The role of managers in team performance.
Berri, David J. ; Leeds, Michael A. ; Leeds, Eva Marikova 等
Introduction: The Role of Managers
The reputation of corporate managers goes through periodic upswings
and downturns. As noted by Ira Horowitz (1994b), Adam Smith argued
managers play an inconsequential role in the performance of a firm.
Specifically, Smith separated the role of the entrepreneur from that of
the manager. In Smith's view, entrepreneurs provide both the
fundamental ideas and capital the organization requires for success.
Beneath the entrepreneur is a group of subordinates that oversees daily
operations. From Smith's perspective, this group of subordinates
does not vary in any significant way from organization to organization.
In essence, the managers of daily operations are little more than
"principal clerks" (Smith, 1976, pp. 54-55). This view of
managers has persisted in the neoclassical model of the firm in which
"top managers are homogeneous ... inputs into the production
process" (Bertrand & Schoar, 2003, p. 1173).
With its emphasis on static equilibrium, neoclassical theory
assumes away any role for managers. In this setting, managers ensure
firms operate in a technically and economically efficient manner. That
is, they extract maximal output from a given set of inputs and minimize
the cost of a given level of output. For a given set of inputs, a given
technology, and given prices, all managers behave in exactly the same
manner.
In contrast to neoclassical economics, the popular press has often
regarded corporate managers, particularly CEOs, with an almost cult-like
devotion. A search of Amazon.com's website showed almost 4,000
entries for Jack Welch, of which about 25 were either books by him or
books whose title featured his name. These works almost uniformly
praised Welch for his leadership of GE. The contrast between economic
theory and popular wisdom reveals a flaw that economists have only
recently begun to address. By focusing on equilibrium, the neoclassical
model overlooks the key role of managers: to seek out and exploit
disequilibria.
The most successful managers take advantage of market
inefficiencies or find previously undiscovered niches. Such managers
thus take on some of the characteristics of entrepreneurs. Unlike
entrepreneurs, however, they work to redirect the inputs of existing
companies rather than create new products or firms. Jack Welch, for
example, did not create any new financial services. He did, however,
transform GE by shifting its focus from manufacturing to financial
services at a time when manufacturing was beginning to decline and the
financial services sector was expanding.
Economic studies of managers have begun to recognize this role of
managers and have sought to quantify their impact on the firms they
head. The studies generally find that managers have a strong impact on
firm policy and profitability. However, these findings are typically the
result of a broad series of interactions between the CEOs, their
"managerial teams," and firms as a whole. Therefore, the
studies can only indirectly infer the contribution of the manager.
Sociologists, in contrast, have long recognized the role played by
managers. For example, Grusky (1961) examined the nature and impact of
managerial succession in firms long before the issue interested
economists. In a follow-up study, Grusky (1963) recognized that sports
are a natural source of data on managerial succession. He noted Major
League Baseball teams provided "reliable and valid measures of
rates of administrative succession and organizational
effectiveness" (Grusky, 1963, p. 21). Because managers in
professional sports teams all pursue identical goals by performing
similar tasks, professional sports are a natural laboratory in which to
investigate the contributions of managers to the performance of their
organizations.
More recent studies in finance and economics have built upon
Grusky's work by isolating the impact of managerial change or of
specific managers. The abundant performance data available in sports
allow researchers to control for the quality of inputs overseen by
managers. Such studies generally adopt Adam Smith's view of
managers as people who rearrange inputs of a given quality. We take this
literature one step further by calculating the impact of managers on the
productivity of individual players. We use this information to determine
the impact of individual managers on team performance.
We compile and analyze a data set measuring the performance of
individual players in the National Basketball Association (NBA) from the
1977-1978 season through the 2007-2008 campaign. We use the mobility of
players and coaches over this period to isolate the impact of coaches on
the teams they direct. Unlike most work on managerial performance, our
study focuses on how managers affect the performance of individual
players. Our results show some managers having reputations for good
management skills may simply be the beneficiaries of the good teams they
coach. This finding suggests recent empirical studies of CEOs may also
be subject to the failure to isolate the behavior of managers.
Next, we show how managers in professional sports behave
entrepreneurially by exploiting inefficiencies and discovering niches.
In the section that follows, we present our measure of player
performance and describe our data. In the Model of Coaching
Effectiveness section, we develop a model of managerial performance,
showing how much a coach contributes to a player's performance.
Then, we present our estimates and use them to evaluate the impact of
coaches on team performance. We finish with the conclusion.
The Economist's View of Management and Coaching
Because of the long-standing view that entrepreneurs, and not
managers, matter, the economics and finance literatures have only
recently begun to quantify the performance of individual managers. Some
studies, such as Chevalier and Ellison (1999) and Bertrand and Schoar
(2003), essentially construct matched-panel data sets allowing them to
track managers as they move from firm to firm. Using such data sets
allows them to separate the performance of managers from the
organizations they head. Unfortunately, a manager's decision to
move is often endogenous. For example, a change in CEOs could be the
result of internal conflict within the organization that adversely
affected the company's performance under the former CEO and whose
resolution boosts the company under the new CEO. The performance of the
firm could thus reflect the underlying conflict leading to the change in
CEOs rather than the behavior of the CEOs themselves. Moreover, when
managers move from one firm to another, they often bring a coterie of
assistants with them. The fact the company effectively hires both the
CEO and his "team" leads to an identification question: does
the firm owe changes in performance to the manager or to the team s/he
heads?
Other studies, such as Bennedsen, Perez-Gonzalez, and Wolfenzon
(2006) and Johnson et al. (1985) base their analyses on truly exogenous
separation: the unexpected deaths of CEOs. However, these data sets
examined small samples or were geographically specific. Moreover,
Johnson et al. took a very different view of managers. In their model,
managers were valuable because they acquired firm-specific skills that
do not necessarily apply elsewhere.
Two books by Michael Lewis demonstrate how managerial initiatives
improve or fail to improve performance in professional sports. In
Moneyball (2003), Lewis shows how Billy Beane, the general manager of
the Oakland Athletics baseball team, exploited inefficiency in the
evaluation of potential major league players. In their analysis of
Moneyball, Hakes and Sauer (2006) demonstrate the Athletics'
success stemmed from the fact they more accurately assessed the value of
players' skills. Other teams had consistently overvalued slugging
percentage (the number of total bases a player advances per at-bat) and
undervalued on-base percentage (the likelihood that a player
successfully reaches base per at-bat). By more accurately assessing
these skills, Beane acquired undervalued players and discarded
overvalued ones. This allowed the Athletics to compete successfully with
teams having much higher payrolls. A less noted point in Moneyball is
the dim view Beane takes of the team's manager. Beane views his
manager similar to Adam Smith, as a principal clerk carrying out the
wishes of his entrepreneurial superiors.
