Arenas. Dennis Coates 1. University of Maryland Baltimore County. Working Paper No September 16, 1997

Transcription

1 The Growth Eects of Sport Franchises, Stadia and Arenas Dennis Coates 1 Brad R. Humphreys University of Maryland Baltimore County Working Paper No September 16, Send correspondence to Coates, Department of Economics, 1000 Hilltop Circle, Baltimore, MD 21250; Internet: coates@umbc.edu.

2 Abstract This paper investigates the relationship between professional sports franchises and venues and real per capita personal income in 37 Standard Metropolitan Statistical Areas in the United States over the period Our empirical framework accounts for the entry and departure of professional football, basketball and baseball franchises, the construction of arenas and stadia, and other sports related factors over this time period. In contrast to other existing studies, we nd evidence that some professional sports franchises reduce the level of per capita personal income in metropolitan areas and have no eect on the growth in per capita income, casting doubt on the ability of a new sports franchise or facility to spur economic growth. We also also nd evidence that results obtained from estimating reduced form relationships, a common practice in the literature, are not robust to alternative statistical model specications. JEL Codes: R58, O18, H70

3 1 Introduction In recent years sports franchises have used their monopoly power to extract rents from state and local governments. As the game goes, a franchise owner declares an existing facility unsuitable. Perhaps it is too old, or too small, or doesn't have enough luxury boxes and suites to raise the necessary revenues to eld a championship caliber team. The owner reminds the local government and business community that many other cities would like tohave a team, and those other cities will build a new stadium. Cities all over the country, desperate for a professional sports team, gear up to convince the owner to move. Often, the promise of a new stadium convinces the owner to stay. Part of this process is the commissioning of economic impact studies which purport to show just how much benet the city or region will reap from a new stadium, a franchise, or both. As Crompton (1995) points out, the results of these studies invariably reect the desires of those who commission them. Advocates of stadia and franchises produce impact studies which nd large economic impacts, translated as benets, from building a stadium or enticing a team to enter the city. Opponents nd small impacts. Robert Baade (1996) has argued, alone and with Richard Dye (1990), that the only way to really assess the impact of stadia and franchises is to compare the economies of the cities or regions where the sports environment has changed with that of cities where it has not changed. Two slightly dierent questions can be asked, but the discussion of sports led growth often does not distinguish between the two. First, does the change in the sports environment change the level of real per capita income? That is, does the average person in the city or region have greater economic resources as a consequence of the change? Second, is the rate of growth of income higher, lower or unchanged because of the change in the environment? The dierence between these questions is important from a policy perspective. A one time boost in income per capita is a good thing, but may not be as valuable as a permanent increase in the rate of economic growth. If proponents of sports as engines of economic growth and vitality are correct, then cities which build stadia or acquire teams should experience either a boost in real per capita income or faster economic growth than other cities and regions or both. If they are not correct, then neither the level nor the growth rate of economic variables across regions will be enhanced by changes in the sports environment. Baade and Dye make a telling point; no matter what the economic impact studies predict, the only way to gauge the actual eects is to develop and estimate a model of the determination of income in the local economy. While developing a structural model of this process is an extremely complex task, reduced form analysis sheds some light on the subject; that is precisely what Baade and Dye and Baade do. Their results are clearly supportive of the position 1

4 that changes to the sports environment are not signicant in the determination of local incomes. This paper takes a new look at the data on local incomes and the role of sports in the determination of those incomes. We are motivated by our belief that there are some potentially signicant methodological aws in the published literature. In short, the specication of the estimating equation is not well designed for answering the questions put to it. We enumerate our criticisms below. Additionally, the existing literature does not do a good job of recognizing the distinct questions noted above. Hence, even with correct methodology, the existing literature is incomplete. The analysis here shows that whether the local economy experiences benecial eects, no eects, or detrimental eects from a new franchise or a new stadium is highly sensitive to the specication of the model. In other words, empirical work on the eects of franchises and stadium construction on local economies is of dubious value. The questionable value of these studies derives from the lack of theoretical modeling of the relationship of the sports environment to the macroeconomic aggregates of interest. The estimated equations are all reduced forms, the interpretation of the coecients from which is problematic at best and misleading at worst. Until models are developed of the regional economy, with the role of the sports environment explicit, and until it is settled which of the two questions from above is being asked, these studies will have little more than public relations value. Section 2 reviews relevant literature on the eects of stadia and franchises on economic growth and development and provides a discussion of the econometric approach to estimating these eects. Section 3 criticizes the existing empirical work and oers alternative specications. We do that to make the contrast between our analysis and the literature as stark as possible. Section 4 discusses the data and the results. Section 5 is a conclusion. 2 The Literature on Sports Franchises and Stadia as Engines of Economic Growth The literature on the role of sports in fostering economic growth and development has two very distinct branches. The rst branch consists of economic impact analyses and case studies to assess the value of a new franchise or a stadium complex to the economic vitality of a city or region. The analysis in this branch is predominantly predictive, that is, it forecasts the aects on the local economy of the arrival of a new franchise or construction of a new stadium or arena, or the consequences of the departure of an existing franchise. 1 Occasionally a decision to provide aid to a sports franchise by local government is examined using costs and benets analysis. The second 1 A related literature examines the eects of large sporting events, such as the Olympic games or World Cup Soccer tournament, on the regional economy. Steiner and Thoni (1996) is an example of this literature. 2

