a wide range of activities and includes the combination of airport facilities, staff, aircraft, capital, and traffic rights. The combination of high degrees of cooperation and commitment makes it a strategic alliance. It should be stressed however, that this does not have to include equity involvement which is not common in the airline industry. There are two distinct features of strategic alliances in the airline industry. One, they include exclusive membership, which means that a member from one alliance cannot be a member of another. Two, they build a marketing entity, which means that they jointly brand and advertise (e.g. by officially naming the alliance). Benefits of strategic alliances are higher than alliances that involve limited levels of collaboration (Iatrou, 2004).
2.3 Airline Alliance Benefits
Airlines operate in a service industry, but unlike other service industries, airlines are subject to high operating costs where large investments in capital and labour are required (Park and Zhang, 2000). Costs can be reduced when partners decide to share these costs. Shared use of resources (staff, airport facilities, sale offices) thus leads to significant cost reductions through economies of scale. Also the common purchase of resources is beneficial to firms. Airlines can commonly buy products in bulk (e.g. fuel and catering) and can bargain lower prices as a result of their improved market power (Doganis, 2006). An alliance also enables airlines to advertise commonly, instead of the promotion of individual routes. The global airline alliances operate under a common brand name, which is easily spread to the public when it involves a large number of partners. It has to be stressed that these cost reductions require a higher level of integration between airlines and are not easily achieved in the short term.
Economics of traffic density (when passengers can be carried more cheaply through lowering unit costs) result from network configuration between alliance partners. The connection of networks offer opportunities for airlines to attract more traffic without the need to operate more flights through which economics of traffic density can be achieved (unit costs will fall) (Iatrou, 2004). Successfully linking the route networks of partners enable airlines to benefit from hub-and-spoke networking which secures traffic feed from partner airlines and enable them to operate larger aircrafts and carry higher numbers of passengers, leading to lower cost per passenger and kilometre. To illustrate, each airline collects traffic and flies it to hub airports and transfers it to spokes or other hubs. The more spokes there are, the more destinations that can be served, and consequently the more travellers that can be attracted.
Taken together, alliance formation can lead to positive externalities for airlines through the access to unique resources and capabilities of other firms, when the alliance is governed correctly and all partners’ interests are aligned. It creates opportunities for rises in passenger volumes, cost reductions and improved profitability. When looking at empirical studies that have investigated the potential impact of airline alliances on different performance measures a majority conclude that alliance formation has a positive impact on airline’s performance (Brueckner, 2001; Iatrou and Skourias, 2001; Oum, et al., 2004; Park and Zhang, 1998; Bissesur and Alamdari, 1998; Park and Zhang, 2000). Oum, Park and Zhang (2000) empirically investigated different effects of alliances and their major finding was that alliances improve overall firm performance, with more pronounced effects when a high level of interfirm cooperation is achieved and when it concerns complementary alliances (non-overlapping networks). The analysis of Bissesur and Alamdari (1998) focused on factors that affect operational success of airline alliances. They conclude with partners’ network size and their compatibility, frequency of service between hubs, flights connection time at the hub and the level of competition as the major factors behind alliances’ operational success. These studies provide evidence that alliances enable airlines to strengthen and maintain their competitive advantage through which they can provide better service to old and new markets.
2.4 Airline Alliances and Passenger Traffic Growth
in terms of traffic and revenues to assess their performance. This approach is adopted in this study, which refers to benefits from alliances in terms of benefits to partner firms and more specifically, considering passenger traffic growth figures. Traffic is the key element in the industry and can be positively influenced by alliance formation, which effect is visible on a short term. Increases in profits, e.g. through cost reductions, are often more time consuming and requires more integration between airlines. Cost reductions can also lead to reductions in fares, which then can attract more passengers resulting in passenger traffic growth.
With a focus on the effects on passenger traffic growth in this research a few relevant studied need to be mentioned. First, the study of Park and Zhang (1998) compared traffic changes on alliance route with non-alliance route for the period 1992-1994. They found traffic increases between 6.8% and 66.8% on alliance routes while non-alliance routes showed decreases or light increases up to 9.1%. Second, Iatrou (2004) focused on the global alliance groupings and concluded, with the use of interviews, surveys and econometrical analysis, that alliances have a positive impact on partner airlines in terms of traffic increases.
