The effect of code-sharing alliances on airline profitability

Journal of Air Transport Management 58 (2017) 50e57

Contents lists available at ScienceDirect

Journal of Air Transport Management journal homepage: www.elsevier.com/locate/jairtraman

The effect of code-sharing alliances on airline profitability Li Zou*, Xueqian Chen 1 College of Business, Embry-Riddle Aeronautical University, Daytona Beach, FL 32114, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 20 May 2016 Received in revised form 30 August 2016 Accepted 15 September 2016

Code sharing and global alliances both have been increasingly adopted by airlines worldwide in recent years. A growing number of airlines, therefore, are embedded in networks of multilateral “coopetitive” (i.e., cooperative, but competitive) relationships that influence their product offering, pricing strategies, operating efficiency, market power, and their overall successes. There has been considerable research analyzing the benefits for airlines from joining global alliances, including bilateral code-sharing partnerships. However, the joint effect of code-sharing and global alliances on airline performance has not been fully investigated. In this paper, we study how the use of code-sharing strategies and their structural embeddedness into global alliances may impact airline performance. Using a unique dataset compiled from Flight Global and Airline Business's Annual Airline Alliance Report, the paper empirically investigates the joint benefits of code-sharing partnerships and global alliances on airline profitability. The results based on a group of 81 airlines during the 2007e2012 period show that the profit margin of an airline is positively associated with the number of code-sharing partners it has. Furthermore, the profit margin gains from code-sharing are greater when an airline has a higher proportion of its codesharing partners in the same global alliance; i.e., allied code-sharing partners. Finally, we find no significant evidence that the percent of comprehensive code sharing partnerships to total partnerships has an impact on profit margin. © 2016 Elsevier Ltd. All rights reserved.

Keywords: Codesharing Airline global alliances Profitability

1. Introduction Code-sharing arrangements, the most common type of airline alliance, rapidly developed in the US domestic airline industry after industry deregulation in 1978 and on international routes by the late 1980s (Dresner, 2010). Under a code-sharing arrangement, one airline can use its designation code on a flight operated by a second carrier. The seats on that flight can be marketed and sold by the first airline either to provide regional connections to complement its own network (i.e., complementary alliance) or to reduce competition by having only one airline actually operate on the route (i.e. parallel alliance). There are several benefits for airlines from developing code-sharing partnerships. For example, through codesharing arrangements, a major hub-and-spoke airline can use the

* Corresponding author. College of Business, Embry-Riddle Aeronautical University 600 S. Clyde Morris Blvd., Daytona Beach, FL 32114, USA. E-mail address: [email protected] (L. Zou). 1 Ms. Xueqian Chen is an undergraduate student in the accelerated MBA program at the College of Business of Embry-Riddle Aeronautical University, and she helped collect data for this project under the supervision of Dr. Li Zou during the spring of 2015. http://dx.doi.org/10.1016/j.jairtraman.2016.09.006 0969-6997/© 2016 Elsevier Ltd. All rights reserved.

flights operated by its regional affiliates or partners to feed traffic from spoke cities to its hub cities, enabling the efficient and successful operation of hub-and-spoke networks. Code-sharing arrangements also allow an airline to expand its service network without committing its own resources; for example, by extending its route coverage to more international destinations that otherwise cannot be served under the restrictive regulatory framework for international air transport. The distinct advantages associated with code-sharing arrangement prompt many global airlines to enter into such partnerships, first at a bilateral level and later evolving into multilateral, more formalized group alliances. In 1989, Wings, the first global alliance, was established, mainly based on cooperation between KLM and Northwest. In 1997, Star Alliance, the largest and most mature alliance, was formed by five core members, including United, Lufthansa, SAS, Air Canada and Thai Airways Intl. One year after the formation of Star Alliance, American Airlines, British Airways, Qantas, Canadian Airlines, and Cathay Pacific teamed together and formed their own alliance e Oneworld. In 2000, Skyteam Alliance was founded by Air France, Delta, Korean Air, CSA, and Aeromexico. Following the merger between Air France and KLM, all the major member airlines of Wings joined Skyteam. With the extinction of

