Stone et al. (1999) empirically tested a conceptual model of factors that affect attendance at sporting events in small colleges. Factor analysis was used to group Likert scale statements into six categories: school identification (the degree to which a person identifies with the school), player identification (the degree to which a person identifies with players on the team), time, being a sports fan, entertainment value, and awareness. In a regression model, awareness, player identification, and school identification were significant in explaining attendance.

Zhang et al. (1997) studied the effect of entertainment options on attendance at minor league hockey games. They found that competitive entertainment options, including movies and television, could have a significant impact on attendance.

METHODOLOGY

The four papers cited in the previous section used Likert-scale variables in questionnaires that were distributed at selected sporting events. The variables used in these papers were the starting point for developing a questionnaire that was likely to reflect the characteristics important for attending Division II football games at the university studied. At the beginning of the study, personal interviews were conducted with approximately twenty students at the university. These interviews were used to modify the variables used in previous studies to reflect variables important to students at this university. The questionnaire is shown in Exhibit I [at the end of this paper- Ed].

The questionnaire was administered after the fifth home game. A convenience sample was used. Because attendance by students represents the bulk of fan support, this study was confined to understanding student behavior. In addition, because students attended the games for free, price was not included in the study. Students were not interviewed at the football games because the researchers wanted to include respondents who did not attend any games. Therefore, students were interviewed in business, humanities, social science, biology, and engineering classes. Of the 220 questionnaires that were completed, 32 had missing data in at least one of the variables, leaving 188 usable responses.

The first part of the questionnaire asks respondents to recall which games they attended, which they knew about in advance, and which they listened to on the radio. The variable ATTEND is total number of games that respondents attended. The variable KNEW is the total number of games that respondents knew about in advance. Both of these variables are subject to response error because respondents may incorrectly recall which games they attended and knew about in advance. Although data on listening to games was recorded as a way to measure behavior, short of attending, this variable was not used in the analysis because very few respondents listened to the games on the radio.

The second part of the questionnaire asks about potentially competitive entertainment, patterned after the work of Zhang et al. (1997). A question about hunting was added because of the importance of hunting, as mentioned by students in the interviews. The third part of the questionnaire is patterned after the works of Wakefield et al. (1996), Tomlinson et al. (1995), and Stone et al. (1999). The fourth part asks five demographic questions, about gender, years at the university, age, marital status, and children. Because only twelve of the respondents were married and only four had children, these variables were not used in the subsequent analysis.

DATA ANALYSIS

The eight Likert scale statements from question #2 in the questionnaire and the fourteen Likert scale statements from question #3 were analyzed using factor analysis to determine the basic, underlying structure of these variables. As described by Hair et al. (1995), the pattern of correlation coefficients, the Bartlett test, and the Measure of Sampling Adequacy (MSA) were used to assess the factorability of the correlation matrix. Because of a low MSA of 0.367, the first statement in question #2 concerning hunting was not used in the factor analysis. Five factors were extracted, based on the criterion of having eigenvalues greater than one. A Scree plot also suggested that five factors were appropriate for extraction. The five factors represented slightly over 63% of the variability in the data.

The factor loadings, after varimax rotation, for the remaining twenty one variables on the five factors is shown in Exhibit II. Based on the pattern of factor loadings, Factor 1 might be described as "Secondary Fan" characteristics. For example, the heaviest loading is for variable Q310, "I attend only if the team has a winning record." Factor 2 could be labeled the "Facilities" factor. The highest loading is for "I think there are good bathroom facilities." Factor 3 measures the "True Fan." The highest loading is for "I attend for the sport itself." Factor 4 is "Other Activities." The highest loading is for "I would rather work out or exercise than attend games." Finally, Factor 5 is a measure of preference for "TV Sports," with the highest loading being for "I would rather watch college football on TV than attend games."

Exhibit II
Factor Loadings for Rotated Factor Matrix

Extraction Method: Principle Component Analysis.
Rotation Method: Varimax with Kaiser Normalization.
Highest absolute factor loadings for each variable are given in bold numbers.
Variable name corresponds to the number on the questionnaire in Exhibit I.

