Major League Duopolsts: When Baseball Clubs Play n Two-Team Ctes Phllp Mller Department of Economcs Mnnesota State Unversty, Mankato September 2006 Abstract: Ths paper focuses on examnng the attendance of MLB teams that play home games n the same metropoltan area duopoly teams. Comparsons were made between the determnants of attendance for duopoly teams and monopoly teams. Whle duopoly and monopoly teams share most of the same determnants, the estmated weghts on some determnants dffer. There s evdence that one duopolst s attendance s negatvely related to the other s performance. Evdence s therefore provded that fans of one team respond to qualty changes n the other team n a cty. Contact Informaton: Phllp Mller Department of Economcs Morrs Hall 150 Mnnesota State Unversty, Mankato Mankato, MN 56001 phllp.mller@mnsu.edu Second draft. An earler verson of ths paper was presented at the 2006 Western Economc Assocaton meetng n San Dego. I wsh to thank all the partcpants at the sesson for ther comments and to Zenon Zygmont for hs comments on an earler verson of the paper. Please do not cte or quote wthout permsson of the author. 1
1. Introducton In June of 2004, Erk Ahlberg descrbed the Chcago baseball scene n the Wall Street Journal: The Chcago Whte Sox have the best record n baseball, and ther best chance n years of endng an 88-year drought of World Seres champonshps. But here n one of Amerca's great sports towns, hardly anyone seems to care. The team has tred almost everythng to lure fans, ncludng half-prce tckets on Mondays, $1 hot dogs, and rovng bands of cheerleaders who gve free tckets to anyone who happens to be wearng a Whte Sox hat or jersey. Stll, the Sox are averagng only 23,000 fans a game -- a tad more than half the capacty of ther South Sde home, U.S. Cellular Feld. When the Sox recently faced another frstplace team, the Los Angeles Angels, only about 20,000 showed up, despte delghtful weather and a 2-for-1 tcket specal. Chcago s not unque n Amercan professonal sports, but t s one of only a handful of ctes that sport more than one franchse n one of the four major leagues. In 2006, four metropoltan areas (New York, Chcago, San Francsco-Oakland, and Los Angeles) had more than one Major League Baseball (MLB) franchse, two metropoltan areas (New York and San-Francsco) had more than one Natonal Football League franchse, two metropoltan areas had more than one NBA franchse (New York and Los Angeles), and New York has three NHL franchses. 2
Durng ts hstory, other ctes have called more than one MLB team the home team, namely Phladelpha, Boston, and St. Lous. But each of these ctes s now a one-team cty. One reason why so few teams n a league call the same cty home s that even wth open entry, fans would be wllng and able to fnancally support only one team. Another reason s nsttutonal. The exclusvty arrangements granted by league offcals to ndvdual franchse owners rase barrers, and thus costs, to any other potental owner seekng to locate a team wthn a certan dstance of an establshed team s stadum. Qurk and Fort (1995) note, for example, the crcle centered at an NFL team s stadum and wth a radus of 75 mles s the exclusve terrtory of that team. Pappas (2002) descrbed some of the barrers facng franchse owners (current and potental) who wshed to move nto another team s terrtory (crca 1999): (E)ther league can move nto a terrtory belongng to a club n the other league, so long as (a) ¾ of the affected league s teams consent; (b) the two parks are at least fve ar mles apart unless the two clubs mutually agree otherwse; (c) the newcomer pays the exstng club $100,000 plus half of any prevous ndemnfcaton to nvade the terrtory; and (d) the move leaves no more than two clubs n the terrtory. Ths provson dates to late 1960, when t was adopted to establsh the terms for the expanson Los Angeles Angels to play n the terrtory clamed by the Dodgers n 1958 Pappas also notes MLB s Amercan League (AL) consttuton requres that a ¾ majorty of current franchse owners must vote n favor of expanson and relocaton. In addton, 3