In The Blind Side (2006), Lewis describes the niche that football
coach Bill Walsh uncovered with the Cincinnati Bengals and later
perfected with the San Francisco 49ers. Walsh developed a new product
that revolutionized his field. Specifically, Walsh created the
"West Coast Offense," where teams rely heavily on quarterbacks
who can respond to what they see on the field and complete short passes
to a variety of receivers. The West Coast Offense transformed the
Bengals and then the 49ers from mediocre teams to dominant offensive
machines, and it greatly enhanced the careers of key players on each
team.
While Michael Lewis's case studies are highly suggestive, they
neither prove nor disprove that coaches and managers in professional
sports systematically affect their teams' performances. A look at
two NBA coaches reveals the difficulty involved in evaluating coaching
performance. Phil Jackson became head coach of the Chicago Bulls of the
NBA in 1988. Over the next nine seasons, the Bulls won 74% of their
regular season contests and six NBA titles. Jackson retired after
winning his sixth title with the Bulls in 1998. His retirement, though,
lasted only one season and in 1999 he became the head coach of the Los
Angeles Lakers. Again, Jackson's team won three consecutive titles.
During his first 14 seasons of coaching, Jackson compiled a record
unmatched in the history of the NBA. He is the only coach with a career
winning percentage greater than .700 and he has won more championships
than any other coach except Red Auerbach.
Using either winning percentage or championships to measure
productivity, Jackson appears to be the best coach in NBA history.
However, Jackson had considerable talent at his disposal. In seven of
Jackson's first nine seasons he coached the incomparable Michael
Jordan. In the 147 games Jordan did not play with the Bulls in 1993-94
and 1994-95, Chicago won 60.5% of its games (http://www.nba.com).While
this record was better than most coaches, it was well below
Jackson's career record.
With the Lakers, Jackson was again blessed with extraordinarily
talented players, particularly center Shaquille O'Neal and guard
Kobe Bryant. When O'Neal was traded to the Miami Heat, the
Lakers' record declined significantly even after Jackson returned
from another year-long retirement. Again, it is hard to separate
Jackson's ability as a coach from the talents of his players.
Phil Jackson's career record stands in stark contrast to that
of Tim Floyd, Jackson's successor in Chicago. Floyd has enjoyed
considerable success as a coach at the collegiate level, winning close
to two-thirds of his games with three colleges. In the NBA, however,
Floyd has had little success. His record with the Bulls before being
dismissed part-way through the 2001-2002 season was a dismal 49-190
(http://www.nba.com).
While Tim Floyd had considerably less success with the Bulls than
Phil Jackson, he also had far fewer talented players on the roster. When
Jackson left, so did almost half the members of the 1997-98 championship
team, including such star players as Dennis Rodman, Scottie Pippen, and
Michael Jordan. At least a portion of Tim Floyd's lack of success
with the Bulls can be attributed to a much shallower talent pool. Some
support for the claim that Floyd was a victim of circumstance comes from
his record in 2003-04 with the New Orleans Hornets. The Hornets won half
their games and advanced to the second round of the playoffs, something
that none of Floyd's teams in Chicago came close to doing. The
contrasting stories of Phil Jackson and Tim Floyd exemplify the
fundamental problem facing those interested in studying the role of
managers in the success of an organization: how can one separate the
performance of management from the performance of the workers?
Previous Economic Studies of Sports Coaches and Managers
The sports economics literature on the contributions of managers
has built on Grusky's estimation in a number of ways.Most notably,
it features more sophisticated techniques. These include an early form
of frontier analysis (Porter & Scully, 1982), generalized least
squares (GLS) (Chapman & Southwick, 1991), hazard models (Ohkusa
& Ohtake, 1996; Scully, 1994), and the Pythagorean Theorem
(Horowitz, 1994a, 1997). The literature also spans a variety of sports,
including college basketball (Clement & McCormick, 1989; Fizel &
D'Itri, 1996) American football (Hadley et al., 2000), and soccer
(Dawson, Dobson, & Gerrard, 2000a, 2000b).
The above studies share two characteristics. First, they attempt to
control for the quality of the talent at the manager's disposal. In
baseball studies, for example, this often takes the form of using
batters' slugging average as an explanatory variable, as first
proposed by Scully (1974). Second, the studies treat talent as
exogenous, as they implicitly assume the role of the manager is to
manipulate inputs of a given quality. Thus, the general form of the
studies can be expressed as
[W.sub.it] = f([A.sub.it], [M.sub.ijt]) + [[epsilon].sub.it] (1)
where [W.sub.it] is the winning percentage of team i in year t,
[A.sub.it] is the inherent ability level of team i in year t, and
[M.sub.ijt] is an indicator variable denoting whether manager j led team
i in year t. Typically, [M.sub.ijt] takes the form of a dummy variable,
while [e.sub.it] is a random error term reflecting unobserved factors.
Kahn (1993) and Ohkusa and Ohtake (1996) are notable exception to
the above framework. Both studies test whether coaches make their
players better. Kahn models Major League Baseball players'
performance as a function of managerial quality. Managerial quality, in
turn, is determined by regressing the manager's salary on his
experience, lifetime winning percentage, and a dummy variable indicating
the league in which he managed. This approach, however, has several
problems. First, because the model relies on an abstract variable called
"managerial quality," Kahn cannot identify the contributions
of specific coaches. Second, if good managers keep their jobs longer,
modeling quality as a function of experience is subject to simultaneity
bias.
Ohkusa and Ohtake (1996) test whether Jovanovic's (1979)
matching hypothesis holds in Japanese baseball. They regress performance
measures on a player's experience and a sequence of managerial
dummy variables. The coefficients reveal the impact of matching player i
with manager j in year t. They find the managerial dummies do not vary
by player, which lead them to reject the hypothesis that individual
players benefit from playing for specific managers.
Measuring Player Performance
We build upon the work by Kahn (1993) and Okhkusa and Ohtake (1996)
by carefully modeling the impact specific coaches had on the
productivity of individual players and on team performance. To start, we
need a measure of player performance.
Studies of baseball have benefitted from a plethora of summary
metrics--such as slugging percentage, OPS (the sum of slugging
percentage and on-base percentage), and linear weights--designed to
measure a baseball player's performance on the field. Researchers
looking at the sport of basketball, though, have far fewer options.
The traditional measure--labeled NBA Efficiency by the
NBA--involves adding together a player's positive statistics
(points, rebounds, steals, assists, and blocked shots) and subtracting
the numbers that detract from wins (turnovers and missed shots). As
Berri (1999, 2008) noted, such an approach fails to account for the
differing impact these statistics have on wins.
