Price Discrimination in the Airline Market: The Effect of Market ConcentrationJoanna Stavins * Federal Reserve Bank of Boston 600 Atlantic Avenue Boston, MA 02106 (617) 973-4217 e-mail: [email protected] November 25, 1996 * Economist, Federal Reserve Bank of Boston. The views expressed in this paper are those of the author and do not necessarily reflect the official views of the Federal Reserve Bank of Boston or the Federal Reserve System. Price Discrimination in the Airline Market: The Effect of Market Concentration Joanna Stavins ABSTRACT Economic theory suggests that a monopolist can price discriminate more successfully than can a perfectly competitive firm. Most real-life markets, however, fall somewhere in between the two extremes. What happens as the market becomes more competitive: Does price discrimination increase or decrease? This paper examines how price discrimination changes with market concentration in the airline market. The paper uses data on prices and ticket restrictions across various routes within the United States, controlling for distances and airport gate restrictions. Price discrimination is found to increase as the markets become more competitive. This paper tests the hypothesis that higher market concentration lowers price discrimination in the airline market. however. However. The existing studies of the airline market show that as market concentration increases. firms have no market power to discriminate by price. Unlike previous studies of airline pricing that focused on the average fare on each flight. provided he has information about consumers’ taste differences and the transaction costs of setting multiple prices do not prevent such pricing strategies. Borenstein and Rose (1994) found a negative effect of market concentration on price dispersion. and Gale (1993) show that price discrimination may increase as the market becomes more competitive. Empirical evidence to support or reject the hypothesis is scant. Theoretical studies contradict that intuition. What happens as the market becomes more competitive: Does price discrimination increase or decrease? From the above statements. it would seem that as market concentration increases. fall somewhere in between the two extremes. Holmes (1989).Price Discrimination in the Airline Market: The Effect of Market Concentration Joanna Stavins In a perfectly competitive market. so does the average price level (Borenstein (1992). At the other extreme. however. Morrison and Winston (1990)). however. so should price discrimination. Price -1- . The airline market is well-suited for empirical analysis of the theory. In the only empirical paper that studied the relationship between the distribution of fares for individual flights and market concentration in the airline market. Borenstein (1985). a monopolist can. few analysts have examined the price-discriminatory mechanisms that airline use. For example. this paper uses individual airline ticket prices and ticket restrictions. in Gale’s theoretical model there is more price discrimination under duopoly than under monopoly. no studies have analyzed the effect of market concentration on price discrimination in the airline market. Most real-life markets. Despite growing literature on airline pricing. The second type is a case where carriers use a rationing device and limit the supply of the cheaper good. For example.e. Travel restrictions attached to cheaper tickets make it costly for consumers to obtain discounts.e.g. or combinations of fares and restrictions attached to the tickets. by offering consumers a range of packages. while Section IV discusses the results of the two models used: one that has fixed price discrimination and another that allows price discrimination to vary with market concentration and with 1 Advance-purchase discounts have cost-based justification as well (see Borenstein and Rose (1994))..discrimination is measured with the effect of ticket restrictions on airfare. Section I explains how air carriers price discriminate and how price discrimination may change with market concentration. flexibility). by restricting the number of discounted seats on each flight. air carriers separate price-sensitive consumers with relatively low disutility from travel restrictions from price-inelastic consumers with high disutility from ticket restrictions. convenience. and second. That way. The variation in airfares can be either justified by cost differences (i.1 This paper