Content uploaded by Jeffrey Cisyk
Author content
All content in this area was uploaded by Jeffrey Cisyk on Jan 23, 2020
Content may be subject to copyright.
Impacts of Performance-Enhancing Drug Suspensions
on the Demand for Major League Baseball∗
Jeffrey Cisyk†
January 1, 2020
∗The author thanks Dr. Pascal Courty immensely for comments.
†University of Victoria; and,
Charles River Associates
1 South Wacker Drive,
34th Floor,
Chicago, IL 60606
jcisyk@uvic.ca.
Impacts of Performance-Enhancing Drug Suspensions
on the Demand for Major League Baseball
Abstract
In 2005, Major League Baseball (MLB) introduced a new policy regarding the use of
performance-enhancing drugs (PEDs) wherein the league would not only suspend, but pub-
licly name any player who tested positive for banned PEDs. Using the estimated television
audience size of MLB games from 2006 to 2012, these PED suspension announcements pro-
vide a unique natural experiment to test how consumers react to news of PED use. This
study finds that PED announcements have two major impacts on the demand for baseball.
First, there is on average an immediate 9.3% reduction in the television audience of the PED
player’s team. Second, the magnitude of the effect gradually decreases over time yet remains
negative and significant for a period 37 days or approximately 33 game-broadcasts. This is
the first study to link PED use to an adverse reaction by consumers in a systematic way
using television audience while controlling for the change in team quality caused by removing
the suspended player from the team.
Keywords: performance-enhancing drugs, doping, baseball, Major League Baseball, televi-
sion audience.
1 Introduction
Professional sports has a very contentious and convoluted history with doping. While consumers
pay great sums of money for the chance to witness an exceptional athletic performance, they
also generally disavow the use of any performance-enhancing drug (PED) in order to achieve
it. A 2005 poll found that 86% of Major League Baseball (MLB) fans say PEDs are either ‘a
serious problem’ or ‘ruining the game’ while only 5% think PEDs ‘make [the] game better.’1
Yet, no matter how outraged a random sample of self-purported baseball fans may seem,
very few studies have found systematic evidence that consumers respond adversely to PED use.
Surveys are also not necessarily guaranteed to be well-representative of individuals who pay to
take in sports. A preferred approach may be to use consumer expenditure on sports as a proxy for
demand, such as ticket sales and home-game attendance, to measure any changes resulting from
news of PED use. This study suggests using the television audience of MLB game-broadcasts to
measure the impact of PED suspensions on the demand for baseball. It is argued that television
audience is a more meaningful proxy for demand for three reasons: (a) due to the nature of the
data, there is potentially a limit to the observed response to news of a PED announcement when
using attendance (‘paid attendance’ measures ticket sales regardless of a ticket-holder’s decision
to physically attend); (b) impacts of a PED announcement can be measured immediately using
television audience (can only be measured beginning the next available home game when using
attendance); and (c) a PED announcement can potentially have a much larger economic impact
on the television audience for the PED player’s team (affects television audience of home and
away game-broadcasts; television audience is much larger than home-game attendance).
The top-left panel of Figure 1 illustrates the first-order impact of a PED announcement on the
television audience of MLB game-broadcasts. Using the pitcher Eliezer Alfonzo as a case study,
deviations from the predicted television audience are plotted for 15 days before and after his
PED suspension announcement (the dashed vertical line indicates the day of the announcement
and the x-axis represents the number of days since the suspension was first announced).2This
high-level evidence suggests that consumers have a quite prompt and pointed reaction to news
of PED use. Yet, since the guilty PED player is also removed from competition, much or all
of this effect could be explained by the change in the quality of the team. However, comparing
the deviation when Eliezer Alfonzo was suspended to that of when he was injured (the top-right
panel of Figure 1), there does not appear much support for this second conjecture.
18% say PEDs are ‘not a problem’; 1% have no opinion - see Gallup, “Baseball Fans Have Little Patience for
Steroid Abuse,” https://news.gallup.com/poll/15379/baseball-fans-little-patience-steroid-abuse.aspx.
2Predicted audience comes from a linear regression of natural logarithm of television audience/attendance and
day of week, month, year, and team fixed effects.
1
-10
-5
0
5
-15 -10 -5 0 5 10 15
Television audience
during PED suspension window
-10
-5
0
5
-15 -10 -5 0 5 10 15
Attendance
during PED suspension window
-10
-5
0
5
-15 -10 -5 0 5 10 15
Television audience
during injury window
-10
-5
0
5
-15 -10 -5 0 5 10 15
Attendance
during injury window
Deviation from Predicted Value (%)
Days since PED suspension/injury announcement
Eliezer Alfonzo - Suspended: 9/14/2011; Injured: 6/9/2007
Impact of PED suspension/injury announcement
Figure 1: Example of the impact of a PED suspension and an injury announcement on proxies
for MLB demand using pitcher Eliezer Alfonzo. Predicted values come from a linear regression
of natural logarithm of television audience/attendance and day of week, month, year, and team
fixed effects.
Shifting the focus from an anecdotal to a systematic relationship, this study estimates the
effect of a PED suspension announcements and finds it has two major impacts on demand. First,
there is on average an immediate 9.3% reduction in the television audience of the team. Second,
the magnitude of the effect gradually decreases over time yet remains negative and significant
for a period of 37 days (approximately 33 game-broadcasts). This negative impact on television
audience is found to exist even after considering the change in quality of the team stemming from
the removal of the PED player while they are suspended. This is the first statistical evidence of
a systematic negative relationship between PED use and television audience.
2
2 Background and Literature Review
Broadly speaking, a PED is any substance taken to increase athletic performance. This includes
a wide range of drugs such as anabolic steroids or amphetamines otherwise used to treat atten-
tion deficit hyperactive disorder. Beginning in 2005, MLB implemented PED testing, selecting
players at random and publicly announcing the names associated with positive test results.
Since 2006, regulations for PED testing have been set forth by the Joint Drug Prevention and
Treatment Program (JDP).3Under the JDP, all MLB players are subject to at least three
random PED tests per year. Within 72 hours of a positive PED test, MLB reveals the name
of the guilty player and said player is immediately suspended without pay. The length of the
suspension depends on the individual’s number of previous positive tests as well as the type of
substance. Although the severity of punishment per offence has since increased, the punishment
regime was consistent from 2006 to 2013 - see Table A1 for details. Most importantly, by pub-
licly announcing the names of the players caught using PEDs, MLB under the JDP becomes an
ideal natural experiment where one can empirically test the effects of exogenous shocks - news of
a positive result of a random PED test - on a dependent variable - the demand for live baseball.
There is already a large and well-developed body of demand estimation of professional sports
dating back to the mid-1970s (Noll, 1974). Traditionally, this demand estimation has been
accomplished by using attendance as a proxy for sports demand - see Borland and MacDonald
(2003) and Villar et al. (2009) for a list of approximately 100 such unique studies. However,
issues with the caveats of home-game attendance in addition to a need for a more varied approach
have led researchers to investigate alternative proxies for demand (Buraimo and Simmons, 2015)
(Tainsky, 2010).
There is also a second relevant body of economic literature regarding issues related to the
use of PEDs. Most of the focus is on understanding what is means to be compliant with
PED regulations and how best to achieve said compliance. Studies can then be split into two
categories on the debate of approaches to achieve compliance (preventative versus punitive)
and most studies make use of a model of an individual’s expected utility as a function of the
risk-weighted costs and benefits of using PEDs (Eber, 2008) (Haugen, 2004) (Maennig, 2002)
(Maennig, 2014) (Maennig et al., 2009).
