Do Reliable Predictors Exist for the Outcomes of NASCAR Races?


This research attempts to ascertain whether factors known prior to a NASCAR race can help to predict the order of finish of that race. We provide evidence in the form of correlation analysis of the order of finish with available quantitative and categorical information collected, and a simple test for the effect of teams (regressions for each races are also available from the authors). Data were collected on 14 races from the 2003 NASCAR Winston Cup (now Nextel Cup Series) schedule.

Many factors influence the outcomes of NASCAR races. The speed and handling of the car, the skill of the driver, and the performance of the pit crew are but a few of the variables that are important determinants of the finish for a particular car and driver. Many variables outside the control of a particular team, such as the behavior of other drivers, weather, cautions, and the like also influence the final order of finish in NASCAR races. A priori, then, it would be anticipated that predicting outcomes in any meaningful way would be problematical.

The goal of this project is to determine whether those objective, measurable, variables known prior to the start of a race are useful in determining the order of the outcome. To this end we have assembled full data sets for 14 different races from the 2003 NASCAR Winston Cup series. The data include the following for each race: the order of finish, pole position, qualifying speed, practice time, the number of team members of a given driver in the race, the finish position in the prior race at that particular venue, the finish position in the immediately preceding race, driver points for the previous year in Winston Cup competition, and laps completed for the previous year in Winston Cup competition. We also have dummy variables to indicate whether it is the rookie year for a driver and whether the driver changed teams for the current year.

A Simple Model

As a first approach to the problem of predicting the order of finish in particular NASCAR races, we offer a simple theoretical model. Order of finish is posited to be functionally related to variable sets reflecting car speed, driver characteristics, team characteristics, performance in related races, and other factors. In functional notation,

F = f(S, D, T, RR, O), where:

F = Order of finish for a particular race

S = Car speed

D = Driver characteristics

T = Team characteristics

RR = Performance in related races

O = Other factors

To be sure, the variable categories listed are not distinct from each other. That is, empirical measures of car speed are certainly related to other categories of variables such as driver and team characteristics. The theoretical model serves to provide a framework for the empirical specification of the model.

Car Speed, Driver Characteristics, and Related Races

The effects of car speed on race outcomes are obvious. Faster cars will, on average, finish better. Also obvious are the effects of driver racing skill and experience. If it is possible to proxy for driver racing skill and experience, such proxies should be related to finish position across races.

Car/driver combinations may also be subject to streakiness in consecutive races and they may also be more successful at particular venues. The empirical variables defined in the following section proxy for these effects.

Team Characteristics

Team characteristics, in particular team size, require additional explanation. It is an empirical fact that multi-car teams have, in recent years, dominated the NASCAR Winston Cup series, and it is commonly believed that multi-car teams have advantages over smaller teams. What particular advantages are possible for multi-car teams?

First, the marginal cost of increasing the speed of a car is likely to be very sharply upward sloping (Allmen, 2001). This is due in part to NASCAR rules regarding car shape, size, aerodynamics, weight, and engine characteristics. While these rules are in place to equalize competition, the existence of this degree of uniformity makes it very difficult and expensive to gain an advantage within the rules. As Bill Elliott, a driver and past owner observes, “It may cost you $5 million to get to the track, but it may cost you an additional $3 million for a few tenths better lap time ….” (Middleton, 2000, p. 37).

A team with more car/driver combinations can apply any found advantage to each of its cars. Such advantages then result in better performances for all cars on the team, and hence greater performance revenues. Consider Figure 1 in which marginal cost (MC) increases sharply as car speed increases and such costs are assumed to be the same for multi-car teams as for single car teams. Since newly discovered speed advantages can be applied to all cars on a multi-car team, those teams can generate greater revenues for the team (MR M = marginal revenue for multi-car teams) than any such advantage generates for a single car team (MR S = marginal revenue for single car teams). Following optimization principles then, multi-car teams would find it worthwhile to achieve a speed of S M, whereas single car teams have incentive to achieve a speed of only S S. If this analysis is correct, multi-car teams would be expected to achieve greater speed in general than single car teams.

Second, it is an empirical fact that larger teams attract greater sponsorship resources, in part because they are more successful. Then, if the sharply increasing marginal costs mean that multi-car teams are more likely to engage in expensive research for given performance benefits and sponsorship revenues depend on performance, the dominance of multi-car teams can be explained (at least in part) by this simple economic analysis.

Figure 1

Figure 1: Multi-car versus single car teams

Third, teams with more sponsorship income are able to offer greater compensation to crewmembers, hire more experienced and specialized team members, such as aerodynamicists, and can more easily afford expensive technology and testing.

Fourth, substantial barriers to success for smaller teams (especially single car teams) may also exist because of scale economies. The advanced technology machinery for making racing parts would be an example of the “lumpy inputs” explanation of scale economies thought to be the most common reason for decreasing long run average cost. Larger teams would then have an advantage since the production of such parts for the team would necessarily be larger in scale.

Other advantages also accrue to multi-car teams. Operationally, multi-car teams also have more test dates available to them at Winston Cup tracks. Hence, more data can be collected and shared among team members when it comes to setting up the cars for races at those tracks. Multi-car teams also have built-in drafting partners, although the NASCAR literature suggests that at the end of the race each driver is “on his own,” (Cotter, 1999; Dolack, 2003; Hinton, 1997; Pearce, 1996, 2003).

Empirical Specification: Races and Data

The data for this project were collected from a variety of Web sites, (Past Race Archive, 2002, 2003), index.html (Statistics Pages from Jayski. 2002), and FS/Auto (Nextel Cup Standings, 2002, 2003). The variable we wish to predict is the order of finish, which is of course, available for each race on the Winston Cup circuit.

Individual Races

The 14 races for which we collected data include short tracks, speedways, super speedways, and a road course. To determine if the same factors are related to order of finish of different races at the same track, we also included both races run at Daytona and both races run at Michigan in 2003. The specific races for which we collected data are: the Brickyard 400 at Indianapolis Motor Speedway, the Food City 500 at Bristol Motor Speedway, the Coca-Cola 600 at Lowe’s Motor Speedway, the Carolina Dodge Dealers 400 at Darlington Raceway, the Daytona 500 at Daytona International Speedway, the Pepsi 400 at Daytona International Speedway, the Virginia 500 at Martinsville Speedway, the Sirius 400 at Michigan International Speedway, the GFS Marketplace 400 at Michigan International Speedway, the Chevy Rock & Roll 400 at Richmond International Speedway, the Aaron’s 499 at Talladega Superspeedway, the Samsung/Radio Shack 500 at Texas Motor Speedway, the Tropicana 400 at Chicagoland Speedway, and the Sirius at the Glen at Watkins Glen International.

The potential explanatory variables for order of finish collected for each race were as follows:

ptime = the practice time closest to race time.

qspeed = the speed at which the car/driver qualified.

pole = position of the car at the start of the race.

points = points scored in the Winston Cup Series for the prior year.

laps = number of laps completed for all Winston Cup races in the prior year.

DNF = did not finish, the number of races in which the driver failed to finish, prior year.

rookie = a dummy variable equal to 1 if the driver was a rookie in 2003, and equal to

0 otherwise.

# drivers = the number of cars/drivers a multi-car owner fields (for 2003, values = 1,2,3,4).

newteam = a dummy variable equal to 1 if the driver was a member of a new team in 2003, and equal to 0 otherwise.

prev = the finish position of the driver in the previous week’s race.

lastyr = the finish position of the driver in the 2002 running of the same race.

Car Speed

The first three variables from the above list, practice time, qualifying speed, and pole position correspond to the car speed category from the model outlined in the previous section. Clearly qualifying speed and pole position are very closely related (since pole position is determined primarily by qualifying speed), however race officials, for reasons such as a rule and/or equipment violation, missing the driver’s meeting, switching to a backup car, an engine change, or a driver change, may alter pole position. For this reason we collected data for both qualifying speed and pole position in case one or the other is a better predictor of race outcomes.

Driver Characteristics

The next four variables, points, laps, DNFs, and rookie, are driver characteristics with the first three representing performance in the prior year, and the variable rookie is a proxy for lack of experience on the Winston Cup circuit. Theoretically, rookies will not have the skill level that existing Winston Cup drivers have developed over the years, nor will they have had exposure to certain tracks that more experienced Winston Cup drivers have competed on in the past.

Team Characteristics

The variables # drivers and newteam correspond to the team characteristics category in the model. The # drivers variable measures the effect of a given owner having multiple cars/drivers or a multi-car team. With respect to the new team variable (newteam), drivers joining a new team will require time to adjust to the way the crew operates, in addition to developing an effective communication style with the crew chief.

Related Race Effects

Related race effects are measured by the variables prev and lastyr. The variable prev is an attempt to proxy for possible streakiness from race to race. That is, are good finishes followed by other good finishes and poor performances followed by poor performance in the following race? The variable lastyr attempts to measure whether a certain racetrack is a better venue for certain car/driver combinations. For example, the dominance of Dale Earnhardt, Incorporated (DEI) at the superspeedways illustrates the expertise a team may develop at specific racing venues (McCarter, 2002). Since 2001, DEI has won 9 out of the 12 races at Daytona International Speedway and Talladega Superspeedway.

Methodology and Expectations

As a first attempt to determine those variables that relate to order of finish, correlation coefficients are computed between order of finish and each of the measured explanatory variables. The following signs are anticipated for the correlation coefficients:

Expected Sign

of coefficient Explanation

equation1 Faster (lower) practice time leads to better finish

equation2 Higher qualify in speed (MPH) leads to better finish

equation3 Better pole position leads to a better finish

equation4 More points from previous year leads to a better finish

equation5 More laps completed from previous year leads to better finish

equation6 More failures to finish leads to poorer finish

equation7 Rookies may be less likely to have better finishes

equation8 Multi-car teams may have better finishes

equation9 Driver on new teams less likely to have better finishes

equation10 Previous race finish positively related to current race finish

equation11 Previous finish at this track positively related to current finish


Note: ρ represents the population correlation coefficient, and f represents finish position, 1 = winner , 2 = second place, etc..


Correlation Analysis

Table 1 in the appendix is the result of the correlation computations. The coefficients in bold are statistically significant at the α = .05 level and consistent with the predicted signs presented in the previous section.

Several results of this exercise are interesting and potentially important for predicting the outcome of NASCAR races. First, considering the columns (how the variables fared across different races), on average the signs of the variables are in accord with expectations (though some are on the whole insignificant). Several of the variables seem to be consistently correlated (linearly) with order of finish across races. For example the number of drivers variable (# drivers) is statistically significant for all races except Darlington and Watkins Glen (even then the coefficients have the predicted sign). Of course that teams with more members tend to be more successful is not a new conclusion—these results support statistically, at the individual race level, the hypothesis that multi-car teams are generally more successful (see the section on team characteristics above). Of the two tracks that did not have statistically significant results with respect to the # drivers variable, the Watkins Glen result might be due to the fact that it is a road race. Watkins Glen is one of only two road courses utilized by Winston Cup, and teams often use substitute drivers with more road racing skills than their full-time driver may possess in these races.

Indicators of drivers’ past successes also are correlated with order of finish. The variable points is statistically significant for 11 of the 14 races and all of these sample correlations have the anticipated sign. Interestingly enough, the three races that did not demonstrate significant results were run at Daytona and Talladega, the two restrictor plate tracks. Similarly laps, which might be interpreted as a measure of driver/car consistency and driver experience, is statistically significant in 7 of the 14 races. The DNF variable seems to explain little in the way of simple correlation with order of finish. In considering this variable, recognize that a driver with more DNFs may have simply competed in more races than another driver. Thus simple correlation, which does not control for levels of other variables, may not be appropriate to measure such effects.

