Inside the nfl: how do minority head coaches measure up ...



2012Hanover CollegeClarence Cledanor[Inside the nfl: how do minority head coaches measure up against the majority?] In 1921, the NFL hired its first minority head coach, Fritz Pollard. Another minority would not be hired as a head coach until the 1979 season when the Oakland Raiders hired Tom Flores. This project will attempt to examine the labor discrimination that existed, and still may exist, in the NFL. Using data prior to the beginning of the 2012 season, this assessment will primarily look to analyze the salaries of NFL head coaches and see if they differ amongst whites and minorities as evidence reveals that, on average, minorities are paid significantly less.Table of ContentsIntroductionPage 3Literature ReviewPage 4ModelPage 8DataPage 11Empirical ResultsPage 11ConclusionPage 15AppendixPage 17BibliographyPage 26Introduction???????????This econometric model will look to delve into the prosperous realm of professional sports and attempt to provide substantial knowledge regarding the state of the minority head coach in professional sports with the main focus being that of the National Football League. Wage and labor discrimination will be the primary issues investigated by this model. More precisely, this model will look to answer the following three questions; 1.) On average, is there a significant gap in salary between those coaches who are of the minority and those who are not? 2.) On average, who are more successful (based on winning percentage), minority head coaches or the majority? 3.) Has the "Rooney Rule",?established in 2003, been successful for minority head coaching candidates? Furthermore, in addition to answering the previous questions, by use of empirical research and statistical evidence, this model, in the overall scheme of the project, will also look to explain the behaviors of employers. For example, this assessment will look to provide information on why employers may, or may not, delegate minorities to managing, or leadership, roles. The essential purpose of this equation will be to provide inferences on NFL head coaches' salaries as well as see what variables, if any, have the more significant impact on them.Cross-sectional data prior to the beginning of the 2012 season is used for each head coach and, therefore, all thirty-two NFL teams. The dependent variable in this model is the salary of each head coach.? The independent variables include head coaching experience, winning percentage, ethnicity, playing experience, and the number of championships won. Intuition, as well as previous literature, indicates that these factors significantly affect the salaries of NFL head coaches.The data gathered for this project comes by way of several sources such as ESPN, official team websites, and the websites of teams' local newspaper. The data concerning the salary of each head coach was, for the most part, obtained from teams' official websites. The data for all of the independent variables was obtained from different credible league sources such as the ones already alluded to.Literature ReviewThe history of minority coaches in professional sports does not possess a long lineage, as, quite frankly, they have not been given as many opportunities in contrast to their counterparts. One would think that for sports where minorities, mainly Hispanics and African-Americans, are practically the playing majority, there would be more minorities in leadership and managing roles. However, such is not the case. Over the eighty-plus year history of the NFL, there have been only twenty-two minority head coaches, with six of the twenty-two only being able to hold the “interim” head coach title. In 1921, the NFL had its first African-American head coach, Fritz Pollard. Another minority would not be hired as a head coach until the 1979 season when the Oakland Raiders hired Tom Flores. Could there be any particular reasons for this? How does the evaluation process for coaching proceed? Does the labor force in professional sports reflect that of the real world labor force? Is there anything being done to counteract this practice? Does it even matter? What follows is a review of literature aimed at analyzing why minority coaches have remained "locked out" for such a long period of time.A review of the literature shows that researchers have observed a common process for the hiring of coaches. For the most part, coaches are hired based upon their experience as well as past successes. However, the basis upon whether or not a coach is hired depends on the preferences of the owner along with social changes in the economy. Goff and Tollison (2008) examined data from the NFL from 1987 through 2007 and revealed that integration was more commonplace in larger populated cities. They also reveal that the hiring process of management