L. Zou, X. Chen / Journal of Air Transport Management 58 (2017) 50e57

Wings, the remaining three global alliances e Star, Oneworld, and Skyteam, continue their growth and expansion, increasing airline membership by 58% from 34 in 2004 to 54 in 2012. With the increasing prevalence of code-sharing partnerships and global alliances, a growing number of airlines, therefore, have recently been embedded in networks of multilateral “coopetitivity”, meaning the coexistence of cooperation and competition among allied partners that influence product offerings, pricing strategies, operating efficiency, market power, and overall performance. On one hand, more and more airlines have formed closer and deeper partnerships with allied airlines in the same global alliance to leverage their joint branding, joint marketing, resource sharing, etc., for potential revenue gains, cost savings, or both. On the other hand, airlines also have developed and maintained their bilateral alliance relationships with non-aligned airlines or even with airlines from rival group alliances. For many airlines, including both aligned and non-aligned carriers, the bilateral code-sharing strategy remains a driver of revenue growth and cost savings. Though there has been much research analyzing the benefits for airlines either from joining global alliances or from building bilateral code-sharing partnerships, an examination of the joint effects of code-sharing and global alliances on airline performance is still unexplored. In this paper, we focus on the question: To what extent are the impacts from code-sharing strategies on an airline's performance moderated by its structural embeddedness (or the lack thereof) into global alliances? Using data collected from Flight Global and the Annual Airline Alliance Summary Report results published by Airline Business, we empirically investigate the combined effects of code-sharing partnerships and global alliances on airline performance. The results based on a group of 81 airlines during the 2007e2012 period show that the profit margin of an airline is positively associated with the number of code-sharing partners it has. Furthermore, the profit margin gains from codesharing partnerships are greater when an airline has a higher proportion of its code-sharing partners in the same global alliance, i.e., allied code-sharing partners. Perhaps, due to methodological limitations, we do not find significant evidence that the level of cooperation moderates the profitability benefits for code-sharing partners from a code-sharing strategy. Nevertheless, our finding that there are joint benefits from code-sharing partnerships and global alliances provides valuable implications for airline management seeking to develop the most rewarding code-sharing partnerships in the context of global alliances. This paper contributes to the existing literature and to airline alliance management in three ways: First, we develop a new construct, namely, allied code-sharing partner, to represent the code-sharing partnership formed between airlines in the same global alliance. This construct is then used to measure the extent of an airline's allied code-sharing partnerships compared to its total number of code-sharing alliances. Through our empirical analysis, we find evidence suggesting that an airline can gain a greater profitability benefit when it increases its code-sharing partnerships with allied airlines. For airlines that are already in the three global alliances (i.e., Star, Oneworld, and Skyteam Alliance), this finding can help them decide whether to develop code-sharing partnerships with allied or non-allied airlines. For those non-aligned airlines that have existing code-sharing arrangements, our findings may be valuable in helping them to choose the most beneficial global alliance to join. Second, although there is empirical research estimating the impact of code-sharing alliances on airline performance, the moderating effect of the depth of code-sharing alliances (measured by the extent of route integration through code-sharing arrangements) is unexplored. In this paper, we develop a convenient instrument to measure the depth of code-sharing arrangements in order to estimate their potentially moderating effects.

51

Finally, we contribute to the existent literature on airline alliances by using a panel dataset that includes a broad sample of airlines varying in operating scale, geographic region, and alliance engagement. In the following section, a literature review is provided. Section 3 introduces our hypotheses. Data descriptions and the empirical models are presented in Section 4. The results are summarized in Section 5. The last section concludes the paper and discusses potential for future research. 2. Literature review Dresner and Windle (1996) examine the early development of airline alliances and code-sharing arrangements and suggest several rationales for the adoption of such strategies by a growing number of international airlines in the 1990s. For most airlines, the primary consideration of forming alliances is to expand global route coverage and serve on international routes that would otherwise be impossible to offer due to legal and regulatory restrictions. Through code-sharing arrangements, the most common type of airline alliance, each of the two partner airlines can market and sell seats under its code on the flights operated by the partner carrier. Such an arrangement may generate revenue to partner airlines through market expansion, traffic feeds, improved connectivity, multiple listings on Computer Reservation System (CRS) screens, etc. Moreover, code-sharing arrangements and other alliance activities may help the partner airlines reduce the cost per passenger because of increased traffic, joint advertising, equipment sharing, etc. Therefore, it is expected that airlines may gain competitive advantages by code-sharing, and that carriers that do not enter into these partnerships are at a disadvantage (Dresner and Windle, 1996). There are various types of code-sharing partnerships. In the US domestic airline industry, the first code-sharing arrangement were formed between major US airlines and their smaller regional partners that provided service on lower density routes, and fed traffic onto mainline routes operated by the major airline. This type of code-sharing arrangement, known as “complementary codesharing,” has become an essential component of the hub-andspoke systems in the US domestic airline industry (Ito and Lee, 2007). On international routes, a similar type of code-sharing arrangement has been adopted by international airlines to connect their route networks and provide seamless connections for passengers traveling from one country to another and flying beyond gateway hubs to inland destinations in the foreign country. In contrast, so-called “parallel code-sharing” arrangements are developed between two airlines that competed on routes prior to forming partnerships (Park, 1997). The study of airfare effects of parallel versus complementary alliances has been a research topic over the last two decades (Yousseff and Hansen, 1994; Oum et al., 1996; Park, 1997; Park and Zhang, 2000; Brueckner, 2001, 2003; Brueckner and Zhang, 2001; Ito and Lee, 2007; Wan et al., 2009; Zou et al., 2011; Gayle and Brown, 2014). The extent of cooperation associated with code-sharing arrangements on international routes also varies depending on whether antitrust immunity is granted or not. With antitrust immunity, the partner airlines are allowed to jointly set airfares and capacity, enabling them to have greater cooperation in terms of airfares, flight scheduling, marketing and capacity adjustments (Dresner and Windle, 1996; Brueckner, 2003). Given the prevalence of code-sharing partnerships and the potential anticompetitive concerns over antitrust immunity for the alliances, Brueckner (2003) develops an in-depth study to investigate the joint airfare effects of code-sharing alliances and antitrust immunity. As an extension of his earlier results (Brueckner and