Measures that account for car/driver speed include pole position (pole), qualifying speed (qspeed) and practice time (ptime). We recognize that pole position and qualifying speed generally measure the same effect. Both are included here to see if one or the other is more closely correlated with order of finish. Based on the sample correlations in Table 1, qspeed is significantly related to order of finish in half of the races and pole in 6 of the 14 races. Practice time (ptime) seems to fare somewhat better—it is significantly related to order of finish in 9 of the 14 races. There may be several reasons for this outcome. The practice times used for statistical analysis were collected from the practice session conducted closest to race time, if all drivers participated in that session. If all drivers did not participate in the last practice session, then practice time statistics were taken from the session run closest to race time in which all drivers practiced. This was done to ensure that the cars would be “set up” in practice as close to race set-up as possible. Since the cars are set-up for race conditions when they practice, it would be expected that the ptime would more closely relate to order of finish than qspeed because the set-up for qualifying is based on two laps at the fastest speed possible. Race day set-up is designed to accommodate consistency and longer runs on the track.

The variable that measures the finish position in the driver’s last race (prev) is statistically related to order of finish in 8 of the 14 races and has the expected sign for all races. This would suggest that driver/car combinations are subject to streakiness, that is, good finishes tend to be followed by other good finishes and vice versa. For only four races is the variable lastyr, the finish position of the driver in the prior running of the race by the same name correlated with the current finish position.

Of the two categorical variables, newteam (equals 1 if the driver joined a new race team for the 2003 season, 0 otherwise) is related to order of finish in 10 of the 14 races and in all races has the anticipated sign. Changing teams, on average, would seem to be related to poorer finishes. On the other hand, rookie status (rookie) was related to finish order only for the first Daytona race and Martinsville.

Again considering the columns in Table 1, the average of the correlation coefficients for each of the explanatory variables across the 14 races is included in the table as the bottom row. A coefficient above 0.25 is generally statistically significant for individual races (again α = .05, one-tailed test, n = 43) On that basis, eight of the variables (laps, points, newteam, pole, # drivers, prev, ptime, and qspeed) are on average statistically (linearly) related to order of finish.

It is also useful to consider the correlations for individual races, i.e., to consider Table 1 by row. For example, at Martinsville order of finish was linearly related to 9 of the 11 variables in the explanatory variable set. The first (June) Michigan race, Richmond, and Chicagoland were linearly related to eight of the explanatory variables. At the other end of the scale, for the two Daytona races, only two of the explanatory variables were correlated with order of finish. One of those variables was the same (# drivers) for both Daytona races. Interestingly, comparing the second (August) Michigan race to the first, only five variables were statistically significant for the second race, but each of those variables was also significant for the first Michigan race. However, relatively strong correlations for pole, practice time and qualifying speed for the first Michigan race were not repeated for the second Michigan race. The reader may examine Table 1 to see that the rest of the races have from three to seven explanatory variables that are statistically significant.

Additional Evidence on Team Effect

The effect of team membership (#drivers) seems to play an increasingly important role in NASCAR (Cotter, 1999; Dolack, 2003; Hinton, 1997; Pearce, 1996, 2003). In 2003, 12 organizations owned and fielded 33 of the 43 cars competing at the majority of NASCAR races. Additionally the Winston Cup Championship has been won by a multi-car team in each of the last 10 years (Pearce, 2003). Therefore, we considered another test of team membership on car/driver success. Using statistics from the entire 2002 and 2003 racing years, a table of results divided into top 10 finishes and finishes out of the top 10 and classified by number of team members was constructed. Table 2 in the appendix shows that teams with four members (the highest number of team members at the start of 2002) had 285 starts and of those, 43.16% resulted in top 10 finishes. The corresponding percentages are 15.55% for three member teams, 29.14% for two member teams, and only 8.68% for drivers without team members. The largest (four member) teams tended to dominate the top 10 finishes. Perhaps surprising is the fact that two member teams had by far the largest number of starts and the second highest rate of top 10 finishes with 29.14%. For three member teams the corresponding percentage was 15.55% and single drivers finished in the top 10 only 8.68% of the time. A simple chi-squared test of independence of the classification of top 10s by number of drivers on a team, yields a χ 2 = 123.9, which allows rejection of the null of independence at α < 0.001. This result confirms the obvious result that the proportion of top 10 finishes does depend on the number of team members.

Table 3 contains the same categories for the 2003 Winston Cup drivers. There was one team with five drivers for the 2003 season, so the table contains an additional column. For the 2003 season, the percentage of top 10 finishes is remarkably constant for the teams with five, four and two members, with 37%, 38% and 32% respectively. Again, teams with three members and especially the single drivers fared less well on the basis of top 10 finishes. Again, the null hypothesis of independence between number of team members and top 10 finishes can be rejected (χ 2 = 106.0), providing statistical confirmation of the already clear evidence that the proportion of top 10 finishes differs by number of team members.


The correlation analysis across 14 races for the 2003 NASCAR Winston Cup series identifies a number of variables that are associated with the order of finish of these races. On average, variables measuring car speed, including practice time, qualifying speed, and pole position are related to the order of finish of races. We also find that prior success on the part of the driver, measured by laps completed in the prior year and points accumulated are also correlated with order of finish. Whether or not the driver was a rookie was, perhaps surprisingly, not on average correlated with finish order across races. There is also some evidence that performances of driver/car combinations are subject to streaks. That is, finish positions in a given race are often correlated with finish positions in the race that follows. Of course these results could simply reflect the fact that some driver/car combinations consistently finish better than others may. Changing teams is correlated with poorer finishes and team size is correlated with better finishes.

The effect of team membership is reinforced by the data in Tables 2 and 3, which classifies top 10 finishes by number of team members. Teams with more members are more successful in terms of top 10 finishes. However, this effect is not monotonic in nature, since two member teams have a larger percentage of top 10s than do teams with three members.

Further research is indicated to test the robustness of these results. Such analysis could include races not in our data set and results from different years of NASCAR racing.



Allmen, P. von. (2001). Is the reward system in NASCAR efficient? Journal of Sports Performance, 2(1), 62-79.

Cotter, T. (1999). Say goodbye to the single-car team. Road & Track, 50(8), 142-143.

Dolack, C. (2003). One is the loneliest number. Auto Racing Digest, 31(6), 66.

Hinton, E. (1997). Strength in numbers. Sport Illustrated, 87(16), 86-87.

McCarter, M. (2002). Stepping up to the plate. The Sporting News, 226(27), 38-39.

Middleton, A. (2000, February). Racing’s biggest obstacle. Stock Car Racing, 34-37.

Past Race Archive, 2002 [Data files]. Available from NASCAR Web site,

Past Race Archive, 2003 [Data files]. Available from NASCAR Web site,

Nextel Cup Standings, 2002 [Data files]. Available from FOXSports Web site,

Nextel Cup Standings, 2003 [Data files]. Available from FOXSports Web site,

Parsons, K. (2002, August 26). Tunnel vision – NASCAR teams’ fortunes are blowing in the wind. The Commercial Appeal, Memphis, TN, p. D9.

Pearce, A. (1996). Fair and square. AutoWeek, 46(50), 40-41.

Pearce, A. (2003). Going it alone. AutoWeek, 53(14), 57-58.

Statistics Pages from Jayski. (2002) [Data files]. Available from Jayski Web site,



Table 1: Correlation coefficients between finish position and the explanatory variables

Explanatory Variables

Race Laps DNF Points newteam Pole #drivers prev lastyr ptime qspeed Rookie?
Indianapolis -0.214 0.091 -0.319 0.342 -0.106 -0.326 0.118 0.435 0.374 0.047 0.088
Bristol -0.209 0.316 -0.350 0.030 0.177 -0.417 0.117 0.407 0.392 -0.190 0.088
Lowe’s -0.209 0.030 -0.291 0.360 0.198 -0.314 0.335 0.085 0.403 -0.283 0.200
Darlington -0.373 0.154 -0.428 0.180 0.176 -0.153 0.170 -0.141 -0.239 -0.234 0.265
Daytona (Feb) -0.044 -0.111 -0.173 0.155 0.243 -0.300 0.274 0.045 -0.106 NA 0.094
Daytona (July) -0.140 0.110 -0.218 0.132 0.006 -0.258 0.167 0.300 0.164 -0.060 0.105
Martinsville -0.462 0.079 -0.540 0.030 0.433 -0.342 0.314 0.352 0.462 -0.496 0.263
Michigan (June) -0.383 0.115 -0.435 0.396 0.596 -0.550 0.478 0.125 0.549 -0.547 0.041
Michigan (Aug) -0.342 -0.297 -0.408 0.380 0.172 -0.356 0.392 0.087 0.237 -0.205 0.212
Richmond -0.367 -0.070 -0.508 0.384 0.406 -0.283 0.300 0.126 0.350 -0.403 0.228
Talladega -0.228 0.088 -0.230 0.306 0.398 -0.335 0.098 0.233 0.080 -0.361 0.170
Texas -0.200 0.085 -0.352 0.282 0.169 -0.393 0.195 -0.178 0.300 -0.159 0.146
Chicagoland -0.339 0.064 -0.413 0.260 0.422 -0.261 0.317 0.246 0.428 -0.464 0.178
Watkins Glen -0.410 -0.298 -0.472 0.406 0.297 -0.244 0.254 0.123 0.285 -0.278 0.239
Average -0.280 0.025 -0.367 0.260 0.256 -0.324 0.252 0.160 0.263 -0.279 0.166


Table 2: Top Ten Finishes by Number of Team Members, 2002 Season

4 member teams 3 member teams 2 member teams One member teams
Top 10 % 43.16% 15.55% 29.14% 8.68%
Total starts 285 328 525 357

Table 3: Top Ten Finishes by Number of Team Members, 2003 Season

5 member teams 4 member teams 3 member teams 2 member teams One member teams
Top 10 % 36.87% 38.19% 22.53% 31.67% 7.66%
Total starts 179 144 395 360 418

The flat MR curves are offered as an approximation. Additional speed should add increasing marginal revenue (as cars move up in finish order, added revenue increases), but since all cars are attempting to increase speed, the possible increases in revenue will be distributed among the competitors.

For example, testing for the aerodynamic properties of a car in a wind tunnel can cost more than $2000 per hour (Parsons, 2002).

T he field is generally set using a combination of timed laps and provisionals. The fastest 36 cars earn a place based on time, while positions 37-43 are determined by a process which may include last season’s final owners standings, current owners standings and former champions. The provisionals are assigned in descending order, beginning with the highest ranking owner in the standings. The lone exception is the Daytona 500, which uses two qualifying races to determine the field. (

In other words, a driver with many laps completed and many DNFs would be expected to fare less well than another driver with many laps completed, but few DNFs.

While the sample size is generally 43 for individual races it is somewhat lower for some individual races, e.g., a race in which a driver/car combination did not run in the race at a particular venue it its previous iteration.

If all four members of a team start in the same race, that would equal 4 starts and if two of those four finish in the top 10, that would be 50% in the top 10.

This procedure can also be described as a test of proportions, that is, we have evidence that the proportions of top 10 finishes differs by number of team members.

2016-10-14T11:50:19+00:00March 4th, 2008|Contemporary Sports Issues, Sports Studies and Sports Psychology|Comments Off on Do Reliable Predictors Exist for the Outcomes of NASCAR Races?

A Secret Shopper Project: Reevaluation of Relationship Marketing Efforts


In the current sport business environment, relationship marketing tactics play a predominant role due to the increased importance of the relationship between professional sport franchises and their customers. Definitions of relationship marketing stress the creating and sustaining of a network between the individual customer and the company (Milne & McDonald, 1998; Mullin, Hardy, & Sutton, 2000; Ping, 1999; Shani & Chalasani, 1992; Wulf, Odekerken-Schroder, & Lacobucci, 2001). For many years, professional sport franchises have been interested in developing fan loyalty or psychological connections with their customers through relationship marketing.