5 branch of this literature, to which this paper belongs, uses cross-section or time-series cross-section data on the economies of cities, regions, or metropolitan areas in an ex post evaluation of the impact of the sports environment on incomes in the area. This branch of the literature focuses on three primary questions. First, does the existence of sports franchises and stadia inuence the trend growth path of the local economy? Phrased dierently, is the growth trend higher in cities with more franchises or newer stadia than in cites with fewer franchises and older venues? Second, do changes in the sports environment induce signicant, if short lived, moves away from trend? Finally, is it eective to use a new stadium as the centerpiece of an urban economic development strategy? 2 In this paper we attempt to answer each of these questions. Crompton (1995) reviews the use of economic impact analyses of sports franchises, stadia and events, the rst branch of the literature mentioned above. He suggests that much of current practice in this regard is mistaken either because it has always been done incorrectly or because those commissioning the studies expect the results to favor the construction of stadia, the holding/hosting of some event or the attraction of a franchise. He describes in detail common errors in the methodology, ranging from the use of inappropriate multipliers to ignoring the substitutability of sports attendance for other expenditures in the budgets of consumers and state or local governments. He does not, like some, argue that economic impact studies are useless, however. He contends that the limitations and misuses of these studies should be made clear to decision-makers, and that correct unbiased studies can be of great help. Mark Rosentraub, with a variety of coauthors, has extensively evaluated the use of sports as a development strategy. Rosentraub and Swindell (1991) examined the decision of Fort Wayne, Indiana to support, in a limited way, the development of a new stadium for a minor league baseball team. The analysis is careful to account for costs and benets of the stadium plan. The authors conclude that the local government correctly oered limited nancial backing for the plan. The failure of Fort Wayne to provide greater subsidies in part explains the owners decision not to relocate to Fort Wayne. Rosentraub, et. al. (1994) assesses the Indianapolis sports led development strategy of the 1970s and 1980s. Many cities have argued that a new stadium was part of a growth and development plan, but Indianapolis is the one city that had a well-articulated strategy along these lines. The authors compare the changes in employment and payrolls between 1977 and 1983 and 1983 to 1989 for Indianapolis with the same variables for cities that ocials from Indianapolis described as 2 Thomas V. Chema (1996) criticizes the Baade style analysis on these grounds. He argues that the aects of suburban stadia will dier from those of the newer stadia that are integrated into the urban growth and renewal plans. 3

6 their competitors. The evidence is that there was some job growth, especially in the service sector, that could be attributed to the sports led strategy. Sports-related jobs increased as a share of all employment by.03%. Given the small size of sports employment (.29% of all employment) this increase is inconsequential. The growth in payrolls rose by about one-quarter of one percent inthe sports related employments. Growth in sports related employment was positively and signicantly correlated with growth in service employment, which includes restaurants, and with hotel and lodging employment. Comparisons with the other cities were less favorable to the sports led strategy. Indianapolis's strategy did not result in more growth than was experienced by other Midwestern communities and did not lead to a concentration of higher paying jobs in the region. Robert Baade and Richard Dye (1990) and by Baade (1996) have undertaken an econometric evaluation of the economic impact of stadia and franchises. Baade and Dye (1990) estimate two empirical models. In the rst model, the real aggregate personal income in an SMSA is explained by population, a time trend, and dummy variables distinguishing years prior to construction or renovation of a stadium from years after renovation or construction, distinguishing years in which a football team is not present from years when one is, and distinguishing years when a baseball team is not present from years in which one is. Their second model explains the SMSA's share of the region's income using the SMSA's share of regional population, along with the trend and dummy variables from their rst model. These analyses have not been supportive of sports or stadium led development strategies. Baade and Dye nd that stadia and new franchises have little discernible aect on the income level of an SMSA - one exception is Seattle, where the Seahawks, a football team, and the Kingdome stadium arrived simultaneously; here the impact of the stadium was positive. Unfortunately, the methodology does not allow for separation of the impact of the stadium from the impact of the arrival of the team. For the other cities, the results indicate that the aects of stadia and franchises on the SMSA's share of regional income are mixed. For all SMSA's except Seattle, the aect of stadia and franchises is signicant and negative; for Seattle it is signicant and positive. In no case is the aect of a baseball franchise distinguishable from zero. 3 Rethinking the Empirical Framework 3.1 Critique of existing empirical models In this section we lay out several criticisms of the existing empirical work on the eects of stadia and franchises on the economic health of a city or region. These criticisms are of three types: either the models are criticized on grounds of economic theory or econometric technique or the variable 4