52

L. Zou, X. Chen / Journal of Air Transport Management 58 (2017) 50e57

Whalen, 2000), Brueckner (2003) finds strong evidence that airline cooperation through either code-sharing or antitrust immunity leads to airfare reductions for interline passengers on international routes. However, the combined airfare reduction effects from code sharing and antitrust immunity were found to be smaller than the separate, individual effects. Such airfare reductions generate further benefits for interline passengers above and beyond the improved convenience and tighter connections associated with code-shared flights. Gayle and Brown (2014) compare the average airfare, traffic and market share on routes connecting the 50 largest cities in the US before and after the implementation of alliances among Delta, Continental, and Northwest Airlines in 2003 and find no statistical evidence for collusive pricing on code-shared routes. Instead, a traffic increase is found on routes where these three airlines have code-sharing arrangements. The demand increase only existed on routes where the joint market share between two allied airlines was greater than 0.49 prior to their code-sharing alliance. According to the authors, the traffic stimulating effects can be attributed to the enhanced opportunity for passengers to accumulate and redeem frequent flier points with code-sharing partner airlines. In other words, the formation of code-sharing partnerships seems to bring additional benefits to airlines that already have a base of customers that are loyal to either one of the partners prior to the alliance. Through code-sharing alliances, loyalty may induce loyalty; thus, the total traffic of the code-sharing partners is greater than the sum of the traffic of the individual airlines. While there has been a general consensus about how codesharing alliances affect airfares in both the domestic and international context, its overall impacts on airline performance are less certain, with mixed findings in the existent literature. For example, in their study of the three alliances formed between US airlines and their European counterparts in the 1990s, including Continental Airlines/SAS, Delta/Swissair, and Northwest/KLM, Dresner et al. (1995) find that not all three alliances generate greater than average increase in traffic and load factor in their markets. Therefore, they conclude that the expected benefits from alliances are not guaranteed, even after the partner airlines realign their route networks for improved connectivity. By contrast, in studying the codesharing alliances between US Air and British Airways and between Northwest and KLM, Gellman Research Associates (1994), using a different methodology, find evidence in support of the positive traffic and revenue effects of these alliances. Their methodology is based on passenger selection outcomes among alternative traveling options, which are then employed to derive values placed by passengers on fares, flight time, connections, codesharing arrangements, etc. Based in part on interview data and airline internal data, the US General Accounting Office (1995) reports large traffic gains and revenue increases for three strategic code-sharing alliances, including Northwest/KLM (formed in 1992), USAir/British Airways (1993), and United/Lufthansa (1994). As for the four regional codesharing alliances examined, two are found to result in modest increases in traffic and revenue, including United/Ansett Australia and United/British Midland, while the other two e Northwest/ Ansett Australia and TWA/Gulf Air fall, are not. According to the US General Accounting Office (GAO, 1995), the majority of the 61 codesharing partnerships have a low degree of route integration at that time and therefore, they are not profitable or long lasting. Morrish and Hamilton (2002) provide a comprehensive review of airline alliances and their impacts on airline performance in both economic and non-economic terms. As Morrish and Hamilton (2002) note, ‘there is no conclusive evidence to date that major airlines have been able to use global alliances to restrict competition and boost their own profitability.’ They further suggest that although alliance partners might experience some increase in

traffic, load factor, and productivity from alliances, these benefits may be partially or even totally offset by greater frequencies and lower airfares, thereby resulting in modest or little profit gains. Pitfield (2007) echoes the view saying that an analysis using data from the US Bureau Transportation Statistics ‘does not yield unambiguous conclusions in accordance with theory or expectation’ about the positive impacts of code-sharing alliances on traffic and market share of partner airlines. These inconclusive findings at the route level may be due to the complexity of supply and demand interactions, unidentifiable route characteristics, and/or diverse regulatory and competitive environments (Pitfield, 2007). The integration between airlines in code-sharing alliances may be extensive, and distinct for different arrangements. Depending on the specifics of each alliance, the market coverage ranges from limited to comprehensive, and from point-specific to regional and strategic. The scope of cooperation may include a variety of areas, such as scheduling, route networks, operations, advertising, frequent flyer programs, etc. As a result, it is necessary to take into account the level and characteristics of code-sharing arrangements when estimating the benefits to the partner airlines. For example, in the study by Oum et al. (2004), the authors categorize alliances into high-level and low-level by the degree of cooperation involved. A high-level alliance involves network-level collaboration, through which the allied airlines link their route networks. By contrast, the collaboration in a low-level alliances only occurs at the route level without combining the entire networks. The authors estimate the impacts of alliances on airline productivity and profitability focusing on the 108 alliances formed among the 22 leading international airlines during the 1986e1995 period. Their results suggest that although alliances in general only lead to the improvement of productivity (not profitability), strategic alliances characterized by high-level cooperation contribute to both higher productivity and greater profitability. Yousseff and Hansen (1994) consider technical efficiency and market power as two primary factors leading to increased profitability for airlines from participation in code-sharing alliances. More specifically, improvement in technical efficiency will contribute to unit cost reduction and can be accomplished through traffic increase, route network consolidation, resource sharing, flight scheduling optimization, and greater outputs in terms of the quality and quantity of connecting services, etc. In addition to the cost saving benefit, the formation of alliances, in general, is viewed as a means for airlines to restrain competition, preclude rivalry, and seek virtual monopoly status. Alliances allow carriers to retain market power which otherwise might be threatened as the longstanding regulatory regimes for air transport are replaced with liberalization over time (Yousseff and Hansen, 1994). On the other hand, the new entrants may also resort to code-sharing alliances as a strategy to help strengthen their market power against the dominant market leader. Using data for 56 airlines during the 1986e1993 period, Park and Cho (1997) find evidence showing that the combined market share of partner airlines is more likely to increase after the formation of code-sharing alliance and such market share gains are higher when the alliance is formed between two relatively new entrants on a route, and when the market has fewer competitors and it is growing quickly. There may be inconsistencies between theoretical predictions and empirical results regarding the cost effects from code-sharing alliances. In a survey conducted by Iatrou and Alamdari (2005) among managers of alliance departments at 28 airlines belonging to the four global alliances in 2002, the respondents, on average, perceive cost to be least affected by alliances, compared to other factors, such as fares, revenue, load factor, and traffic. Moreover, as compared to outcomes such as traffic growth, load factor increases, and revenue growth, cost reduction is cited much less as an

