Economics

Article Abstract 3
Instructions
Your critical abstracts should run to no more than 1 page. They should include the following:
1. Title and author(s)
2. The research question that the author wants to address
3. Main conclusions reached
4. How can this information be applied to the sports market?
5. What did the author(s) miss when evaluating their research question
Your abstract this week should come from one of the following articles:
• Winfree, J, JJ McCluskey, and RC Mittelhammer (2004). “Location and attendance in Major League Baseball”. In: Applied Economics 35, pp. 2117–2124.
I have the article attached for you.

Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=raec20
Download by: [Washington State University Libraries ] Date: 23 January 2016, At: 18:25
Applied Economics
ISSN: 0003-6846 (Print) 1466-4283 (Online) Journal homepage: http://www.tandfonline.com/loi/raec20
Location and attendance in major league baseball
Jason A. Winfree , Jill J. McCluskey , Ron C. Mittelhammer & Rodney Fort
To cite this article: Jason A. Winfree , Jill J. McCluskey , Ron C. Mittelhammer & Rodney
Fort (2004) Location and attendance in major league baseball, Applied Economics, 36:19,
2117-2124, DOI: 10.1080/0003684042000287664
To link to this article: http://dx.doi.org/10.1080/0003684042000287664
Published online: 02 Feb 2007.
Submit your article to this journal
Article views: 373
View related articles
Citing articles: 22 View citing articles
Location and attendance in major league
baseball
JASON A. WINFREE*, JILL J. MCCLUSKEYz,
RON C. MITTELHAMMERz and RODNEY FORTz
Program in Sports Management, University of Michigan, 401 Washtenaw Ave,
Ann Arbor, MI 48109–2214, USA and zSchool of Economic Sciences, Washington
State University, Pullman, Washington, USA
This study uses a travel-cost model to analyse the attendance impacts on Major
League Baseball (MLB) of the closest substitute MLB team. It is found that the
closer two teams are, the lower attendance is at each team relative to two teams that
are farther apart. In addition, when a new team moves into the area of an existing
team, there is an additional initial reduction in attendance for the incumbent team.
This has implications for actions aimed at changing the number of teams in MLB
either by contraction or by possible antitrust approaches that would increase the
number of teams, especially in megalopolis markets. Further, and consistent with
past demand studies, pricing is in the inelastic portion of gate demand and fan
loyalty is a significant contributor to the estimation of gate attendance.
I. INTRODUCTION
It is said that there are three important considerations
in business: location, location, and location. In sports economics,
distance between teams is a very important consideration
for professional sports leagues (and, although
not covered in this paper, college sports conferences). In
1968, when the Kansas City Athletics moved to Oakland,
the attendance of the San Francisco Giants dropped over
32%. The Colorado Rockies have no other Major League
Baseball teams within a 600-mile radius and enjoy large
crowds for every home game. A systematic assessment of
this type of attendance impact is the aim of this paper.
Currently, Major League Baseball (MLB) argues that
many owners face uncertain financial futures due to competitive
imbalance problems. One of the recommendations
made by Levin et al. (2000) is that MLB should allow
owners to move their teams, especially to larger revenue
markets, to reduce the imbalance. Incumbent owners have
made it clear that new teams in their market area are
substitutes that will reduce attendance at the incumbent
team. Indeed, this sensitivity to attendance impacts helps
explain the limited number of teams in so-called ‘megalopolis’
markets.
However, measuring attendance impacts also is important
from a policy perspective. Another approach to their
alleged revenue difficulties has MLB owners and their
Commissioner seriously considering a contraction of the
league by up to two teams. The policy question concerns
fan welfare after contractions. If increases in attendance for
teams proximate to those that are eliminated do not impact
league-wide competitive balance in any meaningful way,
then contraction will make a few owners better off with
primarily negative impacts on fan welfare (for those fans
losing their teams).
Another example of policy relevance concerns a suggestion
originating with Roger Noll and Ira Horowitz in
Congressional Hearings (cited in Fort, 2001, and extended
by Quirk and Fort, 1999). The original Noll-Horowitz
idea is that antitrust laws can be used to create competing
* Corresponding author. E-mail: jwinfree@umich.edu
Applied Economics ISSN 0003–6846 print/ISSN 1466–4283 online # 2004 Taylor & Francis Ltd 2117
http://www.tandf.co.uk/journals
DOI: 10.1080/0003684042000287664
Applied Economics, 2004, 36, 2117–2124
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
leagues in the hope that entry in megalopolis markets would
reduce revenue dispersion and competitive imbalance. The
work here suggests that only careful, case-by-case investigation
will reveal whether attendance impacts are large
enough to reduce revenue imbalance significantly in a
given league.
Demand and support for baseball teams are not only
important to baseball owners and fans, but to state and
local governments. This study also has implications for
taxpayers, who have paid large sums for stadiums and
team subsidies. The public has paid about 64% of professional
sports facilities, which accounts for almost $6 billion
from 1999 to 2003 (Fort, 2002). Local politicians count on
high demand for these professional sports. This study finds
that a given MLB team’s attendance is affected by its
substitute’s location and stadium. An implication of this
finding is that investing in a new stadium might not be a net
gain for the community if it detracts from another team in
the same metropolitan area. For example, it is estimated
that the public will pay $390 million for a new stadium
for the New York Mets that should be completed in
2004. However, if the Mets do not build a new stadium,
many fans may choose to attend Yankee games instead.
It is important to consider the effect of these stadiumfunding
decisions on both the team in question and its
closest substitute.
The principal importance of the attendance demand
analysis in this paper is its examination of how team attendance
is affected by neighbouring teams or the appearance
of a new team close by. Theoretically, the closer the team,
the more fans will be lost to the substitute. In addition,
since one is estimating the demand for attendance at the
gate, there will be findings relevant to past demand observations,
mainly concerning the elasticity of gate pricing and
fan loyalty. At the level of annual observations on teams
used in this paper, one cannot contribute to the analysis of
the uncertainty of outcome hypothesis.
The paper is organized as follows. First, previous studies
of location and baseball demand are briefly reviewed.
Second, the travel-cost model is presented. Third, the
empirical specification is developed and the data are
described. Fourth, the empirical results are presented and
discussed. Conclusions, with applications of the findings
to the policy issues of contraction and competition policy,
complete the study.
II. PREVIOUS STUDIES
A principal tenet of Hotelling’s (1929) location model is
that, all else constant, people will choose to purchase
goods or services from the closest seller. A travel-cost
model (McConnell, 1985) is used to apply Hotelling’s
idea to MLB attendance. The main cost for many baseball
fans is travel cost. Consequently, if there are two teams
that provide essentially the same utility for certain fans,
all else constant, they will attend the game that is closest.
This implies that attendance lost to a substitute team
should be a function of the distance to that substitute.
To the authors’ knowledge, the current study is the first
to use a travel-cost model to explain attendance for professional
sporting events, and no studies examine directly how
baseball attendance is related to substitute possibilities.
Previous demand studies have generated engaging
hypotheses and some interesting findings. One is the recurring
empirical observation that owners set ticket prices in
the inelastic portion of gate demand. Fort (2004) shows a
variety of circumstances where this choice is consistent with
profit maximization and that the conditions hold for MLB
in the late 1980s. It will be of interest to see if this also is
true on the current data.
Another interesting hypothesis is that fan loyalty impacts
attendance. Kahane and Shmanske (1997) show that roster
stability engenders loyalty (higher attendance is observed
for stable rosters, all else constant). Depken (2001) uses
stochastic frontier estimation and interprets teams closest
to the attendance frontier as having the most loyal fans.
Under this interpretation, he finds teams that relocated had
low fan loyalty. It will be noted how the work contributes
to both the inelastic pricing and loyalty issues.
III. THE MODEL
The utility of a baseball fan is a function of the number
of visits to baseball stadiums, the quality of the teams,
the quality of those stadiums, as well as all other goods
consumed. Of course, a fan’s utility can be different for
different teams. It is also reasonable to assume that baseball
fans may desire a variety of stadiums and teams, so
that the home games of all accessible stadiums and teams
are included in the utility function. Fan i faces the following
utility maximization problem:
MaxUiðXi,Vi1, . . . ,Vim, q1, . . . , qmÞ, ð1Þ
subject to a monetary constraint:
Mi þ witwi
¼ Xi þX
m
j¼1
cijVij , ð2Þ
and a time constraint:
t
i ¼ twi
þX
m
j¼1
ðttij
þ tvÞVij , ð3Þ
where Xi is the quantity of the numeraire, Vij is the number
of visits by the ith individual to the jth stadium, qj is the
quality of going to stadium j, Mi is the exogenous income
of individual i, wi is the wage rate of individual i, twi is the
hours worked by individual i, cij is the monetary cost of
going to the jth stadium for the ith individual, t
i is the total
2118 J. Winfree et al.
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
discretionary time for individual i, ttij is the total travel time
for the ith individual going to stadium j, tv is the time spent
during the visit and assumed constant, and m is the number
of major league baseball stadiums.
