The effect of the entry of low-cost airlines on price and ...



The effect of the entry of low-cost airlines on price and passenger trafficMaster ThesisMaster in Economics and BusinessSpecialisation Urban, Port and Transport EconomicsYaxian WuStudent number: 332639Thesis supervisor:Dr. Peran van ReevenDepartment of Applied EconomicsErasmus School of EconomicsErasmus University RotterdamAbstractMany researches about the impact of low-cost airlines are documented since it is one of the most popular topics in airline transport. Generally, the previous findings convince that the entry of low-cost airlines significantly depresses price while increases the passenger traffic. This paper extends to investigate how the low-cost airline impacts the pricing and passenger traffic currently. The sample covers factors in terms of demand, cost and market concentration, using quarterly figures from 1997 to 2010. Moreover, a series of methods including OLS, Fixed effect model and IV estimation, are performed stepwise to explore the most reliable estimation. The low-cost airlines’ entrance does reduce price but at a lower level while the passenger traffic is indirectly affect by the low-cost carriers through price.AcknowledgedFirst of all, I really appreciate my thesis supervisor, Dr. Peran van Reeven for all the help and guidance that he provided. The many thought-provoking discussions and his detailed comments and suggestions were essential in the completion of this work. Without his inspiring and encourage, I would not be able to finish this paper.Furthermore, I would also like to thank some friends who gave me many helps during my writing: Fei Yu, Theun van Vliet, Lu Sun and Wei Li. Thanks for your support both in knowledge and mentality. My deepest thanks are given to my parents for their dedication!Table of Contents TOC \o "1-3" \h \z \u Abstract PAGEREF _Toc307185281 \h iAcknowledged PAGEREF _Toc307185282 \h iiList of Tables and Figures PAGEREF _Toc307185283 \h iii1.Introduction PAGEREF _Toc307185284 \h 12. Literature review PAGEREF _Toc307185285 \h 23.Methodology PAGEREF _Toc307185286 \h 73.1 Sample construction PAGEREF _Toc307185287 \h 73.2 Variables description PAGEREF _Toc307185288 \h 83.3 The estimating equation PAGEREF _Toc307185289 \h 103.3.1 Pooled OLS model PAGEREF _Toc307185290 \h 113.3.2 Fixed effects model PAGEREF _Toc307185291 \h 113.3.3 IV estimation PAGEREF _Toc307185292 \h 123.3.4 Panel IV estimation PAGEREF _Toc307185293 \h 124. Descriptive Analysis PAGEREF _Toc307185294 \h 135. Model results PAGEREF _Toc307185295 \h 195.1 Pooled OLS model PAGEREF _Toc307185296 \h 195.2 Fixed effects model PAGEREF _Toc307185297 \h 215.3 IV estimation PAGEREF _Toc307185298 \h 225.4 Panel IV estimation PAGEREF _Toc307185299 \h 236. Conclusion PAGEREF _Toc307185300 \h 256.1 Comparison with other researches PAGEREF _Toc307185301 \h 266.2 Implications for low-cost airlines PAGEREF _Toc307185302 \h 276.3 Limitation and further research PAGEREF _Toc307185303 \h 28References PAGEREF _Toc307185304 \h 29Appendix 1 Price movement pro and post entry PAGEREF _Toc307185305 \h 31Appendix 2 Passengers movement pro and post entry PAGEREF _Toc307185306 \h 32List of Tables and FiguresTables:Table 2.1 Literature summary 7Table 3.1 Vacation cities in model10Table 3.2 Low-cost carriers in model (with carrier code)10Table 4.1 Descriptive statistics14Table 5.1 OLS regression results20Table 5.2 Fixed effect model results22Table 5.3 Correlation matrix23Table 5.4 2SLS regression results24Table 6.1 Results comparison27Figures:Figure 4.1 The movements of price, passengers and low-cost airlines14Figure 4.2 The distance distribution on top 100 routes / average number of LCC per route15Figure 4.3 One-way prices pro and post entry17Figure 4.4 The passenger traffic pro and post entry18IntroductionMany factors influence the pricing and passenger traffic in air industry. The presence of the low-cost airlines is one of the most popular determinants. Consequently, the impact of low-cost airlines has been much documented. The main conclusions drawn from previous literatures are: the entry of low-cost airlines significantly depresses the price associated with the increase in passenger traffic on the specific routes they joined.Since the first low-cost airline, Southwest Airline, established in 1978, low-cost airlines have captured a huge success in the US as well as European countries. Many famous low-cost airlines like Southwest Airline, AirTran Airways, JetBlue, JetBlue, Ryanair and Virgin, etc. show that the low-cost airlines have grown to be the new strength of development. The Southwest effect, firstly named in 1993 the U.S. Department of Transportation (DOT), indicates the huge impact of the entry of Southwest airline on incumbents in the same serving region. This term is subsequently inferred to describe the general low-cost airlines’ effect. Later on, DOT defined the “low-cost airline service revolution” considering dramatic boost of low-cost airlines (Hüschelrath & Müller, 2011). Most of researches are all theoretically based on the Southwest effects, extending to investigate of the effect of low-cost revolution. Summarised by Wang (2005), the Southwest effects consist of three principals. First of all, the Southwest airline’s entrance brings remarkable passenger enhance. Additionally, the entry of Southwest airline declines passengers travelling on other routes in the same serving region. Furthermore, incumbents attempt to retain the market share on the specific route Southwest airline joined by depressing their price (Ritter, 1993). Taking the Oakland-Ontario airport pair on California corridor as an example, Bennett & Craun (1993) drew several charts illustrating the pricing and traffic movements before and after Southwest’s entry from the second quarter of 1982 to the third quarter of 1992. It was convinced that price declines by 60 percent associated with triple traffic account. Therefore, competitors leave so that Southwest replaces their capacities with the increasing load.This paper extends the research to investigate how the low-cost airline impacts the pricing and passenger traffic under the new circumstance. Considering both time-series and cross-sectional effects, a panel dataset is occupied. The sample covers factors in terms of demand, cost and market concentration, using quarterly figures from 1997 to 2010. Moreover, a series of methods are performed step by step to avoid the systemic drawbacks of previous researches. It starts with the normal OLS regression which gives inadequate explanation about the relations among variables due to the heterogeneity bias. The second step is the fixed effect mode with a better output but still do not sufficiently convince the impact of the low-cost airlines on pricing and passenger traffic. On the top of two basic models, the following step extends to the instrumental variable model. According to the test for endogeneity, the instrumental variable is much more suitable for this dataset. Finally, the panel IV model is installed to deal with the omitted variable bias and causality problem between pricing and demand. The rest of paper is structured as follows: section 2 reviews the literatures relevant to the research question, followed by the sample construction and modelling description in section 3. Then, section 4 illustrates the initial findings descriptively while the results of empirical model are discussed in section 5. Finally, section 6 concludes the paper with comparison as well as suggestions for