A tale of three cities: intra-game ratings in winning, losing, and neutral markets.
Xu, Jie ; Sung, Hojun ; Tainsky, Scott 等
Introduction
Despite the presence of other professional sport leagues including
the National Basketball Association (NBA), National Hockey League (NHL),
and Major League Baseball (MLB), the National Football League (NFL) is
widely considered the most dominant and financially viable North
American sports league. Even the presence of rival major professional
sports leagues has not diminished fan enthusiasm for the NFL, signifying
fans do not perceive there to be any viable substitute. For example,
halfway through the 2013 regular season, NFL games represented the top
18 viewed television programs since the season started in early
September (Brown, 2013).
Previous research has considered the optimal selection of games in
a handful of markets without NFL teams (Tainsky, McEvoy, & Jasielec,
2011). The authors used average game ratings and predictive models to
evaluate network choices in out-of-market game selection. In this
research we examine not only ex ante appraisals of game closeness, but
also intra-game scoring changes as a predictor of ratings dynamics.
Moreover, we test how scoring changes result in divergent outcomes
according to the relationship between the market and scoring margin. To
preview the results, the home market of the losing team and neutral
markets lose a similar proportion of the viewing audience as the final
score margin between teams increases. While the trend of lost viewership
in blowout games exists in the winning team's market, the magnitude
of lost viewership is not nearly as great.
One of the advantages of using television viewership data is the
ability to measure consumer consumption across different times. While
attendance figures can serve as one proxy to assess consumer demand,
attendance for any game is reported by a single measure. Anecdotally,
fans heading for the exits are visible as the game margin and time
remaining point to a probable victor (i.e., reduced outcome
uncertainty), yet these exits are unaccounted for in the reported
attendance figures. To our knowledge, these attendance declines at
various time increments are not routinely measured and would be
cumbersome to collect with the precision that they are tabulated for
television viewership of the same games. We therefore have opted to
utilize television ratings to determine viewer "exits"
attributable to scoring margin. This idea is an extension of
Rottenberg's (1956) uncertainty of outcome hypothesis. As
Rottenberg argued, if "there will be wide variation among teams in
the quality of play, contents will become certain, and attendance will
decline" (p. 247). The current research is novel in the
construction of a metric specifically measuring "certainty"
and "decline" as well as its comparison of the magnitude of
decline across market types.
NFL Television Contracts and Broadcast Scheduling
A brief discussion of NFL television contracts and broadcast
scheduling is provided for context. NFL franchises collectively pool
their television rights and sell those rights nationally to multiple
television networks. The only broadcast rights sold individually by
franchises are for local radio and locally distributed preseason
broadcasts. The majority of the league's games are played on
weekend afternoons, and typically just one nationally telecasted game
each on Sunday, Thursday, and Monday nights. Nationally, four networks
share television broadcasting rights along with DirecTV, which pays for
the rights to broadcast Sunday afternoon games not available on local
affiliates and sells those rights to its customers for a premium.
The overall importance of sport and particularly the NFL has
created various scenarios where major broadcasters have engaged in
bidding wars over property rights oftentimes resulting in premiums being
paid for exclusive rights fees. Furthermore, obtaining the broadcast
rights to NFL games has collectively forced competing networks to pay
premiums to avoid the adverse financial implications of not having
rights to these games (Nicholson, 2007).
Professional sport leagues including the NFL have also implemented
various policies ostensibly facilitating competitive balance.
Collectively, fewer policies specifically targeting game closeness
exist, although it is clearly a league priority to show the best games
to consumers. In NFL markets, it is assumed the most desirable game is
the local game, and consequently local games are broadcast regardless of
whether a higher quality league game is being contested simultaneously.
Still a greater amount of consideration goes into determining the best
among the out-of-market slate of games (Grimshaw & Burwell, 2014).
Furthermore, league broadcasting policy dictates that once an
out-of-market game is deemed uncompetitive, then the network can switch
to a closer, presumably more desirable game contested concurrently. By
contrast, games must be shown in the primary markets of the local
franchise through completion, regardless of score.
Literature Review
The research utilizing television ratings, as opposed to
attendance, as a proxy of demand has evolved from a static average
across the broadcast to include examination of what causes interest to
vary during broadcasts. Paul and Weinbach (2007) studied national
viewership of Monday Night Football games. They first estimated the
quantity of viewers at the start of the game according to a set of
control variables along with absolute quality of the participating
teams, expected game scoring, and relative quality of the participating
teams. The change in ratings was then estimated according to the same
controls along with the identical absolute quality measure, halftime
score differential, and total points scored during the first half. Both
in terms of expectations and actual game qualities, the research showed
fan preference for teams of high absolute quality as well as close and
high-scoring games.
In addition to game closeness, past studies have suggested that an
offensive approach is "regarded to be a more attractive
strategy" (Dewenter & Namini, 2011). This theory is not often
tested directly; however, it is largely supported by evidence from
betting markets and attendance demand. Wagering markets where bettors
have shown a tendency to strongly favor the over have been found in the
National Football League (Paul & Weinbach, 2002), Arena and NCAA
football (Paul & Weinbach, 2005), NBA basketball (Paul, Weinbach,
& Wilson, 2004), and European football feagues (Paul & Weinbach,
2009). Furthermore, scoring has also been shown to be a determinant of
attendance in a variety of leagues including Major League Baseball
(Tainsky & Winfree, 2008), Super 12 Rugby (Owen & Weatherston,
2004), and European football (Falter & Perignon, 2000; Peel &
Thomas, 1992).