L. Zou, X. Chen / Journal of Air Transport Management 58 (2017) 50e57

outcome of code sharing by the participating respondents. In sum, Iatrou and Alamdari (2005) find that code-sharing is the most effective type of alliance for providing airlines distinct revenue benefits in terms of traffic and load factor increases. As for the cost effects, the authors do not observe the same discernable benefits, as widely acknowledged by airline alliance managers. The cost reductions associated with alliances are nonexistent or very small, at least in the short run. Through estimating a structural econometric model, Gayle and Le (2013) decompose the cost effects of codesharing alliances at the route level into three types: short-run marginal cost, medium-to long-run sunk market entry cost, and recurring fixed costs. Using a difference-in-differences estimation methodology, Gayle and Le (2013) focus on the three-way alliances among Delta, Northwest and Continental to investigate the alliance effects in contributing to cost changes at these three airlines. Their results suggest that the Delta-Northwest-Continental alliance allows the three partners to decrease their marginal costs, especially when they have a dominant market presence at origin and destination endpoints, and to decrease their market entry sunk costs. However, the three airlines are found to have increased fixed costs following their alliances, implying that the overall cost effects from alliances are uncertain because of these offsetting consequences. The assessment of the profitability effects of code-sharing alliances needs to take into account several tradeoff, with consideration to the following issues: First, “parallel” and “complementary” code-sharing alliances have distinct effects in terms on traffic, load factor, service, airfare, and revenue. The different outcomes are also contingent upon whether the alliance is endowed with antitrust immunity. Second, the traffic increase resulting from code-sharing alliances may not be sufficient to offset the potential airfare reduction and the incremental costs incurred. Thus, the overall profitability effects cannot be assessed based on changes in traffic or revenue alone. Third, the effects of code-sharing alliances may not be the same across all routes because of distinct route-related market structure and demand characteristics. Moreover, even though partner airlines may win traffic from rival airlines due to alliance agreements, the rival airlines in response may retaliate, punitively, on overlapping routes with the partner airlines and cause the loss of traffic for the partner airlines on some of their noncodeshared routes. Therefore, it is necessary to examine the overall effects of code-sharing alliances not only at the route level, but throughout the network. 3. Hypotheses development As stated above, it is necessary to consider both the potential revenue and cost effects in a code-sharing arrangement for an accurate assessment of its overall profitability. In contrast to a great number of studies investigating revenue effects, empirical studies on the cost effects of code-sharing alliances are very limited. As noted by Gayle and Le (2013), the lack of the research, in part, is due to the difficulty in disaggregating cost data to the route level. Theoretically, the overall effect of alliances on the total cost of airlines (including marginal costs, market entry costs, and fixed costs) may be uncertain. On one hand, code-sharing alliances may help the two partner airlines consolidate their traffic by reducing or eliminating duplicate flights. The practice of code sharing also enables airlines to expand their route networks, optimize flight schedules, and improve connecting services, thereby attracting new passengers and increasing traffic. The increased traffic density can lead to a marginal cost reduction. Moreover, an airline under a code-sharing arrangement can enter new markets by using the flights of its partner airlines. Thus, the airline can avoid making investments in route development, thereby saving costs associated with market entry. On the other hand, to accommodate the

53

increased traffic resulting from code-sharing alliances, the two partner airlines may have to incur costs acquiring baggage handling facilities, airport gates, check-in counters, airplanes, etc., if there is a traffic volume increase beyond their joint capacity. Because of these offsetting effects, Gayle and Le (2013) further suggest that the overall impact on the total costs of partner airlines will be small after considering both the positive and negative forces. This view is consistent with empirical findings in the studies by Goh and Yong (2006), Gagnepain and Marin (2010), and Chen (2000). The survey research by Iatrou and Alamdari (2005), based on the questionnaires filled by alliance managers of 28 airlines participating in Wings, Star, oneworld and Skyteam in 2002 also indicate that cost is perceived to be least affected by the formation of alliances, compared to several other aspects of airline operations, including airfares, load factor, revenue and traffic. Using firm-level data for 10 US airlines from 1994 through 2001, Goh and Yong (2006) estimate a truncated third-order translog cost function and find that ‘ … alliances do appear to lower costs. However, while the impacts are statistically significant, in economic terms the magnitude appears to be immaterial.’ On the revenue side, the majority of the existent literature provides both theoretical arguments and empirical support for the revenue benefits from codesharing alliances. Thus, we propose the following hypothesis. Hypothesis 1. The operating margin of an airline is positively associated with the number of code-sharing partners it has. The level of cooperation between two code-sharing partners can be enhanced when the partners are in the same global alliance. Through sharing airport lounges, joint marketing, branding and frequent flier programs, etc., allied airlines have more opportunities to improve their connecting services involving code-shared flights, and to make their services more seamless. Moreover, the service standards followed by airlines in the same global alliances tend to be more harmonized and consistent. Hence, code-share flights offered by airlines in the same global alliance may be more appealing to passengers, as compared to code-share flights by airlines not in the same global alliances. From the passenger’s perspective, Goh and Uncles (2003) discuss five main benefits that air travelers (especially business travelers) may be aware of when choosing airlines in the same global alliance: “a) greater network access; b) seamless travel; c) transferable priority status; d) extended lounge access; and e) enhanced frequent-flyer program.” The first two can be regarded as benefits associated with codesharing partnerships, while the last three are benefits that are more relevant to global alliances in general. The offering of all these benefits would be most likely when the code-sharing partnership is formed between allied airlines. As outlined by Goh and Uncles (2003), the full awareness of these benefits, and their actual delivery, would increase customer loyalty to a particular airline or alliance group, a source of competitive advantage. In line with this argument, we hypothesize that profitability gains would be greater as an airline increases its code-sharing partnership with airlines in the same global alliance. Hypothesis 2. The operating margin of an airline is positively associated with the percentage of allied code-sharing partners (i.e., those in the same global alliance as the focal airline) it has compared its total number of code-sharing partners. For two airlines in a code-sharing relationship, the more routes that are code-shared, the more integrated are their route networks. On the supply side, the increased number of code-shared routes enables the partner airlines to add more destinations. This enhances the combined route networks, bringing greater efficiencies and cost saving benefits from improved traffic density to the

