UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Airport charge, traffic volume and car rental price : empirical evidence at US airports Shi, Zijun 2014

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2014_september_shi_zijun.pdf [ 1.32MB ]
JSON: 24-1.0102474.json
JSON-LD: 24-1.0102474-ld.json
RDF/XML (Pretty): 24-1.0102474-rdf.xml
RDF/JSON: 24-1.0102474-rdf.json
Turtle: 24-1.0102474-turtle.txt
N-Triples: 24-1.0102474-rdf-ntriples.txt
Original Record: 24-1.0102474-source.json
Full Text

Full Text

AIRPORT CHARGE, TRAFFIC VOLUME AND CAR RENTAL PRICE: EMPIRICAL EVIDENCE AT US AIRPORTS  by Zijun Shi  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES (Business Administration) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  July  2014  © Zijun Shi, 2014 ii  Abstract This paper examines the interaction among airport aeronautical charge, traffic volume, and car rental service charge by employing a cross-section dataset covering 337 airports in the United States. Other determinants of the aeronautical charge are also examined. Using the method of three stage least squares, the main empirical findings are: (1) Car rental price has no significant impact on passenger volume, indicating that rental cost, which is an important cost category of airport concessional goods, does not affect passenger volume of airport. (2) Car rental price has no significant effect on aeronautical charge. (3) Only rental price at airports with a very low transfer rate responds positively to passenger volume. Overall, rental price does not respond to passenger volume or aeronautical charge. (4) Aeronautical charge has a significant negative effect on passenger volume, but it has no significant effect on car rental price.   iii  Preface  This dissertation is original, unpublished, independent work by the author, Zijun Shi.    iv  Table of Contents  Abstract ..................................................................................................................................... ii Preface ..................................................................................................................................... iii Table of Contents ..................................................................................................................... iv List of Tables ........................................................................................................................... vi List of Figures ......................................................................................................................... vii Acknowledgements................................................................................................................ viii Dedication ................................................................................................................................ ix 1 Introduction....................................................................................................................... 1 1.1 Airport management and relationship with airlines ................................................... 2 1.2 Airport revenues ......................................................................................................... 4 1.3 Regulation policies on airport charges ....................................................................... 6 1.4 Car rental service in airports ...................................................................................... 7 1.5 Research questions ..................................................................................................... 7 1.6 Main results ................................................................................................................ 9 1.7 Organization of the thesis ........................................................................................ 10 2 Related Literature ........................................................................................................... 11 2.1 Concession and airport regulation............................................................................ 11 2.1.1 Theories assuming non-existence of concession price effect ........................... 11 v  2.1.2 Theories based on existence of concession price effect ................................... 14 2.1.3 Empirical studies on concession and travel activities....................................... 21 2.2 Car rental pricing ..................................................................................................... 22 2.3 Airport charge determinants .................................................................................... 24 3 Data, Econometric Model and Hypotheses .................................................................... 28 3.1 Data .......................................................................................................................... 28 3.2 Econometric Model .................................................................................................. 29 3.3 Hypotheses ............................................................................................................... 40 4 Methodology ................................................................................................................... 42 4.1 Endogeneity ............................................................................................................. 42 4.2 Three stage least square (3SLS) estimation ............................................................. 44 5 Estimation Results and Discussion ................................................................................. 46 5.1 Car rental pricing ..................................................................................................... 48 5.2 Concession price effect ............................................................................................ 49 5.3 Aeronautical charge determinants ............................................................................ 50 6 Robustness Check ........................................................................................................... 52 7 Conclusion ...................................................................................................................... 63 References............................................................................................................................... 66 Appendix................................................................................................................................. 69 Included Airports and Hub Status ...................................................................................... 69 vi   List of Tables Table 1    Descriptive Statistics .............................................................................................. 31 Table 2    3SLS and OLS Estimation Results ......................................................................... 46 Table 3    3SLS Estimation Results (use measure of "charge per passenger") ....................... 52 Table 4    Group by Percentile of Transferring Passenger Rate ............................................. 53 Table 5    Group by Percentile of Transferring Passenger Rate ............................................. 56 Table 6    Group by Car Types ............................................................................................... 60    vii  List of Figures Figure 1    Histogram of Concession Share .............................................................................. 5 Figure 2    Model Specification .............................................................................................. 29 Figure 3    Histogram of Three Dependent Variables ............................................................. 30 Figure 4    Histogram and Boxplot of Transfer Rates............................................................. 34 Figure 5    Cargo Rate in Airports .......................................................................................... 34 Figure 6    Market Structure of Airport Car Rental ................................................................ 35 Figure 7    Distribution of Herfindahl Index for Airlines ....................................................... 37 Figure 8    Histogram and Boxplot of Southwest-airline Share .............................................. 38 Figure 9    International Traffic Rate ...................................................................................... 38    viii  Acknowledgements First, I would like to express my sincere gratitude to my supervisor Professor Anming Zhang whose patient guidance, support and suggestions have been an immeasurable help throughout my master study. I also owe special thanks to Professor Ting Zhu and Professor Robin Lindsey, for spending valuable time giving me advice and serving as my thesis committee members. I would also like to express my gratitude to Professor David Gillen, who always gives me support and encouragement on my study during the first year. My thanks also goes to Elaine Cho, who has provided lots of detailed administrative supports throughout my master program. Finally, I want to thank my parents, who have supported me both morally and financially throughout my years of education.   ix  Dedication To my parents  1  1 Introduction One of the most striking and consistent trends in the airport sector over the last twenty-five years has been the growing importance of “concession revenues”. The latter includes revenues from retailing, advertising, car rentals, car parking, and land rentals (e.g., Zhang and Zhang 1997 and 2003, Forsyth 2004, and Thompson 2007) as compared to the traditional aeronautical revenues associated with runways, aircraft parking and terminals. Airports worldwide currently derive as much revenue, on average, from concession services as from aeronautical ones (e.g., Zhang and Czerny 2012). A number of studies have elaborated on the relationship between the aeronautical and airport concession services. As discussed in more details below, some authors assume that concession prices can influence passenger demand for flying, while others may question whether such a relationship actually exists. The primary purpose of the present thesis is to determine whether this relationship exists; and if so, what the nature of the relationship is. This is done by empirically examining airports in the United States. In the remainder of this section, we introduce the airport business in the United States for clarifying the background against which we explore the answer to our research questions. Different features of US airports relative to their European counterparts will also be introduced. Air transport market is highly developed and extensively used, to a larger extent in the U.S. than in other parts of the world. The U.S. air transport market dominates the airports industry worldwide with a large proportion of the world's largest airports. 2  1.1 Airport management and relationship with airlines In the United States, almost all airports are publicly owned, by either local or federal government. Some airports are operated directly by government, others by separate public organizations. Examples of these public organizations are multipurpose port authority and airport authority, which are set up by government to endow airports with more autonomy to have them better operated and managed. This form of airport management has been very common in the United States for a long time. Another feature that the US airports specifically have is their relationship with airlines or carriers. Even though US airports are publicly owned, they are operating closely with privately owned airlines. In fact, the airlines take a large share of financial risks of operations and investment at many airports. According to Rigas Doganis (2005): "US airports enter into legally binding contracts known as airport-use agreements which detail the conditions for the use of both airfield and terminal facilities. These contracts are negotiated between the airport and its airline customers. They will specify the fees and rental rates which an airline has to pay and the method by which these are to be calculated." Two basic approaches to measure how airport charges airlines are listed as follows. The first approach is called residual approach. After taking into account other non-aeronautical revenues (e.g. concession revenues), the airlines cover the net costs of operating the airport. In this way, the airlines are taking a big financial risk to guarantee that they will pay enough to the airports such that the airport revenue will always balance out costs. There may be times when passenger volume is low and airlines are getting less and less revenue, 3  the airlines have to make sufficient payment to cover airport net costs as agreed. Therefore, the airports actually don't suffer any losses. The second approach is compensatory approach. This method has airports shoulder almost all risk of operating the airport. Airports charge the airlines fees only to cover the costs of services the airlines use, rather than the net costs of running the airport. To accommodate particular needs of airports, the above two basic approaches are applied by airport managements in different ways, such as a mixed approach which combines both the first and the second methods. The airport-use agreements are traditionally long-term contracts, ranging from twenty to fifty years. However, since the deregulation in the late 1970s, there has been a trend towards shorter term contracts, which implies that the airports and airlines tend to be more flexible in an uncertain air transportation market. Financially, some airports have abandoned the residual approach and adopted a more compensatory approach. This choice more likely applies to financially secure airports which have large market power and do not have to rely on airlines to balance costs.  In many airports, the airlines rent terminals from airports for exclusive use or joint use, which gets the airlines involved in airports' investment decision. For landing rate, most differentiate by signatory and non-signatory airlines rather than international or domestic. Honolulu is among the very few airports that have different rates for international and domestic traffic. Details are specified in use agreements which concern signatory airlines, while signatory airlines all together carry a majority of the airport's total traffic volume. 4  It is to some extent fair to say that airlines are helping airports run the aeronautical operations at many US airports. Airlines construct their own support facilities and maintenance on the land rented from airports. Some even build their own terminals. (Rigas Doganis, 2005) 1.2 Airport revenues Airport revenues are generated from aeronautical operation and commercial or concessionary services. Aeronautical operating revenue includes landing fees, passenger terminal fees, cargo and hangar rentals, aircraft control charges and other aircraft-related fees. Among all the categories, landing fees and terminal fees are the most important ones. Concession revenue arises from non-aircraft-related commercial activities such as duty-free sales, food and beverage, car rental, car parking, accommodation and so on. As discussed above, US airports usually determine concession fees (commercial charges) to maximize revenue, while aeronautical charges are determined by the airport-use agreement, usually on a cost-base tradition.  As traffic grows, concession revenue builds up an increasing percentage of airport's total revenue, which largely comes from expansion of shopping, increased demand of car rental and parking service as well as food and beverage. With revenue information available from the City-data website, Figure 1 shows the share of concession revenue out of the total airport revenue for United States airports in 2005, indicating that a percentage over 50% is commonly observed across all airports. This number even rises above 80% for some airports, most of which are medium to large airports. Moreover, concession revenue has been growing faster than aeronautical revenue. 