On the evaluation of kickers in the national football league.
Berri, David J. ; Schmidt, Martin B.
Introduction
One of the main themes of Adam Smith's seminal book The Wealth
of Nations is the central role specialization plays in determining an
economy's rate of economic growth. For example, Smith describes the
process of specialization within the production of a pin factory in the
following way:
One man draws out the wire, another straights it, a third cuts it,
a fourth points it, a fifth grinds it at the top for receiving the head;
to make the head requires two or three distinct operations; to put it on
is a particular business, to whiten the pins is another ... and the
important business of making a pin is, in this manner, divided into
about eighteen distinct operations, which in some manufactories are all
performed by distinct hands, though in others the same man will sometime
perform two or three of them. (Smith, 1776, p. 6)
Smith, recognizing the value of the division of labor, further
argues that when the "business of making a pin is" divided
into smaller distinct operations, the returns to productivity may be
immense. (1)
This lesson was certainly not lost on many in the production
sector. Henry Ford, for example, used these principles in the production
of the Model T. After incorporating the Ford Motor Company in 1903, Ford
went about revolutionizing the auto manufacturing process. By
introducing the assembly line production to the process of producing
automobiles, Ford was able to increase the production of his Model T
from 1 car every 12 hours to 1 car every 93 minutes in 1914, a rate
nearly 5 times the rate of his competitors. (2)
In the world of sports, though, this message often appears lost.
The best basketball players are asked to dribble, shoot, pass, rebound,
and defend. Baseball players are often prized if they have the five
tools, the ability to hit for average, hit for power, run, throw, and
field. In other words, in basketball and baseball, the nonspecialist is
preferred. (3)
In American football, though, the gains from specialization have
clearly been recognized. Whereas in other sports players are encouraged
to be good at everything, football players focus on a very specific list
of skills. On offense, one player throws the ball, others run, still
others catch, and others only block. (4) This lesson also seems to have
been learned across time as, for example, the day of the two-way player
is now a thing of the past.
Perhaps for no position is this focus on specialization more
important than for the kicker. While other players on an NFL team are
responsible for throwing the ball, carrying the ball, catching the ball,
blocking, or tackling, a kicker has none of these responsibilities. (5)
In sum, many of the activities generally associated with football are
not the responsibility of the kicker. Two jobs are generally assigned to
the kicker. The first is kicking off, with the objective being to kick
the ball as far as possible, thus forcing the opposing team to travel
more yards to score. The second task involves scoring, either via field
goals or extra points.
It is this latter task that generally earns a kicker fame or
infamy. For example, Adam Vinatieri's 48-yard field goal as time
expired proved to be the winning margin in the New England Patriots
triumph over the St. Louis Rams in Super Bowl XXXVI. And Jim
O'Brien--with five seconds left in Super Bowl V--kicked a 32-yard
field goal to give the Baltimore Colts a victory over the Dallas
Cowboys. In contrast, Scott Norwood is remembered for missing a 47-yard
field goal with eight seconds left in Super Bowl XXV against the New
York Giants, the first of four consecutive Super Bowl losses for the
Buffalo Bills.
The ability--or inability--to convert field goal attempts (and
extra point attempts) into points certainly stands out when one thinks
about NFL kickers. Aaron Schatz, though, argued in the New York Times
that this focus was misplaced: (6)
Game-winning field goals are what make kickers famous, but from
season to season it is impossible to tell which kickers will be the
most trustworthy in the closing seconds. Instead of wasting money
on high-priced field-goal kickers, teams would be better off
signing kickers who can be counted on to help their teams
consistently by affecting field position with long kickoffs.
Schatz argument reminds one of the arguments advanced in Moneyball.
In the Michael Lewis classic it was argued that Major League Baseball
teams focused too much attention on slugging percentage and not enough
on a player's ability to get on base (measured via on-base
percentage). In other words, the ability to hit home runs captured the
attention of decision-makers while the ability to draw a walk was
undervalued.
The research of Jahn Hakes and Raymond Sauer (2006, 2007) found
support for the story told by Lewis in Moneyball. Specifically, Hakes
and Sauer found that prior to 2004, player salary was primarily
determined by slugging percentage. On-base percentage--in many of the
years the authors examined--failed to have a statistically significant
impact on player compensation, this despite the fact that the on-base
percentage was a significant determinant of winning.
A similar story has been told about wages in professional
basketball. Berri, Brook, and Schmidt (2007) present evidence that a
player's scoring totals have the largest impact on player salaries.
Factors such as shooting efficiency, rebounds, and turnovers--which play
a larger role in determining team wins--are less important to an NBA
player's compensation.
These stories suggest that decision-makers are drawn to the actions
that are the most dramatic during the course of the game. Factors,
though, whose impact on outcomes is harder to judge by just watching the
game (i.e., walks in baseball and nonscoring actions in basketball)
would then be undervalued. (7)
Following this logic, Schatz is arguing that kickers are likely to
be paid for their ability to score; therefore, the ability to excel with
respect to kickoffs may be undervalued. In order to test this hypothesis
we examine the impact that these two factors (i.e., the ability to score
and kicking off) have on a kicker's salary. If the story of
baseball and basketball also exists in football, then one would expect
that kickers' salaries will be overly influenced by the ability to
score, at least relative to each factors' importance to winning.
In the end we find that a kicker's performance with respect to
scoring (i.e., field goals and extra points) has a relatively larger
impact on player compensation than a kicker's performance on
kickoffs--this despite the fact that kickers' performance with
respect to scoring is highly variable. In contrast, a kicker's
performance with respect to the kicker's other job, kickoffs, is
much more consistent across time. In sum, what we see in the NFL with
respect to the evaluation of kickers is consistent with prior studies of
decision-making in basketball and baseball.