7 specications are criticized. In terms of the econometric technique and underlying economic theory, Robert Baade (1996) estimates the impact of sports franchises and stadia on the economic growth of metropolitan areas using an equation designed to evaluate the extent to which a change in the character of a city's professional sports industry altered a city's real per capita income growth relative to other cities in which no change in the professional sports industry occurred. His empirical model is: KP j=1 y j;t y i;t, K!, y i;t,1, KP j=1 y j;t,1 K! = 0 + 1NT i;t + 2NS i;t + e t (1) where y i;t is real per-capita income in city i at time t, NS it is the number of stadiums less than 10 years old in city i at time t, NT it is the number of professional sports franchises in city i at time t and e t is a stochastic error term. For purposes of exposition, (1) can be rewritten as: (y i;t, y i;t,1), KP j=1 y j;t KP, j=1 y j;t,1 K K! = 0+1NT i;t + 2NS i;t + e t (2) This model has several shortcomings. First, the dependent variable is not the dierence between a given city's growth rate and that of the average city, as Baade's discussion suggests, but is instead the absolute dierence between the city i's year to year change in the level of per capita income and the average year to year change in per capita income over all cities in the sample. The variable needs to be scaled appropriately to measure dierences in growth rates. And growth rates are the variable of interest, because they control for the initial level of per capita income. For example, an identical year to year increase in real per capita income in two cities could signal a dramatic improvement in standards of living in one where the level was very low to begin with, and dismal performance in the other where income per capita was already quite high. Baade's specication of the dependent variable cannot distinguish these cases. Moreover, Baade's discussion of and justication for the inclusion of the second term on the left hand side of (2) is unclear. It seems that he means for this procedure to assure that the impact of the sports environment doesn' t net out, that is, isn't removed by the focus on income growth. The motivation of the model as accounting for the dierences induced by changes in the sports environment is especially important because this formulation is very similar to that which would arise using an event study methodology. In fact, Baade's independent variables capture the eect of dierent levels in the sports environment rather than the eects of changes in the sports 5

8 environment on per capita income. In other words, Baade's approach is not correctly done as an event study because it does not identify the important events, entrance or exit of franchises or construction of a new stadium or arena. Additionally, the approach imposes restrictions on parameters of the model that should be tested; in this case the restriction is that the average level of growth contributes to the overall growth in a city with a weight of one. In the model section which follows we describe the event study technique as well as a second model we employ to assess the eects of the sports environment on the local economy. Second, in the Baade (1996) empirical model - equation (1) above - and the Baade and Dye (1990) empirical model - a closely related empirical specication with the level of real income in city i as the dependent variable and both the population of city i and a deterministic time trend as independent variables - the coecient estimates are likely to suer from either omitted variable bias or measurement error bias. Older cities in the northeast had very poor economic growth during the 1960's, 1970's and 1980's. This period saw substantial migration of business activity to the Sunbelt. These same cities in the northeast tend to have older stadia (and more of them) or to have built new stadia in the suburbs during this period and to have more franchises. Consequently, the variables used to estimate the impact of stadia and franchises potentially capture the eects of the decline of the rust-belt, general urban decline, ight to the suburbs and so on, as well as the intended eects. The coecient on the number of teams would tend to be biased down, in this case. The use of a linear time trend in the Baade and Dye (1990) empirical model induces measurement error in the explanatory variables. The time trend captures a systematic eect which grows at a constant rate over time. It cannot capture business cycle eects which by denition will vary over time with the cycle. These deviations from trend get put into the regression equation error. If the business cycle eects are correlated with any of the explanatory variables, then the coecients on those variables are biased. Third, Thomas V. Chema's (1996) criticism suggests that one should look for a dierence in the impact of stadia developed in the late 1960's and 1970's, which were specically designed to be apart from the city, from those of more recent stadia such as Camden Yards and Jacobs Field which were designed to enhance the central city. One should look also for a dierence between stadia built in the urban core and those built in the suburbs. The literature does neither. Finally, variable denitions are not necessarily well designed to address the important questions. We oer several examples. First, the analysis in Baade and Dye (1990) imposes a constant impact over time of existence of a franchise in a single sport. It is certainly possible that the aects of a franchise might dissipate over time or even grow as the fan base contracts or expands. Second, existing work does not allow for interactions between franchises from dierent sports or multiple 6