5  Figure 1    Histogram of Concession Share  US airports differ from European airports in the proportion of revenues contributed by different form of concession operations. One important feature of US airports is their large proportion of revenue from car rental and car parking services, while duty-free concessions generate relatively lower revenue because international passengers count only a small proportion. As airlines usually rent the airport’s terminal for joint or exclusive use in the US, airports generate more revenue from rents and much less amount from landing fees than their European counterparts. According to Rigas Doganis (2005), US airports tend to have a lower average unit revenues (4.45 in 1989--1990) than European airports (13.5 in 1989—1990), because US airports have lower airport charges. One reason is that they offer less additional services, for example handling. Another important reason is that the large proportion of domestic passengers lead to less concession revenue. Domestic passengers tend to consume less catering and other concession goods, especially duty-free goods. For airports in the United States, car rental and car parking generate a much larger share of concession revenue, while in Europe, revenues from duty-free goods make up a larger proportion.  6  1.3 Regulation policies on airport charges Airports, especially large ones, tend to be location-based monopolies and thus abuse their market power. To prevent them to do so, airport charge regulation should be put into force. There are various forms of airport charge regulations including traditional rate of return regulation, price-cap regulation, cost-based regulation and so on. With two main resources of revenue as mentioned above, price-cap regulation differs in whether to take concession revenue into account when regulating aeronautical charges. Under a single-till form of price-cap regulation, both concession and aeronautical profits are considered to setting price cap. This is a more traditional approach than the other approach called dual-till regulation, which regulates aeronautical charge based on only aeronautical operations. Rate of return (ROR) regulation allows a level of return on a specified rate. As mentioned in Tretheway (2001), ROR regulation has weakness for being expensive to implement, unresponsive and complicated. Price cap regulation is much cheaper to enforce and it also provides incentives for airports to reduce costs. However there is debate on whether single-till or dual-till regulation should be implemented. The US basically adopts a cost-based regulation, which requires the airports to set charges such that the aggregate revenues reflect the cost of services. Given the principles of residual financing, it also seems similar to a single-till price cap regulation. While in other parts of the world, price-cap and ROR regulation are the dominant forms. To make a better choice among different regulations, or to set an optimal value for parameters involved in each regulation, it is essential to have a better understanding of the interaction between concession activities and aviation services. 7  Research or studies conducted on regulations for European airports may not be in accordance with US airports. Compared to European airports, forms of ownership and management of airports in US are quite different. The important difference is that US airports are publicly owned, and thus may not behave competitively or pursue profit as ultimate goal. However, a large proportion of airports in Europe are private entities.  1.4 Car rental service in airports In the United States, most car rental companies derive a significant proportion of their total revenues from airports. For passengers arriving at an airport, the most convenient and accessible rental service they can get is from the car rental companies located inside the airport. To set agencies inside airport terminals, rental companies can attract more airline passengers with a reasonable higher price than those located outside airports. Typically, car rental firms in airports give ten percent of total revenue to the airport, while off-airport car rental firms pay much less based on the cost of offering ground access. This stream of revenue counts around 20% of airports' total concession revenue among all concessionaires. Given the overall number of passengers arriving at the airport, the proportion of transferring passengers out of total passengers differs, which we call transfer rate. The demand for car rental services varies by transfer rates because transferring passengers may not exit the airport and do not need to rent cars. 1.5 Research questions As mentioned above, concession revenue of airports contributes a large proportion of total revenue and are becoming even more profitable than aeronautical operations. For policy makers and managers, it is interesting to know how the concession operation interacts with 8  the traditional aeronautical operation, to make optimal policies or strategies. All these raise several important empirical questions that we investigate in this study. Our research questions in this context include the following: 1. For airports where concession service charges higher price, will the passenger volume be lower, or stay unaffected? The empirical question whether concession service can change the passenger quantity or not, has got little solution in the literature. There are basically two hypotheses. One is that either the price of concessional goods is not a cost significant enough to affect people's travel decision, or people are not aware of the price when making travel decision. Price then has no impact on passenger volume. The other hypothesis is that concession price may affect passenger volume since many frequent passengers, for example business travelers, are fully aware of the cost and benefit associated with concession goods. Thus they will take concession cost into account when making travel decision. Even for some leisure passengers, they may also make a quite comprehensive plan involving consumption of concession goods. This is perhaps the most controversial issue, so one goal of this paper is to check the effect of concession price on passenger volume through employing the car rental price as a representative of airport concession services. 2. Are concession prices correlated with aeronautical charges in some perspective? What other factors may affect airport aeronautical charges? 3. What other factors affect airport passenger volumes and concession prices?  9  1.6 Main results We use a cross-section dataset covering 337 airports in the United States to examine these questions. We combine airport characteristics such as revenue and passenger volume, and car rental market information including the posted rental price (on Orbitz and Expedia websites) and number and identities of rental firms in each airport. By employing three stage least square (3SLS) method, we have the following empirical findings. First, there is no significant effect of car rental price on passenger volume of airports, indicating concession price has no significant impact on demand for aeronautical service. Second, car rental price has no significant effect on aeronautical charge. Third, only rental price in airports with a very low transfer rate responds positively to passenger volume. Overall, rental price does not respond to passenger volume or aeronautical charge. Fourth, aeronautical charge has significant negative effect on passenger volume, and it has no significant effect on car rental price. Finally, the results of determinants of aeronautical charge per passenger are as expected and most are in line with earlier works.  With respect to aeronautical charges, we find the following the significant. There exists a negative effect of passenger quantity on aeronautical charges, possibly because the total charges contain a fixed component. The average distance for arriving flights is positively associated with aeronautical charges, consistent with the tradition of weight-based charges. Higher costs are associated with higher aeronautical charges, in accordance with the cost-base charges adopted by publicly owned airports. Aeronautical charges are higher at airports located in higher-GDP counties or airports that have more total arriving flights.  10  1.7 Organization of the thesis The rest of the thesis is structured as follows. Section 2 reviews the related literature. Section 3 presents the details of data source, econometric model and hypothesis. Section 4 describes the methodology employed. Section 5 discusses the empirical results. Section 6 provides some robustness check. Section 7 concludes.   11  2 Related Literature Our study mainly intersects three domains of literature: airport behavior with concession revenue, car rental pricing and factors affecting airport charges. 2.1 Concession and airport regulation Researchers have created a wealth of theoretical work on the role of concession operations in airport productivity, pricing and regulation policy making. There are basically two streams of theoretical works, with the first stream based on the assumption that people make simultaneous decisions on buying concession goods and flight tickets. Namely, concession price has effect on passengers' travel decision, and we call this effect "concession price effect" in this article. The second stream assuming that people make these two decisions sequentially. That is, travel decision is made independently of the decision of concession consumption. Below we introduce details of some works that are most related to our research interest. 2.1.1 Theories assuming non-existence of concession price effect Zhang and Zhang (1997) built a model to examine whether the profitability of concession operations should be a concern from a social welfare perspective. The authors first analyzed the case where the budget constraints are added to aeronautical and concession operations separately. They found that, without concession operations, the social-marginal-cost pricing would not cover the expenses for running the airport, and this problem will become even severer with higher traffic volume. The model then adds a budget constraint to the aeronautical operations, getting a Ramsey pricing which makes the aeronautical charge deviate from the first-best pricing, namely higher than social marginal cost. 12  The authors then consider the commercial operations together with aeronautical operations, implicitly assuming that the demand for these two services are independent. They believe that the passengers make decision on whether to consume concession goods only after they arrive at the airport and observe the concession prices, rather than before booking flight tickets. Under an "overall break-even constraint", the authors show that the optimal solution calls for a subsidy from commercial activities to aeronautical operations. With this subsidy, the welfare gain, which comes from reducing the markup over marginal cost on aviation side, outweighs the welfare loss from inducing markup over marginal cost on the concession side. Therefore, compared to the Ramsey pricing on aeronautical operations and marginal-cost pricing on the commercial operations, allowing profitability for concession and cross-subsidy from concession revenue to aeronautical revenue is socially desirable. While building up the theoretical model, they mentioned that this argument applies only to the case where the purchases of these two services happen separately in time, but may not be true for some commercial services such as car rental and airport hotel. Our empirical study examines exactly the interaction between car rental service available at airports and the aeronautical operations, and will add more credits to this theoretical study by showing that even car rental is a consumption separated in time from flight ticket purchasing. Zhang and Zhang (2003) considers airports with two different objects: social welfare maximization and profit maximization. This paper focuses on airports' decision on charges and capacity expansion according to these two objects. Theoretically, it is shown that if airports are allowed to earn profits from concession operations, the social welfare will be higher than that without profitable concession activities. The first-best social optimal pricing is not financially feasible, thus two commonly used structures are examined: public airports 13  with revenue constraint and privatized airports with goal of profit maximization. The analysis shows that both types charge a markup over social-marginal cost, but the privatized airports charge higher than public airports, given both would get the same amount of commercial profits. In this paper, the social welfare function is defined as below:  ∫ 𝑄(𝜉, 𝑡)𝑑𝜉∞𝜌+ 𝑃𝑄 − 𝐶(𝑄) − 𝐾𝑟 + 𝑄 [∫ 𝑋(𝜉)𝑑𝜉∞𝜌+ 𝑝𝑋 − 𝑐(𝑋)] (2.1) which is the sum of producer surplus and consumer surplus. For the above formula, the implicit assumption is that the demand for aeronautical service and concession goods are independent, that is, the two consumption decisions are made separately. Here comes to the controversial issue, that is, do passengers take into account the surplus from commercial goods offered inside the airports when deciding whether to buy a ticket? Since the two activities happen separately in time, people may not even consider the potential concession service when making the travel decision. Another possibility is, without arriving at the airport, people may not know the information about the commercial goods at the specific airport, thus they are not able to make the cost and benefit analysis. Therefore, the base of the whole theoretical analysis may not be solid without having more evidence to support this assumption. Lin (2006) builds a new model for competition between hub airports, which formulates non-price competition, price competition among the major airlines in the airport, and the transfer passengers’ consumption of concession goods. Although concession consumption of transfer passengers apparently does not include car rental services which only happen for passengers traveling outside airports, the issue whether concession price has effect on travel decision 14  applies to both transfer and terminal passengers. The author puts explicitly in the paper that he assumes the demand for flights and concession goods are independent: consumers first decide travel decision based on the ticket prices, then decide whether to buy concession goods after arriving at the airport.  Yang and Zhang (2011) examine the price-cap regulation for congested airports where airlines have market power, showing that different form of price-cap regulations (i.e. single-till and dual-till) performs better than the other under different situations. The modeling of the interaction between demand for concession goods and demand for flights is consistent with the papers discussed above: a consumer makes these two decisions sequentially and he (she) buys concession goods only if he (she) has arrived at the airports and his (her) valuation of the concession good is higher than the price.  Gillen and Mantin (2014) investigate the role of concession revenue in the decisions of airport privatization. The results demonstrate that, if the potential for concession revenues are not large enough, the privatization is not recommended. From their model, we can see the flights demand function contains no effect from concession assumption, while concession demand function involves passenger volume. Therefore, the authors follow the same assumption as Lin (2006), Yang and Zhang (2011) and Zhang and Zhang (1997, 2003): passengers make travel decisions and concession goods purchase decisions sequentially.  