It is assumed that the number of visits and quality are
complements if they are for the same team (e.g. a baseball
game is more enjoyable if one is familiar with the players),
while everything else in the utility function are substitutes.
For the large majority of fans who go to see the home
team, the quality variable, q, would be a function of both
the quality of the home team and the quality of the stadium.
It is expected that this quality is effected by variables
such as winning percentage, pennant chances, and the age
of the stadium. The model also assumes that the individual
chooses the amount of time that he or she works at a
constant wage rate. Combining the constraints yields the
following relationship:
Mi þ wit
i ¼ Xi þX
m
j¼1
ðcij þ wiðttij
þ tvÞÞVij , ð4Þ
or
Mi þ wit
i ¼ Xi þX
m
j¼1
ðptj
þ pcj
þ pdidij þ wiðttij
þ tvÞÞVij :
ð5Þ
The monetary cost may also be disaggregated into ticket
price, ptj , price of concessions, pcj , and travel cost per mile
multiplied by the round trip mileage, pdidij . To find the
total cost of the visit, one must consider the monetary
cost and the opportunity cost per visit, which is equal to
the wage rate multiplied by the sum of the travel time and
time spent at the game.
This model implies that if wages or exogenous income
increase, the number of visits should increase. However,
if any of the prices increase, one would expect the number
of visits to decrease. Since a greater distance would increase
the price for the consumer, an increase in distance should
decrease visits. Therefore, if another team moves close by it
should increase visits to the new team, possibly decreasing
visits to the old team.
Maximizing utility (1) subject to constraint (5) yields a
demand equation for the visits to stadium j by individual i:
Vij ¼ Vijðqj ; pij;Mi; dij;ZiÞ; ð6Þ
where demand is a function of quality of the site, prices,
income, and distance to the site among other variables
represented generically by Zi. If quality increases, then so
should the number of visits. Then the aggregate attendance
demand for the team is the sum of all individual demand
equations, as:
Vj ¼X
n
i¼1
Vijðqj; pij ;Mi; dij;ZiÞ: ð7Þ
IV. EMPIRICAL SPECIFICATION AND
THE DATA
The empirical specification of the model in (7) is given
below (descriptive statistics for all variables are presented
in Table 1):
ATTENDANCE
¼0þ1DISTANCEþ2NEWTEAMþ3PRICE
þ4INCOMEþ5POPULATIONþ6WIN%
þ7CHANGEWIN%þ8RUNS9 þ10DIVCHAMP
þ11LOYALTYþ12STADIUMþ13STRIKE
þ14ATTENDGROWTHþ15TREND
þ16ANGELSþ17BLUEJAYSþ18CUBS
þ19DODGERSþ20CARDINALSþ”: ð8Þ
The dependent variable is annual ATTENDANCE, in
hundreds of thousands, for a particular team and ” is a
residual term. The observations cover teams over the years
1963 to 1998, with the number of teams varying from 20 in
1963 to 30 in 1998. The attendance data are for the regular
season, yielding 884 observations.
Two variables novel to this study capture the impacts of
distance between teams (the availability of very close substitutes).
DISTANCE equals the inverse of the distance in
miles to the nearest alternative MLB stadium. The value of
DISTANCE only changes over time for a team if a team
moved or a different team moved close by. For example,
when the Senators left Washington, DC at the end of the
1971 season to become the Texas Rangers, the Baltimore
Orioles DISTANCEvariable changed from 0.0222 to 0.0097
Table 1. Summary statistics
Variable Min Max Mean SD
ATTENDANCE 3.0676 40.5795 16.4907 7.2151
DISTANCE 0.0012 0.1 0.0264 0.0337
NEW TEAM 0 1 0.0362 0.1869
PRICE 2.5728 12.2638 5.9819 1.0735
INCOME 7.0425 20.4951 13.4349 2.4054
POP 0.1317 7.8949 1.5573 1.9581
WIN% 0.3086 0.704 0.5016 0.0686
CHANGEWIN% 0.2349 0.2016 0.0013 0.0699
RUNS 3.29 9.93 6.7775 0.9897
DIVCHAMP 0 1 0.1448 0.3521
LOYALTY 3.0676 44.8335 16.1775 7.1543
STADIUM 0 1 0.0192 0.1374
STRIKE 0 44.8335 1.2813 5.4275
ATTENDGROWTH 0.4065 0.7214 0.0429 0.1637
TREND 1 1296 486.063 399.494
ANGELS 1 0.0407 0.1978
BLUEJAYS 0 1 0.0238 0.1524
CUBS 0 1 0.0407 0.1978
DODGERS 0 1 0.0407 0.1978
CARDINALS 0 1 0.0407 0.1978
Location and attendance in baseball 2119
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
since the Philadelphia Phillies became their closest substitute
MLB team. In this sample, the only simultaneous
single-stadium occupation was the Angels and Dodgers
in Dodger Stadium (1961–1965). The denominator of the
DISTANCE variable was constrained to 0.1 for this special
case, which represented the largest empirical value of the
DISTANCE variable.
NEWTEAM is the second variable used to capture the
impact of distance between teams. NEWTEAM equals one
for the first year a new team moves within 500 miles of an
existing stadium in a given year. A new team might generate
more excitement and be a better substitute for many
people than an existing substitute team. NEWTEAM could
capture an additional initial loss of attendance from league
expansion or relocation. This variable includes teams moving
and expansion years, which includes 1969, 1977, 1993,
and 1998.
Other demand parameters follow a typical demand function
specification (price, income, population and preferences).
PRICE denotes the own team’s price of baseball
tickets in real (regional CPI-adjusted) terms. Prices have
been obtained from four different sources – American
League Red Book (1963), Bruce Domazlicky (1969–1980),
Roger Noll (1975–1988), and the Society for American
Baseball Research (1991–1998). National League Green
Books were not used since older years do not have ticket
price data. Missing data were replaced by interpolation
using the historical rate of growth over the period 1969–
1988. Since there are a variety of seats with different prices
for each baseball game, averaging of seat prices is a difficult
issue to resolve. Coffin (1996) discusses the problems of
estimating an average price. Each data source used its
own system for averaging,1 and when different prices for
the same team and year were available, a simple mean
across all sources was used. Note that ticket prices that
overlapped were very similar in value.
INCOME denotes real (regional CPI-adjusted) statelevel
per capita personal income (in thousands of dollars).
Per capita personal income was obtained from the Bureau
of Economic Analysis website for all US states. Canadian
per capita personal income was used for Montreal and
Toronto. State level data is used to allow for the fact
that people attend baseball games from a variety of geographical
locations.
POP is the population of the city in which a stadium
resides (in millions). City populations are available
from the US Census Bureau website, <www.census.gov>
and Canadian city population was found at <www.demographia.
com>. For years between censuses, a constant
growth rate was applied to interpolate population. The
same method was used for Canadian cities; however,
the censuses were a year later. This variable is used as an
instrumental variable related to the size of market for
sports entertainment.
Preference controls are included for team performance
(winning, run production, championships), loyalty, stadium
quality and work stoppages. WIN% is the team’s regular
season winning percentage. In general, fans prefer to see
teams that win. CHANGEWIN% is the change in winning
percentage for the team from year to year. It is the difference
between the current year’s winning percentage and the
previous (lagged) year’s winning percentage. It is hypothesized
that both winning percentage and its changes
affect the quality of a fan’s experience. If a city has not
had a successful baseball team in quite some time, then
attendance can be relatively more exciting and enjoyable
when the team begins winning more games. However, if a
team wins consistently from year to year, fans may take
winning for granted and become disgruntled when a team
loses.
RUNS is the number of runs that were scored by a team
during a year. Fans have a tendency not only to prefer a
winning team, but also to prefer more rather than fewer
runs scored. However, it was not necessarily assumed that
runs affect attendance linearly. Therefore, in a manner akin
to a Box-Cox transformation, a nonlinear power function
of runs scored was used and was found to be more effective
in explaining attendance.