In sport business, understanding and developing relationship marketing can increase profits. It can also be a solution for some difficulties that professional sport franchises face. Those difficulties in the four big leagues (e.g., NBA, NFL, MLB, and NHL) are as follows (Bovinet, 1999; Howard & Crompton, 2004; James, Kolbe, & Trail, 2002; Mullin et al., 2000):

  • More new franchises (e.g., 17new franchises were added to the four big leagues over the last decade).
  • Increased competition because of new types of sports and leagues (e.g., Arena Football League, Women’s National Basketball Association, Major League Soccer, Action Sports, etc.).
  • Increases in the number of sporting events on TV (e.g., with the growth of specialized sport channels as well as the major networks, sport fans could watch sport programs 24 hours a day).
  • Increases in ticket prices of the big leagues (e.g., the cost of family of four to attend a major league game: MLB-$148.61 (2002-03) which is a 92% increase compared to 1990-91; NFL-$290.41 (2002-03) which is a 90% increase compared to 1990-91; NBA-$254.88 (2002-03) which is an 84% increase compared to 1990-91; NHL-$240.43 (2002-03) which is an 81% increase compared to 1990-91).
  • More entertainment choices (e.g., computer games, DVD movies, HD TV programs, etc.).
  • Reliance on the gate receipts for revenue in some leagues (e.g., NBA and NHL have relatively smaller TV revenues compared to NFL and MLB).
  • Larger facilities and unlikelihood of getting good-view seats.
  • Negative perceptions of players’ behavior and lifestyle.
  • Increases in player salaries.

In addition to these difficulties, sport consumers are less likely to attend a live sporting event, because it is more comfortable to watch games at home without the hassle of traffic, finding parking, and getting back home late. In addition, watching at home offers the benefits of TV replay and analysis (and cleaner restrooms), without having to spend a great deal of money on games that the whole family may not be able to enjoy together (Howard & Crompton, 2004).

As Bovinet (1999) suggested, however, many of these problems could be alleviated by developing communication networks with current and potential customers. Building and maintaining a psychological connection through relationship marketing would allow franchises to increase trust and commitment among their customers. For example, the customers who have a psychological connection with a team are more likely to expect and receive special treatment, such as on-line member services, birthday and anniversary cards, and others.

Although academics have begun to recognize the importance of relationship marketing practices over the last decade (Bovinet, 1999; Bowen, 2003; McDonald &Milne, 1997; Neuborne, 2004; Reinartz & Kumar, 2000; Shani, 1997), systematic and exploratory studies on the theory of relationship marketing in sport business are still lacking.

Purpose of the Study

The purpose of this study was to reevaluate the relationship marketing efforts that the four big leagues (e.g., MLB, NFL, NHL, and NBA) use to communicate with the potential customers.


Sample & Instrument: A Secret Shopper Letter

This study was modeled based on Bovinet’s exploratory study (1999), published in Sport Marketing Quarterly (v8, n3, 1999). During the fall semester, the students in an undergraduate sport marketing class were divided into two teams. Each team sent a secret shopper letter requesting season ticket information to all 122 franchises of the four big leagues (the participants). Specifically, a first group of letters (N=62) to the NFL (n=32) and the NHL (n=30) were sent on January 30, 2004. The second letters (N=60) to the NBA (n=30) and the MLB (n=30) were sent on March 5, 2004. Students obtained address and contact information for each franchise from the team’s Internet sites and through phone directories. A student’s home address, not school address, was used for responses and letters gave no indication of age or other demographic information. The NHL and the NBA were in the regular season, the NFL had just finished its season, and the MLB was due to start its season soon.

Secret Shopper Letter

March 5, 2004


To whom it may concern:

I am in the process of changing job and will be moving into your area within

the coming months. I would greatly appreciate it if you can send me information

on season ticket, team schedule, parking and events that a held aside from sport

games. I am a huge fan of the (TEAM NAME), and I would like to take every

advantage of being able to regularly attend games. Any other information that

I may have excluded and you can provide would be of the utmost help as I am not

very familiar with the area.

Thank you for your consideration.




Table 1 and 2 show the specific results of the study. “Days” indicates how many days it took to receive any type of reply. The last days of reply from the franchises were February 25, 2004 for NFL and NHL and March 28, 2004 for NBA and MLB, respectively. The total length of experiment time was 27 days and total average response time was 13.6 days. Of the four big leagues, MLB had the quickest response rate of any league (M=10.0 days), and the NHL had the tardiest response rate (M=16.6 days). Of the sent letters (N=122), only 63 franchises (52%) responded with any information. Specifically, 22 MLB teams (73%) responded, but only 9 NBA franchises (30%) sent some information.

Table 1

Result of the Response Time (Number of Days) to the Secret Shopper Letter

League/Team Days League/Team Days
Atlanta Falcons 14 Anaheim Mighty Ducks 14
Baltimore Ravens 14 Boston Bruins 13
Buffalo Bills 12 Chicago Blackhawks 20
Chicago Bears 14 Columbus Blue Jackets 21
Cincinnati Bengals 13 Dallas Stars 20
Cleveland Browns 12 Detroit Red Wings 25
Green Bay Packers 20 Minnesota Wild 14
Indianapolis Colts 20 Nashville Predators 14
Jacksonville Jaguars 13 Ottawa Senators 15
Kansas City Chiefs 13 Phoenix Coyotes 13
Minnesota Vikings 14 Pittsburgh Penguins 12
New Orleans Saints 13 San Jose Sharks 14
New York Jets 14 St. Louis Blues 16
San Diego Chargers 13 Tampa Bay Lightening 16
Seattle Seahawks 14 Toronto Maple Leafs 26
Tennessee Titans 17 Washington Capital 13
Average 14.3 Days Average 16.6 Days
Arizona Diamondbacks 13 Atlanta Hawks 6
Baltimore Orioles 21 Boston Celtics 5
Boston Red Sox 9 Charlotte Bobcats 8
Chicago Cubs 6 Chicago Bulls 6
Chicago White Sox 6 Dallas Mavericks 9
Cincinnati Reds 6 Houston Rockets 8
Cleveland Indians 13 Indiana Pacers 23
Colorado Rockies 5 New Jersey Nets 7
Detroit Tigers 6 Utah Jazz 20
Houston Astros 7 Average 10.2 Days
Kansas City Royals 5
Milwaukee Brewers 6
Montreal Expos 7
New York Mets 15
Oakland Athletics 5
Philadelphia Phillies 27
Pittsburgh Pirates 6
San Diego Padres 16
Seattle Mariners 16
St. Louis Cardinals 5
Tampa Bay Devil Rays 5
Toronto Blue Jays 19
Average 10.1 Days

Table 2

Summary of Each League’s Responding Rate

League Sent Responded % of Responses Average Response
NFL 32 Teams 16 Teams 50% 14.3 Days
NHL 30 Teams 16 Teams 53% 16.6 Days
MLB 30 Teams 22 Teams 73% 10.1 Days
NBA 30 Teams 9 Teams 30% 10.2 Days
Total 122 Teams 63 Teams 52% 13.6 Days

These 63 replies show a lower rate than in the previous study (65 replies out of 112 teams; 58%) (Bovinet, 1999). The responses were separated and rated as above expectations, average expectations, and below expectations based on the materials with which the teams replied (e.g., ticket plan, stadium/arena map, cover letter, waiting list application, merchandise catalog, parking information, schedule, business card to contact, and others). For example, if the teams provided more than enough and/or unexpected information such as ticket plans, merchandise catalogs, and fan guides were rated as above expectations. The other materials were considered average or below expectations. Each item of the responses was coded for convenience as follows:

Code A: Ticket plan/seating chart

Code B: Business card/person to contact

Code L: Cover Letter

Code M: Merchandise catalog/fan guide

Code P: Parking information

Code T: Team picture/individual player picture

Code W: Waiting list application

Code X: Stadium/Arena Map

Code Z: Schedule

Code AA: Other (e.g., sticker, key chain, individual phone

call/e-mail, etc.)

Only 15 teams were rated as above expectations (n=15; 12.3%), 32 teamsmet average expectations (n=32; 26.2%), and 16 teams were rated as below expectations (n=16; 13.1%).

Table 3

Category of Response Materials on Expectations

Above Expectations (n=15) Teams: Baltimore Raves (NFL)

Jacksonville Jaguars (NFL)

New York Jets (NFL)

Ottawa Senators (NHL)

St. Louis Blues (NHL)

Cincinnati Reds (MLB)

Cleveland Indians (MLB)

Kansas City Royals (MLB)

Milwaukee Brewers (MLB)

Pittsburgh Pirates (MLB)

Atlanta Hawks (NBA)

Chicago Bulls (NBA)

Dallas Mavericks (NBA)

New Jersey Nets (NBA)

Utah Jazz (NBA)

Types of Response Materials: A, B, L, M, T, X, AA

A, B, M, Interactive CD-Rom

A, M, Z, AA

A, B, L, M

A, B, M, P, AA, Trading Card

A, X, Z, AA, Player Drawing

A, B, L, AA

A, Z, AA, Promotional Flyer, Ticket Order Forms

A, L, Z, AA

A, B, L, Z, AA

A, L, Z, AA

A, L, M, AA

L, M, T, Z, AA, Folder, Pencil, Roster, Comic Book, Bookmark, Magazine, Postcard

A, B, L, Z, AA, E-mail/Phone call

A, B, L, M, Promotional Flyer, Ticket Order Form

Average Expectations (n=32) Teams: Buffalo Bills (NFL)

Chicago Bears (NFL)

Cleveland Browns (NFL)

Green Bay Packers (NFL)

Indianapolis Colts (NFL)

Kansas City Chiefs (NFL)

New Orleans Saints (NFL)

San Diego Chargers (NFL)

Seattle Seahawks (NFL)

Anaheim Mighty Ducks (NHL)

Columbus Blue Jackets (NHL)

Dallas Stars (NHL)

Detroit Red Wings (NHL)

Minnesota Wild (NHL)

Nashville Predators (NHL)

Phoenix Coyotes (NHL)

Pittsburgh Penguins (NHL)

Toronto Maple Leafs (NHL)

Washington Capitals (NHL)

Boston Red Sox (MLB)

Chicago Cubs (MLB)

Chicago White Sox (MLB)

Colorado Rockies (MLB)

Detroit Tigers (MLB)

Houston Astros (MLB)

Montreal Expos (MLB)

New York Mets (MLB)

Philadelphia Phillies (MLB)

Tampa Bay Devil Rays (MLB)

Toronto Blue Jays (MLB)

Charlotte Bobcats (NBA)

Houston Rockets (NBA)

Types of Response Materials: A, B, Z

M, W, AA

A, W, AA

L, M, AA

L, M, AA

A, M, Z

A, B, Z, AFL Schedule

T, Z, AA

A, Z, AA

A, M, Z

A, B, AA

A, M, AA

A, Z, AA

L, Z, AA

T, Z, AA

A, B, Z

L, Z, AA

L, Z, AA

A, Z, AA

A, Z, AA

L, Z, AA

A, Z, AA, Kid’s Club Information

A, Z, AA

A, B, Z

A, M, AA

A, T, AA

A, L, Z

A, Z, AA

A, Z, AA

A, B, Z, Ticket Order Form

B, M, AA

M, Z, Visitor’s Guide

Below Expectations (n=16) Teams (League): Atlanta Falcons (NFL)

Cincinnati Bengals (NFL)

Minnesota Vikings (NFL)

Tennessee Titans (NFL)

Boston Bruins (NHL)

Chicago Blackhawks (NHL)

San Jose Sharks (NHL)

Tampa Bay Lightening(NHL)

Arizona Diamondbacks (MLB)

Baltimore Orioles (MLB)

Oakland Athletics (MLB)

San Diego Padres (MLB)

Seattle Mariners (MLB)

St. Louis Cardinals (MLB)

Boston Celtics (NBA)

Indiana Pacers (NBA)

Types of Response Materials: B, W

A, L

Z (2003 Schedule)

M, Z (2003 Schedule and 2003 Fan Guide)



L, Z

T, Z

A, Z

A, Z



A, Z, Ticket Order Form

A, Z

L, Z


Discussion and Conclusion

Even though this was not a systematic and scientific statistical study, it was an exploratory team project of undergraduate sport management major students to reevaluate a useful concept of sport marketing in professional sport franchises. This study provided some evidence about how current professional franchises manage and treat potential customers. Although the original purpose of this study was simply to count the amount of time it took for the franchises to reply, the results provide some simple but significant directions to professional sport franchises including;

Develop and keep the positive relationship with your current and potential customers:

As Mullin et al. (2000) explained in the concept of the sport consumer escalator, one of the main responsibilities of sport marketers is to keep their loyal customers as heavy users and escalate lower level users into loyal customers. Contacting and relating with the customers is a first step to sell more products and services in the market.