is totally different from the hiring process for players (p. 127). Furthermore, examinations of the regular season win records and of making the playoffs for NFL teams coached by both African-Americans and Caucasians between 1990 and 2002 show that African-American coaches were more successful (Madden, 2010, p.16). Madden gives evidence that supports that teams who hire African-American coaches are more successful than those who do not. However, African-American coaches have not been as successful in the playoffs as only three have won the Super Bowl (p.16). Perhaps this may be a particular reason as to why minorities, for the most part, have not been able to get their foot in the door. Another reason could be that most owners are Caucasians over the age of 60. This insinuates that most of the people whom are in charge come from a time where things were socially different due to social schisms, segregation, civil rights movements, etc. Therefore, they may be unconsciously engaging in labor discrimination against hiring minorities, especially African-Americans.One of the more famous minority NFL coaches, Tony Dungy, went through a series of secondary and assistant coaching jobs before he became a head coach. Dungy was often delegated as a head coaching candidate by his peers when Dungy was an assistant under Chuck Noll in the 1980s with the Steelers, but he would not become a head coach until 1996 when he began his own legacy coaching the Tampa Bay Buccaneers (Bouchette). For some reason, in most cases, minorities must go through numerous assistant coaching positions before even being considered as head coaching candidates. In essence, it is as if minority coaches are disadvantageously evaluated compared to their counterparts. Despite being equally qualified and seemingly more successful, minority coaches, for the most part, just do not get the opportunity. Nevertheless, as Madden (2009) states, “…the results are consistent with African American coaches being held to higher standards to get their jobs in the NFL (p. 1).”One of the more notable movements to bringing awareness to labor discrimination in professional sports is the NFL’s “Rooney Rule.” The Rooney Rule is somewhat the NFL’s version of affirmative action. The Rooney Rule requires NFL teams to interview a specified number of minority candidates for head coaching and senior football operation opportunities whenever there is a vacancy. Prior to its establishment in 2003, there were seven minority head coaches in the history of the NFL, with only two actively coaching during the rules implementation. Madden and Ruther (2010) raise the idea of whether the performance advantage of minority head coaches had been eliminated in the time since the establishment of the Rooney Rule is analyzed. First, just four years after the Rooney Rule implementation there were seven active minority head coaches in the NFL. Evidently, the Rooney Rule is effective (p. 2). However, in accordance with Madden and Ruther, the advantage that minority head coaches had, in terms of being more successful, was now eliminated, as they state, “Since the Rooney Rule was put into place, there are no racial differences in performance among head coaches in the NFL.” There are currently six minority head coaches in the NFL.The sports industry and its related components (managing, coaching, media, marketing, etc.) are a reflection of society. Muster (2001) analyzes the individual industries of Major League Baseball, the National Basketball Association, and the National Football League and compares them to the overall labor force to determine how professional sports fare in creating jobs for minority groups. With the overall labor force becoming more diverse by the moment, the significance of this research was to observe and analyze to what degree these two factions correlate. Muster recognizes that professional athletes have traditionally been male and, mainly, African-American or Caucasian. However, with the emergence of new professional sports organizations for women, opportunities for female athletes have increased. As Muster questions, “But who is working off-field for these organizations? How have sports teams and leagues staffed their organizations? Are they in line with the national labor hiring practices? Or, are they in stark contrast with the real world?” An unfriendly hiring process for minorities could have negative impacts on marketing, participation, popularity, and support for each league. The reason why professional sports industries and society reflect each other, as Muster concludes, is that professional sports are a business first. Therefore, as a business, team officials and league officials must concern themselves with social issues. Minorities in the National Basketball Association, National Football League, and Major League Baseball