University of Northern Iowa University of Northern Iowa
UNI ScholarWorks UNI ScholarWorks
Honors Program Theses Student Work
2016
The determinants of NFL player salaries The determinants of NFL player salaries
Trevor Draisey
University of Northern Iowa
Let us know how access to this document bene7ts you
Copyright ©2016 Trevor Draisey
Follow this and additional works at: https://scholarworks.uni.edu/hpt
Part of the Sports Studies Commons
Recommended Citation Recommended Citation
Draisey, Trevor, "The determinants of NFL player salaries" (2016).
Honors Program Theses
. 213.
https://scholarworks.uni.edu/hpt/213
This Open Access Honors Program Thesis is brought to you for free and open access by the Student Work at UNI
ScholarWorks. It has been accepted for inclusion in Honors Program Theses by an authorized administrator of UNI
ScholarWorks. For more information, please contact [email protected].
Offensive Materials Statement: Materials located in UNI ScholarWorks come from a broad range of sources and
time periods. Some of these materials may contain offensive stereotypes, ideas, visuals, or language.
Running Head: THE DETERMINANTS OF NFL PLAYER SALARIES
THE DETERMINANTS OF NFL PLAYER SALARIES
A Thesis Submitted
In Partial Fulfillment
Of the Requirements for the Designation
University Honors
Trevor Draisey
University of Northern Iowa
May 2016
This Study by: Trevor Draisey
Entitled: The Determinants of NFL Player Salaries
Has been approved as meeting the thesis or project requirement for the Designation
University Honors
______________ ____________________________________________________
Date Dr. Lisa Jepsen, Honors Thesis Advisor, Economics Department
_______________ ____________________________________________________
Date Dr. Jessica Moon, Director, University Honors Program
THE DETERMINANTS OF NFL PLAYER SALARIES
Abstract
Using performance data from 2013 and salary data from 2014 for 426 offensive skill position
players in the National Football League (NFL), this study analyzes the determinants of player
salaries in rookie and veteran NFL contracts. It is the first study to use fantasy football statistics
as a measure of performance across positions. The results indicate that performance, as measured
by fantasy football statistics, is the primary determinant of veteran players’ salaries. Under the
2011 NFL Players Association Collective Bargaining Agreement (CBA), draft position is the
primary determinant of rookie players’ salaries.
THE DETERMINANTS OF NFL PLAYER SALARIES 1
Introduction
The National Football League (NFL) is the largest professional sports league in the
world, earning revenues in excess of 12 billion dollars (Isidore, 2015) and employing almost
1,700 players during the regular season (Davis, 2014). This research studies the factors that
determine the effective salaries of NFL players. A major obstacle to objective compensation
comparisons in past research has been the inability to account for player performance across
positions. This research bridges that gap using fantasy football statistics, which have become an
extremely popular measure of a player’s skill in the eyes of the average fan. This paper provides
a unique contribution to the existing literature because it is the first to incorporate fantasy
football statistics.
Using salary data from a set of 426 offensive skill position players in the 2014 season and
performance data from 2013, I study the correlations of fantasy football statistics and off-field
factors with player salaries and hope to answer the following research question: Is performance
the primary determinant of NFL player salaries? In addition, this paper analyzes the structural
changes to the NFL’s collective bargaining agreement (CBA) with its players union to explain
observed changes in the personnel and contract decisions by NFL franchises. Through this
analysis, this research attempts to answer the secondary research question: How do the
determinants of player salaries differ between veteran and rookie NFL contracts?
By dividing the data set into veteran and rookie contracts, I analyze the differences in
determinants of salaries among players with negotiation powers under the CBA. The results
indicate that performance and draft position are the primary determinants of player compensation
among veteran and rookie contracts, respectively. Fantasy statistics alone explain nearly half of
all variation in cap values among veteran players. Given the explanatory power of this variable,
THE DETERMINANTS OF NFL PLAYER SALARIES 2
excluding it (or a similar measure of performance) from an analysis would yield weak results.
Non-performance related factors, such as race and arrest history, exhibit significant negative
correlations with compensation among veterans, but the relationships are insignificant among
players on rookie contract.
Literature Review
In the National Football League (NFL), the cost of talent makes up the largest portion of
a team’s annual costs, and a team’s success is measured by its ability to win games (Fort, 2011).
All teams in the NFL are subject to the “salary cap”—a maximum amount they can pay for labor.
The salary cap contributes to a competitive balance among the 32 NFL teams by providing all
teams with equal opportunities to acquire top talent (Larsen et al., 2006). Each year, the salary
cap is calculated using a formula established in the collective bargaining agreement (CBA)
between the NFL and the NFL Players Association (NFLPA).
Under the 2006 CBA, the salary cap was computed as all football-related revenues,
minus one billion dollars allocated to the team owners, divided evenly among the 32 teams.
Under this system, players collected about 60 percent of league revenues in the form of salaries
in 2006 (Quinn, 2012). Restructuring the salary cap was a point of emphasis in negotiating a new
CBA after the 2010 season (Quinn, 2012). In the 2011 CBA, the salary cap was restructured as
the sum of 55 percent of revenue from national media contracts, 45 percent of league licensing
revenues, and 40 percent of local team revenues.