Obey the customers:

Some franchises (e.g., New England Patriots, Los Angeles Lakers, Sacramento Kings, and New York Yankees) have sold out all tickets for this year and the following years, thus they may have felt it was not necessary to send ticket information to potential customers. However, even if some teams are sold out of their season tickets, they should send some kind of information with an apology or a letter of appreciation because the potential customer was interested in their products. This sort of communication will also buttress the positive image to potential customers who might become loyal customers in the near future because this relationship depends on unexpected situations such as relocation of a team and decreased winning rates and attendance rates. Depending on the communication style and attitude, customers may react positively or negatively.

Remember that selling tickets is not everything, but rather a cornerstone of selling the team products and services:

Although many teams find their main revenue sources in media and advertising, selling tickets is the reason behind those resources. Increasing gate receipt revenues is desirable not only for the revenues of concessions and novelties, but also for soliciting media and sponsorships.

Develop an information database of customers for a strong positive lifetime relationship:

Shani (1997) has described the differences between database marketing and relationship marketing, finding that the former has a short-term efficiency of marketing effort, while the latter retains customers who can provide the best lifetime value. Although database marketing is a necessary tool to conduct relationship marketing, it does not guarantee successful relationship marketing. Thus, building a database of customer information for relationship marketing is a first step to developing a long-term relationship with current customers. The database should include potential customers who might become heavy users of the products and services in the future. A database allows sport marketers to store information on current and potential customers and provide individualized marketing efforts (McDonald & Milne, 1997).

Realize the costs of serving long-life customers are less than other types of customers and result in willingness to pay higher prices:

Although some businesses spend more to serve long-life customers (Reinartz & Kumar, 2000), the efficiency of serving customers might increase from increasing revenues with long-term relationship customers in sport business. Cumulative revenues from the long-life customers are higher than other types of customers because of their long-term purchases.

One of the interesting results in this study was that MLB teams’ response rate (n=22; 73%) was higher than other leagues, while its response rate was the lowest in Bovinet’s study (1999). This result may be due to the different time frames within which the letters were sent and the different schedules of the leagues. Another interesting point of the study was that a team that is planning to relocate to another city showed great efforts to communicate with potential customers. For example, the New Jersey Nets sent the most information, including a cover letter, new ticket plan and game schedule, promotional event schedule, seating chart, and even made several calls and sent e-mails.

Although this student-organized team project focused on a simple facet of sport marketing, future efforts toward building and retaining customer relationships should be carefully reconsidered in the professional franchises. Keeping positive relationships with current and potential customers directly relates to the survival of any business.

In conclusion, this paper attempted to reevaluate relationship marketing efforts on the part of the major sport leagues and to consider a basic, but very important, marketing strategy in sport business. Although relationship marketing has established itself as an underlying paradigm in industrial and other service businesses, the concept and practice of relationship marketing in sport business is still in its infancy as a mainstream marketing concept. Therefore, professional sport organizations should establish not just a short-term transaction oriented objective, but also a long-term customer relationship goal.

Because of the differences between leagues, some limitations of this study and a recommendation need to be addressed. First, time selection of sending request letters for ticket information should be reconsidered because each league has a different season. For example, the NFL’s response rate (n=16; 50%) was lower than MLB’s (n=22; 73%) and the NHL’s (n=16; 53%) because some NFL franchises may have felt the season was just over and it was useless to spend time and money on mailings. Second, it is necessary to evaluate relationship marketing efforts of minor league professional sport teams because the core of their revenue rests on the gate receipts that rely directly on customer relationships.


  1. Bovinet, J. W. (1999) Customer communication in selected professional sports (MLB, NFL, NHL, NBA): A test. Sport Marketing Quarterly, 8(3), 41-44.
  2. Bowen, J. T. (2003). Loyalty: A strategic commitment. Cornell Hotel and Restaurant Administration Quarterly, 44(5), 31-46.
  3. Howard, D. R., & Crompton, J. L. (2004). Financing sport (2nd ed.). Morgantown, WV: Fitness Information Technology, Inc.
  4. James, J. D., Kolbe, R. H., & Trail, G. T. (2002). Psychological connection to a new sport team: Building or maintaining the consumer base? Sport Marketing Quarterly, 11(4), 215-225.
  5. McDonald, D. A., & Milne, G. R. (1997). A conceptual framework for evaluating marketing relationships in professional sport franchises. Sport Marketing Quarterly, 6(2), 27-32.
  6. Milne, G. R., & McDonald, M. A. (1998). Sport marketing: Managing the exchange process. Sudbury, MA: Jones and Bartlett Publishers.
  7. Mullin, B. J., Hardy, S., & Sutton, W. A. (2000). Sport marketing (2nd ed.). Champaign, IL: Human Kinetics.
  8. Ping, R. A., Jr. (1999). Unexplored antecedents of exiting in a marketing channel. Journal of Retailing, 75(2), 218-241.
  9. Neuborne, E. (2004). Play ball. Sales and Marketing Management, 156(1), 21.
  10. Reinartz, W. J., & Kumar, V. (2000). On the profitability of long-life customers in a non contractual setting: An empirical investigation and implications for marketing. Journal of Marketing, 64(4), 17-35.
  11. Shani, D. (1997). A framework for implementing relationship marketing in the sport industry. Sport Marketing Quarterly, 6(2), 9-15.
  12. Shani, D., & Chalasani, S. (1992). Exploiting niches using relationship marketing. The
  13. Journal of Consumer Marketing, 9(3), 33-43.
  14. Wulf, K. D., Odekerken-Schroder, & Lacobucci, D. (2001). Investments in consumer relationships: A cross-country and cross-industry exploration. Journal of Marketing, 65(4), 33-50.
2015-03-24T09:28:47+00:00March 7th, 2005|Contemporary Sports Issues, Sports Management, Sports Studies and Sports Psychology|Comments Off on A Secret Shopper Project: Reevaluation of Relationship Marketing Efforts

Colors and Cultural Interactions in the Turkish Sport Clubs


The purpose of this study was to determine the colors and color combinations used by the professional soccer clubs and to evaluate the proportion of the cultural interaction between the people and colors in Turkey. The study was carried out on 220 soccer clubs which have 480 colors and 220 color combinations evaluated. The study revealed that the combination mostly preferred by the professional soccer clubs in Turkish soccer leagues is red and white with n=26 (11.8%) followed by green and white with n=25 (11.4%). The color the most frequently observed in professional soccer clubs in Turkish soccer is white by 25.8%. This is followed by red by 18.1% and green by 12.5%. The colors preferred by the soccer clubs are general reflection of the cultural richness of the city, region or of the people they represent. This is especially the case for the clubs with deep rooted historical heritage.

The closed and distant environments are integrated and manifest themselves to the people with colors. A person perceives and identifies his or her environment with colors.

Colors have different meanings in different cultures. A certain color may closely be related to the internal worlds of some people, while it may symbolize a moral enlightenment for the others. The clarification of the role of colors in everyday life entails a comprehensive study. However, it is certain that they have different meanings according to the nation, region or culture involved (Allegos & Allegos, 1999).

For instance, black which symbolizes happiness for the Japanese is the indication of sadness and sorrow in many parts of the world. Yellow is a symbol of power in Chinese culture, where it indicates evil and disease for the Turks ( Ögel, 1991; Genç, 1999). Colors closely affect the feeling of people. People may have experience of feelings such as sorrow, happiness and stimulation through the usage of colors (Budzinski, 1986 ).

Psychologists believe that careful selection and use of colors may accelerate the desired response. This shows that the colors have strong relation with the emotional and physical state of the people. The people who chose black and white are thought to prone emotional problems more than the others, while red indicates viability or rage, yellow shows spaciousness and wideness, blue reflects calmness or protectionism and green means joy or stagnancy (Budzinski, 1986).

Colors are purposely utilized in many ways of life. Architects have utilized the effect of color upon the people for many years. For instance a light colored ceiling shows the room higher than it is and light yellow walls gives the impressions of spaciousness. The use of light green walls in classrooms makes the student to focus their attention on the subject and the use of red in factories and similar facilities increases the output. The use of white color in the medical clinics is believed to aggravate the sorrow and the use of light blue decrease it.

Colors also have a dominant role in commercial sales. For example, light colored cars give the impression of being more spacious than the dark colored ones. The dark colored small vehicles are found more repulsive than the light colored vehicles. The use of red on high bridges serves the purpose of preventing suicides with its stimulating effect. In discos the ground is illuminated with red in lively and rock type music and emotional and relaxing blue color in romantic and slow music (Budzinski, 1986; Hallalı & Nazil, 2001).

It can easily be concluded that the use of colors with or without purpose has certain influences on people. This may be due to both psychological and cultural reasons. In this context, the media prepared taking the stimulating and effect of colors has a very positive effect on the performance of the people and the relations among themselves ( Hallalı & Nazil, 2001).

It is obvious that the colors used in sports have various psychological effects upon the teams and their supporters. However, the colors used by the clubs representing a certain region or city may also have cultural implications. If the target of a sport club is successful and increasing its supporters, the marketability has a prime importance.

The facts that, the colors have stimulating effect on the morals of the people, represent certain social values, their effects on the psychology of people and their use outside the sports ground which creates a commercial value show that colors have an important role in achieving the targets of a club. Therefore, the correct use of colors which have such important impacts upon the lives and performance of the people requires the comprehension of their moral, visual and cultural impacts.

Color is described as phenomena resulted from the visual perception of the rays reflected from the matter ( Hallalı & Nazil, 2001) and radiation with a certain wave length within the visual region of the electromagnetic spectrum ( Active, 2000). Colors have been used for therapeutic purposes for so many years. It was realized that colors had psychological and physical effect upon the people ( Hallalı & Nazil, 2001).

The research shows that the most favorite color of the people is blue followed by red and green. The adult men prefer green, ocean blue, orange and dark purple, and adult women mostly like light purple colors. The favorite colors of children are blue, red, green, yellow and orange (Budzinski, 1986).

Colors are classified as fundamental colors (red, blue and yellow), mid colors obtained by the mixture of at least two colors (purple, orange and green) and neutral colors (white, black and grey) ( Hallalı & Nazil, 2001). The colors have different meaning and different effects.

These can be summarized as follows (Kalmık, 1964; Gabain, 1968; Crozier, 1999; Halis, 2000; Wright, 2000b):

Red. Red is a strong color, which stimulates viability, gives physical courage, has stimulating effect, makes visual impact, and may create stress and aggression. It is related with activation of the impulse of “fight or escape”.

Green . Green creates the feeling of love, relaxation, renewal, dependability, environmental consciousness, balance and peace.

Blue. It has a mainly a relaxing, concentration increasing, assisting in mental relief, and relieving effect. However, it may have a cold and senseless impression as well.

