Vous êtes sur la page 1sur 12

The Impact of the Salary Cap on Competitive Balance

-In the Case of Korean Baseball League-

Keywords: Competitive Balance, Salary Cap, Analysis of Variance by Ranks, Spearman


Rank Correlation Coefficient

Chi-Ung Song
Science and Technology Policy Institute
26F, Specialty Construction Center 395-70
Shindaebang-Dong, Tongjak-Gu
Seoul 156-714, Korea
Tel: 822-3284-1773
cusong@stepi.re.kr

Won Soon Kwon


Hankook University of Foreign Studies
Department of Economics
270 Imun-Dong, Dongdaumun-Gu
Seoul 130-791, Korea
kwon@hufs.ac.kr
June 9, 2008

**The corresponding author is Chi-Ung Song

Abstract

1
We examine whether the salary cap has improved the competitive balance in the KBL
by using the methodology of ‘Spearman Rank Correlation Coefficient’ and ‘Analysis of
Variance by Ranks.’ We find that there are some degrees of association between winning
percentages and salary spending for each season. We also find that winning percentages
of all teams are not statistically significantly different with each other. In conclusion, we
can say that the salary cap might improve the competitive balance in this case.

I. Introduction

2
What is the competitive balance in professional sports leagues? The competitive
balance can be described as the balance of playing strength among different franchises
within a league. Any league that has a relatively fair distribution of playing strengths
among teams would have much better competitive balance than those that do not.
Then, why is this so important in terms of economics? The revenue of a
professional sports franchise is primarily determined by its game quality. If a team loses
games constantly, then it will also lose the attention from fans and public media as well
as revenues from the gate and broadcasting. If more and more teams lose games and
money, then total revenues of the league will decrease as well.
In other words, professional sports leagues will lose money if the distribution of
competitiveness among teams is out of balance, and vice versa. That’s why professional
sports leagues, such as NFL, NBA and NHL, have introduced a salary cap in order to
improve the competitive balance of its own. If there is no limit in salary spending, then
rich franchises located in big cities can collect playing talents as much as they can,
while small franchises will lose playing talents constantly. This will break the
competitive balance, degrade the game qualities, and reduce the total revenues.
Then, did the introduction of salary cap really improve the competitive
balance? Larsen and Fenn (2006) estimate the impact of salary cap on competitive
balance in NFL by using Gini-coefficient and Hefindahl-Hirschman index, and find that
salary cap tends to promote the competitive balance. Diet et al. (2007) introduce a
theoretical model of a team sports, and show that salary cap will increase competitive
balance. However, Barriger et al. (2004) argue that the salary cap did not improve
competitive balance in NFL, and Endo et al. (2003) suggest a similar conclusion with
Barriger et al. from their empirical study on NBA. Both use Gini-coefficient as an
analytical method.
In this paper, we estimate whether the salary cap has improved the competitive
balance in Korean Basketball League (KBL). The KBL is the only professional sports
league in Korea that has introduced a salary cap from the first, and there is virtually no
empirical study that evaluates the impact of salary cap on competitive balance in Korea.
Thus, this study would be one of the first efforts to analyze the impact of the salary cap.
In our study, we introduce the methodologies of ‘Spearman Rank Correlation’ and
‘Analysis of Variance by Ranks,’ while previous researches has mainly utilized Gini
coefficient to evaluate the impact of a salary cap. Therefore, an introduction of other
methodologies would be an additional contribution of our study.