separates cost-based route and carrier effects from discriminatory effects on price in order to test whether market concentration affects price discrimination. time. Section II describes the data used in this paper. The first type of price discrimination is known as second-degree or self-selection price discrimination. Section III presents the model. Consumers choose their preferred version of a product based on their willingness to pay for specific attributes of the good (e. -2- ... The large dispersion in airfares paid by passengers traveling on the same flight has been scrutinized by both consumer groups and the press. Both are combined in this analysis. cost-based) or discriminatory (i. Airlines price discriminate in two ways: first. demand-based). who question the fairness of charging different prices for “the same good”: a seat on a given flight. Saturday-night stayover and advance-purchase requirements are designed to discourage price-inelastic consumers from buying cheaper tickets on a given flight. and product resale must be impossible or costly. For example. thereby making them unattractive to consumers with a high valuation of time or convenience and a low price elasticity 2 The effect could also be measured by time or convenience elasticities of demand. Furthermore.2 Although the resale of airline tickets is possible. Such a market is therefore monopolistically competitive. a firm must have some market power to be able to charge prices above marginal cost. scale economies.the carrier’s market share. consumers differ because of their varying price elasticities of demand. such as blackout days or time-of-day constraints. even if both sell tickets for the Boston-Miami route. They do it by attaching various restrictions to cheaper tickets. the greater the price discrimination on the route. To price discriminate successfully. Barriers to entry arising from sunk costs. and by offering different route networks. Using two different types of ticket restrictions to measure price discrimination. Section V offers conclusions. -3- . In order to price discriminate. Airlines differentiate among themselves by occupying different slots in flight schedules. where the price difference cannot be fully explained by differences in cost. a carrier with a vast number of connections to the West Coast differentiates itself from a carrier flying only along the East Coast. I Airline Price Discrimination Price discrimination means charging different prices to different consumers. firms need to be able to separate consumer groups with different demand elasticities. to prevent arbitrage. the study finds that the more competitive the market. The air transportation market allows for price discrimination. and hub-and-spoke systems give carriers market power even on relatively competitive routes. it involves high search costs and does not eliminate restrictions. the population of consumers must be heterogeneous (otherwise the firm could not separate the market). Figure 1 shows a diagram where the business passengers’ indifference curve (UB ) and the tourist passengers’ indifference curve (UT ) are far enough apart so that the carrier can sell two separate price-restriction combinations. On the other hand. Although some restrictions may lower the airline’s costs by raising expected load factors. -4- . On the one hand. the incentive compatibility constraint is satisfied. the business consumers choose to pay PB. while the tourist consumers choose to pay PT. Arrows indicate the direction of increasing utility. consumers. Consumers maximize their expected utility from flying. time. In this simple two-type model. as the market becomes more competitive.3 The dotted lines indicate the carrier’s indifference curves. leading to lower price discrimination. Carriers’ unique flight schedules (routes. The choice depends on the consumer’s elasticity of demand with respect to convenience. carriers may be forced to charge their marginal cost. flight frequency. but still be able to retain high markups on their captive. price discrimination can increase or decrease. carriers may be forced to charge tourist consumers their marginal cost. They choose between various pricerestriction packages. With more firms in the market. Price discrimination is more likely to exist in markets where the population is dispersed (with a mixture of business and tourist passengers) and the volume of travel is large. such as between low price-high inconvenience and high price-no restrictions combinations. others mainly screen consumers on the basis of their utility functions. and airport dominance) as well as frequent flier plans induce consumers to favor specific carriers. Carriers can successfully price discriminate between different consumer groups only if the difference in demand elasticities is significant. while the solid line connects the observed price-restriction packages available to consumers. or business.of demand. even when fares and ticket restrictions are 3 In other words. or money. 