Yellow. It symbolizes excitement, self confidence, openness, friendship and creativity. Yellow is more rapidly perceived than the other colors, and eyes are more sensitive to it. This makes yellow one of the favorite colors.

White. It is the symbol of health and hygiene. It has gives the feeling of correctness and trust. However, it makes the eyes tired very fast.

Black. It reminds seriousness and dignity. However, it frequently implies sorrow, death and darkness.

Grey. It is colorless and creates no strong feelings.

Purple (Violet). It creates the feeling of pessimism, regret, fear and opposite feelings. It is perceived as the indication of power and divinity.

Orange. It gives the sense of brightness and warmth. It has a powerful activating effect. It creates the feeling of openness and increases the will to live.

Pink. It expresses kindness, softness, happiness and pleasure which are regarded as feminine feelings. That was why the soccer clubs refrain from using this color.

Brown. It induces seriousness, warmth, robustness and support. It mainly addresses the adults. It may occasionally result in melancholic feelings.

Starting from these let us evaluate the color and color combinations used by the soccer clubs for symbolic purposes and investigate their interactions and implications on people.


The research covers 220 soccer clubs playing in the 1 st, 2 nd and 3 rd divisions of Turkish soccer league for the last four seasons. Each sport club is counted once.

Following the investigation of 480 colors and 220 color combinations, their frequency (F) and percentage distributions (%) were evaluated by the use of SPSS 11.0 software. The shirt colors of 220 soccer clubs in Turkish soccer leagues were counted one by one, and than the most frequently used single color by the Turkish soccer clubs was determined. The color combination was also counted and their frequency (F) and percentage distributions (%) were determined.

In addition, the shirt colors of some soccer colors were taken as samples, and they were used for evaluation and comparison purposes.


According to the investigation of the color combination of the soccer clubs different Turkish soccer divisions between 2000 and 2004, Table 1 shows that, the most favorite combination is white-red with n=26 (11.8%) followed by white-green with n=25 (11.4%), white-blue with n=14 (6.36%), red-black with n=15 (6.82%) and red-violet with n=14 (6.36%).

[Table 1. About Here]

The number of colors was found as 480 since some clubs use more than two colors. As seen from Table 2, the most widely used color by the Turkish soccer clubs is white with n=124 (25.8%) followed by red with n=87 (18.1%) and green with n=60 (12.5%).

[Table 2. About Here]


The study revealed that the selection of the colors of the shirt of the clubs is based on incidents, meanings and symbols of life. This is especially the case for the clubs with historical roots. It is observed that the colors which are the indication of cultural values of a certain region, city or a group of people are reflected in the colors of the clubs. However, there are cases which are the opposite of this situation. However, the colors of most of the clubs are seen to be influenced by the value treasured by the region or city of the people. This may be attributed to geographical, local, national, moral and institutional factors. Let us briefly examine each one.Discussion

The study revealed that the selection of the colors of the shirt of the clubs is based on incidents, meanings and symbols of life. This is especially the case for the clubs with historical roots. It is observed that the colors which are the indication of cultural values of a certain region, city or a group of people are reflected in the colors of the clubs. However, there are cases which are the opposite of this situation. However, the colors of most of the clubs are seen to be influenced by the value treasured by the region or city of the people. This may be attributed to geographical, local, national, moral and institutional factors. Let us briefly examine each one.

Geographical Factors. The color of some of the sport clubs in Turkey was determined by taking the geographical location into account. For instance, blue-green colors of Rizespor are based of blue color of the sea and green color of the forests. Bursaspor took its green and white color from the snow of Uludağ Mountain and green of the Bursa plain.

Elazığspor (Claret red-White) took its colors from the claret red-white onyx unique to the region. The colors of Gençlerbirliği were taken from the color of corn poppy of Ankara steppes. Ankaragücü, on the other hand, took its violet and yellow color from the violet color of the famous “misket” grapes of the region where yellow was taken from the famous melon of Ankara as a symbol of power (Kalmık, 1964; MKE Ankaragücü Kulübü , n.d.).

Regional Factors. Diyarbakırspor (green and red) was founded by the merge of the amateur clubs Diclespor and Yıldızspor on 24 June 1968. The club took its green color from Yıldızspor ands red from Diclespor. However these colors are the colors of the out and inside of famous water melon of Diyarbakırspor (Diyarbakırspor, n.d.).

The red and white colors of Kahramanmaraşspor represent the two famous products of the city: ice cream and red pepper. Trabzonspor, on the other hand, took its claret red and blue colors from the scales and the eyes of the famous fish of the region, “hamsi” (Kalmık, 1964; Nntvmsnbc, 2002;, 2003;, 2004b).

National and Moral Factors. Red and black colors of Gaziantepspor represent the blood of 6314 people who died fighting against the French and Armenians during the Independent War and those dark days (Gaziantepspor, 2003). Antalyaspor inspired from the red and white colors of the Turkish flag (Nntvmsnbc, 2002).

The selection of black and yellow for Istanbulspor was to commemorate the students of Istanbul boy’s high school who lost their lives or got wounded in the World War I. Fenerbahçe changes its original colors of yellow and white due to its dislike towards the monarch regime and weakness of these colors. The club later selected yellow and navy blue as its color where yellow represents jealousy felt for Fenerbahçe and navy blue represents the dignity and nobility. The team also successfully uses white for cleanness and clarity, red for love and loyalty and green for success in its symbol and badges ( Kayserilioğlu, 2003;, 2004a).

The initial colors of Beşiktaş were red and white when it was first founded in 1903. However, it was turned into black and white till the ground lost in Balkan war was captured again. It also served to commemorate the ones who went to Balkan war and did not come back. However, the current perception of Beşiktaş’s colors is that winning and loosing are as natural as day and night (, 1995).

The selection of Konyaspor green and white colors was totally based upon moral reasons. The first colors of the club were black and white. However the black color was latter changed into green, the holly color of Islam. This may be attributed to the conservative and religious structure of Konya province.

Samsunspor added black to its original red and white color to remember its players who died in a tragic traffic accident in 1989 (Samsunspor Kulübü, n.d.).

Institutional factors. Siirt Köy Hizmetleri was founded by the Turkish Village Works Office and the club accepted the official color of yellow and blue of the office. Kardemir Demir Çelik Karabükspor, a team of the iron processing town of Karabük, took the red and blue color of melted iron (Batı Karadeniz Haber Ajansı, n.d.).

Yimpaş Yozgatspor (Red-Black) became the sister club of one of the popular teams of 1960s Karagümrük and accepted Karagümrük’s colors.

The investigation of 480 colors of 220 teams as regards to preference of colors showed that (Table 2) white ranks first with 25.8 %. Red comes second with 18.1% followed by green with 12.5%. Blue comes seventh with 7.71%.

The reasons that the white is the most preferred color among the Turkish clubs are that Turkish people have special liking to it and white has cultural implications coming from history. Turkish people regard white as the symbol of cleanness, purity, holiness and experience ( Ögel, 1991). The fact that the white is most widely used color by the Turkish clubs is the indication that it has upper cultural value. Among the sub cultural colors liked by the Turkish nation, red comes second (18.1%), green comes third (12.5%) and yellow comes sixth (8.96%).

In England and Scotland, the most popular color preferred by the soccer teams is blue. This is followed by red ( British Council, 2001). Although first choices are different there is a striking similarity in the second choices. The wide use of red in sports is probably due to its psychological effects.

The study revealed that the least popular color among Turkish soccer clubs is pink (0.21%). This is also the case for the other countries in the world. This is due to the fact that pink creates a soft and feminine image (Allegos & Allegos, 1999).

When we look at the color combinations of soccer clubs, the most popular combination among the Turkish soccer teams is red and white with 26 teams (Table 2).

According to a newspaper report, the most popular color combination among 205 Turkish soccer teams playing in Turkish divisions in 1998-1999 soccer season was green and white with 19 teams. This is followed by red and white with 19 teams and red and black with 13 teams ( Hallalı & Nazil, 2001).

According to research carried out on 43 European countries, the most preferred color combination among the soccer clubs is blue and white with 169 teams (Berke & Utku, n.d.). The research stated that the choice of color combinations after blue and white was red and white (123 teams), red and blue (58 teams), black and white (56 teams), red and black (50 teams), yellow and violet (48 teams) and green and white (46 teams).

According to data listed in Table 2, the most preferred color combinations in Turkish soccer teams after red and white are green and white with 25 teams, blue and white with 14 teams, red and black with 15 teams and red and violet with 14 teams.

This shows that there is a difference of preference between the club colors in Europe and Turkey. This complies well with the fact that the choice of club colors is shaped according to the different symbols and meanings of colors from country to country ( British Council, 2001).

Colors show different cultural meanings in different atmosphere. But generally, it’s similar to the psychological effects of the colors which occur on people. The people in white colored room were observed to be much more restless and leave the location earlier than the people in brown colored room. In the same example the people were given coffee from thermoses with four different colors. They described the coffee from the brown thermos as “dark and strong”, the coffee from red thermos as “sweet”, the coffee from blue thermos as “light with good aroma” and the coffee from the yellow thermos as “bad” (Budzinski, 1986 ).

It was also observed that the pulse rate was accelerated under red light while blue illumination relaxed the muscles, rectified the blood pressure and decreased the pulse rate ( Wright, 2000a).

The firm knowledge psychological effects of colors (, 2003), the effect of colors upon the fans, their ability to establish a cultural value and their capacity of creating commercial activity is of great importance in the selection of the color for the newly founded clubs. It is known that one of the biggest factors for the marketability of a club is strong and visually effective colors (Allegos & Allegos, 1999).


In conclusion, we stress following points. Colors are very important due to their physical effect they induce on people. One of the prime factors which play a determining role on the selection of the clubs by the children is attractive and strong colors (Allegos & Allegos, 1999).

The use of colors with cultural significance may create a feeling of possession and support upon the people. For instance, the claret red-blue colors of Trabzonspor are based the color of the eyes and scales of special fish called “hamsi”, the major product of the city.

The colors to be selected must also be suitable to be used in daily outfits. This will create an additional economical value. The yellow and red colors of Galatasaray are observed almost in every walk of life and became a part of our daily culture. The use of psychological power of red and yellow is important for both to create an economical value and strengthen the support given to the club.

Colors are similar to musical notes. As jazz pianist Thelonius Monk said “there is no wrong note”. What is important is the selection of the appropriate note or color (Wright, 2000c).

We can conclude that the colors of the clubs are the reflectance of the cultural and historical aspects of the people living in different city or region. Therefore the newly established teams must choose their colors taking all these criteria into account.

The soccer clubs with national and international support not only utilize the effect of the colors they selected but they also make the effective use of their symbols. The canary of Fenerbahçe, the lion of Galatasaray and the eagle of Beşiktaş are good examples for that. The investigation of the effects of symbols is the subject of another study.