Methodological Approach and Data
Methodology
The present paper estimates the determinants of an NFL
kicker's salary. In order to estimate these determinants we follow
the literature and estimate the player's marginal contribution to
product quality--in most cases wins--and also include some measure of
star quality as well as team-specific market demographics (Rosen &
Sanderson, 2001). While the marginal contribution to product quality
captures the impact that a player may have on the firm's bottom
line through greater wins, the measures of star quality capture the
impact that the player may have to a team's revenues through
increased revenues due to fans' desire to see the player.
The first to directly estimate a player's marginal product was
Scully (1974). Scully estimated a player's marginal product through
the player's impact on a team's probability of winning. This
approach follows from the argument that a team's revenue stream is
highly correlated with a team's winning percentage. In which case,
one need only estimate the marginal impact of each of a player's
actions on the field on the team's likelihood of winning to capture
the player's marginal impact to the firm's product quality.
Such studies have become common in the literature for baseball and
basketball markets. (8) For example, while Hakes and Sauer (2006)
estimate a positional player's impact on winning through slugging
and on-base percentage, Bradbury (2007) uses the DIPS measures for
pitchers' values and finds a similar result. For basketball, Berri,
Brook, and Schmidt (2007) estimate individual player contribution
through individual statistics such as shooting efficiency, turnovers,
and rebounds. Each of these finds a disconnection, at least for periods,
between player compensation and player contribution.
For the NFL, Ahlburg and Dworkin (1991) and Berri and Simmons
(2009), among others, examine individual player compensation. Ahlburg
and Dworkin examine the relative importance of seniority on player
compensation. They incorporate individual measures of player
performance, both career and in the past season, as well as position,
draft round, and seniority. In the end, the authors find that an
individual player's salary is significantly affected by the
player's seniority and original draft round. Moreover, they find
that a player's career performance explains two to four times as
much of the variance as the player's previous season's
performance. Berri and Simmons (2009), for example, estimate individual
quarterback productivity by incorporating a quarterback's passing
and rushing yards, as well as touchdowns, completions, and interceptions
per attempt.
Krautmann (1999) critiques the Scully approach and argues that
Scully's estimates of player productivity are biased upward as
these productivity measures are only weakly correlated with free agent
player salaries. Krautmann, in contrast, argues that the competition for
player services, which happens when a player reaches free agency, should
move player salaries closer to their unobservable productivities.
Bradbury (2013), however, argues that while Krautmann's
approach has some merit, it also has several significant drawbacks.
Perhaps the most significant is that the studies described above argue
that private decision-makers in both professional baseball and
basketball do appear to price talent inefficiently. Using the Krautmann
method, which assumes that market prices are correct, would make it
fundamentally difficult to examine efficiency in a sport market.
In the end, our approach employs elements of both Scully's and
Krautmann's approach. Like Scully, we are interested in a
kicker's impact on outcomes. And like Krautmann, we regress a
player's salary on performance--in this case, performance on
kickoffs and scoring actions--and a collection of control variables. In
the case of a kicker, the contribution to team wins would likely come
from their two distinct operations: (1) kicking field goals and extra
points and (2) kicking the ball off after scores and at the beginning of
both halves of football. (9) If we assume that the NFL team's goal
is to maximize its total number of wins, then we would expect that a
kicker's salary should largely be determined by these two actions.
(10) This, then, requires that we first convert both the ability to
score (i.e., field goals and extra points) and kickoffs into their
respective importance to winning. Once we have found a common metric for
kickoffs and scoring we can estimate how each factor drives the
compensation of kickers in the NFL.
Measuring kicker performance
While valuing field goals and extra points would seem simple in
that they may be measured in terms of points scored, which is an
essential component of winning, kickoffs are more difficult. What we
require for kickoffs is ameasure that ties kickoff performance to points
(which might then be tied to winning). Specifically, what is an extra
yard in a kick off worth to a team's probability of winning?
Fortunately, for our purposes, empirical studies that tie field
position to winning probabilities have existed for over 40 years. Carter
and Machol (1971, 1978), for example, estimate the expected point
outcome given that the offensive team has a first down at a particular
yard-line. Their estimates however were averages across 10-yard segments
rather than each possible yard outcome. Carroll, Palmer, and Thorn
(1988) extend Carter and Machol's approach and provide expected
point outcomes for each yard-line on a football field.
The results of Carroll, Palmer, and Thorn are then exactly the type
of data we need to estimate the value of a kickoff. The data allows one
to estimate the contribution of a football play directly in terms of how
many points it contributes (EP). For example, Berri and Burke (2012)
describe it this way:
EP is a tantalizing concept for valuing the performance of players
because it can measure the contribution of each play directly in
terms of how many points it contributes. For example, consider a
situation where a QB snaps the ball on a first down and 10 from his
own 30-yard line, worth perhaps 1.1 EP. If he completes a 15-yard
pass, his team now has a first and 10 from its own 45-yard line,
worth perhaps 1.9 EP. In this case, the QB's play has added 0.8
EP to his team's expected net point differential. If instead he
threw an interception, this would give his opponents a first down
at midfield. And this is worth perhaps 2.0 EP for the opponent
and -2.0 EP for the QB's team. The net value of the play would
be negative: -2.0-1.1=-3.1. In this case, the interception
was equivalent to a loss of 3.1 in net EP differential. (p. 146)
There are, however, several shortcomings of these studies. The
first is that they both use data throughout the game. This is
problematic as the incentives are different late in the game--teams
ahead become more conservative, while teams behind are more
aggressive--where urgency becomes more of the issue than scoring
optimization. The second is that these studies assumed that the expected
point outcome is a linear function.