9 franchises from the same sport in a given city. Third, models do not distinguish between cities with existing stadia when a new stadium is constructed and cities without stadia. Fourth, there is no control for single sport versus multiple sport stadia. Fifth, in Baade and Dye (1990) there is little variation in the football and baseball dummies, either because an SMSA had no team in either sport throughout the sample or because it had one of each during the entire period. They estimate the model for each of 9 SMSAs separately and all together, yet four of the SMSAs shownovariation in these variables. Finally, the construction of the explanatory variables imposes equal impacts on both stadium construction and renovation. This is a limitation of the data, but something that can, in principle, be tested. 3.2 Addressing some problems in existing empirical models In this section we describe alternative specications of the relationship between the level and growth rate of real per capita income in a metropolitan area and changes in the number of stadia and baseball, football and basketball franchises. These specications account for the criticisms of the existing models made above and provide evidence on the (lack of) robustness of the results in the literature. We begin with a simple (reduced form) equation which relates the level of economic activity in the metropolitan area in a given year, y i t,toavector of variables describing the economic and business climate in that area during that year, x 3 it, and to a vector of variables which capture the role of stadia and franchises in the determination of economic activity, z it. y it = x it + it + it (3) and are vectors of parameters to be estimated and it is a disturbance term. As an alternative approach to the single equation estimation used in the literature, where (1) or similar equations are estimated separately for each city, we add structure to the disturbance term. For example, let it = e it + v i + u t, where v i is a disturbance specic to SMSA i which persists throughout the sample period, u t is a time t specic disturbance which aects all areas in the same way, and e it is a random shock in jurisdiction i at time t which is uncorrelated across jurisdictions and over time. Estimated this way, the regression purges the eect of national events on each jurisdiction in a given year and generates an SMSA specic impact. In other words, the level of income at any point in time is determined by time and location specic events and the circumstances regarding sport franchises and stadia. 4 3 Among the x it there may bevariables that do not vary over time or across jurisdictions. 4 On the other hand, no variables can be included in the equation which do not vary across SMSA's and across 7

10 In our analysis below x it is limited to SMSA specic time trends or year specic eects and SMSA specic eects. 5 Our vector of sports environment variables is far richer than in that used by Baade and Dye (1990) and Baade (1996). We construct dummy variables indicating the presence of a rst and second football and basketball franchise and a baseball franchise in each of the 37 SMSAs in our sample, the periods following all football, basketball and baseball franchise entries and exits, and the presence of both single-sport and multiple use stadia. We also include that seating capacity of all football, basketball and baseball stadia. Additionally, we include more cities than Baade and Dye (1990) and a longer time period. We also dene the dependent variable to be real per capita income, dropping the population variable as a regressor while still controlling for the size of the SMSA. The second estimation technique we employ is the event study. In the nance and regulation literatures the event study is a common method of addressing questions of the impact of changes in the law or regulations on the value of rms. Stock market information is gathered to track the daily return on stocks in some specic industry and the market return. Dummy variables are constructed for dierent events, say announcements by regulators, passage of legislation, or some other shock to the system. A regression model is estimated in which the deviation of the return on the chosen stock from the market return are explained by the events or announcements. A similar situation is possible here. Suppose that the level of income or the growth of income in a city or metropolitan area is explained by the average level across cities or the average rate of growth across cities plus dummy variables for certain events, say construction and opening of a new stadium, or arrival or departure of a franchise. Statistical signicance of one of these dummy variables indicates that this event explains some of the deviation from the average. The formal event study model is: 3X g it = + g t + k D kit + e it (4) k=1 where g it is either the level of real per capita income or the rate of growth in real per capita income in jurisdiction i at time t, g t is the average level of income or of the growth rate at time t, D kit is a dummy variable indicating the occurrence of event type k in metro area i during time t,, and k are parameters to be estimated and e it is a random error. If k is statistically signicant, time. For example, we cannot use regional dummy variables as regressors as a linear combination of these variables are perfectly collinear. 5 We also estimated models which use the population as an explanatory variable, much like the analysis used by Baade and Dye (1996), but those results are similar to the results in Table 3. 8