The discussion above provides evidence that constellations (as a type of alliance) offer opportunities for airlines to increase their revenues and attract more passengers, through improved services combined with improved efficiency. As a result, airlines should record higher growth in passenger traffic after they joined constellations. Following this logic:
Hypothesis 1a: Airlines benefit from entering alliance constellations in terms of
passenger traffic growth.
enabled by alliance formation. The three alliance groupings under research are idiosyncratic in terms of scale, branding and organization and involve a large number of airlines with a high level of integration. As explained above, airlines benefit through resource sharing where more airlines can build a larger resource base to share from. It has also been stressed that higher levels of integration offers more opportunities for benefits. Basically, the combination of cost reductions and service improvements on a larger scale then conventional alliance types, could lead to higher growth in passenger traffic. As a result, they could have potential for superior results in comparison with the conventional alliance types. Cooperation between multiple airlines, being it with a high level of integration, enables them to attract more passengers since they can offer global seamless services and possible fare reductions as a result of the cost reductions achieved. Where the previous hypothesis focuses on the effect of joining a constellation, the next hypothesis will include airlines that did not join any of the three constellations and states that constellation membership posits a stronger relationship with passenger traffic growth then conventional alliance membership. Therefore:
Hypothesis 1b: Airlines in constellations outperform non-member airlines in terms of
passenger traffic growth.
Hypothesis 1c: Airlines in constellations benefit more from the constellation in their first year of membership.
What is striking in the above is that size is put forward as a driver of constellation benefits. The three global constellations under research are all considered large but also differ in size. This indicates that this could also induce different benefits to firms as result of the different size of the constellation. Although many perspectives are available in the literature on alliance performance, relating to external environment factors, knowledge acquisition or relational aspects, an important stream stresses the importance of partner attributes as predictors of alliance performance. Gomes-Casseres (1994) and Das and Teng (2003) put forward the resource base of the partners as predictors of alliance performance. This view is consistent with the resource based view, which claims that the bundle of resources a firm owns equals its competencies (Lavie, 2006). Resource integration forms the core of alliances and the resourced based view offers a solid basis for understanding how firms develop superior performance through alliances (Das and Teng, 2000). From this view also follows the assumption that a larger resource base provides more opportunities for gaining from that alliance. This indicates that attributes of the constellation could affect the amount of extra passengers attracted as a result of constellation formation.
might be of importance for creating inter- and intra-group passenger traffic growth differences in airline constellations is discussed next.
2.5 Constellation Specific Attributes
Gomes-Casseres (2003) argues that the impact of an alliance on firms is likely to differ depending on the characteristics of the alliance agreement, such as total scale of operation, market reach and organizational structure. Referring to the resource based view, the amount of valuable resources possessed by the alliance result in collective strengths (Das and Teng, 2003). Following, the amount of airlines that decide to cooperate can affect their achievements. Establishing market reach is easier if more members join a constellation, which also leads to more potential for risk sharing. However, the number of members can not assure economies of scale, this depends on the size and type of the members. The real challenge is to put the right pieces together, where members can complement each other. Since the formation of the first alliance constellations, they have proven their dynamical nature. During the last decade they have grown rapidly (see figure 1.2) and while this can be an indication of a constellation’s success, this also comes at some expense with more interests that need to be aligned (Gomes-Casseres, 2003). A large and complex constellation may lead to an inefficient management structure, making corporate governance a challenging issue.
(codesharing) will tend to increase the quality of customer service (Doganis, 2006). Codesharing enables airlines to resemble a ‘single-carrier’ service with respect to check-in and baggage handling (Bamberger et al., 2001). Improved service and lower prices tend to increase demand for coordinated services, consequently leading to passenger traffic growth.
Another important strength of alliance constellations is the use of Frequent Flyer Programs (FFPs), which reward customers who purchase tickets with one airline (or constellation), thereby preventing customers from switching to other airlines easily. Larger multi-partner alliances can offer an attractive Frequent Flyer Program because they offer many opportunities both to earn points and to spend them. As one airline executive puts it: ‘the combined FFP is the glue that holds the alliance together’ (Hanlon: 1996: 57). With the marketing of FFPs airlines especially target business travellers who travel regularly and are less price elastic than leisure travellers, creating higher yields for airlines (Iatrou, 2004).