Romer (2006) allows the expected points given field position to be
nonlinear. Specifically, Romer is examining whether NFL teams were
optimizing with respect to fourth-down decisions. His results indicated
that teams were often kicking the ball (i.e., punting or kicking field
goals) when the optimal choice was to go for the first down.
Similar to the studies mentioned above, Romer's approach was
to estimate, through the use of actual game day data, the value of
taking offensive possession at a particular point on a football field
(i.e., the point value of having a first and 10 from any spot on the
field). Specifically, Romer was estimating not only the points that the
offensive team might score on the ensuing drive, but also the points the
opponent's might score with the field position you're likely
to give them if you don't score, and the points you're likely
to score with the field position they give you after they do or
don't score, and so on.
Figure 1 summarizes Romer's estimates. The figure simply
captures the net outcome (on the game's final score) of taking
possession of the football at a particular point on the field. For
example, taking possession at your own 5-yard line would change a
team's net point total by roughly -1.0.
The shape of curve is generally consistent with what one might
expect. For example, having a first and 10 on your own 20-yard line is
worse, in terms of the final score, than having a first and 10 on your
opponent's 20-yard line--worse by about 3.3 points. Also, the value
of having a first and 10 on your own 15-yard line is zero, suggesting
that a team would be indifferent between having possession on their own
15-yard line or the opposition gaining possession at their 15-yard line.
In other words, both states have the same impact on the chances of
winning.
[FIGURE 1 OMITTED]
Measuring the value of kickers in the NFL
Given the data highlighted in Figure 1, we can estimate the value
of a kickoff. First, we recognize that over the period from 1994 to 2009
an offensive football team starts on approximately their own 30-yard
line following a kick off. There are, however, three outcomes that can
lead the opponent to start someplace else: a touchback, a kick out of
bounds, or a return of a kickoff to another point.
The kicker could kick the ball that enters the end zone and is not
returned; this is referred to as a touchback. The outcome of this play
gives the opponent the ball on the 20-yard line. This outcome moves the
opponent back 10 yards. So a kicker who can kick the ball in such a way
as to get a touchback rather than an average starting point would force
the opposing team to go 10 yards further to score. In terms of
Romer's estimates, moving the football from the 30 back to the 20
is worth 0.556 points to a kicker's teams.
If the kicker kicks the ball out of bounds, this is against the
rules and imparts a penalty to the kicking team; the opponent is given
the ball on the 40-yard line. Again using Romer's estimates of the
cost of moving the opponent from the average outcome, (i.e., the 30-yard
line) to the 40-yard line costs the kicker's team 0.604 points.
Finally, if neither a touchback nor a kick out of bounds occurs,
then the kickoff will be returned. The average return over the period
from 1994 to 2009 was 22 yards. In which case if a kicker manages to
kick the ball past the opponent's eight-yard line (or if he manages
to kick the ball 62 yards), then he has--relative to an average kicker
--given his team a benefit. If the kick, though, is fielded past the
8-yard line then the kick imposes a cost.
Given these three outcomes--and the corresponding values--we can
now evaluate each kicker. To illustrate, let's consider the
performance of Thomas Morstead of the New Orleans Saints in 2009.
Morstead kicked off 101 times that season. Of these kicks, 27 resulted
in a touchback. Given that an average kicker would have only had 11.2
touchbacks, Morstead exceeded the performance of an average kicker by
15.8 touchbacks. As each touchback is worth 0.556 points,
Morstead's additional touchbacks generated 8.8 points for the
Saints. Additionally, Morstead kicked 2 kickoffs out of bounds, which
was 0.7 about the average kicker who would have had only sent 1.3. Given
that each out-of-bounds kick costs the team 0.604 points,
Morstead's relatively bad performance on out-of-bounds kicks cost
the Saints -0.2 points. Finally 72 of his were returned. Morstead's
returned kicks traveled 66.9 yards, or about five yards further than the
kicks of an average kicker. Romer's estimates suggest that saving
five yards on a kickoff is worth 0.289 points. So Morstead's
returned kicks generated 21.4 points for the Saints. If we put all this
together, we see that Morstead's kickoffs were worth 30.1 points
beyond what an average kicker would produce.
Table 1, which reports the best and worst kickers from 1994 to
2009, places this result in some perspective. As one can see,
Morstead's performance in 2009 ranks 4th among the 568 kickers who
have attempted at least 16 kickoffs in a regular season since 1994.
Next we can use Romer's estimates to value the kicker's
other job of kicking field goals and extra points. Whereas the latter is
almost always kicked from the same spot on the field, a field goal can
be attempted anywhere. Data is tracked for how each kicker does from the
19 yards and closer, 20-29 yards, 30-39 yards, 40-49 yards, and beyond
50 yards. With such data in hand, we can determine how the average
kicker did from each distance. And with averages in hand, we can--as we
did in our analysis of kickoffs--ascertain for each kicker how many
points a team could have expected if an average kicker would have
attempted a kicker's field goal attempts from each distance. A
similar calculation was completed for extra points.
To illustrate, consider the performance of Neil Rackers in 2005.