The limitations of NBA Efficiency lead us to employ the measure
detailed in Berri and Krautmann (2006), Berri, Schmidt, and Brook
(2006), and Berri (2008). As these works support, wins in the NBA are a
function of a team's offensive and defensive efficiency; where
efficiency is defined by how many points a team scores and surrenders
per possession. Estimates of the relationship between wins and the
efficiency metrics reveal that points, rebounds, steals, turnovers, and
field goal attempts have virtually the same impact, in absolute value,
on team wins. Free throw attempts and personal fouls have a smaller
effect. Additional regression analysis reveals that both blocked shots
and assists also have a smaller absolute impact. Given these
values--detailed in Table 1--a player's marginal product (PROD) can
be captured simply and accurately, as illustrated by equation (2).
PROD = 3FGM*0.064 + 2FGM*0.032 + FTM*0.018 + MSFG*-0.033 +
MSFT*-0.015 + REBO*0.033 + REBD*0.033 + TO*-0.033 + STL*0.033 +
FTM(opp.)*-0.018 + BLK*0.017 + AST*0.022 (2)
As detailed in Berri (2008), PROD is then adjusted for the
statistics tracked for the team. Then, because players play differing
minutes, we calculate each player's performance per 48 minutes
(ADJP48). (1)
All of the above variables are readily available for players in the
NBA. Using the Sporting News NBA Guide and the Sporting News NBA
Register (various years), as well as
http://www.Basketball-Reference.com, we collected data from the 1977-78
through 2007-08 seasons.
The data set does not include all players in the NBA during this
time period. ADJP48 can be misleadingly high or low for a player
appearing in only a handful of games or playing only a minute or two per
game. To ensure reliable measures of efficiency for each year, we
included only players playing at least 20 games and averaging at least
12 minutes per game. These restrictions yielded 7,887 player
observations. "Player observation" refers to the fact that we
might observe ADJP48 for a given player in multiple seasons.
If every player played for the same coach throughout his career, it
would be impossible to separate player performance from coaching
performance. Fortunately, players frequently change teams through trades
or free agency, and coaches are regularly hired and fired. Of the 7,887
player observations in our sample, 3,595--or 45.6%--were with a new
coach.
While frequent coaching changes were vital for our data set, they
also created a problem. Just as a player with few or brief appearances
might have a misleading ADJP48 value, a coach working with very few
players might have a misleading impact on those players. To minimize
this problem, we include only teams led by coaches who:
* had at least 15 players meeting our minutes and games played
restrictions coming to the coach.
* had at least 15 players meeting our minutes and games played
restrictions leaving the coach.
Given these two restrictions we were left with a sample of 62 head
coaches.
Finding the Best Coach: Moving from the Traditional to the Simple
With our adjusted data, we commenced our search for the best
coaches. Table 2 reports the lifetime coaching records (as of the end of
the 2007-08 season) of the top 20 coaches--in terms of career winning
percentage--in our sample. At the top of our list is Phil Jackson. As
noted, Jackson's teams won 70% of their regular season games. Only
six other coaches in our sample had career winning percentages above
60%. At the bottom of Table 2 is coach Lenny Wilkens, who holds the
career record for regular season wins. His career winning percentage of
53.6%, though, falls far behind the mark of Jackson.
Although not reported in the table, we should note the average
winning percentage in our sample is 50%. Also, if we extended Table 2 to
the end, our sample of 62 coaches is completed by Sidney Lowe and Tim
Floyd. Lowe's career mark in five seasons was 0.257 while
Floyd's was 0.280 over the same number of years.
A difficulty with focusing on career winning percentage is that
wins are ultimately determined by the players on the court.
Consequently, a coach with better players should be able to win more
games. To measure the value of coaches we wish to see how players
perform when the players join--and leave--a specific coach.
Tables 3 and 4 illustrate two simple approaches to seeing how a
coach impacts player performance. Table 3 reports the top 20
coaches--from our sample of 62--who saw the highest percentage of
players get better when they came to the coach. Topping this list is Dan
Issel, who coached for six seasons and won 45.6% of his games as a head
coach. However, of the 15 players who came to play for Issel, 12
improved.
Issel is not the only coach with a losing record to appear in Table
3. Collectively, 11 of the coaches listed lost more than they won.
Again, though, we tried separating the player from the coach. Table 3
suggests although Issel's teams produced a losing record, this was
because of the players not Issel.
Before we posit our conclusions, we also need to look beyond the
simple view of Table 3. One issue with looking at the percentage of
players who improved is that the size of the improvement is not
considered. Table 4 adds that layer of complexity to our study. In Table
4, we assess how many wins the new players coming to a coach produced in
their first year. The coaches are ranked in terms of the average
improvement, with the top 20 coaches reported. Once again Dan Issel tops
the list. On average, new players going to play for Issel produced 3.5
additional wins in their first season.
Table 4 only reports the top 20 coaches--in terms of additional
wins per new player--in our sample. Similarly, the bottom of Table 4
shows six coaches who saw an average of less than one win per new
player. Such a result could suggest most coaches have minimal impacts.
A Model of Coaching Effectiveness
However, before reaching the above conclusion we need to consider
other factors impacting player performance. Beyond coaching, we argued
player performance in a current season was impacted by the list of
factors reported--with the corresponding average value in our sample--in
Table 5.
The first factor listed in Table 5 is the productivity of the
player, or ADJP48. The lagged value of this variable captures the
player's level of human capital at the end of the previous season.
By itself, lagged performance explains 68% of a player's current
performance. As Berri et. al. (2006) demonstrated, NBA players--relative
to player productivity in baseball and football--are quite consistent
over time.
Although basketball players are relatively consistent, we are
interested in why performance does change. Topping the list of factors
causing performance to differ over time is age. We expect when a player
enters the league he will initially get better as he ages, but
eventually time will negatively impact player performance.
Age is not the only physical element altering performance as
injuries will also make a difference. We employ games played, across the
past two seasons, as a proxy for a player's health status. All else
equal, we expect fewer games played indicates more injuries and a lower
level of performance. Of course, more games could also reflect the
coach's opinion regarding a player's productivity. Because
beliefs are based on past performance, including the lagged value of
ADJP48 captures this effect.
Beyond age and injury, the final characteristic of the player we
consider is position. Berri et. al. (2006) and Berri (1999, 2008) also
reported the position a player plays impacts his statistical output.
Consequently, we measured ADJP48 relative to point guards by including
dummy variables for center, power forward, small forward, and shooting
guard.