However, for all the research into the theory of how to best limit the use of PEDs in sports,
very few studies show there are any real consequences to league organizers for failing to do so, i.e.
that consumers react adversely when athletes use PEDs. There is but a small intersection of the
3See MLB Joint Drug Prevention and Treatment Program, http://mlb.mlb.com/pa/pdf/jda.pdf.
3
previous two bodies of literature that attempt to explore consumer behaviour in light of PED
revelations. First, surveys have been utilized to show that consumers have little tolerance for
PED use in sports and suggest that there is at least a potential for PED use to impact consumer
demand (Engelberg et al., 2012) (Solberg et al., 2010). Relatedly, Buechel et al. (2016) illustrate
how such a potential consumer backlash to PED use could actually exacerbate PED prevalence
by incentivising league stake holders to under-report and/or under-test for its use: athletes could
therefore further dope with a low probability of having their transgressions revealed. However,
the intentions of survey respondents can often differ wildly from their own real-world actions
(Brenner and DeLamater, 2016). This issue is compounded by the fact that surveys can often
under-represent the particular subset of the population that actually pay, with money and/or
time, to consume sports (Cisyk and Courty, 2017).
To correct for the lack of external validity and/or representativeness in surveys, Cisyk and
Courty (2017) use home-game attendance as a proxy for the MLB demand and measure the
impact of PED suspension announcements on attendance. The authors find PED announcements
lead to a short-term reduction in attendance which is shown to directly impact the PED player’s
team’s revenue. However, the authors use home-game attendance, which, as eluded to above,
has many constraints hence the need to consider television audience as a proxy for demand (see
Section 3 for a detailed explanation).
Van Reeth (2013) makes the first attempt to link a sport’s television audience and PED use in
a systematic way but ultimately finds an inconclusive and negligible impact. The author makes
use of two television audience metrics of the Tour de France, the average and the maximum
audience size during a broadcast but finds these metrics are not all that dissimilar when used to
explain the larger trend in television viewership. The author is able to explain many patterns
observed in the data but ultimately does not distinguish between the consumer reaction to PED
use and the consumer reaction to the change in quality of the remaining athletes after the PED
user is removed from competition.
Taking a unique approach, Brave and Roberts (2019) illustrate that PED announcements
impact a local MLB team’s non-ticket revenue at a rate of -1.7% per additional PED suspension.
Despite this negative effect, the authors note that due to the impacts PED suspensions have on
player costs, the profit-maximizing number of PED announcements is actually greater than zero
- an issue that is re-visited in Section 5.3.4
4Recall PED suspensions are unpaid and the PED player forfeits his salary for the duration of the suspension.
4
3 Data
The data used herein come from four main sources. The first data source is the estimated
number of viewers, known as the television audience, of each MLB game-broadcast as reported
by the Nielsen Company, or simply Nielsen. Nielsen is recognised as the industry standard for,
among other things, television audience measurements. More specifically, the television audience
is defined as the number of televisions currently tuned to a given broadcast within a predefined
area, known as the designated marketing area (DMA). The sample of television audience used
in this study spans seven MLB regular-seasons from 2006 to 2012, where the 30 MLB teams are
scheduled to play 162 games each season.5
Each MLB team is located within a single DMA. For example, the Atlanta Braves’ local
DMA is Atlanta, Georgia. Typically, each team participating in a game creates its own television
broadcast which is aired only within its local DMA.6As a result, each game ordinarily has two
broadcasts and produces two observations in the data.7Hereinafter, when referring to a specific
team of a game, the ‘local’ audience refers to the audience within the team’s local DMA (i.e.,
the audience of the Atlanta Braves’ game-broadcast in the Atlanta DMA) and the ‘opponent’s’
audience refers to the audience within the opponent’s local DMA (i.e., the audience of the
Philadelphia Phillies’ game-broadcast in the Philadelphia DMA if the Philadelphia Phillies were
to play the Atlanta Braves), regardless of the home/away designation of the two teams.
Note that several games per week are broadcast nationally. Nationally broadcast games are
excluded from the sample because the audience measurement includes viewers outside of either
participating team’s own DMAs. Lastly, the local DMA of the Toronto Blue Jays is outside of
the USA and local audience estimates are subsequently unavailable. The final sample contains
audience information for a total of 29,648 observations. Table 1 displays descriptive statistics
for the television audience variable. Note that Nielsen reports television audience per thousand
of individuals and values are rounded to the nearest thousand. As shown in Table 1, the average
audience size is 95 (thousand) with a standard deviation of 81 (thousand).
As noted in Section 2, numerous previous studies have used home-game attendance as a
proxy for demand. Although in most scenarios the use of attendance is entirely appropriate,
the use of television audience is potentially a superior metric for measuring the impact of PED
5Note the sample does not include games prior to or after the regular-season, i.e. the sample excludes Spring
Training and/or the post-season playoffs.
6Specifically, a team may broadcast any game in which it participates only within a predetermined area
surrounding and including the team’s DMA. For more information see MLB Constitution Article X, §3(a),
http://www.law.uh.edu/assignments/summer2009/25691-b.pdf.
7Note, even if the two participating teams share the same DMA there are still two unique game-broadcasts.
5
Table 1: Descriptive Statistics
Variable Mean SD Min Max
Audience (’000s) 95.21 81.27 1.00 750.00
Playing-Season Suspension 0.02 0.14 0.00 1.00
Off-Season Suspension 0.01 0.08 0.00 1.00
Inactive 0.00 0.06 0.00 1.00
Broadcast of home team 0.50 0.50 0.00 1.00
Predicted Season Wins 80.68 9.31 51.71 103.83
Probability of Winning Game 0.50 0.09 0.21 0.79
Divisional Rival 0.44 0.50 0.00 1.00
Interleague 0.10 0.30 0.00 1.00
Opening Day 0.01 0.08 0.00 1.00
Broadcast Length (minutes) 175.64 27.02 85.00 399.00
In-Market NFL Game 0.02 0.14 0.00 1.00
In-Market NBA Game 0.07 0.25 0.00 1.00
In-Market NHL Game 0.04 0.20 0.00 1.00
Years: 2006 to 2012
Observations: 29,648
announcements for three main reasons. First, the ‘attendance’ measure used in these studies
is often actually ‘paid attendance’ which represents the number of tickets sold to a game as
declared in each games’ official MLB box score. Consequently, after purchase of a ticket, the
consumer is counted as part of the paid attendance regardless of actually attending. This limits
the ability of the empiricist to measure the reaction of consumers if tickets have been sold prior
to a PED announcement. Furthermore, even if attendance was measured by actual number
of individuals attending a game, rational consumers may view their tickets as sunk costs and
attend regardless of what their actions would be if the decision to purchase tickets came after a
PED announcement.
Second, recall audience information is available for most games. This includes both home
and away games of the local team. Therefore, when a PED announcement occurs, it is pos-
sible to measure any potential impact from the day of a PED announcement regardless of if
the PED player’s team is playing a home or an away game. The initial impact measured on
attendance is instead limited to the subsequent home game of the PED player’s team after a
PED announcement.