make up a big percentage of the players. However, of course, the question of whether labor discrimination is practiced has not to do with the players, but rather those in the “front-office.” Muster states, “As far as head coaching positions are concerned, one must remember that these are very exclusive and competitive jobs." Only thirty-two positions are available in the NFL. The six African-Americans who currently hold head coaching positions (18%) clearly mirror the overall U.S. population of this minority group. Muster also expresses, "Are there more qualified African-American candidates to assume these roles? Most certainly. Should the number of African-American head coaches be raised simply to reflect the over-representation of African-American athletes participating in the games? Absolutely not!”It could be the possibility that owners are not hiring minorities purposely, but rather unconsciously. Madden and Ruther (2009) chimes in by stating, "We believe the facts to be consistent with implicit racial discrimination in hiring. [6] NFL teams are not consciously rejecting African American coaches, but are unconsciously or implicitly discounting them in the presence of uncertainty and ambiguity as to who will be a successful coach for the initial hire as an NFL head coach." Muster refutes such notion as he expresses, “If owners are putting the best available talent on the field, and are color-blind enough to bolster their rosters with African-Americans, then it is just as conceivable that they are staffing their front offices with the best talent that they know.” Unfortunately, such notion cannot be accurately determined. As a result, the subject can be endlessly debated.ModelAs previously mentioned, the essential purpose of this equation will be to provide inferences on NFL head coaches' salaries as well as see what variables, if any, have the more significant impact on them. The equation for the model is as follows:*SALi = β0 + β1Expi + β2Winsi + β3Racei + β4Playi + β5SBi + εiWhere:SAL = Salary of head coachExp = Coaching experience; total number of years as head coach in NFLWins = Winning percentage of coach throughout careerRace = Dummy variable indicating whether or not a coach is a minority; 1 = minority, 0 = majorityPlay = Dummy variable indicating whether or not a coach has played in the NFL; 1 = yes, 0 = noSB = Total number of Super Bowl wins in coaching careerε = measurement error due to omitted variables, seasoni = 2012-2013 NFL SeasonThe main hypothesis being tested by this model is that minority coaches are paid differently than white coaches. The null and alternative hypotheses are as follows:H0: β1 ≤ 0; β2 ≤ 0; β3 ≥ 0; β4 ≤ 0; β5 ≤ 0Ha: β1 > 0; β2 > 0; β3 < 0; β4 > 0; β5 > 0The dependent variable (SAL) in the equation is the salary for an NFL head coach during the 2012-2013. This variable will be measured in millions (of dollars). The independent variables represented in the equation are an individual's years of coaching experience, number of wins in their career, ethnicity, playing experience, and number of Super Bowl wins.The first independent variable represented in the equation is coaching experience. More precisely, this variable represents the total years an individual has been coaching in the NFL. This variable is expressed as Exp in the formulated equation. It is expected that the higher an individual's total years of experience or as an individual's coaching experience increases, then the higher his salary. It is apparent that those who have been coaching longer are those with the highest salaries. This notion is perfectly logical as it can be assumed that those coaches who have plenty of experience coaching in the NFL have been able to do so, because they have been successful and are therefore rewarded with higher salaries. Since this notion is presumably verified, then the sign of the coefficient for Exp should be positive.The second independent variable represented in the equation is an individual's winning percentage throughout his head coaching career in the NFL. This variable is expressed as Wins in the formulated equation. It is expected that the higher an individual's winning percentage then the higher his salary. Literature, as well as media, suggests this notion as it can be explained intuitively that those who are more prone to winning are, or should be, rewarded with more lucrative contracts. Assuming this interpretation is correct, the sign of the coefficient for Wins should be positive.The third independent variable represented in the equation is ethnicity. This variable is a dummy variable that indicates if an individual is not Caucasian. This variable is expressed as Race in the formulated equation. It is expected that if a coach is a minority (Race = 1) then there will be an inexplicable decrease in the coach's salary compared