Following Paul and Weinbach (2007), who examined national game
ratings, Tainsky (2010) examined television ratings for NFL games at the
city market level. Historically, while attendance has been used as a
proxy for demand in the sports economic literature, the author argued
sports programming and the growth of media warrant further analysis and
further argued that the use of television broadcasts instead of
attendance allows comparisons of demand in both the home and road
teams' markets. Collectively, several of the identical team
quality, game uncertainty, and market variables influencing attendance
in other sports also affected demand for NFL broadcasts. This idea was
extended to examine viewer demand outside of NFL markets (Tainsky &
McEvoy, 2012) as well as the influence of local games on adjacent
out-of-market
NFL games (Tainsky & Jasielec, 2014). None of the city market
level cited studies measure within-game changes in ratings, relying
exclusively on average demand across the games as the predicted value.
Outside of the NFL context, Alavy, Gaskell, Leach, and Szymanski (2010)
used within-game ratings to examine the relationship between demand and
score in English football (soccer). Like the NFL-based studies, Alavy et
al. (2010) base their initial hypotheses about audience interest on
Rottenberg's UOH. The authors found that, in addition to
uncertainty, audiences possessed distaste for games ending in draws.
Two concepts that have been readily examined in the sport
management literature examining fan behavior include basking in
reflected glory (BIRG) and cutting off reflected failure (CORF).
Specifically, basking in reflected glory represents a phenomenon derived
from the psychology literature when individuals share in the achievement
of others with whom they have an association. Cialdini et al.'s
(1976) seminal study came through in the examination of students of
seven universities who wore school apparel after football victories,
indicating connection to school and individual self-esteem.
Theoretically this is related to the idea of vicarious achievement. When
the team is unsuccessful fans may then cut off reflected failure by
dissociating themselves with the losing organization. This may be
demonstrated by fans not wearing team paraphernalia or attending events
(Campbell, Aiken, & Kent, 2004). Fans tend to sever their connection
or ties to their team due to the recent or prolonged failure (Snyder,
Lassegard, & Ford, 1986). It is to avoid negative association with a
group that has failed in order to protect one's image via the
disassociation. Measuring changes in real-time fan consumption of NFL
contests using Nielsen data could shed insights not only on which groups
of fans will CORF or BIRG, but also how they respond to the overall
competitiveness of the game they are watching.
Data and Method
The television viewership data utilized in this study represent two
years (2008 & 2009) of NFL regular season games. These years were
selected because they represent two years of viewer data prior to the
precipitous growth of DirecTV's Sunday Ticket package (i.e.,
subscriptions rose 34.5% from 2010-2013; DirecTV, 2013). Data for all
games not featuring local teams were collected in all Nielsen Local
People Meter (LPM) markets. Ratings were collected for each 15-minute
period during the contest. Thus for the average three-hour broadcast, 12
unique ratings observations were collected. In engineering a metric to
estimate audience exits, [Ratings.sub.Peak] denotes the game's
highest rating for any 15-minute interval over the course of the
broadcast. [Ratings.sub.Last] was simply the rating during the last
interval collected by the Nielsen Corporation for that game.
[Ratings.sub.Drop], analogous to fan exits, was calculated according the
following formula:
[Ratings.sub.Drop] = ([Ratings.sub.Peak] =
[Ratings.sub.Last])/[Ratings.sub.Peak]
What this variable captures is the proportion of ratings loss
between the maximum number of viewers who were tuned in for any interval
and the number of viewers watching during the game's final
interval. Mathematically if the final ratings interval was the highest
value, RatingsDrop takes the value of zero.
We distinguish between three types of markets in this research.
These are winning market, losing market, and neutral market. Winning
market denotes that of the team winning the game, while losing market
denotes the market of the losing squad at the time of each observation.
All other markets where the game was broadcast were pooled in the
neutral market category. As an example, if Chicago defeated Detroit, the
Chicago market would constitute the winning market, Detroit the losing
market, and the ratings drop of all other cities where the game was
televised would be averaged in the neutral market.
Additionally, score difference intervals were coded based on the
final margin of each game. Because the most common minimum score in
football is three points, we separated the dataset into three-point
intervals (e.g., 0-2, 3-5, 6-8,... 27-29) culminating with the final
interval of all margins 30 or greater.
In this research, we first consider the question of whether the
RatingsDrop was similar across market types and for each interval, where
the null hypothesis of no difference between market types is tested:
[H.sub.0]: The ratings drop is equal across market types.
That is, we test the pairwise comparisons across all games
[RatingsDrop.sub.N] = [RatingsDrop.sub.L]
[RatingsDrop.sub.N] = [RatingsDrop.sub.W]
[RatingsDrop.sub.L] = [RatingsDrop.sub.W]
and for each interval,
[RatingsDrop.sub.N] = [RatingsDrop.sub.L]
[RatingsDrop.sub.N] = [RatingsDrop.sub.W]
[RatingsDrop.sub.L] = [RatingsDrop.sub.W]
where i represents the interval, and N, L, and W represent neutral,
losing, and winning market, respectively.