A specific portion of each player’s contract counts against the team’s total salary cap. A
player’s cap value at the start of a year includes all guaranteed elements in the contract, any
incentives deemed “likely to be earned,” as well as a fraction of the player’s signing bonus
(NFLPA, 2011). Although signing bonuses are paid up-front, for cap purposes they are
THE DETERMINANTS OF NFL PLAYER SALARIES 3
amortized on a straight-line basis over the life of the contract or five seasons, whichever comes
first (NFLPA, 2011). In the case that incentives actually paid in a year exceed those that were
likely to be earned, the excess will be credited against the team’s salary cap in the subsequent
year. Within the constraints of the collective bargaining agreement, the determination of a
player’s salary may depend on a number of factors. For example, the structure of rookie
contracts changed significantly under the 2011 CBA and is quite different than veteran contracts.
Prior to 2011, rookies drafted into the NFL had the potential to sign extremely lucrative
contracts before setting foot on the playing field. During negotiations for the 2011 CBA, both
owners and veteran players wanted to limit the size of rookie contracts (Brandt et al., 2013).
Owners were motivated by major draft busts, like Ryan Leaf and JaMarcus Russell, and veteran
players were unhappy about being out earned by rookies (Brandt et al., 2013). As a result, the
pool of money allocated to rookie contracts shrunk, and rookies were forced into heavily
structured, four-year contracts with fifth-year team options for first round picks (Quinn, 2012).
Players do not have the option to restructure rookie contracts until after the third year of the
contract. Consequently, many young players receive compensation well below their value, and
the value of rookie contracts has dropped significantly since the implementation of the 2011
CBA (Brandt et al., 2013).
Veteran players may not receive the intended benefits from the rookie contract
restructuring. Teams can draft and sign rookie players for relatively lower salaries than veteran
players without the necessity to restructure contracts to reward performance above expectations
(Brandt et al., 2013). There is little incentive for teams to sign more expensive veterans with a
shorter shelf life. To players nearing the end of their careers, this may have very serious personal
financial implications. The rate of bankruptcy among players 12 years out of the league is about
THE DETERMINANTS OF NFL PLAYER SALARIES 4
16 percent, largely independent of the career earnings of the player (Carlson et al., 2015). In
other words, the length and value of a player’s contracts over the duration of his career have little
to no effect on the possibility of bankruptcy post-retirement. Therefore, veteran players have an
incentive to remain in the league to prolong their financial wellbeing, regardless of how much
they earned over their prior contracts. The league and the players union have taken action to
assist veteran players in finding employment with efforts like the veteran combine, where players
go through similar physical testing as rookies entering the draft. However, these efforts have
been largely ineffective, thus far.
A team’s coaches and front office personnel determine the value of each player on a
position-by-position basis. Fort (2011) defined the value of an athlete as the player’s addition to
the team’s winning percentage multiplied by the marginal revenue generated by that player’s
addition to winning percentage. Essentially, this is the dollar value of the team’s improvement as
a result of adding the player. The most obvious measure of a player’s contribution is his
statistics. Comparing player statistics across positions is difficult, however, because not all
positions contribute equally to the outcome of a game or season. For example, a quarterback may
play a larger role in the outcome of each game than a single wide receiver. Comparisons of
statistical performance across positions require a uniform system to convert raw statistics into a
single measure of value.
In recent years, the growing popularity of fantasy football has provided one way to
compare performance across many positions. The comparison is limited to offensive skill
positions, as points are awarded on an individual basis for yards gained and points scored.
Defenses are scored as a team, rather than individually, and offensive linemen do not receive
fantasy scores, so they are excluded from this research. In ESPN standard scoring leagues, one
THE DETERMINANTS OF NFL PLAYER SALARIES 5
point is earned for every ten rushing yards or twenty-five passing yards. In addition, six points
are awarded for a rushing touchdown and four for a passing touchdown. Fumbles and
interceptions count as a loss of two points (ESPN.com/fantasy/football). As part of employers’
valuation processes, they may also consider outside factors that could keep players off the field
entirely and minimize their contributions to team success. Even the best performer on the field is
useless in a situation where he is not allowed to play. One such consideration is arrest history.
USA Today’s online sports database documented over 800 arrests of NFL players since
2000 (www.usatoday.com/sports/nfl/arrests/). Under the league’s personal conduct policy,
players miss significant playing time for criminal offenses. Consider, for example, the case of
Adrian Peterson, the star running back for the Minnesota Vikings. Due to legal issues near the
beginning of the season, Peterson only played one game during the 2014 season. In this case, off-
the-field actions eliminated the marginal productivity he could provide his team. Although the
overall crime rate among NFL players is lower than the crime rate among the general population,
there is some evidence that NFL players commit violent crimes at a higher rate (Leal et al.,
2015). Even if the statistics do not indicate a strong relationship, almost 70 percent of Americans
believe that the NFL has an epidemic of domestic violence (Leal et al., 2015).
The NFL commissioner determines league punishment for off-field behavioral issues,
and he is responsible for maintaining the league’s public image. He has the power to suspend a
player from his team and keep him from playing for an extended period of time (NFLPA, 2011),
and he may use that power to indicate to the public that the league does not tolerate violent
crime. A player with a history of behavioral issues may pose the risk of wasting a team’s cap
space or drawing the ire of fans. Complicating the matter, not all behavioral issues may be
indicated by a player’s arrest history.
