The National Hockey League and Cross-Border Fandom: Fan Substitution and International Boundaries

Brian M. Mills1 and Mark S. Rosentraub2

April 18, 2014

1 Corresponding Author Department of Tourism, Recreation and Sport Management College of Health and Human Performance University of Florida P.O. Box 118208 Gainesville, FL (USA) 32611-8208 Phone: 352-294-1664 Email: bmmillsy@hhp.ufl.edu 2Sport Management School of Kinesiology University of Michigan 1402 Washington Heights Ann Arbor, MI (USA) 48109

Acknowledgements: We are grateful to Jadrian Wooten, Jason Winfree, Brad Humphreys, Brian Soebbing, Frank Stephenson, Liam Lenten, an anonymous referee, and the participants at the Western Economics Association meetings for many helpful comments and suggestions for this paper.

1

The National Hockey League and Cross-Border Fandom: Fan Substitution and International Boundaries

Abstract

This paper uses daily border crossing data within the Niagara region of Ontario and New York to

evaluate the Canadian market for a United States based National Hockey League team, the

Buffalo Sabres. We conservatively estimate that 15 percent of attendees at Buffalo Sabres home

games travel from Canada. This effect is heterogeneous with respect to the opponent country of

origin, with higher levels of game day border crossing associated with a Canadian visiting team.

We also find fan substitution effects between Buffalo and the Toronto Maple Leafs with respect

to both the price of attendance and quality of each team. Implications extend to NHL expansion

near international borders and compensation to incumbent teams both within and across the

national border near where an expansion or relocated team is placed.

Keywords: Hockey, Fan Substitution, International Trade, Travel Costs, League Expansion

2

The National Hockey League and Cross-Border Fandom: Fan Substitution and International Boundaries

“Right now, 15% of our season ticket holders come from Canada.”

-Ted Black, President, Buffalo Sabresi

1 Introduction

This paper uses border-crossing data from the Niagara Peninsula to assess the propensity

of Canadian residents to attend Buffalo Sabres games, and how the price and quality of the

Toronto Maple Leafs—relative to the Sabres—impacts these travel decisions. This inquiry is

vital to the discussion of expansion or relocation by the National Hockey League (NHL). Part of

the expansion debate for any league is the extent to which a new team leads to substitution

effects that undermine the economic position of an existing team. However, research on fan

substitution in professional sports is somewhat limited (Gitter & Rhoads, 2011; Jones &

Ferguson, 1988; Rascher et al., 2009; Winfree, 2009; Winfree & Fort, 2008; Winfree,

McCluskey, Mittelhammer, & Fort, 2004). We therefore add to the literature on fan substitution

across professional sports teams, and examine how this phenomenon occurs across international

borders.

We find that existing teams—the Toronto Maple Leafs and Buffalo Sabres—do compete

for fans both in the price of attendance and the quality of the team on the ice, despite a distance

of 99 miles between their respective home arenas. Canadian hockey fans substitute attendance to

Maple Leafs games by attending Sabres games when there is an increase in the relative price—

and/or decrease in relative quality—of the Leafs. Additionally, our analysis reveals that a

significant portion of the fan base for the Sabres travels from Canada, exhibiting the potential

3

damages that could be claimed by the Sabres under the NHL By-Laws in the case of a relocation

or expansion even across this international border (NHL, 2009).

2 Background and Literature Review

The Buffalo Sabres have a market area that includes the Niagara peninsula’s population

of approximately one million, with nearly 400,000 residents in the nearby St. Catharines-Niagara

area (Statistics Canada, 2013a). The current situation—the overlapping markets of the Toronto

Maple Leafs and Buffalo Sabres—already provides a natural example to model consumer

choices in this area. The close proximity of the two teams, and a relative lack of teams in

Canada, could mean that a large number of supporters of a of U.S.-based hockey team reside

outside the team’s home country. Those living between Toronto and Buffalo may be subject to

the predictions of the Hotelling model: consumers will make their purchase at the most nearby

arena. We address this travel dynamic, along with the relative cost of attendance and the relative

quality of teams as these could differentiate both product (game) quality and price (tickets,

parking, concessions). Consumers may well prefer the closest option, all else equal, but could be

swayed to attend more distant games depending on these price and quality differences. Though

we do not have direct access to game attendee addresses or travel distances, we infer travel to

games through vehicular border crossing data in the Buffalo, New York area, taking a similar

approach to that of Matheson, Peeters, and Szymanski (2012).ii

Particularly relevant to our investigation is the work of Winfree et al. (2004) and the

finding that teams in Major League Baseball (MLB) experience reduced attendance levels when

there was another nearby franchise. Winfree (2009) also found that when a new or relocated

team enters a market, incumbent teams are likely to lose fans. Additionally, Henrickson (2004)

found that the nearby location of other teams may also impact pricing. It is possible that

4

residents of these areas could shift their loyalties to a team located in Canada, raising the

question of whether a U.S.-based team faces barriers within a Canadian market. Understanding

fan substitution in this market area is necessary to the consideration of compensation from a new

team, and whether this is a relevant factor when that team is located across an international

border.

Several groups are interested in having a third NHL team in the region. One location

discussed for a second Ontario Province team is Markham, a northern suburb located 118 miles

from Buffalo. Another possible location is Hamilton, located 42 miles south of home arena used

by the Toronto Maple Leafs and 64 miles from Buffalo. A new team in Hamilton would likely

increase the competition for fans in of both the Maple Leafs and the Sabres, with a market area

that would overlap with both (Figure 1). Though a Hamilton-based team itself would be located

beyond Buffalo’s exclusive market area, the Sabres’ leadership has argued that they are entitled

to a broad interpretation of the league’s guidelines with regard to market areas as much of their

exclusive territory includes large bodies of water.iii,iv

A team in Hamilton could result not only in a closer option for many hockey fans in

Canada, but also provide them with another team from their home country where there could be

a shortage in the supply of professional hockey at its highest level (Keller & McGuire, 2011).

