Determinants of Spectator Attendance 311 Journal of Sport Management, 2002, 16, 311-330 © 2002 Human Kinetics Publishers, Inc. Determinants of Spectator Attendance at NCAA Division II Football Contests Timothy D. DeSchriver University of Massachusetts Paul E. Jensen Drexel University The purpose of this study was to analyze the relationship between spectator attendance at NCAA Division II football contests and selected determinants by estimating multiple economic demand models. The two primary determinants analyzed were winning percentage and promotional activity. Demand models were estimated using OLS and fixed-effect regression analysis. The results suggested that both current and previous year winning percentages are positively related to attendance. Furthermore, it is shown that the effect of previous season winning on attendance diminishes while the effect of current season winning increases as the season progresses. The results also indicated that promotional activities, the number of enrolled students, and market competition significantly affected attendance. Overall, the demand models explained between 37 and 70 percent of the variation in spectator attendance. The findings of this study may aid Division II athletic administrators who are attempting to increase revenues by attracting additional spectators to smallcollege football contests. Introduction and Purpose Millions of spectators attend American intercollegiate sporting events each year. In 2000, over 39 million people attended National Collegiate Athletic Association (NCAA) football games across the four levels of competition (I-A, I-AA, II, and III) (Campbell, 2001).1 Universities such as Penn State and the University of Tennessee attract over 95,000 spectators for each home game, and ticket revenues can eclipse $3 million for a single contest. Even Division II institutions, such as North Dakota State University, average over 12,000 spectators per home contest (Campbell, 2001). While the dollar amount of revenues from football ticket sales T.D. DeSchriver is with the Department of Sport Management, University of Massachusetts, Amherst, MA, and P.E. Jensen is with the Department of Economics and International Business, Drexel University, Philadelphia, PA. 311 312 DeSchriver and Jensen are drastically lower at NCAA Division II institutions they are, in many cases, an equally important source of revenue. Unlike Division I-A institutions, Division II schools generate very little, if any, revenue from sources such as media rights fees, luxury seating, sponsorship, and advertising. Sport consumers are most familiar with Division I-A football. Over 110 institutions compete in Division I-A football, the highest level of competition. Division II football differs significantly from Division I-A. First, the maximum allowable number of grants-in-aid is considerably smaller, 35 versus 85, and competition is focused on a regional level. Division I-A competition is considered to be national in scope. Lastly, in 1999, 64 percent of Division I-A institutions reported a net profit from their football operations while only 23 percent of Division II programs reported a profit (Fulks, 2000). Nine percent of all Division II athletic revenues, excluding institutional support and student fees, are generated by ticket sales to home contests (Fulks, 2000). For these institutions, football is often the primary sport for the generation of ticket sales revenue. As funding from traditional sources such as institutional support and student fees become somewhat tenuous, there is a need for small college athletic administrators to maximize revenue from ticket sales. In addition, increased football ticket sales also lead to increased revenue from sources such as concessions, advertising, and booster donations (Coughlin & Erekson, 1985; Hilkemeyer, 1993). While many Division II athletic administrators are aware of the financial importance of ticket sales, they have very little information on the determinants of spectator attendance at football contests. The purpose of this paper was to develop economic demand models to empirically identify the determinants related to spectator attendance at NCAA Division II football games. The results will aid collegiate athletic administrators in developing marketing strategies to increase spectator attendance and the corresponding revenue. Within this paper, the analysis of those determinants that are controllable by athletic administrators will be emphasized. Athletic administrators have substantial influence over variables such as onfield success, stadium age, and promotional activities. Perhaps the most widely researched relationship has been between on-field success and spectator attendance. Within this paper, a new approach is taken to measure this relationship. Similar to earlier studies, team-winning percentage is used as a proxy for on-field success. Additionally, winning percentage is interacted with a time variable to determine how the relationship between on-field performance and spectator attendance changes over the course of a season. It is logical to assume that the effect of winning percentage will increase throughout the course of a season. A 10-0 team should attract more spectators than a 1-0, holding all else constant. The following research will add to the body of knowledge by determining the magnitude of this effect, if it exists at all. Another unique aspect of this paper is its definition of promotional activity. In most demand models of professional sport attendance, promotional activity has been represented through a single categorical variable that measures the existence of promotions but does not differentiate between them (Marcum & Greenstein, 1985; Wall & Myers, 1989). In the research of McDonald and Rascher (2000) a Determinants of Spectator Attendance 313 series of variables were used to represent promotional activity in Major League Baseball. In general, the variables used by McDonald and Rascher (2000) were related to the frequency and cost of the game-day promotional giveaways such as Bat Day and Hat Day. While these definitions were successful for Major League Baseball, they may be somewhat flawed for intercollegiate football due to the presence of Homecoming. Most American colleges and universities celebrate Homecoming during the football season. This special promotion invites alumni and friends of the university