to if the coach were Caucasian (Race = 0). It can be assumed that due to the low number of minority head coaches in the NFL, minorities, on average, have lower salaries. As a result, the sign of the coefficient for Race should be negative. The fourth independent variable represented in the equation relates to familiarity. This variable is a dummy variable that indicates if a coach had ever played in the NFL. This variable is expressed as Play in the formulated equation. It is expected that if a coach has played in the NFL (Play = 1) then there will be an increase in the coach's salary compared to if the coach did not (Play = 0). Although research has yet to be acquired on this subject, it can be assumed that those coaches who have played in the NFL are more familiar with players which can lead to great camaraderie amongst players and coaches. It can also be assumed that some coaches get their jobs, and salaries, as a result of how successful they were during their playing careers rather than what they have done as a coach. As a result, the sign of the coefficient for Play should be positive.The final independent variable represented in the equation, Super Bowl wins, is probably the most important, in regards to salary.. More precisely, this variable represents the total number of NFL championships a coach has won throughout his career. This variable is expressed as SB in the formulated equation. It is expected that the more championships a coach has won (as a coach's Super Bowl wins increases) then the higher their salary (there will be an increase in the salary of that head coach). This notion is defended by the data and makes sense as it can be expressed that those coaches who complete the ultimate goal deserve to be rewarded. Therefore, the sign of the coefficient for SB should be positive.Data The data presented in the model is cross-sectional from each of the thirty-two head coaches currently in the NFL. The descriptive statistics can be seen in Table 0. Only in one of the variables, WINS, was percentages used to eliminate large variances between head coaches because of differences in the years of experience. Doing this helped limit, and possibly eliminate, the issue of heteroskedasticity. All data was collected from numerous National Football League sources with the most useful being that of ESPN from which the majority of the data for all variables were attained. Fortunately, in regards to data collection, problems such as mismeasurement or missing observations were not encountered during the process.Empirical ResultsThe ordinary least squares method was used several times in the analysis of this model. The initial regression equation of the model was:SALi = 2347925 + 140821.3Expi + 1437283Winsi – 134045.5Racei + 465766.2Playi + 779294SBiIn accordance to the signs of the coefficients, it is apparent that the prior expected judgments were correct. More precisely, as a head coach’s years of coaching experience increases by an additional year then there will be a $140,821 increase in that coach’s salary. Also, a one percentage point increase in a head coach’s winning percentage will result in a $1,437,283 boost in the coach’s salary. In addition, as expected, being a minority head coach has its disadvantages as the model expresses that if a head coach is a minority then there will be a $134,045 decrease in the coach’s salary. In the same sense, it is apparent that if a head coach has had any playing experience in the NFL then they will see a $465,766 boost in their salary. Furthermore, the model expresses that for every Super Bowl won there will be a $779,294 increase in the coach’s salary. Of course, all of these are assumed ceteris paribus. As can be seen in Table 1, the R-squared in this regression equation was 0.57 and the adjusted R-squared was 0.4884. The goodness of fit of the variance in the model can be interpreted as being decent enough since there is not a significant difference between the values of R-squared and the adjusted R-squared. In other words, the model is a good fit and the various variables explain approximately forty-nine percent of the variance in head coach salaries.Other notable inferences that can be made about the model and this particular regression is that the probability for the F-statistic is practically zero which means a conclusion can be made that the model is significant due to the rejecting of the null hypothesis (Ho: β1 = β2 = β3 = β4= β5 0) at the ninety-nine percent confidence level. After conducting a Ramsey Rest Test (Table 3), it is revealed that the regression lacks no omitted variables or specification errors due to the inability to reject the null (at the five percent level of significance) with the calculated F-statistic, 2.195, being lesser than the critical F-statistic, 3.03. Multicollinearity is not much of an issue as