Patriotic factors could indeed impact consumers’ behavior patterns of attendance at Buffalo

Sabres games (Andrijiw & Hyatt, 2009; Hyatt and Andrijiw, 2008; James, Kolbe & Trail, 2002;

Lock, Taylor & Darcy, 2011). It is possible that a team within one’s own home country would

create more civic pride, and a more attractive consumer choice for hockey fans in Ontario. For

example, Jackson (1994) showed the importance of sport—specifically hockey—to Canadians

and their national identity. Additionally, the Americanization of hockey has resulted in some

5

past concern among Canadian fans (Mason, 2002). Though it is difficult to measure the

propensity for current Sabres fans to switch allegiance, a new team in the area could sway fans

with lower levels of identification with Buffalo to games closer to their homes and played in

their country. If so, then a new team located in a home country—relative to one outside of it—

may have additional negative impacts on the Sabres’ fan base.

Finally, while explicit travel costs—such as distance traveled—are important

consumption costs related to attendance at sporting events, international borders can serve as

inconvenient barriers to tourism (Timothy & Tosun, 2003). Despite the relative ease with which

citizens of each country can cross the U.S.-Canadian border, the implicit cost of doing so

increased in 2001 and later in 2009.v Past research has found that these implicit travel costs

impact border crossings and trade (Ferris, 2010; Globerman & Storer, 2009), though there have

been conflicting results (Burt, 2009). This increase in implicit travel costs, paired with possible

loyalty to homeland-based teams, could affect the breadth of the market for the Sabres. We

therefore also assess the impact of the passport requirement—assumed to make border crossings

more cumbersome—on Canadian fan travel to Sabres games. We detail our data modeling

procedure in the following section. Results are presented with a discussion of their respective

economic significance in Section 4, and we round out our analysis with conclusions in and

suggestions for future research in Section 5.

3 Data & Methods

3.1 Data

Game dates and attendance numbers for the NHL were collected from The Hockey

Summary Project (2011), while daily border bridge-crossing data came from the Niagara Falls

Bridge Commission (2011) and the Peace Bridge Historical Traffic Statistics (2011) website.

6

The historical S&P/TSX Composite average (Yahoo, 2013) and Dow Jones Industrial Average

(DJIA; Measuring Worth, 2012) in U.S. dollars was used as a proxy for general economic

stability and growth and indexed by the yearly U.S. inflation rate (Bureau of Labor Statistics,

2013). Exchange rates (Bank of Canada, 2012) were also included in the model. Data were

collected on border crossings from May 2003 through August 2011, and include 7 NHL seasons

(no games were played in the 2004-2005 season as a result of labor discord between owners and

players). Summaries of passenger car and truck crossings by day and month can be found in

Tables 1 and 2. Details on game days for both the Sabres and Maple Leafs can be found in Table

3.vi We note that only passenger cars were counted—commercial traffic was used for a

falsification test—omitting bus traffic from our analysis.vii

We use data from four bridges in the Buffalo area as the dependent variable: the Peace

Bridge, Rainbow Bridge, Whirlpool Bridge, and Lewiston-Queenstown Bridge. The focus was

on cars entering the United States. Commercial trucks, busses, and passenger cars are recorded

separately within the Peace Bridge data. Data collected at the other three bridges note only the

total number of vehicle crossings of any kind for inbound U.S. traffic. We therefore used the

U.S. inbound totals prorated by the proportion shown in the Canada-bound traffic, which were

reported separately, as an estimate of passenger car crossings for these bridges. We also fit the

model to only Peace Bridge traffic to ensure that there were not strong influences of this

estimation of the data, and found similar results with respect to the scale of the coefficients for

inbound traffic. The data aggregated across all bridges were used in the model presented here.viii

The regression model contains dummy controls for the year and day of the week, a proxy

for economic wellbeing within Canada (S&P/TSX Composite), this measure relative to

economic wellbeing in the U.S. (DJIA subtracted from S&P/TSX Composite), and the U.S.-

7

Canadian exchange rate. In order to evaluate the impact of travel cost changes, a dummy

variable for whether a passport is necessary for crossing the border was included in our

regression. Additionally, we used U.S. (YCharts, 2012) and Toronto area (Statistics Canada,

2013b) average monthly per-gallon gasoline prices. We transformed Canadian gas prices to U.S.

dollars-per-gallon in real terms and took the difference between the Toronto and U.S. prices

yielding a relative gas price for Canadian consumers in the model in addition to the absolute

prices paid by those in the Toronto area.

The primary variable of interest was an indicator for the dates of Sabres home games.

The game day dummy variable was interacted with the passport requirement variable to evaluate

decreases specific to hockey travelers as a result of the implicit increase in travel costs. We

include a variable to indicate the impact of the opponent team’s home location, used to evaluate

the effect of hosting a Canadian team on cross border travel and the possibility of rivalry impacts

of interest to Canadian fans (Paul, 2003). This adds two dummy variables to the model: one for

a Canadian opponent, and another unique to the Toronto Maple Leafs. These indicators are not

mutually exclusive, and therefore the total impact of a Sabres game against the Maple Leafs, for

example, would be the sum of the coefficients related to all three variables. In this way, the

economic significance of the difference in opponent impacts can be interpreted by evaluating this

sum. The playoff game indicator is constructed in a similar fashion for all models.

Team quality and cost of attendance were measured relative to the current alternative for

NHL fans in the Niagara Peninsula: the Toronto Maple Leafs.ix To calculate relative cost of

attendance, we used the per-person Fan Cost Index (FCI) from Team Marketing Report (Fort,

2012) in each year and took the difference in the Toronto and Buffalo FCI.x We measured the

relative quality of the Sabres and Maple Leafs in the given season using the difference in each

8

teams’ percentage of possible standings points gained in the given season. We note that relative

FCI and relative quality variables were included in the regression only as an interaction with the

occurrence of a Sabres game. These variables should have no impact on border crossings on

non-game days, and therefore these main effects were not of interest. Finally, an indicator for

concurrent Toronto Maple Leafs home games was used. This was done to control for the

possible competing effects of having two games on the same days or nights within driving

distance for fans. Lastly, we initially included a measure of uncertainty of outcome for each

game employing money line odds transformed into win probabilities for the Sabres, but these

were ultimately excluded as they were highly correlated with the team quality measures for the

Sabres, and were not statistically significant. Descriptions for all variables in the model are

included in the Appendix. xi

A nearly identical model to the one for passenger cars was estimated for commercial

truck traffic in order to evaluate the validity of the passenger car estimations. One would assume

that the occurrence of a hockey game would not impact the number of truck crossings into the

U.S.xii This falsification test confirmed that only the number of passenger cars was impacted by

a scheduled Buffalo Sabres home game and its respective characteristics. We posit that this

result confirms that our Sabres Game dummy variable is not measuring some other phenomenon

correlated with overall border crossings from Canada to the United States.