DIVCHAMP is an indicator variable that equals
one when a team wins a division championship. The
DIVCHAMP variable includes information not necessarily
captured in WIN% since a team can have a high winning
percentage but still not win its division. Prior to 1969, there
were no divisions, so only the two teams that won their
respective leagues are included as the DIVCHAMPs.
The LOYALTY variable is the lagged attendance for
the team (in hundreds of thousands) and is intended to
proxy the fans’ attendance habit persistence. This variable
is lagged one year. For example, Montreal does not seem to
be a ‘baseball town’ and suffers from very low attendance
that persists year in and year out, while attendance in other
cities seems to flourish regardless of how the team plays.
Lagged attendance proxies, both affects. The present version
of ‘loyalty’ varies from that used in other studies
(Kahane and Shmanske, 1997; Depken, 2001), providing
a different dimension for comparison with those works.
STADIUM is an indicator variable used to proxy for
the effect of a new stadium on attendance. New stadiums
tend to increase attendance dramatically, at least for a short
time (and possibly a longer period if new stadium revenues
translate into an increase in team quality). The effect may
be largely captured by the LOYALTY variable, so the
STADIUM variable is used only for the year in which
the new stadium is first opened.
1 Contact the authors for a detailed description of the averaging for each data source.
2120 J. Winfree et al.
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
STRIKE is an indicator variable equal to one multiplied
by the previous year’s attendance in years of work stoppages.
This specification assumes that, while strike effects
can vary across teams, they effectively are proportional to
typical team attendance. There were two years in the sample
when work stoppages had any significant consequence
on season length, namely the players’ strikes of 1981 and
1994 (Fort, 2002). Obviously, attendance will fall if a team
is not playing as many games. About 50 to 60 games were
cancelled for each team during the 1981 and 1994 strikes.
Additional controls also seemed appropriate. ATTENDGROWTH
controls for overall growth in attendance,
league wide. It is measured by the growth rate of league
attendance, omitting the particular team’s attendance being
measured by the ATTENDANCE observation. The variable
is used as a proxy measure for the overall popularity
of baseball. The TREND variable proxies secular shifts
that can occur from a number of different sources, such
as increases in population beyond city limits, increased
availability of leisure time, or shifts in preferences in
favour of people desiring to attend more baseball games.
Examining attendance over time graphically, there appeared
to be an underlying quadratic trend, and to investigate this
systematically, a quadratic time variable is used, equal to
the year squared, so that 1963 equals 1, 1964 equals 4, 1965
equals 9, and so on. The reason that including both the
TREND and ATTENDGROWTH variables is particularly
useful is that ATTENDGROWTH can account for
large and/or irregular increases or decreases in attendance
that TREND cannot. Individual team indicator variables
were also included. These represent fixed effects for these
particular teams that are not otherwise captured by the
model.
It is important to note that other demand variables
relating to the closest substitute may affect a team’s attendance.
For example, the price of a Yankees ticket may
affect the Mets’ attendance. The variables PRICE, WIN%,
CHANGEWIN%, RUNS, DIVCHAMP, and STADIUM
for the nearest team were all initially included in the demand
model and the variables were then weighted by the variables
interacting with the DISTANCE variable. However,
none of these variables was statistically significant, and
they were excluded from the final demand model.
V. EMPIRICAL RESULTS
The results of non-linear least squares estimation (available
by request) were tested for autocorrelation and heteroscedasticity.
No problem was indicated for the former (the
hypothesis of no autocorrelation could not be rejected at
any typical level of type I error following Mittelhammer
et al. 2000, p. 548). However, heteroscedasticity was
present in that the residual variance was dependent on
the level of expected attendance. In order to accommodate
heteroscedasticity, the model was re-estimated via nonlinear
generalized least squares (GLS) with the residual
variance modelled as dependent on the expected level of
attendance. The non-linear GLS results are reported in
Table 2 (the coefficient estimates were not substantially
impacted by the non-linear GLS heteroscedasticity transformation).
The table results show that most of the variance
in attendance is explained by the model, R2¼0.888,
and the predictive fit of the model was quite good, as evidenced
by the mean absolute percent error (MAPE), mean
percent error (MPE), and root mean square error (RMSE)
statistics.
Focusing on the innovation in this paper, the sign of the
DISTANCE and NEWTEAM variables are consistent
with the predictions in Equation 7. As the distance
between two teams increases, the DISTANCE variable
(measured as the inverse distance) between two teams
decreases. The magnitudes seem to be reasonable as well.
Compared to a team with a substitute that is at the sample
average 38 miles away, a team with a substitute that is
28 miles away loses 0.00941 953 500¼18 363 additional
fans per year, and there is an additional decrease of
126 550 fans in the first year that a NEWTEAM substitute
appears.
Table 2. Non-linear generalised least squares estimates
Variable Estimate T-statistic
INTERCEPT 5.7752 4.9094
Distance to substitute
DISTANCE 19.535 4.4309
NEW TEAM 1.2655 3.1238
Demand parameters
PRICES 0.151 1.882
INCOME 0.2153 3.3477
POP 0.3364 4.7727
WIN% 12.557 7.2012
CHANGEWIN% 12.3747 8.6802
RUNS 4.25E09 0.184
RUNS (EXPONENT) 9.1546 3.7676
DIVCHAMP 1.2524 4.5994
LOYALTY 0.7789 41.8821
STADIUM 5.9454 10.1191
STRIKE 0.1873 10.072
Other controls
ATTENDGROWTH 6.5982 11.49
TREND 0.0009 2.3855
ANGELS 1.2692 2.9994
BLUEJAYS 1.4641 2.5348
CUBS 1.2692 2.9795
DODGERS 1.7346 3.5833
CARDINALS 1.3459 3.1935
Diagnostics
R2 0.888
MAPE 12.21
MPE 2.213
RMSE 2.29
Location and attendance in baseball 2121
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
It is instructive to analyze the implications of the model
for a hypothetical team. If we assume a team that is
characterized by all of its continuous variables valued at
the sample means in Table 1, is in the first year of enduring
its new close substitute team (NEWTEAM¼1), has managed
a division championship the first year in a new
stadium (DIVCHAMP¼STADIUM¼1), in a year without
a work stoppage (STRIKE¼0), and that the team is
not one with additional fixed effects, the model predicts
attendance at 2 191 861. We will refer to this as our ‘comparison’
team.
The elasticity of attendance with respect to the distance
variable, evaluated at the sample average values of
ATTENDANCE and DISTANCE, is 0.031, so that a 1%
increase in DISTANCE leads to a 0.031% decrease in
attendance. In terms of physical miles, the sample average
DISTANCE is 0.0264, for an average of 38 miles. A onemile
decrease in miles between stadiums to 37 miles gives a
DISTANCE of 0.0270, an increase of 2.3% in the distance
variable. Thus, a one-mile decrease in the number of miles
between teams, evaluated around the sample average,
will decrease attendance by 2.30.031¼0.070%. For the
‘comparison’ team, each mile below the sample average
distance decreases attendance by 0.000702 191 861¼
1544. Since this ‘comparison’ team construct would be
charging the average ticket price of $5.98, this translates
to ticket revenue losses of $9235 per mile.
The estimation results for insight into two episodes from
MLB’s past are also used. The American League expanded
to include play by the LA Angels beginning in 1961. The
Angels shared Dodger Stadium from 1961–1965, when the
Angels moved to their stadium in Anaheim, 31 miles away.
Recall that the DISTANCE variable was set to 0.1 in this
special case of cohabitation. According to the model, the
owner of the Dodgers could have generated the following
expectation. Relative to the situation that eventually would
prevail for the 1966 season, when the Angels moved into
their new stadium, the first year of cohabitation would
reduce Dodger attendance by 0.06771 953 500¼132 252
per year, plus the first year loss of 126 550, for a total
258 802 lost attendance. In actuality, in 1961, Dodger
attendance fell 440 422 and then immediately bounced
back, but the model suggests that subsequent attendance
would have been even higher without the presence of the
Angels. In the remaining four years of cohabitation,
Dodger attendance would be 132 252 less than it would
be from the 1966 season on. In total, then, over the five
years of cohabitation, the Dodgers could have expected to
lose 258 802þ(4132 252)¼787 810 in attendance. With
the estimated average ticket price for the Dodgers from
1961–1965 being $2.23, the lost attendance represents
about $1.76 million in lost revenue over the period of
cohabitation. Recall that this is at a time when Dodger
star pitchers Don Drysdale and Sandy Koufax were earning
about $60 000 each.