In sum, the value-creating potential of a given constellation is contained in the sum of resources and capabilities that are pooled by member firms (Goerzen, 2005). Thus, the larger the alliance, the greater the aggregated pool of resources and knowledge and the greater the potential for better results. The effect of alliance size on passenger traffic growth originates from cost reductions that lead to cheaper fares, attractive FFPs, and seamless service to a broad range of destinations. All the above can increase customer loyalty and demand. Collectively, these arguments imply that the size of the constellation is an important feature that can influence passenger traffic growth. This leads to the following hypothesis:
Hypothesis 2: Members of an alliance constellation of high aggregated size will
As already highlighted above, it is not only the size of the constellation but also the diversity of resources that can make a constellation successful. Benefits from membership can increase when members hold complementary resources (Park and Zhang, 2000). Alliance members with diverse resources will make them able to combine them in several ways. Complementary route networks are crucial in terms of traffic gains, when more destinations can be marketed to a wider range of people. As a result, a diverse route network can offer many destinations, in combination with an attractive FFP-program, while benefiting from an extensive hub-and-spoke network to offer good flight connections and flight frequencies. It is no surprise that the three largest carriers from the US (American, Delta and United) created three different global alliance groupings where the three largest airlines from Europe (British Airways, Air France/KLM and Lufthansa) have each joined one of them (Grossman, 2007). The same is likely to happen for the major airlines from China; it is expected that China Eastern will join oneworld, China Southern is likely to operate within SkyTeam, and Star Alliance is negotiating with Shanghai and Air China for future membership (Field, 2007).
can not achieve this alone making airline alliance groupings even more valuable. Passenger traffic growth can be best achieved when the services to travellers are improved through the linking of many complementing networks. According to Park and Zhang (2000), travellers prefer constellations that reach globally and offer seamless services. Constellation members are therefore able to improve their competitiveness and be more attractive for travellers. Following, network diversity is another important element for determining membership benefits. The next hypothesis reads:
Hypothesis 3: Members of am alliance constellation with high network diversity will
outperform members of a constellation with low network diversity.
2.6 Member Specific Attributes
The mixed nature of competition and cooperation (co-opetition) within global airline groups and its implications for partners needs to be acknowledged as well (Gilsing and Lemmens, 2007). Above the constellations were compared in terms of aggregated figures but next the focus lies on the individual benefits captured by the members. Heterogeneity of members might induce differentials in membership benefits, even when they belong to the same group. As stated by Gomes-Casseres (1994:6): “Who wins and who loses among competing alliance
groups depends on the competitive advantage that each group of companies collectively builds. Who wins and who loses within a group is a very different matter”. Constellation
An important theory on interfirm partnerships is power dependence theory. This theory basically argues that firms depend on the valuable resources of other firms, resulting in interdependence between firms which determines their relative power. According to the dependency theory, the degree of dependence is related to the amount of resources needed and the alternatives available (Das and Teng, 2003). What are valuable sources when discussing airline constellations? Nohria and Garcia-Point (1991) refer to the size of the firm within the alliance that creates differentials in the way membership benefits are distributed. Others refer to the importance of resources that specific members bring to a network (Gomes-Casseres, 2003). Both perspectives are considered important and will be combined here.
resulting in higher increases in passenger traffic since the larger one cannot benefit from the customer base of smaller partners. However, referring to the dependency theory, this is not likely to be true. In sum, the powerful position of large carriers within a constellation gives them room for capturing a more preferable position, resulting in a relatively higher rise in passenger traffic growth in comparison to smaller carriers (Katz and Saphiro, 1985).
Nevertheless, there are other resources thinkable that could be of great value for the constellation. Also smaller carriers can obtain a powerful position within an alliance group when they have the control over specific valuable resources. Resources, whose withdrawal from the group substantially reduces the benefits captured by members, creates interdependencies between partners and can enable airlines to capture a greater share of constellation benefits. Several resources are of value for the alliance, for example their fleet, country and/or airport access, knowledge (experience), and financial capital. To illustrate, with establishing a global network being mentioned as the core driver behind airline alliances, the destinations an airline serves is mentioned as a valuable resource. Even with so many large airlines allied in alliance constellations, some geographical regions remain underrepresented. The North Atlantic region is well covered but operations remain limited in Africa, South America and certain parts of Asia which are critical for establishing a global route network. Although the carriers from the middle-east show little signs of alliance movement, emerging markets such as Russia, China and India are certainly interesting areas for further expansion.
increases (Lavie, 2006). Powerful partners are likely to get preferred positions in the route network of the constellation. The aim of constellations is to create a flight schedule that offers smooth and frequent connections to as many destinations as possible to attract as many travellers as possible. Consequently, members that fly to ‘important’ destinations will get good feed of passenger traffic in the hub-airports as they obtain good connecting flights in the route network. Through this, it is expected that benefits are not equally allocated among the constellation’s members and that certain members are able to benefit more from the alliance. Consequently, the last hypothesis is presented:
Hypothesis 4: A member controlling valuable resources within its constellation will
outperform members without these resources in terms of passenger traffic growth.
partners are forming the collective strengths of the constellation. The accumulated resources of these firms determine the size of the constellation and airline industry specific, determines how their route network is build. Passenger traffic growth for members is then attained depending on the attributes of the constellation and its position within the group vis-à-vis other members which lead to intra-group differences in passenger traffic growth.