3.2 Modeling Approach

The stationarity of our time series data was assessed to ensure coefficient estimations in

the model were not adversely impacted by unit roots (Davies, Downward, & Jackson, 1995;

Dobson & Goddard, 2001). Three different unit root tests were employed for robustness (Table

4).xiii Each of the two border crossing series of interest—passenger cars and commercial

9

trucks—were found to be stationary. Given this, a generalized additive model (GAM) was used

to evaluate the change in border crossings on Buffalo Sabres game days. The GAM allows a

cyclical (periodic) cubic spline to control for seasonality in border crossing data. This spline

uses the day of the year in order to control for any seasonality in the data—or holiday travel—

and its cyclic nature allows for the endpoints of the year (January 1st and December 31st) to meet

continuously across the years in the data set. This allows for a curvilinear estimation of seasonal

effects within the data. The unknown functional form of this non-parametric component was

estimated through the generalized cross-validation procedure in order to avoid over-fitting to the

data (see Wood, [2000; 2003; 2004; 2006; 2011] for a full explanation of the generalized

additive model estimation). The non-parametric seasonal parameter estimate is additively

combined with a linear model for the parametric coefficient estimates of interest. There are no

traditional parametric coefficients for the seasonal component, as the structure of seasonality

itself is not the main inquiry in this work. The dependent variable in the model is the number of

passenger cars crossing into the U.S. on a given day. The general structure of the generalized

additive model is as follows:

𝑦𝑡 = 𝑿𝒕𝜷 + 𝑓(𝑍𝑡) + 𝜀𝑡.

In this specification, 𝑦𝑡 represents the number of passenger cars crossing into the U.S. on day t.

The terms 𝑿𝒕 and 𝜷 are vectors of the predictor variables and their parametrically estimated

coefficients, respectively, while 𝑓(𝑍𝑡) is a non-parametric spline function of the seasonality

estimated with cyclic cubic splines and generalized cross-validation. This function enters

additively into the linear model, and 𝑍𝑡 is operationalized as the numeric day of the year (from 1

10

through 366). Finally, 𝜀𝑡 represents the error term in the regression model. We note that this

specification excludes monthly fixed effects, given that they are implicit in the day of the year

variable used for the non-parametric component. Specifically, our model estimates the following

equation (where j indexes the year):

𝐶𝑟𝑜𝑠𝑠𝑖𝑛𝑔𝑠𝑡 = 𝛽0 + 𝛽1𝑆𝑃𝑇𝑆𝑋𝑡 + 𝛽2(𝑆𝑃𝑇𝑆𝑋 − 𝐷𝐽𝐼𝐴𝑡) + 𝛽3𝐸𝑥𝑐ℎ𝑎𝑛𝑔𝑒𝑅𝑎𝑡𝑒𝑡 + 𝛽4𝑇𝑜𝑟𝑜𝑛𝑡𝑜𝐺𝑎𝑠𝑡+ 𝛽5𝐺𝑎𝑠𝐷𝑖𝑓𝑓𝑡 + 𝛽6𝑃𝑎𝑠𝑠𝑝𝑜𝑟𝑡𝑡 + 𝛽7𝐿𝑒𝑎𝑓𝑠𝐺𝑎𝑚𝑒𝑡 + 𝛽8𝑆𝑎𝑏𝑟𝑒𝑠𝐺𝑎𝑚𝑒𝑡+ 𝛽9𝑃𝑙𝑎𝑦𝑜𝑓𝑓𝐺𝑎𝑚𝑒𝑡 + 𝛽10𝐶𝑎𝑛𝑎𝑑𝑖𝑎𝑛𝑂𝑝𝑝𝑡 + 𝛽11𝑇𝑜𝑟𝑜𝑛𝑡𝑜𝑂𝑝𝑝𝑡 + 𝛽12𝐹𝐶𝐼𝐷𝑖𝑓𝑓𝑡∗ 𝑆𝑎𝑏𝑟𝑒𝑠𝐺𝑎𝑚𝑒𝑡 + 𝛽13𝑄𝑢𝑎𝑙𝐷𝑖𝑓𝑓𝑡 ∗ 𝑆𝑎𝑏𝑟𝑒𝑠𝐺𝑎𝑚𝑒𝑡+ 𝛽14𝑆𝑎𝑏𝑟𝑒𝑠𝐺𝑎𝑚𝑒 ∗ 𝑃𝑎𝑠𝑠𝑝𝑜𝑟𝑡𝑡 + 𝛽15𝑆𝑎𝑏𝑟𝑒𝑠𝐺𝑎𝑚𝑒 ∗ 𝑇𝑜𝑟𝑜𝑛𝑡𝑜𝐺𝑎𝑠𝑡+ 𝛽16𝑆𝑎𝑏𝑟𝑒𝑠𝐺𝑎𝑚𝑒 ∗ 𝐿𝑒𝑎𝑓𝑠𝐺𝑎𝑚𝑒𝑡 + 𝛽17−22𝐷𝑎𝑦𝑂𝑓𝑊𝑒𝑒𝑘𝑡 + 𝛽23−30𝑌𝑒𝑎𝑟𝑡+ 𝑓�𝐷𝑎𝑦𝑂𝑓𝑌𝑒𝑎𝑟𝑗� + 𝜀𝑡

4 Results and Discussion

The results of the regression models for passenger cars and commercial trucks can be

found in Table 5. Day of week and year fixed effectsxiv estimates are found in Table 6 for both

models, and the seasonality and data fit are visualized in Figure 2.

