Airline Code-Sharing and Capacity Utilization: Evidence from the US Airline Industry.
Following the passage of the Airline Deregulation Act in 1978, the degree of competition among air carriers, increased significantly. Airlines resorted to many ways--including code-sharing--in response to competitive pressures. This article is a retrospective study on the load factor effects of the largest domestic codeshare partnership established in 2003--between Delta, Northwest, and Continental airlines. After controlling for aircraft configuration, carrier, and market characteristics, we find evidence that code-sharing improves the partners' load factors relative to other carriers in all markets combined. However, we find statistically significant negative (positive) codeshare effects on load factor in markets where the codeshare partners competed (did not compete) prior to code-sharing.
Airlines, load factor, codeshare agreement, competition, alliance
JEL classification codes: L93; L13
Airline alliances are marketing arrangements between two or more air carriers allowing them to cooperate substantially across various levels of their operations. An alliance formation is generally a strategy that air carriers use when outright merger or other means of scaling their operations is not an option. Even when other means of scaling operations are available, alliances still provide quicker access to new markets without incurring entry costs (Iatrou and Alamdari 2005). Alliances among airlines can be as simple as reciprocal frequent-flyer programs where passengers can earn and redeem miles across partner airlines, or more comprehensive as code-sharing where an airline can sell seats on flights operated by its codeshare partner (US General Accounting Office 1999).
Codeshare agreements have a long history in the international market for air travel as the airline industry around the world experienced major shocks due to globalization, (de)regulation--cabotage (Chen and Chen 2003). Conversely, alliance formation among US domestic carriers is relatively recent. Over the last couple of decades, we have seen a proliferation of cooperative initiatives between airlines, perhaps the most prominent being codeshare partnerships. The US domestic airline industry witnessed an uptick in code-sharing in the mid-1990s, as major airlines used it (code-sharing) as a strategy to fend off increased competition from low-cost carriers (Mantovani and Tarola 2007). Consequently, in 1995, Northwest and Hawaiian Airlines submitted a codeshare proposal to the transportation authorities (Ito and Lee 2005). Following suit, the six largest US carriers unveiled their plans to codeshare: Continental/Northwest in January 1998 and Delta/United and American/US Airways in April 1998 (Bamberger, Carlton, and Neumann 2004). In October 2001 Alaska and Hawaiian finalized their codeshare agreement while American West and Hawaiian struck a codeshare arrangement in October 2002. Finally, the implementation of similar partnerships ensued with United/US Airways in January 2003 and Delta/Northwest/Continental in June of the same year.
Codeshare agreements between airlines effectively allow a consumer to purchase a codeshare product from one airline (the ticketing carrier) but actually fly segment(s) of the trip on plane(s) operated by carrier(s) that are partner(s) to the ticketing carrier. To facilitate the offering of such interline codeshare products, partner carriers typically coordinate their operations to achieve a seamless integration of their route networks, which potentially may result in more travel-convenient routing across partner carriers' networks (Yimga 2017). (1)
Airline capacity utilization has received much attention in the literature. (2) Airlines have always been concerned about how optimally they can use their planes in order to maximize passenger traffic density and revenue. Recent data from the US Department of Transportation (DOT) show that in May 2016, aggressive capacity growth has produced adverse effects resulting in falling capacity utilizations for many airlines. For example, United Airlines (UA) witnessed the largest fall in load factor (1.494 year over year) while American Airlines (AA) also saw a drop of 0.9 percent in load factor. Load factor is widely recognized as the main measure of operations performance in the airline industry (Bailey et al. 1985).
From an operations standpoint, load factor is an indication of how capacity, or available seat miles (ASM), responds to air travel demand shocks in a specific market. The ASM adjusts to changes in flight frequency or the use of aircraft of different sizes. In few occasions, the ASM may not respond to changes in demand especially in routes with limited frequency where carriers have very little leverage of canceling flights without damaging the overall routing network (Chen and Chen 2003). Usually, lower capacity or ASM leads to a higher load factor, which means more efficient operations and increased economies of passenger traffic density. On the other hand, given the uncertainties (some caused by code-sharing) associated with air travel demand, a high load factor may indicate low flight frequency, suggesting a greater likelihood of seat shortage and potential loss of revenues. Therefore, finding the optimal seat supply on a particular route means dealing with the delicate tradeoff between the advantages of efficient operations and the downsides of low-grade product convenience. The main purpose of this article is to explore these mechanisms by empirically investigating the load factor impacts of a highly contentious codeshare agreement between three major carriers.
In our empirical exercise, we examine the load factor effects of the tripartite codeshare agreement between Delta, Northwest, and Continental airlines. We choose this codeshare alliance for the following reasons: (a) it involves three major full-service carriers; (b) the alliance was the largest to be ever approved in the history of the US airline industry; and (c) the alliance was one of the most legally contentious. We further explore load factor effects of code-sharing based on the existence of pre-alliance competition between the codeshare firms. This allows us to measure differences in capacity utilization in both markets where the partners competed (we also refer to these markets are overlapping, throughout the article) prior to code-sharing, and markets where they did not compete.
Our analysis delivers a number of compelling results. First, we find asymmetric load factor responses to code-sharing. Overall, code-sharing increases load factor. These load factor effects are robust even after we control for aircraft configuration, carrier, and market characteristics. The increased load factor is possibly owing to higher quality product offerings in terms of better variety of seamless connecting flights, greater opportunities for passengers to earn and redeem frequent-flyer miles, and so on (Gayle 2007). Second, and surprisingly, we find statistically significant negative (positive) codeshare effects on load factor in markets where the codeshare partners competed (did not compete). However, upon further analysis, the evidence does not suggest that the codeshare agreement facilitated collusive behavior on the partners' overlapping routes. Besides the policy implications, these results provide managerial significance in terms of codeshare partner selection, network integration, and flight coordination.
The rest of the article is organized as follows. The next section reviews the literature on the effects of airline alliances, followed by an overview of the Delta, Northwest, Continental codeshare agreement. We then discuss the research methodology and estimation technique. Results are reported before our concluding remarks.
The effects of airline alliances in general and code-sharing in particular have received a lot of attention and have been studied by a number of researchers. Oum, Park, and Zhang (1996) provide an excellent review of the effects of airline strategic alliances. The effects that have been examined include welfare and competition, traffic demand, airfare, partner selection, output and profit, and organizational change (Brewer and Hooper 1998; Deprosse and Franke 1998; Gellman Research Association 1994; Nyathi 1995; Oum and Park 1997; Park 1997; Park and Zhang 1998; US General Accounting Office 1995; Youssef and Hansen 1994; Zea and Feldman 1998).
A large number of studies have documented the pricing effects of airline alliances (Bamberger, Carlton, and Neumann 2004; Brueckner 2001, 2003; Brueckner and Whalen 2000; Gayle 2008, 2013; Gayle and Brown 2014; Ito and Lee 2007). These studies argue, on the one hand, that airline cooperation from an alliance pushes airfares down in interline markets because of product complementarity and the mitigation of double marginalization. Brueckner (2003) and Brueckner and Whalen (2000) note that code-sharing allows airlines to eliminate a double markup on itineraries with multiple operating carriers, resulting in lower fares. Ito and Lee (2005) also show code-sharing to be associated with lower fares. On the other hand, an alliance can also reduce competition in markets where the partners' route networks overlap (interhub markets, for instance), and this in turn puts pressure on airfares to rise in these markets. Zou, Oum, and Yu (2011) argue that it is possible that an alliance causes fares to rise even in markets where the partners' route segments are complementary rather than overlapping. They make the argument that the improvement in the quality of interline connections that comes with an alliance would subsequently increase demand.