That season Rackers made 40 of the 42 field goals he attempted. With
each field goal worth three points, Rackers generated 120 points from
his field goals. An average kicker, kicking from the same distances,
would have only generated 96.8 points. So Rackers field goal kicking
generated 23.2 points more than what his team would have seen from an
average kicker. Rackers was also perfect on extra points, a performance
that was 0.3 points better than an average kicker. In sum, Rackers
scored 23.5 more points beyond what we would see from just an average
kicker.
This calculation, though, does not fully capture the impact of a
kicker's scoring. When a kicker misses a field goal, the opposing
team is able to take possession of the ball at the spot where the field
goal was missed. (11) Like the value of kickoffs, holding the ball at
certain points on the field is worth points to the opponent. And again,
Romer's work tells us how many points each position on the field is
worth.
To ascertain this value we need to note where the opponent gains
possession of the ball after each miss. Again, our data does not tell us
the exact distance each kick was attempted. But we do know the ranges
listed above. And from these ranges, we can estimate the location of the
ball and the value of this location to the opponent. For example,
Rackers in 2005 attempted 14 field goals from between 40 and 49 yards.
Given that the goal posts are 10 yards beyond the field of play and the
field goal kicker tends to kick the ball from seven yards behind the
line of scrimmage, a missed field goal from this distance would give the
opponent the ball somewhere between the 37 and 46 yard-line. For the
sake of simplicity, we took the midpoint of this range and assumed a
missed field goal from 40 to 49 yards would give the opponent the ball
at about the 42 yard-line (or 41.5 rounded up). From Romer's work
we know that the opponent can expect to score 1.52 points if they have
the ball at that point. Since Rackers, though, only missed one of these
kicks, he essentially saved his team 5.1 points (or 3.3 x 1.5) on his
attempts from 40 to 49 yards. Similar calculations from each distance
indicate that Rackers saved his team 11.4 points by kicking field goals
at an above average rate in 2005.
In Table 2 we put both calculations with respect to field goals
together. Specifically, we see that from 1994 to 2009, what Rackers did
in 2005 with respect to kicking field goals and extra points was the
best performance by any kicker. On the second half of Table 2, though,
we see that Rackers, in 2001, offered one of the worst performances by a
kicker with respect to kicking field goals and extra points.
Such inconsistency highlights the point by Schatz. Kickers are
simply not very consistent with respect to scoring. To further
illustrate this observation, we examined the correlation between a
kickers performance on field goals in successive seasons. From 1994 to
2009, 375 kickers attempted at least 16 field goals in consecutive
years. With respect to field goal percentage (field goals made divided
by field goals attempted), we only see a correlation of 0.07. And with
respect to scoring points above or below average (including the impact
of field position) we see a correlation of 0.04. In sum, kickers are
quite inconsistent with respect to scoring.
In contrast, we see much more consistency on kickoffs. From 1994 to
2009 we see 404 kickers who kicked off at least 16 times in successive
seasons. If we look at points per kickoff in consecutive years we see a
correlation coefficient of 0.54. As Schatz argued, kickers are much more
consistent with respect to kickoffs.
And when we look at the best and worst performances in Tables 1 and
2, we also see that kickers are capable of producing more points via
kickoffs. All of this suggests that NFL teams should be primarily paying
kickers for kickoffs.
Estimated equation and remaining data
To ascertain how kickoff and scoring impact salaries we employ the
model detailed by equation (1).
lnSAL = [b.sub.0] + [b.sub.1] x KICKOFFVALUE + [b.sub.2] x
SCORINGVALUE + [b.sub.3] x EXP + [b.sub.4] x EXPSQ + [b.sub.5] x LNSMSA
+ [b.sub.6] x CHANGETEAM + [b.sub.7] x VETERAN + [b.sub.8] x DRAFTED +
[b.sub.9] x PROBOWL + [b.sub.10] x OSKR + [e.sub.t] (1)
The dependent variable is a kicker's salary. (12) This value
is weighted by the size of the NFL's salary cap. (13) Salaries were
also logged. It is important to remember that NFL contracts are not
guaranteed. So a kicker who does not perform to expectations can be
easily removed from the team.
Table 3 reports that the average kicker in our sample was paid
about $750,000. Table 3 also reports descriptive statistics for the
factors that might explain a kicker's salary.
The two factors we primarily focus upon in our study of salaries is
Total Kickoff Value and Total Scoring Value. (14) Each of these factors
was calculated according to the above descriptions. It is important to
note that we are using lagged performance. In other words, salary in
2009 is believed to be a function of how the kicker performed in 2008.
In addition to performance, we considered seven control variables.
The first is experience, which we expect to have a positive impact on
compensation early in a player's career. Eventually, though, we
expect further increases in experience to diminish a player's
wages. Additionally, the team's market size is proxied using log
population of the local SMSA (LNSMSA).
Beyond experience and population, we consider four dummy variables.
The first of these is whether or not a kicker is a veteran player.
Veteran players in the NFL are eligible for free agency. (15)
Consequently, these players should see higher salaries.
The next dummy variable is equal to one if a kicker is drafted.
Unlike most positions in the NFL, teams do not often spend draft picks
on kickers. In fact, only 42% of kickers in our sample were actually
drafted by an NFL team. Past research, though, has shown that a
player's draft status can have lingering effects on player
evaluation and compensation. (16) Consequently, it is possible that the
kickers who were drafted are considered better players by
decision-makers independent of actual performance.