the low-cost airlines.2. Literature reviewVarious factors may affect pricing and traffic on air transport industry, such as the efficiency of hub and spoken operational system, the entry of low-cost airlines, the extent of market concentration, and competition from other modes of transportation, etc. (Vowles, 2000 and Wang, 2005). Among these determinants, the entry of low-cost carriers captured most interest of researchers. There are many empirical papers focusing on the effect of low-cost airlines’ entrance on airfare and passenger traffic on US domestic air transport and extensive results have been drawn from them. Since the beginning of the deregulation period, many researchers have being studied the impact of the entry of low-cost carriers in airfare pricing. Bailey et al. concluded that there is the negative significant relation between the new entry and US domestic yield in their book published in 1985. Strassmann used a structural model involved variables of prices, entry and concentration. It is convinced that entry and price are mutually affected. This finding is supported by the fact that decreases in concentration, caused by entry, are associated with a substantial decrease in fares (Strassmann, 1990). In 1992, Whinston and Collins investigated the effect of the entry of People Express to the airfare. They found significant evidence for the negative relationship between price and People Express’ presence, in general low-cost carriers, using the stock data from 1984 to 1985. Bennett and Graun (1993) studied the case of Southwest Airline, and then defined “The Southwest Effect”. In sum, the Southwest effect implies that the Southwest airline’s presence increases the passenger count and lowers the airfare in that particular route with the decrease of passenger traffic in competing routes (Wang, 2005). The common result from these reviewed papers is that the presence of low-cost carriers significantly lowers the price while increasing passenger traffic.Following the previous researches, Windle & Dresner kept on exploring the effect of low-cost carriers’ entrance on the specific market they joined. Windle & Dresner analysed the short and long term effect of the entry of low-cost airlines. They performed both descriptive analysis and econometric models, which give the theoretical basis as well as the methodological basis to this paper. Top 200 US domestic routes data from the third quarter of 1991 to the second quarter of 1994 have been filtered from Origin and Destination Survey published by US Department of Transportation. First, changes in the route after the presence of low-cost airlines have been analysed by the time series in terms of the market concentration which indicated by Herfindahl Index (HI), price and passenger traffic. Southwest taken as representation of low-cost carriers captured the biggest difference among three categories including deregulation and other carriers. It lowers 25% of the market concentration of the route joined, whilst the average decline is 15%. As for the airfare, the price on the route Southwest entered depresses by 48% to pre-entry price and keeps still while the average decrease is 19%. The passenger traffic increases by 300% on the routes Southwest airline enters and 74% on average. Furthermore, two empirical models have been performed to interpret the mathematic relation between selected variables and price. When involved the carrier dummy variable, the presence of low-cost airlines significantly decreases the price on routes, however, the market concentration and route density variables do not impact on price as much as authors expected. Vowles (2000) performed a regression model to explain the variance in airfare pricing. The regression involved several variables assumed have effect on the airfare pricing such as distance, resort, southwest factor, hub, low, market share of low-cost airlines and the market share of the largest carrier in the market. According to the coefficients of model results, each additional presence of low-cost airline is predicted to depress the average airfare by 45.47% which implies that low-cost carriers do have a remarkable effect in pricing. However, Vowles considers that this outcome is not sufficient, because this variable does not measure the percentage of schedule flights offered by low-cost airlines. So the Southwest variable and the market share of the low-cost carriers have been added in the model to compensate the weakness. It convinced that the entry of Southwest airline immediately decrease the price on the specific route by 77.61%.To analyse the extent of the “Southwest Effect” on the pricing of fares in the airline industry, Christine Wang (2005) played a regression using the data in the first quarter of 2004. In the conclusion, the Southwest airline got a significant negative relation with the fare which convinces the expectation of author. To further interpret effect of Southwest presence, Wang performed another regression, however, all data involving Southwest as the reporting carrier was taken out of the dataset. It is interesting that even without the ticket data directly from Southwest, Southwest variable still has one of the biggest impacts in pricing. This finding convinced the Southwest’s entrance do affect the price in the specific airline market it joined.Other related papers extend the research to other impacts of the entry of the low-cost airlines, such as the effect on competitors on the market the low-cost airlines joined, the impact of alternative market in the same serving region, the consumer welfare, the reaction from the established airlines in the market the low-cost airlines entered, and the geographical competition in the whole air transport industry. Although investigations about the impact of the entry of low-cost airlines have been installed in different aspects, they all get the final conclusion related to the pricing effect. That is the low-cost carriers’ entrance definitely decreases the price with the increase of passenger traffic, resulting in the gain of consumer welfare. Dresner et al. (1996) examined the impact of low-cost carriers’ entrance on airfare regarding to alternative routes at the same airport as well as other airports with the same serving area. As a result, the outcome of the regression consolidates the previous conclusion. They further addressed that the presence of low-cost airlines reduces the yields while increases consumer welfare. In addition, to test if the consumer welfare exaggerated, Windle & Dresner (1998) extended their research by analysing the price change after ValuJet airline entering into the hub, Delta. Authors found that Delta lowered the fare on the routes ValuJet has joined without raising the price on the routes involved no low-cost airline which implies that the low-cost carriers’ entrance indeed enhances consumer welfare. Goolsbee and Syverson (2006) found that, on the routes Southwest joined, those incumbents decrease prices considerably. It is interesting that established carriers usually react to the threat of Southwest’s entrance in advance. In other words, they lower airfare on those routes as long as Southwest announced going to join in. This happens because they want to deter Southwest by pre-emptively depressing the price. But the reduction of price increases the passenger traffic on the specific route so that it hardly refuses all the new entrance. Based on the research of Goolsbee and Syverson, Daraban (2007) further examined the incumbent responses