Bilotkach (2005) examines airline alliance formation using transatlantic markets to determine if code-sharing with and without antitrust immunity decreases fares for interline trips equally. The findings suggest that code-sharing and alliance formation both have fare-decreasing effects; however, the code-sharing effect turns out to be more than twice the magnitude of the alliance effect.
Tiernan, Rhoades, and Waguespack (2008) explored the linkage between on-time performance and international airline alliance. Yimga (2017) investigated the on-time performance effects of code-sharing in the US domestic airline industry and found that a codeshare agreement improves the on-time performance of the alliance carriers relative to other airlines. However, this improvement in on-time performance is a function of the degree of competition among the codeshare partners, with the code-share effects on on-time performance being larger in markets where the partners competed in prior to the alliance.
Surprisingly, questions regarding the impact of code-sharing on operations performance measures such as capacity utilization have been left largely unanswered. Are airlines that participate in a codeshare partnership enjoying higher load factors? If so, how large are these load-factor effects? Are these load-factor effects persistent across different types of market? This article attempts to answer these questions by empirically investigating the capacity utilization of the largest domestic alliance--between Delta, Northwest, and Continental airlines--with a particular focus on the codeshare effects in markets where the alliance partners competed and markets where they did not compete prior to code-sharing.
This study contributes to the broad literature of the demand-enhancing potential of airline alliances in general and code-sharing in particular. Inherent features of codeshare arrangements, such as consumers' ability to accumulate and redeem miles across codeshare partners' frequent-flyer programs, build brand loyalty (Evans and Kessides 1993) and increase demand (Chin 2002; Lederman 2008). Lederman (2007) investigated whether enhancements to loyalty programs affect demand by measuring the impact of international frequent-flyer partnerships (FFP) on domestic airline demand. Her findings indicate that enhancements to an airline's FFP are associated with increases in its demand on specifically those routes that depart from airports at which the airline is dominant. Using descriptive analysis of airline operations, Dresner, Flicop, and Windle (1995) evaluate the effects of three trans-Atlantic airline alliances on traffic volume and load factors. Their analysis indicates that in only one of the three agreements have the carriers increased their trans-Atlantic traffic volume and increased their load factors after realigning their route systems to take advantage of the alliance.
Chen and Chen (2003) develop a theoretical model to analyze the impact of different types of global airline alliances on load factor under demand uncertainty. The results derived from their theoretical model were tested using empirical data from the airline industry show that parallel (rather than complimentary) code-sharing results in increased demand.
Background and Definitions
Background of the Delta/Northwest/Continental Codeshare Agreement
On August 23, 2002, Delta, Northwest, and Continental submitted their code-share agreement proposal to the DOT for review. The three legacy carriers request a comprehensive cooperative agreement that involves code-sharing, reciprocal frequent-flyer programs and reciprocal access to airport lounges. The partner carriers claimed, at the time, that the agreement will benefit consumers in the form of seamless service to thousands of new markets, broader availability of low-priced seat, increased frequency of flights, and better time-of-day coverage for travelers, among other benefits. However, the DOT determined that the agreement, if implemented as presented by the three airlines, could result in significant adverse impacts on airline competition unless the airlines agreed to a number of conditions that would limit the likelihood of competitive harm (US Department of Transportation 2003).
Regulators were alarmed by two major issues. First, they expressed concerns about the substantially high combined market share between the three airlines (35%), which was larger than the 23 percent market share of the United/US Airways codeshare agreement that was in operation at the time. Second, they believed that the three carriers had a large number of markets in which their service overlap (3,214 markets) with the potential of affecting nearly 58 million passengers per year (Gayle and Le 2013). These concerns, regulators believed, could potentially facilitate collusion (explicit or tacit) on prices and/or service levels in the partners' overlapping markets.
The codeshare carriers complied with some of the conditions and acknowledged that due to colocation, they will relinquish four gates in Houston (IAH), two gates in Detroit (DTW), five gates in Cincinnati (CVG), and two gates at Dallas Fort Worth (DFW) for common use. Additionally, if the codeshare partners choose to co-locate additional gates at any of the hub airports of any partner or Boston (BOS), the relocating carrier will promptly notify the DOT, and if such co-location creates additional surplus gates (in a manner comparable to the situation at IAH, DTW, CVG or DFW) the partners agree to release any such surplus leased gates.
The Department of Justice's (DOJ) antitrust review concluded that the codeshare agreement would drive down costs for consumers and provide better service, and it approved the alliance on January 17, 2003. It took about six months for the airline partners to meet regulators' conditions, and in June 2003, operating certificates were issued, officially marking the completion of the codeshare partnership.
A couple of terminologies are worth noting for future reference throughout the article. These terms follow the definitions from Gayle and Thomas (2015), among others. A market is defined as a directional origin-destination combination. Following Berry, Carnall, and Spiller (2006) and Gayle (2007), markets are defined directionally, meaning that air travel from San Francisco to Boston is a different market than air travel from Boston to San Francisco, for instance. This allows us to separately account for origin and destination city characteristics. Also, we group cities that belong to the same metropolitan areas when an endpoint (origin or destination) is located in a metropolitan area with multiple airports.
A product is defined as a combination of origin, destination, and airline(s). Each flight has a ticketing carrier and an operating carrier. The ticketing carrier is the airline that issues the flight ticket to consumers while the operating carrier is the airline that owns/operates the aircraft that transports passengers. While all products have both an operating carrier and ticketing carrier, these carriers may be different for a given product.
Since our objective is to study the effects of code-sharing on load factor, we control for virtual codeshare products. A virtual codeshare product has a single ticketing carrier and operating carrier for the itinerary, but the operating and ticketing carriers are different. For example, a flight is ticketed through Delta, but the passenger flies on a Northwest plane. (3)
To measure the codeshare effects on capacity utilization, we estimate reduced-form regression equations of capacity utilization while controlling for product, carrier, airport, and market-structure characteristics. Possible codeshare effects on capacity utilization are identified using a difference-in-differences identification strategy. This strategy enables us to compare pre-post codeshare periods' changes in capacity utilization of the codeshare firms, relative to changes in capacity utilization of other firms over the same pre-post codeshare periods.
Data are gathered from the T100 Domestic Segment, published by the DOT. This dataset is a monthly sample of all flights in the United States. Each observation in the dataset represents a flight and includes information on carrier, origin, destination, aircraft type, transported passengers, freight and mail, available capacity, departures performed, aircraft hours, and load factor.
Moreover, because load factor is only measured for individual flights, we restricted our analysis to nonstop service. We collect monthly data for nonstop domestic flights from June through December of 2002 (pre-codeshare) and 2004 (post-codeshare) since the codeshare agreement between Delta, Northwest, and Continental was approved on August 23, 2003. We choose the sample period for three reasons. First, September 2001 substantially changed both the supply-side and the demand-side incentives for air travel (see, for example, Berry and Jia 2010 and Forbes, Lederman and Tombe 2015). We collected data starting from July of 2002 to mitigate some of the post-9/11 effects. For seasonality reasons, we collected data for the same six months in 2004 for post-codeshare period. Second, the time gap between the pre- and post-codeshare periods was included to account for the alliance implementation lag, which can take up to six months before codeshare products may appear in the DOT's database. Third, the six months before and after the codeshare agreement, though short, helps isolate potential confounding factors that are likely to change market structure (hence affect load factor) over a longer time span.