11 then events of that type inuenced the economic growth of cities; if not, then those events had no impact on city economic growth. Note that the national average level of income or rate of growth is an explanatory variable. This approach lets the data determine the value of. Baade's construction forces to equal one. Note also that what is of interest here is not the number of franchises or the number of stadia, but whether or not a city experienced a change in either of those circumstances. For example, let D1it = 1 if the ith city experienced in year t a loss of a franchise, zero else. Let D2it = 1 if that city inyear t experienced the arrival of a new franchise, zero else. Finally, D3it = 1 if in year t the ith city had a new stadium under construction or opened in the last x years. The event study methodology is employed with the SMSA specic dummy variables as additional regressors. However, because the regression includes the average level of income or the average rate of growth as a regressor, it cannot have the year specic eects. The average variable is the same for all SMSAs during a given year. Hence, inclusion of both year eects and the annual average would not allow estimation of the model because the variables are perfectly correlated. 4 Data and Results In this section we discuss the data and the results of our analysis. The data cover the period 1969 to Income and population data were taken from the Regional Economic Information System, distributed by the U. S. Department of Commerce, Bureau of Economic Analysis. Data on sports franchises and stadia comes from the Information Please Sports Almanac. Table 1 presents variable denitions and descriptive statistics; Table 2 lists the SMSAs and several descriptive statistics for each. The entry, exit and construction variables take onavalue of 1 in each of ten years, the year a franchise moves, or the year a stadium or arena opens, and the nine subsequent years. One might question the choice of this metric as ad hoc. We defend it on the basis of the length of time it takes for the novelty of a new franchise or stadium to wear o, as has been reported in this literature (Baade, 1996), or for the despair from losing a team to subside. 6 One set of entry and departure variables (BBE1, BBE2, FBE1, BAE1, BAE2, BBD1, BBD2, FBD1, BAD1, BAD2) allow for a diering eect on per capita income in each instance of an arrival or departure of a franchise; a second set of entry and departure variables (BBE, FBE, BAE, BBD FBD, BAD) combines these multiple entries and departures, implicitly forcing an equal eect on each event. A bit must be said about the denitions of the variables because they relate to the criticisms 6 Sensitivity tests could be performed on the duration of this eect, but given the observed sensitivity of the empirical results to model specication we suspect that little could be learned from these tests. 9

12 we made of variable specication in the earlier literature. Our rst criticism was that the literature imposes a time invariant eect of franchises. We address this by using both dummy variables indicating the presence of a franchise and the entrance or exit of a franchise in the last ten years. We also allow for both the existence and the entrance and exit of franchises in each of three major professional sports, thus allowing for the eects of a franchise in one sport to be net of the eects of goings on with other sports or other franchises in the same sport. Our specication does not, however, control for any symbiotic or mutually detrimental eects of franchises in more than one sport. We control for construction of new facilities with dummy variables and, combined with the presence of a franchise, which must have had an existing facility, this addresses the criticism of new stadia replacing old stadia versus new stadia where none previously existed. One of the construction variables controls for multiple-sport facilities, as was common in the 1970's. The wide variety of our explanatory variables controls for the gamut of sports environments experienced in the United States. Because we examine the eects of entrance and exit of franchises over a ten year period, few SMSAs have no variation in the explanatory variables. For example, a city which obtained its rst football franchise in 1965 has a value of 1 for FBE1 for 1969 through 1974, and zero thereafter. Table 3 shows the results of estimating equation (1) using a random eects technique; this examines the eects of the sports environment on real income per capita in the SMSAs. 7 Our analysis diers from that in the literature because of the richer set of explanatory variables capturing both the sports environment and non- sports environment. The left panel of the table reports results for the single entry and exit variables model, the right panel contains the multiple exit and entry model. In both cases, the year specic intercepts and SMSA specic time trends are omitted. Each model was estimated with both xed and random eects for each SMSA. A Hausman test indicates that the random eects model is preferred. Our discussion of the results will focus on the single entry and exit variables model because F-tests favor it over either the multiple entry and exit model or a model with no sports variables included. Note that this last statement indicates that even after accounting for lagged real per capita income, population change, city specic time trends and year specic uctuations, real income per capita is inuenced by the sports environment in an SMSA. The fact that the sports environment is found to matter for real income per capita may not be great news for advocates of sport led development however. Among the sports environment variables only four, baseball stadium capacity and capacity squared, basketball arena construction 7 We estimated models like those of Baade and Dye (1990) where population appears on the right-hand side and the dependent variable is the level of real income; these results are available on request. The basic results of these regressions are that the SMSA specic eects are statistically signicant, the entry and exit variables are generally statistically signicant and dier by sport. 10