A box is defined as the market context, arguing that overall industry figures affect the extent to which airlines can benefit from constellation membership. This context is partly determined by the airlines in constellations, which can also be derived from the picture below.
MEMBER ATTRIBUTES Resources Partner Customer Base Destinations served Market access CONSTELLATION ATTRIBUTES Collective Strengths Constellation Size Network Diversity CONSTELLATION PERFORMANCE
Passenger Traffic Growth for Partners
Figure 2.1 Conceptual Model
MARKET CONTEXT
3 RESEARCH METHODOLOGY
Following a deductive research approach, the theories presented above will be tested. A quantitative analysis is appropriate for answering the research question and statistical testing fits the hypotheses as formulated in the previous section. Statistical inference enables to draw conclusions from data and can investigate whether a relationship exists between variables. In order to make these hypotheses testable, data needs to be collected on variables that represent the hypotheses best. In this section the measures and methods are explained.
3.1 Analysis Method
(ratio or interval), predictor variables should be on a ordinal, interval or ratio scale or can be a dummy variable (dichotomous variable on a nominal scale), and the number of cases should substantially exceed the number of predictor variables used (5:1 ratio) (Chen et al., 2003).
The population is assumed to be normally distributed and multivariate and bivariate analysis (including two or more variables) is used to estimate the regression model and specifies the role of the chosen independent variables on explaining passenger traffic growth. A significance level of 0.05 (α = 0.05) is followed, which is the most common level of significance. This can be interpreted as that when a hypothesis is supported; the relationship is valid in 95% of the cases. Since the hypotheses investigating constellation specific attributes are based on a low number of observations which weakens the analysis and lowers the chance of arriving at significant results, a significance level of 0.05 is hard to achieve and therefore a significance level of 0.10 is chosen. Regression analysis requires the definition of a dependent variable, independent variables and also enables the use of control variables to estimate the impact of other predictor variables that are not discussed in the hypotheses. Next, the methodological choices concerning operationalization of the hypotheses are discussed.
3.2 Measurement of Dependent Variable
Passenger traffic constitutes only one of the several non-financial performance criteria thinkable to assess firm performance (passenger load factor, productivity, customer service level etc.) but it is undeniable that all these parameters are some what linked with passenger traffic. The major disadvantage of capacity measures (e.g. passenger load factor) is that alliances are often combined with increasing service frequency, while traffic is likely to respond gradually and could therefore negatively affect load factors (Iatrou and Skourias, 2005). Absolute figures of passenger traffic are inadequate, where passenger growth in percentages ensures comparability with different sized carriers. Therefore, the annual increase in passenger traffic in percentages (YPTGor YCPTG when aggregated constellation figures are
investigated) is the dependent variable in this research, following previous studies conducted by Park and Zhang (2000), Iatrou and Skourias (2005) and Iatrou (2004).
3.3 Measurement of Independent Variables
The four hypotheses posit a relationship between constellation membership and annual passenger traffic growth. Below, the chosen independent variables are discussed.
Membership (XCM): First a dummy variable (nominal) was needed to identify
constellation membership to measure the impact of joining an alliance constellation. A value of 0 represents the period where they were not participating in any of the three investigated constellations and a value of 1 if they were.
Member (XMEM): A variable on nominal level is included to identify non-member
airlines, considering hypothesis 1b. Dummy variables are the only variable on nominal level which can be included in regression analysis, where this variable is either 0 or 1.
First year effect (XYEAR): A dummy variable is employed to identify the first year of
Constellation Size (XCRPK): While there are many variables thinkable that could give
information about constellation size, this research follows a measure that is commonly used for measuring airline size, namely RPK. RPK stands for Revenue Passenger Kilometre, and is measured by multiplying the number of revenue passengers carried on each flight with the flight distance. Constellation size is then calculated as the aggregated RPK of all members during that year. This measure is preferred above the number of members, passenger traffic or fleet size, which do not reckon differences in firm size or route networks.