Players are also part of a team and two characteristics of the team
were also expected to impact individual performance. The first of these
was roster stability, which we expect has a positive impact on a
player's performance. We measure stability as the change in the
number of minutes played by a player's teammates from the previous
season. A priori, greater roster stability makes players more
comfortable with each other, and theoretically this should enhance
performance.
The performance of these teammates also should matter. Previously,
Idson and Kahane (2000) and Berri and Krautmann (2006), among others,
showed a player's teammates affect his performance. In particular,
we expect that as player i's teammates produce more, then player
i's productivity declines. To account for such diminishing returns,
we include the number of wins created by a player's teammates.
Following Grusky's (1963) claim that managerial changes
negatively affect team performance, we anticipate a player's
performance will decline whenever he plays for a new coach. Because we
account for the impact of specific coaches below, this variable is a
dummy variable equaling one if the player has moved to a new coach.
While this paper is predicated on the hypothesis that specific coaches
positively affect a player's performance, we expect the disruption
caused by the coaching change itself to negatively impact player
performance. Similarly, we also expect changing teams can also have a
negative impact on performance. This variable is equal to one if the
team with which a player ends the current season is different from the
team he began the prior campaign.
The factors reported in Table 5 are noted as [X.sub.it] and
[Y.sub.it] in equation (3).
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)
Where
[X.sub.it] = A vector of individual-specific variables
[Y.sub.it] = A vector of team-specific variables and lingering
coaching effects
[DCOACH.sub.ijt] = 1 if player i played for coach j in year t (= 0
otherwise) where j spans the 62 coaches in our data set
[DNC.sub.it] = 1 if player i played for a different coach in year t
than in year t-1 (= 0 otherwise)
As noted, the individual variables in [X.sub.it] consist of the
lagged value of ADJP48, the player's years of experience,
experience squared, the number of games played, and a dummy variable
indicating the player's primary position. The individual variables
in [Y.sub.it] include roster stability, the productivity of teammates,
and the dummy variables for new coach and new team. In addition, we
include the lingering impact of coaching. As we describe below, we are
primarily interested in how a player's performance changes as he
joins and departs to and from a specific coach. Because it is possible
for a coach to impact performance beyond a player's first year on
the team, we also include dummy variables to capture the impact of
coaches in the second and third year.
The second and third year impacts, though, are not the primary
focus of our paper. The two summations in Equation (3) capture the
impact of moving to or away from one of the 62 coaches examined in our
study. The first sum shows how ADJP48 changes for player i when he moves
to one of the coaches in our data set. It interacts the indicator
variable for playing for a new coach in year t with the indicator for
whether the new coach was a specific coach in our study. Thus, if player
i played for a new coach in year t and that coach was one of the 62 we
examined, then [ADJP48.sub.it] changed by [[delta].sub.g] +
[[delta].sub.ijt] where [[delta].sub.g] is the generic effect of playing
for a new coach and is the impact one of the 62 coaches had above and
beyond a generic coach.
If moving to coach j improves player i's performance, then
moving away from coach j could worsen his performance. It is tempting to
hypothesize the impact should be equal and opposite in sign to moving to
a coach. This would be true if the human capital that player i gains
from coach j disappears if not constantly maintained or is specific to
coach j's "system." It would not be the case if coach j
provides player i with lasting skills. The second sum in Equation (3) is
identical to the first sum except [DCOACH.sub.ijt] indicated player i
was with coach j in the previous year. If player i played for a new
coach in year t and the coach he played for in year t-1 was one of the
62 coaches in our data set ADJP48 changed by [[gamma].sub.g] +
[[theta].sub.ijt]. The use of interaction variables in Equation (3) is
similar to the use of differences-in-differences. It is not identical
because the event (moving from one coach to another) is not fixed in
time for all player observations.
Before moving on to discussing our results we acknowledge our data
set is an "unbalanced" panel with a lagged dependent variable,
hence an OLS estimation is inappropriate. Consequently we employed the
Arellano-Bond technique. This method is specifically designed to handle
unbalanced panels and essentially we use it for panel data where there
are empty cells. In this case, we do not have a "balanced
panel" because we do not follow a fixed number of players through
the entire set of years. For example, some players disappear partway
through, while others appear partway through (and may not last until the
end). Consequently, standard panel techniques are not robust. This is
particularly true because people do not disappear from the sample
randomly but by some self-selection procedure (e.g., not good enough to
make the roster).
Estimation Results
The results from estimation of Equation (3) appear in Tables 6A and
6B. For clarity, we split the results into two parts. Table 6A contains
individual player and team variables. Table 6B contains the
statistically significant interaction effects for players joining one of
the 62 coaches and players leaving one of the 62 coaches.
Before discussing the impact of coaches, we first briefly note the
results reported in Table 6A. As expected, current performance is
positively linked to past productivity. This production, though, is
impacted by injury, age, and position played. Specifically, the more
games a player plays, the higher a player's ADJP48. A similar
finding can be told about age early in a player's career. Advances
in age, though, cause performance to decline. We estimate the turning
point occurs at 24.4
years of age.
Finally, consistent with Berri et. al. (2006), we find small
forwards and shooting guards tend to offer less production.
Of the team factors, only the productivity of teammates has the
expected sign and level of significance. Specifically, the more
productive a player's teammates the less production the player will
offer. Although the effect is statistically significant, the impact of
teammates is quite small. The average player--[player.sub.i]--posts an
ADJP48 of 0.302. If [player.sub.i] moves from a team with average
teammates to a team with players whose productivity is two standard
deviations above average, [player.sub.i] will see his ADJP48 value fall
by 0.018. This translates into a decline of only 0.7 wins across an
entire season. In sum, while diminishing returns exists in the NBA, the
actual effect is minimal. While the effect of teammates is small, it
trumps the impact of the remaining team factors. We do not find roster
stability, switching to a new team, or switching to a generic new coach
to have any statistical impact on player performance.
The impact of a generic new coach is quite similar to the effect we
find for most coaches. Table 6B reports the statistically significant
coaching coefficients. When you look at the impact of new, second-year,
and third-year coaches, and also leaving a coach, we find 22 coaches
have a statistically significant impact with respect to one of these
issues. However, since our sample consists of 62 coaches, our results
indicate that for 40 coaches we do not see any statistical impact.
Before we discuss the coaches having a statistically significant
impact, we briefly return to Tables 2-4. These three tables report three
different approaches to ranking coaches. In Table 2 we see the 20
coaches--out of our sample of 62--having the highest career winning
percentage. Of these 20 names, 14 were not found to significantly impact
a new player's performance and 11 names are not listed at all in
Table 6B. Such a result may not surprise since career coaching records
do not separate a coach from his player.