he notes relocaton wthn 100 ar mles of another club must also be approved by that club. 1 Olgopoly and game theory teach us that when two frms wth market power compete n the same product market, the decsons made by one frm fundamentally affect the outcomes felt and the decsons made by the competng frm. There are many nterestng examples of prce wars and other sorts of compettve (and non-compettve) outcomes stemmng from olgopolstc markets. But how do major league duopolsts behave n such stuatons? One answer: such duopolsts, along wth the rest of the league s offcals and franchse owners, cooperate to mnmze the number of home games played n the same cty on the same day. League schedules are arranged so that when one team n a cty s playng at home, the other team that calls the cty home s ether not playng or playng on the road. For example, other than nterleague games, there are no dates durng the 2006 MLB season on whch both duopoly teams n a cty compete n ther home cty on the same day wth one excepton: the Chcago Cubs and Chcago Whte Sox played home games on September 7 th. These games, however, dd not overlap (the Cubs played at 1:20 PM and the Whte Sox played at 7:05 PM). 1 Peter Angelos, the owner of the Baltmore Oroles of MLB, felt that the 2005 move of the Montreal Expos to Washngton DC was an nfrngement on hs exclusve terrtory. Angelos clamed that the Oroles terrtory ncluded parts of Pennsylvana, Maryland, Vrgna, West Vrgna, Delaware, Washngton D.C. and North Carolna (Angelos, 2006). But MLB offcals argued that televson terrtores are the property of MLB. Angelos and MLB offcals reached an agreement n early 2005 that allowed the relocaton of the Expos to Washngton D.C. (and renamed the Natonals) n exchange for majorty ownershp of a new cable televson sports network, the Md-Atlantc Sports Network (MASN), that wll present both Oroles and Natonal games (Fsher, 2005). 4
Although ther home games rarely overlap, duopolsts stll compete for the attenton of the same general group of fans. Whle many wll have strong loyaltes towards one of the teams, others wll be general fans of baseball and may smply want to see the best team play. Consequently, ths exstence of the duopolstc markets poses some nterestng questons for researchers. Does attendance of one team respond to the qualty of the other team? How does one team s attendance vary wth the prcng of the other team s tckets? I explore these questons below. The rest of the paper s organzed as follows: secton 2 descrbes the theoretcal framework, the emprcal model, and the data. Secton 3 presents the emprcal results. Secton 4 concludes. 2. The Theoretcal Framework, the Emprcal Model, and the Data The Theory Franchse owners are assumed to make choces to maxmze profts and are assumed to have some degree of market power n ther local market. I only focus on the demand for a team s tckets durng a gven season (assumed to be the short run). Duopoly teams face off-feld competton from the other local team. Consder two teams that call the same cty home: team and team j. To the extent that fans of team fnd team j s games to be a substtute good, team s attendance wll be a functon of team j s tcket prcng and performance. Let the general demand functon of team s games durng some partcular season be gven by 5
Q ( P, Pop, PCI, Qual, E( Qual ), Msc, φ Qual, λe( Qual ), δp ) = f (1) j j j where Q s the annual attendance and P s the prce of tckets to team s games. Pop s the populaton of the cty n whch team plays. PCI s the per-capta ncome n the cty n whch team plays. Qual and ( ) E Qual are the realzed and expected qualty of team respectvely. In terms of the role that expected and realzed qualty play n determnng attendance, expectatons wll play the largest role n determnng season tcket sales and early-season ndvdual game sales whle realzatons wll play an ncreasngly mportant role n determnng ndvdual games sales as the season progresses. Msc represents a vector of varous mscellaneous factors, such as the age of the stadum and tenure n a cty, that affect attendance. Qual and ( Qual ) j E represent the realzed and expected qualty of team j, the other local j team n the cty. φ and λ are constant scalars whch represent the mportance attached by fans 2 to team j s games and are assumed to be non-negatve. If φ = 0 then the fans of team do not account for the realzed qualty of team j. If λ = 0 then the fans of team do not account for the expected qualty of team j. P j s the prce of admttance to team j s games and δ s a scalar representng the mportance to whch the prce of the other team s games plays n team s demand functon. 2 The term fan s used qute loosely, but not arbtrarly, here. The term fan usually connotes fanatcsm for a partcular team. One could plausbly argue that de-hard fans would not vew the games of another local team as a substtute. On the other hand, there wll be others n both team s drawng area that wll be fans of the sport n general and wll attend games for whch they get the hghest surplus. As used n ths paper, fans refer to general fans of the sport played. 6