13 and basketball team entrance are individually signicant at the 5% level. No franchise variable is signicant and the closest of these to signicance, baseball, carries a negative coecient indicating the presence of the franchise costs the SMSA almost $400 per person per year in real income. The entrance of a basketball franchise carries with it a rise in real per capita income, about $67 per person. 8 Additionally, an increase in the capacity of a baseball stadium, in an SMSA with a baseball franchise, is associated with a rise in real income per capita in the SMSA, though the size of that eect is rather modest. For example, at the mean stadium capacity the additional real income per person of an increase in capacity by 1000 is only about $9.40. Stadium capacity for football and basketball have t-statistics less than.5 in absolute value, clearly indicating that these capacity eects may be ignored. Advocates of new stadia and arenas often argue that these will stimulate the local economy and pay for themselves via multiplier eects. Three of the four construction variables in our analysis have negative coecients and each of them has a larger t-statistic than the lone variable with a positive sign. Basketball construction is signicant at the 5% level and indicates that each person loses almost $73 in each of the ten years subsequent to the construction of the arena. Note that combined with the $67 gain from entrance of a basketball franchise, this variable indicates a net loss of about $6 per person to an SMSA that constructs a new venue for the express purpose of attracting a basketball franchise. Finally, we note the important roles of the lagged real per capita income (t-statistic of 70), the proportionate change in the population of the SMSA (t-statistic of 2.5) reported in the table and the year specic intercepts which are not reported, all but four of which have t-statistics larger than 2 in absolute value. Among the city specic time trends, three are signicant at the 5% level (Los Angeles, San Francisco, and Washington, DC) and one is signicant at the 10% level (San Diego). Seven others have t-statistics over 1, indicating that they increase the adjusted R squared. The bottom line of this discussion is simply that the model does a good job of controlling for the variation in the real per capita income that is not attributable to the sports environment. Table 4 reports the eects of stadium construction, and entrance or exit of franchises in an event study framework. It is important to recognize that the event study methodology rules out the use of the year specic intercepts and the city specic time trend variables; these would be perfectly collinear with the annual average real per capita income (or growth rate in the latter analysis). It is clear from the table that the SMSA specic eects, which are not reported but are available on 8 Russel Sobel suggests that this is an over estimate of the impact of the entrance of a franchise because the players and coaches that move into the SMSA will have much higher than the average income and be very few in number, tending to raise per capita income without increasing income for the average resident. 11

14 request, are jointly signicant. The adjusted R squared rises from.64 to.95 after their inclusion. Most of these eects have t-statistics over one in absolute value, and more than half have t-stats over 2. In this case, F-tests reject the single entry and exit eects for the multiple entry and exit eects model. Consequently, the following discussion focusses on the results in the last column of Table 4. The sports environment variables tell an interesting story. Two of the four construction variables are statistically signicant, one at the 5% level the other at the 10% level, and negative; three of four have negative signs. Construction of a football only stadium reduces per capita income by $153, construction of a baseball only stadium reduces it by $240. The capacity of a baseball stadium has a signicant (at the 10% level) and positive eect on real per capita income, raising it by $56 for each increment of 1000 in stadium capacity. No other capacity variable is close to signicant at conventional levels. Among the franchise variables, only the baseball variable is signicant. According to this result, the presence of a baseball franchise reduces per capita real income in an SMSA by more than $2860. Even counteracting this with the stadium capacity eects, a baseball franchise playing in the average size stadium costs the SMSA more than $850 per person per year. The entrance and exit of franchises also is of little consolation to proponents of sport led development. Among the 11 entrance and exit variables, ve are statistically signicant at the 5% level. Among these, only the entrance of the rst football franchise, and the departure of the second baseball and second basketball franchises have signs favorable to sports as a development tool. The rst football team to enter the SMSA raises per capita real income by $284. The departures of the second baseball or basketball franchise result in a loss in income of about $840 and $430 respectively. On the other hand, entrance of the rst baseball franchise costs the SMSA a little less than $250 per person, and its departure by slightly more than $400 per person. Careful readers will note that the results between Table 3 and Table 4 are quite dierent. We tend to place more trust in the results of Table 3. The reason is simply that we believe that the event study misspecies the relationship, forcing the nation-wide average level of real per capita income to carry too much of the weight. Recall that in the analysis of Table 3 we include year specic eects and city specic time trends. The event study methodology cannot include these variables because of collinearity. But some of their inuence is picked up by the included regressors. For example, the city specic time trends are intended to capture such things as the ight from the north (rust belt) to the south and west (sun belt), and urban decline. These variables would certainly be correlated with entrance and departure of franchises as entrance tends to occur where cities are doing well, departure where they are not. Additionally, few SMSAs experienced entrance or departure of two franchises from a given sport. Consequently, these variables tend to pick up 12