THE DETERMINANTS OF NFL PLAYER SALARIES 6
Weir and Wu (2014) find that an arrest history in college, whether the player is formally
charged or not, correlates to a significant fall in draft position—between 16 and 22 spots. The
fall in draft position does not have a negative impact on NFL performance, however. This is
consistent with the idea that a player’s value falls when he has a history of off-the-field issues,
regardless of performance. Weir and Wu (2014) also find that suspensions enforced for
noncriminal violations, such as team or university violations, lead to a similar drop in draft
position and correlate with worse performance in the NFL. The performance drop-off may be
indicative of attitude issues and difficulty getting along with coaches and teammates. Because
these qualities are purely subjective, teams must make judgments about potential.
All of the aforementioned factors directly relate to a player’s contributions on the field. It
may be possible, however, that factors completely separate from performance influence the
valuation process. Fan discrimination may affect management’s decision to sign an athlete who
could improve the team’s performance (Fort, 2011). When fans prefer to watch players with
certain characteristics unrelated to performance, the marginal revenue generated by a player may
suffer and influence management’s valuation. Kahn (1992) finds evidence consistent with fan
discrimination based on race by studying the percent of white residents in the metropolitan area
in which an NFL team played. He finds that white players receive higher salaries in areas with a
high percentage of white residents, while non-white players see higher salaries in areas with
higher percentages of non-white residents.
Discrimination may also occur at the ownership or teammate level. The wealthy owners
of NFL organizations may value their own preferences over potential lost revenue from
discrimination. Teammate discrimination occurs when players do not wish to interact with
another player for reasons outside of performance. For example, the introduction of Jackie
THE DETERMINANTS OF NFL PLAYER SALARIES 7
Robinson into Major League Baseball exposed significant teammate discrimination. These types
of discrimination are not profit maximizing and are not expected to persist (Fort, 2011). Because
the NFL has little competition in the market for labor, however, owner and teammate
discrimination could persist in the long run (Kahn, 1992).
One factor that may lead to discrimination is the racial identity of the player. Keefer
(2013) finds that black linebackers in the NFL earn ten percent less than their white counterparts.
However, a similar study by Burnett and Van Scyoc (2013) does not find statistically significant
differences in the salaries of rookie wide receivers with respect to race. The effect of a player’s
race on salary is unclear from the literature. Any effects that race may have on player salary are
most likely to occur in the initial hiring process. When making decisions on player retention,
discrimination may play a smaller role due to the fact that players have had an opportunity to
prove themselves for that specific employer (Conlin and Emerson, 2006).
Methodology
For this study, salary data for 426 offensive skill position players are collected from
Spotrac.com, a partner of USA Today Sports Media Group. The highest paid player and average
cap value for each position are listed in Table 1.
Table 1
Highest Paid Player by Position
Only players who have an active contract and filled a spot on the 53-man roster of one of
the 32 NFL organizations at the time the data were collected are included. NFL.com provides
Position
Highest-Paid
Player
Team
Cap Value (2014)
Average Cap Value
by Position
QB
Eli Manning
New York Giants
$20,400,000
$5,275,071
RB, FB
Adrian Peterson
Minnesota Vikings
$14,400,000
$1,734,462
WR
Mike Wallace
Miami Dolphins
$17,250,000
$2,530,006
TE
Jason Witten
Dallas Cowboys
$8,412,000
$1,792,419
THE DETERMINANTS OF NFL PLAYER SALARIES 8
current rosters for all 32 teams and statistics for individual players. I collect data on years of
experience, games played in the 2013 season, and race (white vs. nonwhite) from this website. I
collect data on the number of Pro Bowls to which a player has been selected and whether a
player was a first-round draft pick from Pro-Football-Reference.com.
I can identify players who entered free agency following the 2013 season, including both
restricted and unrestricted free agents, from Scout.com, a Fox Sports affiliate. A player enters
restricted free agency when his contract expires after his third year with a team. In this situation,
the current team is allowed to match any qualifying offer the player has received from other
teams; the player must accept the matching offer. In contrast, unrestricted free agents have
completed a contract of at least four years in length and are allowed to sign with any team.
Commonly, players who demonstrate value above their current contract leverage that position to
renegotiate their contract prior to reaching free agency. These players are not classified as free
agents.
I collect arrest data for all NFL players from the beginning of 2000 to 2013 from the USA
Today online sports database (www.usatoday.com/sports/nfl/arrests). I include only players with
active contracts, so recent cases in which the player was subsequently cut from the team’s roster,
such as Ray Rice, are not included. I collect 2013 fantasy football statistics and team
performance data from ESPN.com. The median household income and percentage of white
residents for the urban areas in which each NFL team play are available from the U.S. Census
website (www.census.gov).
Data and Descriptive Statistics
Teams are subject to a finite salary cap that varies year to year ($133 million in 2014), so
cap value is an accurate representation of the relative value that a team places on each player. For
THE DETERMINANTS OF NFL PLAYER SALARIES 9
this reason, Cap Value is the dependent variable. The average cap value of the players in this
data set is $2,593,659. Of the positions included in this study, quarterbacks have the highest
average cap value by a wide margin. This indicates that teams generally value the quarterback
over other offensive skill positions. Eli Manning is the highest paid player in the data set with a
cap value of $20,400,000 in 2014.
The independent variables are measures of individual characteristics, individual
performance, and team and city characteristics. Arrest is a dummy variable with a value of one if
the player has been arrested at least once from 2000 to 2014 and zero if the player has not been
arrested over that time period. About eight percent of the players in this study have been arrested.