and spatial competition regarding to the entry of low-cost airlines and the conclusion totally demonstrated the previous studies. Likewise, Alderighi, et al. (2004) investigated how the full service carriers respond to the entry of low-cost airlines in terms of airfare pricing. Authors used monthly data in the first quarter of 2004 within the whole Europe, they corroborated that incumbents tend to depress the airfare to against the entry of low-cost carriers. And they further discover that the weak and strong interdependent which implies that direct competition of low-cost airlines also affecting on the established carriers.Most of papers reviewed above are published in the early 2000’s. At that moment, the boost of low-cost carriers was considered as a main growth power for the air transport industry. Nevertheless, the business environment all over the world has changed a lot after decades, especially after the financial crisis. Is the model still fit for the current situation? Is the price effects of low-cost airlines sustained past the initial promotional period? Abda et al. (2011) summarised a new trend for the impact of low-cost airlines growth on domestic traffic using the data of the top 200 largest US airports. They concluded that although low-cost carriers’ market share keeps growing, the extent is much less than before. Generally, the prices on routes with low-cost airlines depress less than those without entries. Meanwhile, people travelling on routes with low-cost airlines’ entrance are more elastic than those without entries, implying that people increase and decrease much more in good years and bad years, respectively. In this paper, empirical model will be performed, using the latest data, to investigate the effect of low-cost carriers’ presence in airfare and passenger traffic in terms of specific routes they entered under the new environment. Overall, to clarify these previous researches, the main literatures reviewed above are listed in a summary table below. Table 2.1 Literature summaryResearchDatasetMethodThe impact of the entry of low-cost airlinesPricePassenger trafficW&D (1995)PanelDescriptivedecrease 19% on averagedecrease 48% (WN effect)increase 182% on averageincrease 300% (WN effect)W&D (1998)Panel3SLSdecrease 53.3% (WN only)decrease 38% (multiple carriers)Vowles (2000)PanelOLSdecrease 45.47% on average decrease 77.61% (WN effect)Alderighi et al. (2004)Cross sectionOLSdecrease 42.58%Wang (2005)Cross sectionOLSdecrease 18.25% (WN effect)G&S (2006)Time seriesOLSdecrease 18.6% at the entry year and keep depressing afterwardsthe magnitude of the quantity response is roughly twice of the fare changesDaraban (2007)Time seriesOLSWN’s entry decrease average fare by 22% while depressing legacy carriers’ price by 17.6%Abda et al. (2011)PanelDescriptivesignificantly depress 5% more on routes with low-cost airlines’ entrance in 2005people are more elastic on routes with low-cost airlines’ entrance, implying increase and decrease more in good years and bad years, respectively MethodologyPreviously, many researches focusing on the impact of the entry are documented. This paper extends the study using panel data, which consider both cross-sectional and time series factors, to evaluate whether the effect of the presence of low-cost airlines on pricing and traffic has changed with time. 3.1 Sample constructionThe main data in this analysis sources from Domestic Airfare Consumer Report which is originally based on the Origin and Destination Traffic Survey conducted by the US Department of Transportation Bureau of Transportation Statistics. This report was first published in June, 1997 by the Department’s Office of Aviation Analysis. The information involves the 1,000 largest domestic city-pair routes covering 75% of all 48 states passengers and 70% of total domestic passengers (Domestic Airfare Consumer Report, 2010). This paper filters the top 100 city-pairs in the domestic US market ranked by the number of passengers in the third quarter of 2010 and matched to other quarters. Besides, data has been also collected from the Bureau of Economic Analysis and previous researches. The panel data consist of repeated observations on certain variables for a number of O-D pairs N at a number of points in time T. Here 100 O-D pairs for 54 points in time are selected from the second quarter of 1997 to the fourth quarter of 2010. Data include price, passengers, distance, income, largest market share, vacation and the presence of low-cost airlines. 3.2 Variables descriptionThe construction of variables is described below. To avoid the bias caused the heteroskedasticity as well as the huge volume variance between different variables and get the sufficient coefficients, all variables with large positive numbers have been transformed into natural log pattern, as seen with ln- prefix such as lnprice, lnpassenger, lndistance and lnincome.Lnprice: Average one-way fares are average prices paid by all fare paying passengers. Therefore, they cover first class fares paid to carriers offering such service but do not cover free tickets, such as those awarded by carriers offering frequent flyer programs (Domestic Airfare Consumer Report, 2010).Lnpassenger: This variable describes the number of passengers travelling on the specific route per day. And it counts both directions into one city-pair, for example, no matter travelling from Chicago to New York or from New York to Chicago, the person will be record in the city-pair of Chicago-New York. The expected relation between price and passenger is negative which implies that the more people travelling, the lower the price is. Lndistance: It shows the non-stop distance between two cities. The numbers used in the analysis are chosen from the fourth quarter of 2010, the latest report. Apparently, distance has the positive predictive relation with the dependent variable, price. Lnincome: The quarterly personal income for states has been collected from the Bureau of Economic Analysis. In order to match the figure of passenger variable, personal income level in both origin and destination cities have been summarised. Then, all figures are adjusted by quarterly inflation rates. It is assumed that people with high income are less affected by the level of price. In other words, the relationship between two variables is supposed as positive. Lg_mktshare: Largest market share represents the market share of the largest carrier on the specific route. This variable is reported in the Domestic Airfare Consumer Report with the name of the largest carrier. The largest market share, at some extent, indicates the concentration of a particular market. The particular market is intensive when the largest carrier takes a high proportion of market share, leaving other small airlines sharing little rest of market. In general, to compensate the loss on other market with fierce competition, the monopolist tends to set up the high price on the market due to the lack of competitor. Consequently, it is assumed that this variable is positive related to price. Vacation: The variable of vacation is a dummy variable to check if a specific city-pair is vacation route. It will be coded 1 when the origin or destination is considered as the vacation place and 0, otherwise. According to previous studies by Windle and Dresner in 1995, vacation cities almost centralised in four regions, including Florida, Hawaii, Nevada and Puerto Rico. Markets between vacation cities usually charge lower airfares implying a negative