back to a campus for a day of fun and activities. While most colleges have many other promotional activities such as Parent’s Day, Youth Day, and Hall of Fame Day throughout the course of the season, few have the success of Homecoming. To account for the uniqueness of Homecoming on most university campuses, two categorical variables were used to analyze the relationship between spectator attendance, Homecoming and other promotional activities. There are several other determinants of spectator demand such as weather, and student enrollment that are not controlled by collegiate athletic administrators and in order to build an accurate demand model these determinants must also be included. By incorporating these uncontrollable determinants a greater percentage of the variation in spectator attendance may be explained and the model may be a more accurate representation of the demand for Division II football contests. Review of Literature There has been a great deal of research devoted to the study of the factors influencing sport consumption, represented by spectator attendance. Researchers such as Pease and Zhang (2001), Iso-Ahola and Hatfield (1986), Courneya and Carron (1992), and Wann, McGeorge, Dolan, and Allison (1994) have focused on the socio-motivational factors which relate to game attendance frequency by individual consumers. Their research has focused on how areas such as spectator aggression, audience facilitation, and team identification have been related to the decisions made by individuals to attend sporting events. While this line of research is important, the focus of this paper is on the determinants that affect aggregate levels of attendance at sporting events. The use of economic demand models has been the most popular method for the analysis of aggregate spectator attendance at collegiate football contests. The majority of research using demand models has been directed toward professional sports such as baseball (Kahane & Shmanske, 1997; McDonald & Rascher, 2000; Noll, 1974; Scully, 1974), football (Hansen & Gauthier, 1989; Siegfried & Hinshaw, 1979), basketball (Burdekin & Idson, 1991; Whitney, 1988; Zhang et al., 1995), soccer (Baimbridge, 1997; Dobson & Goddard, 1992; Peel & Thomas, 1988; Rivett, 1975), and ice hockey (Hansen & Gauthier, 1989; Schofield, 1983; Zhang et al., 1996). Overall, the previous research has found that a positive relationship exists between spectator attendance and the two factors that are emphasized in this paper, promotional activity and on-field success.2 In a review of the literature, few articles were found related to the determinants of spectator attendance at American collegiate sporting events. Two studies 314 DeSchriver and Jensen (Fizel & Bennett, 1989; Kaempfer & Pacey, 1986) focused on the relationship between televised sport and collegiate football attendance at the NCAA Division I-A level. Additionally, Wakefield and Sloan (1995) analyzed the effects of factors such as team loyalty and facility characteristics on the decisions made by individuals to attend college football games. Two additional studies (DeSchriver, 1999; Wells et al., 2000) analyzed the determinants affecting spectator attendance for NCAA Division II football games. However, DeSchriver (1999) analyzed only the individual relationships between attendance and selected determinants through correlation analysis. A full economic demand model was not constructed and, thus, there was no analysis of the effect of the selected determinants as a group on attendance. Wells et al. (2000) constructed a demand model but several important determinants, such as weather and the distance between the two competing institutions, were omitted from the model. In addition, their use of winning percentage as a proxy for on-field success did not attempt to incorporate the effect of how winning changes over the course of a season. Supply and Demand Model for Division II Football In order to study this market and to highlight a somewhat unique feature of this market we first develop a general model of supply and demand. The inverse demand for tickets is given by PD = f(QD, X) (1) where PD is the price of the tickets, QD is the quantity demanded, and X represents a vector of exogenous demand determinants. The marginal cost of production is the cost of admitting another consumer into the event. Thus, the supply side of the market may be MC = g(Qs, Z) (2) where MC is the marginal cost of production, QS is the quantity supplied, and Z represents a vector of exogenous cost variables. It is assumed that suppliers have some market power and that they select quantity to maximize profit ( ). Thus the first order conditions for profit maximization imply (3) Solving equation 3 for the optimal quantity yields Q* = h(X,Z) (4) The optimal price is then obtained by substituting Q* into equation 1. Equation 4 clearly shows that the equilibrium values of price and quantity are functions of X and Z. Therefore, for markets that are appropriately described by equations 1 and Determinants of Spectator Attendance 315 2 we see that P* and Q* are determined simultaneously by the interaction of supply and demand conditions. In such cases one cannot simply estimate demand by regressing observed equilibrium values of Q on P and X as this ignores the effect that Z has on the equilibrium values. This is the classic simultaneous equations problem. When thinking about a model of supply and demand for Division II football tickets, one important simplification can be made to the general model described above. The major costs associated with the event (team expenses, equipment and facility costs) are essentially independent of the number of people that attend the event and thus are fixed costs. Therefore, for all practical purposes the marginal cost of supplying tickets for a game is zero.3 Given this, equation 2 can be replaced with the following MC = 0 (2b) Solving the firm’s profits maximization problem given equations 1 and 2b yields (3b) Solving this for Q yields Q* = h(X) (4b) The critical point to make is that in this model P* and Q* are completely demand determined. The intuition for this result is that with zero marginal costs, profit maximization reduces to revenue maximization, which is independent of supply conditions. Therefore, under these conditions there is no simultaneous equations problem and one may estimate demand by