1995. followed by fares offered at 21 days prior to departure. derived from a hedonic regression of ticket prices on ticket and route characteristics. as well as weekend travel. The data include fares offered for sale at various times before the scheduled travel date. changes in price discrimination involve changes in discounts given to more price elastic consumers. 14 days prior to departure. But in the case of airlines.4 Therefore. if higher competition reduces fares charged to price-elastic travelers. possibly leading to higher price discrimination on more competitive routes. It is therefore associated with raising prices for less elastic consumers. The date was picked to avoid summer or holiday peaks. Price discrimination is usually thought of as a way to extract as much consumer surplus as possible from each group of consumers. Hence. crosselasticities. The data include ticket information for flights on 12 different routes on the same day: Thursday. where price discrimination is measured with marginal implicit prices of ticket restrictions. -5- . given their utility functions (demand elasticities. A variable was constructed to control for the 4 See Borenstein (1989) for a discussion of the effect of airport dominance. price discrimination is exhibited through fare discounts. September 28. II Data This study uses data collected from the electronic version of the Official Airline Guide. Selecting a single day eliminates price differences due to travel on different days of the week. This study tests whether higher market concentration leads to lower price discrimination. Borenstein and Rose (1994) found a negative effect of concentration on price dispersion.) and income. The earliest data include fares offered 35 days prior to departure. etc. In this case. and finally 2 days prior to departure.equal. it may involve an increase in price discrimination—the discount associated with any given ticket restriction may be greater. airlines may have market power in some market segments but not in others. The routes were selected arbitrarily and range from highly concentrated to quite competitive. B Market Concentration 5 One-way fares were multiplied by two. A Ticket Restrictions The data include four restrictions that could be attached to each fare: a cancellation penalty. and other (unspecified). and the name of the carrier. The requirement to stay over Saturday night is least likely to be correlated with cost effects. Each observation contains specific information on a single ticket: the ticket price. the Saturday-night stayover requirement and the number of days of advance ticket purchase requirement were used as proxies for price discrimination. the number of days in advance that purchase is required. ticket restrictions attached to the fare (see below). whether the seat is in a first-class or a coach cabin. The routes selected and carriers serving those routes are listed in Table 1.5 the number of days prior to departure the fare was last offered. Gale and Holmes (1993) found advance-purchase discounts to be a profitmaximizing way to sell tickets. one restriction at a time was included in the estimation. The fares included both round-trip and one-way tickets. To avoid multicollinearity. whether or not a Saturday-night stayover is required.number of days prior to departure the fare was offered. In a theoretical model.6 Since some restrictions may be correlated with the carrier’s costs. 6 -6- . Other studies have found the advance-purchase requirement to be the best tool for airlines to segment their consumers. whether the fare was for a one-way or a round-trip ticket. whether the flight is direct or includes a stop. first class and coach tickets. The restrictions were highly correlated (see Table 2 for correlation coefficients). and a dummy variable indicating whether the origin or destination airport has a controlled number of landing slots. with their means and standard deviations. carriers that are located at one or both ends of the route. The results did not change when either the origin population or the product of the two population measures was used instead. As an alternative measure. (1989) showed that the number of potential competitors has a much smaller effect on ticket prices than the number of carriers actually operating on a given route. Table 3 lists the variables used in the paper. 