A dissertation
Download 1.47 Mb.
|
- Navigate this page:
- 2.4.1. Data
- 2.4.2. Analysis of Data
2.4. MethodologyGiven the data limitations of past studies that used disaggregate data to study airfares, the goal of this paper is to study how certain characteristics of a market affect airfares by using a larger sample of U.S. domestic markets that cover a broad range of market structures. To the best of the authors’ knowledge, the dataset used in this study represents the largest and most comprehensive disaggregate airline pricing database used to research airfares thus far. 2.4.1. DataThe data was collected in the fall of 2007 in collaboration with QL2® Software, one of the major U.S. companies that collects competitive pricing and product information from websites. In order to obtain data for Southwest Airlines, additional webbots were written by an academic team at Georgia Tech in order to supplement the data provided by QL2®. The data collected consists of prices for more than 100 U.S. markets for one month of departure dates, which were selected to represent periods of peak and off-peak demands (i.e., Thanksgiving and early December 2007, respectively). Round-trip and one-way fares were recorded daily for at least 30 days prior to flight departure. Nonstop fares were obtained from each airline’s website, while nonstop and connecting fares were obtained from at least one major online travel agency (Orbitz®, Travelocity® or Expedia®). For an especially detailed explanation of the data collection methodology and compilation, as well as a more specific account of the dataset, the reader is referred to (Pope, Garrow, et al., 2009). A subset of the aforementioned dataset was chosen for data analysis in order to represent a wide variety of interesting market competition structures and airline competition effects, such as monopolies, duopolies, and competitive markets broken down into subcategories representing whether the markets have multi-airport effects and/or LCC presence. In defining these categories of market structures, only airlines that fly nonstop in a market were considered, thus all observations for connecting flights were eliminated. This is in following with the methodology of a number of other researchers (for example, Borenstein and Rose, 1994; Bilotkach, 2005; Bilotkach et al., 2006; Gerardi and Shapiro, 2007; Giaume and Guillou, 2004; Liu and Serfes, 2006; Verlinda, 2005; Verlinda and Lane, 2004) and is done for two reasons. Firstly, this ensures that the analysis is somewhat comparable to those of past studies. Secondly, eliminating connecting tickets makes the analysis far less complicated. This is due to the fact that connecting tickets represent significantly different qualities of service than direct tickets and controlling for the cost differences would be more complex. Additionally, only the lowest fares are included in the observations, which is also comparable to the methodology of other researchers (Bilotkach and Pejcinovska, 2007; Mentzer, 2000; Pels and Rietveld, 2004). Using the lowest observed fare controls for vertical price differentiation. Vertical price differentiation is the difference in prices due to the differing qualities of tickets (such as restricted vs. non-restricted tickets). By controlling for vertical differentiation, the analysis focuses on horizontal price differentiation, which is defined as the difference in prices due to the varying tastes of customers (such as brand preference, aircraft preference, etc). By focusing on the horizontal price differentiation, the analysis can capture the competitive impacts associated with price dispersion. The final dataset that was used in this study consists of 108,632 observations for 62 airport-to-airport markets and 12 airlines. Each observation represents the lowest nonstop, round-trip fare that was offered by each airline flying nonstop in the market on the date that the website was queried and for each specific day of flight departure, assuming a one night stay. The major7 airlines (with airline code) included are: American Airlines (AA), Alaska Airlines (AS), Continental Airlines (CO), Delta Air Lines (DL), Northwest Airlines (NW), United Airlines (UA), and US Airways (US), and the LCCs included are: Air Tran Airways (FL), Frontier Airlines (F9), JetBlue Airways (B6), Spirit Airlines (NK) and Southwest Airlines (WN). In addition, several codeshares are also represented and are denoted by the code of the marketing carrier followed by the code of the operating carrier in parenthesis. The codeshares represented include: AA (AS), AS (AA), NW (AS), UA (US), and US (UA). Table 2.2 lists the airports included in the dataset, along with the airport codes used throughout this paper. Table 2.2: Airport Codes and Names
2.4.2. Analysis of DataWhen analyzing the data, it was apparent that any level of aggregation only served to hide some of the most interesting observations in the dataset. Airline pricing policies in the dataset vary greatly by airline, market, peak/off peak time periods, and the number of days from departure. Thus, it was inappropriate to analyze the data in a way that would aggregate some of these important variables. Because of this, a case study approach was taken instead of a regression type approach. An additional challenge that was encountered was determining which measure to use for price dispersion. Four measures of price dispersion were investigated. These included the standard deviation of fares, the coefficient of variation (standard deviation normalized by the mean), the range of fares (the difference between the highest and lowest fares), and the interquartile range (the difference between the 75th and 25th percentile fares). Each price dispersion measure gave different results about the magnitude of price dispersion in the market. Table 2.3 lists the 62 markets (denoted by the three letters of the origin airport followed by the three letters of the destination airport) used in this analysis, along with the airlines that fly nonstop in each market. Table 2.3 also includes the mean, standard deviation (SD), interquartile range (IQR), coefficient of variation (CV), and range for each market as reference. Table 2.3: Markets, Airlines, and Summary Statistics
Table 2.3: Markets, Airlines, and Summary Statistics (Continued)
Directory: bitstream -> handle -> 1853 handle -> The Economics of Technology Sharing: Open Source and Beyond Josh Lerner and Jean Tirole handle -> The Rise of Digital Curation and Cyberinfrastructure: From Experimentation to Implementation, and Maybe Integration handle -> Mobile augmented reality applications for library services handle -> Libraries, Archives, and Museums as Epistemic Infrastructure1 Download 1.47 Mb. Share with your friends:
|