The purpose of the factor analysis was to use the results in a regression model to explain attendance. As described by Hair et al. (1995) surrogate variables, summated scales, or factor scores might be used for this purpose. For this study, factor scores were used. The independent variables in the model were therefore the five factors described above, using the corresponding factor scores, Q2P1 (Hunting interferes with my attending games), KNEW (the number of games the respondents said that they knew about ahead of time), GENDER, YEARS at the university, and AGE. Descriptive statistics for these variables are given in Exhibit III.

Exhibit III
Descriptive Statistics for Independent Variables

The dependent variable, which is the number of home games attended (ATTEND), is a series of discrete values from 0 to 5. An appropriate regression procedure when the dependent variable is ordinally scaled is ordered probit. Therefore, in order to examine the effects of the independent variables on attendance, an ordered probit procedure was used with ATTEND as the dependent variable and with KNEW, Q2P1, Factor 1, Factor 2, Factor 3, Factor 4, Factor 5, GENDER, YEARS, and AGE as independent variables. Regression results are reported in Exhibit IV. These results were produced using MinitabÔ . The results are statistically significant based on the G statistic, which follows a c2 distribution with the degrees of freedom equal to the number of independent variables (Hosmer and
Lemeshow 1989).

The significant independent variables (a < 0.05) are KNEW and Factors 2, 3, 4, and 5. Because of the way MinitabÔ calculates the coefficients in ordered probit analysis, the reported negative coefficients indicate that an increase in the independent variable tends to be associated with a greater attendance. The pattern of coefficients is as one would expect. Increases in KNEW (knowing about the games in advance), Factor 2 (Facilities), and Factor 3 (True Fan) are associated with increases in attendance. In contrast, increases in Factor 4 (Other Activities) and Factor 5 (TV Sports) are associated with decreases in attendance.

Exhibit IV
Regression Results

Exhibit V
Marginal Probabilities

Reported marginal probabilities may not sum to zero due to rounding.

In linear regression, the estimated coefficients can be interpreted as marginal effects. In ordered probit, the marginal effects must be calculated using the coefficients, and are reported as probabilities. The marginal effects for the significant independent variables are calculated as described in Green (1993) and are shown in Exhibit V. The following illustrates the interpretation of Exhibit V. For each one point increase (i.e., a shift in one standard deviation) in Factor 3, holding all other variables at their mean values, the probability of a student fan attending no games decreases by 30.33%; the probability that one game will be attended increases 6.38%. For each one point increase in Factor 4, holding all other variables at their mean values, the probability of a student fan attending no games increases by 17.76%; the probability that one game will be attended decreases by 3.74%. For KNEW, for every additional game that a fan knew about, holding all other variables at the mean values, the probability of a student fan attending no games decreases 4.7%; the probability of attending one game increases by 0.99%. The other marginal probabilities are interpreted similarly.

DISCUSSION

The marginal probabilities shown in Exhibit V reveal the major causes of student fan attendance at the university studied. Most important is enjoyment of the game itself, as shown by Factor 3 (True Fan). This factor could be thought of as "circumstantial" characteristics, using the Tomlinson et al. (1995) classification. Respondents who scored high on Factor 3 attended in spite of the weather; they attended for the sport itself. Low scores had a depressing effect on attendance. Factor 4 (Other Activities) and Factor 5 (TV Sports) are active (e.g., exercise) and passive (e.g., watch TV) market competitors described by Zhang et al. (1997). Based on the marginal probabilities, as respondents’ scores on these factors increased, the probability of attending games decreased. Not surprisingly, awareness is also important. As awareness of game times increased (KNEW), the probability of attending games also increased. This finding agrees with that of Stone et al. (1999), where awareness was significant in explaining attendance.

The results suggest that in order to improve attendance, this Division II university needs to identify its core market of true football fans and ensure that potential fans know about the home football schedule. Awareness is important in explaining attendance and needs to be managed carefully, especially because promotional budgets tend to be thin at Division II programs. The quality of the facilities (Factor 2) is also important in improving attendance. This factor is a "front room" characteristic, using the Tomlinson et al. (1995) classification. Its importance suggests that the sportscape has a significant impact on attendance.

One way to look at the depressing effect on attendance of market competitors (Factors 4 and 5) is that the football game itself is not entertaining enough to attract customers, except for true fans. At the university studied, the homecoming game had the highest attendance (1,897), even though the weather was extremely cold and blustery that day. The draw of homecoming apparently encouraged marginal fans to attend. Athletic directors at small schools might consider what entertainment or special events could be added in order to encourage attendance by a broader segment of the market, as is currently done in professional and many Division I sporting events. Similarly, care in managing the sportscape should encourage the attendance by those who are not in core market of true football fans.