Berri and Simmons (2009), in a study of NFL quarterbacks, argued
that a football player's compensation might be affected by changing
teams. Specifically, Berri and Simmons found that quarterbacks who
changed teams tended to see their salaries decline. Because the NFL has
a binding salary cap and extensive revenue sharing, it is difficult for
one team to out-bid another team for a player's services. So
players do not often depart teams because they are getting better
offers, but because their current team had decided to let the player
depart (primarily because the original team has lowered their estimation
of a player's value). To control for this effect, a dummy variable
was included that is equal to one if the kicker switched teams.
The final dummy variable also follows from the work of Berri and
Simmons (2009). These authors argued that it was possible that a Pro
Bowl appearance in a player's career could have a lingering impact
in a player's salary. Consequently, a dummy variable is included
that is equal to one if a kicker appeared in the Pro Bowl.
The last factor included addresses the issue of on-side kicks. Once
a kick has traveled 10 yards, the kicking team can take possession of
the ball. To take advantage of this rule, teams will intentionally kick
the ball a bit beyond 10 yards in the hope of recovering the ball. The
advantage of this strategy is the kicking team can retain possession.
The obvious downside is the receiving team can recover the on-side kick
and have very good field position.
Only about 25% of on-side kicks are recovered by the kicking team.
And on average, kickers attempt fewer than 1.5 on-side kicks per season.
So these events are relatively rare. Nevertheless, it is possible that a
kicker who is perceived as proficient with respect to on-side kicks
might receive higher salary offers. Consequently, the number of onside
kicks recovered was included as a factor in our salary equation. (17)
Which kicking gets a kicker paid?
Our first approach in estimating equation (1) is to employ Ordinary
Least Squares. Those results are reported in Table 4, which also
presents what we see when we employ a model with team-specific fixed
effects. (18)
The OLS results indicate that a kicker's salary is only
statistically related to kickoff value, scoring value, and experience.
None of our control variables were found to have any impact on player
salary. In addition, kickoff value is only significant at the 10% level.
And when we turn to the fixed effects model, it is not even
statistically significant at that meager level. (19)
Simple OLS, though, is probably not the ideal approach to
estimating this model. Specifically, we next follow the example of
Hamilton (1997), Leeds and Kowalewski (2001), and Berri and Simmons
(2009) and employ a quantile regression (Koenker, 2005). Such a method
is appropriate when your dependent variable--as is often the case with
respect to salary data in professional sports--fails to follow a normal
distribution. This approach has a number of advantages over simple OLS.
Specifically, it allows us to ascertain the impact of our independent
variables at different points in the distribution. Furthermore, this
approach is less sensitive to outliers and also the issue of
heteroskedasticity.
As one can see, across all the years considered in our study, a
kicker's salary is affected by experience and performance. The
other nonperformance factors, though, are generally insignificant at
every point in the distribution. With respect to performance, scoring
matters at each point of the distribution while the value of kickoffs
only matters for kickers at the ends of the distribution. So kicking off
doesn't appear to matter in the evaluation of every kicker.
When we turn to Table 6, which reports the economic value of
scoring and kickoffs, we can see, scoring generates a higher return.
(20) An additional scoring point consistently adds more to a
kicker's salary than an additional point from kickoffs. Such
results indicate that decision-makers in the NFL are focusing primarily
on a kicker's ability to kick field goals and extra points. The
ability to excel at kickoffs has some value, but this skill is not
considered as valuable at scoring.
Concluding Observations
In discussing a study it is important to note where the research
might progress in the future. With respect to this study, though, rules
might slow that progression. Prior to the 2011 season, the NFL changed
where a kicker kicked off. The movement from the 30-yard line to the
35-yard line changed how often kicks were returned. To illustrate, in
2010, 80.1% of all kickoffs were returned. After the rule change, only
53.5% were returned in 2011 (and 53.2% in 2012). (21) Consequently, it
is possible that this current study will be difficult to replicate in
the future. Because so many kickoffs now result in touchbacks, (22) a
study of kickers in the future will likely be quite different from the
study we present.
This study we present, though, does offer a result consistent with
past studies of compensation in sports. Specifically, past research in
baseball and basketball has found evidence that decision-makers tend to
undervalue factors whose impact on outcomes is not easily ascertained by
simply watching a contest.
The past study that seems most relevant to our current inquiry is
the work of Berri, Brook, and Schmidt (2007) that indicated that scoring
in the NBA dominates a basketball player's compensation. We find a
similar story with respect to kickers. A kicker's scoring appears
to dominate a kicker's compensation in the NFL. But it is kicking
off that appears to have the largest impact on wins in football.
So why is scoring so important? The value of scoring appears most
obvious to those watching football. In contrast, ascertaining the value
of kickoffs requires that someone employ somewhat sophisticated
statistical analysis to ascertain the impact an additional yard from
kickoffs has on team wins. As the aforementioned study of the NBA
indicated, decision-makers in sports tend to have trouble assessing the
impact of actions that require statistical analysis to measure.
A similar story can be told about kickers in the NFL. In the end,
it appears that what drives a kicker's salary is performance on the
field and experience. The other factors--veteran status, changing team,
draft status, pro bowl experience, and onside kicks--do not impact
salary. Decision-makers are only interested in whether or not a person
can do the job and how long that person has done the job. This suggests
the market works efficiently. The only problem is that decision-makers
do not value scoring and kicking off in a fashion consistent with how
these factors impact wins in the NFL.
References
Ahlburg, D. A., & Dworkin, J. B. (1991). Player compensation in
the National Football League. In The Business of Professional Sports. P.
D. Staudohar and J. A. Mangan (Eds.). Champaign, IL: University of
Illinois Press.
Berri, D. J., & Brook, S. (2010). On the evaluation of the
"most important" position in professional sports. Journal of
Sports Economics, 11, 157-171.