15 eects that are specic to one or two SMSAs rather than to some more general phenomenon. Nonetheless, the picture that one gets from this analysis reported in either Table 3 or Table 4 is far dierent from any painted by the advocates of sports led growth. Far from being engines of income increase, these results indicate that at best SMSAs get nothing from their sports franchises, at worst they pay dearly for professional athletic franchises. Interestingly, these results also dier from those in the published literature in this same style. Baade and Dye (1990) and Baade (1996) nd little or no aects, positive or negative. At this point, we turn to an examination of the eects of stadia and professional sports franchises on the growth rate of real per-capita income. By so doing, we intend to address the issue of whether sports and stadia can inuence the rate at which income rises rather than the level of real income in a SMSA. As much of the public debate on the benets and costs of sports and stadia seems to focus on issues pertaining to economic growth, this seems to be the best direction for research in this area. Table 5 reports results of the random eects estimation of the eect of our sports environment variables on the growth rate of real income per capita. The most important information from this analysis is that neither the single nor the multiple entry and exit variables models is supported by the data. That is, F-tests lead to the conclusion that the sports environment variables have no eect on the rate of growth of real income per capita. Indeed, examining Table 5, one sees that few of the sports environment variables have t-statistics over 1 and none is even remotely close to signicant at conventional levels. The lagged value of the growth rate in the SMSA, and the year specic eects provide all the explanatory power in the model. Table 6 shows the results of event study regressions. Recall that there is an annual average rate of growth which is common to all SMSAs in a given year. The idea here is to determine if changes in the sports environment account for any of the discrepancy between the SMSA growth rate and the national average growth rate. As in the case of Table 5, the sports environment variables add nothing as a group to explaining the growth rate of real per capita income in an SMSA once the average growth rate in the nation is controlled for. 5 Conclusions This paper has revisited the issue of metropolitan areas using professional sports as a development strategy. Wehave examined the literature which empirically tests for the inuence of sports and stadia on both the level and the growth rate of real income per capita, nding that literature suspect on methodological grounds. Our approach has been to respecify the relationship between the sports environment and the dependent variable of interest in two ways. First, we propose 13

16 dierent functional forms for the relationship. Second, we redene the independent variables to more accurately address the policy questions. Our empirical results suggest two important conclusions. First, the sports environment does appear to signicantly inuence the level of real income per capita in an SMSA. Unfortunately for proponents of the sports led development strategy, the general nature of this impact is negative. This is a dierent conclusion from that in the literature, which suggests no impact of the sports environment. One possible justication for this negative eect might be the way the sports environment relates to unobservable productivity shocks in an SMSA. For example, presence of a team might induce greater wastage of time as fans spend work time commiserating or celebrating the recent game or handicapping the upcoming contests. Our evidence indicates, moreover, that the size and signicance of the sports environment for the level of real income per capita depends upon the exact specication of the model. Second, the sports environment, or rather changes in that environment, have no impact whatsoever on the growth rate of real income per capita. This latter point is rather comforting. Economic theory suggests that growth in an economy is dependent on expansion of the physical and human capital stocks and on technological change. The link between these fundamentals and the sports environment is tenuous at best. 14