Non-White is a dummy variable equaling one if the player is non-white and zero if the
player is white. About 66 percent of the players are non-white.
Experience, Experience Squared, First Round Pick, Undrafted, Games Played, Percent
Games Started, Pro Bowls, and Fantasy PPG are variables designed to indicate an individual
player’s performance. Experience represents the number of years that a given player has
participated in the NFL. With seventeen years, Peyton Manning has the most experience. The
average number of years of NFL experience is 5.13. Experience Squared represents the square of
the number of years a player has been in the NFL. This variable will control for the effects of
diminishing productivity that would be expected at older ages due to the highly physical nature
of professional football.
First Round Pick is a dummy variable equaling one if the player was selected in the first
round of the NFL draft and zero if drafted in any other round or undrafted. About 18 percent of
the players are first round draft picks. Undrafted is a dummy variable with a value of one for
players who were not selected in the NFL draft. These players were signed as undrafted free
THE DETERMINANTS OF NFL PLAYER SALARIES 10
agents after the draft process concluded. Over 25 percent of the players in this data set went
undrafted.
Games Played represents the number of games in which the player participated during
the 2013 season. The average is nearly 12 games. Percent of Games Started indicates the
percentage of the 16 game season that the player started in 2013. The average percent of games
started was about 39 percent.
Pro Bowls indicates the number of times that a player has been selected to the Pro Bowl
over the course of his career. With thirteen appearances, Peyton Manning has been selected the
most often. The average number of Pro Bowls is 0.57.
Fantasy PPG represents the average number of fantasy points a player scored in ESPN
standard scoring leagues in each game that they appeared, computed as the total number of
fantasy points earned in the 2013 season divided by the number of games in which the player
appeared. Peyton Manning averaged over 25 points per game, the league high for 2013. In
comparison, the league average was less than 5 points per game.
Free Agent is a dummy variable that takes a value of one if a player became a free agent
following the 2013 season and zero otherwise. About 24 percent of the players were free agents
who re-signed with their 2013 team or signed with a different team for the 2014 season.
Team performance data indicates how effectively teams allocate finite cap space to
optimize performance. Playoffs is a dummy variable with a value of one if the player’s team
reached the playoffs in 2013 and zero if it did not. For players who were traded or signed by
another team out of free agency, the team that they played for in 2013 is used instead of their
current team. Each year, 12 out of 32 teams participate in the NFL playoffs. Team Winning
THE DETERMINANTS OF NFL PLAYER SALARIES 11
Percentage indicates the percent of the games won over the course of the 16 game 2013 season.
Each player’s team is determined in the same manner as the Playoffs variable.
Kahn (1992) includes variables to correct for variations in demographics among the cities
where NFL teams are located. Median Income and Percent White Residents describe the
demographics in the urban areas where each team plays. Foxborough, Massachusetts, home of
the New England Patriots, has the highest percentage of white residents at 90.3 percent. In
contrast, Detroit has the lowest percent of white residents: 10.6 percent. San Francisco has the
highest median household income ($73,802), while Detroit has the lowest ($26,325).
Descriptive statistics for the variables are reported in Table 2 for the total data set, as well
as veteran and rookie contract subsets. I define a player on a rookie contract as any player with
two or three years of experience, as contracts cannot be renegotiated until after the third contract
year under the 2011 CBA. I cannot include first-year players because they lack performance
statistics from 2013. A veteran is defined as any player with more than three years of experience.
Model
I use ordinary least squares regression to estimate the effects of the individual, team, and
city variables on the compensation of NFL offensive skill-position players. Veteran and rookie
contracts are modeled separately due to the effects of the 2011 CBA on rookie contract
structures. The variables included in the final models were selected based on theoretical
significance and inclusion in previous literature. To provide insight into the explanatory value of
each independent variable, the dependent variable was also regressed against each independent
variable individually. The R
2
from the individual regressions are presented in appendix Table 1.
THE DETERMINANTS OF NFL PLAYER SALARIES 12
Table 2
Descriptive Statistics
Variable
Mean (Standard Deviation)
Total Data Set Veteran Contracts Rookie Contracts
Cap Value
3,700,403
(4,261,482)
807,931
(760,036)
ln(Cap Value)
14.571
(1.063)
13.412
(0.531)
Fantasy PPG
5.789
(5.272)
3.574
(4.197)
Arrest
0.118
(0.323)
0.031
(0.173)
Nonwhite
0.624
(0.485)
0.718
(0.451)
Experience
6.814
(2.710)
N/A
(Experience)
2
53.741
(45.646)
N/A
First Round Pick
0.228
(0.420)
0.098
(0.298)
Undrafted
0.186
(0.390)
0.356
(0.480)
Pro Bowls
0.871
(1.834)
0.074
(0.305)
Games Played
12.114
(5.059)
11.509
(5.305)
% Games Started
0.475
(0.381)
0.263
(0.313)
Free Agent
0.369
(0.483)
0.025
(0.155)
Team Win %
0.518
(0.191)
0.491
(0.193)
Playoffs
0.430
(0.496)
0.362
(0.482)
% White Residents
0.544
(0.181)
0.542
(0.170)
Median Income
47,000
(11,500)
44,992
(10,610)
ln(Median Income)
10.728
(0.248)
10.687
(0.236)
N 426 263 163
THE DETERMINANTS OF NFL PLAYER SALARIES 13
The model in equation [1] applies to all players and to veterans.