Tables 3 and 4 were an effort to isolate the coach. But the results
were quite similar to what we saw in looking back at Table 2. Table 3
reports the 20 coaches having the highest percentage of player
improvement while Table 4 looks at the 20 coaches who saw the greatest
improvement in their new players. However, in both cases, 70% of the
names listed were not found to have a statistically significant impact
on new player performance. Therefore, once we control for the other
factors impacting player productivity, most of the coaches who
traditionally looked to be effective were found to have little effect on
what a player does when he comes to the coach.
Consequently, it appears what Adam Smith thought about management
in 1776 applies to most NBA coaches today. That is, most coaches do not
statistically impact player performance and subsequently most NBA
coaches are essentially principal clerks.
Although Smith's view applies to most coaches, we did find
some exceptions. In reviewing these exceptions we note that interpreting
the coaching coefficients in Table 6B is complicated. In sports
featuring opposing sides, such as basketball, it is difficult to
separate a player's performance from that of the player opposite
him. A player might score 50 points in a game due to his own outstanding
performance or to a particularly poor job by the player defending him.
Thus, if all coaches do equally well, the overall quality of play could
rise with no change in the standard measures of player performance:
better offensive play makes no more and no less headway against better
defensive play. For a coach to show a significant positive (negative)
coefficient, he must do a particularly good (poor) job relative to other
coaches. Our measure thus differs from that of managers in other
industries, whose success need not come at the expense of other
managers.
Of the 62 coaches in our data set, 14 had a statistically
significant impact on ADJP48 when a player came to the coach. Of these,
Phil Jackson had the greatest impact, with a point estimate of 0.045.
Players who joined a Phil Jackson-coached team saw their ADJP48 increase
by 0.026 more than players who joined a generic coach. Close behind
Jackson were Gregg Popovich and Cotton Fitzsimmons, who increased ADJP48
by 0.042. The remaining 11 coaches listed had a smaller impact. In fact,
the range from the fourth coach listed, Jim O'Brien, and the 12th
coach (Mike Fratello) is similar to what we see between Fitzsimmons and
O'Brien. In other words, although the impact of these coaches is
different from most coaches in our sample (and a generic coach) the
statistically significant impacts are not much different from each
other.
We can see this when we consider the confidence interval of our
estimates.
Drawing a 95 percent confidence interval around the positive
coefficients reported in the first part of Table 6B (Moving to Coach
...) reveals these coaches are not significantly different from the
others. For example, Jackson's confidence interval ranges from
0.020 to 0.070. This range overlaps the range of the last coach--Larry
Brown--listed in Table 6B to have a positive impact on new players
(-0.001 to 0.034). Our inability to distinguish individual coaches'
impacts--even when these impacts differ from zero--is also consistent
with Adam Smith's claim that managers are only "principal
clerks."
As noted, it is possible a coach could impact a player beyond the
first year. Although hypothetical, we did not find much evidence for an
impact beyond the first year with a coach. Specifically, we only found a
positive impact in the second year for Popovich, Jackson, and Don Nelson
and only Jackson had a positive impact in year three.
Eventually, of course, a player leaves a coach. In the last section
of Table 6B we report what happens to players leaving coaches. As noted,
we might expect a player to get worse if a coach is eliciting production
via a specific system. For 10 coaches--out of our sample of 62--we find
evidence that players get worse when they depart the coach. Of the 10
names listed, only three are listed in the first part of Table 6B. In
other words, for only three coaches--Kevin Loughery, Don Nelson, and Jim
O'Brien--we found a player improves when he arrives and then
declines when he departs.
As was the case for our review of the impact of coaches on new
players, constructing a 95% interval around the coefficients describing
the impact of departing a coach shows most of the coefficients are
statistically indistinguishable. The difficulty in distinguishing the
coaches again reinforces the notion that managers do not have much of an
impact on their players or teams.
Because coaches are ultimately judged by how their teams perform,
our final table reports our effort to translate the impact coaches have
on player performance into wins and losses. To do this, we convert the
impact coaches have on ADJP48 into wins. This is simply done by dividing
the coaches' impact on ADJP48 by 48 and multiplying by minutes
played. Specifically, a team plays 48 minutes in a game and 82 games in
a season. Hence, ignoring overtime, a team will play 19,680 minutes in a
regular season. Of these, about 90% are played by players with NBA
experience. If Jackson increases all of the veteran player's ADJP48
by 0.045, then the team will win 16.7 additional wins.
Table 7 shows the results of these manipulations for the coaches
having a statistically significant impact in Tables 6B. It shows that
hiring one of the 14 coaches with a positive effect on ADJP48 adds
significantly to wins. Hiring Jackson, Popovich, or Fitzsimmons can add
more than 15 wins across an entire season. This is enough to transform a
team with a 41-41 record into a 56-26 championship contender. Phil
Jackson provides a natural experiment of sorts. In 2004-05, the Lakers
won 34 games without Jackson. When Jackson returned in 2005-06, the key
performers on the team--Kobe Bryant and Lamar Odom--both had higher
ADJP48s, and the team won 11 more games. While this is less than the 16
wins our model predicts, it is consistent with our prediction. The
difference could be due to roster changes not accounted for by our
model's assumption that only the coach changed.
Conclusion
Basic economics tells us that an appropriate reward system should
be based on an employee's marginal revenue product. In industry, it
should reflect a manager's impact on the company's profits; in
professional sports, it should reflect a manager's contribution to
the team's wins. Unfortunately, it is generally difficult to
separate the performance of the manager from the quality of workers or
athletes whom he supervises. For this reason, coaches in professional
sports are evaluated in terms of the wins and losses of the teams under
their direction. Such an evaluation, though, ignores the fact coaches
work with different endowments of playing talent. This paper measures
the impact coaches have on the performance of their players.
Our point estimates show that some NBA coaches add substantially to
the performance of their players and to the number of games their teams
win. Two of these coaches, Phil Jackson and Gregg Popovich, are
acknowledged as being among the most successful coaches in NBA history,
winning a combined 13 NBA championships. Other coaches we identified had
significantly less success. In fact, of the other coaches having a
positive impact on newly acquired players, only Larry Brown has won an
NBA title. Furthermore, Gene Shue, Isiah Thomas, Kevin Loughery, and
Chris Ford all posted losing records.
Our most surprising finding was that most of the coaches in our
data set did not have a statistically significant impact on player
performance relative to a generic coach. Even the most successful
coaches by our metric--Jackson, Popovich, and Fitzsimmons--were
statistically discernable only from the very worst-rated coaches. We
therefore find little evidence that most coaches in the NBA are more
than the "principal clerks" that Adam Smith claimed managers
were more than 200 years ago.