Our analysis focuses on air travel among the cities within the 48 mainland states of the United States. Following Aguirregabiria and Ho (2012) and Berry (1992), products purchased by fewer than nine passengers transported throughout an entire month were omitted. We report the airlines in our sample in table Ai in the appendix.
Route Level Statistics of DL/NW/CO Codeshare Agreement
Table 1 reports code-sharing activity among the three codeshare partners at the route level. Northwest assumed the role of operating carrier for Continental and Delta on 439 and 472 routes, respectively. Continental operated 245 routes for Delta and 269 routes for Northwest. Delta operated 238 and 283 routes on behalf of Continental and Northwest, respectively. Northwest disproportionately operated for Delta and Continental while Delta and Continental ticketed the majority of the codeshare products.
Table 2 shows a more disaggregated analysis of the code-sharing activity among the three partners in routes where their services overlap. The diagonal of table 2 reports the total directional overlapping routes for each pair. Sixty percent (290) of the routes Northwest operated on behalf of Delta and Continental overlap, reinforcing Northwest's dominant operating carrier role. Ninety overlapping routes were operated by Continental on behalf of the other partners while Delta only operated 57 routes.
Measuring Capacity Utilization
We directly measure capacity utilization by computing the load factor for each product in our sample. We define load factor as the ratio of passenger miles to available seat miles. Following Yimga (2017), among others, our analysis is carried out at the carrier-market-month-year level. Over a given month, a carrier may operate a specific origin-destination multiple times, with different load-factor values. We construct our capacity utilization measure by averaging the load factor of products that belong to a given carrier, in a given market, during a specific month and year. We then collapse the data by carrier-origin-destination-month-year combinations and report the group mean of the codeshare partners and other carriers in figure A1 in the appendix. Figure 1 shows that the distribution of load factor is skewed to the left suggesting that the mean load factor (0.613) is less than the median load factor (0.650).
We average explanatory variables and collapse them using the same approach for consistency. Our final working data set has 147,120 usable observations, where an observation is at the level of carrier-origin-destination-month-year combination.
Table 3 reports a preliminary difference-in-differences estimation of load factors before and after the codeshare agreement for codeshare partners versus other carriers. Table 3 indicates that even though the mean load factor improves overall over the pre-post alliance periods for all carriers, the increase in mean load factor is larger for the codeshare partners' products. We test for the difference in mean load factors between codeshare partners and other carriers in column 4 of table 3 and report the difference-in-differences estimate (bold) in column 1. The statistical significance reported in column 4 suggests that the codeshare agreement improved load factor for the codeshare partners relative to other airlines over the pre- and post-codeshare periods.
It should be acknowledged that the difference-in-differences analysis summarized in table 3 has caveats and provides only rough estimates of the effects associated with code-sharing between the three airlines in all markets combined. For instance, these difference-in-differences calculation does not account for persistent shocks that may differ across markets or the differential changes in overlapping versus nonoverlapping markets. In measuring the load-factor effects associated with code-sharing between the three airlines, a formal econometric model presented below, while not perfect, will do a better job at controlling for potential differences in exogenous control variables across markets.
Our empirical analysis consists of a series of regressions of our dependent variable--capacity utilization (measured by load factor)--on control variables described below.
Our basic reduced-form estimating equation is:
[DepVar.sub.armt] = [alpha] + [beta][X.sub.armt] + [delta][W.sub.armt] + [[lambda].sub.a] + [[mu].sub.m] + [[eta].sub.t] + [origin.sub.r] + [dest.sub.r] + [[epsilon].sub.armt] (1)
where a indexes airlines, r indexes markets, m indexes months and t indexes years. [DepVar.sub.armt] is our dependent variable (load factor), [X.sub.armt] is a vector of control variables, [W.sub.armt] is a vector of dummy variables that captures the codeshare effects. [[lambda].sub.a]'s are airline specific fixed effects, [[mu].sub.m]'s are month fixed effects, [[eta].sub.t]'s are year fixed effects, origin and destination airport specific fixed effects are denoted by [origin.sub.r] and [dest.sub.r], [[epsilon].sub.armt] is the unobserved part of load factor.
The vector [W.sub.armt] includes the following: (1) a dummy variable for codeshare products among Delta, Northwest, and Continental, (2) a post-codeshare time dummy variable, and (3) an interaction term of these dummies. This is equivalent to a difference-in-differences identification strategy where codeshare effects are identified by comparing pre-and post-codeshare periods' changes in load factor of products operated by the codeshare partners, relative to changes in load factor of products operated by other firms over the same pre- and post-alliance periods.
To isolate the effects of code-sharing on load factor, we construct a number of variables to control for factors that may influence flight load factor. First, we control for the level of congestion at an airport by including a dummy variable (Slot Control) in the analysis to control for slot-controlled airports (Yimga 2017). This is a dummy variable coded 1 if the origin and/or destination airports of a given product are slot-controlled. The slot-controlled airports in our data sample are New York LaGuardia, New York Kennedy, Washington National, and Chicago O'Hare.
Second, we control for market distance by including the nonstop flight distance between market city-pairs.
Third, we include a measure of an airline's presence at the market origin city. This variable allows us to draw inference on how economies of passenger traffic density--measured indirectly by the number of departures performed in a market by a carrier during a given month--might affect flight load factor (Chen and Chen 2003).
Fourth, since load factor is also a function of available seats in an aircraft, we construct controls for aircraft types since different types of planes have different sizes and capacities (Chen and Chen 2003). We obtained aircraft type information from the T-100 database of the DOT and report the aircraft types in our working sample in table A2 (see appendix).
Fifth, to control for market structure, we include a dummy variable that equals 1 if the market share of any given carrier in a market is 0.5 or less. This variable captures the competitiveness of a market and was modeled in the spirit of Forbes (2008). This dummy variable is determined by using the market share based on the number of passengers transported in a given directional origin and destination city pair.
Sixth, another measure of competitive conditions we include in our analysis is the Herfindahl-Hirschman Index (HHI). Unlike the competitiveness variable, which does not capture the heterogeneity in market structures since it is a dichotomous variable, the HHI is a continuous measure of market concentration (Gayle 2007). The HHI of each market is calculated by taking the sum of the squared market shares (of passengers transported) of all carriers in a market.
Seventh, we include a series of dummy variables and interaction variables from these dummies to capture the codeshare effects on load factor. We create the dummy variable [T.sup.dnc] to indicate the post-codeshare period and the dummy variable DNC to identify products that are code-shared by Delta, Northwest, or Continental. We also create the dummy [MKT.sup.dnc], which is a zero-one market-specific dummy variable that takes the value of one for origin-destination markets in which any two of the three codeshare partners competed prior to code-sharing.
Finally, we include a series of monthly dummy variables that controls for seasonality in air travel. We report the summary statistics in table 4.
Empirical Results and Discussion
Impact of Measured Determinants of Capacity Utilization
In this section, we report empirical analyses of the impact of some measured determinants and code-sharing on load factor. The results are presented in table 5.
The coefficient estimate on Departures Performed is negative and statistically significant suggesting that higher flight frequency results in lower load factor. Higher flight frequency provides an array of options to consumers which in turn may drive down the average load factor per flight.