relationship between two variables. Cities considered as vacation places are listed in Table 3.1.Table 3.1 Vacation cities in modelRegionCitiesFloridaFort Lauderdale, Fort Myer, Miami, Orlando, Tampa, West Palm BeachHawaiiHilo, Honolulu, Kahylui, KonaNevadaLas Vegas, RenoPuerto RicoSan JuanLow: The last independent variable is the presence of low-cost airline. Likewise, the variable of low is a dummy variable coded 1 if any low-cost carrier participates on the route while 0 on contrary. Table 3.2 shows the list of low-cost airlines involved in this paper (Wikipedia, 2011 & Abda, et al., 2011).Table 3.2 Low-cost carriers in model (with carrier code)Allegiant Air (G4)AirTran Airways (FL)Southwest Airlines (WN)Spirit Airlines (NK)Frontier Airlines (F9)Sun Country Airlines (SY)ProAir Service(P9)Vanguard Airlines (NJ)America West Airlines (HP)Virgin America (VX)American Trans Air (TZ)Western Pacific Airlines(W7)JetBlue Airways (B6)USA3000 Airlines (U5)3.3 The estimating equationIn this paper, the panel dataset has been analysed in four models step by step: Pooled OLS model, Fixed effects model, Instrumental variable (IV) estimation and Panel IV estimation. In first two approaches, two regressions are estimated with the lnprice and lnpassenger as dependent variable, respectively. The independent variables are lnpassenger, lndistance, lnincome, lg_mktshare, vacation and low with the data variable of each quarter and year. The independent variables have been chosen from various related aspects representing a combination of demand, cost and market concentration which influence airlines pricing. The last two approaches use the two stages least squares (2SLS) regression. The construction details will be discussed later in this part.3.3.1 Pooled OLS modelGenerally, the starting point for panel data analysis is Pooled OLS model, so is it in this paper. The pooled OLS estimator treats all the individuals for all time points as a single sample so that the sample gains much bigger size compared to the simple cross-sectional data set. When the sample is sufficiently big, the coefficients of different variables will be assumed infinitely close to the true value. A common equation of pooled OLS model given below (Podestà, 2002):yit = β1 +k=kkβkxkit+eit.yit represents the dependent variable while xit is independent variable. i=1,…,N indicates the number of cross sections while t=1,…,T means the different point of time. k=1,…,K in this case representing the specific explanatory variable. However, when there are differences existing among cross-sectional observations, this model becomes improper on account of the heterogeneity bias caused by the variance of coefficient (Heyman, 2010). 3.3.2 Fixed effects modelConsidering the drawbacks of Pooled OLS model, a panel data model is performed as well. The three common approaches are fixed effects model, random effects model and mixed model. To use which one is naturally depend on different given situations. The fixed effects model imposes time independent effects for each entity that are possibly correlated with the dependent variable. In short, the difference between fixed effect and random effect is that the intercept is constant or not to the independent variables’ intercepts. Hausman test is the post-estimation test usually used to sort out which effects model to choose. In this case, the result of Hausman test indicates that the data collected fits the fixed effects model. The hypothesis of Hausman test is that the estimates for fixed effects model and random effects model have no significant difference. This hypothesis is rejected that implies these two models differ a lot resulting in the selection of using fixed effects model. The generic equation gives as follows (Paap, 2011):yit = α + x’itβ +εit.In the equation, yit represents the dependent variable while xit is independent variable. i=1,…,N indicates the number of cross sections while t=1,…,T means the different point of time.3.3.3 IV estimationAlthough the fixed effect approach recovers the heterogeneity bias in OLS model, it cannot deal with the endogenous problem. In addition, the fixed effect model may cause the omitted variable bias when it automatically ignores time-invariant variables. Moreover, both previous two regressions analysis the relation between price and passenger in one direction while, in fact, the relation is mutually affected. To further extend the model, the IV estimation using 2SLS regression is performed in the third step.Before use the IV estimation, it is necessary to make sure the correlation among variables by using test for endogeneity before installing the model (Shepherd, 2008). If the hypothesis is rejected which infers that the problem of endogeneity exists, the IV estimation gains its advantage, on contrary, may get even worse results than OLS models.The simplest equation for the basic IV method is (Cameron and Trivedi, 2009):y1i= y’2iβ1 + x’1iβ2 + ui, i=1, …, NIn the equation, y1i is the dependent variable while independent variables are consist of endogenous variables (y’2i) and exogenous variables (x’1i). It implies that the errors ui are uncorrelated with x’1i but correlated with y’2i which leads to the inconsistence of β. To fix this endogenous problem, the instrumental variable zi is required. It is assumed that zi fits the restriction that E(ui｜zi)=0. 3.3.4 Panel IV estimationFurthermore, because the dataset is in the panel pattern, the fourth step of Panel IV approach is undertaken. Commonly, the genetic equation for the 2SLS regression is:yit = x’itβ + αi + εit.Likewise, an instrumental variable, zit is required. It assumed that zit meets two assumptions. One is exogeneity while the other is correlated with the time-invariant error (αi) but uncorrelated with the time-varying component of errors implying E(εit｜zit)=0. So the equation represent a consistent estimation regressed of yit on xit with instruments zit (Cameron and Trivedi, 2009). The strength of the instruments impacts the quality of the model as a whole. In other words, the stronger the correlation between the instrumental variable and regressors is, the smaller the IV standard errors are. Once are the instruments too weak, the model is possible to lose the precision as well as get incorrect inference. 4. Descriptive AnalysisThe two sections above demonstrate economic evidences about the effect of the entry of low-cost airlines from documentary and modelling respects. However, after decades, both economic and industry environments have considerably changed. Furthermore, most researches before used cross-sectional model ignoring the effect in time series. Before generating a formal model, descriptive statistics of variables will be analysed in section 5. First of all, Table 4.1 summarised the means of all variables. The average one-way price on a route is 173.99 dollars. The mean distance of top 100 city-pairs is 991.43 miles, implying that the yield per miles is 0.175 dollars. Although the price and distance of average level both increase, the yield is as the same as the result got by Windle & Dresner in 1995. Every day, 2506.62 individuals on average travel between the origin and destination in both directions. The mean level of personal income considering both origin and destination states adjusted by inflation is 586497.54 dollars. As for two dummy variables, 24 of 100 top routes travel between vacation places while 68 percent of city-pairs, on average, have low-cost airlines involved. Table 4.1 Descriptive