8 -7- . Dyer. Following Brueckner. Borenstein (1989) also found the route market share to be a better predictor of ticket price than the airport market share. Graham and Kaplan (1985). However. Borenstein (1992) and Hurdle et al. 7 See Bailey. even if they are not currently operating on the route). Arithmetic means of the origin and destination populations and of per-capita income in the two cities were used as exogenous measures of demand. and Borenstein and Rose (1994). and La Guardia. Some studies have used the number of potential competitors on the route (i. John F.. the degree of tourist traffic was measured by the absolute difference in average January temperatures between the origin and destination.7 Since the data do not include the number of passengers. the number of direct flights on each route was used to calculate each carrier’s market share and the Herfindahl-Hirschman Index (HHI) on each route.8 C Other Variables Other variables include the following route and endpoint characteristics: the distance between the two endpoints. the number of carriers with direct flights on a given route was also used. and Spiller (1992) and Brueckner and Spiller (1994). a dummy variable indicating whether origin or destination city is a major hub.Previous studies have found that using either the number of flights or the number of passengers on a route as a basis for market concentration calculations yields similar results.e. population and per-capita income in both cities. Borenstein (1991). Destination airports with a regulated number of landing slots are Chicago O’Hare. Washington National. Kennedy. an indicator of tourist traffic on each route. Although price dispersion is likely to be correlated with price discrimination. the ratio of the highest to the lowest price). The result is consistent with the Borenstein and Rose (1994) finding. the ratio of standard deviation to the mean. Negative correlation between market concentration and each of the price dispersion measures shows that fares on competitive routes are more dispersed than fares on concentrated routes (see Table 4). where price discrimination is assumed not to vary with market concentration. the more competitive the route. the carrier’s market share. S is the carrier’s market share based on the number of direct flights on the -8- . and other route. The model is a reduced-form regression of airfare on ticket restrictions. In other words. the higher the price dispersion.and ticket-specific factors: Pijk = 0 + 8 1 Rijk + 2 HHIi + 9 3 Sij + 10 4 DISTi + 11 5 DISTSQi + 6 AVGPOPi + 12 7 AVGINCi + DAYSijk + ijk + TEMPi + HUBij + SLOTSi + ONEWAYijk + FIRSTijk + 13 (1) where P is the round-trip airfare. The section below specifies the econometric model used to isolate price discrimination from cost and other factors affecting airfares. The assumption will be relaxed below. market concentration on the route. the above result does not isolate any cost effects. HHI is the Herfindahl index based on the number of each carrier’s direct flights on the route. R is a ticket restriction (a Saturday-night stayover requirement or an advance-purchase requirement). III A Model Restricted Model A restricted model was first estimated.D Price Dispersion Price dispersion on each route was calculated using several different measures (standard deviation. FIRST is a dummy variable equal to 1 for first-class tickets. others vary among the carriers on the route (i. subscript j a carrier.. FIRST. DISTSQ.. TEMP is the absolute difference in mean January temperatures between the origin and destination. HUB is a dummy variable equal to 1 if the carrier has a hub in the origin or destination. while some of the variables vary among individual tickets (i. AVGINC. AVGINC is the average per-capita income in the two cities. P. and subscript k a particular ticket for the carrier on the route. AVGPOP is the average population in the two cities.. SLOTS is a dummy variable equal to 1 if the number of landing slots at either airport is regulated. Equation (2) allows for the effect of price discrimination to vary -9- . ONEWAY. Ticket restrictions are expected to have a negative effect on airfare. DIST. S and HUB).e. respectively. Subscript i denotes route. HHI. The unrestricted model below allows for price discrimination to vary with market concentration: Pijk = + + 0 + Rijk ( 6 0 + 1 HHIi + 7 2 Sij) + 8 1 HHIi + 9 2 Sij + 3 DISTi + 10 4 DISTSQi + 11 5 AVGPOPi + AVGINCi + DAYSijk + TEMPi + HUBij + SLOTSi + ONEWAYijk + FIRSTijk + (2) 12 ijk The variables are defined as above. As the specification shows.e. AVGPOP. TEMP. and DAYS). and SLOTS). some variables vary only among routes (i. and market share to have a positive effect. DIST and DISTSQ are the distance between the two endpoints and distance squared. R.e. ONEWAY is a dummy variable equal to 1 for one-way tickets.route. B Unrestricted Model Price discrimination in the restricted model above is measured by 1 and is assumed not to vary with market concentration. and DAYS indicates the number of days prior to departure the fare was last offered. 10 9 -10- . the availability of gates and planes. frequent flier programs. The effect of a ticket restriction on price (price discrimination) was negative and significant. For example. there may be an omitted-variable bias. concentration in any given city-pair market can be assumed to be fixed. To the extent that ticket prices are correlated with some unobserved. while Table 5b presents the results with the advance-purchase requirement. and 2 the effect of carrier’s market share on price discrimination. HHI is assumed to be exogenous in airfare estimation.05 drop in the ticket price.40 decrease in the ticket price. Entry barriers prevent new carriers from entering city-pair routes (e. IV A Results Restricted Model Equation (1) was estimated using OLS. Increasing the advance-purchase requirement by a day resulted in a $9. on average. then. The carrier dummies control for some of those effects.9 Fixed-effects estimation with carrier dummies was used to control for carrier-specific characteristics. Adding a Saturday-night stayover requirement resulted. a ticket with a 14-day advance purchase requirement cost $131.10 Table 5a presents the results with the Saturdaynight stayover requirement. In the short run. and scale economies in network size). etc. in a $297. Kaplan and Sibley (1983).60 less than a similar ticket on the same Following Graham. and travel agents’ promotion systems raise switching costs and create further scale economies. Airlines set ticket prices based not only on the specific route conditions. whether the Saturday-night stayover or the advance-purchase requirement was used.with market concentration: Pijk/ Rijk = 0 + 1 HHIi + 2 Sij (3) where 1 measures the effect of market concentration. All of these factors create high costs of entry into the airline industry. Computerized reservation systems. incumbent airlines’ hub-and-spoke systems. but also on their overall market position..g. carrier-specific attributes. limited gate access. 12 One-way tickets as well as first-class tickets commanded significantly higher fares than round-trip or coach tickets. In particular. the estimated coefficient on the STAY dummy was -0. A 10 percent increase in a carrier’s market share results in a $50. The estimated coefficients on the other variables were unaffected by the choice of the market concentration measure. the effect of a carrier’s market share on price is positive and statistically significant. B Unrestricted Model The restricted model in Equation (1) assumed that price discrimination did not vary with The effect of ticket restrictions on airfare was similar when a double-log specification was used instead. but the mean per-capita income in endpoint cities was insignificant. the number of carriers with direct flights was used. whether the carrier’s market share was included or not.3 percent increase. leaving only more expensive tickets for sale. The measure is not highly significant. Evaluated at the mean fare. 12 11 -11- . The coefficient on the variable was negative and significant at the 10 percent level (more carriers on a route lead to lower fares). the data allow for examination of how prices change as the departure date gets closer. Since fares were offered at various times prior to departure. possibly because the travel date is in September. cheaper fares disappear. The effect of market concentration was insignificant. As an alternate