54

L. Zou, X. Chen / Journal of Air Transport Management 58 (2017) 50e57

partners. There are also several demand-side benefits to having a greater scope of code-sharing practices. According to Yousseff and Hansen, 1994, the quality and quantity of connecting services may increase because of improved scheduling coordination. As partner airlines have more routes codeshared with each other, the extended route network and will entice passengers to pay higher airfares as a result. Moreover, the benefits to passengers from the joint frequent flier programs offered by partner airlines are multiplied as code-sharing airlines increase the extent of their route integration. Finally, due to passengers’ incomplete information, a large route network can provide an airline with some marketing advantages as it is more likely to be selected by passengers. More specifically, the US GAO, (1995) writes the following while discussing potential factors influencing the benefits for participating airlines from code-sharing alliances.

annual observations from 2007 through 2012, including 136 observations for Star Alliance, 83 for Skyteam, 66 for Oneworld, and 201 for the non-aligned group. Table 1 provides the descriptions for key variables as well as summary statistics. Table 2 presents the correlations between all key variables. Figs. 1e3 present a comparison of average code-sharing characteristics, including the number of code-sharing partners, the percentage of allied code-sharing partners, and the percentage of comprehensive code-sharing partners among airlines in Star, Skyteam, Oneworld, and non-aligned airlines during the 2007e2012 period. The following empirical model is developed to test our three hypotheses.

lnðProfit Marginit Þ ¼ a0 þ a1 Number of CPit þ a2 Percent of Allied CPit

‘The extent to which airlines participating in alliances benefit from them varies greatly and depends on the (1) geographic scope of the code-sharing arrangement, (2) level of operating and marketing integration achieved by the airlines, and (3) agreement between the airlines on how to divide revenues.’

þ a3 Percent of Comprehensive CPit þ a4 lnðLoad Factorit Þ þ a5 lnðPassenger Yieldit Þ þ a6 lnðUnit Costit Þ þ a7 lnðASKit Þ þ a8 Alliance Dummiesit þ εit

Thus, our Hypothesis 3 is the following: Hypothesis 3. The operating margin of an airline is positively associated with the percentage of comprehensive code-sharing partners (i.e., those having a comprehensive code-sharing partnership with the focal airline) compared to its total number of code-sharing partners. In Hypothesis 3, the comprehensive code-sharing partnersrefer to those having ten or more code-shared routes. As explained in the following section, the data we use in the analysis is derived from the annual Airline Alliance Survey report published by Airline Business. The annual survey categorizes the type of code-sharing alliances as limited or comprehensive, with limited code-sharing alliances defined as agreements with up to 10 routes covered and comprehensive agreements with at least 10 routes covered.

(1)

In the regression model, the dependent variable is the operating profit margin for a focal airline i in year t. To test the three hypotheses, we develop three variables for measuring the scope and extent of code-sharing characteristics and the airline's embeddedness in the global alliances. These three independent variables are: (1) Number of code-sharing partners (denoted as Number of CPit for Airline i in year t); (2) The percentage of allied code-sharing partners (denoted as Percent of Allied CPit for Airline i in year t); and (3) The percentage of comprehensive code-sharing partners (denoted ad Percent of Comprehensive CPit for Airline i in year t). Several control variables are included, such as Load Factor, Passenger Yield, Unit Cost, and Available-seat kilometers (representing the total traffic volume the focal airline has in a given year). The Alliance Dummies are included as indicator variables to control for the alliance-specific effects.

4. Data and methodology 5. Results The data for our analysis is mainly collected from FlightGlobal and the Airline Alliance Survey published by Airline Business. In its annual Airline Alliance Survey report, the Airline Business magazine provides a list of codesharing agreements for each member airline of the Star, Skyteam and Oneworld Alliances. In addition, the report provides code-sharing agreement information for a group of leading non-aligned airlines. The code-sharing agreements compiled in the report are relevant to all airlines featured in the Airline Business Top 200 Passenger Rankings list. The alliance survey data is supplemented by airline performance and operating characteristics drawn from FlightGlobal. Using panel data for 81 airlines from 2007 to 2012, we estimate the effects on profit margin to a focal airline from: a) the number of codesharing partners; b) the percentage of allied codesharing partners; and c) the percentage of comprehensive codesharing partners, while controlling for other variables, such as load factor, passenger yield, unit cost, operating scale (measured by ASK), and global alliance membership. During the study period, the number of airlines in the Star Alliance increased from 22 to 27, in the Skyteam Alliance from 12 to 17, and remained 11 in the Oneworld Alliance. In our initial sample, there are a total of 92 airlines, including 24 in Star, 19 in Skyteam, and 11 in Oneword. The remaining 38 are nonaligned airlines. After the exclusion of airlines with missing performance data, our final sample covers 81 airlines with a total of 486