2.1.2 Theories based on existence of concession price effect Czerny (2006) models monopolistic charges, Ramsey charges, single-till and dual-till regulation at a non-congested setting. He develops the two regulation methods as a two-stage game, derives the optimal solution and evaluates their welfare implications. The results show 15  that the private aeronautical charge is increased by the concession revenues relative to the case without concession. More importantly, Czerny (2006) shows single-till regulation dominates dual-till regulation with respect to social welfare. The model that Czerny (2006) built up is based on the following assumptions. First, the airport is uncongested. Second, the airport has fixed cost F≥0 but zero variable cost. Third, concessionaires and airlines are facing perfect competition. Fourth, passengers decide whether to fly based on the total surplus from aeronautical service and concession consumption. Fifth, the author assumes "consumers' reservation prices for aeronautical and commercial services are independently and uniformly distributed on a unit square". With 𝑝₁ , 𝑝₂ denoting the flight charge and concession price respectively, the consumer can get surplus:  𝑉₁ + 𝑉₂ − (𝑝₁ + 𝑝₂) (2.2) where V₁,V₂ are the consumer's willingness to pay for aeronautical service and concession goods. All the concession customers are subset of the total passengers of the airport. With this setting, it is straightforward to see that the passenger will decide to take concession service if V₂ − p₂ > 0, and will fly if V₁ − p₁ +max[0, V₂ − p₂] > 0. Therefore one will choose to fly even if his(her) willingness to pay for the flight is less than the flight ticket as long as the surplus from concession consumption can cover the loss. However, this goes against the assumption of "independent decision" in Zhang and Zhang (2003), thus again concerns the controversial issue mentioned before. It may not be the case that passengers will consider the surplus from consuming concession goods when booking flight tickets. Therefore, at least the fourth assumption may not hold. In our empirical study, the evidence is found to go against this assumption. Namely, concession price doesn't have a significant 16  effect on the passenger volume of an airport, implying people don't actually make decision in the way as the above model describes. Czerny (2013) studies theoretically the behavior of congested airport given the existence of concession revenue. This paper examines two different types of concession services. One is welfare-neutral services such as food and beverage. The demand of this type of concession goods does not depend on traveling activities. Even though we don't travel at all, food and beverage are for sure needed in daily life. The other type, such as car rental services, has a particular important feature: two-sided demand complementarity. The demand for the car-rental-type services strongly depends on traveling activities. Without traveling, people are highly unlikely to use car rentals located at airports. From another direction, a traveler may choose not to travel if the car rental price is too high such that the total cost reaches a prohibitive level. Two models are built up to compare private airports behavior and public airports behavior, where a two-stage game is built to capture the behavior of airline and airport. Firstly, airport sets charge, and secondly, airlines involve in a Cournot game. The welfare-neutral commercial services are found to make airport price regulation obsolete, while the car-rental-type services may still favor the credibility of airport price regulation. For the welfare-neutral goods, Czerny (2013) defines the following variables: τ: airport charge per passenger to the airlines; c₁: airport cost per passenger; 𝑞1𝑖: passenger quantities for each individual airline, ∑ 𝑞1𝑖 = 𝑞1𝑖  where 𝑞1 is the total passenger volume of the airport; 17  𝛷 = 𝜑ℎ − 𝜑𝑎: defined as surplus parameter, where 𝜑ℎ and 𝜑𝑎 are the surplus per consumer from concession consumption outside and inside airport respectively; κ: defined as profit parameter, which is the profit per passenger earned by airport from concession operation; Γ: average congestion costs burdened by airlines; 𝐵𝑅(𝑞1): travel benefits for passengers; 𝑝1: uniform airfare. With the above denotation, the author gives the following equations. Airport profit:  𝛱𝑅 = (𝜏 + 𝜅 − 𝑐₁)𝑞₁ (2.3) Airline profits:  𝜋𝑖𝑅 = (𝑃𝑖𝑅 − 𝜏 − Γ)𝑞1𝑖 (2.4) Consumer surplus:  𝐶𝑆𝑅 = 𝐵𝑅 − (𝑃1𝑅 + 𝛷)𝑞1 (2.5) Welfare:  𝑊𝑅 = 𝐶𝑆𝑅 + 𝛱𝑅 +∑ 𝜋𝑖𝑅𝑖+ ( 𝛷 − 𝜏)𝑞1 (2.6) Solving this two-stage game by backward induction, he finds consistent result with that of Starkie (2001, 2008), Oum et al. (2004) and Zhang and Zhang (2003, 2010) for private airports. Namely, private aeronautical charge is decreasing in commercial revenue, and private airport may charge socially optimally with the existence of welfare-neutral 18  concession services. For public (welfare-maximizing) airports, the welfare-optimal airfare can be equal, below and above social marginal cost per passenger conditional on different surplus parameter values. However, there are two problems with this welfare-neutral model. First, one of the implicit assumptions involved in the model is implied by the demand equilibrium(𝐵𝑅)′ = 𝑝1 +  𝛷, namely, people consider both airfare and concession cost when making travel decision. As the author mentions in the extensions, some economists believe that the concession services don't significantly change passenger quantities. If this is the case, the results for the public airports may not hold. Another problem is the implicit assumption that all airport passengers make concession consumption. In fact, a large proportion of passengers, including the transferring passenger volume, don't demand commercial goods. In our empirical study, we include transfer rate to capture part of this effect. For the car-rental-type services, the two-side demand complementarity between aeronautical and car rental services is the defining feature. New variables are listed below: p₂: car rental price; c₂: constant car rental cost; q₂: car rental quantities; 𝐵𝐶(𝑞1, 𝑞2): benefits for passengers; 𝐷2𝐶: car rental demand; 𝑃1𝐶(𝑞1, 𝑝2): inverse passenger demand. With the above denotation, the following equations are given by Czerny (2013): 19  Airport profits:  𝛱𝑅(𝜏, 𝑝2) = (𝜏 − 𝑐)𝑞₁ + (𝑝₂ − 𝑐₂)𝐷2𝐶(𝑞1, 𝑝2) (2.7) Airline profits:  𝜋𝑖𝐶 = (𝑃𝑖𝐶 − 𝜏 − Γ)𝑞1𝑖 (2.8) Consumer surplus:  𝐶𝑆𝐶 = 𝐵𝐶 − 𝑃1𝐶𝑞1 − 𝑝2𝐷2𝐶  (2.9) Welfare:  𝑊𝐶 = 𝐶𝑆𝐶 + 𝛱𝐶 +∑ 𝜋𝑖𝐶𝑖 (2.10) Solving by backward induction, he finds that, compared to the case without car rental services, this type of concession services can sometimes increase the private aeronautical charge but sometimes can also reduce it. This finding complements the previous findings in literature, which showed a strict negative correlation between commercial operations and private aeronautical charge. The difference comes from the assumptions on the relationship between concession services and airport passenger quantities. Starkie (2001, 2008) and Zhang and Zhang (2003, 2010) assumed that the concession price does not affect passenger volume of the airport, while Czerny (2006, 2013) assumes the other way, that is, the supply or price of concession services can change people's travel decision for their budget constraint. To know which assumption is more close to the reality and shed light on this discrepancy, some empirical evidence may be helpful, and we will provide this evidence in our study. Another problematic assumption is that car rental companies face no competition inside and outside airports. However, we observed that many car rental companies are 20  located inside the airports and offer similar rental services. Therefore it is very likely that to some extent, the car rental firms compete with each other for customers. For this possible competition effect, we include the competition level in the car rental market in our empirical model. Czerny and Lindsey (in progress) generalize the airport case to a broader concept, core-and-side services, which applies not only to airport but also to firms selling both core products and side products. In addition to airports, other examples brought up are railway stations and gas stations. They build a model to examine firms' pricing strategy under both monopolistic and competitive environment, given the demand effect and complementary effect arising from selling side products. The theoretical model includes a consumer utility function with the form  𝑈(𝑞, 𝑠) = 𝑔(𝑞) + 𝛹(𝑞, 𝑠) (2.11) where 𝑔(𝑞),𝛹(𝑞, 𝑠) corresponds to the utility from consuming the core products and side products respectively, while 𝑞, 𝑠 are quantities of core and side products consumed. As consumers buy side services only if they buy the core services, this hierarchy effect is specified by 𝛹(0, 𝑠)= 𝛹(𝑞, 0)=0. Maximizing the utility, it is found that, under monopolistic situation, the side products are priced at marginal cost and mark-up can be earned from the sale of core products. According to this result, the authors argue that the side products increases consumers' willingness to pay for the core services. The implicit assumption underlying this model is that people take into account surplus from both core and side services when making purchasing decision. However, as mentioned before, this may not be true thus the model may not be able to explain the truth. 21  In addition to the above studies, many other contributions have been made to this stream of literature. Starkie (2001, 2008) employs graphs to illustrate that private aeronautical charges decrease with concession operations. Zhang and Zhang (2010) study airport pricing and capacity investment with both commercial and aeronautical activities. They find that private (profit-maximizing) airports tend to over-invest in capacity, even with cost-based regulation. While a public (welfare-maximizing) airport can achieve socially optimal investment in capacity, it may still make inefficient investment under regulation. D'Alfonoso et al. (2013) study airport pricing with consideration of both aeronautical and commercial operations. They include a positive correlation between concession consumption and flight delay as well as the passenger types. Because of the positive externality of delay, it is found that public airports are more likely to induce congestion than private airports. Oum, Zhang and Zhang (2004) examine different price regulations and compare their effects on airport efficiency, considering the existence of concession revenue. They find that the rate-of-return regulation may lead to problem of over-investment in capacity while price-cap regulation tend to cause under-investment. Besides, the under-investment problem under single-till price-cap regulation is more severe than that under dual-till price-cap regulation. In addition, dual-till form associates with a higher total factor productivity than rate-of-return and single-till regulation.  2.1.3 Empirical studies on concession and travel activities From the perspective of travel activities’ effect on concession consumption, Geuens et al. (2004) provide evidence showing that special shopping behavior may happen in airports. They distinguished three types of airport shoppers: shopping lovers, mood shoppers and apathetic shoppers, with only the last type indifferent to shopping inside or outside airports 22  (i.e. independent of travel activities), while the other two types tend to do more shopping for the features and atmosphere in airports. Therefore, it is implied that people’s purchasing behaviors depend on travelling activities.  As Zhang and Czerny (2012) point out, there is little empirical study about whether and to what extent the concession services affect passenger volume. As we can see from the literatures discussed above, this empirical question has important implications for theoretical study and thus policy making. One existing related empirical study is Van Dender (2007) (more details to be discussed in section 2.3). The results show the concession revenues per passenger are declining in the passenger volume. This finding can be explained as that decreasing concession prices lead to higher passenger volume, thus the concession revenues per passenger decrease. Obviously, this argument is in accordance with the assumption that concession prices affect demand for flights. However, it is hard to hold because more passengers usually associate with more concession demand. If the concession demand is a fixed proportion of flights demand, increasing passenger volume does not necessarily lead to decreasing concession revenues per passenger. Therefore, the answer to this empirical question is still unclear.  Our study directly uses concession price rather than per-passenger concession revenues, to get a more precise estimation of the relationship between passenger volume and concession price.  2.2 Car rental pricing Our study also relates to the paper of Khan et al. (in progress), which examines how competition affects the price discrimination in the car rental market in airports. They focus 23  on the in-terminal car rental market for reasons including a clean market definition, a large market-structure variation, and relative ease to get information. Only passengers flying into the airport are targeted, which is the same as our study. These car-rental companies are single-product firms and service quality of a specific type (five types of car are considered) is similar across all companies, therefore the price information is easy to get and analyze. Khan et al. (in progress)'s empirical work on the price levels of car rental services has a similar setting as the first regression equation we have in our study, which has rental price as dependent variable. They collected the posted rental prices and identities of all car rental firms at the airport from Orbitz and Expedia websites. For the variables involved in the price regression equation (the first regression equation in our econometric model), including car rental prices, number of car rental firms, local poverty ratio, local population density, wages, median house value and the number of large corporations with headquarters located in the market, we basically employ Khan et al. (in progress)’s data. Their price data was collected in March 2005 for five popular car types: middle size, economy size, standard size, compact size and full size. Other specialty vehicles are excluded because those are available in only a few airports. The posted prices were collected two weeks in advance. It does not matter much that the posted prices are not necessarily equal to the transacted prices, because the effect of price changes are what we are interested in. In Khan et al. (in progress), they characterize the price-market structure as  𝑙𝑛𝑃𝑚𝑘𝑠𝑡 = 𝛼𝑘 + 𝛽𝑠 + 𝛾𝑡 + 𝑍𝑚𝜆 + 𝛿𝑁𝑚 + 𝜀𝑚𝑘𝑠𝑡 (2.12) where 𝑁𝑚 is the number of firms used as proxy for competition level, 𝑍𝑚 contains the market characteristics, and 𝛼𝑘, 𝛽𝑠, 𝛾𝑡 are fixed effects for firm, type and duration of rental. The authors used number of firms because the market shares cannot be observed. They mentioned 24  an endogeneity problem concerning the market structure because this is not exogenously decided but affected by factors such as cost and demand. Therefore they used population growth, airport traffic growth and house value as instruments to account for this endogeneity problem which is shown to be a significant problem. In our study, we collect more factors that may have impact on the car rental price, including the passenger volume arriving at the airport, the aeronautical charge, connecting passenger rate, cargo rate, international passenger rate and local GDP. Another difference is, as our main interest is in the interaction among car rental price, passenger volume and aeronautical charge, we only focus on one type of car service in each regression, thus car type and rental period are not included in our regression model. We also consider the endogeneity problem concerning the market structure, but the estimation results of our regression system show it is not significant.   