If there exists a difference between winning and losing markets,
then that difference is not attributable to the magnitude of the score
difference, rather only whether the local side is likely to win or lose.
The magnitudes are also compared to neutral markets.
We ran additional tests specifically to address the number of
points deemed by the NFL to constitute an uncompetitive match. The
league has adopted a policy that margins of 18 and greater are
uncompetitive such that the network can switch to a more competitive
out-of-market game outside of primary markets. We therefore tested the
effect of games with such a point disparity against other margins to
gauge whether the short-term ratings effect was statistically different
than other ratings drops. This portion of the analysis was restricted to
neutral markets in corresponding to league policy.
[H.sub.0]: RatingsDrop in games that the margin of victory is 18+
is equal to other intervals. That is, [RatingsDrop.sub.M] =
[RatingsDrop.sub.N], where i- represents intervals less than 18. To test
these hypotheses, we used one-way ANOVA with Tukey's studentized
range test to examine whether the mean differences were significant.
Finally, we examined the impact of final score margin on ratings
according to market type. In addition to RatingsDrop, two additional
measures were created as proxies of change in audience.
WholeGameDifference represented the absolute difference in viewership
between ratings at the beginning and end of the game, while
SecondHalfDifference described viewership difference just during the
second half of the broadcast. These were calculated by the formulae:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
An important difference between these measures and RatingsDrop is
that the Difference variables can take on negative values if the
audience size was larger in the final ratings interval. In so doing the
difference variables possess less of a direct parallel to the
'exits' described by Rottenberg, but nonetheless may add
richness to the examination of the relationship between final score and
audience variation. These variables were included as the dependent
variable in regression analyses for winning, losing, and neutral
markets, respectively.
Estimations were conducted using ordinary least squares with robust
standard errors. It is important to note some of the data decisions made
in order to compare markets with teams to those without as previous
demand studies have estimated either one or the other (e.g., Grimshaw
& Burwell, 2014; Paul & Weinbach, 2007; Tainsky, Xu, Salaga,
& Mills, 2014) but not both. As noted previously, only one
observation was entered for neutral markets regardless of the number of
cities where the game was broadcast so as to not artificially inflate
the number of neutral market observations compared to winning and losing
markets. This prevented the inclusion of income and population as
covariates. Further, the distinction between local and opponent record
is not applicable to neutral markets, and thus the sum of winning
percentages was used in their stead (Paul & Weinbach, 2007). The
estimating equation was given as [RatingsDrop.sub.ij] =
[[beta].sub.0]+[[beta].sub.1] [(TotalTeamQuality).sub.j] +
[[beta].sub.2] [(TotalTeamAge).sub.j] + [[beta].sub.3]
[(AbsSpread).sub.j] + [[beta].sub.4] [(MNF).sub.j] + [[beta].sub.5]
(2009) [[beta].sub.6] [(FinalScoreMargin).sub.j] + ij
for market type i and game j. RatingsDrop was replaced by
WholeGameDifference and SecondHalfDifference and the model was
re-estimated.
Linear restrictions testing was conducted to determine whether the
significance of final score margin varied according to market type.
These were estimated for the full sample as well as for all games in
which the margin of victory was nine points or higher based on the ANOVA
results described more fully in the following section.
Results and Discussion
Market Level Comparison
Summary statistics are presented in Table 1 and represented
graphically in Figure 1. In examining the values and visual trend of
RatingsDrop, there was a distinctly similar and consistent pattern
between losing markets and neutral markets, while that of winning
markets is entirely different. As predicted by the UOH, game interest is
decreased in contests with larger final score margins. This was true for
all three market types. Whereas the general pattern of decreasing
ratings was present in winning markets as well, the scale of the ratings
drop is relatively small when compared to neutral and losing markets.
The pattern is not as evident in the alternative measures, where
audience loss appears to be less for winning markets, but there is more
ambiguity in the summary characteristics in the other market types.
[FIGURE 1 OMITTED]
To test the first hypothesis of whether there were statistically
significant differences among three different market types, we first
compared average RatingsDrop of each market type across all intervals
followed by comparison of the change in ratings of each type of market
for each interval (Table 2). There was a significant difference in
RatingsDrop between neutral markets and winning markets. The same was
true for the comparison of losing and winning markets. However, although
the neutral market ratings drop was slightly larger than the losing
market ratings drop, there was no evidence the difference was
significant.
The same pattern emerged in the comparative analysis of each of the
subsequent interval tests beginning at the 9-11 point interval. In other
words, winning markets were significantly different from neutral markets
and losing markets at every interval from nine points onward, and there
was no difference between the neutral markets and losing markets for
those same intervals. In sum, the pattern of RatingsDrop in neutral
markets and losing markets was similar, but different than that of
winning markets. Although significant differences appeared between
market types in close games, there was no consistency in the direction
or significance when games were tightly contested. Games with larger
final scoring margins were less desirable to fans in neutral and losing
markets than fans in the winning team's market. This trend begins
at a margin of nine points, much smaller than that considered by the
league policy on uncompetitive games. This is not to suggest that the
policy should be revised to reestablish the threshold at nine points,
only to say that we do not find an obvious tipping point in the data at
the 18-point margin.