17 References [1] Baade, Robert A Professional Sports as Catalysts for Metropolitan Economic Development, Journal of Urban Aairs. [2] Baade, Robert A. and Richard F. Dye The Impact of Stadiums and Professional Sports on Metropolitan Area Development, Growth and Change, 21:114. [3] Chema, Thomas V When Professional Sports Justify the Subsidy, a Reply to Robert A. Baade, Journal of Urban Aairs. [4] Crompton, John L Economic Impact Analysis of Sports Facilities and Events: Eleven Sources of Misapplication, Journal of Sport Management, 9: [5] Okner, Benjamin Subsidies of Stadiums and Arenas, in Government and the Sports Business edited by Roger G. Noll. Washington, D. C.: The Brookings Institution. [6] Rosentraub, Mark S., and David Swindell Just Say No? The Economic and Political Realities of a Small City's Investment in Minor League Baseball, Economic Development Quarterly, 5(2): [7] Rosentraub, Mark S., Michael Przybylski, and Daniel R. Mullins Sport and Downtown Development Strategy: If you Build it, Will Jobs Come?, Journal of Urban Aairs, 16(3): [8] Steiner, Michael, and Erich Thoni Sport Mega-Events as an Instrument for Regional Development, presented at the 36th European Congress of the European Regional Science Association, Zurich, Switzerland. 15

18 Table 1 Variable Denitions, Means and Standard Deviations Variable Mean Std. Dev. Denition INCOME Aggregate real income, 1987 dollars POP Population in thousands YRINDEX Linear time trend, 1 in 1969 RPCI Real per capita income BBCAP Baseball Stadia capacity, thousands FBCAP Football Stadia capacity, thousands BACAP Basketball Stadia capacity, thousands Dummy Variables BAE rst basketball franchise entered, last ten years BAE second basketball franchise entered, last ten years FBE rst football franchise entered, last ten years FBE second football franchise entered, last ten years BBE rst baseball franchise entered, last ten years BBE second baseball franchise entered, last ten years BBD rst basketball franchise left, last ten years BBD second basketball franchise left, last ten years FBD football franchise left, last ten years BAD rst baseball franchise left, last ten years BAD second baseball franchise left, last ten years BBCO baseball stadium constructed, last ten years FBCO football stadium constructed, last ten years BBFB baseball / football stadium constructed, last 10 years BACO basketball arena constructed, last ten years BBF baseball franchise present FBF football franchise present BAF basketball franchise present BBE any baseball franchise entered, last 10 years BAE any basketball franchise entered, last 10 years FBE any football franchise entered, last 10 years BBD any baseball franchise left, last 10 years BAD any basketball franchise left, last 10 years FBD any football franchise left, last 10 years

19 Table 2 Mean Values Growth of Real Real Total Real Per-Capita Per-Capita Personal Population Personal Personal Personal City Income (000's) Income Income Income Atlanta Baltimore Boston Bualo Charlotte Chicago Cincinnati Cleveland Dallas Denver Detroit Green Bay Houston Indianapolis Kansas City Los Angeles Miami Milwaukee Minneapolis New Orleans New York Oakland Orange Co Orlando Philadelphia Phoenix Pittsburgh Portland Sacramento St. Louis Salt Lake City San Antonio San Diego San Francisco Seattle Tampa Washington

20 Table 3 Entry and Exit Eects Dependent Variable: Real Per-Capita Personal Income Single Entry and Exit Eects Multiple Entry and Exit Eects Variable Coecent t-stat. Coecent t-stat. C RP CP I, DPOP BBCAP FBCAP BACAP BBCAP FBCAP BACAP BAFR FBFR BBFR BBCO FBCO BBFBC BACO BBE FBE BAE BBD FBD BAD BBE BBE FBE FBE BAE BAE BBD BBD FBD BAD BAD R R

21 Table 4 Event Study Dependent Variable: Real Per-Capita Personal Income Single Entry and Exit Eects Multiple Entry and Exit Eects Common Intercept SMSA Specic Eects Common Intercept SMSA Specic Eects Variable Coecent t-stat. Coecent t-stat. Coecent t-stat. Coecent t-stat. C GRBAR DPOP BBCAP FBCAP BACAP BBCAP FBCAP BACAP BAFR FBFR BBFR BBCO FBCO BBFBC BACO BBE FBE BAE BBD FBD BAD BBE BBE FBE FBE BAE BAE BBD BBD FBD BAD BAD R R

22 Table 5 Entry and Exit Eects Dependent Variable: Growth in Real Per-Capita Personal Income Single Entry and Exit Eects Multiple Entry and Exit Eects Variable Coecent t-stat. Coecent t-stat. C GRP CP I, DPOP BBCAP FBCAP BACAP BBCAP FBCAP BACAP BAFR FBFR BBFR BBCO FBCO BBFBC BACO BBE FBE BAE BBD FBD BAD BBE BBE FBE FBE BAE BAE BBD BBD FBD BAD BAD R R