  1. Active. ( 2000, December). Renkler yalnızca bir ışık mı? Hayır [Are colors just a light?]
  2. [Electronic version]. Bankacılık ve Finans Dergisi, 9. Retrieved June 18, 2003, from
  3. Allegos, A., & Allegos, H. (1999). Colour does matter! An investigation of colour in sport
  4. [Electronic version]. Cyber-Journal of Sport Marketing, 3 (2). Retrieved March 31, 2003, from
  5. Batı Karadeniz Haber Ajansı. (n.d.). Kardemir demir çelik karabükspor t arihçesi [History of the kardemir celik karabukspor ]. Retrieved August 11, 2003, from
  6. Berke & Utku. (n.d.). Forma renklerine göre takımlar ve ülkeleri [ Teams and countries in respect of colors of form ]. Retrieved August 11, 2003, from
  7. (1995). Renkler[Colors]. Retrieved August 11, 2003, from
  8. British Council. (2001). Colours. Retrieved September 5, 2003, from
  9. Budzinski, K. (1986, October). Biz mi renkleri etkiliyoruz renkler mi bizi? [Whether do we effect the colors or do they effect us? ]. Bilim Teknik Dergisi, 227, 6-9.
  10. Crozier, W. R. (1999). The meanings of colour: preferences among hues, Pigment & Resin Technology28(1), 6. Retrieved August 11, 2003, from ProQuest Research Library.
  11. Diyarbakırspor Kulübü. (n.d.). Tarihçe [History ]. Retrieved August 11, 2003, from
  12. Eray, F., & Civitci, Ş. (2000). Psychology effects of garment colors on human. Gazi University Vocational Education Faculty Journal of Vocational Education, 2(3). Retrieved August 11, 2003, from
  13. Gabain, A.V. (1968). Renklerin sembolik anlamları [Symbolical meaning of colors], (S. Yazgan, Trans.). Türkoloji Dergisi, 2(1), 107-113.
  14. Gaziantepspor Kulübü. (2003). Gaziantepspor’ un tarihçesi [History of gaziantepspor ]. Retrieved August 11, 2003, from
  15. Genç, R. (1999). Türk inanışları ile milli geleneklerinde renkler ve sarı kırmızı yeşil[Colors in turkish believs and customs and yellow, red, green]. Ankara: Atatürk Kültür Merkezi Başkanlığı.
  16. Halis, M. (2000). İş yaşamı kalitesi açısından çalışma ortamında insan-renk etkileşimi [Human being-color interaction in working environment as to quality of business life ]. Verimlilik Dergisi, 2, 64-68.
  17. Hallalı, M., & Nazil, H. (2001). İş ortamında renklerin önemi [Importance of colors in working environment]. Standart Dergisi, 473, 64-70.
  18. Kalmık, E. (1964). Renklerin armoni sistemleri [Harmonical systems of colors ], 38-39. İstanbul: İTÜ Matbaası.
  19. Kayserilioğlu, S. (2003). Kuva-i milliye ruhunun halk içindeki sembolüyüz [We are symbols of soul of Kuva-i milliye in the people]. FB Dergisi, 3, 110-113.
  20. (2004a). Fenerbahçe amblem ve renkleri [Emblem and colors of fenerbahce]. Retrieved August 11, 2003, from
  21. (2004b). Trabzon’da futbolun doğuşu [Rising of soccer in trabzon]. Retrieved August 11, 2003, from
  22. MKE Ankaragücü Kulübü . (n.d.). Tarihçe[History]. Retrieved August 11, 2003, from
  23. (2003). Renkler satışları etkiliyor [ Do colors effect the selling]. Retrieved July 08, 2003, from mainPage=1&which= center&no= 1&ref=haberHome&date=08Haziran2003
  24. Nntvmsnbc. (04.19.2002). Gönül verilen renkler [Colors fallen in love]. Retrieved August 11, 2003, from
  25. Ögel, B. (1991). Renkler ve Türkler [ Colors and Turks ]. Türk kültür tarihine giriş[Introduction to the history of Turkish culture ], (Vol.6, pp.377-491). Ankara: Kültür Bakanlığı Yayını.
  26. Samsunspor Kulübü. (n.d.). Samsunspor kulübünün kuruluşu [Establishment of samsunspor ]. Retrieved August 11, 2003, from
  27. Sevim, A. (2000). Futbolun bin bir yüzü[A great many face of soccer]. Retrieved August 11, 2003, from
  28. Yeşil-Beyazlılar Birinci Sırada [ Green- whites are first]. (1998, October 4). Milli Gazete. Retrieved August 11, 2003, from
  29. Wright, A. (2000a). How does it work?. Retrieved August 11, 2003, from
  30. Wright, A. (2000b). Psychological properties of colours. Retrieved August 11, 2003, from
  31. Wright, A. (2000c). Colour affects system. Retrieved August 11, 2003, from

Table 1

Color Combination of Turkish soccer clubs


Combination 1. White Combination 2. Red Combination 3. Green Combination 4. Violet Combination 5. Black Combination 6. Yellow Combination 7. Blue
Colors f % Colors f % Colors f % Colors f % Colors f % Colors f % Colors f %
Red 26 11.8 Green 10 4.55 Black 6 2.73 Yellow 9 4.09 Yellow 9 4.09 Blue 2 0.91 Orange 3 1.36
Green 25 11.4 Black 15 6.82 Yellow 8 3.64 Blue 4 1.82 Blue 1 0.45 Lilac 1 0.45 Claret red 2 0.91
Black 11 5 Violet 14 6.36 Blue 4 1.82 Orange 1 0.45 Orange 3 1.36
Violet 10 4.55 Yellow 11 5 Orange 3 1.36 Purple 1 0.45
Yellow 1 0.45 Blue 5 2.27 Pink 1 0.45 Turquoise 2 0.91
Blue 14 6.36
Orange 2 0.91
Claret red 8 3.64
Purple 7 3.18
Lilac 1 0.45
TOTAL 220 Team

Table 2

Color Ratios of the Turkish soccer clubs

White 124 25.8
Red 87 18.1
Green 60 12.5
Black 47 9.79
Violet 44 9.17
Sari 43 8.96
Blue 37 7.71
Orange 13 2.71
Claret red 10 2.08
Purple 8 1.67
Turquoise 4 0.83
Lilac 2 0.42
Pink 1 0.21
Total 480 100
2015-03-24T09:12:34+00:00March 6th, 2005|Contemporary Sports Issues, Sports Management, Sports Studies and Sports Psychology|Comments Off on Colors and Cultural Interactions in the Turkish Sport Clubs

Intelligence and Football: Testing for Differentials in Collegiate Quarterback Passing Performance and NFL Compensation


This article presents an empirical analysis of the relationships between intelligence and both passing performance in college and compensation in the National Football League (NFL). A group of 84 drafted and signed quarterbacks from 1989 to 2004 was selected for the study. The author hypothesizes that intelligence is the most important and perhaps most rewarded at this position, and a wide variety of passing performance statistics are available to separate the effects of intelligence and ability. The OLS-estimated models reveal no statistically significant relationship between intelligence and collegiate passing performance. Likewise, the author finds no evidence of higher compensation in the NFL for players with higher intelligence as measured by the Wonderlic Personnel Test administered at the NFL Scouting Combine.


Every February hundreds of collegiate football players gather in Indianapolis at the NFL’s scouting Combine, a four-day event designed for NFL scouts to evaluate the talent of the year’s draft eligible players. Many players likely to be drafted in the first round of the NFL draft will refrain from taking the skills and agility tests in Indianapolis, and will schedule “pro days” at their university where they are more comfortable with the atmosphere and expected results of their workouts. All prospects at the Combine undergo X-rays and physicals to address current and past injuries. All players will also take a 12-minute test designed to measure intelligence, the Wonderlic Personnel Test.

First used by a handful of teams in the 1970s, the Wonderlic is a 50-question test designed to be taken in 12 minutes to measure the athlete’s general intelligence (ESPN, 2002; FairTest Examiner, 1995). The score is calculated as the number of questions correctly answered in the allotted time. As a matter of practice, most players do not answer all 50 questions in the time allotted. Wonderlic scores vary by position, though NFL draftees have averaged a score of 19 over the last 20 years. A Wonderlic score of 20 indicates the test taker has an IQ of 100, which is the average intelligence ( The Wonderlic website states that higher scores mean higher intelligence, and intelligence has an impact on playing style and leadership, especially for quarterbacks. The accuracy of this contention will be empirically tested in this article.

This article will also examine the intelligence in two additional models to determine relationship between intelligence and a player’s rookie year compensation and the relationship between intelligence and where an athlete is selected in the NFL draft. This article is particularly relevant in the modern draft era because the millions of dollars spent scouting and signing draftees represent a significant investment by NFL franchises. To the extent that intelligence has an effect on passing ability, NFL franchises may be willing to reward players with such mental abilities. However, if no relationship exists between tested intelligence and performance, then NFL franchises can better utilize resources by focusing on other aspects of player evaluation.

Methodology and Data Sources

The player data for this study were collected from, various NFL draft prospect websites and official university athletics sites. The data include information on 84 drafted NFL quarterbacks who received rookie year salaries between 1989 and 2004. 1 However, due to the limited public availability of Wonderlic scores prior to 1999, most of the quarterbacks used in this study were drafted in the last six years. Although the test results are not officially released, in recent years the Wonderlic scores have been available on the Internet at for many players at the Combine. Salary data were obtained from and between 2001 and 2004 and from USA Today newspaper clippings prior to 2001.

Comparing the distribution of the data used in this study to the total current population of NFL quarterbacks, the author finds a similar proportion of non-white quarterbacks (about 20 percent in each), but a slightly higher proportion of Division 1A quarterbacks in the data (89 percent compared to 80 percent). Such comparisons are important as intelligence tests are frequently found to generate sizable ethnic differences, and such biases could affect the results and interpretation of the model if the sample data is not a representative subset of the true population of NFL quarterbacks (FairTest Examiner, 1995).

Modeling Intelligence and Collegiate Passing Performance

The first relationship this paper addresses is the relationship between intelligence and quarterback passing performance. Using the same set of independent variables in each model, an equation is estimated for each of the dependent variables to measure a quarterback’s passing performance, career passing efficiency (Model I) and total offense per game (Model II). The quarterback’s best collegiate year in terms of total offense per game is used in this analysis.

NFL scouts and coaches assign draft grades to players based on collegiate production and ability. In particular, scouts highly value attributes such as height, quickness, arm strength, vision, leadership and intelligence (CNN/SI, 1998). As such, we develop a model based on the expected contributions of these characteristics (where quantifiable) to a quarterback’s passing performance.

We expect that height may have a positive relationship with passing performance. Taller players can better read defenses, find receivers and avoid having passes deflected by defensive lineman. We expect that a player’s quickness (lower 40-yard dash times) will have a positive relationship with total offense but no effect on passing efficiency. Likewise, NFL scouts expect the quality of a quarterback’s offensive peers to have a positive effect on his passing performance. The models include a DRAFTCLASS variable to control for the general quality of the quarterback’s senior class. DRAFTCLASS is a count variable of players drafted into the NFL during the quarterback’s senior year.

There are no a priori expectations for the relationship between intelligence (as measured by WONDERLIC) and passing performance. Likewise, there are no a priori expectations for the relationship between passing performance and Division 1A football. There are no a priori expectations for the relationship between a quarterback’s race and his passing performance.

The models developed in this section are estimated by OLS without a constant. All dependent and independent variables have been transformed and are centered on the mean of the observed variable to aid in the interpretation of the marginal effects of each explanatory variable. In particular, all effects are for a quarterback with characteristics at the mean, the values of which are shown in Table 1.

Table 1: Summary Statistics of Dependent and Independent Variables

Variable Mean Std. Dev. Min Max
HEIGHT (inches) 74.80 1.45 71.38 77.63
FORTY (seconds) 4.84 0.18 4.36 5.37
DRAFTCAST (# of drafted offensive teammates) 3.08 2.20 0.00 7.00
DRAFTCLASS (# of drafted players from team) 3.69 2.50 1.00 14.00
WONDERLIC 25.45 7.13 10.00 42.00
Division 1A, 1=yes 0.89 0.31 0.00 1.00
Race, 1=non-white 0.20 0.40 0.00 1.00
NCAA (career NCAA passing efficiency) 135.91 12.09 104.35 168.82
TOPGB (total offense per game, best year) 273.57 64.94 109.36 527.20
DRAFT 108.76 82.77 1.00 250.00
REALSAL 769,863 808,343 30,515 2,735,854
NFL ROOKIE RATNG 69.06 27.86 8.80 140.20

As shown in Table 2, the hypothesized model for estimating passing performance does a considerably better job explaining TOPGB than NCAA. In fact, the F Value for the Model I is not statistically significant, suggesting that within the sample of players in this dataset, the model has no explanatory power. This may be the case with a relatively homogenous group of individuals – because all players are similar in ability levels, insignificant variation may exist to accurately estimate the model. This is not the case in Model II where significant variation exists within the TOPGB variable, reflecting not only the quality of the team and the opponents, but also the style of offensive attack.