The level of congestion at the airport is probably one of the most important factors that may explain variations in capacity utilization. The coefficient estimate on Slot Control captures the amount of congestion at slot-controlled airports. The coefficient is negative and statistically significant. This finding is consistent with our expectations since Slot Control captures possible travel inconveniences for passengers due to airport traffic congestion at slot-controlled airports.
Building on existing work examining the relationship between air travel service and market structures, we investigate how market competitive conditions affect carriers' load factor. We control for market-level competition by including the dummy variable Competitive Market. The coefficient estimate on Competitive Market is negative and statistically significant suggesting that carriers that operate in competitive markets experience negative demand shocks.
The coefficient estimate on the HHI is positive and statistically significant, reinforcing the fact that carriers that operate in highly concentrated (less competitive) markets enjoy higher load factors.
The positive and statistically significant coefficient estimate on Distance indicates that market distance is a predictor of load factor, suggesting that carriers that operate longer market distances tend to have fuller flights.
Seasonality is another important factor that explains changes in load factor. To control for seasonality, we included month dummy variables (omitting July for singularity issues). The month dummy variables are statistically significant suggesting that seasonality play a major role in determining load factor.
Overall Capacity Utilization Effects of the Codeshare Agreement
We now turn to the remaining rows of table 5 that contain key variables in examining the effects of the codeshare agreement on load factor. To achieve our ultimate objective of correctly identifying the codeshare effects on load factor, it is imperative to control for the determinants of load factor discussed above. Furthermore, since we are using a difference-in-differences identification strategy, it is also important to control for persistent differences in load factor across carriers (Chen and Gayle 2018). Such controls are especially crucial if the load factor of products offered by the code-share firms are persistently different from load factor of products offered by other firms in our sample. If we fail to control for persistent load factor differences, we may erroneously assign measured differences in load factor to the codeshare agreement.
To measure persistent differences in load factor of codeshare products operated by the partners and other products, we include a dummy variable DNC that identifies products that are code-shared by Delta, Northwest, or Continental. The positive and statistically significant coefficient estimate on DNC suggests that the mean load factor of products associated with Delta, Northwest, or Continental is greater than the mean load factor of products offered by other carriers in the sample.
The coefficient estimate on [T.sup.dnc] measures, on average, how load factor changes over the pre-post codeshare agreement period for products that are not associated with Delta, Northwest, or Continental airlines. The positive coefficient estimate on [T.sup.dnc] indicates that the mean load factor of products operated by airlines other than Delta, Northwest, and Continental airlines increased over the pre--and post-codeshare periods.
Finally, the interaction term [T.sup.dnc] x DNC identifies the difference-indifferences estimate that measures whether load factor of products ticketed/operated by the codeshare carriers changed differently relative to load factor of products ticketed/operated by other airlines. [T.sup.dnc] x DNC captures changes in load factor in Delta/Northwest/Continental products (relative to other products) due to the code-sharing agreement. The coefficient estimate on [T.sup.dnc] x DNC is positive and statistically significant, suggesting that the codeshare agreement caused the mean load factor of Delta/Northwest/ Continental products to rise compared to the mean load factor of other carriers over the pre- and post-codeshare periods. In other words, the codeshare agreement is associated with higher load factor. This result is supported by the preliminary results reported in table 3 and figure At in the appendix, and strengthens the case for economies of passenger traffic density spurred by code-sharing.
Codeshare Effects Accounting for Pre-Codeshare Competition
To measure the load-factor effects of the codeshare agreement in specific markets (markets where the partners competed versus markets where they did not compete), we re-estimate our model with the inclusion of a market dummy variable [MKT.sup.dnc] and present the results in table 6. [MKT.sup.dnc] is a zero-one market-specific dummy variable that takes a value of 1 only for origin-destination markets in which any two of the three codeshare partners competed prior to the alliance.
The effects of the codeshare agreement on load factor in markets where the alliance firms competed before code-sharing is determined by the sum of the coefficients on [T.sup.dnc] and interaction variables [T.sup.dnc] x DNC and [T.sup.dnc] x DNC x [MKT.sup.dnc]. Given that the coefficient estimate on [T.sup.dnc] is not statistically different from zero, the codeshare effects in markets where the partners competed is -0.0106 [o + 0.0185 + (-0.0291)] in column 1 of table 6. Similar negative load factor effects are obtained in columns 2 and 3 of table 6. These negative effects suggest that the Delta/Northwest/Continental codeshare agreement is associated with decreases in load factor in overlapping markets.
The coefficient estimates on [T.sup.dnc] and the interaction variable [T.sup.dnc] x DNC in table 6 have a different interpretation than in table 5. In fact, the sum of these coefficient estimates measures changes in load factor in markets where the codeshare partners did not compete prior to their agreement. The sum of these coefficient estimates suggests the load factor increased in these nonoverlapping markets. Overall, what this suggests is that we find negative (positive) codeshare effects on load factor in markets where the codeshare partners competed (did not compete) prior to code-sharing.
To uncover what is driving this decrease in load factor in competing markets for codeshare partners, we examine and present the percentage changes in load factor, passengers transported and seat supply in table 7. Table 7 shows that load factor increased for both groups of carriers but for different reasons. The codeshare partners witnessed a 4.5 percent increase in load factor in overlapping markets due to relatively larger change in passengers transported versus the change in seats. For other carriers, load factor increased by 5.1 percent because the rate of decrease in available seats (7.7%) outweighed the drop in passengers transported (1.6%). Therefore, the negative load factor effects of code-sharing we observe in these overlapping markets come from the difference in load factor percentage changes between the two groups in column 3 of table 7, despite the fact that passenger volume and seat supply increase (decrease) for codeshare partners (other carriers).
Correcting for Endogeneity
Every empirical study has some endogeneity issues that should at least be discussed and possibly addressed. The fact that carriers decide on the number of departures to operate in a specific market raises selectivity issues. Passengers tend to prefer to fly with a carrier that offers more frequent service and at the same time it is more likely that carriers offer more frequent service on routes with higher passenger volumes. This feedback loop poses a potential endogeneity problem.
To address this concern, we construct instruments for Departures Performed. The instrument variables we use for the two-stage least squares (2SLS) estimation are: (1) available passenger miles and (2) the flight's itinerary distance flown. We estimate the load factor equation using 2SLS and present the results in tables 8 and 9. A Hausman test and Durbin chi-square test both confirm endogeneity by rejecting the exogeneity of Departures Performed at conventional levels of statistical significance. Due to the economics of space, we only report the Hausman test results in tables 8 and 9.
To confirm the validity of instruments used in the 2SLS regression, we estimate first-stage reduced-form regressions of the endogenous variable. First-stage reduced-form regression where we regress Departures Performed against the instruments yields statistically significant coefficients, suggesting that the instruments explain variations in the endogenous variable. [R.sup.2] measure for the regression of Departures Performed against the instruments is 0.6356. We also performed Stock and Yogo's (2005) test for weak instrument. The test rejects the null hypothesis that the instruments are weak, because the test statistic exceeds the critical values of conventional rejection rates of a nominal 5% Wald test. This test indicates that we do not have a weak-instrument problem. The 2SLS estimation results are qualitatively similar to the OLS results.