Third, the measured effect of a PED announcement has the potential to have a larger eco-
nomic impact on television audience than home-game attendance: because the sample contains
information on home and away game-broadcasts, the impact of a PED announcement can be
observed for all affected game-broadcasts of the PED suspension. As illustrated in Figure 1,
despite the fact that each attendance panel makes use of the same dates as its corresponding
6
television audience panel, systemic ‘gaps’ exist in the attendance data simply because a local
team must play games outside of its home DMA where the attendance is attributed to its op-
ponent. Moreover, the television audience represents a far greater number of consumers than
those in attendance: the ratio of television audience to home-game attendance is roughly 6:1.8
Lastly, there are several smaller issues to consider. First, a stadium’s capacity places an upper
bound to the number of tickets sold and therefore an upper bound to attendance. Any measured
effect of a positive demand shock would thereby be limited in its ability to be represented through
attendance. Conversely, it is highly improbable a single television event would be constrained by
an analogous upper bound.9Second, although there is no evidence to support such a practice,
it is not unimaginable that a team would manipulate ticket prices in light of demand shocks,
thereby easing variation in the attendance. For all these reasons above, television audience is
explored to estimate PED impacts on demand.
The second source of data is the moneyline odds of each MLB game collected from Cov-
ers.com. The moneyline odds Mt,s,i are used to calculate the probability team twill win game i
in season s. The moneyline odds are converted into a decimal value of win probability as follows:
prob(wint,s,i ) = (Mt,s,i
Mt,s,i−100 if Mt,s,i <0
100
Mt,s,i+100 if Mt,s,i >0(1)
An adjustment is then made to the win probability of team tand team −tto ensure
prob(wint,s,i ) + prob(win−t,s,i ) = 1, where −tis the opponent of t. This adjustment is necessary
because to the gambling nature of the data: the sum of the decimals odds will be greater than 1
due to the bookmaker’s take or margin (ˇ
Strumbelj, 2016). For simplicity, a basic normalization
is applied as follows:
prob(wint,s,i ) = prob(wint,s,i )
prob(wint,s,i ) + prob(win−t,s,i )(2)
The probability is also used to calculate a proxy for the quality of each team. This proxy is
termed the predicted season wins and measures the expected number of wins of a team for the
entire season prior to playing a given game. It is calculated as as follows:
8The average ratio of each team’s annual attendance to annual television audience is 0.169.
9For context, Nielsen also estimates the percentage of televisions tuned to a particular broadcast, also known
as the broadcast’s ratings. The current record for the highest ratings of a nationally television broadcast is 60.3%
for the M*A*S*H finale, “Goodbye, Farewell and Amen,” aired on February 28, 1983.
7
Predicted Season Winst,s,i =
i−1
X
j=1
(1 |wint,s,j = 1) +
162
X
k=i
prob(wint,s,k ) (3)
Stated alternatively, the predicted season wins is the actual number of wins of a team for
the entire season prior to playing a given game plus the expected future number of wins in the
remainder of the 162-game regular-season.
The third source of data is information on substitutes to MLB game-broadcasts, namely that
of competing sports broadcasts of the National Football League (NFL), the National Basketball
Association (NBA), and the National Hockey League (NHL). When the local MLB team’s DMA
is also home to a substitute sports team, the substitute team is said to be ‘in-market.’ For a list
of in-market teams of the substitute sports leagues for each MLB team see Table A3. Schedules
of substitute sports’ game-broadcasts are collected from Sports-Reference.com and incidences
where a game-broadcast of the local MLB team overlaps with an in-market substitute sports
team are flagged. For example, Table 1 shows that for 7% of the observations, a local MLB
game-broadcast occurred on the same day as an in-market NBA game.
The final source of data is the variable of interest, PED events, from ProSportsTransac-
tions.com. A summary of these events is presented in Table A4. Two types of PED events are
identified based on the timing of the suspension announcement: the first type of PED event, the
off-season suspension, indicates the player was tested and found guilty of PED use outside the
window of the MLB regular-season. There are six of these events in the sample. Recall that PED
testing is random and suspension announcements would therefore be exogenous. However, this
is not to say all suspensions must be exogenous. Off-season suspensions are identified separately
due to the fact that they are always served at the beginning of the subsequent regular-season and
therefore the timing of the game-broadcasts affected by an off-season suspension is endogenous.
The second type of PED event is termed the playing-season suspension, wherein a player
has tested positive for PEDs and their suspension has been announced during the months of
the MLB regular-season. There are 12 playing-season suspensions within the sample period and
2% of all observations are game-broadcasts featuring a local team with a currently suspended
player (see Table 1).
Information is also collected on injury spells of each of the PED players from ProSport-
sTransactions.com where available. This information can be found in Table A4. For seven of
the PED players, an injury spell that is closest in time to their PED event is identified and
matched with the suspension. The injury event represents game-broadcasts for which the PED
8
player did not participate for the local team due to their physical injury.
As mentioned in Section 2, MLB announces the name of a guilty player within 72 hours of a
postive PED test result news of which is then covered and circulated by traditional and social
media. Information on injury events are typically announced through the same channels as PED
events and the speed and extensiveness of the dissemination of these two types of announcements
are expected to be similar.
4 Empirical Framework
Let Adenote the local television audience of a game-broadcast. The following specification is
estimated:
ln(A) = β0+IP ED (βP ED +βe
P ED e) + IOF F βO F F +IIN J (βIN J +βe
IN J e) + βXX+(4)
The variable IPE D is a dummy that takes the value of one if the local team has a player cur-
rently serving a playing-season suspension, and zero otherwise. Within the associated parenthe-
ses, the variable emeasures the time elapsed between the PED announcement and the observed
game-broadcast, measured in days. The impact of the playing-season suspension announce-
ment on the local television audience of the game-broadcast edays later can then be calculated
by summing βP ED +βe
P ED e. The variable IOF F is a dummy that takes the value of one in
game-broadcasts where the local team has a player currently serving an off-season suspension.
Similar to the playing-season suspension, IIN J is a dummy indicating a PED player is currently
injured and the corresponding eis the time since the injury announcement and game-broadcast.
Lastly, Xrepresents a set of control variables which are consistent with past baseball demand
estimation and includes controls for demand cycles (day of the week, afternoon/evening, month,
year, and availability of substitutes) as well as controls for both teams of the game-broadcast
(probability of winning, and local team and opponent fixed effects).
There are several potential issues with this approach regarding endogeneity. First, game-
broadcasts associated with PED announcements that occur in the off-season are not exogenous.
These game-broadcasts are always the first observations of a playing-season. As mentioned
earlier, the off-season PED announcements are treated separately by use of the variable IOF F .
Another area of concern deals with reverse causality: note that PED use increases both
9
demand for baseball and probability of a PED suspension. This conjecture would require PEDs
to have very strong and immediate effects on not only the PED player but on the demand for
the PED player’s team. While both of these aforementioned phenomena occurring simultane-
ously seems unlikely, robustness checks in Section 5.2 explore this issue further to rule out the
possibility of reverse causation.
Lastly, PED announcements may be correlated with team quality. As referenced in Section
2, the suspended player is removed from the team, thereby likely changing the quality of the
PED player’s team. To separate the impact of the PED announcement from the impact of the
change in team quality, another event is studied where the same PED players are removed from
competition for long period - this time in less dubious circumstances - due to injuries. Then,
similar to a difference-in-differences approach, βIN J +βe
IN J eis the effect of the change in team
quality on demand for baseball and (βP ED +βe
P ED e)−(βIN J +βe
IN J e) is the isolated effect of
the PED announcement absent the change in team quality.
5 Results
Estimates of Equation 4 were run on various specifications. Below is a presentation of the core
results which form the basis of the conclusions, additional robustness checks, and a discussion
of the implications of the findings.
5.1 Impacts of PED Announcements on Local Television Audience
Core results of Equation 4 can be found in Table 2 with t-statistics displayed below each esti-
mated coefficient along with asterisks indicating the conventional significance levels. Each model
is reported with heteroskedasticity-consistent standard errors.