well in this model. As can be seen in Table 2, there are no major correlations between any two variables other than that of EXP and SB which logically makes sense as it can be assumed that the longer one has been coaching then the more likely it is that the head coach has won, or has a better chance, of winning a championship. Serial correlation is not of much issue in this model either due primarily to the fact that this model deals with cross-sectional data rather than time-series. However, the Durbin-Watson test statistic of 1.735, being that is close to 2, ensures us that the model lacks any autocorrelation. Heteroskedasticity, due to this model being a cross-sectional study, should be the primary econometric issue present. After running a White test (Table 4) to see if the model has heteroskedasticity, it is apparent that at the ninety-five percent confidence level the null hypothesis cannot be rejected due to the calculated NR2, 10.66, being lesser than the critical NR2, 28.9. As a result, it can be concluded that this model does not have the econometric issue of heteroskedasticity.A series of hypothesis tests were conducted to see which of the independent variables were more significant in relations to salary. The null and alternative hypotheses are as follows:H0: β1 ≤ 0; β2 ≤ 0; β3 ≥ 0; β4 = 0; β5 ≤ 0Ha: β1 > 0; β2 > 0; β3 < 0; β4 ≠ 0; β5 > 0For all tests, the ninety-five percent confidence level was used to determine if the null would be rejected. With p-values of 0.0114 and 0.0256, EXP and SB, respectively, are able to reject the null and it can be concluded that these two variables are significant. The others, however, do not even come close to being significant at a ninety percent level of confidence (Table 1). Therefore, in an attempt to provide a better model, a second regression equation was formulated with the omission of the PLAY variable due to the insignificance of the coefficient. The equation is as follows:SAL = 2406181 + 130736Exp + 1784881Wins – 140134.1Race + 729680.5SBThe omission did not change much. First, the expected signs of the coefficients remained the same. Secondly, there is an apparent increase in the coefficients of all variables other than SB which has an apparent decrease in the magnitude of its coefficient. This reveals that omitting PLAY could be causing some bias in the analysis of the independent variables. Furthermore, this new regression reveals a minor increase in R-squared and adjusted R-Squared (Table 4). Essentially, the goodness of fit of this second regression equation is slightly better than the first regression. Lastly, at the ninety percent level of confidence, it can be concluded that Wins is now statistically significant with a p-value of approximately 0.07 (Table 5).In an attempt to build on the previous regression, a final regression equation was formed with the addition of an interaction term, WINS*RACE. This slope dummy variable accounts for the slopes of the salary variable and the winning percentage variable to be different depending on whether or not a head coach is a minority. It is of high importance, because it allows for an in-model analysis to take place whenever the RACE variable impacts SAL depending on whether or not the coach is a minority or not. The final regression equation is as follows:SAL = 2573911 + 133132.1Exp + 1425597Wins – 2427478Race + 685979.4SB + 4951807Wins*RaceThere is no significant change in the estimated output in comparison to the first two equations (Table 8). However, this last regression equation does provide a revealing fact. A salary equation with both an intercept dummy variable and a slope dummy variable can take on different shapes when graphed with wins on the horizontal axis. By use of algebra, the system of two equations is able to be solved and the following conclusion can be made: if two coaches, one white and the other non-white, begin their head coaching careers at the same time (along with everything else equal) then it is apparent that the white head coach will be paid more due to the intercept dummy. However, there is a point at which the minority head coach can exceed the white coach in salary (due to the magnitude of the slope dummy coefficient) and that is when he surpasses a particular winning percentage.If a head coach is white, the change in salary over the change in winning percentage is equal to $1,425,597. However, if the head coach is a minority, this same change is equal to $6,377,404. This represents the additional returns of winning for a minority head coach. The intercept dummy, however, tells us that initially, when all variables are constant, a minority head coach sees a negative $2,427,478 change in salary (the change in salary over the change in the race variable is 4,951,807 - 2,427,478 = $2,524,329). The point at which the minority head coach would