4.1 Control Variables of Interest

Beginning with our control variables, weekends show increased passenger car crossings,

while fewer commercial trucks cross the border on Saturdays and Sundays. The estimates for

economic prosperity and exchange rates also impact border crossings in expected ways. The

more prosperous the S&P/TSX Composite index the more likely Canadians were to travel across

the border. As the DJIA grows relative to the S&P/TSX, it is also more likely that Canadians

would travel into the more prosperous economic climate. An increase in the price of gas in the

Toronto area was associated with more people crossing the border into the United States. This

11

counterintuitive result may be explained by choices of those Canadians living near the border to

travel to the Buffalo area for shopping and entertainment, rather than traveling to the more

distant market and venues in Toronto. Those near the border, while subject to border crossing

impediments, would require less travel time and expense to get to Buffalo than other areas in

Southern Ontario. Additionally, we find that the relative price of gas significantly impacts

border crossings, with Canadians traveling to the United States more when gas prices are lower

within the U.S. relative to Canada. Border crossings into the U.S decreased by nearly 9 percent,

or approximately 1,120 passenger cars per day, after passport requirements were implemented at

the border. However, commercial trucks were not negatively impacted by this new legislation,

indicating that there was no discernible impact on commercial trade. The positive coefficient

estimated for this may be tied to the ability for drivers to use the Free and Secure Trade (FAST)

pass as official identification in lieu of presenting a passport when crossing the border near the

end of 2009.

4.2 Canadian Interest in the Sabres

The coefficient estimate for Sabres home games implies an increase of about 1,150

additional passenger cars traveling into the U.S. near the Buffalo border on game days.

Interestingly, playoff games do not garner further attention than regular season games from

Canadian fans, and while not statistically significant, the sign of the coefficient is negative. It

may be that playoff games create more excitement for those fans living within the U.S.,

crowding out tickets available for Canadians. When Buffalo’s opponents are divided into U.S.-

based opponents, Canadian opponents, and the Maple Leafs, rather large differences in the

propensity for Canadians to travel to attend Sabres games were observed. Those traveling to the

games across the border do so at a higher rate when the opponent is from Canada, and especially

12

so for the Maple Leafs. More specifically, a Canadian opponent adds greater than 600 more cars

crossing the border than a U.S. opponent, and the Maple Leafs add nearly another 970 cars.xv

This means that while a U.S.-based opponent would result in about 1,150 additional passenger

cars, a non-Toronto-based Canadian team would attract more than 1,750 cars, and the Maple

Leafs attract over 2,700 passenger cars into the U.S. The weighted average of these estimates

based on the number of games against each opponent category is nearly 1,400 per Buffalo home

game day.xvi Finally, the passport requirement did not affect the number of passenger cars

crossing the border that were specifically associated with attendance of Sabres games. This

suggests that the interest in NHL hockey among Canadians has remained relatively unaffected by

the implicit border crossing cost increases.

4.3 Fan Substitution Estimates

Moving to the relative cost of attending games in Toronto and Buffalo, the larger the

discrepancy between the cost of attending a Toronto game (higher) relative to a Buffalo game

(lower), the more Canadians that crossed the border into the U.S. on Sabres game days. The

coefficient estimate equates to approximately 20 additional cars crossing the border with each $1

change in the difference in cost of attendance of the two teams. This impact may be particularly

important for the placement of a new team even closer to Buffalo that is likely to have lower

prices than the Maple Leafs, but requires less hassle crossing an international border. There was

also a significant effect for relative quality of Buffalo and Toronto, implying that the better the

Sabres are relative to the Maple Leafs, the more Canadians that cross the border to attend

Buffalo games. Specifically, an improvement in Buffalo’s points percent by one win relative to

Toronto is estimated at 36 additional cars.xvii

13

The indicator variable for Toronto Maple Leafs home game dates is associated with an

additional 536 border crossings into the United States. At first glance, this seems counter-

intuitive. However, we note that it is likely that people living in the United States that are

attending Maple Leafs games return to the U.S. on the same day as the game was attended.

Therefore, this indicator is likely accounting for those returning from a Maple Leafs home game.

Ultimately, this result highlights a possibility for future research into the U.S. fan base for the

Toronto Maple Leafs. The interaction of a Toronto and Buffalo game—indicating a concurrent

game on the given date—was not significant, but given the direction, could indicate some

equalizing effect in combination with the two main effects of Maple Leafs and Sabres home

games. Again, this effect could be biased downward if those traveling into Canada are less likely

to return on the same day as those traveling into the U.S.

4.4 Economic Significance of Canadians and Fan Substitution

Using the coefficient estimates from our model, further assumptions regarding the make-

up of cars crossing into the U.S. for a hockey game were made. For our discussion here, we

made a conservative assumption that each car with fans going to a Sabres game contained two

passengers (including the driver). We also provide a sensitivity analysis on our assumption of

passengers in Table 7, varying the estimate from a single passenger to three passengers per car.

Multiplying our two passengers per car assumption by the weighted coefficient estimate of all

three Sabres game-related dummy variables implies that approximately 2,776 fans cross the

border from Canada to attend a Sabres home game. Based on this estimate, 15.4 percent of fans

at Sabres games made the trip from Canada, an estimate essentially equal to the claims made by

Sabres President, Ted Black, regarding the percentage of Sabres season ticket holders residing in

Canada (Zeisberger, 2011).xviii The visiting team-specific country coefficients estimate that

14

more than one-fifth of the fans attending Sabres games against a Canadian team—and more than

one-third of those attending games with the Toronto Maple Leafs—came from Canada. Using

Team Marketing Report’s Fan Cost Index (FCI)xix, Canadian attendees are estimated to account

for nearly $6.57 million in attendance related revenues each season, or approximately 8.1 percent

of yearly team revenues (Forbes, 2011).xx

Combining the relative price and quality variables for Toronto and Buffalo with the

assumption of two passengers per car leaves these two estimates at 40 and 72 additional fans,

respectively. Assuming only a single $1 change in the price differential is generated by an

increase in Toronto prices, and only a half win change in opposite directions for each of the two

teams results in the incremental points percent differential, these increases are associated with a

total of approximately $265,000 per year for the Sabres.

Given these estimates, redirecting any significant portion of this revenue to a new team in

Hamilton would likely severely impact revenues for the Sabres, validating the team’s assertion

that a second team in the Greater Toronto area would have a negative effect.