Column 1 reports a simplified version of Equation 4. All control variables have the predicted
sign and are significant while the model is able to explain 80.5% of the variation in the television
audience. Positive coefficients are estimated for dummy variables indicating features of the game-
broadcast such as teams of the same division (+6.7%), interleague play (+7.7%), the first home
game-broadcast of the season (+42.4%), and an evening game-broadcast (+51.0%). All else
equal, game-broadcasts for which the local team is playing at its home ballpark receive a 8.6%
larger television audience. Generally speaking, higher values of predicted season wins indicate
a stronger local team and positively effect the number of viewers at a rate of +3.4% for each
additional expected win. Also, consistent with the uncertainty of outcome hypothesis, the model
10
Table 2: Regression Output
ln(Local Television Audience) (1) (2) (3) (4)
Local Team has a(n)...
Playing-season suspension -0.0346** -0.0978*** -0.0977*** -0.0965***
(-1.975) (-3.537) (-3.534) (-3.487)
Time elapsed 0.0019*** 0.0019*** 0.0019***
(3.188) (3.190) (3.170)
Off-season suspension -0.1298*** -0.1312*** -0.1311*** -0.1297***
(-3.885) (-3.924) (-3.919) (-3.876)
Injury -0.0304 -0.0299
(-0.710) (-0.699)
Time elapsed 0.0026 0.0026
(0.875) (0.882)
Local Team’s Opponent has a(n)...
Playing-season suspension 0.0259
(0.901)
Time elapsed -0.0003
(-0.440)
Off-season suspension 0.0392
(1.068)
Injury -0.0495
(-0.686)
Time elapsed 0.0028
(0.659)
Broadcast of home team 0.0824*** 0.0825*** 0.0826*** 0.0825***
(13.212) (13.231) (13.237) (13.217)
Predicted season wins 0.0330*** 0.0330*** 0.0330*** 0.0330***
(78.466) (78.439) (78.354) (78.346)
Probability of winning game 1.6570*** 1.6542*** 1.6519*** 1.6480***
(5.918) (5.912) (5.904) (5.889)
Probability2-1.1938*** -1.1918*** -1.1895*** -1.1847***
(-4.434) (-4.429) (-4.421) (-4.402)
Divisional rival 0.0652*** 0.0652*** 0.0651*** 0.0651***
(11.815) (11.817) (11.804) (11.807)
Interleague 0.0738*** 0.0738*** 0.0738*** 0.0736***
(6.918) (6.912) (6.912) (6.901)
Opening day 0.3540*** 0.3538*** 0.3537*** 0.3533***
(9.668) (9.660) (9.655) (9.636)
Evening game 0.4121*** 0.4122*** 0.4122*** 0.4122***
(41.484) (41.498) (41.492) (41.490)
Length of broadcast 0.0017*** 0.0017*** 0.0017*** 0.0017***
(17.295) (17.361) (17.358) (17.333)
In-market NFL game -0.4726*** -0.4726*** -0.4727*** -0.4730***
(-17.341) (-17.338) (-17.342) (-17.340)
In-market NBA game -0.0732*** -0.0730*** -0.0730*** -0.0731***
(-5.773) (-5.765) (-5.768) (-5.785)
In-market NHL game -0.0358*** -0.0357*** -0.0357*** -0.0354***
(-2.620) (-2.618) (-2.616) (-2.596)
11
Table 2-Continued: Regression Output
ln(Local Television Audience) (1) (2) (3) (4)
Day of Week FE Yes*** Yes*** Yes*** Yes***
Month FE Yes*** Yes*** Yes*** Yes***
Year FE Yes*** Yes*** Yes*** Yes***
Local Team FE Yes*** Yes*** Yes*** Yes***
Opponent FE Yes*** Yes*** Yes*** Yes***
R20.8051 0.8052 0.8052 0.8052
Observations 29,648 29,648 29,648 29,648
Notes:
[1] ***p<0.01, **p<0.05, *p<0.10.
[2] Model estimated with robust standard errors; t-statistics shown in parenthesis.
[3] Definitions of each variable can be found in Table A2.
suggests television audience peaks when the local team is approximately twice as likely to win
than to lose (win probability of 66.7%) (Rascher, 1999).10 Lastly, as expected, the audience
size for game-broadcasts also decline when substitutes are readily available: this includes NFL
games (-37.7%), NBA games (-7.0%), and NHL games (-3.5%).
However, as for the variable of interest, the model presented in Column 1 implicitly assumes
that the effect of a playing-season PED announcement is constant across the entirety of the
suspension. Under this constant-impact assumption, the model suggests there is but a small
negative impact of a suspension announcement on the television audience.
Instead, the same model is estimated and shown in Column 2, now allowing the impact of
the PED announcement to vary over the length of the suspension as achieved by the inclusion
of the time elapsed variable. The results of the control variables from Column 1 largely remain
unchanged, however Column 2 shows the playing-season PED announcement to now have an
initial 9.3% decline in television audience.11 The time elapsed variable suggests the negative
effect dissipates as time passes the PED announcement at a rate of 1.9 percentage points every
10 days.12 This is an interesting find. Certainly the estimate of the time elapsed variable
could be negative if information of the PED suspension is disseminated slowly, zero if time
does not affect the reaction of consumers, or non-monotonic if there are other factors at play.
Instead, a positive estimate of time elapsed indicates a ‘decaying’ effect of the impact of PED
announcements on television audience. This is consistent with a hypothesis that consumers may
not be able to recall that a player is suspended or may have ended their temporary boycott
(Cisyk and Courty, 2017).
10 1.6570
2×1.1938 = 69.4% ≈2
3
11exp(βP ED )−1 = exp(−0.0978) −1 = 9.3%
1210 ×(exp(βe
P ED e)−1) = 10 ×(exp(0.0019) −1) = 1.9%
12
-15
-10
-5
0
5
Deviation from Counterfactual Audience
Absent PED Announcement (%)
0 10 20 30 40 50
Days Since PED Announcement
Estimated Effect
of a PED Announcement
90% Confidence
Interval
Final Day For Which
Effect is Still Significant
Figure 2: Illustration of the effect of a PED suspension announcement on the local MLB televi-
sion audience - see Column 3 of Table 2.
The full estimation of Equation 4 is shown in Column 3. Similarly, the model finds an
immediate 9.3% decline in television audience the day of a PED announcement which wanes
over time. The negative effect of a PED announcement is found to be significant for 37 days or
approximately 33 game-broadcasts.13 A graphical depiction of this effect is shown in Figure 2
where the black line illustrates the average effect, the blue shaded area represents the confidence
interval, and the red dashed line indicates the final day the effect is statistically significant.
Column 3 also addresses the concern that PED announcements are correlated with team
quality by adding the controls for when the PED player is removed due to injury. As with the
controls for playing-season suspensions, a linear form is imposed on the time elapsed between
injury announcement and the game-broadcast. However, the model suggests the injures to the
PED players have no statistical impact on the television audience. Instead, consumers are
reacting solely to the news of the PED suspension.
Each aforementioned model also estimates the effect of an off-season suspension on television
audience. However, recall that off-season suspension events are identified separately because the
13Each MLB regular-season consists of 162 games played over the span of approximately 180 days, a rate of
162
180 = 0.9 games per day. One would therefore expect 37 ×162
180 ≈33 game-broadcasts to occur within 37 days.
13
affected game-broadcasts are correlated with the beginning of the regular-season. Therefore,
because game-broadcasts are not observed in the same regular-season before the off-season sus-
pension treatment, the estimated coefficient is more difficult to interpret due to the possibility
of other demand shocks that may occur coincidently.