statisticsmeansdminmaxprice173.9973.6356.0550passenger2506.621506.29206.012034distance991.43642.66200.02704income586497.54321291.9475717.21645968lg_mktshare53.6917.8818.8100vacation0.240.430.01low0.680.470.01N5500On the top of the average level, all variables with positive numbers, such as one-way price, the number of passenger travelled per day, personal income by states and the largest market share, vary a lot and much more than it did in 1995. It insinuates that the market is getting competitive and differentiation, resulting in the offset of yields. Moreover, other factors like the sharp increase of oil price are right to explain this counteraction.To investigate the particular impact of the presence of low-cost airlines, several figures are performed below, showing the historical changes of the low-cost carriers and the relationship with other key variables.Figure 4.1 The movements of price, passengers and low-cost airlinesFigure 4.1 shows three key variables, the average one-way price, the passengers travelling per day and the number of low-cost airlines participating on the top 100 routes per year between 1998 and 2010. Figure 4.1 skips the numbers for 1997 and 2009, because the data are collected from the second quarter of 1997 and the data of first quarter of 2009 are entirely missed due to the reporting issues, respectively. In that case, these two figures are not comparable with those in other points of time. The line graph representing the presence of low-cost airlines illustrates a gently increase over 12 years which confirms the background discussed in the second section. In total, 360 low-cost airlines operate in the whole year of 2010. On one hand, the red line indicating the average one-way air fare almost keeps stable, fluctuating between 150 dollars and 200 dollars. Unexpectedly, the price does not go against the increase of low-cost airlines. On the other hand, the line representing the passengers travelling per day is not entirely increasing with the movement of low-cost airlines. Actually, it mildly waves between 2252 and 2740 people per day, reaching the bottom at 2002 and the peak at 2006. It appears to sum up that the price as well as passenger traffic effects of low-cost airlines have sustained past the initial promotional period. However, it is not entirely certain, considering that many other external and internal factors have changed like the sharp increase of oil price.Figure 4.2 The distance distribution on top 100 routes / average number of LCC per routeAfter vertical investigation by time series, Figure 4.2 presents the two variables horizontally based on the distance. On one hand, the blue bars representing the number of city-pairs at that level of miles show the distribution of top 100 flying routes. For instance, there are 2 of the top 100 routes between 2000 miles and 2250 miles, which are Las Vegas-New York and New York-Phoenix. From the bar chart, it is obvious that routes concentrate in relative short distance between 250 miles and 1250 miles with the most of 20 routes scattered in the range of 250 miles to 500 miles. On the other hand, the average numbers of low-cost airlines operating on a route for different distances are illustrated by red bars. Unlikely, Figure 5.2 does not support the negative relationship between the entry of low-cost carriers and the distance concluded in the early research (Windle & Dresner, 1995). Low-cost carriers no longer sorely participate in the short distance routes, but spread to long and popular city-pairs. The route involved the most low-cost airlines distribute around the distance of 1750 miles, at the average level of 47 carriers together operating on one market. However, the time points those low-cost airlines entering the long-distance routes are generally late than they do on the short-distance routes.Unfortunately, the two figures above seem to provide unexpected indications refusing the impact of the entry of low-cost airlines as a whole. To further explore influence of low-cost carriers, the variable of presence of low-cost airlines is deeply analysed. First, 45 of 100 top routes had already got low-cost airlines participating since second quarter of 1997, the beginning of the data collection in this paper. In the meanwhile, there are other 6 city-pairs having no low-cost involved over the whole sample period or the existence of low-cost carriers are too short to be taken into account. For these routes, it is hardly to indicate the entry impact on pricing and passenger traffic. After skipping this kind of routes, 49 city-pairs are left. Then, two line graphs representing the change of price and passengers after the low-cost airlines entering the route are performed. First, according to the Appendix 1, 22 out of 49 routes prove that when the low-cost airlines enter a particular market, the price on such route declines and keep the low level. Moreover, among 22 city-pairs, routes with long-distance capture the deeper impact than those short-distance routes do. In other words, the air fares on the long-distance routes decline more than those on the short routes. The graph (Appendix 1) shows the movements of the price in the four quarters before and after the entry including all 22 city-pairs. Here, Figure 4.3 only takes the route between New York and Seattle as an example. As seen in the figure, at the entry quarter, the price decreases by 40% of the highest pre-entry price, from 441 to 265 dollars. Then, the price keeps stable in the trend of declining although there is a slight stage back at the second quarter after entry. Nevertheless, this result is less than the outcome got by Windle & Dresner in 1995, almost 50% decrease after entry. Besides, the price decline of the city-pair of New York-Seattle is already the largest among the 22 routes depending on the graph (Appendix 1). Figure 4.3 One-way prices pro and post entryFurthermore, the situation of the passenger traffic is illustrated in Appendix 2. It is unpleasant to see that only 16 routes amongst 49 routes are considerably increasing after the low-cost enter such routes. In addition, there is no clue that the entry of low-cost carriers impacts the passenger traffic depending on the distance. How much that the passenger traffic influenced by the entry is random walk among these 16 routes. However, 14 of 16 routes are the routes those also sufficiently affected by the entry in the one-way price figure. It seems that the entry of low-cost airlines has simultaneous effect on both price and passenger traffic. Taking Ft. Lauderdale-New York route as an instance, Figure 4.4 shows the movement of passenger traffic pro and post entry. This route has the most obvious react to the entry of low-cost airlines. The passengers travelling on this route increases 54.2% upon entry, from 4114 to 6343 people per day. Although there is a slight downturn, it gets the peak at the fourth quarter after entry at the number of 6707 people per day, which is 63.4% higher than the lowest point. Also look backwards to the results in the Windle & Dresner’s paper (1995), the entry of Southwest Airline brought 300 percent more traffic to the specific route while all carriers increase 74% of the passenger traffic on the average level.Figure 4.4 The passenger traffic pro and post entryAfter the descriptive analysis, several initial conclusions can be drawn. First, the low-cost airlines keep increasing over the sample period. However, the boost is associated with neither the price declining nor the traffic increasing on average