In the short run, the team s varable costs nclude jantoral, utlty, and mantenance expenses. Fort (2006) argues that team payroll s largely set by contracts fnalzed before a season begns and can essentally be treated as a fxed cost n a short-run analyss lke that beng developed here. Franchse owners are assumed to choose the level of attendance so that the margnal revenue generated by sellng a tcket to the last fan equals the addtonal cost of servng hm. For brevty, capacty s assumed to be rrelevant. The proft-maxmzng team chooses the optmal level of attendance, * * Q, where ( Q ) MC ( Q ) tckets s set at the reservaton prce for the last tcket offered for sale. * MR = and the prce of Note that I make no explct assumpton about the value of the margnal cost of sellng a tcket. The margnal cost of havng one more fan attend a game may be very close to zero, suggestng that proft-maxmzng franchse owners would set tcket prces very close to where demand s unt-elastc. However, numerous studes have shown that tcket prces are set n the nelastc porton of the demand curve for tckets (Fort, forthcomng). Sandy, Sloane, and Rosentraub (2004) explan ths paradox by notng that sports teams are not sngle-product producers but are, nstead, producers of multple products (the game on the feld, concessons, and souvenrs). If so, researchers can model sports teams as frms facng negatve margnal costs and ratonal proft-maxmzng franchse owners wll set tcket prces n the nelastc porton of ther demand curves. The Emprcal Model and the Data 7
In general, ths theoretcal framework suggests the followng general regresson model: Q = α + X β + X θ + ε. (2) j Q s as defned above and X s a matrx of demand factors specfc to team. X j s a matrx of varables specfc to the other team n team s cty. ε s a random error term and α, β, and θ are parameter vectors. The data cover the perod from 1970 to 2003 except for 1989 and 1990 when no tcket prce data was avalable. All U.S. Major League Baseball teams are ncluded n the analyss. Toronto and Montreal are excluded from the analyss because of the lack of metropoltan-specfc data for Toronto and Montreal. All team-specfc productvty data was obtaned from the Lahman database (www.baseball1.com). I assume that team attendance s dependent upon the team s current and prevous performance, SMSA characterstcs, team tcket prces, and other team qualtes as well as year-specfc characterstcs. Team performance s measured by team wnnng percentage. Snce attendance n one year s partly determned by the prevous season s performance, t was necessary to make an assumpton about the prevous year s wnnng percentage for expanson teams playng n ther frst year. I assume, all else equal, that fan expectatons about the performance of an expanson team n ts frst year are the same as fans of a team that won 40% of ts games n the prevous year. In other words, for each expanson team n ts frst year, I set the prevous year s wnnng percentage equal to 0.400. Makng ths assumpton does not change the basc results of the analyss. 8
Team tcket prce data was calculated usng weghted average tcket prce data obtaned from the late Doug Pappas s webste (www.roadsdephots.com) and from past personal correspondence wth Roger Noll. In years where the Pappas and Noll data each had values for each team (1975-1985) I took the average of the values reported n each dataset rather than choose between them. Tcket prce data was only avalable from the Pappas dataset for the years 1970-1974 and 1990-2003. The Noll data only had values for 1986-1988. As noted above, nether source had tcket prce data for 1989 and 1990. Therefore, records for those years were dropped. Lastly, Pappas dd not report a tcket prce for the Tampa Bay Devl Rays for that franchse s expanson year (1998). To be able to nclude ths record n the analyss, I obtaned a tcket prce value from the Team Marketng Report Database (www.teammarketng.com). SMSA populaton and per-capta ncome were obtaned from the Bureau of Economc Analyss Regonal Economc Informaton System (REIS). Other varables ncluded n the analyss are the age of each team s stadum, the age of the team, the number of years each team has been n ts current cty, and a dummy equal to one f a team had been n the playoffs the prevous season: the lagged playoff dummy. Between 1969 and 1994, teams made the playoffs by wnnng ther dvson n a twodvson format. In 1995, MLB went to a three-dvson format and teams could make the playoffs by wnnng ther dvson or by wnnng the wld card the team wth the best record that dd not wn ts dvson. Consequently, the lagged playoff dummy was set equal to one f a team won ts dvson pror to 1995, to one f a team won ts dvson or the wld card n 1995 and thereafter, and to zero otherwse. 9