The expected relationship between a player’s HEIGHT and his passing ability is not found in the passing efficiency model (Model I), suggesting that the collegiate passing performances of this group of NFL-caliber quarterbacks do not vary with height. However, RACE is statistically significant in both models, suggesting that non-white quarterbacks are better than their white peers in terms of passing efficiency and total offense per game. In particular, non-white quarterbacks average 8.9 points higher in passing efficiency and 44.8 more total offensive yards per game than their white peers. DRAFTCAST, the quality of the quarterback’s offensive peers, as measured by the number of offensive teammates drafted during his collegiate career, is statistically significant in the total offense per game model. This effect may reflect the self-selection of better quarterbacks into the top collegiate programs known to produce many NFL-caliber prospects. DRAFTCAST is a count variable with a maximum value of seven. The expected negative relationship is found because in the absence of other offense threats, the team will naturally rely more on the abilities of the quarterback. The estimated coefficient has the interpretation that for each NFL-caliber offensive teammate, a quarterback will average 14 fewer total offensive yards per game.

The coefficients in Model I and Model II reveal no statistically significant relationship between a quarterback’s Wonderlic score and his passing performance in college. It should be noted, however, that considerable differences in both terminology and depth exist between collegiate and NFL playbooks, and a quarterback’s intelligence may affect his passing performance in the NFL even if it does not in college. Model V developed in the next section will test if a quarterback’s intelligence affects his NFL passing efficiency during his rookie year.

Table 2: Passing Efficiency and Total Offense models

(Model I) (Model II)
HEIGHT (inches) 0.439 -4.098
(0.44) (0.84)
FORTY (seconds) 4.57 71.917
(0.52) (1.68)*
DRAFTCAST (# of drafted offensive teammates) 0.621 -14.083
(0.8) (3.73)***
DRAFTCLASS (# of drafted players from team) 0.545 0.419
(0.81) (0.13)
WONDERLIC 0.043 0.366
(0.21) (0.36)
Division 1A, 1=yes -1.115 4.99
(0.66) (0.61)
Race, 1=non-white 8.886 44.836
(2.17)** (2.26)**
Observations 84 84
F Value 1.37 4.58***
R-squared 0.11 0.29
Adjusted R-squared 0.03 0.23

Absolute value of t statistics in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%

Modeling Intelligence and the NFL draft

This section focuses on the relationship between intelligence and quarterback draft position (Model III) and compensation (Model IV). The player’s rookie year salary (REALSAL) is his first-year compensation in 2004 dollars against the salary cap. Despite the fact that the structuring of contracts varies greatly between first-round picks and all other selections, the player’s rookie year salary reflects the team’s relative willingness to pay for a rookie quarterback given the availability of free agent quarterbacks and the limits of the NFL salary cap and team-specific rookie cap. 2 In recent years, NFL franchises have increasingly used incentive clauses, deferred and option bonuses in addition to traditional signing bonuses to defer the cost of signing first-round draft picks. Such a structuring of contracts also insulates the franchises against the risk inherent in both the evaluation of athlete’s abilities and the likelihood of serious injury. The quarterback’s draft position (DRAFT) is his selection number in the draft (e.g., a quarterback selected with the eighth choice in the second round (40th overall pick) would have a draft position of 40.)

If reality does indeed reflect what the previous models suggest (that intelligence has no affect on passing ability), then it is doubtful that a player’s Wonderlic score will affect the dependent variables – DRAFT and REALSAL – in Model III and Model IV. To the extent that his past performance is indicative of his true ability as a quarterback (and there is great debate on this issue), then the quarterback’s draft position and rookie year salary should primarily be explained by themeasures of his collegiate passing performance and the expected ease of his transition to the professional ranks. The models estimated below reveal that only one of the two measures of collegiate passing performance, TOPGB, is statistically significant in both Models III and IV.

The coefficients of the Models III and IV should be of opposite signs due to the inverse relationship between draft position and a player’s salary. This is indeed observed in the comparison of the two models as shown in Table 3. The signs of the coefficients generally reflect the a priori expectations, with some interesting exceptions. Note that while quarterbacks benefit substantially from the number of collegiate NFL-caliber offensive teammates in Model III and Model IV, Model IV suggest that franchises do make efforts to account for the general quality of the player’s offensive (e.g., offensive cast teammates include wide receivers, running backs, and tight ends, but exclude the offensive line) and defensive teammates as reflected in the negative coefficient of DRAFTCLASS. Both DRAFTCAST and DRAFTCLASS represent competing peer effects in the model. While the estimated DRAFTCAST coefficient is of similar sign and significance in previous literature (Mirabile, 2004), it should be emphasized that the datasets for this and the previous study were considerably different, in particular the latter was composed of both drafted and non-drafted athletes.

Table 3: Draft, Salary, and NFL passing models

(Model III) (Model IV) (Model V)
HEIGHT (inches) -14.55 216,192 -4.19
(2.36)** (3.99)*** (1.49)
FORTY (seconds) 200.83 -2,304,356 22.98
(3.68)*** (4.81)*** (0.97)
DRAFTCAST (# of drafted offensive teammates) -12.04 128,819 -2.41
(2.24)** (2.73)*** (0.94)
DRAFTCLASS (# of drafted players from team) 6.15 -80,140 -0.50
(1.50) (2.22)** (0.28)
WONDERLIC -1.08 -1409 -0.57
(0.85) -0.13 (0.96)
Division 1A, 1=yes -39.30 290396 8.16
(3.81)*** (3.20)*** (1.64)
Race, 1=non-white -6.47 -273,860 -2.89
(0.25) -1.21 (0.26)
TOPGB -0.53 5,155 -0.04
(3.27)*** (3.63)*** (0.48)
NCAA -0.21 12,598 0.45
(0.27) (1.84)* (1.16)
Observations 84 84 61
F Value 7.42*** 8.85*** 0.82
R-squared 0.47 0.51 0.12
Adjusted R-squared 0.41 0.46 -0.03

Absolute value of t statistics in parentheses
* significant at 10%; ** significant at 5%; *** significant at 1%

The coefficient on the quarterback’s Wonderlic score is not statistically significant, suggesting that NFL franchises do not select smarter quarterbacks sooner or compensate them better than their peers, ceteris paribus. The regression results show that quarterbacks are compensated not only for than their collegiate passing performance, but for how the quarterback is expected to perform in the NFL. Note these expectations take the form of significant coefficients for a player’s height, time in the 40-yard dash, and level of past competition (Division 1A). Such traits are highly prized for rookie quarterbacks, whose ultimate success is determined by their ability to adjust to the size and the speed of NFL opponents. The conclusion from these coefficients is that either intelligence is not an important factor in drafting and compensating rookie quarterbacks or that concerns about a quarterback’s intelligence raise flags which are consistently investigated and subsequently lowered before the draft. Although the models revealed no compensation for smarter players at the quarterback position, such compensation may indeed exist at other positions where such a wide variety of performance statistics are not readily available.

As suggested previously, a quarterback’s intelligence may affect his passing performance in the NFL even if it does not in college. Model V uses the same group of independent variables utilized in previous models to explain the dependent variable, NFL ROOKIE PASSING, the quarterback’s NFL rookie passing efficiency rating. The formula for passing efficiency in the NFL is different from its NCAA counterpart, with NFL passing efficiency employing different scales and a capped system. Because many rookie quarterbacks receive little or no playing time, this variable is not perfectly observed and could be subject to significant measurement error. Only 61 of the 84 observations have observed passing efficiency statistics during their rookie year. In fact, due to the wide variation in playing time among rookie quarterbacks, the variance in this measure is quite large. The results show a negative but not statistically significant relationship between passing efficiency in the NFL and intelligence as measured by the Wonderlic test.

Discrimination in the NFL Draft

In this section, we will briefly address the literature’s history of NFL discrimination and test whether any evidence of discrimination exists at the quarterback position within the developed models. As has been noted in previous literature (Kahn, 1992), white athletes have historically benefited from a race premium. However, using data from the 1996 season, Gius and Johnson (2000) found evidence of reverse discrimination in the NFL with whites earning 10 percent less than their black peers, results they believe were not previously found because prior investigations used data from the 1970s and 1980s.

We can utilize a group means t-test to determine whether the observed distributions of REALSAL and DRAFT are statistically different for white and non-white quarterbacks. After employing t-test of each of these variables, we can reject the hypothesis of equal means for DRAFT but not REALSAL. In particular the mean DRAFT position for white quarterbacks was 117.3 (with a standard error of 10.2) and the mean DRAFT position for non-white quarterbacks was 75 (with a standard error of 17.5). Although this result is significant at the five percent level, given the relatively small sample size, we should be hesitant to infer the existence of discrimination. Model III and Model IV are better able to test for evidence of discrimination against non-white quarterbacks by controlling for all other differences between the two groups. In fact, neither model reveals evidence of such discrimination directly, though this result may change if the time period of the study were different.

Summary and Conclusions
The market for NFL rookie quarterbacks was examined between 1989 and 2004. Attempts to model passing performance using player and team characteristics revealed statistically significant relationships between a quarterback’s collegiate passing performance and his race and teammates. Intelligence, as measured by the Wonderlic score, was statistically insignificant. Likewise, while expected relationships were found between collegiate passing performance and NFL rookie year salary, the author found no statistically significant relationship between intelligence and compensation or intelligence and draft number after controlling for passing ability. Although the models revealed no compensation for smarter players at the quarterback position, such compensation may indeed exist at other positions where such a wide variety of performance statistics are not readily available. Future studies may endeavor to control for more of the franchise- and league-specific factors that impact the drafting and compensation of collegiate athletes.

This article presents empirical evidence that within the modern draft era, there exists no statistically significant relationship between intelligence and quarterback performance at either the collegiate or professional level. Likewise, more intelligent quarterbacks are neither selected earlier nor compensated more for their mental abilities. Since no statistically significant relationship exists between tested intelligence and performance within the data examined in this study, NFL franchises might better utilize resources by focusing on other aspects of quarterback evaluation.


1. This distinction is made because there are groups of quarterbacks who are drafted but not signed, and likewise a group of undrafted free agent quarterbacks who do sign NFL contracts in their rookie years. The group analyzed in this paper is composed only of players who are both drafted and signed.

2. The rookie cap, the amount of salary cap dollars available to sign rookies, is determined by the number and placement of each team’s draft selections in a given draft.



  1. CNN/SI. (1998) “What NFL teams look for in a player,” CNN/Sports Illustrated. (accessed March 7, 2005).
  2. Duberstein, M.J. (2002) NFL Economics Primer 2002 , National Football League Players Association. (accessed March 7, 2005).
  3. Duberstein, M.J. (2003) Pipeline to the Pros , National Football League Players Association. (accessed March 7, 2005).
  4. ESPN. (2002) “So, how do you score?” (accessed March 7, 2005).
  5. FairTest. (1995) “Testing Pro Football Players.” FairTest Examiner. April 5, 2005).
  6. Gius, M. & Johnson, D. (2000) “Race and Compensation in Professional Football.” Applied Economics Letters, 5, 703-705.
  7. Kahn, L.M. (1991) “Discrimination in Professional Sports: A Survey of the Literature.” Industrial and Labor Relations Review , 44, 395-418.
  8. Kahn, L.M. (1992) “The Effects of Race on Professional Football Players’ Compensation.” Industrial and Labor Relations Review , 45, 295-310.
  9. Mirabile, M. (2004) “The Peer Effect in the NFL Draft.” The Sport Supplement, 12(3). (accessed April 5, 2005).
  10. USA Today. (2005) “ – Football salaries database.” (accessed March 7, 2005).
  11. Whittingham, R. (1992) The Meat Market: The Inside Story of the NFL Draft, New York: MacMillan Publishing Company.
  12. Wonderlic, Inc. (2004) “How Smart is Your First Round Draft Pick?” (accessed December 10, 2004).