It is interesting to note that Factor 1 (Secondary Fan) was not significant in explaining attendance. This factor includes a mixture of front room, back room, and circumstantial variables as described by Tomlinson et al. (1995). It may be that this factor is not significant in explaining attendance because there is little expectation by fans for an exciting band, vibrant cheerleaders, or special event entertainment (all of which loaded heavily on Factor 1). At the university studied, there is a pep band, but not a marching band. The only special event during the season is homecoming.

In the Division II football program studied, there is a core market of true fans. In order to build on this core market, the program needs to manage its limited promotional budget carefully in order to improve awareness. Attracting potential fans that might prefer competitive activities will require managing the sportscape and entertainment in order to attract fans beyond the core market. Whether or not the findings from this study are applicable to other small colleges is a subject for further research. It would be especially interesting to determine the influence of Factor 1 characteristics on attendance. It could be that these characteristics (band, cheerleaders, special events) have a statistically significant impact on attendance at schools that manage them better.

Exhibit I
Football Attendance Survey
All Games are Saturday

The purpose of this survey is to determine why people do or do not attend the home football games. There are no right or wrong answers. If you cannot remember for certain, answer as well as you can. Thank you.

1. For each home game you knew about ahead of time check the "Knew About" box. If you also attended that game check the "Attended" box. If you listened to the game on the radio, check the "listened" box. At the bottom of each column put a total, and if there is no total put a 0.
 

2. Here are some activities that may affect your attendance at home football games. For each statement please circle the appropriate number to indicate your level of agreement or disagreement.

1 – Strongly Disagree , 2 - Disagree, 3 - Neutral, 4 - Agree, and 5 – Strongly Agree

3. Below are some statements that may reflect your attitudes when deciding to attend a home football game. For each statement, please indicate your level of agreement or disagreement.

4. Please tell us about yourself. (Please check the appropriate box or fill in the answer)

Gender: [] Male [] Female

How many years have you attended the university (including this year):

[] 1 [] 2 [] 3 [] 4 [] 5 or more

Age: _______

Are you married? [] Yes [] No

Do you have children? [] Yes [] No

REFERENCES

Bitner, Mary Jo (1992) "Servicescapes: The Impact of Physical Surroundings on Customers and Employees." Journal of Marketing, Vol 56 No 2: 57-71

Greene, W. H. (1993) Econometric Analysis, New York: John Wiley & Sons

Hair, Joseph F. Jr., Rolph E. Anderson, Ronald L. Tatham, and William C. Black (1995) Multivariate Data Analysis, 5th Edition, Englewood Cliffs, New Jersey:
Prentice-Hall

Hosmer, D. W. and S. Lemeshow (1989) Applied Logistic Regression, New York: John Wiley & Sons

Krohn, Franklin B. and Mark Clarke (1998) "Psychological and Sociological Influences on Attendance At Small College Sporting Events." College Student Journal
Vol 32 No 2: 277-287

Lehnus, Darryl L. and Glenn A. Miller (1996) "The Status of Athletic Marketing in Division IA Universities." Sports Marketing Quarterly, Vol 5 No 3: 31-48

Robertson, David and Nigel Pope (1999) "Product Bundling and Causes of Attendance and Non-attendance in Live Profession Sport: A Case Study of the
Brisbane Broncos and the Brisbane Lions." The Cyber-Journal of Sport Marketing, Vol 3 No 1

Stone, George W., Michael A. Jones, and Cameron Montgomery (1999) "A Conceptual Model of the Antecendent Factors Contributing to Fan Support at Small
College Athletic Events." Proceedings of the Atlantic Marketing Association, 1999, 269-283

Wakefield, Kirk L., Jeffrey G. Blodgett, and Hugh J. Sloan (1996) "Measurement and Management of the Sportscape." Journal of Sports Management, Vol 10 No
1: 15-31

Zhang, James J., Dennis W. Smith, Dale G. Pease, and Elizabeth A Jambor (1997) "Negative Influence of Market Competitors on the Attendance of Professional
Sport Games: The Case of a Minor League Hockey Team." Sport Marketing Quarterly, Vol 6 No 3: 31-40