Network Diversity (XCA): With alliance networks opting to offer flights to any
destination in the world; network diversity is measured in the total amount of countries the constellation operates in (ratio). Although this variable does not take specific destinations within countries nor the importance of access to certain countries into account, it can still provide insight in the relevance of network diversity for member performance and was the best option considering data availability.
Control of Valuable Resources (XRES): Variable resources are identified by performing
a qualitative analysis based on information from published articles which can help to identify airlines with valuable resources in a constellation, considering their resource base, the countries they fly to, and access to airports. Using a dummy variable, airlines in possession of valuable resources are identified by a value of 1.
3.4 Control Variables
airline industry in general 4 years of poor performance are followed by 5 years of improved performance (De Heer and Koller, 2000). Therefore, two variables that control for the overall growth in the market are included in this research:
Annual Industry Passenger Traffic Growth (XIG): This variable measures the annual
growth of passenger traffic industry wide (in %). This variable is preferred since it takes into account several factors that could influence passenger traffic growth as determined by the external environment. Doganis (2006) has put forward several factors that influence passenger traffic – income level (GDP), fares, population growth, economic trade – which are all captured in this control variable.
Annual Regional Growth (XRG): Economic conditions can affect the amount of
passengers an airline attracts, where it can not be denied that differences exist between geographical regions. Therefore a control variable that shows the annual growth of passenger traffic per world region (in %) as calculated by the IATA on basis of a sample of carriers in that region has been included, enables controlling for events such as 9/11 where especially North American carriers had to deal with decreasing passenger traffic.
3.5 Data Collection
supplement the data available in the IATA publications. With the three constellations under research being formed between 1997 and 2000, the period 1994-2004 covers years before and after constellation formation. One problem that could not be overcome is that due these data availability till 2004, airlines that decided to join a constellation after 2004 had to be excluded from the research. Furthermore, in the case of airlines that joined an alliance group in the recent years, results on their long term performance within the constellation is missing. However, as investigated by Iatrou (2004), 81% of the airlines interviewed state that they achieved positive results within the first year after joining an alliance constellation.
Data on their membership history is obtained from the airlines’ or constellation’s websites which also provided information about their route network which was needed to calculate the network diversity of the constellation. Two airlines had to be excluded since they either lacked a portion of data or was affected by bankruptcy. Since these airlines were only member of a constellation during a few years, this causes no further great problems for data analysis. In total 32 member airlines remained in the dataset.
4 RESULTS
This section focuses on the quantitative analysis of the impact of airline constellations on passenger traffic growth. The aim is to investigate if there are positive effects on passenger traffic that were created through joining an alliance constellation, and what the drivers of these effects are. It should be stressed that it is methodologically challenging to isolate constellation membership impact from other success factors, but regression analysis allows the use of control variables which can help to explain how passenger traffic growth is driven (see also section 3.1). Following the same sequence of hypotheses stated in the theory section the analysis will be conducted. Only a summary of the statistical outcomes are provided in this chapter where appendix B can be consulted for more details.
4.1 Descriptive Statistics
Variable N Minimum Maximum Mean SD YPTG Airline Annual Passenger Traffic Growth 442 -40.55 77.33 5.68 11.62203 YCPTG Constellation’s Passenger Traffic Growth 19 -0.01 0.10 0.33 0.34 XCM Constellation Membership 442 0 1 0.36 0.479 XMEM Member Airline 298 0 1 0.53 0.5 XYEAR
First Year Membership 158 0 1 0.18 0.388
As can be seen from the table as presented above, the passenger traffic growth of airlines ranges from minus 40.55 % to plus 77.33%. The mean statistic of 5.68% is quite in the center of these values and when taking a closer look of the data it shows a normal distribution of values. This provides evidence that extreme values are not affecting the results. When a high decrease or increase in passenger traffic growth is found, it is checked for whether this was caused by external events, for example a merger, or by opening new routes which makes their results legitimate for statistical analysis. The distribution of the control variables (industry growth and regional growth figures) shows a relatively shorter range in values. This can be explained by the fact that these are averaged numbers. The same pattern occurs for YCPTG
which represent the average passenger traffic growth of the members in the constellation together. Constellation size, which is denoted as XCRPK, ranges in values from 32.006.885 to
74.281.543, with a mean statistic of 48.650.647. The mean statistic is relatively close to the minimum statistic and this can be explained by the fact that of the three constellations under research, Star Alliance is remarkably larger than the SkyTeam and oneworld.