Comparisons within the Neutral Market Data
We then questioned whether the lost viewership in games decided by
the predetermined margin of 18 points was significantly different than
other, smaller scoring margins. Although part one of this study
established that neutral and losing markets possessed similarity in
terms of RatingsDrop, we nonetheless partitioned the sample to include
only neutral narkets--those to which the NFL's broadcast policy on
uncompetitive games applies. To test this, we compared the RatingsDrop
of games decided by 18 or more points with the same three-point
intervals from 0 to 17 used in part one.
The results are shown in Table 3. First, it is clear the general
pattern of reduced viewership in games with higher scoring margins
generally holds through our interval selection. Consequently, the
ratings drop in games decided by 18+ points was indeed significantly
larger than those decided at all intervals below it, statistically
different than the games deemed competitive by league policy.
Notably, the 18+ interval includes games decided by 27-29, and 30+
points, broadcasts that predictably showed the sharpest viewership
decline. Thus, perhaps the ratings drop between the 18+ interval and
higher may be strongly influenced by games that were extremely
uncompetitive. Table 4 shows the comparison of 18-20 interval and other
intervals. The lack of significant differences to the intervals
immediately surrounding it presents a subtle challenge to the idea that
fan perception of game competitiveness changes in games where the
scoring margin is 18 points, but generally upholds that the policy is in
the vicinity of the tipping point between competitiveness and
competitiveness in the eyes of neutral viewers.
Estimating the Influence of Final Margin on Ratings Across Market
Types
The final set of tests utilized results from the previous sections
to examine if there were discrepancies in the impact of scoring margin
on ratings. These take the form of a more traditional approach to
changes in demand, namely regression analysis with final score margin as
a predictor. Table 5 describes the variables included in the analyses.
When examining the entire sample of games, few of the ex ante measures
are significant in predicting RatingsDrop (Table 6). This is reasonable
as these do not change mid-broadcast for those who have already tuned
in. Conversely, the final score difference is significant in all three
market types. The Wald tests showed the magnitude of the effect was
significantly different between winning and losing markets as well as
winning and neutral markets, but did not find for a difference between
losing and neutral markets. The analyses of WholeGameDifference (Table
7) and SecondHalfDifference (Table 8) confirmed the significance of
scoring margin in all three market types, but challenged that no
difference existed between losing and neutral markets. In each of these
estimations there is evidence that the impact of final scoring margin on
ratings was higher in neutral markets, which would lend support to the
NFL policy being restricted to neutral markets.
The previous ANOVA results using RatingsDrop precipitated the idea
that differences existed between winning and all other markets beginning
at a final margin of nine points, but no such differences existed
between neutral and losing markets. Consequently, we partitioned the
data set to include only games decided by nine or more points and
re-estimated the regression models (Tables 9-11). The models
consistently uphold the difference between winning and losing markets,
confirming the earlier finding. The evidence for difference between
winning and neutral markets is relatively weak in the first two models
and absent in the SecondHalfDifference model. None of the three models
support a difference in the impact of final score on ratings between
neutral and losing markets.
Conclusion
We examined how the uncertainty of outcome concept influences
ratings across various television markets. Mixed evidence was found,
essentially suggesting while some viewers are very sensitive to game
scores, others across different markets are minimally influenced.
Specifically, the UOH correctly predicted that demand deceased as
scoring margin increased in each market type. However, there is
considerable macro-level evidence that the magnitude of fans in neutral
and losing markets dropping out of watching games, or exiting, as the
scoring margin increases exceeds that of fans in the markets of winning
squads. Future research could consider the precise scoring margin at the
time of the peak and conclusion of the game to examine whether these are
linked. Other future studies could measure the difference in ratings in
losing and winning markets at the time that broadcasts were cut off in
neutral markets as a point of comparison. The current study did not find
that the NFL's predetermined margin of 18 points represented the
precise juncture at which statistically more fans chose to stop watching
uncompetitive games, although the majority of this analysis supported
that the shift between competitiveness and uncompetitiveness occurred in
the vicinity of this margin. If the policy is intended to curb exits
from uncompetitive games, and exits are comparable in neutral and losing
markets, then the NFL policy must be considering the long-term effects
of switching away from broadcasts in losing markets. Future research can
explore the complexity of the long- and short-run considerations
further.
Moreover, the ideas of BIRGing and CORFing can provide a
perspective of why significant differences exist between ratings drops
in winning and losing markets. The notion that people tend to closely
identify with winning teams and distance themselves from losing teams
originates from the influence of sport identification on self-esteem
(Bee & Kahle, 2006; Pedro, Carmo, & Luiz, 2008; Shih-Hao,
Ching-Yi Daphne, & Chung-Chieh, 2012). This study extends the
research in sport economics and sport management generally, which to
date had only evaluated this regarding the aftermath of games and relied
on self-reported behaviors. In this study we observed that fans of
winning teams are indeed more likely to stay tuned in and
'bask' in their team's victory. Whether fans of the
losing team CORFing truly occurs in the moment or only after the fact is
fodder for further scrutiny. Viewership patterns of fans in losing and
neutral markets showed similarities according to our proxy of exiting,
but the evidence was mixed depending on the specification.