regressing Q on P and X. Econometric Model An econometric demand model was constructed to analyze the effect of selected determinants on game attendance. The determinants were selected based on a review of pertinent literature. The literature separated the determinants into four general categories (Cairns et al., 1986). The following functional form represents these groupings. Attendance = f (economic determinants, demographic determinants, game attractiveness determinants, residual preference determinants) To accurately analyze these relationships, data were collected for three seasons of play (1994, 1996, and 1999). From the data, four demand models were estimated. Individual models were estimated for each of the three seasons and a fourth model was estimated for the three years of data pooled together. This allowed for the relationship between attendance and the determinants to be confirmed by multiple demand models. In addition, the use of multiple years of data allowed for the analysis of changes in spectator attendance across years, holding all else 316 DeSchriver and Jensen constant. The following is a list of definitions for the dependent variable, spectator attendance, and the 13 determinants. Given that some of these determinants were measured through a series of explanatory variables, there were 21 explanatory variables included in the single year models. With the inclusion of two categorical variables for time, 23 explanatory variables were included in the multi-year model. The following is a list of the variables including in the econometric model. Complete descriptions of each variable are located in appendix A. Dependent Variable ATTEND – Game Attendance Number of spectators who attended an individual football game. Economic Determinants GAP - General Admission Ticket Price The price of a general admission ticket to a football contest.4 SAP – Student Admission Policy 1 if enrolled students are admitted free of charge, and 0 otherwise. COMP – Number of Competitors in the Market Area Number of competitors within 50 miles of institution. Demographic Determinants MILES – Miles Between Competing Institutions Distance, in miles, between the cities in which the two competing teams were located. STUDENR – Student Enrollment The number of full-time enrolled students. CITYPOP – City Population The number of residents living in the city where the home team was located.5 Game Attractiveness Determinants WINPER – Home Team Winning Percentage The interaction of winning percentage and the time period variables. PREVWIN – Home Team Winning Percentage in Previous Season The interaction of winning percentage in the previous season and the time period variables. HCPROM/OTHPROM - Promotional Activity Two categorical variables were used to represent promotional activity. One variable accounted for the presence of Homecoming (HCPROM) and the second variable accounted for all other promotional activities (OTHPROM). Residual Preference Determinants WEAT – Weather A categorical variable accounting for the presence of rain or snow. Determinants of Spectator Attendance 317 STADAGE – Age of Stadium The age of the stadium, in years. TP – Time Period The seasons were divided into four three-week time periods. Three categorical variables where used to indicate the time period in which a game was played. Year – Year of Season Two categorical variables were used to indicate the year in which the game was played (1994, 1996, or 1999). Lastly, the issue of stadium capacity as it relates to attendance must be addressed. Some studies (Noll, 1974) have measured attendance as a percentage of stadium capacity. This specification is accurate when measuring demand for individual contests where stadium capacity may limit attendance. This is true of sports where sellouts occur such as the National Football League and the National Hockey League. In these cases, the failure to account for capacity constraints can result in the regression models being misspecified. However, empirical and anecdotal evidence strongly suggests that NCAA Division II football games had a very high level of excess capacity. None of the survey respondents noted that their entire supply of tickets were sold for any games. Therefore, similar to McDonald & Rascher (2000), it was determined that spectator attendance was the most appropriate measure of demand given that stadium capacity was not a binding constraint. Data Analysis The primary method of data collection was the distribution of a survey to the sports information/media relations offices at all NCAA Division II institutions that competed in football. The survey was distributed following the 1994, 1996, and 1999 seasons. In addition, secondary sources such as the US Higher Education Directory, US Census data, and TRIPMAKER computer software were used to obtain data for student enrollment, city population, and miles between competing institutions. Table 1 provides a summary of the respondent’s information for each year, Table 1 Respondent Summary Data Year Number of institutions responding Number of total games played 1994 1996 1999 Total 96 82 82 260 476 411 415 1302 Note: In each of the three years, 147 teams competed in NCAA Division II football. 318 Table 2a DeSchriver and Jensen Summary Statistics Variable Spectator attendance Ticket price (GAP) Student enrollment (STUDENR) City population (CITYPOP) Miles between institutions (MILES) Home team winning %age Home team winning %age in previous season Table 2b Mean SD Minimum Maximum 3,796.5 $5.73 6,034.9 62,424.4 297.7 .496 .524 3,504.4 $1.82 4,589.5 129,208.0 255.9 .364 .245 200 $2 624 1,287 2 0.000 0.000 28,820 $21 35,319 999,999 2,947 1.000 1.000 Bivariate Correlation Matrix ATT Attendance City pop. Stud enr. Ticket price HT Win HT Win %age Miles %age (Prev. sea.) 1.00 City population .13 1.00 Student enrollment .17 .29 1.00 Ticket price .39 .15 –.02 1.00 Miles between institutions .06 .03 .14 –.00 1.00 Home team winning %age .14 –.04 .07 .01 –.02 1.00 HT winning %age in prev. season .21 –.09 .20 –.00 .06 .34 1.00 and Tables 2a and 2b show the summary statistics and bivariate correlations for the key variables. Ordinary least squares (OLS) regression analysis was initially used to estimate the model for each of the three years of data. The three years of data were also combined and the model was estimated for the entire set of 1302 games played. It should be noted