The 1994 players strke resulted n the cancellng of the playoffs that year, so no team lterally won a dvson. However, I treated teams that led ther dvson at the tme the strke began as havng won ts dvson. In other words, snce fans of a team base ther decson to attend games n part on last year s team s performance, ths assumpton s akn to a representatve fan belevng somethng on the order of We would have won the dvson f t hadn t been for that pesky strke! I also nclude dummes equal to one for each of the strke years (1981, 1994, and 1995), each of whch shortened the length of the MLB season. Lastly, all dollar values are expressed n constant 2003 dollars usng the seasonallyadjusted consumer prce ndex for all urban consumers obtaned from the Bureau of Labor Statstcs data webste (stats.bls.gov). 3. Emprcal Results Table 1 presents the means of varables by team type used n the analyss. Compared to the average monopoly team (the only local team n ts partcular cty), the average duopoly team (one of two teams that competes n the same cty) s older, has larger attendance levels, wns more often, plays n a market wth more people and a hgher percapta ncome, and plays n an older stadum. Two sets of regressons were run on the data. Frst, models were ftted separately on the set of monopoly teams and the set of duopoly teams. Performng these regressons allows for a comparson of the determnants of team attendance between monopoly and duopoly teams. To make comparsons between the group of monopoly and duopoly teams, I exclude other team 10
nformaton from regresson on equaton (2) (the term of equaton (2)). X j θ s deleted from the estmaton In the second set of regressons, I ft models solely usng the duopoly team data. In these models, I explore how the performance and tcket prcng of the other team n a cty (referred to as the other team ) mpacts attendance of the team n queston. I thus nclude the expresson X j θ n the estmaton of equaton (2). When performng both sets of regressons, I tested for the presence of random effects usng the Breusch-Pagan (1980) Lagrangan Multpler and the Hausman (1978) tests. All models showed strong evdence of the presence of random effects and I estmated each model by controllng for the presence of random effects. In addton to the Hausman and Breusch-Pagan tests, I tested for the presence of frstorder autocorrelaton (an AR1 process) n each model usng the Wooldrdge (2002) test. Each model sgnfcantly showed the presence of an AR1 process. It s also plausble that team tcket prces and attendance are varables that are smultaneously determned. Therefore, I ran an AR1 random effects two-stage least squares model n each estmaton. I run frst-stage regressons on team tcket prces usng year-specfc dummes for the years not ncluded n the attendance regresson (1970-1980, 1982-1988, 1990-1993, and 1996-2002) as addtonal nstruments, wth 2003 beng the reference year. The results of these frst-stage regressons are avalable upon request. Monopolst vs. Duopolst 11
Table 2 presents the results from performng separate regressons on the monopoly sample and the duopoly sample. The estmated coeffcents on the logarthm of real predcted tcket prces are negatve and sgnfcant n both regressons. In addton, the estmated coeffcents are sgnfcantly greater than -1 (t = 11.57 n the monopoly model and = 8.10 n the duopoly model, both of whch are sgnfcant at less than the 1% level). Therefore, both models present evdence of nelastc prcng at the gate, a fndng consstent wth those of many others who have researched attendance at sportng events (see Fort (forthcomng)). The estmated coeffcents on the logarthm of real per-capta ncome s postve and sgnfcant n each model suggestng that baseball games are normal goods for both types of teams. Note that the estmate for the monopoly regresson s larger than ts counterpart for the duopoly regresson. Ths suggests that growth n real per-capta ncome has a greater mpact on monopoly teams. Ths result s an ndcaton that, n two-team ctes, general fans of baseball have more choces and when addtonal ncome s spent, t s spread among two teams rather than one. The estmated coeffcent on the logarthm of SMSA populaton s nsgnfcant n the monopoly model and postve and sgnfcant n the duopoly regressons. Accordng to these results, a change n SMSA populaton postvely affects attendance at duopolsts games only. Yet, attendance appears to be relatvely populaton nelastc for duopolsts. For nstance, a 1% ncrease n the populaton n an SMSA translates to a 0.6% ncrease n attendance. 12
