FACTORS THAT ARE SIGNIFICANT FOR NBA PLAYER SALARY
INTRODUCTION
The NBA presented a salary ceiling, on the number of cash teams are allowed to use towards their roster, during the 1946/47 season. However, this constraint did not last as it was done away with after a maiden season. In the year 1970, the collective bargaining agreement (CBA) was formulated to set the base earnings, cap players earnings, specify trade rules, and the NBA drafting procedure. The CBA is “a legal contract between the league and the players association that sets up the rules by which the league operates”. Despite this act ahead of its time, a salary ceiling wasn’t reestablished until the season of 1984/85, where each team had 3600000 dollars to use on their whole side. Before the introduction of an earnings ceiling, teams would use an unlimited amount to formulate their team sheet, which was unfair to teams that lacked a lot of money and disadvantaged those who were not prepared to spend a lot of cash on the gifted players. The league introduced salary ceiling to create equality, guarantee balance and weed out the unfair competition in the league. This research paper aims to investigate significant determinants that contribute to the amount of money a player earns as salary. I will determine if age is involved in the under or overpayment of a player. I am also going to identify the determiners behind a player’s pay and investigate if teams have the tendency to incur more expense on a player who scores more.
LITERATURE REVIEW
Literature Review Labor contracts in the NBA differ from the norm and deciding how much money to offer a player always involves a complex decision. Research by Brodman (2009) attempts to tackle this anomaly. He uses multiple regression analysis of the six variables created by Hollinger (2003) (Efficiency, Approximate Value, Versatility, Points per 3 Field Goal, Turnover Ratio and Rebound Rating) to measure a player’s per-minute production and ability (along with other variables such as race, position, age, the team winning percentage, and team payroll) to explain a player’s expected salary. Save for the beginners, Broman evaluates players from the 2006-07 season using their current salary and career statistics. He uses separate variables for career stats and contract year stats to determine whether performance or the most recent season play is more significant. Brodman finds that points per field goal attempt represents the most significant measure, but the more valuable the player is to his team in his contract year, the less he is paid. The second finding is contrary to the expected result. Perhaps more advanced statistics could help explain his finding, or maybe something else is in play. Staw and Hoang (1995) offer an interesting idea that a sunk-cost effect could be in play for many teams. The more resources are poured into a player, the more opportunities he will have to succeed and this, in turn, would likely lead to a higher expected salary compared to later draft picks or minimum-salary players. The authors examine the careers of players coming in through the ranks from 1980 through 1986 and who had played at least two years in the NBA and explained minutes played as a function of statistical production per minute. They also group the box score stats into three categories, scoring, toughness and quickness, and find that scoring is the greatest predictor of playing time, but that draft position is also significant. A study by Grotius, Hill, and Perri (2009) suggests that many teams try to find star talent and fail miserably. They suggest that there are many more four false positives (players who have the potential to be superstars but never become one) than there are actual stars, and this can lead teams to overpay players. The draft exemplifies this, where teams can signal that a player has star potential by picking him early on, but many of the best players in the league aren’t the first pick in their draft. The model finds that higher drafts pick tend to be better players as measured by the efficiency formula (an old model intended to put a single number on production). Still, the model has an R-squared value of between 16 and 17%, which implies that the general managers for teams are not very accurate when evaluating talent. In a separate and less formal setting, Bill Simmons, an ESPN writer for thirteen years and named one of the most influential people in online sports in 2007 by the Sports Business Journal, corroborates this by going through nineteen years worth of NBA drafts and ranking the players based on their actual careers. While his rankings and the exact order can be debated, he finds that the draft position is only a limited predictor of success, that in reality the best talents are fairly randomly drafted (his rankings from the 2011 draft are as follows: 15, 1, 11, 38, 22, 16, 5, 9, 60, 13, 30, 24). While teams have a suboptimal record at evaluating talent, teammates and other external circumstances play a role in a player’s success as well. A paper by Idson and Kahane (2004) shows that a player’s teammates can have a large effect on how that player produces. They regress salary against the points, assists, rebounds, steals, blocks and teammates’ productivity (as proxied by the coach’s years coaching and a career winning percentage). The authors find that while the model is mostly insignificant for an individual, the effects on the teams are significant at the 5% level. No one better exemplifies the idea that circumstances matter than Michael Redd. Redd was stuck behind his all-star teammate Ray Allen for two and a half years in Milwaukee at the start of his career. The first year after Allen left and Redd became a starter marked the first of six straight seasons where Redd averaged over 20 points per game, ending only after he suffered a devastating knee injury. Before he became a starter, Redd signed a contract paying him about $3 million a year. After he became a starter and an all-star, he signed a new contract paying him on average about $15 million a year. Thus, teammates and circumstances matter. Despite the somewhat extreme nature of Michael Redd’s breakout campaign as a back up to an all-star in one year, many players put up bigger and better numbers when given more playing time. Do they then become more valuable players to their teams? Berri (1999) attempts to find a different measure of value by estimating how many wins a player adds to his team per minute over an average replacement. Controversially perhaps, he finds that Dennis Rodman was nearly as valuable or even more valuable than Michael Jordan during the Chicago Bulls 1998- 99 championship run due to his 15.0 rebounds per game average compared to a league average of 8.8 per game for his position and this brings to light the idea that perhaps some statistical measures are overvalued and that more advanced measurements are required, a deficiency addressed by the current study.