The final model shown in Table 2, Column 4, attempts to uncover any spill-over effects of a
PED announcement. This is achieved by adding additional regressors to Equation 4 to control for
when the local team’s opponent has an injury or is currently serving a PED suspension. However,
there appears to be no evidence of a collective response from viewers to an opponent after a
PED announcement - all estimated coefficients of the opponent variables are not significant.
This suggests there is no support for any spill-over effects from PED suspensions and/or injuries
to PED players.
Instead, the impact of a PED announcement is said to be localised (affecting only the local
audience). One possible explanation may be that news of PED announcements may be dissem-
inated more thoroughly within, rather than outside, the local DMA of the PED player’s team.
Another possible explanation may be that fans do not equate watching their local team to pro-
viding support for its opponent. This is not entirely inconsistent with the conjecture that PED
news causes withdrawal of consumer support. Instead, under this hypothesis, consumers would
not necessarily display their displeasure by boycotting their local team because of its opponents’
PED suspension.
5.2 Robustness Checks
Numerous robustness checks are performed. Table 3 displays the most note-worthy variations
while others are described and not shown. Column 1 begins by considering only PED players
for which there is an injury event corresponding to their suspension. Table A4 lists seven such
‘balanced’ events for which there is a valid counterfactual to the playing-season suspension. The
remaining suspension-treated observations are removed from this specification. While the point-
estimates of the playing-season suspension variables are slightly different, following a t-test, they
remain statistically unchanged.
As eluded to in Section 4, there still remains the possibility of reverse causality wherein a PED
player is able to avoid detection long enough to impact demand before the PED announcement.
In this scenario, the negative playing-season suspension coefficient would measure the return of
an ‘artificially inflated’ demand to its otherwise expected level. To address this concern, the
dummy variable for the pre-suspension window of Column 2 takes a value of one for a period of
14
Table 3: Robustness Checks
ln(Local Television Audience) (1) (2) (3) (4) (5) (6)
Playing-season suspension -0.0704** -0.0986*** -0.1133*** -0.0975*** -0.0965*** -0.0976***
(-2.037) (-3.563) (-3.718) (-3.527) (-3.426) (-3.519)
Time elapsed 0.0020*** 0.0019*** 0.0020*** 0.0019*** 0.0019*** 0.0018***
(3.116) (3.194) (3.319) (3.191) (3.296) (3.126)
Off-season suspension -0.1377*** -0.1320*** -0.1316*** -0.1313*** -0.1310*** -0.1321***
(-4.097) (-3.947) (-3.933) (-3.931) (-3.918) (-3.946)
Injury -0.0262 -0.0291 -0.0299 -0.0314 -0.0302 -0.0299
(-0.611) (-0.678) (-0.699) (-0.728) (-0.707) (-0.689)
Time elapsed 0.0027 0.0025 0.0026 0.0026 0.0026 0.0033
(0.907) (0.851) (0.875) (0.886) (0.876) (1.097)
Pre-Suspension Window -0.0200
(-0.810)
WAR 0.0148
(1.192)
MiLB Suspension 0.0038
(0.240)
Second Offence -0.0212
(-0.570)
Reinstatement -0.0276
(-1.447)
Broadcast of home team 0.0823*** 0.0826*** 0.0825*** 0.0826*** 0.0826*** 0.0826***
(13.126) (13.242) (13.233) (13.237) (13.237) (13.245)
Predicted season wins 0.0330*** 0.0330*** 0.0329*** 0.0330*** 0.0330*** 0.0330***
(77.927) (78.450) (78.152) (78.309) (78.353) (78.529)
Probability of winning game 1.6241*** 1.6513*** 1.6540*** 1.6515*** 1.6532*** 1.6559***
(5.783) (5.902) (5.911) (5.902) (5.908) (5.920)
Probability2-1.1598*** -1.1892*** -1.1909*** -1.1893*** -1.1909*** -1.1936***
(-4.292) (-4.420) (-4.426) (-4.419) (-4.425) (-4.436)
Divisional rival 0.0651*** 0.0652*** 0.0651*** 0.0651*** 0.0651*** 0.0651***
(11.745) (11.815) (11.794) (11.803) (11.803) (11.805)
Interleague 0.0722*** 0.0737*** 0.0735*** 0.0738*** 0.0738*** 0.0739***
(6.710) (6.903) (6.891) (6.909) (6.913) (6.923)
Opening day 0.3532*** 0.3538*** 0.3537*** 0.3537*** 0.3536*** 0.3537***
(9.603) (9.707) (9.654) (9.656) (9.654) (9.652)
Evening game 0.4107*** 0.4122*** 0.4123*** 0.4122*** 0.4122*** 0.4122***
(41.313) (41.502) (41.504) (41.492) (41.492) (41.483)
Length of broadcast 0.0017*** 0.0017*** 0.0017*** 0.0017*** 0.0017*** 0.0017***
(17.185) (17.366) (17.342) (17.360) (17.348) (17.377)
In-market NFL game -0.4713*** -0.4727*** -0.4730*** -0.4727*** -0.4727*** -0.4726***
(-17.157) (-17.343) (-17.356) (-17.341) (-17.343) (-17.337)
In-market NBA game -0.0725*** -0.0727*** -0.0733*** -0.0730*** -0.0731*** -0.0731***
(-5.701) (-5.739) (-5.786) (-5.764) (-5.770) (-5.769)
In-market NHL game -0.0378*** -0.0354*** -0.0357*** -0.0356*** -0.0356*** -0.0356***
(-2.769) (-2.601) (-2.616) (-2.609) (-2.606) (-2.607)
15
Table 3-Continued: Robustness Checks
ln(Local Television Audience) (1) (2) (3) (4) (5) (6)
Day of Week FE Yes*** Yes*** Yes*** Yes*** Yes*** Yes***
Month FE Yes*** Yes*** Yes*** Yes*** Yes*** Yes***
Year FE Yes*** Yes*** Yes*** Yes*** Yes*** Yes***
Local Team FE Yes*** Yes*** Yes*** Yes*** Yes*** Yes***
Opponent FE Yes*** Yes*** Yes*** Yes*** Yes*** Yes***
R20.8053 0.8052 0.8052 0.8052 0.8052 0.8052
Observations 29,409 29,648 29,648 29,648 29,648 29,648
Notes:
[1] ***p<0.01, **p<0.05, *p<0.10.
[2] Model estimated with robust standard errors; t-statistics shown in parenthesis.
[3] Definitions of each variable can be found in Table A2.
50 days prior to the PED announcement. Despite the concern, there is no evidence to support
this reverse causality hypothesis: the coefficient of the window variable is insignificant and all
other results remain unchanged. This remains true regardless of the window length considered,
such as 10, 15, or 30 days.
Column 3 addresses any remaining concern of correlation between team quality and PED
announcement by directly controlling for the quality of the PED player. This specification uses
a performance metric named Wins Above Replacement (WAR) which illustrates how important
the PED player is to his respective team prior to suspension. Values of WAR are based on an
individual’s performance relative to an entry-level player where positive values of WAR indicate
greater individual performance.14 If consumers react not to the news of the PED announcement
but to the quality of the suspended PED player, the coefficient on the WAR variable would be
negative and significant (in turn likely causing the coefficient of the playing-season suspension
variable to be insignificant). Instead, the WAR coefficient is found to be statistically no different
than zero and estimates of the playing-season suspension variables are unchanged. Although
not shown, similar results are found when considering the salary of the PED player.