Our second historical episode occurred when the
Senators leftWashington, DC at the end of the 1971 season
to become the Texas Rangers. When the Senators left, the
Baltimore Orioles DISTANCE variable changed from
0.0222 to 0.0097. In terms of miles, the new Philadelphia
Phillies substitute was 103 miles away compared to the
Senators at 45 miles. Again, resorting to the evaluations
at the sample means, Orioles’ attendance would be expected
to increase by 0.01251 953 500¼24 419 per year. With
an estimated Orioles ticket in 1972 being $2.24, that is
an increase of $54 699 for ticket sales alone. Of course,
over time, the effect would accumulate at higher ticket
prices (indeed, the Senators have been gone for 30 years).
The logic of these two examples will be used in the next
section to discuss the implications for contraction and
competition policy aimed at increasing the number of
MLB teams.
Two other outcomes in Table 2 are of interest to those
studying attendance demand. First, calculated at the means,
the price elasticity of attendance demand is 0.055. This is
considerably inelastic and consistent with the rest of the
extant work on attendance demand. Apparently, over the
lengthy sample, owners set ticket prices in the inelastic
portion of attendance demand. The usual explanation
is that attendance-related revenue (parking, concessions,
memorabilia) or the pursuit of subsidies lead to pricing
beyond the maximum of total revenue at the gate.
Second, loyalty as measured by last year’s attendance
significantly affects this year’s attendance. The elasticity
of attendance with respect to last year’s winning percentage
is 0.764. Judging from the relative elasticity estimates,
loyalty has about 14 times the impact that price has on
attendance. This result adds to the growing literature
suggesting that future studies of attendance demand should
account for the effect of fan loyalty.
The effects of the rest of the variables are largely as
expected. Over the sample period, baseball is an income
normal good (elasticity¼0.175) and population effects
are significantly positive. Turning to preferences, performance
(winning percentage, the change in winning percentage,
runs and a division championship) significantly
and positively affects attendance. The year a new stadium
is occupied, attendance increases by 594 500. Strike years
reduce attendance, all else constant, by about 19 000.
Interestingly, team indicator variables for Anaheim,
Toronto, the Chicago Cubs, LA and St Louis all are
significantly positive. Since these are meant to capture
unspecified fixed effects, it should come as no surprise
that we have no specific explanation for this outcome.
However, we do note that Depken (2001) ranked fan
loyalty for each of these teams as relatively strong.
2122 J. Winfree et al.
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
VI CONCLUSIONS
A travel cost model is used to explain attendance for MLB
teams, 1963–1998. At the sample average variable values,
a one-mile increase in distance to a substitute MLB team
increases attendance by about 1544 fans. The first appearance
of a new substitute team reduces attendance by an
additional 126 500 fans. In two interesting examples, sharing
Dodger Stadium with the expansion Angels probably
cost the Dodgers about $1.76 million (1963 dollars) over
a five-year period and the Orioles are better off by about
$54 699 (1972 dollars) per year after the Senators left for
Texas.
In addition, we find evidence of ticket pricing in the
inelastic region of attendance demand, consistent with
past studies. This indicates that attendance-related revenues
and the pursuit of state and local subsidies are important
to MLB owners over the sample period. It also
appears that the fan loyalty is an idea worthy of inclusion
in future studies of attendance demand.
Occasionally, MLB owners desire to expand the market
for baseball and new rivalries can follow the introduction
of a team into a currently occupied territory. However,
even though there are markets that may be able to sustain
additional MLB teams, such an occurrence would have
detrimental attendance effects on the incumbent team.
This leads us to some conclusions on current policy
considerations.
First, consider MLB’s call for contraction. One of the
current candidates, the Minnesota Twins, is largely insulated
from the distance effects analysed in this paper. The
nearest MLB city to Minneapolis/St Paul is Milwaukee,
which is well over 300 miles away. Elimination of the
Florida Marlins is a similar situation, where currently,
Miami is Tampa Bay’s closest MLB substitute, about
281 miles away. If the Marlins are eliminated, Tampa
Bay’s closest MLB substitute becomes the Atlanta Braves,
about 457 miles away. The difference is worth 0.0015
1 953 500¼2930 fans. For current consideration, the Fan
Cost Index (produced by Team Marketing Reports) allows
a fuller appreciation of attendance impacts by estimating
all attendance-related revenues for a family of four. At
Tampa Bay’s Fan Cost Index of $141.60, the additional
2930 fans are worth about $103 722 per year in additional
revenue for the owner of the Devil Rays.
It is difficult to believe that eliminating the Twins or
Montreal Expos will result in any attendance gains at all
for other MLB teams. Since attendance at Expos games is
less than 8000 fans per game, the people of Montreal will
likely not be yearning for baseball if the Expos leave. The
gains derived from contracting either of these teams can
only be in terms of reduced revenue sharing by remaining
teams. Contracting the Marlins would not be very different,
with only a marginal gain for the Devil Rays. For example,
at their 2001 attendance of 1 298 365, and using the same
$141.60 Fan Cost Index, the increase in attendance in
the absence of the Marlins would increase Tampa Bay
attendance-related revenues by only 0.21%.
Finally, consider competition policy that would increase
the number of teams in larger revenue markets, and in
particular, consider meeting the current desire for another
team in the Northern Virginia area. The impact would be
to reverse the attendance gains enjoyed by the Orioles when
the Senators left for Texas (adjusted for all other variables
as in Equation 7). In the first year, the new substitute
would reduce Oriole attendance by 24 419 fans, plus the
first year loss 126 550. At a Fan Cost Index of $141.12 in
Baltimore, first-year losses would be about $5.33 million.
In subsequent years, the annual 24 419 lost fans would cost
the Orioles about $861 502 per year. Competition policy
that would put a team back in the Northern Virginia
area would cost the Orioles (with 2001 attendance
of 2 951 371) about 5% of 2001 attendance-related revenue
in the first year and about 1% of 2001 attendance-related
revenue thereafter. By the foregoing examples, ongoing
attendance impacts are small, even when a team might be
placed fairly close to another (the Orioles example). It
would seem that a look at local TV revenue in the presence
of substitute teams would be required in order to assess the
impact on competitive balance in MLB and in the analysis
of policy considerations, to either reduce or increase the
number of teams.
ACKNOWLEDGEMENTS
The authors thank Bruce Domazlicky and Roger Noll for
portions of the ticket price data.
REFERENCES
Coffin, D. (1996) If you build it, will they come?, in Baseball
Economics: Current Research (Eds) J. Fizel, E. Gustafson
and L. Hadley, Praeger, Westport, CT.
Depken, C. A., III (2001) Fan loyalty in professional sports: an
extension to the National Football League, Journal of Sports
Economics, 2, 275–84.
Fort, R. (2001) Revenue disparity and competitive balance in
Major League Baseball, in Baseball’s Revenue Gap: Pennant
for Sale?: Hearing before the Subcommittee on Antitrust,
Business Rights, and Competition of the Committee on the
Judiciary, United States Senate, 106th Congress, 2nd Session,
21 November 2000, Government Printing Office, pp. 42–52.
Fort, R. (2002) Sports Economics, Prentice Hall, Englewood
Cliffs, NJ.
Fort, R. (2004) Inelastic sports pricing, Managerial and Decision
Economics, 25, 87–94.
Hotelling, H. (1929) Stability in competition, Economic Journal,
39, 41–57.
Kahane, L. and Shmanske, S. (1997) Team roster turnover and
attendance in Major League Baseball, Applied Economics,
29, 425–31.
Location and attendance in baseball 2123
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
Levin, R. C., Mitchell, G. J., Volcker, P. A. and Will, G. F. (2000)
The Report of the Independent Members of the
Commissioner’s Blue Ribbon Panel on Baseball Economics,
Major League Baseball, New York.
McConnell, K. E. (1985) The economics of outdoor recreation, in
Handbook of Natural Resource and Energy Economics, Vol. 1,
North Holland, Amsterdam.
Mittelhammer, R. C., Judge, G. G. and Miller, D. J. (2000)
Econometric Foundations, Cambridge University Press,
Cambridge, UK.
Quirk, J. and Fort, R. (1999) Hard Ball: The Abuse of Power in
Pro Team Sports, Princeton University Press, Princeton,
NJ.