that a White correction was incorporated into the OLS models in order to correct for heteroscedasticity. Determinants of Spectator Attendance 319 Additionally, the data were transformed into logarithmic functional forms when appropriate. Determinants that took the form of categorical variables were not transformed. The transformation allowed for the analysis of the relationship between attendance and the determinants to be in the form of demand elasticities. Given the definitions of the determinants and the logarithmic transformation, the regression equation for each individual year took the following form: log(ATTEND) = 0 + 1[log(GAP)] + 2[log(CITYPOP)] + 3[log(MILES)] + [log(STUDENR)] + 5[log(STADAGE)] + 6[log(COMP)] + 7[log(SAP)] + 4 (WEAT) + 9(HCPROM) + 10(OTHPROM) + 11(TP1) + 12(TP2) + 13(TP3) + 8 [log( WINPER1)] + 15 [log( WINPER 2)] + 16 [log( WINPER3)] + 14 [log( WINPER4)] + 18 [log( PREVWIN1)] + 19 [log(PREVWIN 2)] + 17 [log(PREVWIN3)] + 21[log(PREVWIN4)] + ε 20 where ε is the model error term. The regression equation for the combined data set took a similar form with the addition of two categorical variables representing the year in which the game was played. Findings and Discussion Four demand equations were estimated using multiple regression analysis; one for each year and one for the combined data set. Table 3 summarizes the statistical results from the regression equations. Eleven of the 23 explanatory variables were statistically significant at the .05 level in all of the regression equations. The range of R2’s for the equations was from .368 to .483. Several of the controllable determinants of spectator attendance were statistically significant. It was expected that current season-winning percentage would have little, if any, significance in the early part of the season. But, as the season progressed, winning percentage would increase in significance. For the most part, the statistical results supported this hypothesis. Note that in the four equations that were estimated, current season winning percentage is never significant in the first three weeks of the season but is always significant in the last three weeks of the season. Also note that in general, the magnitude of the coefficients increases as the season progresses. For instance, in the combined case the coefficient increases from .08 in the first time period to .95 in the last time period. From the statistical findings, it appears that on-field success in the current year is positively related to attendance and that the magnitude of this relationship increases as a season progresses. Obviously, all college athletic administrators want to have on-field success. The findings suggest that on-field success has a direct relationship to attendance and, therefore, also has an economic impact on the athletic program. Additional spectators result in additional revenues from ticket sales, concessions, and merchandise. Administrators may wish to invest more resources into areas such as grants-in-aid, recruiting, coaches’ salaries, and equipment if the result is more onfield success. However, it must be noted that this additional investment comes at a cost. Administrators should only invest in these additional resources if they believe that the expected increase in revenue for the athletic department due to winning will offset the additional costs. 320 Table 3 DeSchriver and Jensen Regression Results: Attendance at NCAA Division II Football Games Regression coefficients Variable 1994 1996 1999 Constant 2.712 4.168 5.549 GAP .490** (3.887) Combined 4.309 1.438** (11.870) .814** (5.426) 1.033** (13.001) CITYPOP –.010 (.356) .020 (.710) .032 (1.212) .009 (.563) MILES –.012 (.341) –.184** (4.022) –.102* (2.420) –.101** (4.216) STUDENR .220** (4.593) .156** (3.215) .164** (3.885) .183** (6.637) STADAGE .004 (.124) –.087** (2.950) –.046* (2.333) –.047** (3.025) COMP –.133** (2.996) –.132* (2.537) –.361** (7.821) –.194** (6.803) SAP .561** (5.082) .771** (4.157) –.386** (3.388) .330** (3.970) WEAT .318** (2.940) .507** (4.479) .445** (2.999) .427** (6.116) HCPROM .634** (7.040) .689** (6.875) .586** (6.682) .665** (12.136) OTHPROM .290** (3.889) .284** (3.459) .278** (3.899) .296** (6.711) TP1 .337** (2.827) .769** (6.242) .257* (2.236) .441** (6.640) TP2 .310** (3.655) .650** (5.500) .226** (2.777) .349** (6.468) TP3 .153 (1.711) .263* (2.104) .058 (.648) .097 (1.678) WINPER1 .056 (.326) –.075 (.538) .071 (.424) .083 (.911) WINPER2 .238 (1.699) .418** (2.583) .268* (1.495) .314** (3.349) WINPER3 .835** (3.555) .364 (1.761) –.061 (.487) .218* (1.114) Determinants of Spectator Attendance 321 WINPER4 .954** (3.188) 1.213** (2.723) 1.018** (4.277) 956** (5.424) PREVWIN1 .676* (1.918) 1.160** (4.433) .234 (.574) .583* (1.912) PREVWIN2 .292 (1.471) .532* (2.146) –.158 (.611) .264 (1.558) PREVWIN3 .212 (.775) .373 (1.353) .828** (3.455) .577** (3.284) PREVWIN4 –.096 (.294) .332 (.785) –.030 (.160) .063 (.421) Year (1999) –.132** (2.732) Year (1996) .149 (1.728) n (sample size) R2 F–Statistic 476 .483 20.53** 411 .45 21.01** 415 .368 12.34** 1302 .386 39.54*** Note. Dependent Variable = Log(Attendance); *significant at 5% level, **significant at 1% level (absolute values of t–statistics are in parentheses) The winning percentage of the home team in the previous season was also included in the demand model. It was expected that this determinant would be significant in the early portion of the season and that the relationship would weaken as the season progressed. Early in the season, spectators may use the team’s performance in the previous season as a measure of on-field success because the team has played few, if any, games, during the current season. Again, the regression results were consistent with our expectations. In three of the four cases considered, previous year winning percentage was significant in the first period of the season while in all four cases previous year winning percentage was insignificant in the last period of the season. In addition, the magnitude of the coefficients tended to decrease as the season progressed suggesting that the impact of the previous season’s performance diminished. For example, based on the combined results, an increase in previous-season winning percentage of 10 percent resulted in an expected increase in attendance of 5.8 percent for games played in the first time period but only a .6 percent increase in the last