3
II Data

The salary cap system has been adopted by the KBL from the very beginning of
its own league in 1997. Under the salary cap restriction, the KBL set up the maximum
amount of money that each team can spend for players’ salary and every team has to
keep the financial restrictions. No team can spend more than the ceiling. Now, we
introduce cap ratio that is described as the ratio between the amount of team’s salary
spending and the amount of maximum salary. Then, we are going to analyze the link
between the cap ratio and playing performances among teams.
Theoretically, a salary cap would narrow the spread among teams in total player
salaries, prevent an overzealous owner from monopolizing playing talent and improve
competitive balance.1 It means that the league wants to reduce the influences of
financial power on playing strengths, increase the uncertainty of game outcomes and
make games more exciting by the salary cap restrictions. That is, the salary cap could
enhance the competitive balance.
In general, we can measure the competitive balance of a league by the
distribution of wins among different teams. If the competitive strengths of teams are in
good balance, then the league would have an even distribution of wins. Thus, we are
going to look the distribution of teams’ winning percentages in every single season.
Our data consists of the winning percentage and the cap ratio of teams in the
regular season games from 1997-1998 to 2006-2007 seasons. Actually, the KBL has
started its league from 1997 season but there were only 8 teams at that time. It was
1997-1998 seasons when ten teams started to participate in regular season games. Thus,
we analyze the impact of a salary cap on the competitive balance during 1997-1998 and
2006-2007 season games that all of ten teams have been in the races.
We use the winning percentage of each team as a proxy of a winning record and
it is described as the number of winning games divided by the number of total games.
Meanwhile, we use the salary ratio of individual team as a proxy of the salary spending
and it is described as the amount of players’ salary divided by the amount of salary cap.
The salary cap is the maximum amount of salary that teams can spend in a single
season.
Table 1 shows the information about the average of winning percentages and
salary ratios of individual teams during 1997-1998 and 2006-2007 seasons. The more
detailed seasonal data on teams’ performances, such as numbers of wins, winning
percentage, salary spending, can be found in KBL website.

1
Barriger et al. (2004)

4
<Table 1> Teams’ winning percentages and salary ratios in average

Teams Total games Winning percentage Salary ratio


Mobis Phoebus 504 0.494 0.887
KT&G Kites 504 0.476 0.887
Dongbu Promy 504 0.544 0.909
Daegu Orions 504 0.488 0.898
KTF Magic Wings 504 0.466 0.796
Black Slammer 504 0.437 0.908
KCC Egis 504 0.563 0.967
Samsung Thunders 504 0.530 0.926
LG Sakers 504 0.540 0.941
SK Knights 504 0.462 0.880
 data resources: www.kbl.or.kr & www.history.kbl.or.kr
 Data from 1997-1998 to 2006-2007 seasons.
 Black Slammer – Incheon Black Slammer

<Table 2> the Salary Cap imposed by KBL


(In Million Won)
Season Salary cap
1997-1998 900
1998-1999 900
1999-2000 945
2000-2001 1,000
2001-2002 1,050
2002-2003 1,150
2003-2004 1,250
2004-2005 1,375
2005-2006 1,500
2006-2007 1,600
 data resources: www.kbl.or.kr & www.history.kbl.or.kr

Table 2 shows the salary cap, i.e., a maximum amount of money that can be
spent for players’ salary of teams.

III. Analytical Methodology

3-1. Spearman Rank Correlation

5
At first, we use the ‘Spearman Rank Correlation Coefficient’ to examine the
strength of a link between the winning percentage and the cap ratio. The formula for the
‘Spearman Rank Correlation Coefficient’ is as follows,
6 × ∑d 2
Rs = 1 − …………………………………………………………………... (1)
n( n 2 −1)
In order to calculate the coefficient (Rs), we rank both sets of data (winning
percentage and cap ratio) from the highest to the lowest for each season. In the group of
winning percentage, a team with the best winning percentage gets a rank value of 1 and
the worst team gets the largest number of rank value. The same procedure is applied to
the group of cap ratio. A team with the highest cap ratio gets a rank value of 1 and a
team with the lowest spending gets the largest number of rank value.
Then, we obtain differences of rank values between two ranks for same team.
Next, we square the difference and get d2 of equation (1). In addition, we have 10 teams
in the KBL, so we have 10 rakings as a whole. Thus, the number of rankings (n) is 10 in
equation (1).
Once we calculate the correlation coefficient (Rs) for a single season, we can
measure the degree of line association between two sets of ranks in that season. The
guideline for the measurement is listed in table 3.