Berri, D. J., Brook, S. L., & Schmidt, M. B. (2007) Does one
need to score to score? International Journal of Sports Finance, 2,
190-205.
Berri, D. J., & Simmons, R. (2009). Race and the evaluation of
signal callers in the National Football League. Journal of Sports
Economics, 10, 23-43.
Bradbury, J. C. (2013). What is right with Scully estimates of
player's marginal revenue product. Journal of Sports Economics, 14,
87-96.
Bradbury, J. C. (2007). Does the baseball labor market properly
value pitchers? Journal of Sports Economics, 8, 616-632.
Carroll, B., Palmer. P., & Thorn, J. (1988). The hidden game of
football. New York, NY: Grand Central Publishing.
Carter, V., & Machol, R. E. (1971). Operations research on
football. Operations Research, 19, 541-544.
Carter, V., & Machol, R. E. (1978). Optimal strategies on
fourth down. Management Science, 24, 1758-1762.
Hakes, J. K., & Sauer, R. D.(2006). An economic evaluation of
the moneyball hypothesis. Journal of Economic Perspectives, 20, 173-186.
Hakes, J. K., & Sauer, R. D. (2007). The moneyball anomaly and
payroll efficiency: A further investigation. International Journal of
Sports Finance, 2, 177-189.
Hamilton, B. H. (1997). Racial discrimination and professional
basketball salaries in the 1990s. Applied Economics, 29, 287-296.
Koenker, R. (2005). Quantile regression. Cambridge, England:
Cambridge University Press.
Krautmann, A. C. (1999). What's wrong with Scully-estimates of
player's marginal revenue product. Economic inquiry, 37, 369-81.
Leeds, M. A., & Kowalewski, S. (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, 244-256.
Romer, D. H.(2006) Do firms maximize: Evidence from professional
football. Journal of Political Economy, 114, 340-65
Rosen, S.,& Sanderson, A.(2001). Labour markets in professional
sports. Economic Journal, Royal Economic Society, 111, 47-68.
Schatz, A. (2006, November 12). Keeping score: N.F.L. kickers are
judged on the wrong criteria. New York Times.
Scully, G. W. (1974). Pay and performance in Major League Baseball.
American Economic Review. 65, 915-930.
David J. Berri [1] and Martin B. Schmidt [2]
[1] Southern Utah University
[2] College of William and Mary
Endnotes
(1) In Smith's example, he argues that one worker producing a
pin in isolation would be lucky to produce 1 pin a day. In contrast,
separating out the tasks may result in the production of nearly 240
pins.
(2) See www.econedlink.org.
(3) This is not to say that specialists do not exist, they
certainly do; rather players who have multiple skills are generally held
in greater regard than those with fewer skills.
(4) Specialization is such an important part of the game that the
rules prohibit those who block from catching the ball.
(5) This is not quite accurate as a kicker may have some slight
responsibility tackling an opposing player on kickoffs and blocked field
goals.
(6) Schatz, Aaron. (2006). "Keeping Score: N.F.L. Kickers Are
Judged on the Wrong Criteria." (November 12). New York Times.
(7) One should note that Berri and Brook (2010), in a study of
hockey goalies, also reported evidence that decision-makers do not
evaluate performance correctly in hockey.
(8) For a review of many of these, see Kahn (2000).
(9) Data on kickoffs and scoring can be found at NFL.com.
(10) One could alternatively assume profit maximization as the
team's goal. The kicker's salary would still be determined by
these two factors in the driving force behind revenues is a team's
total number of wins.
(11) We would like to thank Evan Osborne for making this
observation.
(12) Salary data can be found at the website of USA Today. USA
Today does not report data for the salaries of individual players after
the 2009 season.
(13) The value of the NFL's salary cap can be found at
Football 101, a site maintained by Mark Lawrence. The specific website
we looked at was the following: http://football.calsci.com/
SalaryCap3.html.
(14) In order to be included in our sample a kicker had to kick
field goals (and extra points) and kickoff. So a kicker like Thomas
Morstead--who only kicked the ball off in 2009--was not included in our
sample. In all, 422 observations were employed from 1995 to 2009. Of
these, we had salary data for 316 kickers.
(15) As Berri and Simmons (2009) note, 'NFL players are
broadly eligible for free agency after four seasons of experience. After
three years, players have restricted free agent status in which teams
holding the player's contract are allowed to make offers that at
least match those available on the free agent market. Experimentation
revealed that the impact of veteran or free agent status does not depend
on whether we use three or four years as the qualifying period."
(16) Berri and Simmons (2009) found that draft position impacts a
quarterback's compensation well into a quarterback's career.
(17) We also included dummy variables for each year considered in
the study (i.e., 1996 to 2008).
(18) With team-specific fixed effects, our market size variable was
dropped.
(19) We also considered whether performance from two years ago
affected a player's current salary. The inclusion of performance
from two years ago was not significant, and the model with performance
from two years ago had a much lower r-squared. Although the model we
report has an unusually low r-squared for a salary model in sports,
adding performance from previous years does not increase explanatory
power.
(20) Since the model is semi-logged, the slope coefficient is found
by multiplying the dependent variable by the estimated coefficient. We
employ the average value of the dependent variable across the sample
considered in making this calculation.
(21) This data is reported at NFL.com.
(22) According to NFL.com, there were 416 touchbacks in the regular
season in 2010. In 2011, this number rose to 1,120. And in 2012, the
number was 1,156.