The model, Eq. (1), is estimated using a random generalized least square (GLS) procedure. The estimation results are presented in Table 3. Because of the nature of the panel data, we cannot assume that the effort terms in Eq. (1) are independent and identically distributed, and thus results based on ordinary least square (OLS) estimation would be vulnerable to issues such as heteroscedasticity, serial correlation and within panel or contemporaneous correlation across the panel. To detect these potential concerns, we first run the Breusch-Pagan/Cook-Weisberg test to check if heteroscedasticity is present. The results (c2(1) ¼ 570.52, p ¼ 0.0000) strongly reject the null hypothesis of homoscedasticity. In addition, the results from the Woodridge test (F(1, 52) ¼ 2.216, p ¼ 0.1426) fail to reject the null hypothesis of no autocorrelation. Considering our use of a relatively small panel with a large number of airlines, we use the random GLS estimation for its data efficiency and better fit. The results are summarized as follows. In Model 1, the effect of the number of code-sharing partners on operating profit margin is estimated while controlling for variables such as load factor, passenger yield, unit cost, available seatkilometers, and airline global alliance status. The coefficient for Number of Codesharing Partners is found to be positive (¼0.014) and highly significant (p < 0.001), suggesting that operating margin is positively related to the number of code-sharing partners,

L. Zou, X. Chen / Journal of Air Transport Management 58 (2017) 50e57

55

Table 1 Variable description and descriptive statistics. Variable

Description

Mean (Std. Dev.)

Profit margin Number of code-sharing partners Percent of allied code-sharing partners Percent of comprehensive code-sharing partners Unit cost ($) Yield ($) Load factor (%)

The operating profit margin (i.e., EBIT/Operating revenue) for airline i in year t The number of code-sharing partners airline i has in year t The percentage of code-sharing partners that are allied partners of airline i in year t The percentage of code-sharing partners that have comprehensive codesharing arrangements with airline i in year t The total operating expenses per available seat mile (CASM) for airline i in year t The total passenger yield (Cents/RPMs) including scheduled and non-scheduled passenger services The total RPMs divided by the total ASMs including both scheduled and non-scheduled revenue services operated by airline i in year t The total available seat kilometers for passenger services incurred by airline i in year t

0.012 (0.099) 14.07 (7.48) 0.61 (0.17) 0.61 (0.17)

Total ASKs (million)

0.094 (0.029) 0.104 (0.034) 0.742 (0.103) 58,137 (63,175)

Table 2 Correlation matrix for key variables.

1. 2. 3. 4. 5. 6. 7. 8.

Profit margin Number of code-sharing partners Percent of allied code-sharing partners Percent of comprehensive code-sharing partners Unit cost Yield Load factor Total ASKs

1

2

3

4

5

6

7

8

1.00 0.07 0.11 0.05 0.14 0.02 0.24 0.02

1.00 0.19 0.13 0.32 0.17 0.25 0.34

1.00 0.37 0.03 0.05 0.02 0.05

1.00 0.06 0.09 0.42 0.53

1.00 0.89 0.10 0.01

1.00 0.27 0.10

1.00 0.51

1.00

20

50%

18

45%

16

40%

14

35%

12

30%

10

25%

8

20%

6

15%

4

10%

2

5% 0%

0 2007

2008

2009

2010

2011

2012

2008

2009

2010

2011

Avg. # of codesharing partners of airlines in Star

Avg. % of comprehensive codesharing partnership for airlines in Star

Avg. # of codesharing partners of airlines in Skyteam

Avg. % of comprehensive codesharing partnership for airlines in Skyteam

Avg. # of codesharing partners of airlines in Oneworld

Avg. % of comprehensive codesharing partnership for airlines in Oneworld

Avg. # of codesharing partners of non-aligned airlines

Avg. % of comprehensive codesharing partnership for non-aligned airlines

Fig. 1. The average number of codesharing partners.

0.7 0.6 0.5 0.4

0.3 0.2 0.1 0 2008

2009

2010

2011

2012

Fig. 3. The average percentage of comprehensive codesharing partnership.

0.8

2007

2007

2012

Avg. % of allied of codesharing partners of airlines in Star Avg. % of allied of codesharing partners of airlines in Skyteam Avg. % of allied of codesharing partners of airlines in Oneworld

Fig. 2. The average percentage of Allied codesharing partners.

consistent with Hypothesis 1. More specifically, the results reveal that an airline will increase its profit margin by 0.14% with an addition of 10 code-sharing partners. The estimated coefficients for

most control variables are statistically significant and show expected signs. In particular, Load Factor and Passenger Yield are found to have positive effects on profitability whereas Unit Cost is negatively related to profit margin. The coefficient for available seat-kilometers is found statistically insignificant suggesting that operating scale by itself may not impact an airline's profitability after taking into account the effects from unit cost, load factor, and passenger yield. Since the estimation of Model 1 is based on a complete set of 81 airlines including those members of a global alliance, the regression results for the dummy variables representing global alliances status indicate that compared to non-allied airlines, only member airlines of Oneworld, on average, have greater profitability. The other two global alliances (i.e., Star and Skyteam) do not directly impact the profitability of their member carriers, on average. In Model 2, the variables, Number of code-sharing partners and Percent of Allied Code-sharing Partners, are both included as explanatory variables in estimating profit margin. The inclusion of Percent of Allied Code-sharing Partners is necessary for testing