[
1
]

(
 
)
=
0
+
1
∗  +
2
∗  +
3
∗ ℎ +
4
∗ 
+
5
∗ 
2
+
6
∗  +
7
∗  +
8
∗  +
9
∗  +
10
∗ % +
11
∗ %ℎ +
12
∗ 
(

)
In this model, the dependent variable is the natural log of the cap value. Cap value does
not represent all of the cash a player may receive in a given year, but it provides an excellent way
to compare salaries because it captures the value of a player relative to the total team salary cap.
Signing bonuses are amortized on a straight-line basis over the term of a player’s contract, even
though the player may receive the entire bonus up front (NFLPA, 2011). Variables to measure
performance are included in model [1] because veteran players have had the opportunity to
negotiate a salary based on performance. Strong correlations among Games Started, Games
Played, and Fantasy PPG motivated the decision to include only the latter variable to minimize
the effects of multicollinearity in the model. Similar concerns about the correlation between
Team Win Percentage and Playoffs motivated the decision to include only the former.
Model [2] is designed to capture the factors that determine Cap Value in rookie contracts.
[
2
]

(
 
)
=
0
+
1
∗ ℎ +
2
∗  +
3
∗  +
4
∗  +
5
∗ %ℎ +
6
∗ 
(

)
Like model [1], ln(Cap Value) serves as the dependent variable. The difference between
the two models is the absence of individual and team performance variables in model [2].
THE DETERMINANTS OF NFL PLAYER SALARIES 14
Because players cannot renegotiate rookie contracts until after the third year, performance cannot
exert positive influence on rookie contract cap value. At the inception of the contract,
performance is an unknown. Therefore, cap value may only vary with performance to the extent
that future performance is explained by known factors at the inception of the contract. Free
Agent is included in this model because teams have the option to release a player at any time, so
a player with two or three years of experience may still be classified as a free agent.
Predicted Signs of Coefficients
For each variable in the models, I specify the expected sign of the coefficient in the fitted
model. Expected signs are determined by examining the findings of past literature and applying
the theoretical relevance of each variable. A positive sign indicates that I expect an increase in
the value of the variable to correlate with an increase in the natural logarithm of Cap Value. A
negative sign indicates that I expect an increase in the value of the variable to correlate with a
decrease in the natural logarithm of Cap Value.
Assuming that League punishment following a player’s arrest affects playing time and
productivity, a previous arrest should negatively affect player salary. The most extreme case of a
negative impact on salary—a team releasing a player—is not accounted for in this study, as only
players currently occupying a roster spot are included in the data set. A team may retain a player
after being arrested if it values his contributions to the team enough to compensate for the
negative publicity they may receive. Until contract renegotiations, the cap value of players in this
situation may not be affected.
A survey conducted to measure the public attitude toward Michael Vick’s criminal
punishment and reinstatement to the League indicates that white respondents tend to support
harsher punishment of players who have been arrested (Piquero et al., 2011). As a result, arrests
THE DETERMINANTS OF NFL PLAYER SALARIES 15
may have a larger effect on salary among teams located in cities with a high percentage of white
residents. Overall, there is no evidence to suggest that arrest would have a positive effect on
salary. I expect a negative coefficient for the Arrest variable.
In the market for linebackers from 2001-2009, Keefer (2013) find evidence consistent
with differences in salary based on race. The results suggest that white linebackers receive ten
percent higher salaries on average than their black counterparts. Kahn (1992) shows that the
correlation between race and salary is dependent on the racial demographics of the urban area
where the team plays. He finds that white players tend to earn more than non-white players in
areas with a high percentage of white residents, while non-white players tend to earn more in
areas with a high percentage of non-white residents. In contrast, Gius and Johnson (2000)
conclude that white players earn ten percent less than black players. Another recent study by
Burnett and Van Scyoc (2013) finds no differences in the salaries of rookie wide receivers in the
NFL based on race. As this data set closely resembles my subset of rookie contracts, I expect to
see a similar lack of correlation between race and compensation among players in their rookie
contracts. To the extent that correlation exists, I would expect the sign of Nonwhite to be
negative, as found in the previous literature.
Following the NFL’s new collective bargaining agreement in 2011, the value of rookie
contracts has dropped significantly (Brandt et. al., 2013). In addition, all new rookie contracts
last for four years and cannot be renegotiated until after the completion of three full seasons.
Consequently, many young players receive compensation well below their values (Brandt et. al.,
2013). As a result, I expect Experience to have a positive coefficient in the total and veteran data
sets, as players with the skill level to continue playing beyond their rookie contracts sign new,
more lucrative contracts following their third or fourth year in the league. Following labor
THE DETERMINANTS OF NFL PLAYER SALARIES 16
economics theory, I expect the sign of Experience Squared to be negative as players face
diminishing productivity at older ages. Experience and Experience Squared are not included in
the model for rookie contracts because each variable would have only two possible values.