Discussion on Managerial and Policy Implications
From a managerial perspective, entering into a codeshare agreement is also a risk-pooling endeavor as the partner airlines use their collective capabilities to deal with uncertainty in demand. Demand uncertainty may increase (or decrease) load factor depending on the likelihood that the potential seat shortage of one codeshare partner can be offset by another partner's excess capacity. Therefore, accurately forecasting demand patterns and flight load factor is crucial for managers because doing so has serious implications on codeshare partner selection, network integration, and how partner airlines coordinate their flight schedules. These implications also call for determining the optimal seat supply on a particular route, as code-share partners deal with the seemingly delicate tradeoff between the advantages of efficient operations and the shortcomings of lower service convenience (Brenner 1982).
From a policy perspective, the negative load factor effects we found in overlapping markets (markets in which the potential partners' own flights were already competing between the origin and destination cities before the codeshare agreement) may be suggestive of potential collusive behavior on service levels, as the DOT feared, among the codeshare partners. However, upon further investigation, we found that passenger traffic for the codeshare partners actually increased in these overlapping city pairs, indicating no evidence that the codeshare partners engaged in collusive behavior.
The main objective of this article is to empirically investigate the load factor effects of the Delta/Northwest/Continental codeshare agreement, with a particular focus on differences in load-factor changes in the partners' overlapping and nonoverlapping markets.
We find that the codeshare agreement between Delta, Northwest, and Continental is associated with improved load factor for the alliance firms relative to other carriers. We also find asymmetric load-factor responses to code-sharing in markets where the partners competed versus markets where they did not compete. We find statistically significant negative (positive) codeshare effects on load factor in markets where the codeshare partners competed (did not compete). Furthermore, we find no evidence that these codeshare partners facilitated collusion on service. The findings of this study provide new insights into managerial and policy issues surrounding airline alliances.
Though our results are compelling, there are other important market effects that we did not explore that may be fertile topics for future research. Another interesting extension would be to analyze possible entry deterrence effects of this codeshare agreement. Also, since we found negative load factor effects in overlapping markets, can we quantify the consumer welfare implications? Exploring these questions may provide a better understanding of the distributional effects of code-sharing in particular and alliances in general.
Embry-Riddle Aeronautical University
Embry-Riddle Aeronautical University
This research is funded in part by a grant from the Embry-Riddle URI (Undergraduate Research Institute) Ignite and Faculty Research Development Program. For very helpful comments and suggestions, we thank co-editor Yoshinori Suzuki, two anonymous referees, Adin Herzog, Anne Boettcher, and Robin Sobotta. Any remaining errors are our own.
(1.) Interline indicates that at some point in the trip, when passengers change planes they also change carriers.
(2.) An airline's capacity utilization measures how efficiently an airline fills seats on its planes. We use airline's load factor (passenger-miles as a proportion of available seat-miles) to proxy for capacity utilization.
(3.) Gayle (2008) and Ito and Lee (2007) provide a more comprehensive analysis of the code-sharing practice.
Aguirregabiria, V., and C.-Y. Ho. 2012. "A Dynamic Oligopoly Game of the US Airline Industry: Estimation and Policy Experiments.'7ournal ofEconometrics 168:156-73.
Bailey, E. E., D. R. Graham, D. P. Kaplan, and D. P. Graham. 1985. Deregulating the Airlines. Vol. 10. Cambridge, MA: MIT Press.
Bamberger, G. E., D. W. Carlton, and L. R. Neumann. 2004. "An Empirical Investigation of the Competitive Effects of Domestic Airline Alliances." Journal of Law and Economics 47 (1): 195-222.
Berry, S. 1992. "Estimation of a Model of Entry in the Airline Industry." Econometrica 60:889-918.
Berry, S., M. Carnall, and P. T. Spiller. 2006. "Airline Hubs: Costs, Markups and the Implications of Customer Heterogeneity." In Advances in Airline Economics, vol. 1, Competition Policy and Antitrust, ed. D. Lee, 183-213. North Holland: Elsevier.
Berry, S., and P. Jia. 2010. "Tracing the Woes: An Empirical Analysis of the Airline Industry." American Economic Journal: Microeconomics 2 (3): 1-43.
Bilotkach, V. 2005. "Price Competition between International Airline Alliances." Journal of Transport Economics and Policy 39 (2): 167-90.
Brenner, M. A. 1982. "The Significance of Airline Passenger Load Factors." In Airline Economics, edited by G. W. James. Lexington, MA: Lexington Books.
Brewer, M. A., and P. Hooper. 1998. "Strategic Alliances Among International Airlines and Their Implications for Organizational Change." Working Paper, Institute of Transport Studies, University of Sydney.
Brueckner, J. K. 2001. "The Economics of International Codesharing: An Analysis of Airline Alliances." International Journal of Industrial Organization 19 (10): 1475-98.
--. 2003. "International Airfares in the Age of Alliances: The Effects of Code-sharing and Antitrust Immunity." Review of Economics and Statistics 85 (1): 105-18.
Brueckner, J. K., and W. T. Whalen. 2000. "The Price Effects of International Airline Alliances."Journal of Law and Economics 43 (2): 503-46.
Chen, F. C. Y., and C. Chen. 2003. "The Effects of Strategic Alliances and Risk Pooling on the Load Factors of International Airline Operations." Transportation Research PartE: Logistics and Transportation Review 39 (1): 19-34.
Chen, Y., and P. G. Gayle. 2019. "Mergers and Product Quality: Evidence from the Airline Industry." International Journal of Industrial Organization 62:96-135.
Chin, A. T. 2002. "Impact of Frequent Flyer Programs on the Demand for Air Travel." Journal of Air Transportation 7 (2): 53-86.
Deprosse, H., and M. Franke. 1998. "One for All and All for One: Threats and Opportunities Posed by the Partnering Principle." In Handbook of Airline Marketing, edited by G. F. Butler and M. R. Keller. New York: McGraw-Hill.
Dresner, M., S. Flicop, and R. Windle. 1995. "Trans-Atlantic Airline Alliances: A Preliminary Evaluation."Journal of the Transportation Research Forum 35 (1): 13-25.
Evans, W. N., and I. N. Kessides. 1993. "Localized Market Power in the US Airline Industry." Review of Economics and Statistics 75 (1): 66-75.
Forbes, S. J. 2008. "The Effect of Air Traffic Delays on Airline Prices." International Journal of Industrial Organization 26 (5): 1218-32.
Forbes, S. J., M. Lederman, and T. Tombe. 2015. "Quality Disclosure Programs and Internal Organizational Practices: Evidence from Airline Flight Delays." American Economic Journal: Microeconomics 7 (2): 1-26.
Gellman Research Association. 1994. "A Study of International Airline Code-Sharing."
Report submitted to Office of Aviation and International Economics, US Department of Transportation, Washington, DC.
Gayle, P. G. 2007. "Airline Code-Share Alliances and Their Competitive Effects ."Journal of Law and Economics 50 (4): 781-819.
--. 2008. "An Empirical Analysis of the Competitive Effects of the Delta/Continental/ Northwest Code-Share Alliance." Journal of Law and Economics 51 (4): 743-66.
--. 2013. "On the Efficiency of Codeshare Contracts Between Airlines: Is Double Marginalization Eliminated?" American Economic Journal: Microeconomics 5 (4): 244-73
Gayle, P. G., and D. Brown. 2014. "Airline Strategic Alliances in Overlapping Markets: Should Policymakers Be Concerned?" Economics of Transportation 3 (4): 243-56.
Gayle, P. G., and H. B. Le. 2013. "Airline Alliances and Their Effects on Costs." Manuscript, Kansas State University.
Gayle, P. G., and T. Thomas. 2015. "Product Quality Effects of International Airline Alliances, Antitrust Immunity, and Domestic Mergers." Review of Network Economics 14 (1): 45-74.