measure of market concentration. To test whether the effect of time to departure varied by market concentration on the route.20 increase in ticket price (a 4.221. Indeed.221 × 1166 = -$257. As it comes closer to the departure date.11 In the first regression. calculated at the mean).route without the requirement. Larger population in either origin or destination led to higher fares. Waiting an additional day to buy a ticket raises the fare by around $1. The coefficient on the interaction term was insignificant. Flights with relatively more tourist traffic have somewhat lower fares. the coefficient on the DAY variable was negative in all specifications. controlling for other ticket attributes.69. DAY was interacted with HHI. the discount for Saturday-night stayover equals: P/ STAY = lnP/ STAY × P = -0. the carriers’ competition for tourist consumers (i.4 HHIi . calculated at the 25th.market concentration. The results of the regression with the Saturday-night stayover requirement are shown in Table 6a. The higher the market concentration on a route.80 25th percentile 50th percentile 75th percentile The results are consistent with the hypothesis that as more carriers operate on a given route. the higher the market concentration on a route.13 The table below shows the effect of a Saturday-night stayover ticket restriction on airfare.0 Sij (4) Equation (4) indicates that for a given level of market share.1529 0. 13 -12- . that is.363. The Effect of a Saturday-Night Stayover Requirement on Ticket Price by Market Concentration HHI Price discount for Saturdaynight stayover ($) 0. and 75th percentiles of HHI for the sample at the mean value of market share. the correlation coefficient is only 0.. 50th. while fares Although HHI and S are positively correlated. the lower the price discrimination. consumers with elastic demand) increases.33.2511 -424.1640 0.11 -208. the lower the effect of the restriction on airfare. the lower the price discount for a Saturday-night stayover restriction.8 + 2196. The unrestricted model in Equation (2) was also estimated with fixed carrier effects.e. Therefore the effect of one of the variables can be evaluated without altering the level of the other.49 -400. The estimated price discrimination is as follows: Pijk/ Satijk = -700. price discrimination is higher on routes with more competition and lower market concentration.9 HHIi . holding cost effects constant.charged to business consumers (i. The effect of market share on price discrimination is the opposite: the higher the carrier’s market share on a given route.8 + 174.. The price discrimination effect derived from the estimated equation is as follows: Pijk/ Advijk = -33. The price discrimination effect is measured in this case as the discount on airfare for increasing the advance-purchase requirement by one day. and 75th percentiles of HHI for the sample at the mean value of market share. As a result. consumers with inelastic demand) stay high. 50th. The table below shows the average effect on airfare of increasing an advance-purchase requirement by one day. it would show that individual carriers with more market power on a route price discriminate more.e.72. -13- . If a table similar to the one above was constructed for different levels of market share holding market concentration constant. the number of days of advance purchase requirement was used instead of the Saturday-night stayover requirement as the measure of restriction to quantify the degree of price discrimination (Table 6b). the larger the price discrimination by the carrier. To test whether the result is robust. calculated at the 25th.8 Sij (5) Using the advance-purchase requirement does not change the result that price discrimination decreases with market concentration on a route. holding market concentration constant (although the result was not statistically significant). The effect of market concentration is independent of whether market share is included or not—the results were almost identical (and statistically not significantly different) when the interaction with market share variable was omitted from the regression. 71.14 V Summary and Conclusion Using the effect of individual ticket restrictions on airfare as a measure of price discrimination by air carriers.0 S ij.1529 0. However. 