1 / 26

New Approaches to Player Valuation: Analyzing How Wins Generate Revenue for Major League Baseball Teams

New Approaches to Player Valuation: Analyzing How Wins Generate Revenue for Major League Baseball Teams. Graham Tyler Brown University ’12 Honors Thesis in Economics. Unique Nature of Baseball . Everything that baseball players do on the field is recorded

kyle
Download Presentation

New Approaches to Player Valuation: Analyzing How Wins Generate Revenue for Major League Baseball Teams

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. New Approaches to Player Valuation: Analyzing How Wins Generate Revenue for Major League Baseball Teams Graham Tyler Brown University ’12 Honors Thesis in Economics

  2. Unique Nature of Baseball • Everything that baseball players do on the field is recorded • Easier than in almost all other professions to quantify performance • Defining product as wins, we can determine a players product for any given period of time • If we know how wins generate revenue, we can determine a players marginal revenue product—the amount of revenue he was responsible for generating—for a period of time

  3. Motivation • If we can estimate a player’s marginal revenue product we can use this information to analyze various economic phenomena and test economic theories across a wide range of topics • Monopsonistic exploitation (Raimondo, 1983) and owner collusion (Bruggink and Rose, 1990) • Presence of a winner’s curse in auctions (Cassing and Douglas, 1980) and (Burger and Walters, 2006) • Human capital theory of labor markets (Blass, 1992) • Status of MLB owners as rational, profit maximizing, and having essentially perfect information

  4. Previous Methods for MRP Estimation • Anthony Krautmann’s method based on the economic theory of markets • Commercial approach utilized by analysts such as Vince Gennaro (Diamond Dollars) and Nate Silver (Baseball Between the Numbers)

  5. Scully Model (Gerald Scully, 1974) • Estimate player’s added wins • New sabermetric methods have given us Wins Above Replacement (WAR) • Estimate the added revenue from a win through regression of revenue on team wins • MRP=WAR*wins

  6. Adding Complexity to the Revenue-Wins Model • Nonlinearity • Wins increasing probability of making playoffs more valuable than additional wins at lower win totals • Team Specificity • Wins probably more valuable for some teams

  7. Market Size, Pay, and PerformanceJohn D. Burger and Stephen K. Walters(2003) • Linear spline function at a threshold number of wins to capture nonlinearity • Interaction between market size and wins to capture increased marginal revenue from wins for teams in larger markets • Implies team variation in returns to winning is entirely due to variation in market size

  8. Regression Model • Used data released from 2000 Blue Ribbon Panel hired by MLB to assess growing disparity in local revenues between teams and the effect on competitiveness • Regression Model: • TRit = α1 + α2t+ β0Mit + β1Mit • Wit + β2Mit • W2it + β3Stadiumit + β4Ageit + εit • TR = real local team revenue in millions of 1999 dollars; • t = time trend • M = market size, measured as metropolitan area population in millions • W = regular season wins • W2 = wins above the contention threshold of 84 • Stadium = a binary dummy variable that takes a value of one if the team has a new or significantly renovated stadium • Age = number of years since the opening of a new or renovated stadium

  9. Hypotheses and Assumptions of My Research • Assumptions • Revenue-win relationship is nonlinear and varies by team • Hypothesis 1: Price is an omitted variable when not included in the revenue-win model • If price is set in anticipation of team performance it is correlated with revenue and with actual wins • Hypothesis 2: Team variation cannot fully be explained by market demographic factors • Hypothesis 3: Teams can be grouped by similar returns to winning and there will be variation in the revenue-win relationship by group among the three different CBAs of the last 15 years

  10. Hypothesis Testing Strategy • H1 and degree of team variation tested through Model 1: • Revenueit=0Teami + 1 Teami *WinPctit + 2Teami*WinPct2it1/2 + 3Priceit + 4Stadiumit-1 + 5Stadiumit-2 + 6Newit-1 + 7Newit-2 +8Playoffsit-1 + 9t + it • F-test of team win coefficients tests team variation in returns to winning • F-test of team win coefficients in unrestricted version vs. version restricting price=0 tests H1

  11. Hypothesis Testing Strategy Continued • H3 Tested through a categorical version of Model 1: • Revenueit=0Categoryj*CBAt+ 1Categoryj *WinPctit*CBAt + 2Categoryj*WinPct2it1/2*CBAt + 3Priceit + 4Stadiumit-1 + 5Stadiumit-2 + B6Newit-1 + B7Newit-2 + 8Playoffsit-1 + 9t + it • If we can group similar teams this clustering should produce clearer estimates that will vary within group across the three time periods if the CBAs had a significant effect on the return to winning

  12. Replication of Burger and Walters Results

  13. Pooled Teams Version of Model 1

  14. Results of Full Specification of Model 1 • Validation of H1 (significant difference between win coefficients in restricted vs. unrestricted models) • Significant team variation • Many insignificant win coefficients (large SE’s) • Some teams look as expected • Some teams have significant negative linear coefficient OR significant negative threshold coefficient • For ARI this negative win coefficient persists in specification 4 (when threshold variable is omitted) • For some teams nonlinear specification appears to make sense and threshold coefficient is significant, for others this may not be the case

  15. Robustness Check: Categorized Version of Model 1 • Teams categorized through two methods • 1) 5 groups based on teams with similar estimates for returns to winning • 2) 5 groups determined theoretically based on knowledge of teams’ fan bases and historical performance • Both groupings result in very few significant win coefficient estimates • Indicates it is very difficult to group teams based on returns to winning • No definitive evidence CBAs had significantly different effect on returns to winning

  16. Variation Across Teams in Returns to Winning

  17. Testing H2 • Model 2 tests how much of the team variation in the returns to winning can be explained by a team’s market characteristics: • Coefficienti=0 + 1Marketi + 2Incomei + 3Divisioni + 4Distancei + 5Sportsi + i,

  18. Results of Model 2 Estimation: Linear Combination of Win Coefficients

  19. Implications of Model 2 Results • Market size explains almost none of the variation • Picks up effects of “pro sports” variable and “distance” variable when these are individually left out • About 40% of team variation cannot be explained

  20. What DO We Know? • Price Matters (sometimes) • ARI is the least affluent U.S. market in the sample • Expected winning  perceived higher demand  higher prices  reduced attendance/revenue due to high consumer price elasticity of demand  high price, high wins (if wins correlated with predicted wins), low revenue • Model estimates average effect of price across teams with significant positive coefficient • Wins coefficient for ARI is then its deviation from the average price effect across all teams, not actually the returns to winning

  21. What DO We Know? • Perpetual winning or losing reduces the returns to winning

  22. What DO We Know? • There appears to be a capacity constraint due to limited number of seats in the stadium

  23. What DO We Know? • Market size is not the determining factor • Anecdotally: LAA vs. LAD, CHC vs. CWS, SF vs. OAK (very different estimates, baseline revenues, and returns to winning for teams within the same market) • Seems that in larger markets with two teams there is an established team with a large fan base and baseline revenue vs. a team trying to expand fan base with more variation in revenue (and thus in some cases higher returns to winning)

  24. Caveats • Small sample size and specification of the model

  25. Caveats Continued • Potential unobserved factors • Television deals • Lagged effect of winning (CWS) • Historical relationship with fans • Can this be created over time?

  26. Future Research • Relationship between winning and valuation of franchise as a whole • Teams typically sell for 2-2.5x revenue

More Related