4.5 Expansion and Relocation Implications

With these results come important implications for expansion, particularly when the sport

in question is entrenched in the culture of the area (Mason, 2002). Concerns with potential

losses in revenue and fans are not without merit. Recent history has produced disagreements

over the rights to areas near an incumbent professional team due to rather favorable antitrust

treatment of sports leagues. The same could be said for any place that borders another team’s

exclusive market area irrespective of whether they are in the same country. Professional sports

leagues like the NHL and MLB have historically supported plans to compensate teams when

their market areas have overlapped with those supporting the new franchise. For example, in

15

1992, the owners of the Anaheim Mighty Ducks (the Disney Corporation) agreed to compensate

the Los Angeles Kings, with half of a $50 million expansion fee paid directly to the Kings

(Lapointe, 1992). More recently, MLB’s Oakland Athletics have experienced resistance from

the San Francisco Giants with respect to their proposal to move to San Jose.

As discussions continue regarding the placement of an NHL team near Buffalo, the

impact to the Sabres should be an important consideration with respect to exclusive market rights

and compensation for them as an incumbent team. Our results highlight that this is the case

despite being across an international border. While we cannot directly address fan loyalty to a

new home-country team, the finding that there are measureable changes in border crossings

depending on the home country of the visiting team in Buffalo highlights that there could be

some nationalistic preference in this context. However, this estimate would also require an

incorporation of the extent to which any short-term losses were offset by increased fan interest

from new rivalry between the three nearby teams, which could ultimately elevate their longer-

term revenues.

5 Conclusions

We address the structure of the Buffalo Sabres’ market for home attendance from fans

living across international borders and into Canada, and the choice made between attending

Sabres games rather than Toronto Maple Leafs games. We find that Buffalo’s Canadian fan base

provides a significant portion of gate revenues. In addition, the relative price and quality of the

Sabres and Maple Leafs impacted border crossings associated with game attendance. This study

provides new evidence of within-league fan substitution adding to previous efforts, with

important insight into future inquiries with regard to fan substitution. For example, a new team

could cannibalize fans that usually cross the border to see hockey. Further research is necessary

16

to measure the effect of fan interest in rivalries induced by this proximity. It may also be

relevant to investigate cross-border fandom in other areas such as Detroit, Seattle, and

Vancouver. Additionally, future investigation could evaluate cross-league substitution across

international borders. For example, the Buffalo Bills and Buffalo Sabres may compete with the

Maple Leafs (as noted here) or other Canadian teams such as MLB’s Toronto Blue Jays, the

National Basketball Association’s Toronto Raptors, or the Hamilton Tiger Cats and Toronto

Argonauts of the Canadian Football League. This would be relevant in the context of both gate

receipts, as well as television demand for professional sport (Tainsky & McEvoy, 2012).

17

References Andrijiw, A.M. & Hyatt, C.G. (2009). Using optimal distinctiveness theory to understand

identification with a nonlocal professional hockey team. Journal of Sport Management, 23, 156-181.

Breunig, J. (2012). “New Media Summit,” Sabre Noise. Retrieved June 30,

2012 from: http://sabrenoise.com/2012/06/12/new-media-summit-2/. Bank of Canada. (2012). Daily noon exchange rates: 10-year lookup. Retrieved January 20,

2012 from: http://www.bankofcanada.ca/rates/exchange/10-year-lookup/. Bureau of Labor Statistics. (2013). CPI Inflation Calculator. Retrieved August 1, 2013 from:

http://www.bls.gov/data/inflation_calculator.htm. Burt, M. (2009). Tighter border security and its effect on Canadian exports. Canadian Public

Policy, 35, 149-169. Davies, B., Downward, P. & Jackson, I. (1995). The demand for rugby league: Evidence from

causality tests. Applied Economics, 27, 1003-1007. Dobson, S. & Goddard, J. (2001). The economics of football. Cambridge, UK: Cambridge

University Press. Elliot, G., Rothenberg, T.J., & Stock, J.H. (1996). Efficient tests for an autoregressive unit root.

Econometrica, 64, 813-836. Ferris, J.S. (2010). Quantifying non-tariff trade barriers: What difference did 9/11 make to

Canadian cross-border shopping? Canadian Public Policy, 36, 487-501. Forbes. (2011). NHL Team Valuations: #21 Buffalo Sabres. Retrieved November, 2011 from:

http://www.forbes.com/lists/2010/31/hockey-valuations-10_Buffalo-Sabres_313362.html.

Fort, R. (2012). Rodney Fort’s Sports Business Data: NHL Fan Cost Index. Retrieved July 20, 2012 from https://sites.google.com/site/rodswebpages/codes.

Gitter, S.R. & Rhoads, T.A. (2011). Determinants of minor league baseball attendance. Journal

of Sports Economics, 11, 614-628.

18

Globerman, S. & Storer, P. (2009). Border security and Canadian exports to the United States: Evidence and policy implications. Canadian Public Policy, 35, 171-186.

Hyatt, C.G. & Andrijiw, A.M. (2008). How people raised and living in Ontario became fans of

non-local National Hockey League teams. International Journal of Sport Management and Marketing, 4, 338-355.

Henrickson, K. (2012). Spatial competition and strategic firm location. Economic Inquiry,

50(2), 364-379. Jackson, S.J. (1994). Gretzky, crisis, and Canadian identity in 1988: Rearticulating the

Americanization of culture debate. Sociology of Sport Journal, 11, 428-446. James, J.D., Kolbe, R.H., & Trail, G.T. (2002). Psychological connection to a new sport team:

Building or maintaining the consumer base? Sport Marketing Quarterly, 11, 215-225. Jones, J.C.H. & Ferguson, D.G. (1988). Location and survival in the National Hockey League.

The Journal of Industrial Economics, 36, 443-457. Keller, T. & McGuire, N. (2011). The new economics of the NHL: Why Canada can support 12

teams. Report of the Mowat Policy Innovation Centre, School of Public Policy & Governance, University of Toronto, April 11, 2011. ISBN: 978-0-9867464-7-5.

Lapointe, J. (1992). NHL to add teams in Miami, Anaheim: Huizenga, Disney high-profile

owners. Retrieved March 11, 2014 from http://articles.baltimoresun.com/1992-12-11/sports/1992346112_1_huizenga-miami-team-hockey.

Lock, D., Taylor, T., & Darcy, S. (2011). In the absence of achievement: The formation of new

team identification. European Sport Management Quarterly, 11, 171-192. Mason, D. (2002). “Get the puck outta here!”: Media transnationalism and Canadian identity.