period of the season. Both of the explanatory variables representing promotional activities were statistically significant and positively related to attendance. The Homecoming and other promotion variables were significant at the .01 level for each of the four regression equations. The presence of Homecoming increased attendance by 322 DeSchriver and Jensen approximately twice as much as the presence of other types of promotions. The statistical findings support the claim that promotional activities are significant and positively related to attendance. However, it must be noted that only 40.4 percent of the non-Homecoming games had some type of promotional activity. Based on these findings, it is recommended that Division II football marketers investigate the initiation of special activities such as Parents’ Day, Hall of Fame Day, and other types of promotions. But, these promotions must be analyzed from an economic viewpoint. The additional cost of these activities to the athletic department should not be greater than the additional revenue that can be generated from the increase in ticket sales, concessions, and merchandise sales. Stadium age and the number of miles between the two competing teams were also found to be significant for three of the four demand equations, 1994 being the exception in both instances. As expected, the age of the facility and the distance between schools were both negatively related to attendance. The impact of stadium age was relatively small as a 10 percent increase in age only reduced attendance by .47 percent in the combined data set. The effect of distance traveled was approximately twice as large as the impact of stadium age. For example, a 10 percent increase in distance traveled, decreased attendance by approximately 1 percent. This finding may be important with respect to the development of game schedules and/or the selection of conference affiliation. Efforts should be made to schedule non-conference games against teams that are close in proximity. Also, there may be a situation in which an institution must make a decision on conference affiliation. While many factors may go into this decision, the effect on attendance for football games would likely be one of the criteria. The findings here suggest that a conference with teams in close geographic proximity will positively influence football attendance. Two additional determinants were significantly related to attendance in the regression equations. Student enrollment and weather were both positively related to spectator attendance. For every 10 percent increase in student enrollment, there was a corresponding increase in attendance of approximately 2 percent. Thus, holding all else constant, the largest Division II institutions had the highest levels of attendance. Also, given the definition of the weather variable, the presence of rain or snow for a game had a negative relationship with attendance. Obviously, college athletic administrators have no control over enrollment size or weather but it is important for them to recognize that these variables influence attendance. Two institutions may have a similar profile for all of the model variables, but if they have different enrollments, the administrators should expect a difference in football attendance. Time was another determinant that was included in the demand model. This determinant was measured in two ways. First, the time within a season was captured through three categorical variables to determine if attendance changed throughout the course of a single season. Secondly, given that data was obtained for three different seasons, two categorical variables were used in the combined data set to determine if there were any changes in attendance across seasons. The statistical findings show that the higher levels of attendance occurred in the first Determinants of Spectator Attendance 323 two time periods, or six weeks, of the season. All of the regression coefficients for the first two time periods were significant in the four demand equations suggesting that attendance tends to decrease as the season progresses, holding all else constant. This finding may occur because as the season progresses, the weather becomes considerably colder in many regions of the country. As a result, some consumers may opt to stay at home and consume football via media outlets such as radio and television. One of the two season variables was statistically significant. The categorical variable for 1999 was significant and negatively related to attendance. Thus, holding all else constant, the 1994 and 1996 seasons had higher levels of attendance than 1999. College football administrators should be somewhat concerned with this finding. However, it should be stressed that this data set is limited to three years. While the findings suggest that attendance in 1999 was below the 1996 and 1994 levels, this may not signal a long-term trend. Additional years of data are needed to investigate this possibility. With that said, there may be some changes in the collegiate football market that have led to this finding. The late 1990’s have seen increases in both sports television networks and the number of televised games. These findings may suggest that consumers are watching televised collegiate games instead of attending Division II games. Ticket pricing policy was another determinant that was included in the demand model. As stated earlier, ticket policy was represented through two variables; a categorical variable indicating whether or not students are admitted free and a second variable measuring the dollar amount charged for general admission. As expected free admission for students had a positive and significant effect. Surprisingly, the general admission price variable was also found to be positive and significant at the .001 level for all of the regression equations. At first glance this positive relationship between price and attendance may seem strange, but in this case it is actually quite reasonable. First, note that maximizing profits with zero marginal costs is equivalent to revenue maximization. In order to maximize revenue, all schools would simply set ticket prices such that the equilibrium values of price and quantity correspond to the mid point of the demand curve they face. Second, also note that most of the variation in our data set is cross-sectional variation as the data set has a