Finally, we consider the benefits of policies enacted to create
greater parity in the league. Although this research underscores the
benefits of competitive balance and, specifically, close games, it also
suggests that were there to exist uncompetitive games, all else equal,
the league as a whole is better off when large markets are not on the
losing end of blowouts. Furthermore, although out-of-market game
selection is important, it is all the more vital in large markets. Thus
the underlying tension between competitiveness and the advantages of
large markets described by Rottenberg (1956) and formally modeled by
El-Hodiri and Quirk (1971) still exists. Even within the domain of
creating parity, equal is more important in some markets than others.
From a practical standpoint the size of the exit effect should be
considered along with simulations and the other variables that have been
shown to be significant in previous studies of out-of-market interest in
optimizing broadcast selections.
References
Alavy, K., Gaskell, A., Leach, S., & Szymanski, S. (2010). On
the edge of your seat: Demand for football on television and the
uncertainty of outcome hypothesis. International Journal of Sport
Finance, 5, 75-95.
Bee, C. C., & Kahle, L. R. (2006). Relationship marketing in
sports: A functional approach. Sport Marketing Quarterly, 15, 102-110.
Brown, M. (2013). Halfway through season, NFL has 18 of the
most-watched TV programs since Sept 5. Retrieved from
http://www.forbes.com/sites/maurybrown/2013/11/07/halfwaythrough-season-nfl-has-18-of-the- most-watched-tv-programs-since-sept-5/
Campbell, R., Aiken, D., & Kent, A. (2004). Beyond BIRGing and
CORFing: Continuing the exploration of fan behavior. Sport Marketing
Quarterly, 13, 151-157.
Cialdini, R. B., Borden, R. J., Thorne, A., Walker, M., Freeman,
S., & Sloan, L. (1976). Basking in reflected glory: Three (football)
field studies. Journal of Personality & Social Psychology, 34,
366-375.
Dewenter, R., & Namini, J. E. (2013). How to make soccer more
attractive? Rewards for a victory, the teams' offensiveness, and
the home bias. Journal of Sports Economics, 14, 65-86.
DirecTV (2013). 2013 annual report of DirecTV. Retrieved from
http://investor.directv.com/files/doc_financials/annual/DIRECTV%202013%20Annual%20Report.pdf
El-Hodiri, M., & Quirk, J. (1971). An economic model of a
professional sports league. The Journal of Political Economy, 79,
1302-1319.
Falter, J. M., & Perignon, C. (2000). Demand for football and
intramatch winning probability: An essay on the glorious uncertainty of
sports. Applied Economics, 32, 1757-1765.
Grimshaw, S. D., & Burwell, S. J. (2014). Choosing the most
popular NFL games in a local TV market. Journal of Quantitative Analysis
in Sports, 10, 329-343.
Nicholson, M. (2007). Sport and the media: Managing the nexus.
London, UK: Elsevier Press.
Owen, P. D., & Weatherston, C. R. (2004). Uncertainty of
outcome and Super 12 Rugby Union attendance: Application of a
general-to-specific modeling strategy. Journal of Sports Economics, 5,
347-370.
Pedro, D., Carmo, L., & Luiz, M. (2008). Fandom affiliation and
tribal behaviour: A sports marketing application. Qualitative Market
Research: An International Journal, 11, 17-39.
Paul, R. J., & Weinbach, A. P. (2002). Market efficiency and a
profitable betting rule: Evidence from totals on professional football.
Journal of Sports Economics 3, 256-263.
Paul, R. J., Weinbach, A. P., & Wilson, M. (2004). Efficient
markets, fair bets, and profitability in NBA totals 1995-96 to 2001-02.
Quarterly Review of Economics and Finance 44, 624-632.
Paul, R. J., & Weinbach, A. P. (2005). Bettor preferences and
market efficiency in football totals markets. Journal of Economics and
Finance, 29, 409-415.
Paul, R. J., & Weinbach, A. P. (2007). The uncertainty of
outcome and scoring effects on Nielsen ratings for Monday Night
Football. Journal of Economics and Business, 59, 199-211.
Paul, R. J., & Weinbach, A. P. (2009). Are behavioral biases
consistent across the Atlantic? The over/under market for European
soccer. The Journal of Gambling Business and Economics, 3, 89-101.
Peel, D. A., & Thomas, D. A. (1992). The demand for football:
Some evidence on outcome uncertainty. Empirical Economics, 17, 323-331.
Salaga, S., & Tainsky, S. (2013). The effects of outcome
uncertainty, scoring, and pregame expectations on Nielsen ratings for
Bowl Championship Series games. Journal of Sports Economics.
Siegfried, J. J., & Hinshaw, C. E. (1979). The effect of
lifting television blackouts on professional football no-shows. Journal
of Economics and Business, 32, 1-13.