exceed the white head coach in salary is when he achieves a winning percentage of approximately forty-nine percent. A more detailed and comprehensive assessment might look to include the number of years it may take a minority coach to exceed the Caucasian coach as well as the details of contractual agreements (amongst other things) to accurately explain this notion.ConclusionIt is more than likely that this model does not include all the possible factors that could affect the salaries of NFL head coaches. Some other factors that could have been helpful to this model are; 1.) The inclusion of all head coaches since 1990 as a way of being able to compare and contrast statistics amongst white and minority head coaches along a wider period and 2.) To go along with the previous idea, the inclusion of a dummy variable that measured the impact the Rooney Rule has had on the NFL since 2003. The variable simply would have measured minority head coaches amongst each other by distinguishing if a minority coach was hired prior to 2003 or after. With these two inclusions, the success of minority head coaches in comparison to that of whites could have been analyzed. The difference in salary as well as the number of minority hires since, and prior to, 2003 could have been analyzed, too. However, due to the inability to acquire such information, this model was unable to include such factors. The reason for wanting to include such factors comes by way of the assertion, "The evidence on fire and rehire, when combined with the other evidence of greater successes for African American coaches presented in Madden (2004), is consistent with implicit racial discrimination in the NFL during the 1990-2002 period (Madden & Ruther 2009)." In the end, this assessment serves as a decent model that, for the most part, explains the overall factors that determine the salaries of head coaches in the NFL. The R-squared and adjusted R-squared in all presented regressions are rather favorable. The F-statistics in all presented regressions also reveal that this model is significant. Fortunately, econometric issues such as functional form, omitted variables, multicollinearity, serial correlation, and heteroskedasticity are not a problem in this model. However, the insignificance of the coefficient for RACE in all presented regressions is puzzling. Reasons for such results could include the exclusion of some other variable that relates to ethnicity and/or the sample size, or it could very well be the case that ethnicity does not play a role in the salaries given to NFL head coaches. If all information were able to be acquired, it would be interesting to see the affect the aforementioned inclusions would have had on this model.Appendix.Table 1.Regression results Dependent Variable: SALMethod: Least SquaresDate: 11/14/12 Time: 23:45Sample: 1 32Included observations: 32VariableCoefficientStd. Errort-StatisticProb.??C2347925.556282.84.2207400.0003EXP01140821.358206.222.4193520.0228WINS1437283.1250355.1.1495010.2608RACE-134045.5571540.0-0.2345340.8164PLAY465766.2514493.30.9052910.3736SB779294.0381456.12.0429450.0513R-squared0.570960????Mean dependent var4174063.Adjusted R-squared0.488452????S.D. dependent var1754148.S.E. of regression1254612.????Akaike info criterion31.08991Sum squared resid4.09E+13????Schwarz criterion31.36474Log likelihood-491.4386????Hannan-Quinn criter.31.18101F-statistic6.920076????Durbin-Watson stat1.734561Prob(F-statistic)0.000315Table 2.Correlation TableSALEXP01WINSRACEPLAYSBSAL?1.000000?0.670735?0.496089-0.092485?0.015986?0.605656EXP01?0.670735?1.000000?0.463853-0.077102-0.178340?0.582798WINS?0.496089?0.463853?1.000000-0.005791?0.175591?0.318260RACE-0.092485-0.077102-0.005791?1.000000?0.021592-0.096471PLAY?0.015986-0.178340?0.175591?0.021592?1.000000-0.197288SB?0.605656?0.582798?0.318260-0.096471-0.197288?1.000000Table 3.Ramsey RESET Test:F-statistic2.195031????Prob. F(3,23)0.1160Log likelihood ratio8.056844????Prob. Chi-Square(3)0.0449Test Equation:Dependent Variable: SALMethod: Least SquaresDate: 12/07/12 Time: 03:20Sample: 1 32Included observations: 32VariableCoefficientStd. Errort-StatisticProb.??C-3675686927580073-1.3327330.1957EXP01-4795873.3028452.-1.5836060.1269WINS-4395009230380229-1.4466680.1615RACE5200551.3085214.1.6856370.1054PLAY-156809269960919.