Journal of Sport and Social Issues, 26, 140-167. Matheson, V., Peeters, T., & Szymanski, S. (2012). If you host it, where will they come from?

Mega-events and tourism in South Africa. University of Antwerp, Faculty of Applied Economics Working Paper. Retrieved from http://ideas.repec.org/p/ant/wpaper/ 2012015.html on October 20, 2012.

Measuring Worth. (2012). Daily closing values of the Dow Jones Average in the United States,

May 2, 1885 to present. Retrieved January 20, 2012 from: http://measuringworth.com/ DJA/index.php.

19

NA. (2011). The Peace Bridge Historical Traffic Statistics. Retrieved September 2, 2011 from: http://www.peacebridge.com/index.php?option=com_wrapper&view=wrapper&Itemid=687.

NA. (2011). Niagra Falls Bridge Commission Traffic Statistics. Retrieved September 2, 2011

from: http://niagarafallsbridges.com/traffic_statistics.php3.

NA. (2011). The Hockey Summary Project. Retrieved September 5, 2011 from: http://sports.groups.yahoo.com/group/hockey_summary_project/.

NHL. (2009). NHL By-Laws. In U.S. Bankruptcy Court for the District of Arizona. (2009).

Case Number 2:09-bk-09488-RTBP, declaration of William L. Daly. Paul, R.J. (2003). Variations in NHL attendance: The impact of violence, scoring, and regional

rivalries. American Journal of Economics and Sociology, 62, 345-363. Rascher, D.A., Brown, M.T., Nagel, M.S., & McEvoy, C.D. (2009). Where did National Hockey

League fans go during the 2004-2005 lockout? An analysis of economic competition between leagues. International Journal of Sport Management and Marketing, 5, 183-195.

Statistics Canada (2013a). Focus on Geography Series, 2011 Census: Census metropolitan area

of St. Catharines – Niagara, Ontario. Retrieved August 6, 2013 from: http://www12. statcan.gc.ca/ census-recensement/2011/as-sa/fogsspg/Factscmaeng.cfm?LANG=Eng& GK=CMA&GC=539.

Statistics Canada. (2013b). Gasoline and fuel oil, average retail prices by urban centre.

Retrieved August 6, 2013 from: http://www.statcan.gc.ca/tables-tableaux/sum-som/l01/cst01/econ154a-eng.htm.

Tainsky, S., & McEvoy, C.D. (2012). Television broadcast demand in markets without local

teams. Journal of Sports Economics, 13, 250-285. Timothy, D.J. & Tosun, C. (2003). Tourists’ perceptions of the Canada-USA border as a barrier

to tourism at the International Peace Garden. Tourism Management, 24, 411-421. Winfree, J.A. (2009). Fan substitution and market definition in professional sports leagues.

The Antitrust Bulletin, 54, 801-822. Winfree, J.A. & Fort, R. (2008). Fan substitution and the 2004-05 NHL lockout. Journal of

Sports Economics, 9, 425.434. Winfree, J.A., McCluskey, J.J., Mittelhammer, R.C., & Fort, R. (2004). Location and attendance

in Major League Baseball. Applied Economics, 36, 2117-2124.

20

Wood, S.N. (2011). Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. Journal of the Royal Statistical Society (B), 73, 3-36.

Wood, S.N. (2006). Generalized additive models: An introduction with R. Chapman Hall, Taylor & Francis Group, LLP: Boca Raton, FL.

Wood, S.N. (2004). Stable and efficient multiple smoothing parameter estimation for generalized additive models. Journal of the American Statistical Association, 99, 673-686.

Wood, S.N. (2003). Thin-plate regression splines. Journal of the Royal Statistical

Society (B), 65, 95-114.

Wood, S.N. (2000). Modelling and smoothing parameter estimation with multiple quadratic penalties. Journal of the Royal Statistical Society (B), 62, 413-428.

Yahoo. (2013). Yahoo Finance: S&P/TSX composite Index. Retrieved August 6, 2013 from:

http://finance.yahoo.com/q/hp?s=^GSPTSE&a=00&b=1&c=2003&d=06&e=1&f=2013&g=d

YCharts. (2012). U.S. Retail Gas Price. Retrieved January 20, 2012 from:

http://ycharts.com/indicators/gas_price. Zeisberger, M. (2011). “Sabres banking on Canadian fans,” Toronto Sun. Retrieved June 30,

2012 from: http://www.torontosun.com/sports/hockey/2011/04/15/18013671.html.

21

TABLE 1: Average Daily Passenger Automobile Crossings (2003-2011) Peace Whirlpool Lewiston-Queenstown Rainbow Total

January 5,252 286 2,689 3,218 11,445 February 5,737 293 2,841 3,408 12,279 March 6,351 333 3,465 3,804 13,953 April 6,274 380 3,702 4,282 14,638 May 7,045 487 4,043 4,530 16,105 June 7,873 524 4,158 5,060 17,615 July 9,816 627 4,866 6,390 21,699 August 10,217 676 5,106 6,760 22,759 September 7,785 529 4,044 5,039 17,397 October 7,172 423 3,818 4,757 16,170 November 6,599 355 3,712 4,249 14,915 December 5,994 330 3,290 3,814 13,428 Peace Whirlpool Lewiston-Queenstown Rainbow Total Sunday 8,089 449 4,696 5,741 18,975 Monday 6,647 418 3,664 4,231 14,960 Tuesday 6,137 413 3,160 3,759 13,469 Wednesday 6,284 424 3,175 3,840 13,723 Thursday 6,843 427 3,516 4,127 14,913 Friday 8,439 510 4,298 4,989 18,236 Saturday 8,314 462 4,417 5,940 19,133 Peace Whirlpool Lewiston-Queenstown Rainbow Total 2004 7,741 232 3,706 4,836 16,515 2005 7,643 290 4,116 4,912 16,961 2006 7,582 299 3,997 4,835 16,713 2007 7,254 395 4,001 4,663 16,313 2008 6,872 435 3,756 4,441 15,504 2009 6,498 465 3,506 4,004 14,473 2010 6,507 559 3,547 4,393 15,006

22

TABLE 2: Average Daily Commercial Truck Crossingsa

Month Averagea Week Day Average Year Average

January 2,646 Sunday 1,392 2004 2,965 February 2,821 Monday 3,254 2005 3,046 March 2,962 Tuesday 3,749 2006 2,951 April 2,880 Wednesday 3,864 2007 2,957 May 2,931 Thursday 3,720 2008 2,796 June 3,011 Friday 2,792 2009 2,441 July 2,628 Saturday 965 2010 2,549 August 2,803 September 2,922 October 2,922 November 2,847 December 2,461

a. Includes all bridges.