<Table 3> the degree of line association


Scope Relationship
1.0 Perfect positive
0.8 ~ 1.0 Strong positive
0.5 ~ 0.8 Moderate positive
0 ~ 0.5 Weak positive
0 No association
0 ~ -0.5 Weak negative
-0.5 ~ -0.8 Moderate negative
-0.8 ~ -1 Strong negative
-1 Perfect negative

The null hypothesis to test would be “there is no association between two sets
of data.” If the Rs value equals to zero, then the null hypothesis is accepted. Otherwise,
it is rejected.
By calculating the correlation coefficient for every single season, we can figure
out how the coefficient as well as the association of two ranks would vary across the

6
time period.

3-2. Analysis of Variance by Ranks2

If the competitiveness of individual teams is in good balance, then there would


be an even distribution of wins across teams. Consequently, the winning percentage of
each team would not be statistically significantly different with one another.
In this study, we examine whether there are statistically significant differences
in the wining percentages among different teams in KBL. We use ‘Analysis of Variance
by Ranks’ (Stekler, 2001) in order to determine whether all winning percentages are
evenly distributed. The precise steps are as follows.
First, we rank each team by its winning percentage for each season. A team
with the best winning percentage gets a rank value of 1. Thus, the larger the value of
rank the less percentage are the teams’ winning. This process is repeated for each season
from 1997-1998 to 2006-2007. We have 10 sets of ranks as well as 10 sets of seasons.
If the winning percentages of all teams are equal, each team’s rank will have the
same expected value over all seasons. Using the ‘Analysis of Variance by Ranks’, it is
possible to test the null hypothesis that all teams in KBL have equal ranks by using a
chi-square test. A rejection of the null hypothesis means that the average rankings differ
statistically among teams, and that all winning percentages are not evenly distributed.
Some wins more, while others win fewer games.
The analytical model is listed as equation (2) and equation (3). We have 10
seasons from 1997-1998 to 2006-2007 seasons, so a number of time period (N) is 10.
The number of team (I) is 10 and the number of rankings (n) is also 10.
10
S i = ∑ Ris , i = 1,2,3,...., I ……………………………………………………………...
s =1

(2)
I

∑[ S i − 1 / 2( N )( n + 1)]
………………………………………………………...…
S= i =1

[ ( N )(n − 1)(n + 1) / 12]


(3)
The distribution of this statistic is approximately chi-square with n-1 degree of
freedom. We reject the null hypothesis if either the calculated chi-square value is greater
than the critical value for a given level of significance or the probability of obtaining a
calculated value for a given degree of freedom is less than the level of significance. The

2
Song (2005)

7
rejection of the null hypothesis indicates that forecasters have statistically significantly
different winning percentages.

IV. Empirical Results

In table 4, we have the results of ‘Spearman Rank Correlation Coefficient’ that


are calculated for each season.

<Table 4> Spearman Rank Correlation Coefficient


Season Coefficient (Rs) Association
1997-1998 0.6 Moderate positive
1998-1999 0.515 Moderate positive
1999-2000 0.661 Moderate positive
2000-2001 0.212 Weak positive
2001-2002 0.164 Weak positive
2002-2003 0.297 Weak positive
2003-2004 0,564 Moderate positive
2004-2005 -0.091 Weak negative
2005-2006 -0.382 Weak negative
2006-2007 -0.018 Weak negative
Average 0.2522 Weak positive
• Average = sum of Rs/number of season

As we can see from table 4, we can reject the null hypothesis of no association
between two sets of data. There are certain degrees of association between the winning
percentage and the cap ratio even though they are not strong. Thus, we can say that the
playing strength is somehow affected by the amount of salary spending.
Then, does the competitive balance exist during the course of 10 seasons in the
Korean Basket League (KBL)? If test result from the ‘Analysis of Variance by Ranks’
indicate that the average rankings differ statistically among teams, then all winning
percentages are not evenly distributed and the playing strength has been out of balance.
From the empirical test using the ‘Analysis of Variance by Ranks’, we get the
calculated chi-square value of 10.30. This value is less than the critical value (16.92) of
chi-square distribution at the 0.05 level with 9 degrees of freedom. In addition, the
probability of obtaining the calculated value of 10.30 with 9 degrees of freedom is far

8
greater than 0.05.
As a result, we do not reject the null hypothesis of equal winning percentages at
the 0.05 level of significance. It means that the playing strengths are balanced and all
winning percentages are evenly distributed. Thus, we can conclude that there exists the
competitive balance in the KBL.