Column 4 adds a control for PED events of players who, at the time of announcement, were
members of a Minor League Baseball (MiLB) team.15 Among these events, only those featuring
MiLB players with some MLB experience (either prior to or post-suspension) are considered -
see Table A4 for this list of MiLB PED suspension events. However, no significant effect is found
from a MiLB PED suspension announcement. While this may be unexpected, it is not all that
entirely surprising as a similar result has already been illustrated with the spill-over effect, or
14See MLB, ”Wins Above Replacement (WAR),” http://m.mlb.com/glossary/advanced-stats/wins-above-
replacement.
15Note MiLB teams are usually (although not always) affiliated with a MLB team and consist of individuals
vying for positions on said MLB team.
16
lack thereof, with PED events of the local team’s opponent.16 Related to the result that local
consumers do not respond to their opponent’s transgressions, fans may not be aware of the local
team’s MiLB PED suspension (MiLB teams are often in very different DMAs than its MLB
affiliate17) or do not equate support for the local MLB team with support for the associated
MiLB team. Note that while various specifications of a decay effect were explored, none were
found to be significant.
Several PED players within sample also have a second PED suspension.18 By adding a
dummy for a PED player’s second suspension, Column 5 explores the consumer’s reaction to
recidivism. This estimated effect could plausibly have either sign: a negative value indicating
a stronger condemnation for repeated indiscretions; a positive value - possibly even negating
the playing-season suspension effect - indicating apathy and indifference to a known PED user.
Instead, results suggest consumers have the same reaction to repeat offences.
The final column of Table 3 tests for the consumer reaction to the end of the PED suspension.
Similar to Column 2, the model adds a dummy variable for the post-suspension window of 50
days after the PED player’s reinstatement. Oddly enough, consumers do not undergo a second
round of (temporary) boycotts when the PED player returns to the team. This is an interesting
find, although it is not entirely inconsistent with the conjecture that consumers may not recall a
player is suspended (the hypothesised reason there exists the decay effect given a playing season
suspension). These results hold after varying the reinstatement window length and considering
decay.
Various forms of standard error estimators are considered in order to address any remaining
concern of correlation of the error term that may lead to overstating the impact of the results.
As mentioned in Section 5, all specifications report heteroskedasticity-consistent standard errors,
however there remains alternative options to explore. This includes clustering standard errors
by opponent-year or opponent-series.19 This also includes bootstrap and jackknife standard-
error estimation techniques. In each case, the significance of the results is unchanged and the
inference thereof remains the same.
Other specifications are explored - such as unilaterally removing each PED suspension event
from the estimation sample or considering alternative definitions of in-market substitute sports
- all with similar results. One such specification asks if fans react stronger to PED suspension
16See Column 4 of Table 2
17In fact, unless otherwise granted permission, no MiLB team may play its home games within the home
territory of any MLB team - see MLB Rule 52(a)(4), https://registration.mlbpa.org/pdf/MajorLeagueRules.pdf.
18Alfonzo, Mota, Perez, and Ramirez - see Table A4.
19Typically, a home-away pair play in two to four consecutive games known as a ‘series.’
17
announcements when the team is experiencing great success - a reaction perhaps fueled by the
belief their team’s performance was driven by PED use. This is done by splitting events by
the prominence of the PED player’s team using the predicted season wins at the time of the
suspension event. Results suggest fans have equal reactions regardless of the team’s performance,
however local fans may not be the correct reference group to investigate this concern. Instead,
one may want to consider the viewership patterns of individuals with no obvious allegiance to
one team, such as consumers outside of all DMAs of MLB teams or consumers within DMAs
with multiple MLB teams, and observe any substitution away from the PED player’s team or
away from the sport entirely. This is left for further research.
Another item for future research would be investigate alternative definitions of the control
group, i.e. injury spells. While there appears to be no effect on demand from injury announce-
ments of PED players, the average length of these events is not necessarily equal to that of the
PED suspensions. Other events worthy of exploration could be consumer reaction to other types
of suspensions. For example, in 2015, MLB introduced a formal policy on domestic violence that
has already brought about ten suspensions against MLB players ranging in length from 15 to
100 games.20
Lastly, although the sample covers seven years of MLB television viewership, there are con-
siderations to make for external validation. Recall that the entirety of the sample occurs under
one punishment regime; however, the punishment regime has since been toughened and bans
for the same offence are now much longer.21 Assuming the rate of PED suspensions since
the end of the sample has not declined dramatically, there would necessarily be a higher fre-
quency of suspension-treated game-broadcasts in the period since 2013. Within sample, 2% of
game-broadcasts carry a PED treatment and these relatively rare events may provide large infor-
mational value to consumers. However, it is not entirely obvious that more suspension-treated
game-broadcasts would have the same average effect. Future research should explore the PED
suspension announcement across the various punishment regimes to test for time-dependent
recidivism and consumer habituation to PED announcements.
20See MLB-MLBPA Joint Domestic Violence, Sexual Assault, and Child Abuse Policy,
http://riveraveblues.com/wp-content/uploads/2015/08/Domestic-Violence-Policy.pdf; see also ProSport-
sTransactions.com.
21As of 2014, the PED punishments are 80 games for a first infraction, 162 games for a second, and lifetime for
a third - see MLB Joint Drug Prevention and Treatment Program, http://mlb.mlb.com/pa/pdf/jda.pdf.
18
5.3 Protecting the financial integrity of baseball
While clandestine use of PEDs may actually be a benefit to the commercial viability of sports,
league organisers may still have strong economic incentive to enforce anti-PED regulations (Pre-
ston and Szymanski, 2003). Because public news of PED use is generally not well received, one
such incentive would be to stop the outflow of consumer support and protect the sport’s financial
integrity (Mazanov and Connor, 2010). For example, as illustrated by Cisyk and Courty (2017),
news of a PED announcement results in a loss of close to $743,000 in forgone ticket revenue to
the PED player’s team.
However, the decline in television audience does not necessarily provide the same incentive:
consider that each MLB team can bargain with a local television station for the rights to locally
broadcast each of its regular-season games. Typically, a team enters into a large contract with
often a single television station for rights to broadcast its games. The dollar value of its contract
is primarily predicated on the number of viewers the local television station expects to receive
and the value of advertisement it can sell during the game-broadcasts.
Information on each team’s television contract can be found in Table A5. There are two
important features to note of each television contract, the first of which is the team’s ownership
share of the local television station. While the average share is 22%, 13 teams have zero owner-
ship of the local television station and, therefore, have little incentive in preserving the financial
integrity of the game-broadcasts for the remainder of its current contract.
Second, television contracts are often very lengthy - the largest observed value in Table A5 is
30 years. In the short-run, these contracts become fixed costs for the television station regardless
of realised advertisement revenue. Bargaining for a new television contract may include updated
information on the financial integrity of the team but, as already found by Brave and Roberts
(2019), teams may forgo long-term profits for short-terms gains from PED use. Thus, such
lengthy contracts suggest that a team need not have a large discount rate to greatly discount a
decline in future television contract revenue.
6 Summary and Conclusion
This study finds when a PED suspension is announced, the PED player’s team on average expe-
riences an 9.3% decline in television audience in the subsequent game-broadcast. The magnitude
of this negative response to the announcement begins to wane over time yet is statistically sig-
19
nificant for at least 37 days. This study also finds there is no effect on television audience when
the same players are removed from their teams due to injuries suggesting that consumers are
responding solely to the PED announcement and not to the change in the talent featured in the
game-broadcast.
This study also finds that while there is a strong localised response to PED suspension, there
is no evidence of a collective response, i.e., the PED player’s opponent does not see a decline in
its own television audience. A similar finding that consumers do not react to PED suspensions
of the local team’s affiliated MiLB players suggest news of PED transgressions does not travel
far outside the local team’s DMA or that fans do not equate support for their local team with
support for the opponent and/or MiLB team.