14 -14- .8 25th percentile 50th percentile 75th percentile As with the Saturday-night stayover requirement above. the results were similar.9 Sij. the smaller the effect of the restriction on airfare. by Market Concentration HHI Price discount for each day of Advance Purchase ($) 0.7 + 323.The Effect of an Advance Purchase Requirement on Ticket Price. intuition suggests that price discrimination should increase with market concentration. Since firms in perfectly competitive markets cannot price discriminate.0 HHIi 619.1 -1. each carrier’s unique market position (route schedule.2511 -19. Travelers buying unrestricted tickets tend to prefer a particular carrier. while for the advance-purchase requirement: Pijk/ Advijk = -46. The derived price discrimination equation for the Saturday-night stayover requirement was: Pijk/ Satijk = -806. airport dominance..0 -17. frequent flier plan) enables it to retain market power with respect to its business (inelastic) consumers.1640 0. the higher the market concentration on a route. Therefore.2 HHIi . but not tourist (elastic) consumers. while higher market share leads to more price discrimination by a carrier. both direct and connecting flights). In both cases higher market concentration leads to lower price discrimination. this study finds that price discrimination decreases with market concentration.e. carriers When market share and market concentration were based on the total number of flights on a route instead of just direct flights (i.1 + 4765. even on more competitive routes. The result remains almost identical when the interaction of the advance-purchase requirement with market share is omitted from the regression. while retaining their market power in the other market segment. As a result. Even when carriers face competition on a route. -15- . the more competitive routes have more price discrimination. they effectively compete only for the price-elastic segment of the market.on competitive routes are forced to lower their tourist fares. but they are able to maintain high markups on their business fares. Richard L. Graham. Werden. 1992. _____. Daniel P. Sibley. 135-46. 389-93. 1989. Gloria J. Severin. Thomas J. _____. 1990.” American Economic Review. 16 (Autumn). Brueckner. and Daniel P. 1992. 1983. Rose. Kaplan. Spiller. Hurdle. pp. Holmes. Gregory J. vol. vol. 79 (March). 37 (October). 8 (August). Potential Entry. 380-97. 1989. pp.” Bell Journal of Economics.S. 653-83. vol.” American Economic Review. Steven A. Holmes.” RAND Journal of Economics. 38 (December). Ian L. pp. “The Effects of Third-Degree Price Discrimination in Oligopoly. Airline Competition.: MIT Press. 1985. 106 (November). “Hubs and High Fares: Dominance and Market Power in the U. 118-38. Borenstein. pp. Deregulating the Airlines. and David S. Gale. “Competition and Price Dispersion in the U. David R. pp. 1994..” Quarterly Journal of Economics. Joskow. “The Evolution of U. 1991. pp. “Price Discrimination in Free-Entry Markets. vol. “Fare Determination in Airline Huband-Spoke Networks. 1985. “Price Dispersion in a Market with Advance-Purchases. pp..” Journal of Political Economy. and Pablo T. 20 (Autumn). “Economics of Traffic Density in the Deregulated Airline Industry. pp. Nichola J. vol. 1989. 6 (Spring). _____. 1237-66. Mass. Williams. Gale. vol. 14 (Spring). Airlines. 1994. -16- . Graham. Brueckner. Ian L. and Clifford Winston.” RAND Journal of Economics. Jan K. “Advance-Purchase Discounts and Monopoly Allocation of Capacity. 309-33. pp. vol. vol.S. Cambridge. 1993. 344-65. “The Dynamics of Airline Pricing and Competition. Airline Industry. 83 (March). vol. Johnson. Kaplan. 80 (May). pp. Borenstein. vol. 23 (Autumn). 45-73. and Pablo T. “Concentration. Airline Industry.References Bailey. Dyer. Jan K. “The Dominant-Firm Advantage in Multi-Product Industries: Evidence from the U.” Journal of Economic Perspectives. pp.S. 451-64.. and Michael A.” American Economic Review. vol. pp. and Thomas J. 379-415. vol. “Efficiency and Competition in the Airline Industry. pp. and Performance in the Airline Industry. 1993.. 244-50. Andrew S.S. Severin and Nancy L. Spiller. 102 (August).” Review of Industrial Organization. 