The coeffcents on current and lagged team wnnng percentage are both postve and hghly sgnfcant n each model. In short, fans want to watch a wnnng team. The estmated coeffcent for current wnnng percentage s larger than that for the lagged counterpart. Ths suggests what have you done for me lately? s the more mportant determnant of whether to attend games relatve to what dd you do for me last year? for both sets of teams. In addton, the estmated coeffcents on current and lagged wnnng percent are larger n the duopoly regresson suggestng that duopolst fans are more responsve to the team s on-feld performance. The results suggest that whether a team made the playoffs the prevous season postvely and sgnfcantly mpacts attendance of monopoly teams but the sgnfcance s weak. For duopoly teams, the estmated coeffcent on the lagged playoff dummy s postve but nsgnfcant suggestng that duopolst attendance varaton s already explaned by changes n current and past wnnng percent. The coeffcent on the age of the stadum and ts quadratc term are negatve and postve respectvely and both are sgnfcant n the monopolst regresson. But n the duopolst regresson, nether estmate s sgnfcant. Only two of the duopoly teams moved nto new stadums durng the sample perod (the Chcago Whte Sox n 1991 and the San Francsco Gants n 2000). The other 6 duopolsts played n stadums opened on or before 1966. The monopolst regressons suggest that attendance ncreases when a team receves a new stadum but then begns to fall off over tme, but ths drop also tends to level off as the stadum ages. Therefore, snce the average duopolst plays n an older 13
stadum (relatve to the monopolst counterpart) any statstcal attendance effect of stadum age has possbly dmnshed. The results suggest that as a monopoly team ages, attendance drops off over tme, but the declne gradually dsspates. Ths s a smlar effect as that found wth the age of monopoly team stadums. For duopoly teams, the estmated coeffcent on the lnear age of the team s negatve yet nsgnfcant, but the coeffcent on the quadratc term s postve and sgnfcant. The results suggest, therefore, that as duopoly teams age, attendance ncreases at a quadratc rate. Both sets of results suggest that there may be some hstorcal value n older teams. The results suggest that as a monopoly team spends more tme n a cty, ts attendance rses but at a decreasng rate. Ths could be an ndcaton of the buldng of fan loyalty for these teams. No such effect s present for the duopoly model. Whle the Dodgers (1958 from New York), the Gants (1958 from New York), the A s (1968 from Kansas Cty) and the expanson Mets (1962) and Angels (1961). The other three teams date back over 100 years as of 2003. The nsgnfcance of the number of years a duopolst has spent n the current cty s potentally due to the newness of the average duopoly havng worn off by the tme the sample perod began. In each model, the coeffcent on the 1981, 1994, and 1995 strke dummes are negatve and sgnfcant showng that for those years, the shortened season, not surprsngly, resulted n lower attendance levels for both monopolsts and duopolsts. 14
Overall, the estmates n Table 2 suggest the determnants of monopolst and duopolst attendance dffer both n terms of the factors partcular to one group s attendance levels, but also n terms of the weghts gven to partcular factors. Fgure 1 presents a dagram of estmated attendance levels for the average monopolst and for the average duopolst usng the coeffcents from model 4 from Table 2. The average monopolst s the hypothetcal local monopoly team that has the average monopolst statstcs gven n Table 1 for team wnnng percentage, age of the stadum, age of the team, and years n the current cty. For brevty, I also assume that the team dd not make the playoffs and that the attendance measure was from 2003. The real tcket prce was calculated usng the same parameter estmates from the frst stage estmaton used n the regressons presented n table 2. The average duopolst s smlarly defned. At a wnnng percent of 0.400, the monopolst draws 2.1 mllon fans whle the duopolst draws 2.5 mllon fans. At a 0.500 wnnng percent, the average monopolst draws 2.2 mllon fans whle the average duopolst draws just under 2.8 mllon fans. Ths 100- pont ncrease n lagged wnnng percent led to roughly 100,000 more fans attendng the average monopolst s games, an average of approxmately 1,200 fans for each of the assumed 81 home games. For the duopolst, however, the ncrease n attendance was just over 300,000, an average of just over 3,700 per home game. When the average team s lagged wnnng percentage ncreases to 0.600, the team, f t s a monopolst, draws an addtonal 2,100 fans per game whle the average duopolst draws an addtonal 4,000 fans per game. Note the average duopolst outdraws the average monopolst for all 15