NFL teams spend first round draft picks on players they believe possess the potential to
have the largest impact on the future of the organization. As such, they should place a larger
value on players selected early in the draft resulting in a positive coefficient. I expect First
Round Pick to be positive in the rookie contract subset, as first round draft picks reflect high
performance expectations. The salary restrictions imposed on rookies in the 2011 collective
bargaining agreement may mitigate this effect in the total data set, as the decrease in average
rookie contract value post-2011 CBA may reduce the average salary of first round picks in the
total data set. I expect this variable to be weaker in predicting veteran contracts, as these players
have performed well enough to warrant new or continued contracts regardless of draft position.
In all cases, I expect the sign of coefficient for First Round Pick to be positive.
I expect the Undrafted variable to take the opposite sign of the First Round Pick variable,
and I expect the magnitude to be nearly the same. I expect undrafted players among the rookie
contract to have especially small contracts, as they must prove they are capable of performing at
the same level as players deemed to have more potential in the draft. Much like the First Round
Pick variable, I expect the effect of Undrafted to diminish in the total data set and veteran subset,
as compensation decisions move toward measured, rather than expected, NFL performance.
Standout players at each position are selected to participate in the Pro Bowl each year.
Many Pro Bowl appearances suggest that a player has performed extraordinarily well over an
extended period of time and may warrant a higher salary. I expect a positive correlation between
Pro Bowls and Cap Value.
THE DETERMINANTS OF NFL PLAYER SALARIES 17
Players who outperform their contracts have the opportunity to leverage their
performance to negotiate for higher salaries in free agency. Such performance may also enable a
player to negotiate a contract extension that would keep him from entering free agency at all.
The frequency of these contract extensions may reduce the number of highly valuable players
entering free agency and so reduce the positive effects on salary. With the new collective
bargaining agreement limiting the value of rookie contracts, teams may find it a more efficient
allocation of cap space to draft and sign multiple rookies instead of signing one veteran from free
agency (Brandt et al., 2013). As a result, free agency may be negatively correlated with salary.
The sign of the coefficient is difficult to predict in the veteran subset because I cannot measure
the prevalence of contract renegotiation. Because players on rookie contracts do not have this
renegotiation option, I expect the sign of Free Agent to be negative within the rookie subset. A
team has no incentive to release a player on a rookie contract into free agency unless he performs
below the value of the contract and show little future potential. Such expectations would likely
result in a smaller contract with a new team or the player leaving the league entirely.
Fantasy football provides a unique way to evaluate an offensive player’s contribution on
the field. By awarding points for yards and touchdowns and penalizing players for turnovers,
fantasy football allows cross-positional comparisons of performance. I expect Fantasy PPG to be
positively related to player cap value.
Exceptional team performance should correlate to higher salaries. Assuming a successful
team is able to identify the players who make the largest contribution, those players should
receive a relatively high salary. Offensive skill position players may receive more credit for a
team’s success, as their contributions are easily observed. I expect Team Winning Percentage to
be positively related to player salary.
THE DETERMINANTS OF NFL PLAYER SALARIES 18
As outlined by Kahn (1992), Percent White Residents may affect salaries for white and
non-white players differently. The sign of the coefficient may depend on the distribution of white
and non-white players in certain urban areas. Median Income should have a minimal effect on
salary because each team is operating under the same salary cap, and it is in each team’s best
interest to efficiently allocate the entirety of the cap space.
Results and Discussion
Table 3 contains the results from the ordinary least squares regression used to estimate
the empirical models specified in [1] and [2]. Model R
2
in Appendix Table 1 represents the R
2
value when the natural log of cap value is analyzed against that single regressor.
Fantasy PPG is statistically significant (p<.01) in both models. The result from the
veteran contract model indicates that a one-point increase in fantasy points per game correlates
with a 10.6 percent increase in cap value.
1
The Fantasy PPG model R
2
(.46) indicates that this
variable explains nearly half of the variation in the natural log of cap value for veteran players.
Of the variables considered in the veteran contract model, Fantasy PPG has the greatest
explanatory value by a wide margin. The veteran contracts model provides the clearest image of
the true correlation between fantasy performance and cap value because it excludes rookie
contracts, which cannot be influenced by fantasy performance.
Arrest is statistically significant (p<.05) in both models. The influence is clearest in the
veteran data set. The model for veteran contracts indicates that a past arrest correlates with a 26.1
percent decrease in cap value.
1
Because the dependent variable is the natural log of cap value, coefficients are interpreted as

− 1.