Further robustness checks are considered such as PED player performance controls, recidi-
vism, and reaction to the PED player’s reinstatement from suspension: all return similar results.
Other specifications raise important questions for further research such as the reaction to a PED
suspensions from fans who have other MLB options available or consumer reaction in light of
suspensions stemming from the relatively new MLB policy on domestic violence.
The findings of this study are consistent with the hypothesis that consumers do care about
PED use in sports. While PEDs are said to be a potential benefit (greater chance to see an
exceptional athletic performance) as well as a potential hindrance (loss of consumer support) to
the financial integrity of sports, this study certainly points to the latter, although it has been
illustrated that teams may not fully internalise the cost of its own player’s PED suspension.
20
7 Appendix
References
Borland, J., and R. MacDonald. 2003. Demand for sport. Oxford review of economic policy
19:478–502.
Brave, S. A., and K. A. Roberts. 2019. The Competitive Effects of Performance-Enhancing
Drugs: MLB in the Posttesting Era. Journal of Sports Economics 20:747–781.
Brenner, P. S., and J. DeLamater. 2016. Lies, damned lies, and survey self-reports? Identity as
a cause of measurement bias. Social psychology quarterly 79:333–354.
Buechel, B., E. Emrich, and S. Pohlkamp. 2016. Nobodys innocent: the role of customers in the
doping dilemma. Journal of Sports Economics 17:767–789.
Buraimo, B., and R. Simmons. 2015. Uncertainty of outcome or star quality? Television audience
demand for English Premier League football. International Journal of the Economics of
Business 22:449–469.
Cisyk, J., and P. Courty. 2017. Do fans care about compliance to doping regulations in sports?
The impact of PED suspension in baseball. Journal of Sports Economics 18:323–350.
Eber, N. 2008. The performance-enhancing drug game reconsidered: A fair play approach.
Journal of sports economics 9:318–327.
Engelberg, T., S. Moston, and J. Skinner. 2012. Public perception of sport anti-doping policy
in Australia. Drugs: education, prevention and policy 19:84–87.
Haugen, K. K. 2004. The performance-enhancing drug game. Journal of sports economics
5:67–86.
Maennig, W. 2002. On the economics of doping and corruption in international sports. Journal
of Sports Economics 3:61–89.
Maennig, W. 2014. Inefficiency of the anti-doping system: Cost reduction proposals. Substance
use & misuse 49:1201–1205.
Maennig, W., et al. 2009. Pecuniary disincentives in the anti-doping fight. Economic Analysis
and Policy 39:349.
21
Mazanov, J., and J. Connor. 2010. Rethinking the management of drugs in sport. International
journal of sport policy 2:49–63.
Noll, R. 1974. Attendance and Price Setting. Government and the Sports Business. RG Noll.
Washington, DC. The Brookings Institution .
Preston, I., and S. Szymanski. 2003. Cheating in contests. Oxford review of economic policy
19:612–624.
Rascher, D. 1999. A test of the optimal positive production network externality in Major League
Baseball. Sports Economics: Current Research.
Solberg, H. A., D. V. Hanstad, and T. A. Thøring. 2010. Doping in elite sport-do the fans
care? Public opinion on the consequences of doping scandals. International Journal of Sports
Marketing and Sponsorship 11:2–16.
ˇ
Strumbelj, E. 2016. A comment on the bias of probabilities derived from betting odds and their
use in measuring outcome uncertainty. Journal of sports economics 17:12–26.
Tainsky, S. 2010. Television broadcast demand for National Football League contests. Journal
of Sports Economics 11:629–640.
Van Reeth, D. 2013. TV demand for the Tour de France: The importance of stage characteristics
versus outcome uncertainty, patriotism, and doping. International Journal of Sport Finance
8:39.
Villar, J. G., P. R. Guerrero, et al. 2009. Sports attendance: a survey of the literature 1973-2007.
Rivista di Diritto e di Economia del lo Sport 5:112–151.
22
Table A1: PED Suspension Punishments, 2006 to 2013
Length of Suspension
Positive Tests Non-Stimulant PEDs Stimulants
1st 50 games Follow-up Testing
2nd 100 games 25 games
3rd Up to lifetime 80 games
4th - Up to lifetime
Source: Major League Baseball’s Joint Drug Prevention and Treatment Program.
Table A2: Variable Descriptions
Variable Name Description
Playing-season Takes a value of 1 when a PED player from the team is currently serving a suspension
suspension announced during the regular-season, 0 otherwise. Time elapsed refers to the number
of days since the announcement of the suspension.
Off-season Takes a value of 1 when a PED player from the team is currently serving a suspension
suspension announced outside the regular-season, 0 otherwise.
Inactive Takes a value of 1 when a PED player from the team is inactive due to an injury, 0
otherwise. Time elapsed refers to the number of days since the PED player was placed
on the disabled list.
Local team is home team Takes a value of 1 when the local team is the home team, 0 otherwise.
Predicted season wins The expected number of total season wins of the local team.
Probability of winning game The expected probability the local team wins the game.
Probability2The squared value of probability of winning game.
Divisional rival Takes a value of 1 when the local team and non-local team are within the same
division of the same league, 0 otherwise.
Interleague Takes a value of 1 when the local team and non-local team are not within the same
league, 0 otherwise.
Opening day Takes a value of 1 for the first game-broadcast in each season where the local team
is the home team, 0 otherwise.
Evening game Takes a value of 1 when the game is played after 18:00, local time.
Length of broadcast Length of the game-broadcast, measured in minutes.
In-market NFL game Takes a value of 1 when the game-broadcast occurs on the same day as a game of
an NFL team in the local team’s DMA.
In-market NBA game Takes a value of 1 when the game-broadcast occurs on the same day as a game of
an NBA team in the local team’s DMA.
In-market NHL game Takes a value of 1 when the game-broadcast occurs on the same day as a game of
an NHL team in the local team’s DMA.