The economic theories outlined here provide a sound base for the current study. However, concrete evidence in the form of data is required to draw any conclusions going forward. The next section presents the data and model used to provide such proof.
RESEARCH METHODOLOGY
Introduction
In this chapter, the procedures and techniques used in conducting the study are described here. It comprises of the following: research design, collection of data, methods of data analysis.
Research Design
The research design refers to the overall strategy to be adopted, to integrate the different components of the study coherently and logically, thereby ensuring the research problem and research questions will be adequately addressed.
Thus, the function of the research design is to ensure the evidence obtained enables one to address the research problem logically effectively, and as unambiguously as possible therefore a key element is required to draw firm and convincing conclusions.
This study will take up a correlational research design to answer the research problem which is to investigate the relationship that exists between NBA’s players’ salary as a dependent variable and selected independent variables of age, agesq, points, blocks, steals, assists, rebounds, g, ws48, per, usg, vorp, pg, sf, eastern, round 1, American.
Methods of data collection
This study made use of secondary data; hence data was obtained from basketball-reference.com
Operation definition of variables
Dependent variable
Salary- The salary an NBA player earns.
Independent Variables
Some definitions of independent variables
Age- The age of an NBA player
Points- Points scored by a player in an NBA game
Blocks- Blocks performed by a player
Pg-Whether, a player, plays point guard
Sf-Whether, a player, plays small forward
Sg-Whether, a player, plays shooting guard
Data analysis
Data was analyzed using computer software Stata package 15.1. To accomplish the desired objectives, the technique of multiple regression was employed. To find the coefficients of the model, one fits the following equation
y = β0+β1×11+β2×2+β3×3+…
The βs are coefficients while xs represent the independent variables
I will fit several regression models until I get the best optimal model and that will be the final model
To find the independent variables that have a significant impact on the dependent variable y, we will look at the p-value, if it is lower than 0.05 then it is substantial, the closer to zero the p-value is, the better.
DATA ANALYSIS
Regression 1
This analysis contains the following variables height, weight, age, points, blocks, steals, assists, rebounds, ft, fg3, fg, mp, ows, dws, ws, pg, sg, sf, pf, c, round1 and american
note: pg omitted because of collinearity
age, points, steals, assists, rebounds ft, fg3, fg and round1 have significant effect on salary
y = β0+β1x,1+β2×2+β3×3+β4×4+β5×5+β6×6+β7×7+β8×8+β9×9+β10×10+ β11×11 + β12×12 + β13×13 + β14×14 + β15×15 + β16×16 + β17×17 + β18×18 + β19×19 + β20×20 + β21×21
We apply this model with coefficients in the column coef. replacing β1 to β2, x1 to x21 being replaced by height, weight, age, points, blocks, steals, assists, rebounds, ft, fg3, fg, mp, ows, dws, ws, sg, sf, pf, c round1 american variables respectively
Regression 2
This regression analysis contains the following variables age, agesq, points, blocks, steals, assists, rebounds, g, ws48, per, usg, vorp, pg, sg, sf, pf, eastern, round1 and american.