119-39. vol. Morrison. Elizabeth E.” RAND Journal of Economics.” Journal of Industrial Economics.” Journal of Law and Economics. David R. Figure 1: Market Segmentation by Air Carriers Ticket Price Ticket Restrictions -17- . TW.Continental DL . CO. NW.Southwest YX . Louis Los Angeles Chicago Orlando Washington Carriers a AA. HP. US AA. NW. TW. DL. DL. DL. NW AA. UA AA. DL. YX AA. US AA. UA.Northwest TW .America West NW . CO.Midwest Express -18- . NW. NW. Louis Destination Portland San Francisco Cleveland St. DL. NW. DL. TW. TW. US. UA. CO. WN a AA .USAir WN . NW. UA. HP.Tower Air HP . US. UA. HP. CO. CO.American CO . TW. HP. US AA. DL. UA. NW. CO. Louis Memphis Houston Minneapolis St. YX AA. CO. UA AA.TWA UA . TW. UA. UA. DL.United US .Table 1: Routes and Carriers Serving Each Route Origin Atlanta Boston Boston Boston Dallas Denver Detroit Milwaukee New York Philadelphia Pittsburgh St.Delta FF . TW. US CO. CO. DL. NW. TW. US. UA AA. FF. DL. YX AA. CO. UA. 8542 1.0000 0.9463 0.0000 0.0000 0.8992 Advance Purchase Saturday Night Other 1.0000 -19- .Table 2: Correlation Coefficients among Ticket Restrictions Ticket Restrictions Cancellation Penalty Advance Purchase Requirement Saturday Night Stayover Requirement Other Cancellation Penalty 1.6823 1.7573 0.7160 0. 24 0.93 0.49 6.54 11.20 13.15 0.05 0.02 21244.41 0.10 0.49 0.39 0.18 0.22 0.49 0.37 0.60 Standard Deviation 866.48 0.14 12.15 0.61 0.57 2957.26 0.44 0.85 0.15 0.21 1611.77 0.49 -20- .50 883.21 0.15 0.07 0.16 0.46 1773.91 20.41 4.40 5498.Table 3: Descriptive Statistics Description Round-Trip Airfare Number of Days Prior to Departure One-Way Direct Flight First Class Cancellation Penalty Advance Purchase Requirement (# days) Saturday Night Stayover Requirement Other Restriction Distance Average Population Average Per Capita Income Difference in Mean January Temperatures Market Share (Direct Flights) Market Share (All Flights) HHI (Direct Flights) HHI (All flights) Hub Dummy Restricted Slots Dummy Mean 1165. 4892 -0.P (20%)) / Median P Max / P Min HHI -0.5599 -0.6549 -21- .5082 -0.5197 -0.P (20%) Standard Deviation / Mean (P (80%) .Table 4: Correlation Between Market Concentration and Price Dispersion Price Dispersion Measure Standard Deviation P (80%) . 89 241.70 390.67 72.22 -0.01 -2.15 Standard Error 288.17 -37.80 -1.30 -3.04 0.07 0.81 201.98 -3.95 -5.89 -0.93 2.68 501.83 30.49 -22- .99 0.62 -1.34 28.05 3.01 0.00016 0.93 -1.50 -1.15 869.46 0.33 -230.67 -297.69 56.49 5.33 56.40 2172 0.16 11.02 665.94 t-Statistic -1.15 0.000051 0.52 0.01 1.23 28.05 -224.Table 5a: Restricted Model with the Saturday Night Stayover Requirement Coefficient Intercept Saturday Night Stayover Requirement HHI Market Share Distance Distance Squared Average Population Average Per Capita Income January Temperature Hub Dummy Slots Dummy One-Way First Class Number of Days Prior to Departure N R2 F -560. 70 380.000051 0.95 t-Statistic -2.01 0.20 -1.59 28.55 -0.01 1.14 0.83 -2.79 -9.97 4.02 -0.85 30.26 3.14 2138 0.46 241.92 0.14 1.Table 5b: Restricted Model with the Advance Purchase Requirement Coefficient Intercept Advance Purchase Requirement HHI Market Share Distance Distance Squared Average Population Average Per Capita Income January Temperature Hub Dummy Slots Dummy One-Way First Class Number of Days Prior to Departure N R2 F -826.82 0.07 2.55 0.93 0.32 -2.99 30.40 132.94 -216.96 27.91 -3.75 72.04 0.15 0.19 -23- .10 851.35 -1.83 203.00012 0.44 Standard Error 284.72 -0.44 -58.69 873.36 28.05 -2.15 195. 52 -1.14 0.06 0.91 27.89 -1.16 -3.92 71.92 t-Statistic -2.46 6.52 2172 0.08 -700.77 0.15 -3.31 27.34 0.21 4.01 0.99 337.89 -0.94 574.84 2196.39 259.59 Standard Error 281.24 -2.24 2.02 31.72 364.14 292.33 660.56 223.26 -1.03 -2.000050 0.72 12.81 2.29 -1.11 54.47 -32.79 66.30 -1.88 -10.53 0.95 -729.01 1.66 -24- .12 869.00016 0.40 -264.Table 6a: Unrestricted Model with the Saturday Night Stayover Requirement Coefficient Intercept Saturday Night Stayover Requirement Saturday Stayover × HHI Saturday Stayover × Market Share HHI Market Share Distance Distance Squared Average Population Average Per Capita Income January Temperature Hub Dummy Slots Dummy One-Way First Class Number of Days Prior to Departure N R2 F -811.41 -362.57 6. 98 28.67 2.68 -1.81 -433.08 26.98 Standard Error 279.56 -2.13 2138 0.28 0.01 0.15 0.00 458.17 258.01 -1.05 0.66 222.71 346.11 -8.00 -0.33 28.00011 0.01 1.25 867.94 t-Statistic -3.91 -44.66 0.38 2.000050 0.06 4.48 -3.02 -0.21 -25- .Table 6b: Unrestricted Model with the Advance Purchase Requirement Coefficient Intercept Advance Purchase Requirement Advance Purchase × HHI Advance Purchase × Market Share HHI Market Share Distance Distance Squared Average Population Average Per Capita Income January Temperature Hub Dummy Slots Dummy One-Way First Class Number of Days Prior to Departure N R2 F -867.71 -0.97 24.09 3.85 -72.84 -2.02 -213.09 28.29 30.89 6.53 -1.83 174.70 -33.29 71.87 30.42 0.18 3.65 -1.85 851.