Snyder, C. R., Lassegard, M., & Ford, C. E. (1986). Distancing
after group success and failure: Basking in reflected glory and cutting
off reflected failure. Journal of Personality & Social Psychology,
51, 382-388.
Tainsky, S. (2010). Television broadcast ratings for National
Football League contests. Journal of Sports Economics, 11, 629-640.
Tainsky, S., & McEvoy, C. (2012). Television broadcast demand
in markets without local teams. Journal of Sports Economics, 13,
245-260.
Tainsky, S., McEvoy, C. D., & Jasielec, M. (2011, October).
Viewership maximization in non-NFL markets. Annual Conference of the
Sport Marketing Association. Houston, TX.
Tainsky, S., & Winfree, J. A. (2008). Financial incentives and
league policy: The example of Major League Baseball's steroid
policy. European Sport Management Quarterly, 8, 67-81.
Tainsky, S., Xu, J., Salaga, S., & Mills, B. M. (2014).
Spillover benefits to local enthusiasm: Increases in league-wide
interest as a consequence of local sports team competitiveness. Journal
of Economics and Business, 74, 1-10.
Tainsky, S., Xu, J., & Zhou, Y. (2012). Qualifying the game
uncertainty effect: A game-level analysis of NFL postseason broadcast
ratings. Journal of Sports Economics, 13, 250-265.
Wu, S. H., Tsai, C. Y. D, & Huang, C. C. (2012). Toward team or
player? How trust, vicarious achievement motive, and identification
affect fan loyalty. Journal of Sport Management, 26, 177-191.
Endnotes
(1) Salaga and Tainsky (2013) used the interval just after the
halftime break as 'halftime.' We have followed this
established practice in this research.
Jie Xu [1], Hojun Sung [2], Scott Tainsky [1], and Michael Mondello
[3]
[1] University of Illinois
[2] University of Florida
[3] University of South Florida
Jie Xu is a doctoral candidate in the Department of Recreation,
Sport & Tourism. Her research focuses on the economics of sport.
Hojun Sung is a doctoral candidate in the Department of Tourism,
Recreation & Sport Management. His research interests include demand
for sport and professional sport league structure.
Scott Tainsky, PhD, is an associate professor in the Department of
Recreation, Sport & Tourism. His research program is focused on
sports demand and the industrial organization of leagues.
Michael Mondello, PhD, is an associate professor in the Department
of Marketing. His research interests include stadium financing,
contingent valuation method (CVM), employee compensation, analytics, and
ticket pricing.
Table 1. Summary Statistics for Winning, Losing, and Neutral Markets
Mean SD
Winning RatingsDrop 0.04366 0.07480
WholeGameDifference -0.30924 0.250169
SecondHalfDifference -0.11029 0.1370
N 353
Losing RatingsDrop 0.134455 0.13751
WholeGameDifference -0.121931 0.29759
SecondHalfDifference 0.000976 0.19335
N 352
Neutral RatingsDrop 0.14579 0.143627
WholeGameDifference -0.44329 0.647595
SecondHalfDifference -0.19449 0.41917
N 417
Min Max
Winning RatingsDrop 0 0.508333
WholeGameDifference -2.44 0.358696
SecondHalfDifference -0.58586 0.508333
N
Losing RatingsDrop 0 0.654321
WholeGameDifference -1.70588 0.58209
SecondHalfDifference -0.64336 0.533333
N
Neutral RatingsDrop 0 0.656638
WholeGameDifference -3.51313 0.603349
SecondHalfDifference -3.33333 0.703535
N
Table 2. Market Level Comparisons Using Ratings Drop
Market Difference Simultaneous 95%
Comparison Between Confidence Limits
Means
Overall N--L 0.019 -0.001 0.04
N--W 0.103 0.082 0.124 ***
L--W 0.084 0.062 0.105 ***
Interval 0-2 N--L 0.028 -0.011 0.068
N--W 0.022 -0.016 0.059
L--W -0.007 -0.047 0.034
Interval 3-5 N--L 0.022 -0.006 0.05
N--W 0.039 0.01 0.067 ***
L--W -0.007 -0.047 0.034
Interval 6-8 N--L 0.037 0.005 0.069 ***
N--W 0.071 0.039 0.104 ***