lagged wnnng percentages and the rate of ncrease of attendance s larger for the average duopolst. I now turn the reader s attenton to an examnaton of duopoly team attendance levels. Duopolst vs. Duopolst Table 3 presents regresson results on varous models estmated solely wth the duopoly team sample. These models allow the reader to examne the relatonshp between one cty s duopolst s attendance and the cty s other duopolst s tcket prces and past season s performance. Note that these models nclude nether the quadratc stadum age term nor the lnear and quadratc terms on the number of years n the cty. These varables were deleted from these models because they were nsgnfcant n every model. Other than the deleton of these varables and the ncluson of the cty s other duopolst s tcket prces and past performance, the models are the same as those presented n Table 2. The regresson results suggest the estmaton procedure fts each model well. In models 1 and 2, the estmated coeffcents of the logarthm of real tcket prces are postve but each s nsgnfcant. In model 3, ts coeffcent s negatve but nsgnfcant. The predcted tcket prces do not explan changes n attendance over and above that explaned by the other varables. The coeffcent on the logarthm of real per-capta ncome s postve and sgnfcant, suggestng that baseball games are normal goods n duopoly ctes. Models 1 and 3 each gve weak evdence that attendance may be ncome unt-elastc. In model 2 the 16
hypothess that attendance s unt-elastc wth respect to ncome cannot be rejected 3. Models 5 and 6 suggest that attendance s ncome elastc at least at the 5.6% level of sgnfcance. Each model also shows that there s a postve relatonshp between SMSA populaton and duopolst attendance but that attendance s nelastc wth respect to changes n populaton. For nstance, accordng to model 1, a 1% change n populaton leads to a 0.48% change n team attendance all else equal. Not surprsngly, the results suggest attendance s postvely and sgnfcantly affected by current and past team performance. Also, as n the monopolst vs. duopolst comparson, current wnnng percent has the greater mpact on attendance for duopoly teams and the estmates for both varables are farly robust across models. The estmated coeffcents on the age of the stadum are negatve n every regresson but not sgnfcant n any model. The quadratc term on team age s postve but nsgnfcant n each model. These results suggest that changes n attendance are not explaned by changes n the age of the stadum for duopoly teams. All three strke year dummes are negatve and sgnfcant n each model. Ths suggests, not surprsngly that duopolst attendance was lower n the strke years relatve to nonstrke years. Now I turn the reader s attenton to the mpact of a team s other local compettor s prcng and performance on the attendance of the team n queston. The other team s 3 The t-values on the hypothess test that the estmated coeffcents on the log of per-capta ncome were dfferent from 1 for models 1-3 are respectvely as follows: 1.66 (p-value = 0.098), 1.61 (p-value = 0.109), and 1.74 (p-value = 0.083). 17
the team that shares the same home cty as the team n queston. For nstance, the New York Yankees s the other team to the New York Mets and vce-versa. There s some evdence that lagged performance and prces of the other team negatvely mpact attendance of the team n queston. In models 1 and 2, the logarthm of the other team s tcket prce s negatvely related to the attendance of the team n queston. In other words, accordng to models 1 and 2, when (for example) the Cubs rase tcket prces fewer fans go to Whte Sox games. Does ths mean that Cubs and Whte Sox games are complementary goods? Ths s not lkely the case. Ths anomaly can be explaned by the logc that hgher prces are postvely correlated wth expected team performance and other amentes found at a team s ballpark. For nstance, f the Cubs mproved the amentes at Wrgley Feld, ths would be expected to draw some potental attendees away from Whte Sox games, but would also result n hgher tcket prces to Cubs games. In other words, Cubs tcket prces and Whte Sox attendance are correlated but changes n Cubs tcket prces do not per-se cause changes n Whte Sox attendance. Indeed, as evdenced by model 1 (compared to model 3), addng the other team s tcket prce to the regresson causes the coeffcent of the other team s lagged wnnng percentage to become nsgnfcant. The results also suggest that one duopolst s attendance s relatvely nsenstve 4 to changes n the other team s tcket prces. Ths s expected snce most fans don t base ther attendance decsons upon competng team prces. But some apparently do. 4 The t-statstcs for testng whether the coeffcents on the logarthm of the other team s tcket prce s greater from -1 (.e. nelastc cross-prce elastcty) are 5.37 for model 1 and 5.15 for model 2. 18