THE DETERMINANTS OF NFL PLAYER SALARIES 19
Table 3
Regression Results
Dependent Variable—ln(Cap Value)
Variable
Coefficient (Standard Error)
Total Data Set Veteran Contracts Rookie Contracts
Constant
13.228**
(1.349)
11.917**
(1.963)
15.241**
(1.349)
Fantasy PPG
0.080**
(0.007)
0.101**
(0.009)
N/A
Arrest
- 0.261*
(0.103)
- 0.302*
(0.128)
NA
Nonwhite
- 0.161**
(0.060)
- 0.184*
(0.087)
0.034
(0.058)
Experience
0.483**
(0.038)
0.448**
(0.076)
N/A
(Experience)
2
- 0.026**
(0.003)
- 0.025**
(0.005)
N/A
First Round
Pick
0.512**
(0.082)
0.379**
(0.110)
1.236**
(0.090)
Undrafted
- 0.137*
(0.067)
- 0.058
(0.107)
- 0.247**
(0.480)
Pro Bowls
0.132**
(0.029)
0.118**
(0.034)
N/A
Free Agent
- 0.374**
(0.073)
- 0.383**
(0.088)
- 0.519**
(0.174)
Team Win %
0.017
(0.152)
- 0.133
(0.225)
N/A
% White
Residents
0.156
(0.183)
0.173
(0.264)
- 0.060
(0.176)
ln(Median
Income)
- 0.097
(0.131)
0.041
(0.187)
- 0.172
(0.130)
Adjusted R
2
0.7119 0.6303 0.6191
F-Statistic F(11,414)=88.51** F(11,251)=38.23** F(6,156)=44.89**
N 426 263 163
**Significant at the one-percent level
*Significant at the five-percent level
THE DETERMINANTS OF NFL PLAYER SALARIES 20
The Nonwhite variable provides one of the more interesting results in this research. The
models indicate that race is statistically significant in the total (p<.01) and veteran (p<.05)
contract data sets, but it is not statistically significant in the model for rookie contracts. The
model for veteran contracts indicates that nonwhite players receive, on average, 16.8 percent less
than their white counterparts. I suspect that this result may be partially due to the fact that 13 of
the top 15 cap values in the veteran contract data set belong to white starting quarterbacks. The
density of high-value, white players at the top of the distribution exaggerates the magnitude of
the coefficient.
Experience and Experience Squared remain very consistent across both models. Each
additional year of NFL experience correlates to a 62.1 percent increase in cap value in the model
for the total data set. The negative coefficient for Experience Squared reflects diminishing
productivity in later years. These results are statistically significant (p<.01).
As expected, being a first round draft pick has a strong, positive correlation with cap
value. This variable, when considered alone, explains over half of the variation in the natural log
of cap values of rookie contracts. It has, by far, the most explanatory power of the variables in
the model for the rookie data set. Among rookie contracts, first round draft picks receive, on
average, 244 percent larger cap values than players not drafted in the first round. For veterans,
first round picks receive 46.1 percent more than their counterparts on average. The variable is
statistically significant (p<.01) across all models.
The correlation between being undrafted and compensation manifests in the opposite
direction of first round picks. It is important to note that the result is only statistically significant
in the model for rookie contracts (p<.01), where going undrafted correlates with a 21.9 percent
THE DETERMINANTS OF NFL PLAYER SALARIES 21
lower cap value. The lack of significance in the veteran model indicates that players who prove
themselves worthy of a contract do not face lower contract values in subsequent contracts.
As expected, Pro Bowls are positively correlated with cap value, with a 12.5 percent
increase for each additional Pro Bowl invitation among veteran contracts. This result is
statistically significant (p<.01).
Free agency following the 2013 season correlates with a 31.8 percent decrease in cap
value. This result is statistically significant (p<.01) and indicates that the most valuable free
agents tend to renegotiate contracts prior to expiration, rather than enter free agency. Among
rookie contracts, free agents received, on average, 40.5 percent lower cap values in the following
year. This relationship is consistent with the assertion that teams only release players from rookie
contracts when they significantly underperform. Teams have an incentive to renegotiate contracts
with players who outperform their current contracts before those players enter free agency. Such
a tactic prevents a player from abandoning the team for more money elsewhere. For players who
underperform in their contracts, teams are not as aggressive. Therefore, underperforming players
are heavily represented in the free agent market, and free agents receive, on average, a pay cut
between 30 and 40 percent.
Team winning percentage is not statistically significant in determining a player’s cap
value. In addition, neither the percentage of white residents nor the median household income in
the area where teams play have statistically significant effects on cap value. The insignificance
of team performance and metropolitan demographics may be a result of the standardized salary
cap by which all teams must abide, regardless of success or location.
THE DETERMINANTS OF NFL PLAYER SALARIES 22
Conclusions
Past research into the determinants of NFL player salaries struggles to compare players
across positions because of the differences in statistical performance measures. This research is
the first to use fantasy football statistics as a formal measure of cross-positional performance.
Using salary data from 2014 and the previous year’s individual and team performance
information, this research analyzes the determinants of NFL player salaries in rookie and veteran
contracts.
The results indicate that race and arrest history are significant factors in determining a
veteran player’s cap value but have little to no significant effects on the cap values of players on
rookie contracts. The results also suggest that, while race and arrests are significant, the greatest
variation in an NFL player’s cap value comes from the player’s individual performance and
career factors. Players receive, on average, 11 percent larger cap values for each additional
fantasy point per game. The statistical significance of these results and the raw explanatory
power of the variable indicate that fantasy football statistics closely mirror the real-life
performance judgments made in the front offices of NFL organizations.
In rookie contracts, draft position is the primary determinant of salary. First round draft
picks receive significantly higher salaries, while undrafted players receive much less. These
differences persist throughout the length of rookie contracts due to the limits on negotiation in
the 2011 collective bargaining agreement.
References
Brandt, Andrew, Richard O’Brien, and Mark Bechtel. 2013. “Boxed In, Squeezed Out.” Sports
Illustrated, May 27. Accessed January 6, 2015.
http://www.si.com/vault/2013/05/27/106326494/boxed-in-squeezed-out.
Burnett, Nancy, and Lee Van Scyoc. 2013. “Compensation Discrimination for Wide Receivers:
Applying Quantile Regression to the National Football League.” Journal of Reviews on
Global Economics 2: 433-441. Accessed January 3, 2015.
http://dx.doi.org/10.6000/1929-7092.2013.02.30.