2.3 Airport charge determinants The other stream of literature this study contributes to is factors affecting aeronautical charge. Although there are plenty of empirical studies on determinants of airline pricing, we can only find the following four papers studying airport pricing. Van Dender (2007) study used a panel data of 55 US airports from 1998 to 2002. He did not employ a panel-data econometric analysis because of a lack of variation in some important variables, so he conducted a cross-section study with six simultaneous equations, using three stage least square regression. The dependent variables of these six equations are airport fare, aeronautical charge, concession revenue per passenger, passenger volume, departures and delays. According to the regression results, he found that the effect of departures on 25  aeronautical charge is negative. With potential competition coming from airports nearby, the charges are lower. A higher aeronautical charge is observed for airports with higher airline concentration level, except for airports dominated by Southwest Airlines, which is a low-cost carrier. A reason given in Van Dender (2007) is that the airlines transfer their profits, which result from market power, to airports. The results also show positive effects of average distance and the share of international departures on aeronautical charge, which is due to the weight-based charging principle. From the equation with concession revenue as dependent variable, it is found that the concession revenues per passenger is negatively correlated with the passenger volume. This effect is in favor of the notion that lower concession price leads to higher passenger volume. Bel and Fageda (2010)'s main interest is in the impact of privatization and regulation on charges of European airports. They employed a cross-section dataset of 100 large airports in Europe, using two stage least square regression to deal with the endogeneity problem coming from the demand function. Bel and Fageda's results imply that the private and non-regulated airports charge higher than regulated or public airports. They also found that airport pricing would not be affected by the regulation mechanism (e.g. ROR, price-cap). Their findings indicate that airports charge higher if the passenger volume is higher, which they explained with a greater monopoly rents or higher costs. As to the empirical findings on airport charge, Bel and Fageda found that the potential competition from nearby airports, airline concentration, existence of low-cost carriers and share of domestic passengers all had a negative effect on aeronautical charges. Bilotkach et al. (2012) have similar interests with Bel and Fageda (2010): the effect of regulation and privatization on airport charges, and the factors affecting airport charge. They 26  employed a panel-data of 61 airports in Europe spanning from 1990 to 2007, which distinguishes their study from the previous two papers. Van Dender (2007) didn't analyze the regulation issue since airports in US are publicly owned and operated based on cost. However, European aviation market endows Bel and Fageda (2010) and Bilotkach et al. (2012) the context to empirically examine the effect of regulation and privatization, since airports in Europe adopt various operational policies. To estimate the regression model, Bilotkach et al. fully employed panel-data econometric technique by using Generalized Method of Moment (GMM). Their empirical findings indicate a lower aeronautical charge under single-till regulation. On average, both privatization and ex-post regulation lead to lower aeronautical charges, while price-cap regulation does not have significant effect on aeronautical charges. For other non-policy factors, airports acting as hubs tend to charge higher, while the potential competition of nearby airports does not seem to have significant impact on aeronautical charges. Comparing the results of Bel and Fageda (2010) and Bilotkach et al. (2012), we can see the contradiction in the impact of privatization on airport charges. While a positive effect is found in the former, a negative one is observed in the latter. Also, Bel and Fageda did not get any significant effects of different regulation mechanisms on airport pricing, while Bilotkach et al. showed a lower charge under single-till regulation and non-significant effect of price-cap regulation. Another difference is on the effect of competition from nearby airports: negative effect in the former and nonsignificance in the latter. As Bilotkach et al. explained, the reason of the contradiction may come from the different data structure and methodologies used in these two studies. 27  Choo (2014) focused on the factors affecting aeronautical charges. He analyzed a panel-data of 59 airports in United States spanning from 2002 to 2010, using different estimation methods including panel-data random effect, Hausman Taylor, fixed effect and EC2SLS. The empirical findings reveal a cross-subsidization effect from concession operations to aeronautical operations. Choo (2014) also found a significant substitution effect for the landing fees and terminal fees charged by airports. That is, a higher terminal fee is set with a lower landing fee and vice versa. Hub airports charge higher fees, which is consistent with Bilotkach et al. (2012). A higher international traffic also significantly leads to a higher charge. This is similar to the findings of Van Dender (2007) and Bel and Fageda (2010). Two possible reasons given by the author are that the cost to serve international flights is higher and international traffic is more competitive in terms of transport modes than others (e.g. rail, road). Dominance of low-cost carriers don't significantly lower aeronautical charge, but Southwest-dominated airports are found to charge lower.   28  3 Data, Econometric Model and Hypotheses 3.1 Data  Our data was collected from the following publicly available sources. The first major piece of data includes passenger volume, connecting passengers, distance of flights, number of flights, cargo volume, international passengers and airline involved. These are all collected from the database of Bureau of Transportation Statistics, specifically T-100 Market (All Carriers) and T-100 Segment (All Carriers). Only the information for year 2005 was collected for accordance with the availability of revenue and rental price data. We only employ the arriving passenger volume for airports as destination. One reason is the symmetry of quantities for passengers arriving and departing an airport. The other reason is that passengers arriving at an airport are the potential consumers of the car rental services provided in the airport. The total arriving passenger volume is available from T-100 Segment (All Carriers), while the terminal passenger volume is recorded in T-100 Market (All Carriers). The second piece of data includes the airports’ aeronautical revenues, concession revenues and operating costs. I collected these revenue and cost from the City-data website. According to the website, they correspond to the rows “aeronautical operating revenue”, “non-aeronautical operating revenue” and “operating expenses”. The most recent available data is year 2007 for some airports and year 2008 for the others.  What we need in this study is the information in year 2005.  Our last major source of data is provided by Khan et al. (in progress). They collected the posted car rental prices and identities of all car rental firms at the airport from Orbitz and Expedia websites. The price data was collected in March 2005 for five popular car types: 29  middle size, economy size, standard size, compact size and full size. Other specialty vehicles are excluded because those are available in only a few airports. The posted prices were collected two weeks in advance. They also obtained the demand and cost factors for rental firms from the database of Census, Bureau of Labor Statistics(BLS) and Compustat. The information collected by Khan et al. (in progress) and included in our first price regression equation are: car rental prices, number of car rental firms, local poverty ratio, local population density, wages, median house value, the number of large corporations with headquarters located in the market, and the percentage of population in the market that uses public transportation to get to work. Apart from the above, we also collected the county-level GDP for airports located in different county from the website of Bureau of Economic Analysis (BEA).  3.2 Econometric Model Since the three variables that we are interested in are airports' aeronautical charge, car rental price, and airport passenger volume, which decide each other simultaneously, the estimated model is a system of simultaneous equations with these three endogenous variables. We have them summarized in Figure 2. Figure 2    Model Specification (1) laavg_price (lpsg05   lcharge   competition   transfer_pct   cargo_r   f_rate   lgdp   lpopden  logpop   povertyratio   wage   num_hq   lhouse   pubratio)  (2) lpsg05 (laavg_price   lcharge   lgdp   lpopden   logpop   povertyratio   holiday   hub   num_p  num_hq   num_line   lav_dist   wage)   (3) lcharge (lpsg05   laavg_price   lfrt05   lc_rev   transfer_pct   hhi   num_line   num_p   hub   f_rate  lav_dist  cost   s_w) NOTE: bold indicates endogeneity 30  The three main variables that we are interested in are airport aeronautical charge, traffic volume and car rental service price, whose histograms are shown in Figure 3.  Figure 3    Histogram of Three Dependent Variables   31  Each variable involved in the regression equations above is briefly summarized in table 1. Table 1    Descriptive Statistics   Variable Brief explanation Summary Endogenous variables Obs. Mean Std.Dev.    Min Max laavg_price log average price across all car types and rental firms  337      3.996978     .1987396       3.329821      4.55598 lpsg_05 log arriving passengers in 2005  337      11.96008      2.500058      6.807935      17.56281 lcharge_2 log average aeronautical revenue per passenger  337      2.572683      1.074599      0.8284116    6.966433  lcharge log average aeronautical revenue per output(in terms of weights of total passengers and cargo)  337      -2.852631    1.097725       -4.577419    1.668115 Exogenous variables      transfer_pct the percentage: transferring passengers out of total passengers   337       0.0510325        0.1076124           0            0.6682959 cargo_r the percentage: total cargo weight over total weight of passengers and cargo  337       0.1002747      0.147196            0            0.8384494 f_rate the percentage: foreign passengers over all passengers  337       0.0118261      0.0414081          0            0.4582302 competition the number of  rental firms at that airport  337      5.323442       3.0559               1            17 lgdp log county-level GDP  337      10.40101      0 .20455       9.820269     11.52268 lpopden log local population density  328      5.15759         1.694706     0.9707085    9.960421 logpop log population  328      12.0922        1.314694      8.794067     16.06884 povertyratio local poverty ratio  337      0.1257378    0.0432336    0.027341     0.3260119 holiday dummy for the airport to be located at a vacation destination  337      0.1068249    0.3093498          0                1 wage local average retail wage  328      16.29048      1.994856      12.33397     28.82603 lhouse log median house value  328      11.16303      0.4208773    10.51867     12.74811 num_hq number of large corporations that have their headquarters located in the market  337       7.830861      21.0137              0               136 pop_change population growth rate from 1990-2005  328      0.1390911   0.1502647  -0.1495123    0.9691277 grthrate passenger volume  growth rate from 2002-2004  337      1.13495        1.930378    -0.9481379     16.00517 hub dummy for hub airport  337      0.1038576    0.3055293           0                  1 num_p number of airports located in the same county  337      1.148368       0.501728            1                  5 num_line number of airlines flying into that airport  337      866.3145       1715.301            8                11234 lav_dist log average distance of flight arriving at this airport  337      6.284859       0.7174405    2.833213      7.976054 lfrt_05 log total freight in 2005  316       13.87108         4.145457                0               21.96237 hhi HHI for airlines in this airport  337      0.4599778     0.3114021    0.0520644          1 cost unit cost proxy for airport cost  337      0.2822182     0.7250399    0.0090009     7.224199 s_w dummy for Southwest Airlines dominated airport  337      0.0652819     0.2473899          0                     1 pubratio percentage of population in the market that uses public transportation to get to work 337        0.023705      0.0504353          0         0.4735137 32  (1) We start from the first equation. Rental price for each car type of different rental companies can be gotten from Orbitz and Expedia web sites. Khan et al. (in progress) retrieved this piece of information and we directly use it. For the first regression equation, we use the average one-day rental price of mid-size car across all companies located inside the airport terminal as dependent variable, and we will check different car types in the robustness check section. This is a reduced form equation which captures the equilibrium outcome of the car rental market at the airport. The explanatory variables capture three streams of factors: demand, supply and cost. First, the demand of rental service is characterized by airport passenger volume, transfer rate, cargo rate, international passenger rate, local GDP, population density, population, poverty ratio, number of large corporations that have headquarters located in the market, and  percentage of population in the market that uses public transportation to get to work.  The first four variables capture the number of passengers who arrive at the airport and potentially need to rent a car. The last factor addresses the outside option faced by the passengers, which is to take public transit. We expect that the more convenient the public transit is, the more competition there is thus the lower price rental companies will charge. The rest capture the level of prosperity of the destination, which may affect passengers’ decision on whether a car is necessary for their travel. Usually, the more prosperous a place is, the larger its size will be, and the better road condition there will be. For the supply side, as we cannot observe the exact market, the number of rental companies located in the airport is a proxy used to capture the market structure, namely the level of competition. The costs faced by car rental firms, which also affect the pricing, are explained by retail wage and 33  median house value. Airport's aeronautical charge may also affect concession price, because the U.S. airports employ a cost-based regulation which consider both concession and aeronautical revenues, thus may put pressure on concession operation if they cannot get enough aeronautical revenue to balance the costs.  Airport's aeronautical revenue and concession revenue are collected from City-data website, where the former includes landing fees, terminal fees, apron charges, tiedowns, security reimbursement and so on. Airport's aeronautical charge is measured from the total aeronautical revenue divided by passenger volume. We will also try another measure in the section of robustness check. Passenger volume is the number of passengers that arrive at the airport. As there is approximately a symmetry between the passenger volume arriving and those departing from an airport, we don't include the departure side. We can use "T-100 market" dataset to get this information. The number of transferring passengers can be gotten by subtracting the above passenger volume from "T-100 segment" volume which is the total number of passengers arriving at this airport, both as final destination and transfer stop. The transfer rate, namely the percentage of transferring passengers out of total passengers, is constructed from the transferring passenger volume divided by the total number. In our dataset, the highest transfer rate is 66.83% and the lowest is zero. As shown below, figure 4 is the histogram and boxplot of transfer rates across all airports. 34  Figure 4    Histogram and Boxplot of Transfer Rates   Cargo rate measures the percentage of total cargo weight over total weight of both passengers and cargo. With a high cargo rate, an airport is more likely to concentrated on freight transit, thus the airport structure and policy on concession operation may be different from those with low cargo rate. In this study, we count 200 pounds for each passenger to calculate the cargo rate. If the cargo rate is too high, much noise will be diluting the effect of concession operations on both passenger and airport behaviors which we can observe. Therefore we drop airports with cargo rates more than 90%. Figure 5 shows the distribution of cargo rate across all airports. Among our sample, more than ninety percent of airports have cargo rates less than 30% and the mean value is around 10%. Figure 5    Cargo Rate in Airports   35  The market structure of car rental companies in each airport is conventionally captured by Herfindahl Index or market concentration ratio. However the market shares cannot be observed, so the number of rental firms is employed, which can be got from Orbitz and Expedia web sites. The firms include both national and local car rental firms. We can see from the distribution of firm numbers shown in Figure 6, the market structures are well diversified in the airports we observe. Figure 6    Market Structure of Airport Car Rental  County-level GDP can be got from Bureau of Economic Analysis. Population density and population was retrieved from Census. All others can be got from Bureau of Labor Statistics (BLS) and COMPUSTAT.  (2) In the second equation, factors affecting passenger volume come from the following perspectives. Firstly, local GDP, population density, population, poverty ratio, holiday dummy, number of large corporations, retail wage capture how attractive the destination is 36  and thus affect the passenger volume arriving at the airport. For example, holiday dummy, with value of one if the airport is located at a holiday destination, usually endows the destination more appealing and thus attracts many travellers from far away. Secondly, we use number of airports located in the same county as proxy of competition faced by each airport. Thirdly, number of airlines flying into the airport, hub dummy and average distance of flights arriving at the airport capture the throughput capacity of the airport, a more direct measure of the passenger volume. Fourthly, the aeronautical charge captures the ticket price, which is a big concern for passengers to decide fly or not. Lastly but most importantly, the effect from rental price is what we most care about. We want to see if people take concession price into consideration when making travel decisions.  Hub dummy has value of one if the airport is a hub and otherwise zero. Average distance of the arriving flights is a proxy for weight of airplane, because flights flying from far away are more likely to have larger aircrafts and more fuel. All these data can be collected from the "T-100 market" dataset.  (3) The explanatory variables involved in the third equation includes the following aspects. The demand for air transport services are captured by passenger volume, freight volume, and number of flights arriving at the airport. These variables capture the costs faced by airports: transfer rate, airport operating cost, average distance of arriving flights, international passenger rate and dummy for Southwest Airline’s dominance. Airport market structure can be explained by Herfindahl Index (HHI) for airlines at the airport, number of airports located within the same county, and international passenger rate. Concession revenue affects 37  aeronautical charge because there may be a cross-subsidization from concession side to aeronautical side.  We include freight because it is also charged by airports. HHI is calculated according to the total weight of passengers and freights that each airline carries, which can be used to capture concentration of airlines at each airport thus the bargaining power of airports. This information is also derived from the T-100 market dataset. The percentage of foreign passengers over all passengers (international traffic rate). We can see from figure 7 that there is a wide cross-section of HHI for airlines at airports, corresponding to different airline market structures. Figure 7    Distribution of Herfindahl Index for Airlines   The dummy for Southwest Airlines dominated airport has value of one if more than thirty percent of the airport's output is operated by Southwest Airlines, as fares and aeronautical revenue are lower for airports dominated by Southwest. As shown below, Figure 8 is the histogram and boxplot of the ratio of Southwest-airline output over airport's total output. 38  Figure 8    Histogram and Boxplot of Southwest-airline Share   The international traffic rate is calculated according to whether the origin airports are located in other countries. For those flying from abroad, we sum up as international traffic volume. More international flights may associate with a large airplane size, higher airport operational cost and higher airport bargaining power, thus have effect on the aeronautical charge. In our sample, we have the following distribution of international traffic rate, showing a high concentration within 0 to 2.5%, which is in accordance with the high proportion of domestic traffic in US air transport market. Figure 9    International Traffic Rate  39  We have concession revenue as one potential determinant of airport aeronautical charge, but not another simultaneous equation with concession revenue as dependent variable and charge as independent variable. The reason comes from how the airport charge is decided in practice. The determination consists of two sequential steps (Zhang and Zhang, 1997). In order to maximize concession revenue, the airport authorities first set rates for the concession operations, after which aeronautical charges are set to balance revenue shortfall if there is any. Therefore, concession revenue affects aeronautical charge but not the other way around. To satisfy the model identification requirements, we need at least one shifter for each equation. In other words, there must be at least one factor in each equation that does not affect the other two dependent variables. For the first equation, the competition in car rental market affect the rental price, but it is hard to think of any direct or reasonable effect from competition faced by rental firms on the passenger volume and charge on the airport side. In the second equation, the holiday dummy makes the destination appealing thus more travellers are attracted, but has no any direct effect on airport charge. Also, only through its effect on passenger volume, the rental price may be affected. Namely there is no direct effect from holiday dummy on rental price. One factor among the exclusive explanatory variables of aeronautical charge is airport operating cost. Especially for the U.S. airports which adopt cost-based charging tradition, costs directly offer airports a charging basis. On the other hand, airport operating costs do not have any direct effect on passenger volume and rental price from any reasonable perspective.  Therefore, each of the three equations has its own economic meaning, and as a whole are identified as a simultaneous system.  40  For all the airports included in T-100 market dataset, we only keep those with complete information for the variables explained above. Some airports don't have revenue information available, while some don't have rental service or operating expense data available. For these airports, there is no sense that can be made out of them, thus we can just drop them. Airports included are listed in appendix A. 3.3 Hypotheses Our interest lies in the interaction among airport aeronautical charge, traffic volume and car rental service price, as well as the determinants of aeronautical charge. Also, we are curious to see some other factors that may affect these three dependent variables. With simple demand-and-price relationship, we expect that the rental price will be increasing in passenger volume, since a higher passenger volume brings higher demand for concessional goods. We don't have clear clue about the relationship between car rental price and aeronautical charge. A substitution effect may be possible. Given a fixed budget, passengers will consume less concession goods given a high aviation expenditure, thus put downward pressure on concession price. As for the effect of concession price on passenger volume, there are two hypotheses. One is that either price of concessional goods is not significant enough to affect people's travel decision, or people are not aware of prices when making travel decision, so that prices have no effect on passenger quantity of airport. The other hypothesis is that concession price may affect passenger volume since many frequent passengers are fully aware of the surplus associated with concession goods. Therefore they will take concession cost into account when making travel decision. Even for some leisure passengers, they may also make a 41  detailed plan involving consumption of concession goods based on their budget. Aeronautical charge is expected to negatively affect the passenger volume, again, for the reason of demand-and-price trend. All the other factors in the second equation are expected to affect passenger volume positively except number of airports located in the same county, number of headquarters of big companies and poverty ratio. For the last equation with aeronautical charge per passenger as dependent variable, we expect passenger volume to negatively affect the unit aeronautical charge. The reason is that the total charge is conventionally weight based, thus it contains a big fixed component for each flight. The weight-based argument also applies to our hypothesis that the average distance of flights arriving at the airport will have positive impact on charge. The higher the airport cost is, the higher the airport charge will be. This is in fact the case for airports in US, where airports are publicly owned, thus charge airlines to cover operating cost, but not for profit. Charges may increase with airline concentration which we capture by HHI, probably because of a strong vertical relationship between airports and airlines. Another possible reason is that the oligopolistic airline market lead to high aeronautical charges. The Southwest Airlines dominated airports are expected to have lower aeronautical charge. For the percentage of foreign passengers, we don't have clear expectation. Given the cost has been controlled, the possible effect comes from the fact that airports with higher international flights may have larger bargaining power thus charge airlines higher fees. However, airports in US are cost-based and not operating solely for profits, so they may not leverage this bargaining power to earn more money.   42  4 Methodology For convenience and simplicity of methodology clarification, we write the econometric model shown in Figure.1 as below.  y₁ =  α + y₂α₀ + y₃α₁ + X₁α₂ + ε₁   y₂ =  β + y₃β₀ + y₁β₁ + X₂β₂ + ε₂ (4.1)  y₃ =  γ + y₁γ₀ + y₂γ₁ + X₃γ₂ + ε₃  where y₁, y₂, y₃ are the column vector of observations on each of the three jointly dependent variables. X₁, X₂, X₃ are the other covariates for each equation. α₀, α₁, α₂, β₀, β₁, β₂, γ₀, γ₁, γ₂ are the corresponding coefficients. ε₁, ε₂, ε₃ are the column vector of disturbances. These three equations correspond respectively to the regression equations with rental price, passenger volume and aeronautical charge as dependent variables. 4.1 Endogeneity As described above, we have a system of three simultaneous equations. Each equation is called a structural equation because each has a ceteris paribus, behavioral, causal interpretation.  Simultaneity is known to be one of the three sources of endogeneity, with the other two being measurement error and omitted variables. Generally, covariates that are simultaneously determined with the dependent variables are correlated with the error term. Therefore if we used OLS for each regression equation, the estimators would be biased and inconsistent. Now we explain how the endogeneity comes into being by using a simple two-equation simultaneous model as follows. 43   𝑋𝑖 = 𝛼 + 𝛼₀𝑌𝑖 + 𝑣𝑖 (4.2)  𝑌𝑖 = 𝛽 + 𝛽₀𝑋𝑖 + 𝑢𝑖 Solving the two equations, we have the reduced form of the model  𝑋𝑖 =𝛼 + 𝛼₀𝛽1 − 𝛼₀𝛽₀+𝑣𝑖 + 𝛼₀𝑢𝑖1 − 𝛼₀𝛽₀ (4.3)  𝑌𝑖 =𝛽 + 𝛽₀𝛼1 − 𝛼₀𝛽₀+𝛽₀𝑣𝑖 + 𝑢𝑖1 − 𝛼₀𝛽₀ It is obvious to see that 𝑋𝑖 is correlated with 𝑢𝑖, and 𝑌𝑖 is correlated with 𝑣𝑖  for  𝐶𝑜𝑣(𝑋𝑖, 𝑢𝑖) = 𝐶𝑜𝑣 (𝑣𝑖 + 𝛼₀𝑢𝑖1 − 𝛼₀𝛽₀, 𝑢𝑖) (4.4)  =𝛼₀1 − 𝛼₀𝛽₀𝑉𝑎𝑟(𝑢𝑖)  𝐶𝑜𝑣(𝑌𝑖, 𝑣𝑖) = 𝐶𝑜𝑣 (𝛽₀𝑣𝑖 + 𝑢𝑖1 − 𝛼₀𝛽₀, 𝑣𝑖) (4.5)  =𝛽₀1 − 𝛼₀𝛽₀𝑉𝑎𝑟(𝑣𝑖) So the OLS estimator is biased and inconsistent for this two-equation model. Similarly, for our three-equation simultaneous model, if we perform the regression equation by equation, we will get a biased and inconsistent OLS estimator. Another possible endogeneity problem comes from the observed car-rental market structure. As mentioned in Khan et al. (in progress), there may be unobserved cost and demand shocks that affect car-rental firms' entry decision. Namely, the market structure, or the proxy variable we adopt--number of car-rental firms in the airport--is not exogenously given but endogenous. Using the same instrumental variables (population growth rate, traffic growth rate, house value) used in Khan et al., we employ the Durbin-Wu-Hausman tests to test if this 44  endogeneity problem is significant. The test results suggest that this endogeneity issue does not exist in our context. Given this test result, using this instrumental variable approach would be less efficient than the case without using it. Therefore, there is no need to use these instrumental variables to control the "omitted variable" problem, because it is actually undesirable for its inefficiency. What we need to do next is to deal with the endogeneity caused by simultaneity. 4.2 Three stage least square (3SLS) estimation If we have enough instruments, one method that we are familiar with is to apply two-stage least square method (2SLS) to estimate each equation of system (4.1). As mentioned in Wooldridge (2002), the reason why we abandon single-equation analysis but seek system methods is to pursue higher efficiency. By estimating α₀, α₁, α₂, β₀, β₁, β₂, γ₀, γ₁, γ₂ jointly (i.e. by system procedure), we can usually get more efficient estimators. Acknowledging this, we first need to check if our econometric model is identified. In other words, we need to figure out whether the structural parameters are estimable. In order to have identification for the model, for each equation (say equation-1), we must have at least one exogenous variable(shifter) in each of the other two equations(say equation-2, equation-3), and these exogenous variables must not appear in the covariates of equation-1. That is to say, there must be some factors that exogenously shift equation-2 and equation-3 but not equation-1. Looking into our regression model, it is easy to see that all three equations can be identified. We can find variables such as "average distance of flight arriving at the airport", "airport operating cost" not included in X₁ of the first equation but appear as exogenous variables in X₂ and X3. Also, variables like "number of car rental firms" and "airport operating cost" are included as pre-given in X₁ and X3, while they are not in X₂. For the third equation with 45  aeronautical charge as dependent variable, "number of car rental firms" and "vacation destination" are not included but appear in the covariates X₁, X₂. After checking out the identification issue, then comes the question on what specific methodology to use to account for the endogeneity problem contained in the three simultaneous equations. As we know, generalized method of moments (GMM) can be generally applied to multiple-equation system. GMM estimator, with a symmetric and positive semi-definite matrix, is consistent and asymptotically normal under certain conditions (Wooldridge, 2002). With homoscedastic errors, a particular weighting matrix can be used to get a special-case GMM estimator: three stage least square estimator (3SLS), which is also consistent and asymptotically normal. From another perspective as mentioned in Hayashi (2000), if the instruments are the same for every equation, a simplified formula can be achieved: the 3SLS estimator. In our model, we can use all the exogenous variables involved in the system as the same instruments for each equation. There are three reasons for our usage of 3SLS to estimate the three-equation model. First, there is a long history of using 3SLS in simultaneous equations model, much earlier than the introduction of GMM approach. Second, under homoscedasticity assumption for the system, 3SLS may have a better finite sample behavior than the GMM estimator. Third, as we mentioned in the literature review, Van Dender (2007) used 3SLS to deal with simultaneous equations model and had quite well performed estimates. In the next section, we will analyze the 3SLS estimation results produced by Stata commands.   