Iatrou, K, and F. Alamdari. 2005. "The Empirical Analysis of the Impact of Alliances on Airline Operations." Journal of Air Transport Management 11 (3): 127-34.
Ito, H., and D. Lee. 2005. "Domestic Code Sharing Practices in the US Airline Industry." Journal of Air Transport Management 11 (2): 89-97.
--. 2007. "Domestic Code Sharing, Alliances, and Airfares in the US Airline Industry." Journal of Law and Economics 50 (2): 355-80.
Lederman, M. 2007. "Do Enhancements to Loyalty Programs Affect Demand? The Impact of International Frequent Flyer Partnerships on Domestic Airline Demand." RAND Journal of Economics 38 (4): 1134-58.
--. 2008. "Are Frequent-Flyer Programs a Cause of the 'Hub Premium'?" Journal of Economics and Management Strategy 17 (1): 35-66.
Mantovani, A., and O. Tarola. 2007. "Did the Entry of Low Cost Companies Foster the Growth of Strategic Alliances in the Airline Industry?" Rivista di Politico Economica 97 (1): 189-220.
Nyathi, Z. M. 1995. "Modeling Strategic Alliance Partner Choice in International Airline Network." Working Paper, Institute of Transport Studies, University of Sydney.
Oum, T. H., and J. Park. 1997. "Airline Alliances Current Status Policy Issues and Future Directions." Journal of Air Transport Management 3 (3): 181-95.
Oum, T. H., J. Park, and A. Zhang. 1996. "The Effects of Airline Codesharing Agreements on Firm Conduct and International Air Fares." Journal of Transport Economics and Policy 30 (2): 187-203.
Park, J. 1997. "The Effects of Alliances on Market and Economic Welfare." Transportation Research Part E: Logistics and Transportation 33 (3): 181-95.
Park, J., and A. Zhang. 1998. "Airline Alliances and Partner Firm's Outputs." Transportation Research Part E: Logistics and Transportation 34 (4): 245-55.
Shen, C. 2015. "Code Sharing and Merger: Continental, Delta and Northwest." Frontiers of Economics in China 10 (4): 643-64.
Stock, J. H., and M. Yogo. 2005. "Asymptotic Distributions of Instrumental Variables Statistics with Many Instruments." In Identification and Inference for Econometric Models, vol. 6, edited by D. W. K. Andrews and J. H. Stock, 109-20. Cambridge: Cambridge University Press.
Tiernan, S., D. Rhoades, and B. Waguespack. 2008. "Airline Alliance Service Quality Performance--An Analysis of US and EU Member Airlines." Journal of Air Transport Management 14 (2): 99-102, doi:10.1016/j.jairtraman.2008.02.003.
US Department of Transportation. 2003. "Termination of Review Under 49 U.S.C. 41720 of Delta/Northwest/Continental Agreements." Federal Register 68 (15): 3293-99, https://www.gp0.g0v/fdsys/pkg/FR-2003-01-23/pdf/03-1528.pdf.
US General Accounting Office. 1995. International Aviation: Airline Alliances Produce Benefits, but Effect on Competition Is Uncertain. Washington, DC: US Government Printing Office.
--. 1999. "Aviation Competition: Effects on Consumers from Domestic Airline Alliances Vary," Report #GAO/RCED-99-37. Washington, DC: US Government Printing Office.
Yimga, J. 2017. "Airline Code-Sharing and Its Effects on On-Time Performance"Journal of Air Transport Management 58:76-90.
Youssef, W., and M. Hansen. 1994. "Consequences of Strategic Alliances between International Airlines: The Case of Swiss Air and SAS." Transportation Research Part A: Policy and Practice 28 (5): 415-31.
Zea, M., and D. Feldman. 1998. "Going Global: The Risk and Rewards of Airline Alliance-Based Network Strategies." In Handbook of Airline Marketing, edited by G. F. Butler and M. R. Keller. New York: McGraw-Hill.
Zou, L., T. H. Oum, and C. Yu. 2011. "Assessing the Price Effects of Airline Alliances on Complementary Routes." Transportation Research Part E: Logistics and Transportation Review 47 (3): 315-32.
Caption: Figure 1 Load factor distribution.
Table 1/Number of Code-Sharing Routes Ticketing Carrier Operating Carrier CO DL NW CO 245 269 DL 238 283 NW 439 472 Source: Adapted from Shen 2015. Table 2/Number of Overlapping Code-Sharing Routes Ticketing/Operating CO/DL CO/NW DL/CO DL/NW NW/CO NW/DL CO/DL 238 CO/NW 439 DL/CO 39 245 DL/NW 290 472 NW/CO 11 90 269 NW/DL 57 72 283 Source: Adapted from Shen 2015. Table 3/Difference-In-Differences Estimation Results, Pre- and Post-Codeshare Outcome Mean Load S. Err. [absolute P > [absolute Variable Factor (1) (2) value of] (3) value of] (4) Before Control 0.504 Treated 0.608 Diff(T-C) 0.104 0.003 38.36 0.000 *** After Control 0.539 Treated 0.662 Diff(T-C) 0.124 0.003 44.01 0.000 *** Diff-in-Diff 0.019 0.004 4.97 0.000 *** Notes: Treated (T) denotes the codeshare partners DL, NW and CO. Control (C) denotes other carriers. *** p < 0.01; ** p < 0.05; * p < 0.10 Table 4/Summary Statistics Continuous Variables Mean Std. Dev. Min Max Load factor (a) .613 .186 .002 1 No. of departures 37 51.572 0 1424 performed (per product per month) Nonstop flight distance 587308 580.515 0 4962 (miles) HHI (Herfindahl- 2.351 1.915 .171 19.718 Hirschman Index) Categorical Variables Frequency Slot-controlled 483 markets (b) Non-slot-controlled 14,831 markets Competitive markets 5,048 Noncompetitive markets 10,266 Codeshare products 4,084 by DL, NW and CO (DNC) Overlapping markets 1,946 between DL/NW/CO (MKT (DNC)) Number of Products 35,225 Number of Markets (c) 15,314 (a) Ratio of passenger miles to available seat miles. (b) Dummy variable coded 1 if origin and destination airports are slot-controlled. (c) A market is a directional origin-destination combination. Table 5/OLS Regression Results for Effect of Code-Sharing on Load Factor (1) (2) (3) Departures Performed (log) -0.0058 *** -0.0057 *** -0.0172 *** (0.0003) (0.0003) (0.0003) Slot Control -0.0343 *** -0.0343 *** -0.0078 (0.0067) (0.0067) (0.0061) Competitive