L--W 0.034 0.001 0.067 ***
Interval 9-11 N--L 0.037 -0.016 0.091
N--W 0.093 0.04 0.146 ***
L--W 0.056 0 0.111 ***
Interval 12-14 N--L -0.008 -0.065 0.049
N--W 0.133 0.077 0.189 ***
L--W 0.141 0.085 0.197 ***
Interval 15-17 N--L 0.003 -0.069 0.076
N--W 0.134 0.061 0.207 ***
L--W 0.131 0.056 0.205 ***
Interval 18-20 N--L -0.011 -0.096 0.075
N--W 0.171 0.082 0.259 ***
L--W 0.181 0.092 0.27 ***
Interval 21-23 N--L 0.002 -0.075 0.078
N--W 0.18 0.102 0.258 ***
L--W 0.178 0.1 0.257 ***
Interval 24-26 N--L 0.033 -0.072 0.139
N--W 0.174 0.064 0.284 ***
L--W 0.141 0.034 0.248 ***
Interval 27-29 N--L 0.049 -0.075 0.173
N--W 0.178 0.056 0.3 ***
L--W 0.129 0.005 0.253 ***
Interval 30 Plus N--L 0.062 -0.029 0.154
N--W 0.217 0.12 0.313 ***
L--W 0.154 0.055 0.253 ***
Note: Comparisons significant at the 0.05 level are indicated by ***
Table 3. Comparison within Neutral Markets (Interval 18+ vs. Others)
Interval Comparison Difference Simultaneous
Between 95% Confidence Limits
Means
I18+ V.S. I0-2 0.169642 0.127636 0.211648 ***
I18+ V.S. I3-5 0.160347 0.131883 0.188811 ***
I18+ V.S. I6-8 0.146217 0.115413 0.177021 ***
I18+ V.S. I9-11 0.102770 0.066361 0.139179 ***
I18+ V.S. I12-14 0.080582 0.042235 0.118929 ***
I18+ V.S. I15-17 0.057320 0.012166 0.102474 ***
Note: Comparisons significant at the 0.05 level are indicated by ***
Table 4. Comparison within Neutral Markets (Interval 18-20 vs. Others)
Interval Comparison Difference Simultaneous 95%
Between Confidence Limits
Means
I18-20 V.S. I0-2 0.16138 0.10300 0.21977 ***
I18-20 V.S. I3-5 0.15278 0.10486 0.20070 ***
I18-20 V.S. I6-8 0.13889 0.08931 0.18848 ***
I18-20 V.S. I9-11 0.09622 0.04238 0.15006 ***
I18-20 V.S. I12-14 0.07150 0.01582 0.12718 ***
I18-20 V.S. I15-17 0.04824 -0.01321 0.10969
I18-20 V.S. I21-23 0.03183 -0.02805 0.09171
I18-20 V.S. I24-26 -0.00227 -0.06718 0.06264
I18-20 V.S. I27-29 -0.00736 -0.07412 0.05941
I18-20 V.S. I30+ -0.03659 -0.09569 0.02251
Note: Comparisons significant at the 0.05 level are indicated by ***
Table 5. Description of Variables in the Regression Equations
Variable Description
TotalTeamQuality Combined winning percentage of teams
TotalTeamAge Combined age of competing teams
AbsSpread Absolute value of the closing betting line
MNF Indicator variable, 1=Monday Night Football game,
0=otherwise
2009 Indicator variable, 1=2009 season, 0=otherwise
FinalScoreMargin Difference in final score between competing teams
Table 6. Regression and Linear Restrictions Testing Across Market
Types Using Ratings Drop
Winning Losing Neutral
Market Market Market
TotalTeamQuality 0.00338 0.0247 0.0312
(0.0131) (0.0243) (0.0217)
TotalTeamAge 0.000236 ** 0.000222 0.000283 *
(0.0001) (0.0001) (0.0002)
AbsSpread 0.00018 0.00116 -0.000217
(0.0008) (0.0015) (0.0017)
MNF 0.0766 *** 0.122 *** 0.121 ***
(0.0229) (0.0301) (0.0261)
2009 -0.0081 -0.0152 -0.00341
(0.0071) (0.0111) (0.0111)
FinalScoreMargin 0.00241 *** 0.00816 *** 0.00836 ***
(0.0005) (0.0007) (0.0007)
Constant -0.0157 -0.0319 -0.022
(0.0183) (0.0314) (0.0317)
N 353 352 417
[R.sup.2] 0.18 0.4077 0.3885
W vs. N L vs. N W vs. L
(p) (p) (p)
TotalTeamQuality
TotalTeamAge
AbsSpread
MNF
2009
FinalScoreMargin 50.0385 0.0432 52.5772
0.0000 0.8353 0.0000
Constant
N
[R.sup.2]
Table 7. Regression and Linear Restrictions Testing Across Market
Types Using Whole Game Ratings Difference
Winning Losing Neutral
Market Market Market
TotalTeamQuality 0.143 *** 0.0603 0.246 **
(0.0406) (0.0455) (0.1140)
TotalTeamAge 0.00151 *** 0.000784 ** 0.00260 ***
(0.0003) (0.0003) (0.0008)
AbsSpread 0.000483 -0.00323 -0.013
(0.0038) (0.0032) (0.0096)
MNF 0.155 *** 0.163 *** 0.273 ***
(0.0497) (0.0516) (0.0938)
2009 0.000705 -0.0125 -0.0148
(0.0235) (0.0242) (0.0564)