The dummy variables are dichotomous in nature, which means that they can either have a value of 0 or 1. The mean statistics of these variables reveal the distribution of values among the cases. For example, XYEAR shows a mean statistic of 0.18 which indicates that the
majority of the cases were valued with zero. This is not surprising, where for this specific variable only the year an airline decided to join a constellation was numbered as 1. Since this study explores multiple years of membership, it is obvious that the value of 0 was assigned more often to the dataset.
4.2 Constellation Membership (H1a)
The model for hypothesis 1a is estimated as follows:
YPTG = a + β1*XCM + β 2*XIG + β 3*XRG+ ε2
The regression equation states that constellation membership, industry growth and regional growth respectively are predicting the dependent variable, in this case airline’s passenger traffic growth. Using the enter method (meaning that each independent variable is included in the model) a significant model emerges (p = 0.000). In statistics, the term significant is used to indicate that the evidence for the tested hypothesis has reached the chosen significance level (in this case below the α of 0.05) (Moore and MacCabe, 2006). The p-value stands for the probability of the observed outcome, where the evidence for the hypothesis is stronger when the p-value is smaller (Moore and MacCabe, 2006). A statistically significant model indicates that the results did not occur purely by chance and that a valid relationship exists between the variables.
The multiple correlation coefficient (denoted as R2) reveals the correlation between the observations and the predicted values and is a descriptive measure between 0 and 1 (Moore and MacCabe, 2006). For hypothesis 1a, the R2 is 0.123, which indicates that 12.3% of variation in passenger traffic growth (YPTG) is explained by the selected predictor variables.
Coefficients (beta) indicate the change of the dependent variable, considering the other variables being constant. For example, airline passenger traffic growth increases when industry growth increases (beta = 0.161). The results on the predictor variables are shown below:
Predictor Variable Beta p
XCM(constellation membership) -0.179 0.002
XIG(industry growth) 0.161 0.008
XRG(regional growth) 0.185 0.003
A significant model does not automatically mean that each variable has a significant impact on the dependent variable. Therefore, it is important to check the p-values of each predictor variable. The p-values on the predictor variables are below the significance level of 0.05 which indicates that the data fits the model well, where all predictor variables have a significant impact on predicting YPTG. The low p-values in combination with a significant
regression model indicate that the correlation between predictor variables is not causing multicolinearity3 (Chen et al., 2003). The model is then estimated as follows:
YPTG = 4.682 + -0.179*XMEM + 0.161*XIG + 0.185*XRG+ ε.
The betas of the predictor variables are standardized, which means that they employ the same standard of measurement and can be compared to determine which of these variables is more important in relation to the dependent variable, in this case XIG (Bryman and Cramer, 2001).
In a significant model, it concludes that alliance constellation membership decreases passenger traffic growth. This contradicts with the findings of previous studies, which has found increases in traffic for member airlines. Industry growth and regional growth are both confirmed contributors to passenger traffic growth of airlines. A R2 of 0.123 indicates that more factors are at play determining passenger traffic growth, which is not unexpected for
something as complex as passenger traffic growth. The selected control variables are not sufficient enough to control for these external factors. Referring to approaches from other researchers that have assessed passenger traffic growth as a result of alliance membership have included variables such as service levels (e.g. fares) and/or economic variables (e.g. personal income levels) (Iatrou, 2004; Park and Zhang, 2000). Unfortunately, these factors could not be included in this study considering the time and resources available. The relatively low predicting power of the model does not provide a solid basis for interpreting the results, but according to these findings H1a needs to be rejected.
4.3 Members vs. Non-Members (H1b)
The previous test has examined airlines in the period before and after joining the constellation. The following test includes the annual passenger traffic growth of non-member airlines to be able to compare these two groups. In other words, member airlines are compared with non-members. The data of members during their non-member years are excluded. The regression equation is depicted below and introduces XMEM which resembles the dummy
variable that identifies member and non-member airlines. Both control variables are included since airlines joined in different years, where there results could have been influenced by the growth in the industry or in the region.