References
Bellotti, B. (1992). The points created basketball book 1991-92.
New Brunswick, N.J.: Night Work Publishing.
Bennedsen, M., Perez-Gonzalez, F., & Wolfenzon, D. (2006). Do
CEOs matter? Unpublished manuscript.
Berri,D. (2008). A simple measure of worker productivity in the
National Basketball Association. In B.Humphreys & D.Howard (Eds.)
The business of sport (pp. 1-40).Westport, CT.: Praeger.
Berri, D. (1999). Who is most valuable? Measuring the player's
production of wins in the National Basketball Association.Managerial and
Decision Economics, 20(8), 411-427.
Berri, D., Brook, S., Fenn, A., Frick, B., & Vicente-Mayoral,
R. (2005). The short supply of tall people: Explaining competitive
imbalance in the National Basketball Association. Journal of Economics
Issues, 39(4), 1029-1041.
Berri, D., & Krautmann, A. (2006). Shirking on the court:
Testing for the disincentive effects of guaranteed pay. Economic
Inquiry, 44(3), 536-546.
Berri, D., Schmidt, M., & Brook, S. (2006). The wages of wins:
Taking measure of the many myths in modern sports. Palo Alto, CA:
Stanford University Press.
Bertrand, M., & Schoar, A. (2003).Managing with style.
Quarterly Journal of Economics, 118(4), 1169-1208.
Blass, A. (1992). Does the baseball labor market contradict the
human capital model of investment? The Review of Economics and
Statistics, 74(2), 261-268.
Bradbury, J. C. (in press). Peak athletic performance and ageing:
Evidence from baseball. Journal of Sports Sciences.
Chapman, K., & Southwick, L. (1991). Testing the matching
hypothesis: The case of Major League Baseball. American Economic Review,
81(5), 1352-1360.
Chevalier, J., & Ellison, G. (1996). Are some mutual fund
managers better than others? Cross-sectional patterns in behavior and
performance. Journal of Finance, 54(3), 875-899.
Clement, R.,& McCormick, R. (1989).Coaching team
production.Economic Inquiry, 27(2), 287-304.
Dawson, P., Dobson S., & Gerrard B. (2000a). Stochastic
frontiers and the temporal structure of managerial efficiency in English
soccer. Journal of Sports Economics, 1(4), 341-362.
Dawson, P., Dobson S., & Gerrard B. (2000b). Estimating
coaching efficiency in professional team sports: Evidence from English
Association football. Scottish Journal of Political Economy, 47(4),
399-421.
Fizel, J., & D'Itri, M. (1997). Managerial efficiency,
managerial succession and organizational performance. Managerial and
Decision Economics, 18(4), 295-308.
Fort, R., Lee, Y., & Berri, D. (2008). Race and coaching
efficiency in the NBA. International Journal of Sport Finance, 3(2),
84-96
Grusky, O. (1961) Corporate size, bureaucratization, and managerial
succession. The American Journal of Sociology, 67(3), 261-69.
Grusky, O. (1963).Managerial succession and organizational
effectiveness. The American Journal of Sociology, 69(1), 21-31.
Hadley, L., Poitras, M., Ruggiero, J., & Knowles, S. (2000).
Performance evaluation of National Football League teams. Managerial and
Decision Economics, 21(2), 63-70.
Hakes, J., & Sauer, R. (2006). An economic evaluation of the
Moneyball hypothesis. Journal of Economic Perspectives, 20(3), 173-185.
Heeren, D. (1994). Basketball abstract 1994-95 edition.
Indianapolis, IN: Masters Press.
Horowitz, I. (1994a). Pythagoras, Tommy Lasorda, and me: On
evaluating baseball managers. Social Science Quarterly, 75(1), 413-419.
Horowitz, I. (1994b). On the manager as principal clerk. Managerial
and Decision Economics, 15(5), 187-194.
Horowitz, I. (1997). Pythagoras's petulant persecutors.
Managerial and Decision Economics, 18(4), 343-344.
Idson, T., & Kahane, L. (2000). Team effects on compensation:
An application to salary determination in the National Hockey League.
Economic Inquiry, 38(2), 345-357.
Johnson, B.,Magee, R., Nagarajan, N., & Newman, H. (1985) An
analysis of the stock price reaction to sudden executive death:
Implications for the management labor market. Journal of Accounting and
Economics, 7(1-3), 151-174.
Jovanovic, B. (1979). Job matching and the theory of turnover.
Journal of Political Economy, 87(5), 972-990.
Kahn, L. (1993). Managerial quality, team success, and individual
player performance in Major League Baseball. Industrial and Labor
Relations Review, 46(3), 531-547.
Lee, Y., & Berri, D. (2008). A re-examination of production
functions and efficiency estimates for the National Basketball
Association. Scottish Journal of Political Economy, 55(1), 51-66.
Lewis, M. (2003) Moneyball: The art of winning an unfair game. New
York: W.W. Norton.
Lewis, M. (2006). The blind side: Evolution of a game. New York:
W.W. Norton.
Ohkusa, Y., & Ohtake, F. (1996). The relationship between
supervisor and workers--The case of professional baseball in Japan.
Japan and the World Economy, 8, 475-488.
Porter, P., & Scully, G. (1982). Measuring managerial
efficiency: The case of baseball. Southern Economic Journal, 48(3),
642-650.
Scott, F., Long, J., & Sompii, K. (1985). Salary vs. marginal
revenue product under monopsony and competition: The case of
professional basketball. Atlantic Economic Journal, 13(3), 50-59.
Scully, G. (1994).Managerial efficiency and survivability in
professional team sports. Managerial and Decision Economics, 15(5),
403-411.
Scully, G. (1974). Pay and performance in Major League Baseball.
American Economic Review, 64(6), 917-930.
Smith, A. (1776). An inquiry into the nature and causes of the
wealth of nations (1963 edition). Chicago: R.D. Irwin.
Zimbalist, A. (1992). Baseball and billions. New York: Basic Books.
David J. Berri [1], Michael A. Leeds [2], Eva Marikova Leeds [3],
and Michael Mondello [4]
[1] Southern Utah University
[2] Temple University Japan
[3] Moravian College and Temple University Japan
[4] The Florida State University
David J. Berri, PhD, is an associate professor of economics in the
Department of Economics and Finance. His current research focuses on the
economics of sports, specifically the topics of consumer demand,
competitive balance, and worker productivity.
Michael A. Leeds, PhD, is a professor of economics in the Fox
School of Business and Management. His research focus is in the area of
sports economics.