large cross section, about 140 schools, while the time series dimension is relatively small with 3 years of data. Thus, the positive relationship between price and quantity does not suggest that each school faces an upward sloping demand curve, but rather it is a result of the nature of the data set and the behavior of the schools. For simplicity sake imagine that all the school’s individual demand curves are parallel. Larger schools have demand curves that are shifted out to the right and small schools have demand curves that are shifted in to the left. If each school selects price at the mid point of their demand curve, the cross section of equilibrium prices and quantities traces out a line with a positive slope. Therefore, the positive relationship between price and quantity is not problematic, but it does suggest that differences across schools are important. The next step in the analysis was to investigate the significance of these differences through the use of a fixed effects model. 324 Table 4 DeSchriver and Jensen Fixed Effects Regression Model Robust Coefficient Standard Error Constant GAP CITYPOP MILES STUDENR STADAGE COMP SAP WEAT HCPROM OTHPROM TP1 TP2 TP3 WINPER1 WINPER2 WINPER3 WINPER4 PREVWIN1 PREVWIN2 PREVWIN3 PREVWIN4 Year (1999) Year (1996) 9.011 .238 .022 –.124** –.316 .015 –.013 –.124 .486** .654** .209** .424** .330** .047 .011 .112 .426** .824** .348 .342* .331 .225 –.133* –.239 .151 .074 .028 .326 .658 .163 .142 .078 .055 .050 .072 .061 .063 .098 .097 .140 .184 .240 .140 .188 .135 .054 .133 n (sample size) R2 F–Statistic 618 .70 30.79** Variable t–statistic 1.577 .301 4.368 .970 .511 .080 .875 6.216 11.971 4.167 5.860 5.406 .756 .116 1.154 3.036 4.466 1.454 2.438 1.763 1.658 2.451 –1.801 Note. In order to save space, we are not reporting the estimates for the 39 institution fixed effects coefficients. It can be reported that the results of the F–test for the group of fixed effect coefficients was highly significant. The F(39,555) = 19.45 and was significant at the .01 level. A fixed effects model is estimated because the previous results suggest that there are important unobservable differences across schools. For instance, an important factor in determining attendance may be tradition. Some schools have rich traditions in football and thus draw large crowds. If these differences are accounted for, then it is expected that there should not be a significantly positive relationship between price and attendance. To do this, the model was reestimated Determinants of Spectator Attendance 325 with a categorical variable for each school. The fixed effects results are reported in Table 4. The first thing to note about the results is that the coefficient on price is much smaller and now statistically insignificant. Secondly, it is noted that the fixed effects results are otherwise qualitatively equivalent to the OLS results. From this one can conclude the following. In the OLS case price was significant because it was correlated with unobservable individual school characteristics that affect demand. Once these fixed effects are accounted for price is no longer significant. As a result, there is not much we can say about the price elasticity of demand based on our study. If one were interested in examining the effects of price changes on attendance then a data set with a longer time series and more variation in price over time would be required. Conclusions and Recommendations The purpose of this paper was to analyze the relationship between spectator attendance at NCAA Division II football games and a set of selected determinants such as on-field success and promotional activity. The relationships were analyzed through the development of economic demand models using log-linear multiple regression analysis. The results form OLS and fixed effects models were qualitatively equivalent and the models explained between approximately 40 to 70 percent of the variation in attendance. Both on-field success, as measured by winning percentage, and promotional activity were found to be positively related to attendance. One of the primary contributions of this paper is that it examined how the effect of winning on attendance changes throughout the season. By interacting winning percentage with time, it was shown that the effect of winning in the current season increases over the course of a season while the impact of the previous season’s winning percentage on attendance decreased as the season progressed. Attendance was also positively influenced by promotional activities such as Homecoming. Given this, new promotional activities may be an effective strategy for administrators who are attempting to increase attendance. In closing, this study has shown that it is possible to explain the variation in collegiate sport spectatorship through the use of a tool that has been primarily used for professional sport. The need for additional research concerning the determinants of spectator demand for collegiate sport is evident. Similar models may be constructed for other collegiate sports, at all levels, depending upon the availability of data. It may be of interest to determine the relationship between spectator attendance in other sports like basketball and hockey with determinants such as winning percentage, promotional activity, and ticket price. Lastly, we note that this study has focused on the effects that promotions and on-field success have on attendance, and due to the nature of our data set, we were unable to study the price elasticity of demand. Administrators might be interested in the implications of various pricing policies and thus future research based on a data set with a longer time series would help examine this issue. 