YPTG = a + β1*XMEM + β 2*XIG + β 3*XRG+ ε4
In the previous test airlines were investigated in the pre- and post- constellation period, where in this analysis airlines in the post- constellation period are compared with non-member
airlines. Again, a significant model is found, showing a p-value of 0.000. With this model, passenger traffic growth is explained by the predictor variables with a R² of 0.108 indicating that 10.8% of passenger traffic growth is explained by the predictor variables. The results on the predictor variables are shown below:
Predictor Variable Beta p
XMEM (membership) -0.091 0.102
XIG (industry growth) 0.251 0.000
XRG (regional growth) 0.122 0.040
Constellation membership does not have a significant impact on predicting YPTG, but overall
the data fits the model well. These results lead to the following model:
YPTG = 2.180 + -0.091*XMEM + 0.251*XIG + 0.122*XRG+ ε.
The negative impact of constellation membership on airline’s annual passenger traffic growth is still present, but the beta of -0.091 shows that the negative contribution of constellation membership to YPTG is more positive than in H1a (-0.179). However, these coefficients are
standardized and comparison of beta values between different regression models is not possible. Overall, these results lead to the rejection of H1b. Again, the value for R²indicates that other variables are likely to be affecting YPTG. Other factors that are assumed to be of
4.4 First Year Impact (H1c)
In order to investigate the impact of alliance constellation membership over time, considering the statement of airlines that the first year of membership has leaded to the best improvement in results, a dummy variable is employed to define the first year of participation. The dataset now only considers the member airlines during their constellation membership period.
This model is defined as:
YPTG = a + β1*XYEAR + β 2*XIG + β 3*XRG+ ε
The multiple regression analysis (see also appendix B) results in a significant model (p = 0.003), with a R²of 0.088. The betas and p-values of the predictor variables are shown below.
Predictor Variable Beta p
XYEAR(first year effect) 0.071 0.368
XIG(industry growth) 0.260 0.020
XRG(regional growth) 0.040 0.632
Estimation of the associated p-values for the individual regression coefficient reveals that only XIG is significant (p = 0.02). That is, only XIG makes a significant contribution to the
model, where XYEAR and XRG are highly insignificant. This does not mean that the predictor
variables XYEAR and XRG do not contribute to the model, but these results deserve further
variables can be excluded from the model in order to arrive at a better analysis. Running a stepwise regression reveals a R2 of 0.081 if only XIGis included as a predictor variable. The
first model is able to explain 8.8% with 3 predictor variables, where the second model explains 8.1% with only one predictor variable. Normally, the R2 rises when more predictor variables are included, and the slight increase as experienced here indicates that the second model is stronger and the analysis concludes with industry growth as the only valid predictor of YPTG. Although the beta XYEAR shows a positive beta value in model 1, indicating that the
first year of constellation membership increased passenger traffic performance in comparison with the years thereafter, these results are not significant and hypothesis 1c can not be supported. The analysis has led to the following regression equation:
YPTG = 1.040 + .0.284 *XIG + ε.
Taken the three hypotheses investigated together, the expected pattern has not risen from the analysis of above. Where previous studies have focused on specific routes to examine passenger traffic increases, this study looks at the overall passenger traffic growth of an airline and also compares it with non-member airlines and their non-member years. The test results show a significant higher performance for airlines in their non-member years or for non-member airlines. Furthermore, it has been proven that alliance membership does not lead to significant higher results in the first year of participation.
passenger traffic into account is based on a sample of carriers, being averaged numbers. The high external disturbances that occurred after 2000 (e.g. 9/11) might not be cancelled out by these averaged numbers. Especially major US airlines suffered from 9/11, being in serious trouble and close to bankruptcy, where regional carriers were less affected since they offered services to specific markets and aimed at business travelers who ‘have to fly’ (Bisignani, 2007). Also, since data after 2004 could not be analyzed, including the years after 2004 could have led to different results. Over the last few years the airline industry has boosted and restored its health with increasing passenger traffic, creating a different setting for alliance constellations. Focusing on the non-member sample, it is observed that 17 of the 20 largest airlines are participating in constellation, which automatically leads to relatively smaller firms in the non-member sample. As has been mentioned before, firm size was put forward as a driver raising passenger traffic growth and could therefore have affected the results.
As a final remark it should be mentioned that although an increase in passenger traffic can be interpreted as a positive outcome, this does not cover the whole story. For example, Alitalia is making record losses while their passenger traffic performance is considerably decent. Further analysis is needed to give a more detailed conclusion about the performance implications of alliance networks. Next, the analysis will continue by focusing on constellation attributes.