Eva Marikova Leeds, PhD, is an associate professor of economics and
business. Her research interests include financial institutions,
development economics, and sport economics.
Michael Mondello, PhD, is an associate professor in the Department
of Sport and Recreation Management. His research interests focus on
financial and economic issues of sport organizations.
Endnote
(1) For details on ADP48 one is referred to Berri et. al. (2006)
and Berri (2008). We would note here that the team adjustments involve
incorporating the team variables listed in Table 1. Following Scott,
Long, and Sompii (1985), these variables are allocated across players by
minutes played. Such an adjustment accounts for team defense and team
pace. Additionally, as detailed in the aforementioned works, performance
is also adjusted for the blocked shots and assists of teammates.
Finally, we adjusted each player's ADJP48 by the average value in
each season. This was done by subtracting the average value from each
season from each player's value. Then the average value across all
31 seasons was added. This last step was done to adjust for the change
in pace we see across the seasons in our sample.
Table 1: The Impact of Various Statistics Tracked for Players and Teams
on Wins in the NBA
Player Variables Marginal Value
Three Point Field Goal Made (3FGM) 0.06438
Two Point Field Goal Made (2FGM) 0.03179
Free Throw Made (FTM) 0.01758
Missed Field Goal (MSFG -0.03337
Missed Free Throw (MSFT) -0.01500
Offensive Rebounds (RBO) 0.03337
Defensive Rebounds (RBD) 0.03325
Turnovers (TOV) -0.03337
Steal (STL) 0.03325
Opponent's Free Throws Made (DFTM) -0.01752
Blocked Shot (BLK) 0.01744
Assist (AST) 0.02228
Team Variables Marginal Value
Opponent's Three Point Field Goals Made (D3FGM) -0.06414
Opponent's Two Point Field Goals Made (D2FGM) -0.03168
Opponent's Turnovers (DTOV) 0.03325
Team Turnover (TMTOV) -0.03337
Team Rebounds (TMRB) 0.03325
Note: These estimates are based on the model detailed in Berri (2008).
The data employed to estimate the Berri (2008) model can be found at
Basketball-Reference.com and in various issues of The Sporting News NBA
Guide. The specific years used to estimate the Berri (2008) model began
with the 1987-88 NBA season and ended in 2007-08.
Table 2: The Top 20 Coaches from 1977-78 to 2007-08
Ranked in terms of Career Winning Percentage (after the 2007-08 season)
minimum 15 qualified players come to coach, 15 qualified players depart
coach
Rank Coach Years Games
1 Phil Jackson 17 1,394
2 Gregg Popovich 12 934
3 K.C. Jones 10 774
4 Pat Riley 24 1,904
5 Paul Westphal 7 426
6 Rick Adelman 17 1,315
7 Jerry Sloan 23 1,806
8 Flip Saunders 13 983
9 Chuck Daly 14 1,075
10 George Karl 20 1,493
11 Jeff Van Gundy 11 748
12 Don Nelson 29 2,234
13 Rick Carlisle 6 492
14 Rudy Tomjanovich 13 943
15 Larry Brown 23 1,810
16 Mike Fratello 17 1,215
17 Del Harris 14 1,013
18 Doug Moe 15 1,157
19 Doug Collins 8 619
20 Lenny Wilkens 32 2,487
Winning
Rank Wins Losses Percentage
1 976 418 0.700
2 632 302 0.677
3 522 252 0.674
4 1,210 694 0.636
5 267 159 0.627
6 807 508 0.614
7 1,089 717 0.603
8 587 396 0.597
9 638 437 0.593
10 879 614 0.589
11 430 318 0.575
12 1,280 954 0.573
13 281 211 0.571
14 527 416 0.559
15 1,010 800 0.558
16 667 548 0.549
17 556 457 0.549
18 628 529 0.543
19 332 287 0.536
20 1,332 1,155 0.536
Note: These records are reported by the Sporting News NBA Register and
Basketball-Reference.com [http://www.basketball-reference.com/
coaches/].
Table 3: The Top 20 Coaches
Ranked by percentage of players who improve upon coming to the coach
Rank Coach Players Improved
1 Dan Issel 15 12
2 Don Casey 22 15
3 Jim O'Brien 24 16
4 Phil Jackson 41 26
5 Mike Schuler 19 12
6 Mike Dunleavy 51 32
7 Rick Carlisle 24 15
8 John Lucas 31 19
9 Tom Nissalke 18 11
10 Byron Scott 30 18
11 Kevin Loughery 42 25
12 Doc Rivers 27 16
13 Bob Hill 29 17
14 Doug Moe 24 14
15 Isiah Thomas 24 14
16 Rick Pitino 19 11
17 Bill Fitch 39 22
18 Doug Collins 32 18
19 Wes Unseld 16 9
20 Cotton Fitzsimmons 38 21
Percentage Winning
Rank Improved Percentage
1 80.0% 0.464
2 68.2% 0.357
3 66.7% 0.517
4 63.4% 0.700
5 63.2% 0.530
6 62.7% 0.478
7 62.5% 0.571
8 61.3% 0.401
9 61.1% 0.388
10 60.0% 0.487
11 59.5% 0.417
12 59.3% 0.508
13 58.6% 0.514
14 58.3% 0.543
15 58.3% 0.456
16 57.9% 0.466
17 56.4% 0.460
18 56.3% 0.536
19 56.3% 0.369
20 55.3% 0.518
Table 4: The Top 20 Coaches Again
Ranked by how many additional wins a new player produces
Rank Coach Players Improved
1 Dan Issel 15 12
2 Wes Unseld 16 9
3 Don Casey 22 15
4 Mike Schuler 19 12
5 Phil Jackson 41 26
6 Jim O'Brien 24 16
7 Doug Moe 24 14
8 Isiah Thomas 24 14
9 Tom Nissalke 18 11
10 Cotton Fitzsimmons 38 21
11 Rick Carlisle 24 15
12 Eric Musselman 20 10
13 Doug Collins 32 18
14 John Lucas 31 19
15 Rick Pitino 19 11
16 Doc Rivers 27 16
17 Bill Fitch 39 22