level of the top 100 city-pairs. Second, the figure also rejects the outcome from previous researchers that the low-cost airlines tend to participate only in the short-distance market. Finally, it should be admitted that the presence of low-cost airlines does influence the pricing and traffic on the particular route they entered but the impact is levelling off. However, the deep interpretation will be performed in the next section using several steps of formal statistical models.5. Model resultsDepending on the descriptive statistics, the impact of the presence of low-cost carriers still exists but on a lower level. In details, only 22 percent routes illustrate that the entry of low-cost airlines is associated with the price decrease. And the extent of price decline varies based on the distance of routes resulting in the longer routes have deeper decrease. Using the quarterly data from 1997 to 2010, this section will interpret the statistical meaning of data and analysis the result of panel data models. As described in the modelling section, the regressions for panel data will be performed in four steps.5.1 Pooled OLS modelThe beginning step of analysis is the Pooled OLS model. Table 5.1 lists the results of two OLS models. The first column is the regression with the variables of lnprice as the dependent variable while the second column model regresses the variable of lnpassenger. The standard errors have been adjusted for both regressions by the cluster of group number. R2 which interprets how much of the variability in the actual values explained by the model are 69.36% and 22.94%, respectively. In particular of the first regression, four of the six cross sectional independent variables are significant, including lnpassenger, lndistance, vacation and low. It is logical that the price increases with the flight distance. The price decreases when the route involves the vacation origin or destination whilst the number of passengers negatively influences the price on the particular route. Moreover, the presence of low-cost airlines sufficiently depresses the price on the route been joined. Usually, rich people are less elastic to price, so it is assumed that the airlines may set higher price in states those with higher personal income. Nevertheless, this sort of price discrimination is not existed as the lnincome is not significant related to lnprice. Meanwhile, the monopoly power is neither inferred according to the non-significant relation between the indicator of largest market share and price. Table 5.1 OLS regression results(1)(2)lnpricelnpassengerlnpassenger-0.0650***(-9.81)lndistance0.433***-0.0809***(71.16)(-5.29)lnincome-0.0528***0.246***(-8.96)(22.60)lg_mktshare-0.000468*-0.0109***(-2.26)(-29.68)vacation-0.276***0.129***(-44.85)(8.02)low-0.211***-0.0808***(-25.63)(-4.96)lnprice-0.250***(-9.70)_cons3.506***6.768***(36.18)(40.79)N53615361G100100t statistics in parentheses* p < 0.05, ** p < 0.01, *** p < 0.001When it comes to the results of second column, only three cross sectional variables are significant, including lnincome, lg_mktshare and lnprice. It is apparently that low price attracts more passengers. Rich people travel more since they have more disposable personal income. Furthermore, people prefer travelling on the routes with less concentration which offer them more choices. Unfortunately, the rest variables such as lndistance, vacation and low are not significantly related to the number of passengers travelling. The expected relation between flight distance and the number of passengers is negative. People prefer short travels considering the comfort, time as well as cost. Conventionally, the vacation places appear to attract more passengers inferring a positive relation between them. In addition, it has been convinced in the descriptive analysis section that the entry of low-cost airlines increases the passenger traffic. Here, to explain the non-significant relations, the travel objective change might be a reason. Currently, with the globalisation and liberalisation of sky, people increasingly travel for business. In that case, the impacts of distance, vacation and low become ambiguous.5.2 Fixed effects modelNevertheless, the results of OLS regression are likely to be biased depending on the inherent drawbacks of the model which discussed in the previous section of modelling. To check the reliability of the results, a further step is undertaken using the fixed effects model. Likewise, two fixed effects models with dependent variables as lnprice and lnpassenger, respectively, are performed with robust standard errors as follows. This adjustment of the standard errors prevents the inaccurate individual variances misleading the result by weighting them less. After the modification, all coefficients keep the same sign, however, with smaller t estimations. Two variables, lndistance and vacation, are skipped leaving rest independent variables all significant. It is due to the limitation of the fixed effects model. The model assists in controlling for unobserved heterogeneity when this heterogeneity is constant over time and correlated with independent variables. This constant can be removed from the data through differencing, for example by taking a first difference which will remove any time invariant components of the model (Wikipedia, 2011). Variables of distance and vacation are such constants that do not change over time. So the model omitted these two variables automatically.Table 5.2 shows the details of the results. On one hand, in the estimation for lnprice, most of the expectations are fulfilled. All variables except lg_mktshare are significant. As expected, the variable of lnpassenger has negative relation with the lnprice, implying that more passengers travelling on a specific route decreases the price. It is kind of promotion so that airlines earn small profit but with quick turnover in order to compensate other unpopular routes. Lnincome is positively related to the lnprice which hints that the people living in the richer region are willing to pay higher price for travelling. Finally, the variable representing the presence of low-cost airlines is negatively related to the dependent variable, lnprice. In the previous descriptive analysis, the impact of the entry of low-cost carriers has appeared to be proved that is getting weak and ambiguous. The statistical evidence given by the model convinces that, all other things being equal, the entry of low-cost carriers on a specific route sufficiently decline the price. As for the lg_mktshare variable, it has no significant relation with the lnprice when considering the date variables. On the other hand, the estimation for lnpassenger also gets much more significative results than OLS model does. Most variables get significant results, however, the presence of low-cost airlines which is the main research objective, keeps showing no influence to the passenger traffic. Generally, coefficients are too small to give sufficient interpretation.Table 5.2 Fixed effect model results(1)(2)lnpricelnpassengerlnpassenger-0.361***(-4.97)lnincome0.965***1.366***(3.92)(3.93)lg_mktshare0.00145-0.00315**(1.89)(-2.80)low-0.0600**0.0381(-3.26)(1.95)lnprice-0.730***(-13.50)_cons-4.621-6.108(-1.52)(-1.37)N53615361G100100t statistics in parentheses* p < 0.05, ** p < 0.01, *** p < 0.0015.3 IV estimationThe two steps before do not give satisfactory outcomes. As described in the modelling section, an IV model with the 2SLS regression is installed. At the beginning, it is very