Lastly, model 4 presents regresson results where a frst-stage regresson was run on the logarthm of the other team s tcket prce. Predcted values were calculated and ncluded n a second stage regresson to control for potental smultanety bas between one local duopolst s attendance and the other s tcket prce. When the compettor s predcted tcket prce s added to the regresson, ts coeffcent s postve, suggestng that the games of local duopolsts are ndeed substtutes for one another, but the estmate s nsgnfcant. The other team s lagged wnnng percent s negatvely related to the attendance of the average duopolst provdng further evdence that fans of one team respond to the qualty of the other team. Fgure 3 presents a dagram showng attendance estmates for the average duopolst evaluated at varyng wn percents of the competng team n the cty. In addton to the assumptons that defne the average duopolst, I assume ths team won 50% of ts games n the past season. The estmates were calculated usng model 3 from Table 3. The attendance estmate of the average duopolst falls as the other team s lagged wnnng percent ncreases. For nstance, the average duopolst draws approxmately 1.84 mllon fans when the other team has a lagged wnnng percentage of 0.500. Ths attendance estmate drops to 1.78 mllon when the competng team has a lagged wnnng percentage of 0.600. However, snce the evdence on the effect of compettor team s wnnng percentage s mxed, care must be taken when nterpretng the dagram. 4. Concluson Ths paper examnes the attendance of MLB teams that play home games n the same metropoltan area (San Francsco Gants/Oakland A s, Los Angeles Dodgers/Anahem 19
Angels, Chcago Cubs/Chcago Whte Sox, and New York Mets/New York Yankees). One reason why so few ctes call more than one team the home team s because some markets can only proftably support one team. Another reason s because of exclusve terrtores granted to franchse owners by MLB offcals. Grantng these terrtores rases the costs that a current (or potental) franchse owner would ncur should he want to move hs team to another team s exclusve market. Comparsons were made between the determnants of attendance for the so-called duopoly teams and the rest of the U.S. MLB teams that do not share ther home markets wth other MLB teams (so-called monopoly teams). I found that the whle duopoly and monopoly teams share most of the same determnants, the estmated weghts on some determnants dffer. In examnng the set of duopoly teams, there s evdence that one team s attendance s dependent upon the other team s performance. 20
References Ahlberg, Erk (2004) Whte Sox Are Hot, So Why Are Fans In Chcago So Blasé? Wall Street Journal, June 15, 2005 Angelos, Peter (2006), Testmony of Peter G. Angelos Before the House Commttee on Government Reform U.S. House of Representatves Aprl 7, 2006, http://reform.house.gov/uploadedfles/oroles%20- %20Angelos%20Testmony.pdf Breusch, T. and A. Pagan (1980) The Lagrange Multpler Test and ts Applcatons to Model Specfcaton n Econometrcs, Revew of Economc Studes 47, 239-253 Fsher, Erc (2005), TV Deal Partners Oroles, Natonals, Washngton Tmes 04/01/2005, http://www.washngtontmes.com/natonal/20050401-012526- 8027r.htm Fort, Rodney (2006), Sports Economcs, Pearson, Upper Saddle, NJ Fort, Rodney. Forthcomng. "Inelastc Prcng At the Gate? A Survey." In Wladmr Andreff, Jeffrey Borland, and Stefan Szymansk (eds.) The Handbook on the Economcs of Sport (Northampton, MA: Edward Elgar Publshng, Inc.). Hausman, J.A. (1978), Specfcaton Tests n Econometrcs, Econometrca 46, 1251-1271 Pappas, Doug (2002), Insde the Major League Rules, http://roadsdephotos.sabr.org/baseball/02-4rules.htm 21
Qurk, James and Rodney Fort (1995), Hard Ball: The Abuse of Power n Pro Team Sports, Prnceton Unversty Press, Prnceton N.J. Sandy, Robert, Peter J. Sloane, and Mark Rosentraub (2004), The Economc of Sport An Internatonal Perspectve. Palgrave, Houndmlls, Basngstoke, Hampshre Wooldrdge, J. M. (2002). Econometrc Analyss of Cross Secton and Panel Data. 22
Table 1 Means of Varables All Teams Monopoly Teams Duopoly Teams Varable Mean Std. Dev. Mean Std. Dev. Mean Std. Dev. Home Attendance 1,835,244 767584.7 1,802,747 760803.8 1,903,663 778736.6 Real Per-Capta Income (BY = 2003) $29,506 5541.735 $28,402 4977.009 $31,829 5947.188 SMSA Populaton 5,158,896 4429549 2,901,260 1163896 9,912,279 4977883 Wnnng Percentage 0.501 0.0705472 0.497 0.0716924 0.509 0.0674309 Age of Stadum 27.5 23.41299 23.9 21.7676 35.0 24.96436 Age of Team 68.2 39.37214 63.7 41.0863 77.7 33.63489 Years n Current Cty 51.1 39.40202 51.0 40.67212 51.2 36.66077 Real Tcket Prce (BY = 2003) $12.92 4.094625 $12.89 4.084657 $12.99 4.122862 Sample Sze 795 539 256 23