age, agesq, points, blocks, steals, assists, rebounds, g, per, vorp, pg, sg, pf and round1 have significant effect on salary
y = β0+β1x,1+β2×2+β3×3+β4×4+β5×5+β6×6+β7×7+β8×8+β9×9+β10×10+ β11×11 + β12×12 + β13×13 + β14×14 + β15×15 + β16×16 + β17×17 + β18×18 + β19×19
We apply this model with coefficients in column Coef. replacing β1 to β 19, x1 to x19 being replaced by age, agesq, points, blocks, steals, assists, rebounds, g, ws48, per, usg, vorp, pg, sg, sf, pf, eastern, round1, american
β0 is the constant
Regression model 3
From the first regression I had a total of 21 independent variables namely: height, weight, age, points, blocks, steals, assists, rebounds, ft, fg3, fg, mp, ows, dws, ws, pg, sg, sf, pf, c, round1, and american
But I found that the independent variables of significant effect on the dependent variables are age, points, steals, assists, rebounds ft, fg3, fg and round1, therefore I will use these variables to fit model 3
Our third model becomes
y = β0+β1×1+β2×2+β3×3+β4×4+β5×5+β6×6+β7×7+β8×8+β9×9
Where y-Dependent variable y
β0-Constant
β1- β9 are the coefficients of independent variables in the Coef. Column
x1-x9 represent the independent variables age, points, steals, assists, rebounds ft, fg3,fg and round1 respectively
Regression 4
From the second regression I had a total of 19 independent variables namely: age, agesq, points, blocks, steals, assists, rebounds, g, ws48, per, usg, vorp, pg, sg, sf, pf, eastern, round1, american
But I found that the independent variables of significant effect on the dependent variables are age, agesq, points, blocks, steals, assists, rebounds, g, per, vorp, pg, sg, pf and round1, therefore, I will use these variables to build model 4
The fourth model becomes
y = β0+β1×1+β2×2+β3×3+β4×4+β5×5+β6×6+β7×7+β8×8+β9×9+ β10×10+ β11×11 + β12×12 + β13×13 + β14×14
Where y-Dependent variable y
β0-Constant
β1- β14 are the coefficients of independent variables in the Coef. Column
x1-x14 represent the independent variables age, agesq, points, blocks, steals, assists, rebounds, g, per, vorp, pg, sg, pf and round1
Final Regression model
Looking at model 3 and model 4, we see that the independent variables age, points, assists, rebounds, round 1, g, per, vorp have a significant effect on the dependent variable salary, therefore I will use these variables to fit the final model
This model becomes:
Salary= 63.90514age+ 50.33193points+ 58.97752assists+ 68.71918rebounds+ 169.4335round1 -3.183174g+ -21.91137per+ 101.7727vorp+ -1507.759
CONCLUSION
The purpose of this study was to identify determinants of an NBA player’s salary in an attempt to determine how much control players can exert over their expected salary and how much is due to factors out of their control such as height and age. With the salary cap expected to rise in the coming years, players can maximize their income if they know what teams look for in players. At the same time, the sides can identify overvalued or undervalued skills in an attempt to beat the market. The analysis revealed that teams most highly value statistics that contribute directly to wins. Steals, points, rebounds were easily the most significant variables, and assists were not far behind that group. This paints a scene of the NBA as a league driven by results.
In the future, research should include the effect on career earnings, not just I a season, investigating what skills translate best into career longevity. Also, an analysis of each statistical measure would be beneficial for studies such as this: looking at the volume of points per game versus efficiency and the effects on annual and career earnings, This study, in particular, could be refined by using more advanced data, such as synergy stats, offensive and defensive ratings, and more. An investigation into whether career stats or contract year stats are influential should also be considered. The specific determinants of an athlete’s salary will likely never be fully unlocked due to the human nature involved in all such contracts, but this study is a good step in that direction.
References
- Asteriou, S. G. Hall, Applied econometrics, Palgrave Macmillan (2011)
- Cohen, Statistical Power Analysis for the Behavioral Sciences, 2nd edition (1988)
Investopedia, Hypothesis-testing, http://www.investopedia.com/exam-guide/cfa-level-1/ quantitative-methods/hypothesis-testing.asp#ixzz4ecUhPYTa 2017-04-20
- Kennedy, A guide to econometrics. 6. ed., Oxford: Blackwell (2008)
- Lang, Elements of Regression Analysis, KTH Mathematics (2015)
- Coon, NBA Salary Cap FAQ, Collective Bargaining Agreement (2016) http://www. cbafaq.com/salarycap.htm#Q5 2017-05-03
- Gaines, The NBA is the highest-paying sports league in the world, Business Insider (2015) http://www.businessinsider.com/sports-leagues-top-salaries-2015-5?r=US&IR=T& IR=T 2017-05-09
- S. Totty, M. F. Owens, Salary Caps and Competitive Balance in Professional Sports Leagues, Journal for Economic Educators (2011) https://www.researchgate.net/publication/ 227458677_Salary_Caps_and_Competitive_Balance_in_Professional_Sports_Leagues 2017-05-09
- Wallace, The Advantages of Salary Caps, Small Business – Chron.com (2011) http: //smallbusiness.chron.com/advantages-salary-caps-18682.html 2017-05-09
- M. Wooldridge, Introductory Econometrics: A Modern Approach, South-Western, 5th Edition (2013)
- Montgomery, E. Peck, G. Vining: Introduction to Linear Regression Analysis, WileyInterscience, 5th Edition (2012)
NBA.com, (2015) http://www.nba.com/2015/news/07/08/nba-salary-cap-2016-official-release/ 2017-05-04
NBA.com, FAQ (2016) http://www.nba.com/news/faq 2017-05-05
NBA.com, (2016) http://www.nba.com/2016/news/07/02/nba-salary-cap-set/ 2017- 05-04 C. Neiger, How Salary Caps Changed Sports, Investopedia (2010) http://www.investopedia. com/financial-edge/0910/how-salary-caps-changed-sports.aspx 2017-05-09