46  5 Estimation Results and Discussion We have the 3SLS and OLS estimation results listed in table 2. The results of these two methods are obviously different, indicating there is a big endogeneity problem. Therefore we only analyze the estimation results of 3SLS. In correspondence with the three streams of literatures in section 2, we analyze the results from the following three perspectives. Table 2    3SLS and OLS Estimation Results Equation 3SLS OLS (1) Average car rental price Coef. Coef.    lpsg_05 0.007 0.051***  (0.38) (3.74) lcharge -0.005 0.021  (-0.14) (1.52) transfer_pct -0.453 -0.677  (-0.64) (-0.91) cargo_r 0.487 0.243  (1.21) (0.55) f_rate 0.675** 0.411  (1.96) (1.21) competition -0.027*** -0.035***  (-3.26) (-4.25) lpopden 0.021* 0.005  (1.74) (0.41) logpop 0.042** 0.005  (2.26) (0.25) povertyratio -0.146 -0.201  (-0.52) (-0.68) wage 0.003 -0.005  (0.46) (-0.65) num_hq -0.001 0.000  (-1.05) (0.23) lgdp 0.171** 0.094  (2.01) (1.08) lhouse -0.034 -0.048  (-0.96) (-1.23) pubratio 0.222 0.300  (0.90) (1.04)       47  Equation 3SLS OLS (2) Passenger volume Coef. Coef. laavg_price 0.686 0.298  (0.60) (1.18) lcharge -1.642*** -0.831***  (-15.63) (-17.91) lgdp 2.041*** 1.504***  (4.86) (4.07) lpopden 0.168*** 0.171***  (3.18) (3.14) logpop 0.274*** 0.481***  (3.43) (5.87) povertyratio -0.733 -1.045  (-0.57) (-0.77) holiday 0.186 0.332*  (1.13) (1.71) hub 0.741** 0.357  (2.27) (1.29) num_p 0.026 0.022  (0.15) (0.15) num_hq -0.006* -0.011***  (-1.70) (-3.15) num_line 0.0002*** 0.000***  (3.73) (6.58) lav_dist 0.550*** 0.751***  (4.64) (9.15) wage 0.060** 0.108***  (2.00) (3.42)    (3) Aeronautical charge Coef. Coef. lpsg_05 -0.404*** -0.530***  (-5.15) (-8.05) laavg_price 1.664** 0.421**  (2.08) (2.04) lc_rev 0.042 -0.040**  (0.96) (-2.23) lfrt_05 -0.029* -0.518  (-1.93) (-1.36) transfer_pct -0.482 0.243***  (-1.58) (4.75) hhi 0.065 -0.564**  (0.37) (-2.50) num_line 0.0001** 0.000*  (2.16) (1.91) num_p 0.042 -0.051  (0.39) (-0.61) 48  Equation 3SLS OLS (3) Aeronautical charge Coef. Coef. hub 0.326 0.261  (1.48) (1.21) f_rate 0.251 3.005**  (0.22) (2.25) lav_dist 0.209** -0.052  (2.50) (-0.61) cost 0.153** 0.406***  (2.12) (4.52) s_w -0.036 0.139  (-0.26) (0.77) lgdp 0.942*** 0.785***  (3.81) (3.51)    1. t-statistics in parentheses 2. Significance: *** 0.1%, ** 1%, *5% 3. The three equations correspond to the three-equation simultaneous regression model, with dependent variables the natural logarithm of rental price, natural logarithm of passenger volume of the airport and natural logarithm of aeronautical charge.  4. All independent variables listed in the first column are explained in table 1.    5.1 Car rental pricing Firstly, let's check the first regression equation. The effect of passenger volume on car rental price is not significant, which is possible since more passengers don’t necessarily mean more demand for car rental services. Namely, the concession is not simply a fixed proportion of the aeronautical demand. Another possible reason is that the car-rental market at the airport is fully competitive, therefore all rental firms are charging cost-based price. Or it may be that airports have transferring passengers. So the rental companies may not tell whether their service's demand will increase based on the mere fact that there are more passengers arriving at the airport. We will discuss this transfer rate issue in the robustness check section. As expected, concession price is not significantly affected by aeronautical charge, empirically implying that the pricing decisions for concession are made first, independent of aeronautical charging decision. 49  Some other results that worth mentioning are the positive effect from percentage of international passengers and the negative effect from the number of rental companies. One possible reason is that international passengers are less familiar with the local public transit system, thus rely more on car rental services. Also, international travellers are less likely to have friends or relatives to pick them up. Therefore they have less bargaining power since there is less outside options for them. The number of rental companies is a proxy for level of competition in the market. From table 2, we see that the higher the competition level is, the less each company can charge consumers. Not surprisingly, rental prices are significantly higher with higher local GDP, population and population density. It is common to see services are more expensive in a more prosperous city economically or culturally. 5.2 Concession price effect For the second equation, it is interesting to see a nonsignificant effect of car rental price on passenger volume. This provides the first empirical evidence to answer whether concession price will affect passenger volume of airports. From this result, we can see concession price is not big enough to affect passengers' travel decision, or they are not aware of the surplus of concession service when making the travel decision. As introduced in the literature review section, this piece of our empirical results adds credits to the theoretical works of Starkie (2001, 2008), Zhang and Zhang (1997, 2003, 2010), Lin(2006), Yang and Zhang (2011) and Gillen and Mantin (2014), which assumed that people's decision of travel and consumption of concessional goods are made independently and separately in time. For Czerny (2006, 2013) and Czerny and Lindsey (in progress), they assume that passengers will decide to travel as long as the total surplus from both flying and concession consumption is positive. 50  That is, they implicitly assume that there exists effect from concession price on passenger volume, which is not consistent with our empirical finding. It is in accordance with our expectation that aeronautical charge negatively affects the passenger volume--the more expensive the trip is, the less people will travel. Local GDP, population and population density positively associate with passenger volume, which quite goes along with our observation that cultural, economic or political centers are themselves places of attraction thus have quite busy airports. Hub airports are observed to have more traffic, which is why these airports are called hub, the collection and distribution center of traffic. As expected and observed, given all other factors fixed, more flights flying into the airports means more passengers. It is also within our expectation that if the average distance of flights arriving at the airport is longer, the airplanes are probably of larger size, thus carrying more passengers. 5.3 Aeronautical charge determinants For the last equation with aeronautical charge as dependent variable, we see a negative effect of passenger volume on airport charge, which is anticipated. With a big fixed component for each flight, more passengers will lead to a lower charge per person. The rental price here has no effect on charge. The reason why car rental price has no significant effect on aeronautical charge may be that aviation side does not have direct connection with concession side financially, or only care the total concession revenue it can get for covering cost. However, airports don't care what pricing strategies the concessionaires use. This explanation is quite reasonable since car rental price doesn't affect the passenger volume of airports, thus airports have no reason to interfere with the car rental pricing. From our results, concession revenue doesn't have cross-subsidization effect on the aeronautical revenue, at least revenue per 51  passenger. This result is in accordance with that of Bilotkach et al. (2012) but against Choo (2014). The transfer rate is not correlated with charge, which is intuitive. Usually, the greater HHI is, the more bargaining power the airlines will have, thus the airport obtain less market power. But for cost-based charge, there's no more can be bargained down, so the effect of HHI is absent. With all other factors controlled, number of flights arriving at the airport has positive impact on charge, which is also consistent with the weight-based charge tradition. As expected, average distance of flights is positively correlated with aeronautical charge, because airplanes flying from far away are more likely to be larger, thus heavier fixed component. The operation cost positively affect charge, which is the consequence of cost-base charging policy. We can see that local GDP also has a positive estimated coefficient. This is ubiquitous for goods to be more expensive in a richer city. Those variables that don't have significant effect include number of airports in the same county, hub dummy and percentage of international passengers, are due to the cost-base charging conduct and airports are operating as not-for-profit entities and don't have the incentive to earn higher markup. Southwest dominated airports don't significantly have lower charge, but when we increase the threshold for an airport to be Southwest-dominant (such as fifty percent rather than thirty percent), we find a negative effect. So the low-cost effect depends on how we define "dominance" in this context. Some effects such as that of hub and passenger volume don't agree with all previous studies, one reason is the different market (i.e. US airport market rather than European market) we are considering, the other reason is we are not using total charge but charge per passenger.   52  6 Robustness Check There are different measures of aeronautical charge in the literature. Previously, we use charge per output as the measure for aeronautical charge. Output is calculated as the total weights of passengers and cargo, where we count 200 pounds for each passenger. Here we also tried another measure, that is, charge per passenger. The 3SLS results using this measure can be found in Table 3, which is consistent with our results before. Table 3    3SLS Estimation Results (use measure of "charge per passenger") Equation (1) Average car rental price Equation (2) Passenger volume Equation (3) Aeronautical charge Variables Coef. Variables Coef. Variables Coef. lpsg_05 -0.004 laavg_price -0.090 lpsg_05 -0.484***  (-0.20)  (-0.08)  (-6.10) lcharge_2 -0.024 lcharge_2 -1.653*** laavg_price 1.200  (-0.66)  (-16.25)  (1.56) transfer_pct -0.598 lgdp 1.985*** lc_rev 0.069  (-0.90)  (4.64)  (1.61) cargo_r 0.286 lpopden 0.156*** lfrt_05 -0.010  (0.73)  (2.87)  (-0.73) f_rate 0.775** logpop 0.286*** Transfer_pct -0.511*  (2.35)  (3.50)  (-1.73) competition -0.026*** povertyratio -1.067 hhi -0.026  (-3.19)  (-0.81)  (-0.15) lpopden 0.022* holiday 0.101 num_line 0.0002***  (1.75)  (0.64)  (4.22) logpop 0.047** hub 0.177 num_p 0.007  (2.47)  (0.55)  (0.07) povertyratio -0.184 num_p -0.013 hub 0.017  (-0.64)  (-0.08)  (0.08) wage 0.004 num_hq -0.006* f_rate -0.028  (0.53)  (-1.95)  (-0.03) num_hq -0.001 num_line 0.0004*** lav_dist 0.273***  (-1.12)  (6.08)  (3.29) lgdp 0.205** lav_dist 0.626*** cost 0.126*  (2.40)  (5.42)  (1.87) lhouse -0.045 wage 0.057* s_w -0.007  (-1.33)  (1.83)  (-0.05) pubratio 0.227   lgdp 0.923***  (0.95)    (3.86) 53  1. t-statistics in parentheses 2. Significance: *** 0.1%, ** 1%, *5% 3. Variables are defined the same as those in table 2.   Since those transferred passengers don't use car rental service, the results may be different for airports with different level of transfer rate. We can see from the summary statistics that the maximum rate is 66.83% and minimum is zero. Therefore we group airports into several groups according to the percentile of transfer rate. We tried respectively five groups, four groups, and three groups. The results for five groups are listed is Table 4 and four groups, and the results for three groups are listed in Table 5. We have interesting findings from Table 4 and Table 5. Airports with the lowest transfer rate out of five groups and four groups have a significant positive effect from passenger volume on car rental price. This effect becomes nonsignificant in the three-group case. This phenomenon implies that car rental price is adjusted according to the passenger volume only when transfer rate is low enough. Because in those airports, the signal of increasing demand is clearer. In addition to this change, all other results basically agree with our previous findings. Table 4    Group by Percentile of Transferring Passenger Rate                  Percentile Equation     0~20% 20%~40% 40%~60% 60%~80% 80%~100% Average car rental price Coef. Coef. Coef. Coef. Coef. lpsg_05 0.155** -0.089 -0.089 0.113 0.009  (3.03) (-1.65) (-1.65) (1.96) (0.19) lcharge_2 0.107* -0.039 -0.039 -0.121 0.021  (2.36) (-1.06) (-1.06) (-1.66) (0.51) transfer_pct 332.901 -248.818 -248.818 18.375 -0.153  (0.76) (-1.72) (-1.72) (1.62) (-0.12) cargo_r 2.776 -0.502 -0.502 0.615 -1.446  (1.94) (-0.33) (-0.33) (0.75) (-1.20) f_rate -0.285 4.659 4.659 1.321* -5.126  (-0.11) (1.42) (1.42) (2.09) (-1.77) competition -0.037 -0.045** -0.045** -0.090*** 0.006  (-1.58) (-2.99) (-2.99) (-5.13) (0.22) 54                   Percentile Equation 0~20% 20%~40% 40%~60% 60%~80% 80%~100% Average car rental price Coef. Coef. Coef. Coef. Coef. lpopden -0.005 0.003 0.003 0.008 -0.036  (-0.16) (0.11) (0.11) (0.33) (-1.16) logpop -0.032 0.034 0.034 0.009 0.097  (-0.69) (0.97) (0.97) (0.16) (1.83) povertyratio -1.195 0.243 0.243 0.303 -1.005  (-1.37) (0.45) (0.45) (0.44) (-1.41) wage 0.001 -0.013 -0.013 0.014 -0.002  (0.05) (-0.93) (-0.93) (0.52) (-0.13) num_hq 0.006 -0.002 -0.002 -0.001 0.001  (0.59) (-0.94) (-0.94) (-0.44) (0.54) lgdp 0.007 0.355* 0.355* 0.055 -0.310  (0.03) (2.00) (2.00) (0.25) (-1.60) lhouse -0.108 0.021 0.021 -0.003 -0.350*  (-1.24) (0.26) (0.26) (-0.04) (-2.18) pubratio 0.534 -0.995 -0.995 -0.853 0.342  (0.71) (-1.16) (-1.16) (-1.17) (0.15) Passenger volume Coef. Coef. Coef. Coef. Coef. laavg_price 2.417 -0.974 -0.974 -0.224 0.398  (1.71) (-1.32) (-1.32) (-0.35) (0.31) lcharge_2 -1.003*** -0.787*** -0.787*** -1.443*** -0.786***  (-6.64) (-9.84) (-9.84) (-8.50) (-7.29) lgdp 1.598 1.299** 1.299** 1.532* 1.693  (1.67) (2.68) (2.68) (2.08) (1.71) lpopden 0.152 0.151* 0.151* 0.048 -0.067  (1.13) (2.28) (2.28) (0.60) (-0.53) logpop 0.134 0.224* 0.224* 0.435** 1.003***  (0.65) (2.47) (2.47) (2.61) (4.44) povertyratio -0.389 -0.570 -0.570 -0.163 2.677  (-0.10) (-0.37) (-0.37) (-0.08) (0.77) holiday 0.594 0.369 0.369 0.081 1.088  (1.33) (1.40) (1.40) (0.35) (1.08) hub -8.089**   -0.041 0.700  (-3.03)   (-0.11) (1.03) num_p 0.522 -0.356 -0.356 -0.396 -0.089  (1.61) (-1.57) (-1.57) (-1.48) (-0.25) num_hq -0.040 -0.003 -0.003 -0.007 -0.010  (-1.01) (-0.51) (-0.51) (-1.56) (-1.28) num_line 0.003*** 0.002*** 0.002*** 0.000*** 0.001***  (4.51) (7.00) (7.00) (4.74) (3.87) lav_dist 0.159 0.588*** 0.588*** 0.756*** 0.255  (0.84) (3.50) (3.50) (4.78) (1.45) wage -0.008 0.071* 0.071* 0.087 0.034  (-0.12) (2.19) (2.19) (1.25) (0.38) 55                   Percentile Equation 0~20% 20%~40% 40%~60% 60%~80% 80%~100% Aeronautical charge Coef. Coef. Coef. Coef. Coef. lpsg_05 -0.693*** -0.915*** -0.915*** -0.590*** 0.071  (-3.93) (-5.46) (-5.46) (-4.47) (0.50) laavg_price 0.107 -0.903 -0.903 0.414 -0.811  (0.12) (-0.98) (-0.98) (0.80) (-0.93) lc_rev 0.153 0.301** 0.301** 0.219* -0.046  (1.39) (2.83) (2.83) (2.00) (-0.65) lfrt_05 -0.016 0.046 0.046 0.028 -0.080**  (-0.53) (1.73) (1.73) (0.97) (-2.82) transfer_pct -35.196 76.152 76.152 -1.262 -1.039*  (-0.19) (1.32) (1.32) (-0.25) (-2.24) hhi -1.217** 0.193 0.193 0.353 -0.589  (-2.64) (0.73) (0.73) (1.13) (-1.73) num_line 0.002*** 0.001*** 0.001*** 0.000 0.000*  (3.51) (5.24) (5.24) (1.88) (2.04) num_p 0.162 -0.145 -0.145 -0.180 0.457*  (0.79) (-0.96) (-0.96) (-1.15) (2.45) hub -2.120   0.004 -0.384  (-1.25)   (0.02) (-0.78) f_rate -15.696 12.003 12.003 2.221 -17.914  (-1.63) (1.48) (1.48) (1.48) (-1.44) lav_dist -0.326* 0.272 0.272 0.262* -0.411**  (-2.42) (1.34) (1.34) (2.07) (-3.03) cost 1.461*** 2.309*** 2.309*** 0.291 3.044***  (3.83) (7.50) (7.50) (1.07) (8.67) s_w 0.315   0.109 -0.302  (0.72)   (0.63) (-1.11) lgdp 1.192** 1.459*** 1.459*** 0.239 0.062  (2.67) (4.70) (4.70) (0.59) (0.12)             1. t-statistics in parentheses 2. Significance: *** 0.1%, ** 1%, *5% 3. Transferring passenger rate ranges from 0.00% to 66.83%. Columns are divided by percentile of transfer rate, not exact value of transfer rate. Variables are defined the same as table 2. 