important to select the instruments. Table 5.3 shows the correlation matrix. Suppose that lnprice is correlated with the time-varying component of the error, then the simple fixed effect model becomes inconsistent and the variable of lnprice is needed to be instrumented. In this case, the instrumental variable is needed to be highly correlated with pricing but not directly determine the number of passenger travelling. Therefore, there are two choices among variables according to the correlation matrix. One is the variable of lndistance while the other is low. However, in the fixed effect model, the variable of lndistance has been dropped for the time-invariance. as for the variable of low, its correlation with lnprice and lnpassenger are -0.32 and 0.01, respectively. They differ a lot. Moreover, the correlation with lnpassenger is not significant under 95% confidence interval. Hence, the variable of low representing the presence of low-cost airlines is the most proper instrumental variable. It implies that the low-cost airlines impacts passenger traffic by influencing the pricing level. Table 5.3 Correlation matrix　lnpricelnpas~rlndis~elninc~elg_mk~evacat~nlowlnprice1　　　　　　lnpassenger-0.14*1　　　　　lndistance0.72*-0.05*1　　　　lnincome0.08*0.20*0.08*1　　　lg_mktshare-0.28*-0.27*-0.42*0.12*1　　vacation-0.23*0.12*0.11*-0.22*-0.12*1　low-0.32*0.01-0.12*-0.03*-0.010.10*1* p < 0.05Since the instrumental variable has been selected, a basic IV estimation is performed. Considering that the main purpose of this step is to make sure which method (OLS or 2SLS) is more suitable for the database, the regression results are not reported. Besides, a test for endogeneity is performed after running 2SLS regression. And the hypothesis that the variables are exogenous has been rejected implying the IV estimation is much more reliable than OLS method. 5.4 Panel IV estimationThe three steps above are exploring the most suitable approach for the dataset. Finally, the panel IV method is considered as the best choice which can either fix the endogenous problem or interpret the mutual causality between price and demand. Furthermore, p-value of Hausman test is 0.9727 which accept the hypothesis. In other words, the panel IV estimation should be run under the random effect model. Hence, it avoids the omitted variable bias in the second step caused by the fixed effect model. Table 5.4 presents the outcomes.Table 5.4 2SLS regression results1st Stage2nd Stagelnpricelnpassengerlnprice-1.066***(-13.57)low-0.102***(-16.37)lndistance0.470***0.359***(9.84)(4.97)lnincome0.187***0.685***(3.93)(10.85)lg_mktshare0.003***-0.00206***(13.89)(-4.54)vacation-0.211**0.0452(-2.79)(0.46)_cons-0.5731.826*(-0.86)(2.13)N53615361G100100t statistics in parentheses* p < 0.05, ** p < 0.01, *** p < 0.001Similarly, the first column indicates the results of first stage while the second stage outputs are recorded in the second column. And the R2 is equal to 9% which is considered as quite acceptable in panel IV estimation. In the first stage, all independent variables are significant with expected sigh. Especially, it is pleasant to see that the largest market share has significant positive effect on pricing. Although the coefficient (0.003) is lower than other variables, the market concentration does affect the air fare sufficiently. The more intensive the market is, the higher price is. It is fit for the theoretical evidence that when the monopolist controls a specific market, they have power to increase the price to compensate their business on other depressing markets (Wang, 2005). The presence of low-cost airline keeps the negative relation with the price. Because low is a dummy variable, it can be concluded in details that one more low-cost airline enters the specific route, the price decreases by 10.2%. As for the second stage, lnprice which is instrumented by low negatively influence the number of passengers. Passengers are elastic so that they tend to switch from expensive routes to cheap ones. The mathematical relation is 1 dollar decrease on price leads to 1.07 more passengers. However, the vacation loses the significant relation with passenger traffic. There are several reasons may explain it. On one hand, from the statistic aspect, lnprice from the current period is instrumented. However, it is not an external variable added especially for IV estimation, but one of variables in previous models. If the time-varying errors are independent, then it is not suitable to be a valid instrument (Cameron & Trivedi, 2009). On the other hand, in terms of reality, the purpose of travelling changes a lot recently. People fly to different places not only for travelling but also for business. Passengers cannot choose destination when they travel on business. Meanwhile, those traditional resorts are dropping attraction for tourists pursue diversification and customisation nowadays. In sum, four econometric methods have been used step by step in this section. Beginning with the simplest OLS model, the results are not acceptable due to the inherent method drawback. The fixed effect model solves the problem of heterogeneity but leaves the omitted variable bias. Furthermore, to deal with the endogenous problem and mutual causality, IV estimation is used in the third step. In the end, the panel IV estimation resolves all problems and gives the final output of this paper. The impact of the entry of low-cost airlines on pricing is significant negative. Nevertheless, the presence of low-cost airlines does not directly affect the passenger traffic any more. However, it shows additive effect among the presence of low-cost airlines, price and passenger traffic through the recurrence relation. One additional low-cost airline enter the route reduce 10.2% of price. 1 dollar depressing of price brings 1.07 more passengers to the route. Consequently, it infers an additional presence of low-cost airline increases passengers travelling on that route by 10.9%. 6. ConclusionAs one of the most popular topic in the airline transport, the impact of the entry of low-cost airlines on pricing and passenger traffic has been documented a lot. This paper extends the previous researches to evaluate the influence in the current circumstance. Having the descriptive as well as the modeling analysis above, this section will conclude the findings with comparison with other researches and advices for the low-cost airlines. Some limitations and further research suggestion are given in the end of this paper. 