Carson, Kyle, Joshua Kim, Annamaria Lusardi, and Colin Camerer. 2015. “Bankruptcy Rates
Among NFL Players With Short-Lived Income Spikes,” National Bureau of Economic
Research. Accessed October 13, 2015. http://www.nber.org/papers/w21085.
Conlin, Michael, and Patrick Emerson. 2006. “Discrimination in Hiring Versus Retention and
Promotion.” Journal of Law, Economics and Organization 22, no. 1: 115-136. Accessed
September 9, 2015. http://dx.doi.org/10.1093/jleo/ewj005.
Davis, Wade. 2014. “A Numbers Game: Making an NFL Roster.” Huffington Post, October 28.
Accessed March 26, 2016. http://www.huffingtonpost.com/wade-davis-jr/a-numbers-
game-making-an-nfl-roster_b_5731630.html.
Ducking, Johnny, Peter Groothuis, and James R. Hill. 2014. “Compensation Discrimination in
the NFL: An Analysis of Career Earnings.” Applied Economics Letters 21, no. 10: 679-
682. Accessed January 10, 2015. http://dx.doi.org/10.1080/13504851.2014.884688.
Fort, Rodney. 2011. Sports Economics. 3rd Ed. Upper Saddle River, NJ: Pearson Education, Inc.
Gius, Mark, and Donn Johnson. 2000. “Race and Compensation in Professional Football.”
Applied Economic Letters 7, no. 2:73-75. Accessed January 11, 2015.
http://dx.doi.org/10.1080/135048500351843.
Isidore, Chris. 2015. “NFL Revenue: Here Comes Another Record Season.” CNN Money,
September 10. Accessed March 26, 2016.
http://money.cnn.com/2015/09/10/news/companies/nfl-revenue-profits/.
Kahn, Lawrence M..1992. “The Effects of Race on Professional Football Players’
Compensation.” Industrial and Labor Relations Review 45, no. 2: 295-310. Accessed
January 6, 2015. http://dx.doi.org/ 10.1177/001979399204500207.
Keefer, Quinn and Andrew Wesley. 2013. “Compensation Discrimination for Defensive Players:
Applying Quantile Regression to the National Football League Market for Linebackers,”
Journal of Sports Economics 14, no. 1: 23-44. Accessed January 8, 2015.
http://dx.doi.org/10.1177/1527002511413288.
Larsen, Andrew, Aju Fenn, and Erin Spenner. 2006. “The Impact of Free Agency and the Salary
Cap on Competitive Balance in the National Football League.” Journal of Sports
Economics 7, no.4: 374-390. Accessed January 8, 2015.
http://dx.doi.org/10.1177/1527002505279345.
Leal, Wanda, Marc Gertz, and Alex R. Piquero. 2015. “National Felon League?: A Comparison
of NFL Arrests to General Population Arrests.” Journal of Criminal Justice 43, no. 5:
397-403. Accessed March 30, 2016. http://dx.doi.org/10.1016/j.jcrimjus.2015.08.001.
Leeds, Michael A., and Sandra Kowalewski. 2001. “Winner Take All in the NFL: The Effect of
the Salary Cap and Free Agency on the Compensation of Skill Position Players” Journal
of Sports Economics 2, no. 3: 244-256. Accessed January 5, 2015. http://dx.doi.org/
10.1177/152700250100200304
Mahoney, Josh. 2009. “Does Race Affect the Salaries of NFL Players?” Honors Thesis.
University of Northern Iowa.
NFL Players Association, Collective Bargaining Agreement. August 4, 2011. Accessed
September 9, 2015. https://nfllabor.files.wordpress.com/2010/01/collective-bargaining-
agreement-2011-2020.pdf.
Piquero, Alex, Nicole Leeper Piquero, Marc Gertz, Thomas Baker, Jason Batton, and J.C.
Barnes. 2011. “Race, Punishment, and the Michael Vick Experience.” Social Science
Quarterly 92, no. 2: 535-551. Accessed January 9, 2015. http://dx.doi.org/
10.1111/j.1540-6237.2011.00781.x.
Quinn, Kevin. 2012. “Getting to the 2011-2020 National Football League Collective Bargaining
Agreement.” International Journal of Sport Finance 7:141-15. Accessed September 15,
2015.
Weir, Kendall, and Stephen Wu. 2014. "Criminal records and the labor market for professional
athletes: The case of the National Football League." Journal of Sports Economics 15:
617-635. Accessed September 20, 2015. http://dx.doi.org/10.1177/1527002513487739.
Appendix Table 1
Individual Model R
2
Dependent Variable—ln(CapValue)
Variable
Model R
2
Total Data Set Veteran Contracts Rookie Contracts
Fantasy PPG 0.4384 0.4592 --
Arrest 0.0029 0.0043 --
Nonwhite 0.0135 0.0112 0.0061
Experience 0.3548 0.1485 --
(Experience)
2
0.2831 0.1324 --
First Round Pick 0.2489 0.1964 0.5505
Undrafted 0.0810 0.0300 0.1705
Pro Bowls 0.2848 0.2555 --
Free Agent 0.0005 0.1025 0.0188
Team Win % 0.0025 0.0006 --
% White Residents 0.0002 0.0004 0.0005
ln(Median Income) 0.0000 0.0014 0.0068
N 426 263 163