Author Information

McDonald P. Mirabile

2747 South Glebe #405
Arlington, VA 22206


2017-07-06T07:51:13+00:00March 5th, 2005|Contemporary Sports Issues, Sports Management, Sports Studies and Sports Psychology|Comments Off on Intelligence and Football: Testing for Differentials in Collegiate Quarterback Passing Performance and NFL Compensation

NBA Salaries: Role Players and Superstars


This study uses quantile regressions to evaluate salaries of NBA players in the 2001–2002 season separately for big men and guards. Quantile regression allows the measurement of the return to player attributes for guards and big men at different salary (skill) levels along the distribution of NBA salaries.

І. Introduction

The salaries of National Basketball Association (NBA) players constitute the primary component of costs of NBA franchise operation. Despite this, there is little theoretical or empirical groundings for the salary determining process of players. This analysis estimate players’ salaries, separating players into two groups: big men (forwards and centers in the League) and guards. Additionally, the study utilizes quantile regression procedures to estimate salaries of big men and guards. Quantile regression allows the measurement of the effect of performance attributes on players’ salaries at different points along the salary distribution. Indeed, the returns to attributes may also differ greatly by player position for low-salary, low-skill, bench-warmers on a team relative to high-salary, high-skill superstars.

II. Data and Results

Individual performance statistics of the 409 NBA players listed in the end of the 2001–2002 season rosters of the 29 NBA franchises were taken from the official website of the NBA. Performance statistics are utilized for games played during the regular season, excluding the preseason and playoffs.

Table 1 presents OLS estimation of earnings of NBA players in the 2001–2002 season. The coefficient on exp in column 1 reveals that each year of experience significantly increases earnings of NBA players by 36.8 percent. Findings of large returns to experience are expected given that the individual player salary cap is based on years of experience in the league (Staudohar, 1999 ). An increase of 1 assist per game significantly increases earnings by 7.4 percent. Additionally, an increase in average block shots of 1 per game results in significant earnings increase of 16.3 percent whereas, an increase in the average number of points scored per game by 1 point significantly increases salaries of NBA players by 4.4 percent. Earnings of blacks differ insignificantly from their white counterparts. These findings are consistent with recent evidence of no employer earnings discrimination of NBA players (Kahn, 2000). The findings for the coefficient on big-men reveal the NBA premium for centers and forwards from their size that is not a result of observed performance statistics. Big men significantly earn 18 percent more than guards, after accounting for other performance attributes.

Tables 2 and 3 present the results of the quantile regressions for guards and big men, respectively. The coefficient on exp is large and significant for guards at each skill level; however, the coefficients are much larger for players at intermediate skill levels relative to low-skill players and superstars at the tails of the distribution. Indeed, the returns to each year of NBA experience is 52, 51, and 53 percent for guards at the 25 th, 50 th, and 75 th quantiles, respectively — but, only 34 percent for low-skill players at the 15 th quantile and superstars at the 90 th quantile. This finding indicates greater returns to human capital accumulation for intermediate level, role players (occurring in the middle of the earnings distribution) relative to low-skill players and superstars. This could reflect team owners’ willingness to reward guards who are role players for doing the subtle things necessary to win that are not accounted for in performance statistics — for instance, diving on the floor to save a ball from going out of bounds.

Low-skill guards (15 th quantile) are unique in being significantly rewarded for steals (a 1 percent increase in steals per game increases their earnings by 62 percent) and discounted for rebounds (a 1 percent increase in rebounds per game decreases earnings by 13 percent). The negative returns to rebounding for low-skill guards may appear odd at first glance; however, in the transition from offense to defense, one of the primary responsibilities of guards is to be the first players back on defense. Thus, guards prevent the other team from scoring fast-break points. If guards remain on their offensive end of the court for rebounds, the opposing team is likely to score easy fast-break points. Moreover, low-skill guards’ decreased returns to rebounding may indirectly capture the negative effects for rebounding and not getting back on defense when the opposing team fast-breaks. Furthermore, the negative effect diminishes as skill level increases — possibly indicating that higher skilled players are more athletic or have superior decision-making abilities, which allows them to discern the opportune times to remain on the offensive end for rebounds versus getting back on defense. Scoring points is significantly rewarding for guards at all levels of the earnings distribution.

Table 3 presents the regression results for the quantile earnings estimates for big men. Similar to guards, the returns to each season of NBA experience for big men is greater for players at intermediate skill levels relative to low-skill players and superstars at the tails of the distribution. Additionally, blocking shots is a very lucrative attribute for intermediate and high-skill big men, but not for their lower skill counterparts—medium- and high-skill big men receive a significant premium (between 15.6 and 25.7 percent for a 1 percent increase in blocks per game) but the returns for low-skill big men are smaller in magnitude and insignificant. In contrast, low-skill big men receive the highest returns to rebounding (12.6 percent) — with the returns to rebounding declining in magnitude and significance as skill level increases (a significant increase of 10.2 and 7.4 percent increase at the 25 th and 50 th quantiles, respectively, and insignificantly increasing at higher skill levels). Low-skill big men (in the 15 th quantile) are unique as the only big men who are significantly rewarded for assists or penalized for increased average steals. A 1 percent increase in average assists increases their earnings by 16.6 percent. However, a 1 percent increase in average steals reduces their earnings by 10.4 percent. A possible explanation for low-skill big men’s salary discount for steals is that in a half-court defense, big men are often positioned near the basket — constituting the last line of defense in preventing easy lay-ups for the opposing team. If a big man goes for a steal (rather than staying in between the basket and the offensive player) the big man may be unsuccessful in stealing the ball — leaving the offensive player near the basket for an uncontested lay-up. Moreover, the low-skill, big-men discount for steals may reflect the negative effects of going for steals, but failing to steal the ball. Also, the negative effects diminish as skill level increases, possibly indicating that higher skilled big men are more athletic or have better decision-making abilities which allow them to discern the opportune times to go for steals versus remaining in defensive position (between the offensive player and the basket).

IV. Conclusion

We utilize quantile regressions to examine salaries of NBA players — allowing for a different distribution of earnings for big men and guards. Our findings provide further insights into the salary structure of NBA players — as players with a range of skills and playing different positions receive significantly different salary rewards and penalties for performance attributes.


  1. Dupree, David. “2001–02 NBA Salaries” USA Today, (November, 15, 2001).
  2. Hamilton, Barton H. “Racial Discrimination and Professional Basketball Salaries in the 1990s.” Applied Economics 29 (March 1997): 287–96.
  3. Kahn, Lawrence M. “The Sports Business as a Labor Market Laboratory.” Journal of Economic Perspectives 14 (Summer 2000): 75–94.
  4. Koenker, Roger and Gilbert Bassett. “Regression Quantiles.” Econometrica 46 (January 1978): 33-50.
  5. Nance, Roscoe. “10-Year Vets Bide Time on Market.” USA Today, (July 31, 2003).
  6. Staudohar, Paul. “Labor Relations in Basketball: The Lockout of 1998–99.” Monthly Labor Review 122 (April 1999): 3–9.

 Table 1

Salaries of NBA Players in 2001–2002 Season

Variables Coefficient t-statistic
Constant -1.282*** (-6.74)
Exp 0.313 *** (10.93)
Exp 2 -0.014** (-7.83)
Assists 0.071** (2.43)
Blocks 0.151 *** (3.63)
Steals 0.052 (0.94)
Rebounds 0.062 *** (2.68)
Pts/Game 0.043 *** (4.09 )
Free throw 0.244 (1.11)
Black -0.010 (-0.12)
Big-man Sample size 0.188*409 (1.88)
R ^ 2Adj- R^ 2 .602.593

Note: * (**, ***) denotes significance at the .10 (.05, .01) level.


Table 2

Salaries of Guards in the NBA in the 2001–2002 Season

Variable 15 th 25 th 50 th 75 th 90 th
Constant -1.698***(-5.25) -1.499***(-3.48) -1.334***(-3.80) -1.502***(-3.18) -1.032***(-2.99)
Exp .29409***(4.76) .4221***(6.51) .4157***(6.90) .4306***(4.87) .2933***(5.07)
Exp 2 -.0151***(-4.24) -.0220***(-5.98) -.0190***(-5.16) -0.0191***(-3.50) -.0117***(-3.75)
Assists .0442(0.99) .0196(0.36) .0409(0.92) .0539(0.81) .0433(0.90)
Blocks .1016(1.22) .1201(1.41) .1553**(2.20) .1894**(2.01) -.0310(-0.46)
Steals .4859**(2.31) .3570(1.54) -.0096(-0.04) .0132(0.04) .0972(0.44)
Rebounds -.1458*(-1.69) -.0530(-0.57) -.0267(-0.31) .0229(0.18) .0613(0.79)
Points/game .05452***(2.71) .0549***(2.60) .0745***(3.51) .0407**(2.17) .0468**(2.33)
Free throw .2081(0.71) -.0577(-0.13) .1919(0.47) .9472*(1.78) .3756(0.92)
Black -.1205(-0.61) -.1637(-0.63) -.1927(-0.93) -.0164(-0.06) .1667(0.78)

Note: t-statistics are in parenthesis, * (**, ***) denotes significance at the .10 (.05, .01) level.

Table 3

Salaries of Big Men in the NBA in the 2001–2002 Season

Variable 15 th 25 th 50 th 75 th 90 th
Constant -1.682***(-5.80) -1.756***(-6.60) -1.149***(-3.76) -.1874(-0.75) .4952*(1.79)
Exp .2360***(4.54) .2947***(6.01) .3255***(6.59) .3027***(6.87) .2440***(5.66)
Exp 2 -.0090***(-2.77) -.0133***(-4.36) -.01479(-4.61) -.0129***(-4.48) -.0089***(-3.45)
Assists .1534*(1.78) .1143(1.38) .0011(0.01) .0736(1.01) .0417(0.42)
Blocks .1221(1.30) .1161(1.38) .2289***(2.75) .1453**(2.03) .2267***(3.44)
Steals -.1094*(-1.83) -.0637(-1.01) .0456(1.28) .0447(1.48) .0137(0.42)
Rebounds .1186***(2.99) .0979**(2.51) .0720*(1.94) .0242(0.87) .0024(0.09)
Points/game .0453**(2.09) .0407*(1.83) .0490**(2.29) .0294*(1.86) .0281*(1.75)
Free throw .0093(0.02) .3001(0.79) .3276(0.85) -.0313(-0.10) .2718(0.75)
Black -.14579(-0.92) .0602(0.43) -.0425(-0.33) .1155(1.02) -.0967(-0.82)

Note: t-statistics are in parenthesis, * (**, ***) denotes significance at the .10 (.05, .01) level.



See Koenker and Bassett (1978) for a review of this procedure.

During the 2001–2002 NBA season there were 29 teams. Since then, one team has changed locations, the Charlotte Hornets moved to New Orleans and, another, the Charlotte Bobcats, was added in the 2004–2005 season.

Data for this analysis were obtained from, last viewed April 30, 2004.

The marginal impact of a characteristic on the salaries of the group in question is found by taking the exponential of the estimated coefficient minus one and multiplying by 100.

2015-03-24T08:26:18+00:00March 4th, 2005|Contemporary Sports Issues, Sports Management, Sports Studies and Sports Psychology|Comments Off on NBA Salaries: Role Players and Superstars