Market -0.0182 *** -0.0182 *** -0.0450 *** (0.0008) (0.0008) (0.0008) HHI 0.0027 *** 0.0026 *** -0.0003 (0.0002) (0.0002) (0.0003) Distance (log) 0.0471 *** 0.0470 *** 0.0481 *** (0.0006) (0.0006) (0.0007) DNC 0.0206 *** 0.0202 *** 0.0183 *** (0.0021) (0.0021) (0.0019) [T.sup.dnc] 0.0008 0.0003 0.0007 (0.0008) (0.0008) (0.0007) [T.sup.dnc] X DNC 0.0095 *** 0.0101 *** 0.0103 *** (0.0024) (0.0024) (0.0020) August -0.0056 *** -0.0051 *** (0.0013) (0.0011) September -0.0099 *** -0.0088 *** (0.0013) (0.0011) October -0.0097 *** -0.0093 *** (0.0013) (0.0011) November -0.0083 *** -0.0089 *** (0.0013) (0.0011) December -0.0114 *** -0.0116 *** (0.0013) (0.0011) Constant 0.1736 *** 0.1801 *** 0.1950 *** (0.0178) (0.0178) (0.0312) Aircraft Type Fixed Effects x x x Market Orig. Fixed Effects x Market Dest. Fixed Effects x Carrier Fixed Effects x Number of Observations 147120 147120 147120 [R.sup.2] 0.46 0.46 0.62 The equations are estimated using ordinary least squares (OLS). Fixed effects are incrementally included in each specification but were not reported for brevity. Note: Standard errors are in parentheses. *** p < 0.01; ** p < 0.05; * p < 0.10 Table 6/OLS Regression Results for Effect of Code-Sharing on Load Factor in Overlapping Markets (1) (2) (3) Departures Performed -0.0059 *** -0.0058 *** -0.0173 *** (log) (0.0003) (0.0003) (0.0003) Slot Control -0.0413 *** -0.0414 *** -0.0066 (0.0068) (0.0068) (0.0061) Competitive Market -0.0199 *** -0.0199 *** -0.0450 *** (0.0008) (0.0008) (0.0008) HHI 0.0024 *** 0.0024 *** -0.0002 (0.0002) (0.0003) (0.0003) Distance (log) 0.0474 *** 0.0473 *** 0.0479 *** (0.0006) (0.0006) (0.0007) DNC 0.0186 *** 0.0182 *** 0.0191 *** (0.0021) (0.0021) (0.0020) [T.sup.dnc] 0.0008 0.0003 0.0007 (0.0008) (0.0008) (0.0007) [T.sup.dnc] x DNC 0.0185 *** 0.0191 *** 0.0162 *** (0.0026) (0.0026) (0.0023) [MKT.sup.dnc] 0.0155 *** 0.0155 *** -0.0022 (0.0014) (0.0014) (0.0015) [T.sup.dnc] x DNC x -0.0291 *** -0.0290 *** -0.0189 *** [MKT.sup.dnc] (0.0038) (0.0038) (0.0032) August -0.0056 *** -0.0051 *** (0.0013) (0.0011) September -0.0099 *** -0.0088 *** (0.0013) (0.0011) October -0.0096 *** -0.0093 *** (0.0013) (0.0011) November -0.0083 *** -0.0088 *** (0.0013) (0.0011) December -0.0115 *** -0.0116 *** (0.0013) (0.0011) Constant 0.1734 *** 0.1798 *** 0.1952 *** (0.0178) (0.0178) (0.0312) Aircraft Type Fixed Effects x x x Market Orig. Fixed Effects x Market Dest. Fixed Effects x Carrier Fixed Effects x Number of Observations 147120 147120 147120 [R.sup.2] 0.46 0.46 0.62 Notes: The equations are estimated using Ordinary Least Squares (OLS). Fixed effects are incrementally included in each specification but were not reported for brevity. Standard errors are in parentheses. *** p < 0.01; ** p < 0.05; * p < 0.10 Table 7/Percentage Changes in Overlapping Markets across Carrier Groups Average Percentage Change Number of Number of Load passengers (1) Seats W Factor (3) Codeshare Partners 8.3 2 4.5 Other Carriers -1.6 -7.7 5.1 Table 8/2SLS Regression Results for Effect of Code-Sharing on Load Factor (1) (2) Departures Performed (log) 0.0319 *** 0.0318 *** (0.0005) (0.0005) Slot Control -0.0381 *** -0.0379 *** (0.0023) (0.0023) Competitive Market -0.0048 *** -0.0049 *** (0.0008) (0.0008) HHI -0.0017 *** -0.0018 *** (0.0003) (0.0003) Distance (log) 0.0465 *** 0.0464 *** (0.0006) (0.0006) DNC 0.0133 *** 0.0129 *** (0.0022) (0.0022) [T.sup.dnc] 0.0023 *** 0.0016 * (0.0008) (0.0008) [T.sup.dnc] x DNC 0.0120 *** 0.0127 *** (0.0025) (0.0025) August -0.0066 *** (0.0013) September -0.0121 *** (0.0013) October -0.0120 *** (0.0013) November -0.0113 *** (0.0013) December -0.0142 *** (0.0013) Constant 0.1430 *** 0.1510 *** (0.0186) (0.0186) Aircraft Type Fixed Effects X X Market Orig. Fixed Effects Market Dest. Fixed Effects Carrier Fixed Effects Endogeneity Test: F(1,147000) = F(1,146995) = [H.sub.0]: Departures Perf. 10194.5 *** 10148.2 *** is exogenous (p = 0.0000) (p = 0.0000) Number of Observations 147120 147120 [R.sup.2] 0.41 0.41 (3) Departures Performed (log) 0.0216 *** (0.0006) Slot Control -0.0557 *** (0.0046) Competitive Market -0.0162 *** (0.0009) HHI -0.0017 *** (0.0003) Distance (log) 0.0559 *** (0.0007) DNC 0.0216 *** (0.0020) [T.sup.dnc] 0.0011 (0.0007) [T.sup.dnc] x DNC 0.0134 *** (0.0021) August -0.0061 *** (0.0011) September -0.0111 *** (0.0011) October -0.0120 *** (0.0011) November -0.0115 *** (0.0011) December -0.0135 *** (0.0011) Constant 0.2046 *** (0.0325) Aircraft Type Fixed Effects X Market Orig. Fixed Effects X Market Dest. Fixed Effects X Carrier Fixed Effects X Endogeneity Test: F(1,145194) = [H.sub.0]: Departures Perf. 6671.01 *** is exogenous (p = 0.0000) Number of Observations 147120 [R.sup.2] 0.58 Notes: The equations are estimated using two stage least squares (2SLS). Fixed effects are incrementally included in each specification but were not reported for brevity. Standard errors are in parentheses. *** p < 0.01; ** p < 0.05; * p < 0.10 Table 9/2SLS Regression Results for Effect of Code-Sharing on Load Factor in Overlapping Markets (1) (2) Departures Performed (log) 0.0317 *** 0.0316 *** (0.0005) (0.0005) Slot Control -0.0409 *** -0.0407 *** (0.0024) (0.0024) Competitive Market -0.0059 *** -0.0060 *** (0.0009) (0.0009) HHI -0.0018 *** -0.0019 *** (0.0003) (0.0003) Distance (log) 0.0467 *** 0.0466 *** (0.0006) (0.0006) DNC 0.0121 *** 0.0116 *** (0.0022) (0.0022) [T.sup.dnc] 0.0023 *** 0.0017 ** (0.0008) (0.0008) [T.sup.dnc] x DNC 0.0188 *** 0.0195 *** (0.0028) (0.0028) [MKT.sup.dnc] 0.0109 *** 0.0108 *** (0.0015) (0.0015) [T.sup.dnc] x DNC x -0.0220 *** -0.0220 *** [MKT.sup.dnc] (0.0039) (0.0039) August -0.0066 *** (0.0013) September -0.0120 *** (0.0013) October -0.0120 *** (0.0013) November -0.0112 *** (0.0013) December -0.0142 *** (0.0013) Constant 0.1427 *** 0.1506 *** (0.0186) (0.0186) Carrier Fixed Effects x x Aircraft Type Fixed Effects x x Market Orig. Fixed Effects Market Dest. Fixed Effects EndogeneityTest: F(l, 146998) = F(l,146993) = [H.sub.0]: departures perf. 10165.3 *** 10119.6 *** is exogenous (p = 0.0000) (p = 0.0000) Number of Observations 147120 147120 [R.sup.2] 0.41 0.41 (3) Departures Performed (log) 0.0217 *** (0.0006) Slot Control -0.0558 *** (0.0046) Competitive Market -0.0164 *** (0.0009) HHI -0.0016 *** (0.0003) Distance (log) 0.0564 *** (0.0007) DNC 0.0217 *** (0.0020) [T.sup.dnc] 0.0011 (0.0007) [T.sup.dnc] x DNC 0.0197 *** (0.0024) [MKT.sup.dnc] 0.0113 *** (0.0016) [T.sup.dnc] x DNC x -0.0202 *** [MKT.sup.dnc] (0.0034) August -0.0061 *** (0.0011) September -0.0111 *** (0.0011) October -0.0120 *** (0.0011) November -0.0115 *** (0.0011) December -0.0135 *** (0.0011) Constant 0.2030 *** (0.0325) Carrier Fixed Effects x Aircraft Type Fixed Effects x Market Orig. Fixed Effects x Market Dest. Fixed Effects x EndogeneityTest: F(l,145192) = [H.sub.0]: departures perf. 6661.55 *** is exogenous (p = 0.0000) Number of Observations 147120 [R.sup.2] 0.58 Notes: The equations are estimated using two stage least squares (2SLS). Fixed effects are incrementally included in each specification but were not reported for brevity. Standard errors are in parentheses. *** p < 0.01; ** p < 0.05; * p < 0.10 Table A1/Airlines in Sample Code Airlines AA American Airlines (a) AS Alaska Airlines (a) B6 JetBlue Airways CO Continental Air Lines (a) DL Delta Air Lines (a) EV Atlantic Southeast HP America West Airlines MQ American Eagle F9 Frontier Airlines (a) FL AirTran Airways (a) G4 Allegiant Air HA Hawaiian Airlines HP America West Airlines N7 National Airlines NJ Vanguard Airlines NK Spirit Air Lines NW Northwest Airlines (a) OO SkyWest (a) QX Horizon Air RP Chautauqua Airlines SM Sunworld International Airlines SY Sun Country Airlines TZ ATA Airlines (a) UA United Air Lines (a) US US Airways (a) WN Southwest Airlines YX Midwest Airlines * Carrier is involved in virtual codeshare product. Table A2/Aircraft Types in Sample Code Aircraft 030 Cessna 180 033 Cessna 185A/B/C Skywagon 034 Helio H-250/295/395 035 Cessna C206/207/209/210 Stationair 036 Cessna 172 Skyhawk 040 De Havilland DHC2 Beaver 042 De Havilland DHC3 Otter 079 Piper PA-32 (Cherokee 6) 080 Piper PA-18 (Super-Cub) 084 Piper PA-28 (Cherokee) 091 Float/Amphib Turbine 094 Land-Turbine 111 Beech King Air 90 117 Beech Baron (55 Series) 122 Cessna C-310 Series 125 Cessna C-402/402A/402B 131 Pilatus Britten-Norman BN2/A Islander 133 Beech 65/65A-8O/65B-8O (Oueen Air) 160 McDonnell Douglas DC-3/A/C.C-47/B 170 Grumman C-21A (Goose) 190 Piper PA-23-250 (Aztec/ Apache) 194 Piper PA-31 (Navajo)/T-1020 195 Piper PA-34/39 (Twin Commanche) 201 Pilatus Britten-Norman BN2ATrislander 315 Bell B-206A 360 Robinson R44 404 Beech C99 405 Beech 1900 A/B/C/D 406 Beech 200 Super Kingair 412 Casa/Nurtanio C212 Aviocar 416 Cessna 208 Caravan 417 Cessna 406 Caravan II 430 Convair CV-580 441 Aerospatiale/Aeritalia ATR-42 442 Aerospatiale/Aeritalia ATR-72 449 Dornier 328 450 Fokker Friendship F-27/ Fairchild F-27/A/B/F/J 456 Saab-Fairchild 340/B 461 Embraer EMB-120 Brasilia 467 Swearingen Metro III 469 British Aerospace Jetstream 31 471 British Aerospace Jetstream 41 478 Piper T-1040 479 Pilatus PC-12 482 De Havilland DHC8-400 Dash-8 483 De Havilland DHC8-100 Dash-8 484 De Havilland DHC8-300 Dash 8 485 De Havilland Twin Otter DHC-6 486 Shorts Harland SC-7 Skyvan 489 Shorts 360 491 De Havilland DHC8-2000 Dash-8 602 FokkerF28-4000/6000 Fellowship 612 Boeing 737-700/700LR/ Max 7 614 Boeing 737-800 616 Boeing 737-500 617 Boeing 737-400 619 Boeing 737-300 620 Boeing 737-100/200 621 Boeing 737-200C 622 Boeing 757-200 623 Boeing 757-300 624 Boeing 767-400/ER 625 Boeing 767-200/ER/EM 626 Boeing 767-300/300ER 627 Boeing 777-200ER/200LR/233LR 628 Canadair RJ-100/RJ-100ER 629 Canadair RJ-200ER /RJ-440 630 McDonnell Douglas DC-9-10 631 Canadair RJ-700 632 Domier328Jet 634 Boeing 737-900 635 McDonnell Douglas DC-9-15F 640 McDonnell Douglas DC-9-30 645 McDonnell Douglas DC-9-40 650 McDonnell Douglas DC-9-50 654 McDonnell Douglas DC9 Super 87 655 McDonnell Douglas DC9 Super 80/MD81/82/83/88 656 McDonnell Douglas MD-90 676 Embraer-140 690 Airbus Industrie A300B/ C/F-100/200 691 Airbus Industrie A300-600/R/ CF/RCF 692 Airbus Industrie A310-200C/F 694 Airbus Industrie A320-100/200 696 Airbus Industrie A330-200 698 Airbus Industrie A319 699 Airbus Industrie A321 710 Boeing 727-100 715 Boeing 727-200/231A 730 McDonnell Douglas DC-10-10 732 McDonnell Douglas DC-10-30 733 McDonnell Douglas DC-10-40 735 McDonnell Douglas DC-10-30CF 740 McDonnell Douglas MD-11 760 Lockheed L-1011-1/100/200 765 Lockheed L-1011-500 Tristar 816 Boeing 747-100 817 Boeing 747-200/300 819 Boeing 747-400 835 Avroliner RJ85 868 British Aerospace BAe-146-300 676 Embraer-140 690 Airbus Industrie A300B/ C/F-100/200 691 Airbus Industrie A300-600/R/ CF/RCF 692 Airbus Industrie A310-200C/F Figure A1 Mean group load factor. Pre vs. Post Code-sharing Other Carriers Load Factor Before 0.50 After 0.54 Codeshare Partners Load Factor Before 0.61 After 0.66 Note: Table made from bar graph.
|Printer friendly Cite/link Email Feedback|
|Author:||Yimga, Jules; Gorjidooz, Javad|
|Date:||Sep 22, 2019|
|Previous Article:||Achieving Financial Performance in Uncertain Times: Leveraging Supply Chain Agility.|
|Next Article:||A Pioneering Approach to Reducing Fuel Cost and Carbon Emissions from Transportation.|