-1.5742450.1291SB-2711893016920601-1.6027170.1226FITTED^29.54E-067.10E-061.3433190.1923FITTED^3-1.10E-121.01E-12-1.0966580.2841FITTED^44.65E-205.16E-200.9018300.3765R-squared0.666456????Mean dependent var4174063.Adjusted R-squared0.550441????S.D. dependent var1754148.S.E. of regression1176142.????Akaike info criterion31.02564Sum squared resid3.18E+13????Schwarz criterion31.43787Log likelihood-487.4102????Hannan-Quinn criter.31.16228F-statistic5.744557????Durbin-Watson stat1.909705Prob(F-statistic)0.000448Table 4.Heteroskedasticity Test: WhiteF-statistic0.360677????Prob. F(18,13)0.9767Obs*R-squared10.65811????Prob. Chi-Square(18)0.9083Scaled explained SS6.736504????Prob. Chi-Square(18)0.9922Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 11/19/12 Time: 14:23Sample: 1 32Included observations: 32Collinear test regressors dropped from specificationVariableCoefficientStd. Errort-StatisticProb.??C4.86E+111.12E+120.4331380.6720EXP011.75E+121.40E+121.2460650.2347EXP01^2-3.68E+104.61E+10-0.7978940.4393EXP01*WINS-1.96E+123.00E+12-0.6534840.5248EXP01*RACE-3.48E+118.52E+11-0.4082430.6897EXP01*PLAY-2.06E+102.95E+11-0.0698450.9454EXP01*SB2.12E+114.25E+110.4981920.6267WINS-5.80E+121.41E+13-0.4102480.6883WINS^26.50E+122.39E+130.2713820.7904WINS*RACE2.06E+132.47E+130.8368490.4178WINS*PLAY3.47E+121.84E+130.1882480.8536WINS*SB2.49E+132.46E+131.0123440.3298RACE-8.44E+129.11E+12-0.9265070.3711RACE*PLAY1.87E+125.59E+120.3344370.7434RACE*SB-5.42E+127.99E+12-0.6783060.5095PLAY-1.25E+129.91E+12-0.1257140.9019PLAY*SB-3.74E+114.43E+12-0.0844880.9340SB-1.50E+131.54E+13-0.9714650.3490SB^2-1.23E+121.74E+12-0.7060630.4926R-squared0.333066????Mean dependent var1.28E+12Adjusted R-squared-0.590381????S.D. dependent var1.80E+12S.E. of regression2.27E+12????Akaike info criterion60.02403Sum squared resid6.68E+25????Schwarz criterion60.89431Log likelihood-941.3845????Hannan-Quinn criter.60.31250F-statistic0.360677????Durbin-Watson stat1.697726Prob(F-statistic)0.976659Table 5.Dependent Variable: SALMethod: Least SquaresDate: 11/19/12 Time: 14:27Sample: 1 32Included observations: 32VariableCoefficientStd. Errort-StatisticProb.??C2406181.550698.84.3693220.0002EXP01130736.056938.932.2960740.0297WINS1784881.1185958.1.5050120.1439RACE-140134.1569587.5-0.2460270.8075SB729680.5376235.21.9394270.0630R-squared0.557436????Mean dependent var4174063.Adjusted R-squared0.491871????S.D. dependent var1754148.S.E. of regression1250413.????Akaike info criterion31.05845Sum squared resid4.22E+13????Schwarz criterion31.28747Log likelihood-491.9351????Hannan-Quinn criter.31.13436F-statistic8.502028????Durbin-Watson stat1.637109Prob(F-statistic)0.000142Table 6.Ramsey RESET Test:F-statistic1.407064????Prob. F(3,24)0.2650Log likelihood ratio5.184619????Prob. Chi-Square(3)0.1588Test Equation:Dependent Variable: SALMethod: Least SquaresDate: 12/07/12 Time: 05:20Sample: 1 32Included observations: 32White Heteroskedasticity-Consistent Standard Errors & CovarianceVariableCoefficientStd. Errort-StatisticProb.??C-3989543016233434-2.4576090.0216EXP01-4484171.1601206.-2.8004950.0099WINS-5736786421605131-2.6552890.0139RACE5101531.1962204.2.5998990.0157SB-254404178816476.-2.8855540.0081FITTED^21.00E-053.99E-062.5058330.0194FITTED^3-1.21E-125.75E-13-2.1061030.0458FITTED^45.33E-202.97E-201.7943970.0854R-squared0.623633????Mean dependent var4174063.Adjusted R-squared0.513859????S.D. dependent var1754148.S.E. of regression1223059.????Akaike info criterion31.08393Sum squared resid3.59E+13????Schwarz criterion31.45036Log likelihood-489.3428????Hannan-Quinn criter.31.20539F-statistic5.681067????Durbin-Watson stat1.736279Prob(F-statistic)0.000586Table 7.Heteroskedasticity Test: WhiteF-statistic1.058159????Prob. F(13,18)0.4459Obs*R-squared13.86173????Prob. Chi-Square(13)0.3837Scaled explained SS8.351858????Prob. Chi-Square(13)0.8199Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 11/19/12 Time: 14:29Sample: 1 32Included observations: 32Collinear test regressors dropped from specificationVariableCoefficientStd. Errort-StatisticProb.??C4.31E+118.51E+110.5063490.6188EXP012.03E+128.53E+112.3801130.0286EXP01^2-2.15E+102.18E+10-0.9862580.3371EXP01*WINS-2.90E+121.45E+12-2.0006170.0607EXP01*RACE-5.14E+115.19E+11-0.9917700.3345EXP01*SB9.16E+102.60E+110.3515460.7293WINS-7.20E+127.71E+12-0.9341630.3626WINS^21.17E+131.04E+131.1203390.2773WINS*RACE1.94E+131.60E+131.2128380.2409WINS*SB2.96E+131.81E+131.6361210.1192RACE-6.61E+124.95E+12-1.3347680.1986RACE*SB-7.50E+124.98E+12-1.5058290.1495SB-1.60E+131.16E+13-1.3825950.1837SB^2-1.27E+121.12E+12-1.1292840.2736R-squared0.433179????Mean dependent var1.32E+12Adjusted R-squared0.023809????S.D. dependent var1.74E+12S.E. of regression1.72E+12????Akaike info criterion59.48760Sum squared resid5.34E+25????Schwarz criterion60.12886Log likelihood-937.8016????Hannan-Quinn criter.59.70016F-statistic1.058159????Durbin-Watson stat1.591157Prob(F-statistic)0.445860Table 8.Dependent Variable: SALMethod: Least SquaresDate: 11/19/12 Time: 14:52Sample: 1 32Included observations: 32VariableCoefficientStd. Errort-StatisticProb.??C2573911.560498.84.5921800.0001EXP01133132.156343.652.3628590.0259WINS1425597.1206703.1.1813980.2481RACE-2427478.1891511.