326 DeSchriver and Jensen References Baimbridge, M. (1997). Match attendance at Euro 96: Was the crowd waving or drowning? Applied Economics Letters, 4, 555-558. Burdekin, R.C., & Idson, T.L. (1991). Customer preferences, attendance and the racial structure of professional basketball teams. Applied Economics, 23(1B), 179-186. Cairns, J.A., Jennett, N., & Sloane, P.J. (1986). The economics of professional team sports: A survey of theory and evidence. Journal of Economic Studies, 13(1), 3-80. Campbell, R. M. (2001, January 15). College football attracts second-highest attendance total. The NCAA News, pp. 1, 13. Coughlin, C.C., & Erekson, O.H. (1985). Contributions to intercollegiate athletic programs: Further evidence. Social Science Quarterly, 66(1), 195-202. Courneya, K.S., & Carron, A.V. (1992). The home advantage in sport competitions: A literature review. Journal of Sport & Exercise Psychology, 14, 13-27. DeSchriver, T.D. (1999). Factors affecting spectator attendance at NCAA Division II football contests. International Sports Journal, 3(2), 55-65. Dobson, S.M., & Goddard, J.A. (1992). The demand for standing and seated viewing accomodation in the English Football League. Applied Economics, 24(10), 1155-1163. Fizel, J.L., & Bennett, R.W. (1989). The impact of college football telecasts on college football attendance. Social Science Quarterly, 7(4), 980-988. Fulks, D.L. (2000). Revenues and expenses of divisions II and III intercollegiate athletics programs: Financial trends and relationships-1999. Indianapolis, IN: National Collegiate Athletic Association. Hansen, H., & Gauthier, R. (1989). Factors affecting attendance of professional sports events. Journal of Sport Management, 3(1), 15-32. Hilkemeyer, F. (1993). Food for thought. Athletic Business, 17(5), 39-43. Iso-Ahola, S.E., & Hatfield, B. (1986). Pyschology of sports: A social psychological approach. Dubuque, IA: Wm. C. Brown. Jennett, N. (1984). Attendances, uncertainty of outcome and policy in Scottish League Football. Scottish Journal of Political Economy, 31(2), 176-198. Kaempfer, W.H., & Pacey, P.L. (1986). Televising college football: The complementarity of attendance and viewing. Social Science Quarterly, 67, 176-185. Kahane, L., & Shmanske, S. (1997). Team roster turnover and attendance in major league baseball. Applied Economics, 29, 425-431. Marcum, J.P., & Greenstein, T.N. (1985). Factors affecting attendance of Major League Baseball: II. A within-season analysis. Sociology of Sport Journal, 2(4), 314-321. Mawson, M.L., & Bowler, W.T., III. (1989). Effects of the 1984 Supreme Court ruling on the television revenue of NCAA Division I football programs. Journal of Sport Management, 3, 79-89. McDonald, M. & Rascher, D. (2000). Does bat day make cents? The effect of promotions on the demand for Major League Baseball. Journal of Sport Management, 14, 8-27. Noll, R.G. (1974). Attendance and price setting. In R.G. Noll (Ed.), Government and the sports business (pp. 115-157). Washington, DC: The Brookings Institute. Pease, D.G., & Zhang, J.J. (2001). Socio-motivational factors affecting spectator attendance at professional basketball games. International Journal of Sport Management, 2(1), 31-59. Peel, D., & Thomas, D. (1988). Outcome uncertainty and the demand for football: An analysis of match attendances in the English Football League. Scottish Journal of Political Economy, 35(3), 242-249. Determinants of Spectator Attendance 327 Ragan, J.F., & Thomas, L.B. (1990). Principles of microeconomics. San Diego: Harcourt Brace Janovitch. Rivett, P. (1975). The structure of league football. Operational Research Quarterly, 26(4), 801-812. Schofield, J.A. (1983). Performance and attendance at professional team sports. Journal of Sport Behavior, 6(4), 196-206. Scully, G.W. (1974). Pay and performance in Major League Baseball. The American Economic Review, 64(6), 915-930. Siegried, J.J., & Hinshaw, C.E. (1979). The effect of lifting television blackouts on professional football no-shows. Journal of Economics and Business, 32(1), 1-13. Wakefield, K.L., & Sloane, H. J. (1995). The effects of team loyalty and selected stadium factors on spectator atttendance. Journal of Sport Management, 9(2), 153-172. Wall, G.V., & Myers, K. (1989). Factors influencing attendance: Toronto Blue Jays games. Sports Place International, 3(1/2), 19-33. Wann, D.L., McGeorge, K.K., Dolan, T.J., & Allison, J.A. (1994). Relationships between spectator identification and spectators’ perceptions of influence, spectators’ emotions, and competition outcome. Journal of Sport and Exercise Psychology, 16, 347364. Wells, D.E., Southall, R.M., & Peng, H.H. (2000). An analysis of factors related to attendance at Division II football games. Sport Marketing Quarterly, 9(4), 203-211. Whitney, J.D. (1988). Winning games versus winning championships: The economics of fan interest and team performance. Economic Inquiry, 26(4), 703-724. Zhang, J.J., Pease, D.G., Hui, S.C., & Michaud, T.J. (1995). Variables affecting spectator decisions to attend NBA games. Sport Marketing Quarterly, 4(4), 29-39. Zhang, J.J., Smith, D.W., Pease, D.G., & Mahar, M.T. (1996). Spectator knowledge of hockey as a significant predictor of game attendance. Sport Marketing Quarterly, 5(3), 4148. Notes 1 For the remainder of the paper, the term football will be used to describe American football. 2 All earlier research that has included winning percentage as a determinant of spectator attendance has found a significantly positive relationship. Earlier work such as McDonald and Rascher (2000), Marcum and Greenstein (1983), and Wall and Myers (1989) found a significant positive relationship between attendance and promotional activity. 3 This is true only when the capacity constraint of the stadium is non-binding. This is the case in the vast majority games in our sample. 4 Ticket prices were adjusted for inflation using the GDP deflator index. 5 1990 US Census data were utilized for this variable. 6 Whitney (1988) also had success with use of logarithmic transformations of spectator attendance data sets. 7 Note that for the fixed effects estimation we only included the schools that are in the sample in all three years and thus there were 40 schools used in this portion of the analysis. We also reestimated our GLS model with this smaller sample and the results are qualitatively equivalent to the results reported in Table 2. This suggests that the results are not sensitive to changes in the sample of schools considered. 8 Ticket prices were adjusted for inflation using the GDP deflator index. 328 DeSchriver and Jensen 9 1990 US Census data were utilized for this variable. It must be noted that .49 was subtracted from each winning percentage prior to it being multiplied by the time period categorical variables. This was done to differentiate between a game that was not played in a specific time period, and would therefore have a value of 0, and a game where the home team had a winning percentage of 0. Statistically, this has no effect on the results because every winning percentage was reduced by the same value, .49. An example may help in describing the formulation of this determinant. Suppose a game was played in the first time period and the home team had a .500 winning percentage prior to the game. In this instance, the WINPER1 variable has a value of .01, and the WINPER2, WINPER3, and WINPER4 variables have a value of 0. 