56

L. Zou, X. Chen / Journal of Air Transport Management 58 (2017) 50e57

Table 3 The random-effects GLS estimation results for Ln (operating profit margin). Independent variable

Model I

Model II

Model III

Model IV

Number of code-sharing partners Percent of allied code-sharing partners Percent of comprehensive code-sharing partners Ln (Load factor) Ln (Passenger yield) Ln (Unit cost) Ln (ASKs) Star alliance Skyteam alliance Oneworld alliance Constant No. of observations Wald c2 Prob > c2 Overall R2

0.014*** (0.004)

0.017*** (0.005) 0.226* (0.140)

0.014*** (0.004)

2.297*** (0.274) 1.761*** (0.122) 1.997*** (0.125) 0.011 (0.031) 0.071 (0.074) 0.053 (0.068) 0.162* (0.088) 0.946** (0.371) 338 300.69 0.0000 0.35

1.865*** (0.397) 1.945*** (0.152) 2.202*** (0.165) 0.009 (0.036) 0.351*** (0.105) 0.273*** (0.101)

0.016*** (0.005) 0.256* (0.146) 0.126 (0.157) 1.887*** (0.399) 1.935*** (0.152) 2.187*** (0.166) 0.002 (0.038) 0.332*** (0.106) 0.263*** (0.099)

1.086** (0.465) 232 196.59 0.0000 0.30

0.027 (0.100) 2.286*** (0.278) 1.762*** (0.122) 2.001*** (0.126) 0.013 (0.032) 0.077 (0.077) 0.056 (0.069) 0.162* (0.088) 0.938** (0.374) 338 300.29 0.0000 0.34

1.176** (0.472) 232 194.80 0.0000 0.31

The numbers in parentheses are standard errors of coefficients. *** Significant at 0.01 level; ** Significant at 0.05 level; * Significant at 0.1 level.

Hypothesis 2. As can be seen from Column 3 of Table 3, this newly added variable has a positive (¼0.226) and moderately significant (p < 0.10) coefficient, suggesting that the higher the proportion of allied code-sharing partners an airline has, the greater its profit margin. Thus, Hypothesis 2 is supported. The estimation results for other control variables are similar to those found in Model 1. Considering that Model 2 is estimated using data only for airlines in the three global alliances, the default case for comparison is member airlines of Oneworld. Therefore, the negative and significant coefficients for Star and Skyteam can be explained as an indication that among the three global alliances, members of Oneworld, on average, have higher profit margins than their counterparts in Star and Skyteam. To test Hypothesis 3, the variable Percent of Comprehensive Codesharing Partners is incorporated as a regressor, along with Number of Code-sharing Partners, in Model III. Consistent with the findings from Models I and II, the coefficient for Number of Code-sharing Partners is positive (¼0.014) and highly significant (p < 0.001). However, the variable representing the level of code-sharing partnership, Percent of Comprehensive Code-sharing Partners, is positive (¼0.027) but insignificant, suggesting that an airline can enhance its profitability through increasing the number of code-sharing partners without regard to whether the partnership is comprehensive or limited. Though this result does not support Hypothesis 3, the conclusion may be circumspect given limitations in the variable, Percent of Comprehensive Code-sharing Partners, used for measuring the extent of code-sharing partnership. In Model IV, we include all the three alliance variables and other control variables. The results are very similar to those found in Models I-III. In summary, our empirical results validate the performance benefits for an airline from developing more code-sharing partnerships with other airlines. Moreover, profitability gains for partner airlines resulting from their code-sharing partnerships are found to be higher when the two partners are airlines in the same global alliance. Hence, for an airline planning to develop codesharing partners, it is optimal to choose those in the same global alliance for greater operating margin improvement. Overall, the implication from this study assures airline management of the performance benefits from joining global alliances in conjunction with forming code-sharing partnership with allied carriers. 6. Conclusions, implications and limitations In this paper, we investigate the impacts of code-sharing alliances on an airline's operating margin. Using panel data for 81 major airlines from 2007 through 2012, we find evidence showing

that there is a highly significant and positive relationship between the number of code-sharing partners and an airline's operating margin. This result supplements the findings of Iatrou and Alamdari (2005) based on their survey data that code-sharing alliances provide airlines with significant revenue gains but provide few cost impacts. From a practical perspective, our results provide confidence to managers to increase their use of code-sharing alliances as a strategy for enhanced profitability, consistent with the recent trend in the airline industry. Take Etihad Airways as an example. Although the airline has not joined a global alliance, it has formed a growing number of code-sharing alliances with other carriers in recent years. According to the article “Etihad uses partnership to beat 2011” from Airline Business magazine (Bonnassies, 2012), Etihad Airways had two code-sharing partners in 2008 contributing to 1% of the airline's revenue. By 2012, it had thirtyfive code-sharing arrangements with thirty-five airlines accounting for 19% of the airline's total revenue. Moreover, the results indicate a statistically significant and positive association between operating margin and the percentage of code-sharing arrangement with allied airlines. This finding has two important managerial implications. First, for airlines in a global alliance, it suggests that the benefits from code-sharing arrangements will be amplified as they develop more code-sharing partnerships with their existent allies in the global alliance. The positive association between greater profit margin and a higher percentage of allied code-sharing partners is consistent with the finding by Oum et al. (2004) that there is no substantial profit benefits for airlines in global alliances unless the alliances involve some highlevel cooperation. It can be shown from our study that through code-sharing partnerships, allied airlines in global alliances can develop a higher level of cooperation for greater profit gains. Therefore, for an airline that is already participating in one of the three global alliances, the implication is that it will benefit more if it chooses those allied airlines in the same global alliance as its codesharing partners. On the other hand, for airlines that are not members of global alliances, it may be beneficial for them to consider joining the global alliance that contains the greatest number of their existent code-sharing partners. As suggested by our results, through global alliances allied code-sharing partners may develop a higher level of operating and marketing integration, and thus the profit margin benefits from code-sharing partnership can be enhanced. Finally, consistent with previous literature, we find differential performance among the three global alliances. Member airlines of Oneworld appear to have a greater operating margin, on average, than those in the Star and Skyteam Alliances. The use of