18 Mike Dunleavy 51 32
19 Jim Lynam 37 19
20 Stan Albeck 38 21
Increase
Percentage Increase in Wins
Rank Improved in Wins per Player
1
2 80.0% 52.9 3.5
3 56.3% 49.8 3.1
4 68.2% 64.0 2.9
5 63.2% 44.2 2.3
6 63.4% 80.6 2.0
7 66.7% 44.8 1.9
8 58.3% 44.0 1.8
9 58.3% 33.9 1.4
10 61.1% 25.0 1.4
11 55.3% 48.9 1.3
12 62.5% 30.1 1.3
13 50.0% 22.8 1.1
14 56.3% 35.0 1.1
15 61.3% 30.6 1.0
16 57.9% 17.9 0.9
17 59.3% 24.1 0.9
18 56.4% 34.2 0.9
19 62.7% 37.3 0.7
20 51.4% 19.1 0.5
55.3% 18.6 0.5
Table 5: Means of Relevant Variables
Variable Mean
Productivity of Player (ADJP48) 0.301
Age 27.141
Games Played Past Two Seasons 141.8
Center 0.207
Power Forward 0.200
Small Forward 0.196
Shooting Guard 0.201
Productivity of Teammates (TMWP48) 0.097
Roster Stability 0.690
New Team 0.289
New Coach 0.456
Note: Player data can be found in the Sporting News NBA Guide (various
years), the Sporting News NBA Register (various years),and
Basketball-Reference.com
Table 6A: Estimated Coefficient for Non-Coaching Independent Variables
Independent Variable Coefficient Standard Error
AdjP48, lagged * 0.1588 0.0355
Age * 0.0465 0.0064
Age Squared * -0.0010 0.0001
Games Past Two Seasons * 0.0006 0.0001
Center 0.0070 0.0113
Power Forward -0.0004 0.0099
Small Forward *** -0.0143 0.0084
Shooting Guard * -0.0179 0.0069
Productivity of Teammates (TMWP48) * -0.2996 0.0449
Roster Stability 0.0080 0.0069
New Team -0.0025 0.0025
New Coach -0.0033 0.0035
Independent Variable z-statistic
AdjP48, lagged * 4.4700
Age * 7.2700
Age Squared * -8.3800
Games Past Two Seasons * 8.1800
Center 0.6200
Power Forward -0.0400
Small Forward *** -1.7000
Shooting Guard * -2.5800
Productivity of Teammates (TMWP48) * -6.6800
Roster Stability 1.1600
New Team -0.9900
New Coach -0.9300
* Significant at 1% level ** Significant at 5% level *** Significant at
10% level
Table 6B: The Coaches with a Statistically Significant Impact on
Player Performance
Moving To Coach ... Coefficient Standard Error
Phil Jackson * 0.045 0.013
Gregg Popovich * 0.042 0.016
Cotton Fitzsimmons * 0.042 0.013
Jim O'Brien ** 0.032 0.013
Gene Shue * 0.030 0.011
Don Nelson ** 0.030 0.012
Flip Saunders * 0.028 0.011
Isiah Thomas ** 0.028 0.014
Rick Pitino *** 0.027 0.016
Stan Albeck ** 0.026 0.011
Kevin Loughery ** 0.026 0.010
Mike Fratello ** 0.022 0.011
Chris Ford ** 0.020 0.011
Larry Brown ** 0.017 0.009
Matt Guokas * -0.046 0.014
Second Year with Coach ...
Gregg Popovich * 0.031 0.012
Phil Jackson ** 0.026 0.012
Don Nelson *** 0.028 0.014
Bob Hill * -0.046 0.014
Third Year with Coach ...
Phil Jackson * 0.055 0.011
Moving away from Coach ...
Doug Collins * -0.034 0.012
Bernie Bickerstaff * -0.033 0.012
Jim O'Brien ** -0.031 0.015
Paul Silas *** -0.028 0.014
Jack Ramsay ** -0.026 0.013
Doug Moe *** -0.025 0.013
Kevin Loughery ** -0.025 0.011
Rick Carlisle ** -0.023 0.011
Don Nelson ** -0.023 0.009
Paul Westhead *** -0.022 0.012
Chris Ford *** 0.025 0.015
Isiah Thomas ** 0.036 0.014
Moving To Coach ... z-statistic
Phil Jackson * 3.550
Gregg Popovich * 2.610
Cotton Fitzsimmons * 3.170
Jim O'Brien ** 2.510
Gene Shue * 2.650
Don Nelson ** 2.580
Flip Saunders * 2.700
Isiah Thomas ** 2.000
Rick Pitino *** 1.700
Stan Albeck ** 2.240
Kevin Loughery ** 2.520
Mike Fratello ** 1.970
Chris Ford ** 1.860
Larry Brown ** 1.880
Matt Guokas * -3.210
Second Year with Coach ...
Gregg Popovich * 2.650
Phil Jackson ** 2.120
Don Nelson *** 1.950
Bob Hill * -3.350
Third Year with Coach ...
Phil Jackson * 4.840
Moving away from Coach ...
Doug Collins * -2.830
Bernie Bickerstaff * -2.630
Jim O'Brien ** -2.070
Paul Silas *** -1.940
Jack Ramsay ** -2.060
Doug Moe *** -1.940
Kevin Loughery ** -2.220
Rick Carlisle ** -2.120
Don Nelson ** -2.480
Paul Westhead *** -1.800
Chris Ford *** 1.710
Isiah Thomas ** 2.570
* Significant at 1% level ** Significant at 5% level *** Significant
at 10% level
Table 7: Another View of the Top NBA Coaches
Ranked by impact of coach on player performance
Moving To Coach ... Coefficient Estimated Wins
Phil Jackson 0.045 16.7
Gregg Popovich 0.042 15.5
Cotton Fitzsimmons 0.042 15.5
Jim O'Brien 0.032 11.7
Gene Shue 0.030 11.2
Don Nelson 0.030 10.9
Flip Saunders 0.028 10.5
Isiah Thomas 0.028 10.4
Rick Pitino 0.027 9.8
Stan Albeck 0.026 9.5
Kevin Loughery 0.026 9.4
Mike Fratello 0.022 8.0
Chris Ford 0.020 7.6
Larry Brown 0.017 6.1
Matt Guokas -0.046 -16.9
Second Year with Coach ...
Gregg Popovich 0.031 11.3
Phil Jackson 0.026 9.7
Don Nelson 0.028 10.4
Bob Hill -0.046 -17.0
Third Year with Coach ...
Phil Jackson 0.055 20.4
Moving away from Coach ...
Doug Collins -0.034 -12.6
Bernie Bickerstaff -0.033 -12.1
Jim O'Brien -0.031 -11.6
Paul Silas -0.028 -10.2
Jack Ramsay -0.026 -9.6
Doug Moe -0.025 -9.3
Kevin Loughery -0.025 -9.3
Rick Carlisle -0.023 -8.7
Don Nelson -0.023 -8.7
Paul Westhead -0.022 -8.0
Chris Ford 0.025 9.1
Isiah Thomas 0.036 13.2