2124 J. Winfree et al.
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016

Economics

Article Abstract 3
Instructions
Your critical abstracts should run to no more than 1 page. They should include the following:
1. Title and author(s)
2. The research question that the author wants to address
3. Main conclusions reached
4. How can this information be applied to the sports market?
5. What did the author(s) miss when evaluating their research question
Your abstract this week should come from one of the following articles:
• Winfree, J, JJ McCluskey, and RC Mittelhammer (2004). “Location and attendance in Major League Baseball”. In: Applied Economics 35, pp. 2117–2124.
I have the article attached for you.

Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=raec20
Download by: [Washington State University Libraries ] Date: 23 January 2016, At: 18:25
Applied Economics
ISSN: 0003-6846 (Print) 1466-4283 (Online) Journal homepage: http://www.tandfonline.com/loi/raec20
Location and attendance in major league baseball
Jason A. Winfree , Jill J. McCluskey , Ron C. Mittelhammer & Rodney Fort
To cite this article: Jason A. Winfree , Jill J. McCluskey , Ron C. Mittelhammer & Rodney
Fort (2004) Location and attendance in major league baseball, Applied Economics, 36:19,
2117-2124, DOI: 10.1080/0003684042000287664
To link to this article: http://dx.doi.org/10.1080/0003684042000287664
Published online: 02 Feb 2007.
Submit your article to this journal
Article views: 373
View related articles
Citing articles: 22 View citing articles
Location and attendance in major league
baseball
JASON A. WINFREE*, JILL J. MCCLUSKEYz,
RON C. MITTELHAMMERz and RODNEY FORTz
Program in Sports Management, University of Michigan, 401 Washtenaw Ave,
Ann Arbor, MI 48109–2214, USA and zSchool of Economic Sciences, Washington
State University, Pullman, Washington, USA
This study uses a travel-cost model to analyse the attendance impacts on Major
League Baseball (MLB) of the closest substitute MLB team. It is found that the
closer two teams are, the lower attendance is at each team relative to two teams that
are farther apart. In addition, when a new team moves into the area of an existing
team, there is an additional initial reduction in attendance for the incumbent team.
This has implications for actions aimed at changing the number of teams in MLB
either by contraction or by possible antitrust approaches that would increase the
number of teams, especially in megalopolis markets. Further, and consistent with
past demand studies, pricing is in the inelastic portion of gate demand and fan
loyalty is a significant contributor to the estimation of gate attendance.
I. INTRODUCTION
It is said that there are three important considerations
in business: location, location, and location. In sports economics,
distance between teams is a very important consideration
for professional sports leagues (and, although
not covered in this paper, college sports conferences). In
1968, when the Kansas City Athletics moved to Oakland,
the attendance of the San Francisco Giants dropped over
32%. The Colorado Rockies have no other Major League
Baseball teams within a 600-mile radius and enjoy large
crowds for every home game. A systematic assessment of
this type of attendance impact is the aim of this paper.
Currently, Major League Baseball (MLB) argues that
many owners face uncertain financial futures due to competitive
imbalance problems. One of the recommendations
made by Levin et al. (2000) is that MLB should allow
owners to move their teams, especially to larger revenue
markets, to reduce the imbalance. Incumbent owners have
made it clear that new teams in their market area are
substitutes that will reduce attendance at the incumbent
team. Indeed, this sensitivity to attendance impacts helps
explain the limited number of teams in so-called ‘megalopolis’
markets.
However, measuring attendance impacts also is important
from a policy perspective. Another approach to their
alleged revenue difficulties has MLB owners and their
Commissioner seriously considering a contraction of the
league by up to two teams. The policy question concerns
fan welfare after contractions. If increases in attendance for
teams proximate to those that are eliminated do not impact
league-wide competitive balance in any meaningful way,
then contraction will make a few owners better off with
primarily negative impacts on fan welfare (for those fans
losing their teams).
Another example of policy relevance concerns a suggestion
originating with Roger Noll and Ira Horowitz in
Congressional Hearings (cited in Fort, 2001, and extended
by Quirk and Fort, 1999). The original Noll-Horowitz
idea is that antitrust laws can be used to create competing
* Corresponding author. E-mail: jwinfree@umich.edu
Applied Economics ISSN 0003–6846 print/ISSN 1466–4283 online # 2004 Taylor & Francis Ltd 2117
http://www.tandf.co.uk/journals
DOI: 10.1080/0003684042000287664
Applied Economics, 2004, 36, 2117–2124
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
leagues in the hope that entry in megalopolis markets would
reduce revenue dispersion and competitive imbalance. The
work here suggests that only careful, case-by-case investigation
will reveal whether attendance impacts are large
enough to reduce revenue imbalance significantly in a
given league.
Demand and support for baseball teams are not only
important to baseball owners and fans, but to state and
local governments. This study also has implications for
taxpayers, who have paid large sums for stadiums and
team subsidies. The public has paid about 64% of professional
sports facilities, which accounts for almost $6 billion
from 1999 to 2003 (Fort, 2002). Local politicians count on
high demand for these professional sports. This study finds
that a given MLB team’s attendance is affected by its
substitute’s location and stadium. An implication of this
finding is that investing in a new stadium might not be a net
gain for the community if it detracts from another team in
the same metropolitan area. For example, it is estimated
that the public will pay $390 million for a new stadium
for the New York Mets that should be completed in
2004. However, if the Mets do not build a new stadium,
many fans may choose to attend Yankee games instead.
It is important to consider the effect of these stadiumfunding
decisions on both the team in question and its
closest substitute.
The principal importance of the attendance demand
analysis in this paper is its examination of how team attendance
is affected by neighbouring teams or the appearance
of a new team close by. Theoretically, the closer the team,
the more fans will be lost to the substitute. In addition,
since one is estimating the demand for attendance at the
gate, there will be findings relevant to past demand observations,
mainly concerning the elasticity of gate pricing and
fan loyalty. At the level of annual observations on teams
used in this paper, one cannot contribute to the analysis of
the uncertainty of outcome hypothesis.
The paper is organized as follows. First, previous studies
of location and baseball demand are briefly reviewed.
Second, the travel-cost model is presented. Third, the
empirical specification is developed and the data are
described. Fourth, the empirical results are presented and
discussed. Conclusions, with applications of the findings
to the policy issues of contraction and competition policy,
complete the study.
II. PREVIOUS STUDIES
A principal tenet of Hotelling’s (1929) location model is
that, all else constant, people will choose to purchase
goods or services from the closest seller. A travel-cost
model (McConnell, 1985) is used to apply Hotelling’s
idea to MLB attendance. The main cost for many baseball
fans is travel cost. Consequently, if there are two teams
that provide essentially the same utility for certain fans,
all else constant, they will attend the game that is closest.
This implies that attendance lost to a substitute team
should be a function of the distance to that substitute.
To the authors’ knowledge, the current study is the first
to use a travel-cost model to explain attendance for professional
sporting events, and no studies examine directly how
baseball attendance is related to substitute possibilities.
Previous demand studies have generated engaging
hypotheses and some interesting findings. One is the recurring
empirical observation that owners set ticket prices in
the inelastic portion of gate demand. Fort (2004) shows a
variety of circumstances where this choice is consistent with
profit maximization and that the conditions hold for MLB
in the late 1980s. It will be of interest to see if this also is
true on the current data.
Another interesting hypothesis is that fan loyalty impacts
attendance. Kahane and Shmanske (1997) show that roster
stability engenders loyalty (higher attendance is observed
for stable rosters, all else constant). Depken (2001) uses
stochastic frontier estimation and interprets teams closest
to the attendance frontier as having the most loyal fans.
Under this interpretation, he finds teams that relocated had
low fan loyalty. It will be noted how the work contributes
to both the inelastic pricing and loyalty issues.
III. THE MODEL
The utility of a baseball fan is a function of the number
of visits to baseball stadiums, the quality of the teams,
the quality of those stadiums, as well as all other goods
consumed. Of course, a fan’s utility can be different for
different teams. It is also reasonable to assume that baseball
fans may desire a variety of stadiums and teams, so
that the home games of all accessible stadiums and teams
are included in the utility function. Fan i faces the following
utility maximization problem:
MaxUiðXi,Vi1, . . . ,Vim, q1, . . . , qmÞ, ð1Þ
subject to a monetary constraint:
Mi þ witwi
¼ Xi þX
m
j¼1
cijVij , ð2Þ
and a time constraint:
t
i ¼ twi
þX
m
j¼1
ðttij
þ tvÞVij , ð3Þ
where Xi is the quantity of the numeraire, Vij is the number
of visits by the ith individual to the jth stadium, qj is the
quality of going to stadium j, Mi is the exogenous income
of individual i, wi is the wage rate of individual i, twi is the
hours worked by individual i, cij is the monetary cost of
going to the jth stadium for the ith individual, t
i is the total
2118 J. Winfree et al.
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
discretionary time for individual i, ttij is the total travel time
for the ith individual going to stadium j, tv is the time spent
during the visit and assumed constant, and m is the number
of major league baseball stadiums.