23

TABLE 3: Summary of Game Dates Buffalo Toronto Concurrente

Sunday 13 0 0 Monday 19 35 7 Tuesday 28 92 16 Wednesday 63 14 3 Thursday 27 30 3 Friday 88 1 0 Saturday 49 113 33 Buffalo Toronto Concurrent

January 37 42 9 February 46 40 8 March 55 52 9 April 12 19 5 May 0 0 0 June 0 0 0 July 0 0 0 August 0 0 0 Septembera 0 0 0 October 35 43 10 November 52 42 5 December 50 47 16 Buffalo Toronto Concurrent

2003b 19 15 4 2004c 22 26 5 2005 19 22 4 2006 41 40 11 2007 40 37 9 2008 44 39 7 2009 42 44 12 2010 38 41 7 2011d 22 21 3 a. We do not include preseason games in our data. b. Our sample begins in May of 2003, after the 2002-2003 season. c. The 2004-2005 season was cancelled due to the owners’ lockout of the players. d. Our sample ends just before the 2011-2012 season. e. All concurrent games are also accounted for within each team’s individual counts.

24

TABLE 4: Unit Root Test for Daily Canada to U.S. Border Crossing Passenger Cars Commercial Trucks

Unit Root Test No Trend Trend No Trend Trend

ADFa -5.191 (8)*** -5.264 (8)*** -4.763 (7)*** -6.322 (6)***

PPb -16.807 (8)*** -17.126 (8)*** -11.525 (8)*** -13.422 (8)***

DFGLSc -3.292 (8)*** -4.274 (8)*** -25.939 (8)*** -26.107 (8)*** “***” denotes rejection of the presence of a unit root at the 99% confidence level. a. Lag chosen by the Schwartz Information Criterion. b. Lag chosen by the Newey-West procedure. c. Lag chosen by the modified AIC method.

25

TABLE 5: Regression Estimates for Passenger Cars and Commercial Trucks

“***”, “**”, “*” refer to 99%, 95%, and 90% significance levels. Dollar amounts in 2011 constant U.S. dollars. a. 000’s b. Coefficient refers to change in crossings with a one-cent increase in the U.S.-to-Canadian exchange rate. c. Coefficient for one cent change in price. d. Coefficient is interpreted as the impact of these game-specific variables only on game days.

Passenger Cars Commercial Trucks

Constant 14284.4*** 2144.3*** (2473.93) (486.10)

S&P/TSX Comp.a 155.3** 25.5* (72.47) (14.24)

(S&PTSX – DJIA)a -207.0** 16.5 (102.87) (20.21)

Exchange Rateb 26.5* 6.4** (15.64) (3.07)

Toronto Gas Pricec 5.7*** 0.4 (1.79) (0.35)

Relative Gas Pricec 9.3*** 0.07 (3.03) (0.60)

Passport -1117.7*** 155.6*** (215.80) (42.40)

Sabres Game 1154.4*** 5.0 (190.05) (37.41)

Canadian Opp. 608.1** 3.1 (259.13) (51.00)

Toronto Opp. 965.7** -28.9 (424.78) (83.66)

Playoff Game -375.5 115.5 (434.96) (85.58)

Sabres × Relative FCId 19.7*** 1.0 (7.27) (1.43)

Sabres × Relative Qualityd 2.9* 0.2 (1.45) (0.29)

Leafs Game 536.1*** -21.6 (133.34) (26.23)

Sabres × Passport -646.1 -40.3 (423.10) (83.27)

Sabres × Tor. Gas Price -0.6 -0.3 (2.23) (0.44)

Sabres × Leafs -371.2 30.9 (286.87) (46.49)

Adjusted R2 0.870 0.924

N 3,045 3,045

26

TABLE 6: Year and Week Day Fixed Effects Estimates Passenger Cars Commercial Trucks

Sundaya 975.3*** -1414.2*** (117.98) (23.19)

Monday -3092.9*** 448.7*** (118.17) (23.24)

Tuesday -4671.8*** 951.9*** (120.07) (23.61)

Wednesday -4451.5*** 1071.6*** (115.94) (22.79)

Thursday -3155.0*** 932.2*** (117.12) (23.01)

Saturday 924.0*** -1832.0*** (120.83) (23.75)

2004b -1208.9*** -127.5*** (155.92) (30.63)

2005 -876.3*** -84.9** (198.97) (39.10)

2006 -1231.5*** -226.2*** (264.28) (51.93)

2007 -1924.3*** -256.0*** (284.92) (55.99)

2008 -2516.4*** -409.5*** (358.79) (70.50)

2009 -2278.0*** -787.5*** (437.50) (85.97)

2010 -1776.0*** -777.8*** (464.20) (91.22)

2011 -1678.6*** -781.1*** (463.21) (91.02)

“***”, “**”, “*” refer to 99%, 95%, and 90% significance levels, respectively. a. Day coefficients imply change relative to Fridays. b. Year coefficients imply change relative to 2003.