V. Conclusion

From the ‘Spearman Rank Correlation Coefficient’, we find that there are some
degrees of association between winning percentages and salary spending in the KBL.
Still, the estimated association has not been so strong.
Meanwhile, we find that winning percentages of all teams are estimated as not
statistically significantly different with one another. Thus, we can say that the playing
strengths are distributed quite evenly, and there exists the competitive balance.
From these two empirical findings, we may conclude that the salary cap might
improve the competitive balance in the case of KBL.
In the future, we are going to examine the impact of salary cap on the
competitive balance by using the Gini coefficient as well.

Appendix
<Table A-1> Ranking by the winning percentages

9
Y1 Y2 Y3 Y4 Y5 Y6 Y7 Y8 Y9 Y10

Orions 5 10 8 10 1 1 3 6 6 4

KTF 7 9 9 8 7 4 8 4 4 3

SK 10 8 2 3 2 10 7 8 9 7

Samsung 9 6 3 1 8 5 5 5 2 5

KT&G 8 7 5 4 6 8 9 3 7 6

Mobis 3 2 6 9 10 6 10 7 1 1

Dongbu 4 4 4 7 9 3 1 1 3 8

Slammer 6 3 10 5 4 7 4 10 10 9
s

KCC 1 1 1 6 3 9 2 2 5 10

LG 2 5 7 2 5 2 6 9 8 2

 Y1 = 1997-1998 seasons and Y10 = 2006-2007 seasons

<Table A2-> Ranks by the salary ratios

Y1 Y2 Y3 Y4 Y5 Y6 Y7 Y8 Y9 Y10

10
Orions 4 10 9 8 4 1 1 7 5 8

KTF 7 9 10 10 10 10 8 8 7 1

SK 10 2 5 7 2 9 10 6 4 7

Samsung 8 5 4 5 1 5 9 1 6 5

KT&G 3 4 3 4 8 8 7 10 1 10

Mobis 1 3 1 1 3 7 5 9 10 9

Dongbu 9 6 7 6 9 6 4 5 8 4

Slammer 6 8 8 9 7 3 2 3 9 3
s

KCC 2 1 2 3 6 2 3 2 2 6

LG 5 7 6 2 5 4 6 4 3 2

Reference

Barriger, A. K., J. W. Sharpe, B. P. Sullivan, and P. M. Sommers. (2004), “Has a Salary


Cap in the NFL Improved Competitive Balance?” Discussion Paper No. 04-02,

11
Middlebury College.

Dietl, H., M. Lang, and A. Rathke. (2007), “The Effect of Salary Caps in Professional
Team Sports on Social Welfare”, Working Paper Series ISSN 1660-1157, Institute for
Strategy and Business Economics, University of Zurich.

Endo, M. G., K. M. Florio, J. B. Gerber, and P. M. Sommers. (2003), “Does a Salary


Cap Improve Competitive Balance?” Atlantic Economic Journal 31(4), 388.

Hibon, M and H. O. Stekler (2003), “The M-3 Competition: Statistical Tests of the
Results”, INSEAD Working Paper.

Larsen, A., and A. J. Fenn (2006), “The Impact of Free Agency and the Salary Cap on
Competitive Balance in the National Football League”, Journal of Sports Economics 7,
374-390.

Makridakis, S. and M. Hibon (2003), “The M-3 competition: Results, Conclusion and
Implications”, Special Issue, International Journal of Forecasting 16(4), 451-476.

Song, C. (2005), “An Evaluation of Judgmental and Statistical Model Forecasts” Ph. D.
Dissertation, The George Washington University.

Stekler, H. O. (1987), “Who Forecasts Better?”, Journal of Business and Economic


Statistics 5, 155-158.

Stekler, H. O. (2001), “The M-3 Competition: The Need for Formal Statistical Tests”,
International Journal of Forecasting 17(4), 576-577.

12

Vous aimerez peut-être aussi