It is assumed that the number of visits and quality are
complements if they are for the same team (e.g. a baseball
game is more enjoyable if one is familiar with the players),
while everything else in the utility function are substitutes.
For the large majority of fans who go to see the home
team, the quality variable, q, would be a function of both
the quality of the home team and the quality of the stadium.
It is expected that this quality is effected by variables
such as winning percentage, pennant chances, and the age
of the stadium. The model also assumes that the individual
chooses the amount of time that he or she works at a
constant wage rate. Combining the constraints yields the
following relationship:
Mi þ wit
i ¼ Xi þX
m
j¼1
ðcij þ wiðttij
þ tvÞÞVij , ð4Þ
or
Mi þ wit
i ¼ Xi þX
m
j¼1
ðptj
þ pcj
þ pdidij þ wiðttij
þ tvÞÞVij :
ð5Þ
The monetary cost may also be disaggregated into ticket
price, ptj , price of concessions, pcj , and travel cost per mile
multiplied by the round trip mileage, pdidij . To find the
total cost of the visit, one must consider the monetary
cost and the opportunity cost per visit, which is equal to
the wage rate multiplied by the sum of the travel time and
time spent at the game.
This model implies that if wages or exogenous income
increase, the number of visits should increase. However,
if any of the prices increase, one would expect the number
of visits to decrease. Since a greater distance would increase
the price for the consumer, an increase in distance should
decrease visits. Therefore, if another team moves close by it
should increase visits to the new team, possibly decreasing
visits to the old team.
Maximizing utility (1) subject to constraint (5) yields a
demand equation for the visits to stadium j by individual i:
Vij ¼ Vijðqj ; pij;Mi; dij;ZiÞ; ð6Þ
where demand is a function of quality of the site, prices,
income, and distance to the site among other variables
represented generically by Zi. If quality increases, then so
should the number of visits. Then the aggregate attendance
demand for the team is the sum of all individual demand
equations, as:
Vj ¼X
n
i¼1
Vijðqj; pij ;Mi; dij;ZiÞ: ð7Þ
IV. EMPIRICAL SPECIFICATION AND
THE DATA
The empirical specification of the model in (7) is given
below (descriptive statistics for all variables are presented
in Table 1):
ATTENDANCE
¼0þ1DISTANCEþ2NEWTEAMþ3PRICE
þ4INCOMEþ5POPULATIONþ6WIN%
þ7CHANGEWIN%þ8RUNS9 þ10DIVCHAMP
þ11LOYALTYþ12STADIUMþ13STRIKE
þ14ATTENDGROWTHþ15TREND
þ16ANGELSþ17BLUEJAYSþ18CUBS
þ19DODGERSþ20CARDINALSþ”: ð8Þ
The dependent variable is annual ATTENDANCE, in
hundreds of thousands, for a particular team and ” is a
residual term. The observations cover teams over the years
1963 to 1998, with the number of teams varying from 20 in
1963 to 30 in 1998. The attendance data are for the regular
season, yielding 884 observations.
Two variables novel to this study capture the impacts of
distance between teams (the availability of very close substitutes).
DISTANCE equals the inverse of the distance in
miles to the nearest alternative MLB stadium. The value of
DISTANCE only changes over time for a team if a team
moved or a different team moved close by. For example,
when the Senators left Washington, DC at the end of the
1971 season to become the Texas Rangers, the Baltimore
Orioles DISTANCEvariable changed from 0.0222 to 0.0097
Table 1. Summary statistics
Variable Min Max Mean SD
ATTENDANCE 3.0676 40.5795 16.4907 7.2151
DISTANCE 0.0012 0.1 0.0264 0.0337
NEW TEAM 0 1 0.0362 0.1869
PRICE 2.5728 12.2638 5.9819 1.0735
INCOME 7.0425 20.4951 13.4349 2.4054
POP 0.1317 7.8949 1.5573 1.9581
WIN% 0.3086 0.704 0.5016 0.0686
CHANGEWIN% 0.2349 0.2016 0.0013 0.0699
RUNS 3.29 9.93 6.7775 0.9897
DIVCHAMP 0 1 0.1448 0.3521
LOYALTY 3.0676 44.8335 16.1775 7.1543
STADIUM 0 1 0.0192 0.1374
STRIKE 0 44.8335 1.2813 5.4275
ATTENDGROWTH 0.4065 0.7214 0.0429 0.1637
TREND 1 1296 486.063 399.494
ANGELS 1 0.0407 0.1978
BLUEJAYS 0 1 0.0238 0.1524
CUBS 0 1 0.0407 0.1978
DODGERS 0 1 0.0407 0.1978
CARDINALS 0 1 0.0407 0.1978
Location and attendance in baseball 2119
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
since the Philadelphia Phillies became their closest substitute
MLB team. In this sample, the only simultaneous
single-stadium occupation was the Angels and Dodgers
in Dodger Stadium (1961–1965). The denominator of the
DISTANCE variable was constrained to 0.1 for this special
case, which represented the largest empirical value of the
DISTANCE variable.
NEWTEAM is the second variable used to capture the
impact of distance between teams. NEWTEAM equals one
for the first year a new team moves within 500 miles of an
existing stadium in a given year. A new team might generate
more excitement and be a better substitute for many
people than an existing substitute team. NEWTEAM could
capture an additional initial loss of attendance from league
expansion or relocation. This variable includes teams moving
and expansion years, which includes 1969, 1977, 1993,
and 1998.
Other demand parameters follow a typical demand function
specification (price, income, population and preferences).
PRICE denotes the own team’s price of baseball
tickets in real (regional CPI-adjusted) terms. Prices have
been obtained from four different sources – American
League Red Book (1963), Bruce Domazlicky (1969–1980),
Roger Noll (1975–1988), and the Society for American
Baseball Research (1991–1998). National League Green
Books were not used since older years do not have ticket
price data. Missing data were replaced by interpolation
using the historical rate of growth over the period 1969–
1988. Since there are a variety of seats with different prices
for each baseball game, averaging of seat prices is a difficult
issue to resolve. Coffin (1996) discusses the problems of
estimating an average price. Each data source used its
own system for averaging,1 and when different prices for
the same team and year were available, a simple mean
across all sources was used. Note that ticket prices that
overlapped were very similar in value.
INCOME denotes real (regional CPI-adjusted) statelevel
per capita personal income (in thousands of dollars).
Per capita personal income was obtained from the Bureau
of Economic Analysis website for all US states. Canadian
per capita personal income was used for Montreal and
Toronto. State level data is used to allow for the fact
that people attend baseball games from a variety of geographical
locations.
POP is the population of the city in which a stadium
resides (in millions). City populations are available
from the US Census Bureau website, <www.census.gov>
and Canadian city population was found at <www.demographia.
com>. For years between censuses, a constant
growth rate was applied to interpolate population. The
same method was used for Canadian cities; however,
the censuses were a year later. This variable is used as an
instrumental variable related to the size of market for
sports entertainment.
Preference controls are included for team performance
(winning, run production, championships), loyalty, stadium
quality and work stoppages. WIN% is the team’s regular
season winning percentage. In general, fans prefer to see
teams that win. CHANGEWIN% is the change in winning
percentage for the team from year to year. It is the difference
between the current year’s winning percentage and the
previous (lagged) year’s winning percentage. It is hypothesized
that both winning percentage and its changes
affect the quality of a fan’s experience. If a city has not
had a successful baseball team in quite some time, then
attendance can be relatively more exciting and enjoyable
when the team begins winning more games. However, if a
team wins consistently from year to year, fans may take
winning for granted and become disgruntled when a team
loses.
RUNS is the number of runs that were scored by a team
during a year. Fans have a tendency not only to prefer a
winning team, but also to prefer more rather than fewer
runs scored. However, it was not necessarily assumed that
runs affect attendance linearly. Therefore, in a manner akin
to a Box-Cox transformation, a nonlinear power function
of runs scored was used and was found to be more effective
in explaining attendance.