27

TABLE 7: Sensitivity Analysis of Estimates of Attendees per Passenger Car

U.S. Opponent Canadian Opponent Toronto Maple Leafs Weighted Average % of Attendance

𝜷𝑼.𝑺. = 𝟏,𝟏𝟓𝟒.𝟒 𝜷𝑪𝒂𝒏 = 𝟔𝟎𝟖.𝟏 𝜷𝑻𝒐𝒓 = 𝟗𝟔𝟓.𝟕 𝜷𝑾 = 𝟏,𝟑𝟖𝟖.𝟏 𝑨𝒗𝒈.𝑨𝒕𝒕𝒆𝒏𝒅 = 𝟏𝟖,𝟎𝟒𝟕

Passengers 𝜷𝑼.𝑺. + 𝜷𝑪𝒂𝒏 = 𝟏,𝟕𝟔𝟐.𝟓 𝜷𝑼.𝑺. + 𝜷𝑪𝒂𝒏 + 𝜷𝑻𝒐𝒓 = 𝟐,𝟕𝟐𝟖.𝟐

1 1,154 1,762 2,728 1,388 7.7

1.5 1,732 2,644 4,092 2,082 11.5

2 2,309 3,525 5,457 2,776 15.4

2.5 2,886 4,406 6,821 3,470 19.2

3 3,463 5,288 8,186 4,164 23.1

28

FIGURE 1: 50-Mile Radius for Buffalo and Toronto

29

FIGURE 2: Seasonal Regression Components and Border Crossings (2003-2011)

30

Appendix Crossings The daily number of passenger car (truck) crossings into the United

States SPTSX Daily closing S&P/TSX Composite Index in 000s in inflation

adjusted U.S. 2011 dollars (centered) SPTSX – DJIA Difference in the daily closing S&P/TSX Composite and daily

closing Dow Jones Industrial Average in 000’s in inflation adjusted U.S. 2011 dollars; (SPTSX minus DJIA)

ExchangeRate U.S. – Canada Exchange Rate TorontoGas Toronto average monthly regular gas price in U.S. 2011 dollars GasDiff Difference in absolute monthly gas price between Toronto and U.S.

average gas price per gallon, in inflation adjusted 2011 US$0.01; (Toronto minus U.S. Average)

Passport Indicator that passport is required to enter U.S. SabresGame Indicator that there is a Sabres home game on the given day PlayoffGame Indicator that the Sabres home game is a playoff game CanadianOpp Indicator that the Sabres game is against a Canadian-based NHL

opponent TorontoOpp Indicator that the Sabres game is against the Toronto Maple Leafs FCIDiff Difference in per-person Fan Cost Index between Buffalo and

Toronto (centered) in yearly inflation adjusted 2011 U.S. dollars; (Toronto minus Buffalo)

QualDiff Difference in possible standings points percent in the season of the current game for Buffalo and Toronto (centered); (Buffalo minus Toronto)

LeafsGame Indicator that there is a Maple Leafs home game on the given day SabresGame*Leafs Interaction of SabresGame and LeafsGame indicator variables SabresGame*Passport Interaction of SabresGame and Passport indicator variables SabresGame*TorontoGas Interaction of SabresGame and TorontoGas variables DayOfWeek Day of Week Fixed Effects Year Yearly Fixed Effects

31

Notes i Zeiseberger (2011) ii The authors use incoming flights to estimate travel to the Cricket and FIFA World Cups in South Africa. iii Sabres team president Ted Black: “Every team in the league has a 50 mile radius fan base”…“I guess about 45% of our potential fan base is under water or in another country,” (Breunig, 2012). iv The NHL By-Laws state that, “Any such consent by the Member Clubs may be made subject to reasonable and appropriate conditions, including payment to the League of a transfer fee”…”and/or payment of an indemnification fee (or fees) to reflect the goodwill developed by a neighboring member (or members) in the new location,” (NHL, NHL By-Laws, Section 36, Transfer of Franchise Location, 2009, pp. 115). v On June 1, 2009 restrictions increased and anyone crossing the U.S.-Canada border was required to have a passport, or qualified NEXUS or FAST pass. This was specific to land and sea travel, as airline passengers were required to have a passport when entering the U.S. beginning in January of 2007. vi We note that one of the Sabres home games was played in Ralph Wilson Stadium against the Pittsburgh Penguins, allowing many more attendees than the capacity at their arena. We examined models with and without this data point included, and it did not significantly impact our coefficient estimates. Therefore, we include this game. vii It is difficult to know the number of passengers on any individual bus. viii The results of the Peace Bridge only model are available upon request. ix Minor league hockey was not included although it is noted that fans could attend Bulldogs’ games in Hamilton as well as games involving other teams. x This variable is centered for easier interpretation. xi We also evaluated the impact of the number of Canadian born players on the Sabres roster to gauge interest in players from the home country of those crossing the border. However, there was no significant effect of this variable. We therefore leave this variable out of our analysis. xii We also fit the model using diesel gas prices, but there was no substantive difference in the results of the regression. Therefore, we proceeded with using an identical model to the passenger car regression, with regular gas prices as a proxy for overall price changes in fuel. xiii We employ the Augmented Dickey-Fuller Test (ADF), Phillips-Perron test (PP) and Dickey-Fuller Generalized Least Squares test (DFGLS; Elliot, Rothenberg, & Stock, 1996). For the ADF test, we use Schwartz Information Criterion (SIC) for lag selection, while the PP test uses the Newey-West method. The DFGLS procedure chooses optimal lags from Modified Akaike’s Information Criterion (MAIC). xiv We also fit our model with a yearly time trend in lieu of fixed effects, and found similar results. These are available upon request. xv Because the coefficients indicate team-specific effects, the overall effect is the sum of coefficients pertaining to the game characteristics. Therefore, using the point-estimates, a U.S.-based visitor attracts approximately 1,150 passenger cars from Canada, a Canadian opponent attracts approximately 1,763 cars, and the Maple Leafs attract approximately 2,728 cars. xvi In our data set, there were 229 games against U.S.-based franchises, 57 against Canadian-based franchises other than the Maple Leafs, and 24 games against the Maple Leafs in Buffalo. Therefore, the weighted average estimate is 𝛽𝑊 = (229∗1,154)+(57∗1,763)+(24∗2,728)

229+57+24≈ 1,388.1.

32

xvii A single win is worth approximately 0.0122 percentage points in the standings, or a 12.2 unit change in our transformed points percent variable. Therefore, the coefficient estimate multiplied by this change results in our additional passenger car estimate. xviii During a large portion of the period under analysis, Buffalo reported selling out nearly all of their games, with an average attendance level of 18,047 (a sold out arena is reported as 18,690 fans). We calculate percentages based on the average attendance over this period. xix FCI from the 2011-2012 season are used, and therefore estimates are in 2011 dollars. Average spending per fan in this season was $57.75 at Buffalo Sabres home games. The team plays 41 regular season home games each season. xx Team revenues were estimated at $81 million.