10 Appendix A Description of Variables Dependent Variable ATTEND – Game Attendance The number of spectators who attended an individual football game, as reported by the host institution. Economic Determinants GAP - General Admission Ticket Price The price of a general admission ticket to a football contest.8 It was expected that a positive relationship existed between price and attendance. SAP – Student Admission Policy Responding institutions had two distinct policies with respect to the admittance of full-time students to football games. In some cases, students were admitted free of charge, while in other instances students were required to purchase a ticket. A categorical variable was used to represent this variable with 1 representing no payment for admission, and 0 otherwise. It was expected that a positive relationship existed between student admission policy and attendance. COMP – Number of Competitors in the Market Area The number of NCAA Division I and II institutions that competed in football and were located within 50 miles of the responding institution. It was expected that a negative relationship existed between COMP and attendance. Demographic Determinants MILES – Miles Between Competing Institutions The number of miles between the cities in which the two competing football teams were located. It was expected that a negative relationship existed between MILES and attendance. STUDENR – Student Enrollment The number of full-time enrolled students, undergraduate and graduate, who attended the host institution during the year in which the game was played. It was expected that a positive relationship existed between student enrollment and attendance. Determinants of Spectator Attendance 329 CITYPOP – City Population The number of residents living in the city where the home team was located.9 The relationship between city population and attendance was expected to be positive. Game Attractiveness Determinants WINPER – Home Team Winning Percentage The winning percentage for the home team was captured through four variables. This allowed for the analysis of the effect of winning percentage on attendance over the course of a season. A series of four categorical variables were used to represent the time period in the season when a specific game was played. Each of the three seasons analyzed had games played over a span of 12 weeks. The weeks were separated into four time periods, three weeks in each period. One categorical variable was assigned for each time period. The categorical variables took a value of 1 if the game in question was played during that time period, and 0 otherwise. For each game, the winning percentage of the home team prior to the game was interacted with the four time period categorical variables.10 By representing winning percentage through this method, both the size and change over time of the relationship between winning percentage and game attendance was analyzed. It was expected that a positive relationship existed between winning percentage and attendance and that this relationship would strengthen during the course of a season. PREVWIN – Home Team Winning Percentage in Previous Season This determinant was constructed in a manner that was similar to WINPER. Again, this was done to determine how the effect of the home team’s on-field success in the previous season would effect attendance throughout the following season. PREVWIN was represented through a series of four variables. It was expected that a positive relationship would expect between winning percentage in the previous season and attendance. In addition, this relationship was expected to weaken throughout the course of a season. Promotional Activity This determinant was measured through the use of two categorical variables. The first variable (HCPROM) measured the existence of homecoming and had a value of 1 if the game was played as part of an institution’s homecoming celebration, and 0 otherwise. A second categorical variable (OTHPROM) was used to capture the effect of other types of promotions such as parents’ day, youth day, and senior citizens’ day on game attendance. It has a value of 1 if any promotion other than homecoming occurs, and 0 otherwise. A positive relationship was expected between both promotional variables and spectator attendance. The relationship was expected to be stronger for the HCPROM variable. Residual Preference Determinants WEAT – Weather The weather conditions were represented by a categorical variable. It took the value of 0 if rain or snow fell during the time period consisting of two hours prior to the start of the football game and/or during the football game, 1 otherwise. A 330 DeSchriver and Jensen positive relationship was expected between good weather conditions and attendance. STADAGE – Age of Stadium The age of the stadium, in years, where the game was played. The negative relationship was expected between the age of the stadium and attendance. TP – Time Period As mentioned above, each of the three seasons was separated into four time periods. Each time period consisted of three weeks. Three categorical variables where used to analyze the relationship between time and game attendance. The time period categorical variable took the value of 1 for games played during that time period, and 0 otherwise. All of the time period variables took the value of 0 for games played during the fourth time period. The expected relationship between the time of the season in which the game was played and attendance was unclear, a priori. Year – Year of Season For the demand model including data from all three of the years (referred to as the combined data set), two categorical variable were included to account for the year of the season. These variables allowed for the analysis of the relationship between the season in which the game was played (1994, 1996, or 1999) and spectator attendance. For the year 1994, both of the categorical variables were 0. The expected relationship between the year of the game and attendance was unclear prior to the statistical analysis.
© Copyright 2026 Paperzz