4.5 Constellation Size (H2)
by hypothesis 2, members of a large aggregated constellation are able to outperform members of smaller constellations. Next, the aggregated performance of the members will be compared with the alliance size in terms of aggregated RPKs (revenue passenger kilometre) to examine whether these two concepts are linked. By operating a regression analysis, it is investigated whether aggregated RPK is affecting aggregated constellation passenger traffic growth. A significance level of 0.10 is followed in this case, because the low number of observations makes it rather impossible to achieve significant results with an α of 0.05. The regression model is:
YCPTG = a + β1*XCRPK + β 2*XIG + ε
Consequently, the regression can be run. The regression model is significant (p = 0.001) and the model predicts an estimated 56% of aggregated constellation performance, which is considerably high with two independent variables. The results on the predictor variables are shown below:
Predictor Variable Beta p
XCRPK(constellation size) 0.069 0.070
XIG(industry growth) 0.719 0.001
The coefficients (beta) indicate that XCRPK and XIG are of predicting value for constellation
passenger traffic growth (YCPTG). However, the relationship between constellation size and
but would not have been significant on the 0.05 level. XIG is highly significant and is a strong
contributor to YCPTG. The regression model is formulated as:
YCPTG = 0.002 + 0.069*XCRPK + 0.719*XIG + ε.
Concluding, the results reveal a positive correlation between XCRPK and YCPTG, following
hypothesis 2 is supported. It concludes that a larger constellation in terms of RPK is beneficial for member firms in order to raise passenger traffic growth. However, these results have to be interpreted with care. Statistical testing as employed above is based on probabilities, making it impossible to guarantee that the decision for rejecting or accepting the hypothesis is the right one. The low number of cases analyzed does not provide a solid ground for statistical analysis and a larger number of observations are preferred although impossible due to the aggregated figures analyzed in this case.
4.4 Network Diversity (H3)
Source: Airline Business
Performing a regression analysis where the number of countries an airline group serve as a predictor of YCPTG, the outcome shows a significant model (p = 0.001), where the predicting
power of the model is 55.8% (see appendix B for more details). The regression equation is build as following:
YCPTG = a + β1*XCa + β 2*XIG + ε.
However, the number of countries a constellation serves (XCA) is not a strong predictor of
YCPTG while XIG is. The results on the predictor variables are presented in the table below,
where the p-value of XCA is slightly above a significance level of 0.05.
Predictor Variable Beta p
XCA(countries active) 0.053 0.054
XIG(industry growth) 0.744 0.000
However, this analysis is based on a small amount of observations which makes it hard to generate grounded evidence - with an α of 0.05 - for a relationship based on quantitative analysis. Therefore, the significance level of 0.10 is accepted here and following XCA can be
interpreted as a valid predictor of YCPTG. However, as already explained in the discussion of
the previous hypothesis, these results should be interpreted with care. The equation is:
YCPTG = -0.07 + 0.053*XCa + 0.774*XIG + ε.
In sum, hypothesis 3 can be supported on basis of these results, although the effect of the number of countries a constellation serves is limited compared to the effect of annual average industry passenger traffic growth. While industry growth is based on data of around 700 airlines, it seems that the 32 airlines included in this research resemble industry growth extremely well. Concluding, these results suggest that constellation membership and constellation attributes are not important predictors of passenger traffic growth.
4.7 Control of Valuable Resources (H4)
aspects, such as firm size, access to important airports/countries, image, their fleet or the year they joined the constellation. Since this is hard to investigate with quantitative analysis, a qualitative analysis is needed in order to identify airlines that have control of valuable resources that can be of particular importance for the position within a constellation.
An in-depth analysis for each constellation is needed. On basis of articles about the constellations as published in different studies or articles, the powerful airlines in each constellation are discussed below. Starting with the Star Alliance, the recently joined airline South African resembles an important step for Star Alliance to expand their activities on the African continent. This can also be seen from figure 4.2 which shows the presence of the three alliance constellations in different world regions according to the countries they operate in. Only the years 200 and 2007 are depicted, giving a short overview of the presence and growth of the three constellations. The increase of destinations in Africa has more than doubled for Star Alliance in these years. Unfortunately, South African joined the Star Alliance
2000 0 5 10 15 20 25 30 35 40 45
Africa Americas Asia Asia Pacific Europe
world regions c o u n tr ie s a c ti v e Star Alliance Oneworld SkyTeam 2007 0 5 10 15 20 25 30 35 40 45
Africa Americas Asia Asia Pacific Europe
world regions c o u n tr ie s a c ti v e Star Alliance Oneworld SkyTeam
