From 1996-2000 the New York Yankees
won four out of five World Series; in 2000 their payroll was $92.5 million, more
than five times as high as the Minnesota Twins’ $16.5. In that time frame, none
of the teams in the bottom half of payroll spending won a single postseason
game.[1]
But in 2002 and 2003 the Oakland A’s, one of the lowest-spending teams in
baseball, made the playoffs both years, at one point winning twenty straight
games in a row. What had changed?
In 2002, responding to the lack of
competitiveness of small-market teams, Major League Baseball instituted
revenue-sharing. Under that system every
team contributed 31% of their revenue to a common pot, with the money
redistributed to every team; as well, central funds from national broadcasts were
redistributed. At the same time a luxury tax was instituted, fining teams that exceeded
a certain threshold of spending. While revenue-sharing was phased in, another
trend was taking baseball by storm: sabermetric analysis, more popularly known
as moneyball. This approach, popularized by Billy Beane, the General Manager of
the Oakland A’s, allowed small-market teams to compete with their richer
big-market brothers. While the success of the A’s was certainly impressive, the
question remained: had the competitive balance really been restored? Just how
much had revenue sharing and moneyball changed baseball? Post-2002, did
small-market teams become more competitive because of revenue sharing, and did
the cost of a win go down because of the moneyball effect?
To analyze this I am using a
time-series cross-sectional analysis for the years 1985-2014. The dependent
variable is winning percentage (wins divided by total games). The independent
variable is salary share (team payroll divided by league payroll) which
controls for the league-wide increases in payroll over the thirty-year period
as well as inflation. A dummy variable was created to differentiate between the
pre-2002 and post-2002 periods.
Figure 1 below shows the
distribution of salary share. It is relatively rare for a team to command more
than 6% of the total salary of the league, and only three times has the share
been greater than 8% (the Yankees from 2004-2006).
In figure 2 below the relationship is visualized by year,
with 2002 indicated as the year that revenue sharing went into effect.
Somewhat counterintuitively, it
would appear that revenue sharing had little effect on the distribution of
salary share by year. In fact, the opposite would appear to be true. The spread
becomes greater after revenue sharing, with the Yankees hitting three of their
highest years of salary share after revenue sharing went into effect. This can
be explained by the inauguration of the Yankees’ YES television network, which
launched in 2002 and is now valued around $3.4 billion, providing the Yankees
with a great deal more money to expand payroll.[2]
The effects of revenue-sharing and
moneyball post-2002 are shown in Table 1 below:
Table
1. Salary and Win-Loss
|
|
|
|
Win Percentage
|
(1)
|
|
Share of
Salary
|
0.02***
|
|
|
(7.26)
|
|
Post-2002
|
0.00
(0.33)
|
|
|
|
|
Constant
|
0.43***
|
|
|
(42.85)
|
|
R-Squared
|
0.130
|
|
Observations
|
858
|
t statistics in
parentheses. * p < 0.05, ** p < 0.01,
*** p < 0.001
The above model indicates that share
of salary as a percentage of the league salary is correlated with win
percentage, which should come as no surprise. A 1 unit increase in salary share
corresponds to a .02 unit increase in team win percentage. To give an example,
in the year 1985 the total league salary was $262 million; that means that
spending an extra $2.6 million (about the difference between Baltimore and the
Yankees that year) correlated to 3.24 extra wins on average. In that year three
out of the four divisions in baseball were won by less than three games,
meaning those 3.24 extra wins are quite valuable (whether they were worth $2.6
million is for a GM to decide). By 2014, with a total league salary of $3.19
billion, those same 3.24 wins would cost $31.9 million, or about the price of
Alex Rodriguez.
The model also indicates that there
is no significant difference between the periods before and after 2002. The
interaction effect was negligibly small and not significant at all. This
indicates that revenue sharing and moneyball did little to impact the lack of
competitive balance in baseball. The effect of moneyball was short-lived, since
other teams (especially big-market ones) quickly adopted sabermetrics and thus
reduced the edge for small-market teams like Oakland.
One major issue with the findings
is the above-mentioned phenomenon of regional sports networks like the Yankee’s
YES network. This was an extra revenue stream that was instituted exactly
concurrent with revenue sharing, making it difficult to assess the impact of
revenue sharing. Nonetheless, the findings show clearly that salary is
correlated with win percentage. Neither the effect of moneyball, nor of the
revenue sharing, was enough to significantly change that relationship. These
findings support the notion that salary caps (like those employed in basketball
and football) create more parity and a less direct relationship between money
and wins. That is bad for fans of the Yankees, but good for everyone else.
Appendix:
Table
1. Summary Statistics
|
|
|
|
|
|
|
|
Mean
|
Std.
Dev.
|
Min.
|
Max.
|
|
Win Percentage
|
0.50
|
0.07
|
0.27
|
0.72
|
|
Salary Share
(%)
|
3.49
|
1.26
|
0.32
|
9.52
|
|
N
|
858
|
|
|
|
Do File:
* Preamble
set
more off
ssc
install estout
cd
"G:\"
**merge databases
collapse (sum) salary, by (yearid teamid)
import delimited C:\Users\ad9538a\Desktop\Teams.csv, clear
keep yearid teamid franchid w l era g r
drop if year <=1984
merge 1:1 yearid teamid using G:\Stata\Blogpost_2\Salaries_collapsed.dta
** generate variables
gen winprct = w/g
gen runspergame = r/g
egen totalsalary = sum(salary), by (yearid)
gen salary_share = (salary/totalsalary) * 100
gen salarymil = (salary/1000000)
**transform
gen dummyyear =.
replace dummyyear = 0 if year < 2002
replace dummyyear = 1 if year >= 2002
** time-series reg
encode franchid, gen (team)
sort team yearid
xtset yearid team
** year v. salary graph & histogram
histogram salary_share, frequency ///
title("Figure
1. Salary Share Histogram", size(large) color(black)) //
twoway (scatter salary_share year, xline(2002)), ///
title("Figure
2. Salary Share by Year", size(large) color(black)) ///
note("Red
line indicates revenue sharing") legend(off) //
xtitle("year")///
ytitle("Salary
Share (%)") //
** win percentage vs. percentage of total salary of the
league
xtreg winprct salary_share if year<2002
xtreg winprct salary_share if year>=2002
xtreg winprct salary_share i.dummyyear
c.salary_share#dummyyear
*** Summary Statistics
estpost sum winprct salary_share, listwise
esttab
using sum2.rtf, cells("mean(fmt(2)) sd(fmt(2)) min(fmt(2))
max(fmt(2))") ///
title("Table
1. Summary Statistics") nonumber replace
***** Pretty Table
eststo clear
eststo:
quietly reg winprct salary_share i.dummyyear c.salary_share#dummyyear //
esttab
using file_nam3.rtf, replace label ///
stats(r2
N, fmt(3 0) labels("R-Squared" "Observations")) ///
legend
varlabels(_cons Constant) ///
b(2)
t(2) ///
title({\b
Table 2. Salary and Win-Loss})
eststo
clear
[1]
Levin, Richard; Mitchell, George; Volcker, Paul; Will, George. “The Report of
the Independent Members of the Commissioner’s Blue Ribbon Panel on Baseball
Economics”. (July 2000). http://www.mlb.com/mlb/downloads/blue_ribbon.pdf
No comments:
Post a Comment