L. Zou, X. Chen / Journal of Air Transport Management 58 (2017) 50e57

comprehensive code-sharing partnerships, however, is not found to have the expected positive effect on an airline's operating margin. It is important to note that this insignificant result may be due to the limited measurement we employ for comprehensive code-sharing arrangements. Future research could develop an alternative variable to represent in a more precise manner the geographic scope of a code-sharing arrangement between partner airlines. To achieve this goal, the use of the data detailing code-sharing agreements at the route level would be very helpful. References Bonnassies, O., 2012. Etihad uses partnerships to beat 2011. Airl. Bus. 28 (9), 16. Brueckner, J.K., Whalen, W.T., 2000. The Price effects of international airline alliances. J. Law Econ. 43, 503e545. Brueckner, J.K., 2001. The economics of international codesharing: an analysis of airline alliances. Int. J. Ind. Organ. 19, 1475e1498. Brueckner, J.K., Zhang, Y., 2001. A model of scheduling in airline networks: how a hub-and-spoke system affects flight frequency, fares and welfare. J. Transp. Econ. Policy 35 (2), 195e222. Brueckner, J.K., 2003. International airfares in the age of alliances: the effects of codesharing and antitrust immunity. Rev. Econ. Statistics 85 (1), 105e118. Chen, M., 2000. The effects of strategic alliances on airline profitability. Econ. Issues 5, 87e95. Dresner, M., Flicop, S., Windle, R., 1995. Trans-Atlantic airline alliances: a preliminary evaluation. J. Transp. Res. Forum 35 (1), 13e25. Dresner, M., Windle, R., 1996. Alliances and code-sharing in the international airline industry. Built Environ. 22 (3), 201e211. Dresner, M., 2010. The Economics of Airline Alliances. Critical Issues in Air Transport Economics and Business. Published by Routledge. Gagnepain, P., Marin, P.L., 2010. The effects of airline alliances: what do the aggregate data say? SERIEs 1, 251e276. Gayle, P.G., Brown, D., 2014. Airline strategic alliances in overlapping markets:

57

should policymakers be concerned? Econ. Transp. 3, 243e250. Gayle, P.G., Le, H.B., 2013. Airline Alliances and Their Effects on Costs. Kansas State University. Working Paper. Goh, K., Uncles, M., 2003. The benefits of airline global alliances: an empirical assessment of the perceptions of business travelers. Transp. Res. A 37, 479e497. Goh, M., Yong, J., 2006. Impacts of code-share alliances on airline cost structure: a truncated third-order translog estimation. Int. J. Ind. Organ. 24, 835e866. Iatrou, K., Alamdari, F., 2005. The empirical analysis of the impact of alliances on airline operations. J. Air Transp. Manag. 11 (3), 127e134. Ito, H., Lee, D., 2007. Domestic code sharing, alliances, and airfares in the US airline industry. J. Law Econ. 50 (2), 355e380. Morrish, S.C., Hamilton, R.T., 2002. Airline alliances e who benefits? J. Air Transp. Manag. 8 (6), 401e407. Oum, T.H., Park, J.H., Zhang, A., 1996. The effects of airline codesharing agreements on firm conduct and international air fares. J. Transp. Econ. Policy 30, 187e202. Oum, T.H., Park, J.H., Kim, K., Yu, C., 2004. The effect of horizontal alliances on firm productivity and profitability: evidence from the global airline industry. J. Bus. Res. 57, 844e853. Park, J.H., 1997. The effects of airline alliances on markets and economic welfare. Transp. Res. Part E 33 (3), 181e195. Park, N.K., Cho, D.S., 1997. The effect of strategic alliance on performance: a study of international airline industry. J. Air Transp. Manag. 3 (3), 155e164. Park, J.H., Zhang, A., 2000. An empirical analysis of global airline alliances: cases in North Atlantic market. Rev. Ind. Organ. 16, 367e383. Pitfield, D.E., 2007. The impact on traffic, market shares and concentration of airline alliances on selected European-US routes. J. Air Transp. Manag. 13, 192e202. US General Accounting Office, 1995. International Aviation: Airline Alliances Produce Benefits, but Effect on Competition Is Uncertain. GAO/RCED-95e99, April, 1995. Wan, X., Zou, L., Dresner, M., 2009. Assessing the price effects of airline alliances on parallel routes. Transp. Res. Part E 45 (4), 627e641. Yousseff, W., Hansen, M., 1994. The consequences of strategic alliance between international airlines: the case of Swissair and SAS. Transp. Res. Part A 28 (5), 415e431. Zou, L., Oum, T.H., Yu, C., 2011. Assessing the price effects of airline alliances on complementary routes. Transp. Res. Part E 47 (3), 315e332.