David J. Berri is a professor of economics in the Department of
Economics and Finance at Southern Utah University. His research
primarily examines the economics of sports, with a specific focus on
behavioral economics, worker productivity and compensation, and
competitive balance.
Martin B. Schmidt is a professor in the Department of Economics at
the College of William and Mary, where he specializes in macroeconomics
and the economics of sport.
Table 1: The Best and Worst on Kickoffs
Kickoff
Top 10 Yards Touch
kicking kickers Year Kickoffs ave. -backs
Mitch Berger 1998 112 70.2 3
Pat McAfee 2009 80 70.0 21
Morten Andersen 1998 90 68.4 2
Thomas Morstead 2009 101 67.7 27
Morten Andersen 1995 82 68.9 2
Rhys Lloyd 2008 88 67.8 0
Stephen Gostkowski 2009 91 67.8 21
John Hall 1998 84 67.8 1
Mitch Berger 1996 71 68.9 0
Matt Prater 2009 78 68.4 28
Bottom 10
kicking kickers
Chris Jacke 1994 83 54.5 0
Steve Christie 2000 72 53.6 4
Al Del Greco 1995 81 55.7 8
Kris Brown 1999 73 56.0 0
Mike Hollis 1995 66 54.8 1
Al Del Greco 1994 55 52.8 3
Neil Rackers 2002 64 55.0 0
Craig Hentrich 2003 91 57.7 4
Cary Blanchard 2000 54 53.8 2
John Kasay 1995 70 56.8 1
Kickoff
Value points
Top 10 Out-of- per above
kicking kickers Bounds kickoff ave.
Mitch Berger 0 0.433 48.5
Pat McAfee 1 0.421 33.6
Morten Andersen 1 0.340 30.6
Thomas Morstead 2 0.298 30.1
Morten Andersen 1 0.356 29.2
Rhys Lloyd 0 0.329 29.0
Stephen Gostkowski 0 0.300 27.3
John Hall 0 0.314 26.4
Mitch Berger 0 0.368 26.1
Matt Prater 0 0.334 26.0
Bottom 10 Kickoff
kicking kickers points
below
average
Chris Jacke 0 -0.549 -45.6
Steve Christie 2 -0.599 -43.1
Al Del Greco 4 -0.428 -34.7
Kris Brown 0 -0.457 -33.3
Mike Hollis 1 -0.492 -32.5
Al Del Greco 2 -0.578 -31.8
Neil Rackers 0 -0.491 -31.4
Craig Hentrich 2 -0.339 -30.9
Cary Blanchard 0 -0.568 -30.7
John Kasay 0 -0.398 -27.9
Table 2: The Best and Worst on Kicking Field Goals and Extra Points
Field
Top 10 Total goal Field Extra pt
scoring kickers Year pts attempts goal % attempts
Neil Rackers 2005 140 42 95.2% 20
Gary Anderson 1998 164 35 100.0% 59
Mike Vanderjagt 2003 157 37 100.0% 46
Jason Hanson 2008 88 22 95.5% 25
Sebastian Janikowski 2009 95 29 89.7% 17
Cary Blanchard 1996 135 40 90.0% 27
Pete Stoyanovich 1997 113 27 96.3% 35
Jeff Wilkins 2003 163 42 92.9% 46
Joe Nedney 2005 97 28 92.9% 19
Al Del Greco 1998 136 39 92.3% 28
Field
Bottom 10 Total goal Field Extra pt
scoring kickers Year pts attempts goal % attempts
Seth Marler 2003 90 33 60.6% 30
Joe Nedney 1996 89 29 62.1% 35
Kris Brown 2001 124 44 68.2% 34
Wade Richey 2001 89 32 65.6% 26
Neil Rackers 2000 57 21 57.1% 21
Steve McLaughlin 1995 41 16 50.0% 17
Neil Rackers 2001 74 28 60.7% 23
Kris Brown 2009 106 32 65.6% 43
Todd Peterson 2002 61 21 57.1% 25
Doug Pelfrey 1999 81 27 66.7% 27
Pts Value of Total
Top 10 Extra beyond missed value of
scoring kickers pt % ave. field goals field goals
Neil Rackers 100.0% 23.5 11.4 34.9
Gary Anderson 100.0% 20.5 9.0 29.6
Mike Vanderjagt 100.0% 18.8 7.9 26.7
Jason Hanson 96.2% 16.1 10.4 26.5
Sebastian Janikowski 100.0% 16.1 8.1 24.2
Cary Blanchard 100.0% 14.3 7.4 21.7
Pete Stoyanovich 97.2% 14.1 7.1 21.2
Jeff Wilkins 100.0% 13.9 6.9 20.8
Joe Nedney 100.0% 13.5 6.3 19.8
Al Del Greco 100.0% 13.4 4.8 18.2
Pts Value of Total
Bottom 10 Extra beyond missed value of
scoring kickers pt % ave. field goals field goals
Seth Marler 100.0% -19.3 -7.8 -27.2
Joe Nedney 97.2% -17.1 -8.0 -25.1
Kris Brown 91.9% -17.5 -6.3 -23.9
Wade Richey 100.0% -16.5 -6.2 -22.7
Neil Rackers 100.0% -15.4 -6.6 -22.0
Steve McLaughlin 100.0% -16.0 -5.8 -21.8
Neil Rackers 95.8% -16.5 -5.3 -21.8
Kris Brown 97.7% -16.1 -5.4 -21.6
Todd Peterson 96.2% -16.1 -5.4 -21.4
Doug Pelfrey 100.0% -16.2 -4.6 -20.8
Table 3: Descriptive Statistics for Factors Employed in Equation (1)
Standard
Variable Observations Average Deviation
Real salary 316 $750,241.90 $746,830.00
Kickoffs 422 65.92 21.34
Kickoff yards 422 4,126.37 1,389.75
Touch backs 422 7.31 6.12
Out-of-bounds kicks 422 1.08 1.04
