Returns to Skill in Professional Golf

Returns to Skill in
Professional Golf
A Quantile Regression Approach
Leo H. Kahne
International Journal of Sport Finance, 2010
Overview
 Scully’s paper was the first to link player skills with
compensation in MLB
 Spurred a number of golf related studies
 There are a host of papers studying the returns to
skill in professional golf
 Focus on the importance of driving distance, iron
play, greens in regulation and putting
 Past studies have determined that importance of
driving has increased but putting remains the
most important skill
 Most recent studies prior to 2010 focus on the
fact that skill does not equal earnings
 Skill=Score=Rank=Earnings
Purpose
 To examine the linkage between professional
golfers’ earnings and their skills with the use of
quantile regression with data from 2004-2007
 Past studies did not account for the highly
positively skewed distribution of earnings
 Small group of highly talented golfers
 Tiger Woods, Vijay Singh, Phil Mickelson – 49/135
first place wins in ‘04-’07 (36%)
 Nonlinear payout structure
 Distribution of earnings:
 First place: 18% of purse
 Second place: 10.8% of purse
 Third place: 6.8% of purse
 Fourth place: 4.8% of purse
cbssports.com
Model
 Past models used simple conditional mean
regression estimation (OLS – Ordinary Least-Squares)
 Causes difficulties due to non normality of the errors
 New Model: Quantile Regression
 Can provide more understanding to the effects of
various covariates on earnings
 Model:
 Yi – dependent variable, real earnings per PGA event
 xi – vector containing five variables expected to explain
golf earnings
 B – vector of coefficients to be estimated
 Ei – error term
 q – specific quantile associated with the equation (5
quantiles)
Model, cont.
 Five Variables (vector xi)
 Greens in Regulation
 Positive coefficient expected
 Putting Average
 Negative coefficient expected
 Save Percentage (Up & Downs)
 Positive coefficient expected
 Yards per Drive
 Positive coefficient expected
 Driving Accuracy
 Positive coefficient expected
 Also measures experience, height and weight
Method
 Took summary of data to prove that methods
such as OLS can be influenced by outlier effects
 Tiger Woods – he greatly affected every estimated
coefficent except Driving Accuracy
 Looking to get a more
realistic estimation for
the “typical”
professional golfer
Method
 Took the OLS results and compared them to five
different quantiles:
 q10 (lowest earnings)
 q25
 q50
 q75
 q90 (highest earnings)
Results
 All five variables had expected coefficient signs,
except driving accuracy
 Negative coefficient could be attributed to tradeoff
between driving accuracy and distance
 Estimated coefficients for Greens in Regulation,
Putting Average & Save Percentage most
significant at 1% level
 Estimated coefficients for Yards per Drive is more
statistically significant at the 10th and 25th quantile
regressions and then becomes less significant
 Driving Distance is more important for those on the
lower end of earnings
Results, cont.
 From OLS results, we see that an increase in one
percentage point on greens in regulation,
increases earnings by $7,485
 In Quantile Regression results, we see that an
increase in one percentage point in GIR
increases earnings by $4,111
 Due to skewness and outliers in earnings measure
 The coefficient becomes more significant as we
move from the lower quantiles to the higher
quantiles, implying that an increase in greens in
regulation has a greater positive affect on
expected earnings for the better golfers
Results, cont.
 OLS results predict an increase in earnings by
$700,082 for reducing putting average by one
stroke
 Quantile regression predicts a $374,041increase
for reducing putting average by one stroke
 If golfers in the 10th quantile of earnings decrease
average by one stroke, expected earnings will
increase by $182,000
 If golfers in the 90th quantile of earnings decrease
average by one stroke, expected earnings will
increase by $717,000
 Again, due to nonlinear payout structure
Results, cont.
 This conditional quantile regression takes it a step
further and will help professionals realize not just
what is important, but which important aspects
are the best for them to focus on given their pay
rank
 For example, a golfer in the 25 quantile can increase
earnings by $5,739 by reducing putting average by
one standard deviation, which is approximately a
32% increase in earnings per event
 A golfer in the 75 quantile will see approximately a
$9,973 increase by reducing putting average by one
standard deviation, however this is only a 16%
increase in earnings per event
Thoughts
 ‘04-’07 did not provide the most interesting data
due to the strong outliers, most specifically Tiger
Woods
 Would be interesting to see the results now with
the huge amount of upcoming talent and new
winners almost every week
 Would also be interesting to see how skill and
tournament earnings affect sponsorship earnings
per player