DIVCHAMP is an indicator variable that equals
one when a team wins a division championship. The
DIVCHAMP variable includes information not necessarily
captured in WIN% since a team can have a high winning
percentage but still not win its division. Prior to 1969, there
were no divisions, so only the two teams that won their
respective leagues are included as the DIVCHAMPs.
The LOYALTY variable is the lagged attendance for
the team (in hundreds of thousands) and is intended to
proxy the fans’ attendance habit persistence. This variable
is lagged one year. For example, Montreal does not seem to
be a ‘baseball town’ and suffers from very low attendance
that persists year in and year out, while attendance in other
cities seems to flourish regardless of how the team plays.
Lagged attendance proxies, both affects. The present version
of ‘loyalty’ varies from that used in other studies
(Kahane and Shmanske, 1997; Depken, 2001), providing
a different dimension for comparison with those works.
STADIUM is an indicator variable used to proxy for
the effect of a new stadium on attendance. New stadiums
tend to increase attendance dramatically, at least for a short
time (and possibly a longer period if new stadium revenues
translate into an increase in team quality). The effect may
be largely captured by the LOYALTY variable, so the
STADIUM variable is used only for the year in which
the new stadium is first opened.
1 Contact the authors for a detailed description of the averaging for each data source.
2120 J. Winfree et al.
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
STRIKE is an indicator variable equal to one multiplied
by the previous year’s attendance in years of work stoppages.
This specification assumes that, while strike effects
can vary across teams, they effectively are proportional to
typical team attendance. There were two years in the sample
when work stoppages had any significant consequence
on season length, namely the players’ strikes of 1981 and
1994 (Fort, 2002). Obviously, attendance will fall if a team
is not playing as many games. About 50 to 60 games were
cancelled for each team during the 1981 and 1994 strikes.
Additional controls also seemed appropriate. ATTENDGROWTH
controls for overall growth in attendance,
league wide. It is measured by the growth rate of league
attendance, omitting the particular team’s attendance being
measured by the ATTENDANCE observation. The variable
is used as a proxy measure for the overall popularity
of baseball. The TREND variable proxies secular shifts
that can occur from a number of different sources, such
as increases in population beyond city limits, increased
availability of leisure time, or shifts in preferences in
favour of people desiring to attend more baseball games.
Examining attendance over time graphically, there appeared
to be an underlying quadratic trend, and to investigate this
systematically, a quadratic time variable is used, equal to
the year squared, so that 1963 equals 1, 1964 equals 4, 1965
equals 9, and so on. The reason that including both the
TREND and ATTENDGROWTH variables is particularly
useful is that ATTENDGROWTH can account for
large and/or irregular increases or decreases in attendance
that TREND cannot. Individual team indicator variables
were also included. These represent fixed effects for these
particular teams that are not otherwise captured by the
model.
It is important to note that other demand variables
relating to the closest substitute may affect a team’s attendance.
For example, the price of a Yankees ticket may
affect the Mets’ attendance. The variables PRICE, WIN%,
CHANGEWIN%, RUNS, DIVCHAMP, and STADIUM
for the nearest team were all initially included in the demand
model and the variables were then weighted by the variables
interacting with the DISTANCE variable. However,
none of these variables was statistically significant, and
they were excluded from the final demand model.
V. EMPIRICAL RESULTS
The results of non-linear least squares estimation (available
by request) were tested for autocorrelation and heteroscedasticity.
No problem was indicated for the former (the
hypothesis of no autocorrelation could not be rejected at
any typical level of type I error following Mittelhammer
et al. 2000, p. 548). However, heteroscedasticity was
present in that the residual variance was dependent on
the level of expected attendance. In order to accommodate
heteroscedasticity, the model was re-estimated via nonlinear
generalized least squares (GLS) with the residual
variance modelled as dependent on the expected level of
attendance. The non-linear GLS results are reported in
Table 2 (the coefficient estimates were not substantially
impacted by the non-linear GLS heteroscedasticity transformation).
The table results show that most of the variance
in attendance is explained by the model, R2¼0.888,
and the predictive fit of the model was quite good, as evidenced
by the mean absolute percent error (MAPE), mean
percent error (MPE), and root mean square error (RMSE)
statistics.
Focusing on the innovation in this paper, the sign of the
DISTANCE and NEWTEAM variables are consistent
with the predictions in Equation 7. As the distance
between two teams increases, the DISTANCE variable
(measured as the inverse distance) between two teams
decreases. The magnitudes seem to be reasonable as well.
Compared to a team with a substitute that is at the sample
average 38 miles away, a team with a substitute that is
28 miles away loses 0.00941 953 500¼18 363 additional
fans per year, and there is an additional decrease of
126 550 fans in the first year that a NEWTEAM substitute
appears.
Table 2. Non-linear generalised least squares estimates
Variable Estimate T-statistic
INTERCEPT 5.7752 4.9094
Distance to substitute
DISTANCE 19.535 4.4309
NEW TEAM 1.2655 3.1238
Demand parameters
PRICES 0.151 1.882
INCOME 0.2153 3.3477
POP 0.3364 4.7727
WIN% 12.557 7.2012
CHANGEWIN% 12.3747 8.6802
RUNS 4.25E09 0.184
RUNS (EXPONENT) 9.1546 3.7676
DIVCHAMP 1.2524 4.5994
LOYALTY 0.7789 41.8821
STADIUM 5.9454 10.1191
STRIKE 0.1873 10.072
Other controls
ATTENDGROWTH 6.5982 11.49
TREND 0.0009 2.3855
ANGELS 1.2692 2.9994
BLUEJAYS 1.4641 2.5348
CUBS 1.2692 2.9795
DODGERS 1.7346 3.5833
CARDINALS 1.3459 3.1935
Diagnostics
R2 0.888
MAPE 12.21
MPE 2.213
RMSE 2.29
Location and attendance in baseball 2121
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
It is instructive to analyze the implications of the model
for a hypothetical team. If we assume a team that is
characterized by all of its continuous variables valued at
the sample means in Table 1, is in the first year of enduring
its new close substitute team (NEWTEAM¼1), has managed
a division championship the first year in a new
stadium (DIVCHAMP¼STADIUM¼1), in a year without
a work stoppage (STRIKE¼0), and that the team is
not one with additional fixed effects, the model predicts
attendance at 2 191 861. We will refer to this as our ‘comparison’
team.
The elasticity of attendance with respect to the distance
variable, evaluated at the sample average values of
ATTENDANCE and DISTANCE, is 0.031, so that a 1%
increase in DISTANCE leads to a 0.031% decrease in
attendance. In terms of physical miles, the sample average
DISTANCE is 0.0264, for an average of 38 miles. A onemile
decrease in miles between stadiums to 37 miles gives a
DISTANCE of 0.0270, an increase of 2.3% in the distance
variable. Thus, a one-mile decrease in the number of miles
between teams, evaluated around the sample average,
will decrease attendance by 2.30.031¼0.070%. For the
‘comparison’ team, each mile below the sample average
distance decreases attendance by 0.000702 191 861¼
1544. Since this ‘comparison’ team construct would be
charging the average ticket price of $5.98, this translates
to ticket revenue losses of $9235 per mile.
The estimation results for insight into two episodes from
MLB’s past are also used. The American League expanded
to include play by the LA Angels beginning in 1961. The
Angels shared Dodger Stadium from 1961–1965, when the
Angels moved to their stadium in Anaheim, 31 miles away.
Recall that the DISTANCE variable was set to 0.1 in this
special case of cohabitation. According to the model, the
owner of the Dodgers could have generated the following
expectation. Relative to the situation that eventually would
prevail for the 1966 season, when the Angels moved into
their new stadium, the first year of cohabitation would
reduce Dodger attendance by 0.06771 953 500¼132 252
per year, plus the first year loss of 126 550, for a total
258 802 lost attendance. In actuality, in 1961, Dodger
attendance fell 440 422 and then immediately bounced
back, but the model suggests that subsequent attendance
would have been even higher without the presence of the
Angels. In the remaining four years of cohabitation,
Dodger attendance would be 132 252 less than it would
be from the 1966 season on. In total, then, over the five
years of cohabitation, the Dodgers could have expected to
lose 258 802þ(4132 252)¼787 810 in attendance. With
the estimated average ticket price for the Dodgers from
1961–1965 being $2.23, the lost attendance represents
about $1.76 million in lost revenue over the period of
cohabitation. Recall that this is at a time when Dodger
star pitchers Don Drysdale and Sandy Koufax were earning
about $60 000 each.