Total kickoff value 417 -1.13 11.89
Field goals made 422 22.24 7.05
Field goals attempted 422 27.64 8.00
Field goal percentage 422 0.80 0.09
Extra points made 422 31.78 11.22
Extra points attempted 422 32.20 11.28
Total scoring value 422 0.38 6.62
Experience 422 5.82 4.66
Population 422 4.26 4.16
(in millions)
Dummy variable for 422 0.61 0.49
veteran player
Dummy variable for 422 0.44 0.50
drafted player
Dummy variable for 422 0.12 0.33
changing team
Dummy variable for 422 0.06 0.25
Pro Bowl player
Onside kicks attempted 422 1.42 1.37
Onside kicks recovered 422 0.35 0.63
Variable Minimum Maximum
Real salary $31540.4 $4806760.0
Kickoffs 10.0 112.0
Kickoff yards 551.0 7220.0
Touch backs 0.0 29.0
Out-of-bounds kicks 0.0 5.0
Total kickoff value -43.1 29.2
Field goals made 1.0 40.0
Field goals attempted 1.0 46.0
Field goal percentage 0.3 1.0
Extra points made 2.0 74.0
Extra points attempted 2.0 74.0
Total scoring value -19.3 23.5
Experience 0.0 20.0
Population 1.12 18.32
(in millions)
Dummy variable for 0.0 1.0
veteran player
Dummy variable for 0.0 1.0
drafted player
Dummy variable for 0.0 1.0
changing team
Dummy variable for 0.0 1.0
Pro Bowl player
Onside kicks attempted 0.0 9.0
Onside kicks recovered 0.0 4.0
* -- Salary data is taken from the website of the USA Today.
Performance data can be found at the website of NFL.com
Table 4: OLS Estimation of Equation (1)
OLS
Variable Standard
Coefficient Errors
Kickoff value 0.008 *** 0.005
Scoring value 0.042 * 0.009
Experience 0.217 * 0.058
Experience, squared -0.011 * 0.003
Population, logged -0.066 0.166
Veteran player -0.017 0.214
Drafted player -0.057 0.105
Changing team 0.016 0.144
Pro Bowl -0.113 0.190
On-side kicks recovered -0.052 0.073
Constant term 12.911 * 1.105
Observations 316
R-squared 0.244
Fixed Effects
Variable Standard
Coefficient Errors
Kickoff value 0.008 0.005
Scoring value 0.042 * 0.009
Experience 0.214 * 0.061
Experience, squared -0.010 * 0.003
Population, logged -- --
Veteran player -0.012 0.218
Drafted player -0.063 0.131
Changing team 0.036 0.167
Pro Bowl -0.056 0.216
On-side kicks recovered -0.046 0.085
Constant term 12.443 * 0.143
Observations 316
R-squared 0.229
* --significant at the 1% level
** --significant at the 5% level
*** --significant at the 10% level
Table 5: Quantile Regressions of the Log of Real Salary
Quantiles
Variables 0.1 0.25 0.5
Kickoff value 0.011 -0.005 0.008
0.007 0.005 0.007
Scoring value 0.032 ** 0.049 * 0.043 *
0.013 0.010 0.012
Experience 0.305 * 0.200 * 0.214 **
0.080 0.064 0.087
Experience, squared -0.018 * -0.010 * -0.010 **
0.004 0.003 0.004
Population, logged -0.134 0.143 0.073
0.238 0.201 0.251
Veteran player 0.074 0.141 -0.006
0.270 0.216 0.316
Drafted player -0.374 ** -0.141 -0.153
0.159 0.118 0.151
Changing team 0.108 -0.084 -0.091
0.228 0.179 0.234
Pro bowl 0.371 -0.164 -0.227
0.304 0.227 0.303
On-side kicks recovered 0.059 -0.105 -0.005
0.101 0.098 0.117
Constant term 12.035 * 10.964 * 12.077 *
1.568 1.336 1.663
Observations 316 316 316
Pseudo R-squared 0.202 0.161 0.141
Quantiles
Variables 0.75 0.9
Kickoff value 0.013 ** 0.022 *
0.005 0.007
Scoring value 0.037 * 0.040 **
0.010 0.017
Experience 0.267 * 0.191 **
0.076 0.093
Experience, squared -0.014 * -0.011 **
0.004 0.005
Population, logged 0.019 -0.343
0.210 0.285
Veteran player -0.108 -0.344
0.286 0.392
Drafted player -0.038 -0.052
0.126 0.181
Changing team -0.037 0.068
0.194 0.238
Pro bowl -0.014 -0.340
0.247 0.355
On-side kicks recovered -0.103 -0.091
0.095 0.114
Constant term 12.827 * 16.161 *
1.397 1.869
Observations 316 316
Pseudo R-squared 0.129 0.099
*--significant at the 1% level
**--significant at the 5% level
***--significant at the 10% level
Table 6: The Economic Value of Scoring and Kickoffs
Quantile Scoring Scoring Kickoff Kickoff
coefficient value coefficient value
10% 0.032 $2,892 not significant
25% 0.049 $9,422 not significant
50% 0.043 $23,430 not significant
75% 0.037 $46,574 0.013 $16,249
90% 0.040 $97,880 0.022 $52,842