FinalScoreMargin 0.00982 *** 0.0174 *** 0.0271 ***
(0.0013) (0.0013) (0.0028)
Constant -0.724 *** -0.457 *** -1.209 ***
(0.0539) (0.0663) (0.1770)
N 353 352 417
[R.sup.2] 0.2293 0.3786 0.2082
W vs. N L vs. N W vs. L
(p) (p) (p)
TotalTeamQuality
TotalTeamAge
AbsSpread
MNF
2009
FinalScoreMargin 30.4910 9.8532 17.1274
0.0000 0.0017 0.0000
Constant
N
[R.sup.2]
Table 8. Regression and Linear Restrictions Testing Across Market
Types Using Second Half Ratings Difference
Winning Losing Neutral W vs. N
Market Market Market
TotalTeamQuality 0.0385 0.0157 0.0649
(0.0264) (0.0318) (0.0659)
TotalTeamAge 0.000582 *** 0.00026 0.00131 **
(0.0002) (0.0002) (0.0006)
AbsSpread 0.00218 -0.00166 -0.00668
(0.0016) (0.0021) (0.0062)
MNF 0.0705 *** 0.0882 ** 0.256 ***
(0.0233) (0.0388) (0.0369)
2009 -0.00587 -0.00191 -0.0352
(0.0122) (0.0155) (0.0368)
FinalScoreMargin 0.00712 *** 0.0116 *** 0.0182 ***
(0.0007) (0.0008) (0.0020)
Constant -0.306 *** -0.184 *** -0.569 ***
(0.0354) (0.0427) (0.1070)
N 353 352 417
[R.sup.2] 0.3042 0.3861 0.221
Winning L vs. N W vs. L
(p) (p) (p)
TotalTeamQuality
TotalTeamAge
AbsSpread
MNF
2009
FinalScoreMargin 27.5734 9.4676 17.6786
0.0000 0.0021 0.0000
Constant
N
[R.sup.2]
Table 9. Regression and Linear Restrictions Testing Across Market
Types Using Ratings Drop for Margins of 9+
Winning Losing Neutral
Market Market Market
TotalTeamQuality -0.00502 0.00233 0.0507
(0.0224) (0.0415) (0.0464)
TotalTeamAge 0.000319 ** 0.000183 0.000447
(0.0002) (0.0002) (0.0003)
AbsSpread 0.00113 0.00088 0.00175
(0.0013) (0.0022) (0.0032)
MNF 0.0938 *** 0.168 *** 0.172 ***
(0.0313) (0.0319) (0.0425)
2009 -0.0151 -0.0256 0.0373
(0.0115) (0.0165) (0.0232)
FinalScoreMargin 0.00214 *** 0.00613 *** 0.00528 ***
(0.0008) (0.0010) (0.0017)
Constant -0.0109 0.0627 -0.0131
(0.0315) (0.0565) (0.0778)
N 194 194 142
[R.sup.2] 0.1499 0.2678 0.189
W vs. N L vs. N W vs. L
(p) (p) (p)
TotalTeamQuality
TotalTeamAge
AbsSpread
MNF
2009
FinalScoreMargin 2.7406 0.1794 10.1400
0.0978 0.6719 0.0015
Constant
N
[R.sup.2]
Table 10. Regression and Linear Restrictions Testing Across Market
Types Using Whole Game Ratings Difference for Margins of 9+
Winning Losing Neutral
Market Market Market
TotalTeamQuality 0.115 ** 0.0961 0.260 *
(0.0492) (0.0630) (0.1410)
TotalTeamAge 0.000883 ** 0.000269 0.00185
(0.0004) (0.0004) (0.0016)
AbsSpread 0.00963 *** 0.00236 0.0153
(0.0029) (0.0038) (0.0148)
MNF 0.172 *** 0.206 *** 0.365 ***
(0.0582) (0.0496) (0.0910)
2009 -0.00748 -0.0219 0.0839
(0.0259) (0.0290) (0.0978)
FinalScoreMargin 0.00642 *** 0.0115 *** 0.0162 ***
(0.0014) (0.0017) (0.0059)
Constant -0.619 *** -0.336 *** -1.124 ***
(0.0638) (0.0836) (0.3360)
N 194 194 142
[R.sub.2] 0.2205 0.257 0.1119
W vs. N L vs. N W vs. L
(p) (p) (p)
TotalTeamQuality
TotalTeamAge
AbsSpread
MNF
2009
FinalScoreMargin 2.6032 0.5838 5.2032
0.1066 0.4448 0.0225
Constant
N
[R.sub.2]
Table 11. Regression and Linear Restrictions Testing Across Market
Types Using Second Half Ratings Difference for Margins of 9+
Winning Losing Neutral
Market Market Market
TotalTeamQuality 0.0236 0.0528 0.0626
(0.0345) (0.0498) (0.0747)
TotalTeamAge 0.000436 * 0.0000112 0.00000993
(0.0002) (0.0003) (0.0007)
AbsSpread 0.00426 ** 0.00000182 0.00764
(0.0017) (0.0028) (0.0103)
MNF 0.0816 *** 0.135 *** 0.258 ***
(0.0307) (0.0404) (0.0501)
2009 -0.00593 -0.00117 0.0645
(0.0159) (0.0209) (0.0539)
FinalScoreMargin 0.00485 *** 0.00774 *** 0.00727
(0.0010) (0.0011) (0.0045)
Constant -0.237 *** -0.108 * -0.305 **
(0.0460) (0.0635) (0.1190)
N 194 193 139
[R.sup.2] 0.1933 0.2226 0.0764
W vs. N L vs. N W vs. L
(p) (p) (p)
TotalTeamQuality
TotalTeamAge
AbsSpread
MNF
2009
FinalScoreMargin 0.2793 0.0104 3.7687
0.5971 0.9187 0.0522
Constant
N
[R.sup.2]
