Race and the Evaluation of Signal Callers in the National Football League

1
Race and the Evaluation of Signal Callers in the
National Football League

• David J. Berri (California State University,
  Bakersfield)
• Rob Simmons (Lancaster University Management
  School)

2
NFL Labour Market

• Entry of players normally by college draft
• Free agency normally only after 4 years
• Some monopsony power for owners
• Binding salary cap 55-64 of designated team
  revenues allowable for team wage bill
• Multi-period contracts with many contingency
  clauses
• Player trades but not usually for cash

3
Concern About Black Representation in NFL

• Few black quarterbacks though increasing in
  number 1/35 in 1971, 3/43 in 1994, 11/36 in 2006
• NFL wants to encourage more black head coaches-
  Rooney rule for head coach hires
• No black owners (yet)
• Segregation Blacks dominate specialist offense
  defence Whites dominate quarterback, offensive
  line kickers
• Is there salary discrimination for black QBs?

4
Quarterback Rating(Completion rate -
0.3)/0.2 (Yards/Pass attempt -3)/4
Touchdowns/Pass attempt/0.05 (0.095 -
Interceptions/Pass attempt)/0.04100/6Maximum
score 158.2 (?)Really this is a
passing-efficiency measure
5
Weakness of QB Rating

• Does not include rush yards
• And black QBs rush more than white QBs
• Are black QBs compensated for this aspect of
  their performance?
• An alternative measure that does include QB rush
  yards is QB score
• All yards (3all plays) (30all turnovers)
• Simpler measure, black QBs more productive than
  white on this measure- both per game and per play

6
Literature on Pay Discrimination in
SportsBasketball Hamilton, App Econ 1997
uses quantile regressions to show whites earn
less than blacks at lower end of pay distribution
while blacks earn less than whites (18) at top
endBodvarsson/Partridge, Lab Econ 2001 no
evidence of employer discrimination but find
evidence of white co-worker nonwhite customer
discriminationNFL Kahn, ILRR 1992 some
evidence of customer discrimination- white
(black) salaries vary positively with whites
(blacks) in local metropolitan areas
7
Salary Statistics Pass Attempts 0, 1995-2006
8
Kernel Density White Salary
9
Kernel Density Black Salary
10
Black-White Salary DistributionBlack
skewness 0.72 kurtosis 2.49White skewness
1.07 kurtosis 3.38
11
How to Measure Performance?

• Quarterback rating
• Quarterback score- includes rush yards
• Detailed measures e.g. pass yards, touchdowns,
  fumbles, interceptions
• Just pass yards and rush yards

12
?The ModelLOG SALARY ?0
?1PERFORMANCE ?2PROBOWL ?3EXPERIENCE
?4EXPERIENCE SQ ?5POP ?6VETERAN ?7CHANGE
TEAM ?8OFFENSE SALARY ?9DRAFT ROUND 1
?10DRAFT ROUND 2 ?10BLACK ?11BLACKPERFORMANCE
error
13
Estimation IssuesSeparate black/white
equations not enough observationsBLACK
interaction termsQuantile regression- since
salary distributions are skewed
14
Control variables for salary

• Standard concave shape of salary-experience
  profile, at median salary is maximised at 8 to 9
  years
• High draft round has positive effect on salary,
  as does veteran (free agency)
• Changing team lowers salary
• Local population size has no effect on salary
  (revenue sharing in NFL)
• Greater total offense salary is associated with
  higher QB salary
• Career pass attempts is a significant predictor

15
Quantile Regression of Log Real Salary
16
With QB score
17
With full performance measures
18
With pass yards rush yards only
19
Conclusions

• Few opportunities for black quarterbacks until
  recently
• Quarterback salary depends on draft position,
  veteran status, change of team, experience and
  number of career pass attempts
• Important to identify an appropriate measure of
  performance QB SCORE is suitable
• Black quarterbacks have less dispersed and less
  skewed salary distribution relative to blacks
• Little evidence of wage discrimination evaluated
  at means
• Rush yards are not rewarded as a performance
  measure- although this is a special attribute of
  black QBs
• But some evidence that black quarterbacks get
  lower salaries than whites from quantile
  regressions when BLACK is interacted with PASS
  YARDS
• This discrimination result applies at median and
  above