Our second historical episode occurred when the
Senators leftWashington, DC at the end of the 1971 season
to become the Texas Rangers. When the Senators left, the
Baltimore Orioles DISTANCE variable changed from
0.0222 to 0.0097. In terms of miles, the new Philadelphia
Phillies substitute was 103 miles away compared to the
Senators at 45 miles. Again, resorting to the evaluations
at the sample means, Orioles’ attendance would be expected
to increase by 0.01251 953 500¼24 419 per year. With
an estimated Orioles ticket in 1972 being $2.24, that is
an increase of $54 699 for ticket sales alone. Of course,
over time, the effect would accumulate at higher ticket
prices (indeed, the Senators have been gone for 30 years).
The logic of these two examples will be used in the next
section to discuss the implications for contraction and
competition policy aimed at increasing the number of
MLB teams.
Two other outcomes in Table 2 are of interest to those
studying attendance demand. First, calculated at the means,
the price elasticity of attendance demand is 0.055. This is
considerably inelastic and consistent with the rest of the
extant work on attendance demand. Apparently, over the
lengthy sample, owners set ticket prices in the inelastic
portion of attendance demand. The usual explanation
is that attendance-related revenue (parking, concessions,
memorabilia) or the pursuit of subsidies lead to pricing
beyond the maximum of total revenue at the gate.
Second, loyalty as measured by last year’s attendance
significantly affects this year’s attendance. The elasticity
of attendance with respect to last year’s winning percentage
is 0.764. Judging from the relative elasticity estimates,
loyalty has about 14 times the impact that price has on
attendance. This result adds to the growing literature
suggesting that future studies of attendance demand should
account for the effect of fan loyalty.
The effects of the rest of the variables are largely as
expected. Over the sample period, baseball is an income
normal good (elasticity¼0.175) and population effects
are significantly positive. Turning to preferences, performance
(winning percentage, the change in winning percentage,
runs and a division championship) significantly
and positively affects attendance. The year a new stadium
is occupied, attendance increases by 594 500. Strike years
reduce attendance, all else constant, by about 19 000.
Interestingly, team indicator variables for Anaheim,
Toronto, the Chicago Cubs, LA and St Louis all are
significantly positive. Since these are meant to capture
unspecified fixed effects, it should come as no surprise
that we have no specific explanation for this outcome.
However, we do note that Depken (2001) ranked fan
loyalty for each of these teams as relatively strong.
2122 J. Winfree et al.
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
VI CONCLUSIONS
A travel cost model is used to explain attendance for MLB
teams, 1963–1998. At the sample average variable values,
a one-mile increase in distance to a substitute MLB team
increases attendance by about 1544 fans. The first appearance
of a new substitute team reduces attendance by an
additional 126 500 fans. In two interesting examples, sharing
Dodger Stadium with the expansion Angels probably
cost the Dodgers about $1.76 million (1963 dollars) over
a five-year period and the Orioles are better off by about
$54 699 (1972 dollars) per year after the Senators left for
Texas.
In addition, we find evidence of ticket pricing in the
inelastic region of attendance demand, consistent with
past studies. This indicates that attendance-related revenues
and the pursuit of state and local subsidies are important
to MLB owners over the sample period. It also
appears that the fan loyalty is an idea worthy of inclusion
in future studies of attendance demand.
Occasionally, MLB owners desire to expand the market
for baseball and new rivalries can follow the introduction
of a team into a currently occupied territory. However,
even though there are markets that may be able to sustain
additional MLB teams, such an occurrence would have
detrimental attendance effects on the incumbent team.
This leads us to some conclusions on current policy
considerations.
First, consider MLB’s call for contraction. One of the
current candidates, the Minnesota Twins, is largely insulated
from the distance effects analysed in this paper. The
nearest MLB city to Minneapolis/St Paul is Milwaukee,
which is well over 300 miles away. Elimination of the
Florida Marlins is a similar situation, where currently,
Miami is Tampa Bay’s closest MLB substitute, about
281 miles away. If the Marlins are eliminated, Tampa
Bay’s closest MLB substitute becomes the Atlanta Braves,
about 457 miles away. The difference is worth 0.0015
1 953 500¼2930 fans. For current consideration, the Fan
Cost Index (produced by Team Marketing Reports) allows
a fuller appreciation of attendance impacts by estimating
all attendance-related revenues for a family of four. At
Tampa Bay’s Fan Cost Index of $141.60, the additional
2930 fans are worth about $103 722 per year in additional
revenue for the owner of the Devil Rays.
It is difficult to believe that eliminating the Twins or
Montreal Expos will result in any attendance gains at all
for other MLB teams. Since attendance at Expos games is
less than 8000 fans per game, the people of Montreal will
likely not be yearning for baseball if the Expos leave. The
gains derived from contracting either of these teams can
only be in terms of reduced revenue sharing by remaining
teams. Contracting the Marlins would not be very different,
with only a marginal gain for the Devil Rays. For example,
at their 2001 attendance of 1 298 365, and using the same
$141.60 Fan Cost Index, the increase in attendance in
the absence of the Marlins would increase Tampa Bay
attendance-related revenues by only 0.21%.
Finally, consider competition policy that would increase
the number of teams in larger revenue markets, and in
particular, consider meeting the current desire for another
team in the Northern Virginia area. The impact would be
to reverse the attendance gains enjoyed by the Orioles when
the Senators left for Texas (adjusted for all other variables
as in Equation 7). In the first year, the new substitute
would reduce Oriole attendance by 24 419 fans, plus the
first year loss 126 550. At a Fan Cost Index of $141.12 in
Baltimore, first-year losses would be about $5.33 million.
In subsequent years, the annual 24 419 lost fans would cost
the Orioles about $861 502 per year. Competition policy
that would put a team back in the Northern Virginia
area would cost the Orioles (with 2001 attendance
of 2 951 371) about 5% of 2001 attendance-related revenue
in the first year and about 1% of 2001 attendance-related
revenue thereafter. By the foregoing examples, ongoing
attendance impacts are small, even when a team might be
placed fairly close to another (the Orioles example). It
would seem that a look at local TV revenue in the presence
of substitute teams would be required in order to assess the
impact on competitive balance in MLB and in the analysis
of policy considerations, to either reduce or increase the
number of teams.
ACKNOWLEDGEMENTS
The authors thank Bruce Domazlicky and Roger Noll for
portions of the ticket price data.
REFERENCES
Coffin, D. (1996) If you build it, will they come?, in Baseball
Economics: Current Research (Eds) J. Fizel, E. Gustafson
and L. Hadley, Praeger, Westport, CT.
Depken, C. A., III (2001) Fan loyalty in professional sports: an
extension to the National Football League, Journal of Sports
Economics, 2, 275–84.
Fort, R. (2001) Revenue disparity and competitive balance in
Major League Baseball, in Baseball’s Revenue Gap: Pennant
for Sale?: Hearing before the Subcommittee on Antitrust,
Business Rights, and Competition of the Committee on the
Judiciary, United States Senate, 106th Congress, 2nd Session,
21 November 2000, Government Printing Office, pp. 42–52.
Fort, R. (2002) Sports Economics, Prentice Hall, Englewood
Cliffs, NJ.
Fort, R. (2004) Inelastic sports pricing, Managerial and Decision
Economics, 25, 87–94.
Hotelling, H. (1929) Stability in competition, Economic Journal,
39, 41–57.
Kahane, L. and Shmanske, S. (1997) Team roster turnover and
attendance in Major League Baseball, Applied Economics,
29, 425–31.
Location and attendance in baseball 2123
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016
Levin, R. C., Mitchell, G. J., Volcker, P. A. and Will, G. F. (2000)
The Report of the Independent Members of the
Commissioner’s Blue Ribbon Panel on Baseball Economics,
Major League Baseball, New York.
McConnell, K. E. (1985) The economics of outdoor recreation, in
Handbook of Natural Resource and Energy Economics, Vol. 1,
North Holland, Amsterdam.
Mittelhammer, R. C., Judge, G. G. and Miller, D. J. (2000)
Econometric Foundations, Cambridge University Press,
Cambridge, UK.
Quirk, J. and Fort, R. (1999) Hard Ball: The Abuse of Power in
Pro Team Sports, Princeton University Press, Princeton,
NJ.
2124 J. Winfree et al.
Downloaded by [Washington State University Libraries ] at 18:25 23 January 2016