56  Table 5    Group by Percentile of Transferring Passenger Rate Percentile  Equation 0~33.33%   33.33%~66.67% 66.67%~100%  Percentile  Equation 0~25%   25%~50%   50%~75%   75%~100%   Average car rental price  Coef.  Coef.  Coef. Average car rental price  Coef.  Coef.  Coef.  Coef.          lpsg_05 -0.043 0.032 0.036 lpsg_05 0.138** 0.013 0.005 -0.040  (-1.10) (0.91) (0.97)  (2.91) (0.33) (0.13) (-0.95) lcharge_2 -0.037 0.006 -0.008 lcharge_2 0.145*** 0.023 -0.071 -0.046  (-1.10) (0.09) (-0.20)  (3.39) (0.65) (-1.03) (-1.06) transfer_pct -7.093 -17.569 0.308 transfer_pct 423.942* -8.885 -30.515* 0.058  (-0.08) (-1.04) (0.31)  (1.99) (-0.25) (-2.10) (0.05) cargo_r 0.380 -0.596 0.753 cargo_r 1.740 0.447 -0.039 -0.816  (0.34) (-0.99) (1.11)  (1.39) (0.40) (-0.07) (-0.77) f_rate 1.922 0.954* -0.783 f_rate -1.081 1.116* 0.937 -0.254  (1.21) (2.35) (-1.02)  (-0.45) (2.24) (1.81) (-0.11) competition -0.019 -0.049*** -0.018 competition -0.026 -0.032** -0.068*** -0.026  (-1.22) (-3.84) (-0.94)  (-1.34) (-2.63) (-4.06) (-1.17) lpopden 0.024 0.003 0.010 lpopden -0.018 0.039 0.002 -0.014  (1.18) (0.12) (0.51)  (-0.70) (1.72) (0.07) (-0.48) logpop 0.025 0.026 0.059 logpop -0.002 -0.028 0.061 0.194***  (0.87) (0.73) (1.31)  (-0.04) (-0.86) (1.34) (3.91) povertyratio -0.170 0.286 -0.584 povertyratio -0.987 0.190 -0.393 -1.411  (-0.33) (0.66) (-0.94)  (-1.30) (0.41) (-0.70) (-1.93) wage -0.008 -0.037** -0.005 wage 0.002 -0.004 -0.036* 0.014  (-0.68) (-2.71) (-0.31)  (0.14) (-0.27) (-2.14) (0.75) num_hq -0.001 -0.001 0.000 num_hq 0.006 0.000 -0.001 -0.002  (-0.33) (-0.55) (0.23)  (0.62) (0.31) (-0.98) (-1.00) lgdp 0.488*** 0.213 -0.375* lgdp -0.005 0.098 0.205 -0.196  (3.38) (1.63) (-2.17)  (-0.03) (0.73) (1.18) (-1.00)          57  Percentile  Equation 0~33.33%   33.33%~66.67% 66.67%~100%  Percentile  Equation 0~25%   25%~50%   50%~75%   75%~100%   Average car rental price  Coef.  Coef.  Coef. Average car rental price  Coef.  Coef.  Coef.  Coef. lhouse -0.051 0.099 -0.053 lhouse -0.130 0.052 0.051 -0.281  (-0.77) (1.34) (-0.69)  (-1.49) (0.80) (0.72) (-1.87) pubratio 0.837 -0.612 0.668 pubratio 0.808 -0.711 -0.063 0.724  (1.38) (-1.68) (0.72)  (1.12) (-1.43) (-0.12) (0.68) Passenger Volume  Coef.  Coef.  Coef. Passenger Volume  Coef.  Coef.  Coef.  Coef.          laavg_price -3.279* 0.483 1.412 laavg_price 1.041 -2.763* -0.366 -2.774  (-2.18) (0.56) (1.03)  (0.80) (-2.13) (-0.50) (-1.45) lcharge_2 -0.837*** -1.584*** -1.042*** lcharge_2 -1.018*** -0.935*** -1.592*** -0.870***  (-7.54) (-12.10) (-8.69)  (-7.55) (-7.45) (-14.32) (-6.42) lgdp 2.537*** 1.536** 1.989* lgdp 1.803* 1.233 2.482*** 0.564  (3.47) (2.74) (2.15)  (2.33) (1.86) (4.45) (0.42) lpopden 0.195* 0.236** -0.037 lpopden 0.165 0.351** 0.024 0.028  (2.44) (3.07) (-0.44)  (1.85) (3.26) (0.33) (0.21) logpop 0.140 0.200 0.786*** logpop 0.157 0.049 0.389*** 1.166***  (1.31) (1.60) (4.31)  (1.11) (0.27) (3.33) (3.98) povertyratio -1.579 -3.028* -0.529 povertyratio 0.184 -4.979* -0.211 -4.087  (-0.80) (-2.15) (-0.19)  (0.07) (-2.34) (-0.15) (-0.89) holiday 0.296 -0.156 -0.116 holiday 0.686 -0.105 -0.076 0.001  (0.97) (-0.71) (-0.34)  (1.93) (-0.32) (-0.44) (0.00) hub -2.625 0.238 0.112 hub -6.963*** 1.506 0.179 -0.187  (-1.79) (0.73) (0.25)  (-3.82) (1.55) (0.65) (-0.30) num_p 0.224 0.012 0.015 num_p 0.201 -0.108 -0.347 0.083  (0.99) (0.04) (0.06)  (0.76) (-0.29) (-1.52) (0.23) num_hq -0.007 -0.006 -0.012** num_hq -0.047 -0.008 -0.006 -0.013  (-0.95) (-1.19) (-2.60)  (-1.48) (-1.12) (-1.82) (-1.52) 58  Percentile  Equation 0~33.33%   33.33%~66.67% 66.67%~100%  Percentile  Equation 0~25%   25%~50%   50%~75%   75%~100%   Passenger Volume  Coef.  Coef.  Coef. Passenger Volume  Coef.  Coef.  Coef.  Coef. num_line 0.002*** 0.0003*** 0.0004*** num_line 0.003*** 0.0002 0.0004*** 0.0005*  (5.91) (4.72) (3.80)  (6.32) (1.45) (7.86) (2.42) lav_dist 0.356* 1.155*** 0.563*** lav_dist 0.260 1.557*** 0.863*** 0.322  (2.33) (5.15) (3.84)  (1.63) (5.87) (6.20) (1.59) wage -0.017 0.081 0.079 wage 0.020 0.107 0.028 0.074  (-0.43) (1.56) (1.21)  (0.43) (1.47) (0.59) (0.70)          Aeronautical charge  Coef.  Coef.  Coef. Aeronautical charge  Coef.  Coef.  Coef.  Coef.          lpsg_05 -0.669*** -0.592*** -0.533*** lpsg_05 -0.630*** -0.516 -0.639*** -0.129  (-5.40) (-4.24) (-4.19)  (-5.03) (-1.76) (-6.50) (-0.87) laavg_price -0.737 0.485 2.294* laavg_price 0.308 0.113 0.285 -0.604  (-0.76) (0.65) (2.50)  (0.39) (0.11) (0.50) (-0.61) lc_rev 0.141* 0.143 0.134* lc_rev 0.084 0.144 0.135 0.057  (2.01) (1.48) (1.98)  (1.18) (0.77) (1.86) (0.72) lfrt_05 -0.015 -0.019 0.006 lfrt_05 0.008 -0.016 0.008 -0.045  (-0.64) (-0.76) (0.25)  (0.32) (-0.52) (0.34) (-1.51) transfer_pct 70.349 5.363 -0.907* transfer_pct 39.782 -31.081 1.227 -1.510**  (1.71) (0.63) (-2.37)  (0.43) (-1.00) (0.25) (-3.06) hhi -0.487 -0.088 -0.288 hhi -0.653* -0.425 0.115 -0.427  (-1.73) (-0.26) (-0.96)  (-2.04) (-1.19) (0.35) (-1.14) num_line 0.001*** 0.0002* 0.0003** num_line 0.001*** 0.000 0.0003*** 0.0002  (4.97) (2.41) (2.61)  (4.91) (0.30) (4.62) (1.50) num_p 0.074 0.010 0.077 num_p 0.056 -0.264 -0.135 0.306  (0.51) (0.05) (0.45)  (0.33) (-1.46) (-0.86) (1.51) hub -1.685 0.041 -0.380 hub -1.881 0.719 0.145 -0.358  (-1.65) (0.14) (-1.12)  (-1.57) (0.96) (0.64) (-0.85) 59  Percentile  Equation 0~33.33%   33.33%~66.67% 66.67%~100%  Percentile  Equation 0~25%   25%~50%   50%~75%   75%~100%   Aeronautical charge  Coef.  Coef.  Coef. Aeronautical charge  Coef.  Coef.  Coef.  Coef. f_rate -5.010 0.745 1.472 f_rate -12.880 1.204 -0.281 -3.722  (-0.81) (0.55) (0.46)  (-1.63) (0.70) (-0.22) (-0.35) lav_dist -0.009 0.658** 0.066 lav_dist -0.105 0.967*** 0.510*** -0.322*  (-0.08) (3.01) (0.54)  (-0.92) (3.57) (4.06) (-2.06) cost 1.730*** -0.004 1.048*** cost 1.541*** 2.499*** -0.010 2.531***  (6.59) (-0.07) (3.87)  (5.79) (4.25) (-0.21) (6.33) s_w 0.095 -0.039 -0.133 s_w -0.000  0.045 -0.263  (0.27) (-0.14) (-0.64)  (-0.00)  (0.28) (-1.04) lgdp 1.170** 1.363*** 0.990 lgdp 1.280*** 0.902* 1.338*** 0.128  (3.19) (3.70) (1.81)  (3.54) (2.06) (3.99) (0.22)          1. Same notation as those of table 4.  60  The rental prices for different types of car may also behave differently. We group observations into five groups with different types of rental cars including middle, standard, economic, full and compact. The results are listed in Table 6. Table 6    Group by Car Types                        Car Type Equation Middle Economy Standard Full Compact Average car rental price Coef. Coef. Coef. Coef. Coef.       lpsg_05 0.002 0.023 0.071*** -0.012 0.011  (0.10) (0.96) (3.50) (-0.71) (0.51) lcharge_2t -0.019 0.018 0.134*** 0.003 -0.017  (-0.51) (0.39) (3.35) (0.10) (-0.44) transfer_pct -1.165 -2.361** -0.684 -0.126 -1.285  (-1.54) (-2.24) (-0.78) (-0.23) (-1.61) cargo_r 0.272 0.347 0.106 0.136 0.480  (0.68) (0.65) (0.24) (0.41) (1.09) f_rate 0.737** 0.730* -0.118 0.628** 0.898**  (2.24) (1.77) (-0.34) (2.08) (2.53) competition -0.029*** -0.033*** -0.011 -0.013* -0.035***  (-3.44) (-3.11) (-1.26) (-1.83) (-3.76) lpopden 0.020 -0.001 -0.002 0.025** 0.015  (1.61) (-0.03) (-0.15) (2.42) (1.08) logpop 0.048** 0.043* 0.009 0.044*** 0.037*  (2.41) (1.73) (0.44) (2.82) (1.67) povertyratio -0.242 -0.354 -0.166 -0.200 -0.126  (-0.83) (-0.97) (-0.54) (-0.84) (-0.39) wage 0.002 -0.000 -0.008 0.007 -0.000  (0.25) (-0.02) (-0.96) (1.16) (-0.04) num_hq -0.001 -0.001 -0.000 -0.001 -0.001  (-0.90) (-1.30) (-0.52) (-1.38) (-0.87) lgdp 0.212** 0.208* -0.059 0.114 0.263***  (2.46) (1.94) (-0.64) (1.53) (2.82) lhouse -0.046 -0.082* 0.006 -0.008 -0.079**  (-1.34) (-1.80) (0.16) (-0.29) (-2.04) pubratio 0.237 0.465 0.158 0.150 0.113  (0.96) (1.36) (0.55) (0.80) (0.41)       Passenger  volume Coef. Coef. Coef. Coef. Coef. lavg_price 0.310 -0.040 4.965*** 0.262 -0.517  (0.28) (-0.05) (4.61) (0.20) (-0.49) lcharge_2 -1.663*** -1.654*** -1.719*** -1.564*** -1.701***  (-16.50) (-16.70) (-17.21) (-15.00) (-16.63) 61                         Car Type Equation Middle Economy Standard Full Compact Passenger  volume Coef. Coef. Coef. Coef. Coef. lgdp 1.818*** 2.031*** 1.706*** 1.774*** 2.019***  (4.14) (5.03) (4.21) (4.09) (4.72) lpopden 0.136** 0.095* 0.087* 0.187*** 0.130**  (2.43) (1.89) (1.83) (3.19) (2.47) logpop 0.290*** 0.248*** 0.191*** 0.325*** 0.241***  (3.32) (3.20) (2.59) (3.69) (2.89) povertyratio -1.131 -1.140 -0.813 -1.463 -0.939  (-0.87) (-0.98) (-0.75) (-1.03) (-0.76) holiday 0.153 0.112 0.131 0.103 0.084  (0.93) (0.73) (0.88) (0.63) (0.55) hub      0.168 0.211 -0.016 0.134 0.244  (0.52) (0.67) (-0.05) (0.42) (0.75) num_p -0.091 -0.074 -0.006 -0.067 -0.020  (-0.52) (-0.43) (-0.04) (-0.38) (-0.11) num_hq -0.005 -0.005* -0.005* -0.007** -0.005*  (-1.64) (-1.78) (-1.66) (-2.01) (-1.66) num_line 0.0004*** 0.0004*** 0.0003*** 0.0004*** 0.0004***  (6.05) (7.07) (5.39) (5.62) (6.42) lav_dist 0.640*** 0.709*** 0.663*** 0.640*** 0.661***  (5.54) (6.08) (6.06) (5.46) (5.64) wage 0.047 0.037 0.042 0.068** 0.039  (1.38) (1.25) (1.47) (2.06) (1.18)       Aeronautical charge Coef. Coef. Coef. Coef. Coef.       lpsg_05 -0.504*** -0.541*** -0.487*** -0.477*** -0.495***  (-6.49) (-6.50) (-5.92) (-5.44) (-6.29) lavg_price 1.184* 0.374 3.480*** 2.841*** 0.378  (1.72) (0.65) (4.52) (2.94) (0.55) lc_rev 0.078* 0.092* 0.034 0.054 0.070  (1.82) (1.82) (0.65) (1.22) (1.56) lfrt_05 -0.009 -0.008 -0.002 -0.010 -0.009  (-0.62) (-0.59) (-0.16) (-0.60) (-0.70) transfer_pct -0.315 -0.126 -0.230 -0.705** -0.357  (-1.04) (-0.41) (-0.85) (-1.99) (-1.25) hhi -0.038 -0.041 0.086 -0.061 0.007  (-0.23) (-0.25) (0.52) (-0.33) (0.04) num_line 0.0002*** 0.0002*** 0.0001*** 0.0002*** 0.0002***  (4.38) (4.78) (2.96) (3.67) (4.39) num_p -0.028 -0.018 0.011 -0.036 0.008  (-0.27) (-0.17) (0.11) (-0.35) (0.08) hub 0.026 0.059 -0.033 0.016 0.065  (0.12) (0.28) (-0.17) (0.08) (0.29) 62                         Car Type Equation Middle Economy Standard Full Compact Aeronautical charge Coef. Coef. Coef. Coef. Coef. f_rate 0.134 0.578 0.303 -0.997 0.315  (0.13) (0.59) (0.32) (-0.77) (0.31) lav_dist 0.284*** 0.326*** 0.320*** 0.285*** 0.303***  (3.44) (3.87) (3.98) (3.47) (3.57) cost 0.116* 0.071 0.077 0.122* 0.099  (1.75) (1.12) (1.28) (1.68) (1.53) s_w 0.002 -0.008 -0.119 -0.022 -0.009  (0.01) (-0.07) (-0.90) (-0.15) (-0.07) lgdp 0.848*** 1.096*** 0.906*** 0.739*** 0.986***  (3.53) (4.67) (4.01) (2.86) (4.09)       1. t-statistics in parentheses 2. Significance: *** 1%, ** 5%, * 10% 3. Variables are defined the same as those in table 2, except the car rental prices are the average prices for each type of car across all rental firms at the airport.  According to the results, for standard-size car, the aeronautical charges and car rental prices are positively correlated, implying a compensation-like strategy. The rental price for standard-size car and passenger volume are also positively correlated. The reason behind this phenomenon is not well known, but may be due to an unusually high demand for standard-size car than other types. Therefore the owners of rental companies adopt a different strategy for this highly demanded products, to elicit higher profits by following the airports’ charging trend and passenger quantity changes.  Again, except for the differences discussed above for the standard-size group, the results are overall consistent with our previous results    63  7 Conclusion Based on a cross-section dataset covering 337 airports in the United States, this thesis examines the interaction among airport aeronautical charges, traffic volume, and car rental prices. The primary contribution is to answer the important empirical question on whether airport concession services can affect the passenger volume. Namely, we provide policy makers and researchers empirical evidence to better understand whether the decision of buying flight tickets and the decision of purchasing concession goods are made simultaneously or sequentially. The results also have implication on how concessionaires (car rental company managers in this context) make pricing strategies according to the traffic flow and airport charges. We also contribute to the literature on determinants of the aeronautical charges.  We have the following main empirical findings. First of all, car rental price has no significant impact on passenger volume, indicating that rental cost, which is an important cost category of airport concessional goods, does not affect passenger volume of airport. Secondly, aeronautical charge has significant negative effect on passenger volume, and it has no significant effect on car rental price. Thirdly, only rental price in airports with a very low transfer rate responds positively to passenger volume, but rental price overall does not respond to passenger volume or aeronautical charge. The significant factors affecting aeronautical charges we find are: negative effect of passenger quantity, the average distance for arriving flights (positively associated with aeronautical charges), positive association with costs, local GDP (positive effect), and number of flights flying into the airport (positive effect).  64  The evidence that concession price has no effect on passenger volume is particularly instructive for airports management. Traditionally, people think airports serve the airlines’ needs and the airlines own passengers, which is the “one-side” view. Now with the growing importance of concession operations, airports are regarded as two-sided platforms, placing itself between airlines and passengers. In this situation, airlines and passengers are two groups of customers for airports, where each one will be better off if the other thrives. Then the airports also benefit from the profits earned from both sides, which constitute two streams of revenues: aeronautical source and concession source. Therefore, when it comes to the issue of airport regulation or pricing strategy, if you try to structure the regulation on one side you have to pay attention to the impact on the other. However, with our evidence in 2005 with car rental services, we know that if airports want to maximize the concession profits at the price level, they do not need to worry about the effect on the aeronautical or the airline side, because concession pricing will not have significant impact on the demand for flight tickets. So at least in this situation, managers don’t need to consider too much about the two-sided effect. So this is one important managerial implication.  In addition, as elaborated in the section of related literature, among the abundant theoretical studies, some investigate airport regulation, some examine the airport pricing, and some discuss the issue of privatization, all of which address different aspects of airport management.  They all make assumptions about the effect of concession price on passengers’ travel decision. Some assumed there exists such an effect while the others assumed the other way. Therefore, our empirical evidence implies that those theoretical models which assume non-existence of such effect are more likely to correctly capture the situation in real world. 65  Our work has several limitations that can be explored for future research. First, we currently have only gathered cross-section data in 2005, which is limited in the capability to illustrate causal relationship between the variables we are interested in. Also, recent years have seen growing availability of online reservation, which may also bring in new factors affecting how people make travel and rental decision. This can be resolved by obtaining data of multiple years, which cannot be obtained now since the needed data has not been published. Second, even though car rental revenue accounts for a big share (around 20%) of total airport revenue, more information about other concession services such as car parking may add credibility to the results of this study. Future work on factors explaining the attractiveness of a destination can also be explored, thus the regression model can be further improved, specifically the equation with passenger volume as dependent variable. For example, the number of hotels and restaurants can also be used as explanatory variables in the second equation. These two factors are traditionally used in the gravity models, which capture the attractiveness of a specific area. Due to the limitation of data, these factors are not included, so future research can also make progress from this perspective.   