23
Table A3: Available Substitutes in Each DMA, 2006 to 2012
DMA MLB NFL NHL NBA
Atlanta, GA Atlanta Braves Atlanta Falcons Atlanta Thrashers Atlanta Hawks
Baltimore, MD Baltimore Orioles Baltimore Ravens
Boston-Manchester, MA Boston Red Sox New England Patriots Boston Bruins Boston Celtics
Chicago, IL Chicago Cubs, Chicago Bears Chicago Blackhawks Chicago Bulls
Chicago White Sox
Cincinnati, OH Cincinnati Reds Cincinnati Bengals
Cleveland-Akron-Canton, OH Cleveland Cleveland Browns Cleveland Cavaliers
Dallas-Ft. Worth, TX Texas Rangers Dallas Cowboys Dallas Stars Dallas Mavericks
Denver, CO Colorado Rockies Denver Broncos Colorado Avalanche Denver Nuggets
Detroit, MI Detroit Tigers Detroit Lions Detroit Red Wings Detroit Pistons
Houston, TX Houston Astros Houston Texans Houston Rockets
Kansas City, KS-MO Kansas City Royals Kansas City Chiefs
Los Angeles, CA Los Angeles Angels of Anaheim Los Angeles Kings, Los Angeles Clippers
Los Angeles Dodgers Anaheim Ducks Los Angeles Lakers
Miami-Ft. Lauderdale, FL Miami (Florida) Marlins Miami Dolphins Florida Panthers Miami Heat
Milwaukee, WI Milwaukee Brewers Green Bay Packers Milwaukee Bucks
Minneapolis-St. Paul, MN Minnesota Twins Minnesota Vikings Minnesota Wild Minnesota Timberwolves
New York, NY
New York Mets New York Giants New York Islanders New York Knicks,
New York Yankees New York Jets New York Rangers Brooklyn (New Jersey) Nets
New Jersey Devils
Philadelphia, PA Philadelphia Phillies Philadelphia Eagles Philadelphia Flyers Philadelphia 76ers
Phoenix-Prescott, AZ Arizona Diamondbacks Arizona Cardinals Arizona Coyotes Phoenix Suns
Pittsburgh, PA Pittsburgh Pirates Pittsburgh Steelers Pittsburgh Penguins
San Diego, CA San Diego Padres San Diego Chargers
San Francisco-Oakland- Oakland A’s, Oakland Raiders, San Jose Sharks Golden State Warriors
San Jose, CA San Francisco Giants San Francisco 49ers
Seattle-Tacoma, WA Seattle Mariners Seattle Seahawks Seattle SuperSonics
St. Louis, MO St. Louis Cardinals St. Louis Rams St. Louis Blues
Tampa-St. Petersburg-Sarasota, FL Tampa Bay (Devil) Rays Tampa Bay Buccaneers Tampa Bay Lightning
Washington-Hagerstown, DC-MD Washington Nationals Washington Washington Capitals Washington Wizards
24
Table A4: Suspension and Injury Events, 2006 to 2012
Playing-Season Events Suspension Injury
Player PositionaTeam Date LengthbTeam Date Lengthb
Yusaku Iriki P NYM 4/28/2006 50
Jason Grimsley RP ARI 6/12/2006 50
Juan Salas RP TBR 5/7/2007 50
Neifi Perez (1) IF DET 7/6/2007 25
Neifi Perez (2) IF DET 8/3/2007 80
Manny Ramirez (1) OF LAD 5/7/2009 50 LAD 4/23/2010 14
Edinson Volquez SP CIN 4/20/2010 50 CIN 5/21/2009 10
Manny Ramirez (2) OF TBR 4/8/2011 100 LAD 7/3/2010 9
Eliezer Alfonzo (2) C COL 9/14/2011 48cSFG 6/9/2007 8
Guillermo Mota (2) RP SFG 5/7/2012 100 SFG 8/23/2010 13
Freddy Galvis IF PHI 6/19/2012 50 PHI 8/17/2012 43
Melky Cabrera OF SFG 8/15/2012 50
Bartolo Colon SP OAK 8/22/2012 50 OAK 6/23/2012 10
Off-Season Events Suspension
Player PositionaTeam Date Lengthb
Carlos AlmanzardRP TEX 10/4/2005 10h
Guillermo Mota (1) RP NYM 11/1/2006 50
Mike CameroneOF SDP 10/31/2007 25
Dan Serafini RP COL 11/27/2007 50
Jose Guillen OF KCR 12/6/2007 15
J.C. Romero RP PHI 1/6/2009 50
Minor League Events Suspension
Player PositionaTeam Date Lengthb
Felix HerediafRP NYM 10/18/2005 10h
Ramon Ramirez P CIN 4/11/2006 50
Nerio Rodriguez P PIT 5/19/2006 50
Abraham Nunez OF SFG 5/24/2006 50
Yamid Haad C SFG 5/31/2006 50
Daniel McCutchen P NYY 8/8/2006 50
Francisco Cruceta P TEX 5/9/2007 50
Lino Urdaneta P NYM 5/16/2007 50
Angel Salome C MIL 7/24/2007 50
Ryan Jorgensen C CIN 9/7/2007 50
Luther Hackman P TEX 10/30/2007 50
Jordan Schafer OF ATL 4/8/2008 50
Eliezer Alfonzo (1) C SFG 4/30/2008 50
Humberto Cota C COL 5/28/2008 50
Jorge Sosa P SEA 8/21/2008 50
Runelvys Hernandez P HOU 9/6/2008 50
Henry Owens P MIA 11/11/2008 50
Sergio MitregP MIA 1/6/2009 50
Pablo Ozuna IF PHI 6/11/2009 50
Prentice Redman (1) OF LAD 6/25/2010 50
Pablo Lopez IF WSN 7/27/2010 50
Prentice Redman (2) OF LAD 7/27/2010 100
Omar Quintanilla IF COL 8/11/2010 50
Matt Kinney P SFG 8/24/2010 50
Kevin Frandsen IF PHI 5/11/2011 50
Mark Rogers SP MIL 8/19/2011 25
25
Table A4-Continued: Suspension and Injury Events, 2006 to 2012
Events Not Considered Suspension
Player PositionaTeam Date LengthbReasoning
Matt Lawton OF NYY 11/2/2005 10hReleased before serving
suspension in 2006.
Jay Gibbons OF BAL 12/6/2007 15 Released before serving
suspension in 2008.
Marlon Byrd OF - 6/25/2012 50 Free agent at time of suspension.
Notes:
aC = Catcher; IF = Infielder; OF = Outfielder; SP = Starting Pitcher; RF = Relief Pitcher; P = Pitcher
(Relief/Starting unspecified).
bMeasured in number of games, unless noted elsewhere.
c100-game suspension was reduced to 48 due to procedural issues with test samples.
dServed suspension with Atlanta Braves at start of 2006 season.
eServed suspension with Milwaukee Brewers at start of 2008 season.
fServed suspension with Cleveland’s minor league affiliate at start of 2006 season.
gServed suspension with New York Yankees’ minor league affiliate at start of 2010 season.
hSuspension length measured in days.
Source: ProSportsTransactions.com
Table A5: Television Contract Information, 2016
Annual Average
Value Length Ownership Market Size Audience
($ Millions) (Years) (%) (Thousands) (Thousands)
Los Angeles Dodgers 334.0 25 100 5,613.5 117.3
New York Yankees 190.0 30 20 7,384.3 293.5
Los Angeles Angels 150.0 20 25 5,613.5 62.6
Seattle Mariners 100.0 18 71 1,818.9 64.1
Philadelphia Phillies 100.0 25 25 2,949.3 170.4
Boston Red Sox 80.0 N/A 80 2,366.7 145.3
Houston Astros 80.0 20 0 2,215.7 20.6
Texas Rangers 80.0 20 10 2,588.0 154.1
Arizona Diamondbacks 75.0 20 N/A 1,812.0 63.0
San Francisco Giants 70.0 25 30 2,502.0 126.2
St. Louis Cardinals 66.7 15 30 1,243.5 95.2
Chicago Cubs 65.0 16 20 3,484.8 75.0
New York Mets 52.0 25 65 7,384.3 170.6
Chicago White Sox 51.0 16 20 3,484.8 75.8
Detroit Tigers 50.0 10 0 1,845.9 167.0
San Diego Padres 50.0 20 20 1,075.1 20.0
Oakland A’s 47.6 21 0 2,502.0 32.8
Washington Nationals 46.0 23 18 2,359.2 57.6
Baltimore Orioles 46.0 23 82 1,085.1 52.6
Cleveland 40.0 10 0 1,485.1 64.7
Minnesota Twins 40.0 12 0 1,728.1 77.5
Atlanta Braves 35.0 20 0 2,326.8 83.7
Cincinatti Reds 30.0 10 0 897.9 75.4
Pittsburgh Pirates 25.0 10 0 1,165.7 72.7
Milwaukee Brewers 24.0 7 0 902.2 44.9
Kansas City Royals 20.0 12 0 931.3 35.1
Colorado Rockies 20.0 10 0 1,566.5 42.3
Tampa Bay Rays 20.0 10 0 1,806.6 87.0
Miami Marlins 18.0 15 0 1,621.1 31.6
Note: N/A = Not Available.
Sources: FanGraphs.com; Nielsen.
26