Table 2 Regresson Results: Local Monopoly Teams vs. Local Duopoly Teams Monopoly Teams Duopoly Teams Log of Real Predcted Tcket Prce -0.4115721*** -0.2936702* 0.1219387 0.1596771 Log of Real Per-Capta Income 1.428038*** 0.9802367*** 0.1475865 0.2435428 Log of SMSA Populaton 0.0606684 0.6067568*** 0.093797 0.1343458 Team Wnnng Percent 1.872046*** 2.14254*** 0.1214603 0.1771258 Prevous Season's Wnnng Percent 0.799557*** 1.191964*** 0.141648 0.2131422 Made Playoffs Last Season 0.0376675* 0.043755 0.0204445 0.0305148 Age of Stadum -0.0148847*** -0.0015022 0.0026406 0.0061649 Age of Stadum Quadratc 0.0001264*** 0.0000103 0.0000355 0.0000819 Age of Team -0.0114733*** -0.0067167 0.004428 0.0062018 Age of Team Quadratc 0.0000752** 0.0001008** 0.0000343 0.0000425 Years n Cty 0.0112976** -0.0018487 0.0046172 0.0065309 Years n Cty Quadratc -0.0000701** -0.0000057 0.0000356 0.0000505 d1981-0.5529822*** -0.4462372*** 0.0339275 0.0486666 d1994-0.2566532*** -0.2963222*** 0.0363055 0.0552647 d1995-0.2645292*** -0.2314099*** 0.0363575 0.0551549 Intercept -1.204926-6.454069** 1.62124 3.226431 R-sq: wthn 0.7104 0.7128 between 0.6139 0.6134 overall 0.6608 0.6632 Sample Sze 539 256 Hausman Test 28.61*** 21.08** Breusch 454.58*** 354.94*** Wooldrdge test for autocorrelaton 39.026*** 52.687*** Wald ch2(16) = 814.97*** 396.32*** *** Sgnfcant at the 1% level or better ** Sgnfcant at the 5% level up to but not ncludng the 1% level * Sgnfcant at the 10% level up to but not ncludng the 5% level #Instrumented Varable for both Models: Log of Real Tcket Prce Duopoly Instruments: Log of Real Per-capta ncome; log of populaton; prevous season WPCT; age of stadum and ts quadratc term, age of team and ts quadratc term; years n cty and ts quadratc term; d1981; d1994; d1995 Addtonal Duopoly Instruments: Dummes for Each Year 1970-2002 except 1981, 1994, 1995, and 2001 Monopoly Instruments: Log of Real Per-capta ncome; prevous season WPCT; age of stadum and ts quadratc term; d1981; d1994; d1995 Addtonal Monopoly Instruments: Dummes for Each Year 1975-2002 except 1981, 1994, 1 24
Table 3 Local Duopoly Teams Model 1 2 3 4 Log of Real Predcted Tcket Prce 0.0254882 0.0718726-0.2310648-0.3539765* 0.2170681 0.215201 0.1556991 0.2050121 Log of Real Per-Capta Income 1.362915*** 1.351672*** 1.383433*** 1.495374*** 0.2184732 0.2187147 0.2208648 0.2447267 Log of SMSA Populaton 0.4782099*** 0.4731949*** 0.4965529*** 0.5189809*** 0.1222669 0.1215503 0.127755 0.1295042 Team Wnnng Percent 2.129112*** 2.123568*** 2.123344*** 2.137754*** 0.1740717 0.1744715 0.1741515 0.17553 Prevous Season's Wnnng Percent 1.216004*** 1.223843*** 1.293775*** 1.349598*** 0.1914787 0.1918742 0.1857571 0.1951537 Age of Stadum -0.0038819-0.0033154-0.0056725-0.0072527 0.005024 0.0050195 0.0049708 0.0051727 Age of Stadum Quadratc 0.0000558 0.0000493 0.0000768 0.000098 0.0000637 0.0000636 0.0000634 0.0000662 d1981-0.4537341*** -0.453333*** -0.4539674*** -0.4620531*** 0.0483111 0.048439 0.0482217 0.0500626 d1994-0.2901802*** -0.2933088*** -0.2900909*** -0.2888161*** 0.0548823 0.0549827 0.0548393 0.0551893 d1995-0.2195256*** -0.2196516*** -0.2194669*** -0.2261887*** 0.0545915 0.0547351 0.0545442 0.0557925 Log of Compettor's Tcket Prce -0.2431912* -0.295945** - - 0.1410564 0.1366202 Log of Compettor's Predcted Tcket Prce - - - 0.2910507 0.3826438 Prevous Season's Wnnng Percent of Compettor -0.2495167 - -0.3244884* -0.4214988** 0.1736355 0.1677121 0.2089301 Intercept -8.341433*** -8.265154*** -8.787791*** -10.72386*** 2.815397 2.814034 2.870868 3.534708 R-sq: wthn 0.7226 0.7205 0.7088 0.7179 between 0.3663 0.3721 0.3335 0.3293 overall 0.5764 0.5785 0.549 0.5493 Sample Sze 256 256 256 256 Hausman Test for Random Effects: 123.73*** 52.39*** 24.92*** 28.19*** Breusch Pagan LM test for RE 674.97*** 866.59*** 857.98*** 863.91*** Wooldrdge test for autocorrelaton 43.796*** 49.413*** 44.965*** 44.15*** Wald ch2(13) 393.65*** 389.71*** 386.58*** 391.53 *** Sgnfcant at the 1% level or better ** Sgnfcant at the 5% level up to but not ncludng the 1% level * Sgnfcant at the 10% level up to but not ncludng the 5% level #Instrumented Varable: Log of Real Tcket Prce Instruments for own tcket prce: Log of Real Per-capta ncome; log of populaton; prevous season WPCT; age of stadum and ts quadratc term, age of team and ts quadratc term; years n cty and ts quadratc term; d1981; d1994; d1995 Addtonal Instruments: Dummes for Each Year 1970-2002 except 1981, 1994, 1995, and 2001 Instruments for compettor's tcket prce: compettor's prevous season's wnnng percent; d1982; d1986-d1988; d1991-d1993; d1996-d2001 25
Fgure 1 Avg. Monopolst Attendance vs. Avg. Duopolst Attendance 6,000,000 5,000,000 Estmated Attendance 4,000,000 3,000,000 2,000,000 1,000,000 0 0.000 0.100 0.200 0.300 0.400 0.500 0.600 0.700 0.800 0.900 1.000 Lagged WPCT Monopoly Duopoly 26
Fgure 2 Attendance Estmates-Duopolst vs. Compettor's Lagged Wn Percent 2,400,000 2,200,000 Duopolst's Estmated Attendance 2,000,000 1,800,000 1,600,000 1,400,000 1,200,000 1,000,000 0.000 0.100 0.200 0.300 0.400 0.500 0.600 0.700 0.800 0.900 1.000 Compettor's Lagged Wn Percent 27