66  References  Bel, G., Fageda, X., 2010. Factors explaining charges in European airports: competition, market size, private ownership and regulation. Journal of Regulatory Economics 37, 142--161. Bilotkach, V., Clougherty, J.A., Mueller, J., Zhang, A., 2012. Regulation, privatization, and aeronautical charges: panel data evidence from European airports. Journal of Regulatory Economics 42, 73--94. Choo, Y.Y., 2014. Factors affecting aeronautical charges at major US airports. Transportation Research Part A 62, 54-62. Czerny, A.I., 2006. Price-cap regulation of airports: single-till versus dual-till. Journal of Regulatory Economics 30, 85--97. Czerny, A.I., 2012. Public versus private airport behavior when concession revenues exist. Economics of Transportation 2, 38-46. Czerny, A., Lindsey, R. Multiproduct competition with core and side services. Working paper. University of British Columbia. WHU -- Otto Beisheim School of Management. D'Alfonso, T., Jiang, C., Wan, Y., 2013. Airport pricing, concession revenues and passenger types. Journal of Transport Economics and Policy 47, 71-89. Doganis, R., 2005. The airport business. Routledge. Forsyth, P., 2004. Locational and monopoly rents at airports: creating them and shifting them. Journal of Air Transport Management 10, 51–60. 67  Geuens, M., Vantomme, D., Brengman, M., 2004. Developing a typology of airport shoppers. Tourism Management 25, 615–622. Hayashi, F., 2000. Econometrics. Princeton University Press. Khan, R., Singh, V., Zhu, T. Price discrimination and competition in the auto rental industry. Working in progress. Lin, M.H., 2006. Hub-airport competition: connecting time differentiation and concession consumption. Australian Economic Papers 45 (4), 299–317. Oum, T.H., Zhang, A., Zhang, Y., 2004. Alternative forms of economic regulation and their efficiency implications for airports. Journal of Transport Economics and Policy 28, 217--246. Starkie, D., 2001. Reforming UK airport regulation. Journal of Transport Economics and Policy 35, 119--135. Starkie, D., 2008. Aviation Markets: Studies in Competition and Regulatory Reform. Ashgate. Thompson, B., 2007. Airport retailing in the UK. Journal of Retail and Leisure Property 6, 203–211. Tinbergen, Jan, 1962. Shaping the World Economy: Suggestions for an International Economic Policy, Journal of Farm Economics, Vol. 46, No. 1 (Feb., 1964), pp. 271-273. Van Dender, K., 2007. Determinants of fares and operating revenues at US airports. Journal of Urban Economics 62, 317--336. 68  Wooldridge, J.M., 2002. Econometric analysis of cross section and panel data. MIT Press, Cambridge. Yang, G., Zhang, A., 2011. Price cap regulation at congested airports. Journal of Regulatory Economics 39, 293–312. Zhang, A., Zhang, Y., 1997. Concession revenue and optimal airport pricing. Transportation Research Part E 33, 287--296. Zhang, A., Zhang, Y., 2003. Airport charges and capacity expansion: effects of concessions and privatization. Journal of Urban Economics 53, 53--75. Zhang, A., Zhang, Y., 2010. Airport capacity and congestion pricing with both aeronautical and commercial operations. Transportation Research Part B 44, 404--413. Zhang, A., Czerny, A., 2012. Airports and airlines economics and policy: An interpretive review of recent research. Economics of Transportation 1, 15-34. Zellner, A., Theil, H., 1962. Three-stage least squares: simultaneous estimation of simultaneous equations. Econometrica 30, 54-78. 69  Appendix Included Airports and Hub Status Airport Code Airport Name City Hub ANC Ted Stevens Anchorage Intl Anchorage 1 ATL Hartsfield Jackson Atlanta Intl Atlanta 1 BNA Nashville Intl Nashville 1 BOS General Edward Lawrence Logan Intl Boston 1 BWI Baltimore Washington Intl Baltimore 1 CAE Columbia Metropolitan Columbia 1 CLE Cleveland Hopkins Intl Cleveland 1 CLT Charlotte Douglas Intl Charlotte 1 CMH Port Columbus Intl Columbus 1 COS City Of Colorado Springs Muni Colorado Springs 1 CVG Cincinnati Northern Kentucky Intl Cincinnati 1 DAY James M Cox Dayton Intl Dayton 1 DCA Ronald Reagan Washington Natl Washington 1 DFW Dallas Fort Worth Intl Dallas-Fort Worth 1 DTW Detroit Metro Wayne Co Detroit 1 FLL Fort Lauderdale Hollywood Intl Fort Lauderdale 1 HNL Honolulu Intl Honolulu 1 IAD Washington Dulles Intl Washington 1 IAH George Bush Intercontinental Houston 1 JFK John F Kennedy Intl New York 1 LAX Los Angeles Intl Los Angeles 1 LGA La Guardia New York 1 MCI Kansas City Intl Kansas City 1 MCO Orlando Intl Orlando 1 MDW Chicago Midway Intl Chicago 1 MEM Memphis Intl Memphis 1 MKE General Mitchell Intl Milwaukee 1 MSP Minneapolis St Paul Intl Minneapolis 1 OGG Kahului Kahului 1 ORD Chicago Ohare Intl Chicago 1 PDX Portland Intl Portland 1 PHL Philadelphia Intl Philadelphia 1 PHX Phoenix Sky Harbor Intl Phoenix 1 PIT Pittsburgh Intl Pittsburgh 1 RDU Raleigh Durham Intl Raleigh-durham 1 SAF Santa Fe Muni Santa Fe 1 70  Airport Code Airport Name City Hub SEA Seattle Tacoma Intl Seattle 1 SFO San Francisco Intl San Francisco 1 SLC Salt Lake City Intl Salt Lake City 1 SRQ Sarasota Bradenton Intl Sarasota 1 STL Lambert St Louis Intl St. Louis 1 TPA Tampa Intl Tampa 1   ABE Lehigh Valley Intl Allentown 0 ABI Abilene Rgnl Abilene 0 ABQ Albuquerque International Sunport Albuquerque 0 ABR Aberdeen Regional Airport Aberdeen 0 ACK Nantucket Mem Nantucket 0 ACT Waco Rgnl Waco 0 ACV Arcata Arcata CA 0 AEX Alexandria Intl Alexandria 0 AGS Augusta Rgnl At Bush Fld Bush Field 0 AHN Athens Ben Epps Airport Athens 0 ALB Albany Intl Albany 0 ALS San Luis Valley Regional Airport Alamosa 0 ALW Walla Walla Regional Airport Walla Walla 0 AMA Rick Husband Amarillo Intl Amarillo 0 AOO Altoona Blair Co Altoona 0 APF Naples Muni Naples 0 APN Alpena County Regional Airport Alpena 0 ATW Appleton Appleton 0 ATY Watertown Regional Airport Watertown 0 AUG Augusta State Augusta 0 AUS Austin Bergstrom Intl Austin 0 AVL Asheville Regional Airport Asheville 0 AVP Wilkes Barre Scranton Intl Scranton 0 AZO Kalamazoo Kalamazoo 0 BDL Bradley Intl Windsor Locks 0 BED Laurence G Hanscom Fld Bedford 0 BFF Western Nebraska Regional Airport Scottsbluff 0 BFL Meadows Fld Bakersfield 0 BGM Greater Binghamton Edwin A Link Fld Binghamton 0 BGR Bangor Intl Bangor 0 BHB Hancock County - Bar Harbor Bar Harbor 0 BHM Birmingham Intl Birmingham 0 BIL Billings Logan International Airport Billings 0 BIS Bismarck Municipal Airport Bismarck 0 BJI Bemidji Regional Airport Bemidji 0 BLI Bellingham Intl Bellingham 0 71  Airport Code Airport Name City Hub BMI Central Illinois Rgnl Bloomington 0 BOI Boise Air Terminal Boise 0 BPT Southeast Texas Rgnl Beaumont 0 BQK Brunswick Golden Isles Airport Brunswick 0 BRD Brainerd Lakes Rgnl Brainerd 0 BRL Southeast Iowa Regional Airport Burlington 0 BRO Brownsville South Padre Island Intl Brownsville 0 BTM Bert Mooney Airport Butte 0 BTV Burlington Intl Burlington 0 BUF Buffalo Niagara Intl Buffalo 0 BUR Bob Hope Burbank 0 BZN Gallatin Field Bozeman 0 CAK Akron Canton Regional Airport Akron 0 CDC Cedar City Rgnl Cedar City 0 CEC Del Norte County Airport Crescent City 0 CEZ Cortez Muni Cortez 0 CGI Cape Girardeau Regional Airport Cape Girardeau 0 CHA Lovell Fld Chattanooga 0 CHO Charlottesville-Albemarle Charlottesville VA 0 CHS Charleston Afb Intl Charleston 0 CID Cedar Rapids Cedar Rapids 0 CKB Harrison Marion Regional Airport Clarksburg 0 CLL Easterwood Fld College Station 0 CLM William R Fairchild International Airport Port Angeles 0 CMI Champaign Champaign 0 COD Yellowstone Rgnl Cody 0 COU Columbia Rgnl Columbia 0 CPR Natrona Co Intl Casper 0 CRP Corpus Christi Intl Corpus Christi 0 CRW Yeager Charleston 0 CSG Columbus Metropolitan Airport Columbus 0 CVN Clovis Muni Clovis 0 CYS Cheyenne Rgnl Jerry Olson Fld Cheyenne 0 DAB Daytona Beach Intl Daytona Beach 0 DAL Dallas Love Fld Dallas 0 DBQ Dubuque Rgnl Dubuque IA 0 DDC Dodge City Regional Airport Dodge City 0 DEN Denver Intl Denver 0 DHN Dothan Rgnl Dothan 0 DIK Dickinson Theodore Roosevelt Regional Airport Dickinson 0 DLH Duluth Intl Duluth 0 DRO Durango La Plata Co Durango 0 72  Airport Code Airport Name City Hub DSM Des Moines Intl Des Moines 0 DUJ DuBois Regional Airport Du Bois 0 EAR Kearney Municipal Airport Kearney 0 EAT Pangborn Field Wenatchee 0 EAU Chippewa Valley Regional Airport Eau Claire 0 EKO Elko Regional Airport Elko 0 ELM Elmira Corning Rgnl Elmira 0 ELP El Paso Intl El Paso 0 ENA Kenai Muni Kenai 0 ERI Erie Intl Tom Ridge Fld Erie 0 ESC Delta County Airport Escanaba 0 EUG Mahlon Sweet Fld Eugene 0 EVV Evansville Regional Evansville 0 EWB New Bedford Regional Airport New Bedford 0 EYW Key West Intl Key West 0 FAI Fairbanks Intl Fairbanks 0 FAR Hector International Airport Fargo 0 FAT Fresno Yosemite Intl Fresno 0 FAY Fayetteville Regional Grannis Field Fayetteville 0 FLG Flagstaff Pulliam Airport Flagstaff 0 FLO Florence Rgnl Florence 0 FNT Bishop International Flint 0 FOD Fort Dodge Rgnl Fort Dodge 0 FOE Forbes Fld Topeka 0 FSD Sioux Falls Sioux Falls 0 FSM Fort Smith Rgnl Fort Smith 0 FWA Fort Wayne Fort Wayne 0 GCC Gillette-Campbell County Airport Gillette 0 GCK Garden City Rgnl Garden City 0 GEG Spokane Intl Spokane 0 GFK Grand Forks Intl Grand Forks 0 GGG East Texas Rgnl Longview 0 GJT Grand Junction Regional Grand Junction 0 GLH Mid Delta Regional Airport Greenville 0 GNV Gainesville Rgnl Gainesville 0 GPT Gulfport-Biloxi Gulfport 0 GRB Austin Straubel Intl Green Bay 0 GRI Central Nebraska Regional Airport Grand Island 0 GRR Gerald R Ford Intl Grand Rapids 0 GSO Piedmont Triad Greensboro 0 GSP Greenville-Spartanburg International Greenville 0 GTF Great Falls Intl Great Falls 0 73  Airport Code Airport Name City Hub GTR Golden Triangle Regional Airport Columbus Mississippi 0 GUC Gunnison - Crested Butte Gunnison 0 HGR Hagerstown Regional Richard A Henson Field Hagerstown 0 HHH Hilton Head Hilton Head 0 HII Lake Havasu City Airport Lake Havasu City 0 HLN Helena Rgnl Helena 0 HOT Memorial Field Hot Springs 0 HOU William P Hobby Houston 0 HPN Westchester Co White Plains 0 HRO Boone Co Harrison 0 HSV Huntsville International Airport-Carl T Jones Field Huntsville 0 HTS Tri State Milton J Ferguson Field Huntington 0 HVN Tweed-New Haven Airport New Haven 0 HYS Hays Regional Airport Hays 0 ICT Wichita Mid Continent Wichita 0 IDA Idaho Falls Rgnl Idaho Falls 0 IFP Laughlin-Bullhead Intl Bullhead 0 ILM Wilmington Intl Wilmington 0 IND Indianapolis Intl Indianapolis 0 INL Falls Intl International Falls 0 IPL Imperial Co Imperial 0 IPT Williamsport Rgnl Williamsport 0 ISN Sloulin Fld Intl Williston 0 ISP Long Island Mac Arthur Islip 0 ITH Ithaca Tompkins Rgnl Ithaca 0 ITO Hilo Intl Hilo 0 IYK Inyokern Airport Inyokern 0 JAC Jackson Hole Airport Jacksn Hole 0 JAN Jackson Evers Intl Jackson 0 JAX Jacksonville Intl Jacksonville 0 JBR Jonesboro Muni Jonesboro 0 JHW Chautauqua County-Jamestown Jamestown 0 JLN Joplin Rgnl Joplin 0 JMS Jamestown Regional Airport Jamestown 0 JNU Juneau Intl Juneau 0 JST John Murtha Johnstown-Cambria County Airport Johnstown 0 KOA Kona Intl At Keahole Kona 0 LAN Capital City Lansing 0 LAR Laramie Regional Airport Laramie 0 LAS Mc Carran Intl Las Vegas 0 LBB Lubbock Preston Smith Intl Lubbock 0 LBE Arnold Palmer Regional Airport Latrobe 0 74  Airport Code Airport Name City Hub LBF North Platte Regional Airport Lee Bird Field North Platte 0 LBL Liberal Muni Liberal 0 LEB Lebanon Municipal Airport Lebanon 0 LFT Lafayette Rgnl Lafayette 0 LGB Long Beach Long Beach 0 LIH Lihue Lihue 0 LIT Adams Fld Little Rock 0 LMT Klamath Falls Airport Klamath Falls 0 LNK Lincoln Lincoln 0 LNS Lancaster Airport Lancaster 0 LRD Laredo Intl Laredo 0 LSE La Crosse Municipal La Crosse 0 LWB Greenbrier Valley Airport Lewisburg 0 LWS Lewiston Nez Perce Co Lewiston 0 LYH Lynchburg Regional Preston Glenn Field Lynchburg 0 MAF Midland Intl Midland 0 MBS Mbs Intl Saginaw 0 MCE Merced Municipal Airport Merced 0 MCN Middle Georgia Rgnl Macon 0 MCW Mason City Municipal Mason City 0 MDT Harrisburg Intl Harrisburg 0 MEI Key Field Meridian 0 MFR Rogue Valley Intl Medford Medford 0 MGM Montgomery Regional Airport  MONTGOMERY 0 MGW Morgantown Muni Walter L Bill Hart Fld Morgantown 0 MHK Manhattan Reigonal Manhattan 0 MKG Muskegon County Airport Muskegon 0 MKK Molokai Molokai 0 MKL Mc Kellar Sipes Rgnl Jackson 0 MLB Melbourne Intl Melbourne 0 MLI Quad City Intl Moline 0 MLU Monroe Rgnl Monroe 0 MOB Mobile Rgnl Mobile 0 MOD Modesto City Co Harry Sham Modesto 0 MOT Minot Intl Minot 0 MQT Sawyer International Airport Marquette 0 MSL Northwest Alabama Regional Airport Muscle Shoals 0 MSN Dane Co Rgnl Truax Fld Madison 0 MSO Missoula Intl Missoula 0 MSY Louis Armstrong New Orleans Intl New Orleans 0 MVY Martha\\'s Vineyard Vineyard Haven MA 0 MWA Williamson Country Regional Airport Marion 0 75  Airport Code Airport Name City Hub MWH Grant Co Intl Grant County Airport 0 MYR Myrtle Beach Intl Myrtle Beach 0 OAJ Albert J Ellis Jacksonville NC 0 OAK Metropolitan Oakland Intl Oakland 0 OKC Will Rogers World Oklahoma City 0 OMA Eppley Afld Omaha 0 ONT Ontario Intl Ontario 0 ORF Norfolk Intl Norfolk 0 ORH Worcester Regional Airport Worcester 0 OTH Southwest Oregon Regional Airport North Bend 0 OWB Owensboro Daviess County Airport Owensboro 0 OXR Oxnard - Ventura County Oxnard 0 PAH Barkley Regional Airport PADUCAH 0 PBI Palm Beach Intl West Palm Beach 0 PDT Eastern Oregon Regional Airport Pendleton 0 PFN Panama City Bay Co Intl Panama City 0 PGA Page Municipal Airport Page 0 PGV Pitt-Greenville Airport Greenville 0 PHF Newport News Williamsburg Intl Newport News 0 PIA Peoria Regional Peoria 0 PIB Hattiesburg Laurel Regional Airport Hattiesburg/Laurel 0 PIH Pocatello Regional Airport Pocatello 0 PIR Pierre Regional Airport Pierre 0 PNC Ponca City Rgnl Ponca City 0 PNS Pensacola Rgnl Pensacola 0 PQI Northern Maine Rgnl At Presque Isle Presque Isle 0 PRC Ernest A Love Fld Prescott 0 PSM Pease International Tradeport Portsmouth 0 PSP Palm Springs Intl Palm Springs 0 PUB Pueblo Memorial Pueblo 0 PUW Pullman-Moscow Rgnl Pullman 0 PVD Theodore Francis Green State Providence 0 PWM Portland Intl Jetport Portland 0 RAP Rapid City Regional Airport Rapid City 0 RDD Redding Muni Redding 0 RDG Reading Regional Carl A Spaatz Field Reading 0 RDM Roberts Fld Redmond-Bend 0 RHI Rhinelander Oneida County Airport Rhinelander 0 RIC Richmond Intl Richmond 0 RIW Riverton Regional Riverton WY 0 RKD Knox County Regional Airport Rockland 0 RKS Rock Springs Sweetwater County Airport Rock Springs 0 76  Airport Code Airport Name City Hub RNO Reno Tahoe Intl Reno 0 ROA Roanoke Regional Roanoke VA 0 ROC Greater Rochester Intl Rochester 0 ROW Roswell Intl Air Center Roswell 0 RST Rochester Rochester 0 RSW Southwest Florida Intl Fort Myers 0 SAN San Diego Intl San Diego 0 SAT San Antonio Intl San Antonio 0 SAV Savannah Hilton Head Intl Savannah 0 SBN South Bend Rgnl South Bend 0 SBY Salisbury Ocean City Wicomico Rgnl Salisbury 0 SCE University Park Airport State College Pennsylvania 0 SDF Louisville International Airport Louisville 0 SGF Springfield Branson Natl Springfield 0 SHD Shenandoah Valley Regional Airport Weyers Cave 0 SHR Sheridan County Airport Sheridan 0 SHV Shreveport Rgnl Shreveport 0 SJC Norman Y Mineta San Jose Intl San Jose 0 SJT San Angelo Rgnl Mathis Fld San Angelo 0 SLN Salina Municipal Airport Salina 0 SMF Sacramento Intl Sacramento 0 SNA John Wayne Arpt Orange Co Santa Ana 0 SPI Abraham Lincoln Capital Springfield 0 SPS Sheppard Afb Wichita Falls Muni Wichita Falls 0 STC Saint Cloud Regional Airport Saint Cloud 0 SUN Friedman Mem Hailey 0 SUX Sioux Gateway Col Bud Day Fld Sioux City 0 SWF Stewart Intl Newburgh 0 SYR Syracuse Hancock Intl Syracuse 0 TEX Telluride Telluride 0 TLH Tallahassee Rgnl Tallahassee 0 TOL Toledo Toledo 0 TRI Tri-Cities Regional Airport BRISTOL 0 TTN Trenton Mercer Trenton 0 TUL Tulsa Intl Tulsa 0 TUP Tupelo Regional Airport Tupelo 0 TUS Tucson Intl Tucson 0 TVC Cherry Capital Airport Traverse City 0 TWF Magic Valley Regional Airport Twin Falls 0 TXK Texarkana Rgnl Webb Fld Texarkana 0 TYR Tyler Pounds Rgnl Tyler 0 TYS Mc Ghee Tyson Knoxville 0 77  Airport Code Airport Name City Hub UIN Quincy Regional Baldwin Field Quincy 0 VCT Victoria Regional Airport Victoria 0 VLD Valdosta Regional Airport Valdosta 0 VPS Eglin Afb Valparaiso 0 WDG Enid Woodring Regional Enid 0 WRL Worland Municipal Airport Worland 0 WST Westerly State Airport Washington County 0  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items