-1.2833540.2107SB685979.4373687.71.8357020.0779WINS*RACE4951807.3909071.1.2667480.2165R-squared0.583162????Mean dependent var4174063.Adjusted R-squared0.503001????S.D. dependent var1754148.S.E. of regression1236642.????Akaike info criterion31.06106Sum squared resid3.98E+13????Schwarz criterion31.33588Log likelihood-490.9769????Hannan-Quinn criter.31.15216F-statistic7.274871????Durbin-Watson stat1.426366Prob(F-statistic)0.000223Table 9.Ramsey RESET Test:F-statistic0.754934????Prob. F(3,23)0.5308Log likelihood ratio3.005377????Prob. Chi-Square(3)0.3908Test Equation:Dependent Variable: SALMethod: Least SquaresDate: 12/07/12 Time: 05:29Sample: 1 32Included observations: 32VariableCoefficientStd. Errort-StatisticProb.??C-2613399537392659-0.6989070.4916EXP01-2697899.3274242.-0.8239770.4184WINS-2706940334999464-0.7734230.4471RACE47230359587252440.8042600.4295SB-1427971616872864-0.8463130.4061WINS*RACE-971848381.20E+08-0.8098700.4263FITTED^26.03E-068.34E-060.7231130.4769FITTED^3-7.33E-131.21E-12-0.6062740.5503FITTED^43.27E-206.31E-200.5180040.6094R-squared0.620528????Mean dependent var4174063.Adjusted R-squared0.488538????S.D. dependent var1754148.S.E. of regression1254506.????Akaike info criterion31.15464Sum squared resid3.62E+13????Schwarz criterion31.56688Log likelihood-489.4742????Hannan-Quinn criter.31.29129F-statistic4.701326????Durbin-Watson stat1.460589Prob(F-statistic)0.001637Table 10.Heteroskedasticity Test: WhiteF-statistic0.939150????Prob. F(15,16)0.5462Obs*R-squared14.98282????Prob. Chi-Square(15)0.4527Scaled explained SS9.649036????Prob. Chi-Square(15)0.8412Test Equation:Dependent Variable: RESID^2Method: Least SquaresDate: 12/07/12 Time: 05:28Sample: 1 32Included observations: 32Collinear test regressors dropped from specificationVariableCoefficientStd. Errort-StatisticProb.??C2.47E+118.84E+110.2794320.7835EXP012.22E+129.05E+112.4530470.0260EXP01^2-1.92E+102.29E+10-0.8363900.4153EXP01*WINS-3.33E+121.54E+12-2.1600990.0463EXP01*RACE-2.82E+126.33E+12-0.4454940.6619EXP01*SB8.35E+102.72E+110.3074340.7625EXP01*(WINS*RACE)5.54E+121.47E+130.3758890.7119WINS-7.19E+128.07E+12-0.8902270.3865WINS^21.27E+131.09E+131.1615110.2625WINS*RACE1.46E+131.44E+140.1018990.9201WINS*SB3.08E+131.88E+131.6322050.1222WINS*(WINS*RACE)-3.28E+132.51E+14-0.1305460.8978RACE-4.03E+111.90E+13-0.0212070.9833RACE*SB-3.38E+122.06E+13-0.1644070.8715SB-1.65E+131.21E+13-1.3697890.1897SB^2-1.26E+121.17E+12-1.0771030.2974R-squared0.468213????Mean dependent var1.24E+12Adjusted R-squared-0.030337????S.D. dependent var1.76E+12S.E. of regression1.79E+12????Akaike info criterion59.57111Sum squared resid5.13E+25????Schwarz criterion60.30398Log likelihood-937.1377????Hannan-Quinn criter.59.81403F-statistic0.939150????Durbin-Watson stat1.536135Prob(F-statistic)0.546165Works CitedAllen, W., & Chadwick, C. (2012). Performance, Expectations, and Managerial Dismissal: ?Evidence from the National Football League. Journal Of Sports Economics, 13(4), ?337-363. Altonji, G. J., & Banks, M. R. (1999). Race and Gender in the Labor Market. Journal of ?Economic Literature, 3144-3251. Retrieved from , E. (2007). Tony Dungy A Man of Grace. Pittsburgh Post-Gazette. Retrieved from ?, M., & Emerson, P. (2006 April). Discrimination in Hiring versus Retention and ?Promotion: An Empirical Analysis of within-Firm Treatment of Players in the NFL. ?Journal of Law, Economics, & Organization, 22.1, 115-136. Retrieved from ?, B.L., & Tollison, R.D. (2008) Racial Integration of Coaching: Evidence from the NFL. ?Journal of Sports Economics,10.2,127-140. Retrieved from ?, L., Levine, D. I., & Leonard, J. (2011). Racial Bias in the Manager-Employee ?Relationship: An Analysis of Quits, Dismissals, and Promotions at a Large Retail Firm. ?Journal Of Human Resources, 46(1), 26-52. Hadley, L., Poitras, M., Ruggiero, J., & Knowles, S. (2000 March). Performance Evaluation of ?National Football League Teams. Managerial and Decision Economics, 21.2, 63-70. ?Retrieved from , L. (1991 April). Discrimination in Professional Sports: A Survey of the Literature.Industrial and Labor Relations Review, 44.3, 395-418. Retrieved from ?, J. (2004). Differences in the Success of NFL Coaches by Race, 1990-2002 ?Evidence ofLast Hire, First Fire. Journal of Sports Economics, 5.1,6-19. Retrieved ?from , J., & Ruther, M. (2010 July). Has the NFL's Rooney Rule Efforts 'Leveled the Field' for African American Head Coach Candidates?. PSC Working Paper Series, 0-20. Retrieved from , J., & Ruther, M. (2009). Reply to: Differences in the Success of NFL Coaches by Race: A Different Perspective. Journal Of Sports Economics, 10(5), 543-550.Muster, B. (2001) Minority Hiring Practices in Professional Sports. The Sport Journal, 1543-?9518. Retrieved from , B. L., Solow, J. L., & Walker, T. B. (2011). Moving on Up: The Rooney Rule and?Minority Hiring in the NFL. Labour Economics, 18(3), 332-337.?doi: Dwindling Number of Black Head Coaches in Big-Time College Football. (2003). The ?Journal of Blacks in Higher Education, 38, 50-52. Retrieved from ?? ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download