6.1 Comparison with other researchesIn the literature review section, a summary table has been drawn to briefly demonstrate results found by previous researchers. To clarify the comparison, the table is restored with findings of this paper (Table 6.1). Obviously, the conclusion drawn from this paper generally supports the previous findings. Initially, the low-cost airlines’ entrance does reduce price associated with the increase of the number of passengers. However, the differences are in two main aspects. First, the impact is sufficient level off. From the 38% to 10.2%, the negative influence on price is apparent shrink. Meanwhile, the entry of low-cost airlines does not directly affect the number of passengers any more, but mediately affect through price. The reasons leading to these differences might be as follows. First, the data collected in this paper is from the second quarter of 1997 to the fourth quarter of 2010. It covers 14 years with 54 quarters which is much longer and newer than any other literatures reviewed in section 2. The papers reviewed are mostly published in the early 2000’s. At that moment, the boost of low-cost airlines was considered as a main growth power for the air transport industry. Nonetheless, the business environment all over the world has changed, especially after the financial crisis. Hence, the new data gives the new trend of the impact of low-cost airlines. Depending on Table 6.1, given the same method (OLS), the impact on price has about halved from 1998 to 2007. On top of the time changes, method is another reason for the different conclusion. As seen in the Table 6.1, although some of them used the panel data, they only used OLS method which has inherent drawbacks when analysis the panel dataset. In this paper, this kind of drawbacks has been resolved stepwise by advanced approaches, including fixed effect model and IV estimation. Based on the results of IV estimation, the impact on pricing has been halved again, from 22% to 10.2%. As indicated in previous sections, the conclusions of this paper are more reliable than others.Table 6.1 Results comparisonResearchDatasetMethodThe impact of the entry of low-cost airlinesPricePassenger trafficW&D (1995)PanelDescriptivedecrease 19% on averagedecrease 48% (WN effect)increase 182% on averageincrease 300% (WN effect)W&D (1998)Panel3SLSdecrease 53.3% (WN only)decrease 38% (multiple carriers)Vowles (2000)PanelOLSdecrease 45.47% on average decrease 77.61% (WN effect)Alderighi et al. (2004)Cross sectionOLSdecrease 42.58%Wang (2005)Cross sectionOLSdecrease 18.25% (WN effect)G&S (2006)Time seriesOLSdecrease 18.6% at the entry year and keep depressing afterwardsthe magnitude of the quantity response is roughly twice of the fare changesDaraban (2007)Time seriesOLSWN’s entry decrease average fare by 22% while depressing legacy carriers’ price by 17.6%Abda et al. (2011)PanelDescriptivesignificantly depress 5% more on routes with low-cost airlines’ entrance in 2005people are more elastic on routes with low-cost airlines’ entrance, implying increase and decrease more in good years and bad years, respectively This paperPanelOLSone more entry decreases 21.1% of price on averageone additional entry decreases passengers by 8.08%This paperPanelFEone more entry decreases 6% of price on averageone additional entry increases passengers by 3.81%This paperPanel2SLSone additional entry decreases 10.2% of price1% price reduction leads to 1.07% more passengersone more entry increases passengers by 10.9%6.2 Implications for low-cost airlinesAccording to the conclusion, low-cost airlines are losing impact on pricing and passenger traffic. It convinces the conventional strategy for growth may not continue to optimistically work in the future. Hence, how to maintain cost advantages and capture new opportunities become the most important tasks for low-cost carriers (Bundgaard et al,. 2006). To keep cost advantage, low-cost airlines need increase the efficiency of using fuel. Most of low-cost carriers are still using very old and simplex fleet which are very low efficient of using oil. The dramatically increasing fuel price no doubt burdens the low-cost airlines. Since it is not possible to control the fuel price, the only way to reduce the cost is to utilize fuel more efficient. As for the new growth opportunities, low-cost airlines have already entered the most highly profitable city pair routes. Competition on those routes is getting fierce and expensive. So they are supposed to seek new growth points on those hub cities served by weakened legacy carriers, or international destinations might be another choice. 6.3 Limitation and further researchCompared with previous researches, this paper gives a more reliable interpretation for the impact of entry of low-cost airlines on pricing and passenger traffic. However, there is still a limitation and further researches are required.Looking backward to the descriptive analysis (Section 4), Figure 4.3 shows the price changes pro and post entry. The dummy variable of low is taken into account at the entry quarter. However, the depressing impact has already existed from two quarters before the entry. It is the same with Figure 4.4 which indicates the movements of passengers pro and post entry. The number of passengers enhances since two quarters before the entry. These two figures support the conclusion of Goolsbee & Syverson (2008) that incumbents tend to deter low-cost airlines by pre-emptively depressing the price, however, the entry cannot be avoided due to the increase on passenger flow.This lagged dummy variable appears to influence the model results. So for the further researches, they are supposed to consider taking into account the entry point in advance, for example, put the dummy at the point of two quarter before the entry.References[1] Bailey , Graham & Kaplan (1985), Deregulating the Airlines, London, England[2] Doganis, R. (2006), The Airline Business, Bodmin, Great Britain[3] Cameron & Trivedi (2009), Microeconometrics using Stata, Texas, United States[4] Strassmann (1990), Potential competition in the deregulated airlines, The Review of Economics and Statistics, Vol. 72, No. 4 (Nov., 1990), pp. 696-702[5] Whinston & Collins (1992), Entry and Competitive Structure in Deregulated Airline Markets: An Event Study Analysis of People Express, The RAND Journal of Economics, Vol. 23, No. 4 (Winter, 1992), pp. 445-462[6] Bennett & Craun (1993), The airline deregulation evolution continues: The Southwest effect, presented by Office of Aviation Analysis, United States[7] Windle & Dresner (1995), The short and long run effects of entry on U.S. domestic air routes, Transportation Journal, Winter 1995, 35, 2, pp. 14[8] Windle & Dresner (1999), Competitive responses to low cost carrier entry, Transportation Research, Part E 35 (1999), pp. 59-75[9] Dresner, Lin & Windle (1996), The impact of low-cost carriers on airport and route competition, Paper presented at American Economic Association Meetings, Maryland, United States[10] Vowles (2000), The effect of low fare air carriers on airfares in the US, Journal of Transport Geography, 8 (2000), pp. 121-128[11] Wang (2005), The effect of a low cost carrier in the airline industry, Paper presented at MMSS Honors Seminar, United States[12] Goolsbee & Syverson (2008), How do incumbents respond to the threat of entry? Evidence from the major airlines, The Quarterly Journal of Economics, Retrieved from [13] Abda (2011), Impacts of LCC growth on domestic traffic and fares at largest US airports, Journal of Air Transport Management, 18 (2011) pp. 21-25[14] Daraban (2007), Low cost carrier entry, incumbent responses and spatial competition in the US airline industry (PhD Thesis, The Florida State University, United States)[15] Alderighi, Cento, Nijkamp & Rietveld (2004), The entry of low-cost airlines, Unpublished paper, Tinbergen Institute Discussion Paper TI 2004-074/3.[16] Bundgaard, Bejjani & Helmer (2006), Strategic report for Southwest airlines, Out of the Box Consulting, Pandora Group[17] Paap, R. (2011), Lecture 4: Applied Microeconometrics [PowerPoint slides], Rotterdam, The Netherlands, Erasmus University Rotterdam[18] US Department of Transportation (2011), Airline Origin and Destination Survey (DB1B), Retrieved from [19] Bureau of Economic Analysis (2011), Regional Data, Quarterly State Personal Income: Personal income (SQ1), Retrieved from 1 Price movement pro and post entryAppendix 2 Passengers movement pro and post entry ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download