Determining the Value of NFL Quarterbacks
Total Page:16
File Type:pdf, Size:1020Kb
1 Determining The Value Of NFL Quarterbacks Economics Honors Thesis Steve Alexander 2008 2 Table of Contents Variable Reference Guide Introduction Related Research Results Conclusion Summary Statistics Appendix Works Cited 3 Variable Reference Guide (Every statistic is specific to a particular season/year) Ht Height in inches Wt Weight in pounds Age Age in years Exp Years of NFL experience (Equals 1 if player is a rookie) G Games attended (Quarterback did not necessarily play) GS Games where Quarterback was designated as a starting player Pcomp Number of pass completions Patt Number of passes attempted PcompPct Pcomp/Patt Pyds Number of passing yards PTD Number of passing touchdowns scored Interceptions Number of interceptions thrown Sck Number of times sacked SckY Number of yards lost as a result of being sacked Rate Passer Rating (this is explained elsewhere in the paper) Ratt Number of rushing attempts Ryds Number of rushing yards RTD Number of rushing touchdowns scored Fum Number of fumbles FumL Number of fumbles lost BaseSalary Annual base salary in dollars (there is a minimum salary level explained later) SignBonus Signing bonus in dollars as a lump sum in the year it was negotiated OtherBonus Roster, report, workout, and other bonuses in dollars TotalSalary BaseSalary + SignBonus + OtherBonus CapValue BaseSalary + OtherBonus + Pro‐Rated Signing Bonus 4 Introduction This study is comprised of several closely related goals. First and foremost, an attempt will be made to predict the future earnings of National Football League quarterbacks based on their performance in previous seasons, as judged by multiple key statistics. With these statistics, efforts will also be made to predict a team’s future record and to see exactly how vital these players are to their teams. The reader is assumed to have a basic understanding of the game of football. The quarterback position is unique. Historically, it has been regarded as the most critical role. The pay many of these players receive reflects that fact well. The quarterback is the first player to receive the football when it is put into play and the decisions he makes can alter the team’s odds of success or failure drastically. He must be hyper‐aware of every one of the other 21 players on the field—genetically gifted super‐athletes with lightning‐quick speed. He can throw the ball, run with the ball, or give the ball to another player. From memory, he can call and execute the complex plays the coach has chosen, or he can modify these calls after evaluating the opposing defense with audibles. Quarterbacks represent the majority of Most Valuable Player Award recipients and are remembered for years after the rest of their teammates have been forgotten. 5 One would therefore imagine that a quarterback’s pay would be highly sensitive to his on‐field performance, and perhaps the performance of his team would be affected likewise. The players whose perceived values are most directly related to measurable statistics are probably quarterbacks. For instance, this study would be difficult or impossible if it concerned offensive linemen because their performance is not easily quantifiable. Other positions can be similarly confounding, such as defensive backs. A great defensive back may not have outstanding statistics, such as a high amount of interceptions, because the ball will not often be thrown in their direction. For these reasons, I have chosen to study the quarterback position over all others. I compiled a list of every quarterback on the NFL’s payroll from the 2000 season to the end of the most recent (2007) season. I then gathered their individual statistics for each season from the official NFL records database. Next, using U.S.A Today’s database, I was able to find 6 figures for each player’s annual compensation and bonuses. Finally, using the STATA program, I attempted to create meaningful regressions. Primarily, four methods were used to regress these time‐series panel data. To control for such a high level of endogeneity and fluctuation of the variance of the error term, the fixed effects method was employed. Three dummy‐variable approaches were also utilized. The first used a dummy variable representing each year (season). The second involved 96 dummy variables, each assigned to a specific quarterback. The third used 32 dummy variables, each assigned to a specific team. Before the results of this study are shown, it is necessary for a quick primer on NFL pay schemes. Each player receives an annual salary, and there are minimum salary levels that vary with the number of years a player has been in the NFL. This minimum level is determined by the Collective Bargaining Agreement (CBA), which is negotiated by the NFL Players Association, a union of which every NFL player is a member. Years of Experience Minimum Salary 0 $285,000 1 $360,000 2 $435,000 3 $510,000 4‐6 $595,000 7‐9 $720,000 10+ $820,000 7 Players may also receive signing bonuses when they commit to a contract and other performance‐based bonuses. This makes a player’s total salary numbers fluctuate wildly from year to year, as signing bonuses can be very large and are counted as a lump‐sum payment in the year they were signed for. For this reason I have chosen to use the CapValue variable as my main salary statistic rather than the TotalSalary variable. The CapValue variable pro‐rates the amount of any signing bonuses over the years for which the player has contractually agreed to play. An NFL team may not pay more than a certain level of compensation to its players as a whole. This level is called the salary cap, and the CapValue statistic is the amount of compensation used for the purposes of calculating whether a team is at or under the cap. Because choosing to pay a quarterback more means choosing to pay other players less, the CapValue statistic gives an accurate measurement of the degree to which a team values their quarterback. Quarterbacks who have been drafted but have not yet played in their first season are allowed to negotiate their compensation only with the team that drafted them. For this reason, pay levels for rookies tends to be extremely low relative to more experienced quarterbacks who have negotiated another contract. 8 Related Research During the course of this study, some academic research that is directly relevant was found. Far more research was found that was tangentially relevant, but still interesting. When dealing with rookies, Hendricks, DeBrock, and Koenker found that the sooner draftees were allowed to negotiate with other teams, the lower the demand for athletes with more uncertain futures. These athletes might include players from lesser‐known schools or players with a history of injuries. These researchers also found that the visibility of the football program that a player came from was significant and positively correlated with their success in the draft. However, players from less visible programs seemed to have better careers over the long term, as their salaries were less likely to fall over time. These players received less pay initially, though. In my research I expected to find large returns to experience. Clark and Hall found that teams with a greater number of veteran players were of a better quality and had more success. The number of veterans was positively correlated with the level of competition between teammates. This is because the more veterans a team has, the less the likelihood of an open position. The increased competition for the spot drives the starter and the backups to perform better than they otherwise would. Clark and Hall also found that the preseason (which is not discussed in this study due to its nature) is a good thing, as it increases competition between teammates and therefore improves team quality. To 9 perform their research, these individuals used a team‐dummy method identical to the method performed in this study. To discuss the research done by Leeds and Kowaleski, it is necessary to again refer to the CBA. The current CBA went through a major overhaul in 1993. These researchers found that dramatic shifts in quarterback pay resulted. After this overhaul, the top quartile of quarterbacks (ranked according to pay) was rewarded more for starting games than for their performance in those games. In other words, performance was de‐emphasized relative to merely starting in games. For the lowest quartile of quarterbacks, however, the basis of pay stayed basically the same as before the new CBA. For these quarterbacks, there was a far greater emphasis on performance than for the higher‐ranked players. These low‐ranked quarterbacks could dramatically improve their pay by performing better. In my research I expected to find a large return to starting games, as this study might suggest. 10 Results Determining Cap Value as a Result of Passing Touchdowns Scored The primary way a quarterback scores points is by passing the football. Thus, the PTD statistic should prove especially relevant in determining the value of the quarterback. My first approach was via the fixed effects method. 1. CapValue Coefficient Std. Error P>|t| L1.PTD 108,527.1 15,302.3 0.000 R2=0.4692 #Obs=367 #Groups=84 I started simply by regressing the number of passing touchdowns scored in the previous season (L.PTD) on the cap value of the quarterback in the next season. With a high level of significance, it can be said that scoring one extra passing touchdown is predicted to increase a quarterback’s compensation in the next season by approximately $108,527. So a very large premium is placed on scoring an additional passing touchdown.