NCAA College Basketball Television Viewership: Does Preference for Outcome Uncertainty Change Throughout the Season?
Kang, Byungju ; Salaga, Steven ; Tainsky, Scott 等
NCAA College Basketball Television Viewership: Does Preference for Outcome Uncertainty Change Throughout the Season?
Introduction
While there has been an increase in the number of consumers
watching sports on smartphones, desktops, and tablets, these still
represent a small fraction of the total audience, with traditional
television viewership constituting 89% of Baby Boomer viewers and over
80% of Generation X viewers (Statista, 2017). (1) Thus, it is fair to
state that sports viewership takes place primarily in two different
domains--live attendance at the arena and via television broadcast.
Figuring prominently in the landscape of sports television viewership in
the United States is National Collegiate Athletic Association Division I
Men's Basketball (NCAAM). A recent extension of the television
broadcasting rights to the men's postseason tournament cost CBS and
Turner an average of $1.1 billion annually (Brady, 2016). The widespread
interest in the sport is corroborated by the American Gaming
Association, who estimated $9.2 billion was wagered on March Madness
games in 2016 (American Gaming Association, 2016). These figures not
only serve as indicators of the degree of consumer interest in live
college basketball, but also demonstrate the importance of television
revenues to the NCAA and its member institutions.
Similar to other contexts, such as North American professional
leagues (e.g., Lee & Fort, 2008; Mills & Fort, 2014; Soebbing,
2008; Sung & Mills, 2017) and European football (e.g., Alavy,
Gaskell, Leach, & Szymanski, 2010; Buraimo & Simmons, 2015;
Feddersen & Rott, 2011; Schreyer, Schmidt, & Torgler, 2018),
there is a need to understand factors influencing consumer interest in
the live sport product. Much of this stream of the sport economics
literature centers on contest quality. The absolute and relative quality
of the competing teams figure prominently in existing research, with
implications for league policy with respect to competitive balance and
outcome uncertainty. Considering the attention paid to both quality and
uncertainty and how each impacts consumer interest, the lack of
comprehensive examination of NCAAM is a conspicuous gap in the
literature.
Whereas the distribution of talent in North American professional
sports leagues is artificially constrained by various mechanisms (i.e.,
salary cap, luxury tax), no such spending constraints exist in college
sports. It could be argued that conferences essentially serve as
regulators of relative quality, yet there is still considerable
variation in team quality within a given conference. (2) Further,
nonconference play comprises over one-third of games and no scheduling
restrictions exist. Consequently, basketball powerhouse Duke University
from the Atlantic Coast Conference, which generated $39.40 million in
basketball-specific annual revenues in 2016, plays teams like Elon
University of the Colonial Athletic Association, who generated less than
five percent of that amount, or $1.88 million (U.S. Department of
Education Office of Postsecondary Education, 2018). Thus, it is possible
that conclusions reached regarding the importance of relative quality in
the context of other leagues are not representative of the preferences
of college basketball consumers, given the natural variation in revenues
and hence playing talent that are inherent in the sport. Given this, if
college basketball viewers have a preference for increased anticipated
outcome uncertainty, then policy measures to remedy imbalance are worthy
of consideration.
Literature Review
This study contributes to two key strands of literature in sports
economics: work estimating the determinants of television viewership in
sport, and investigation into the relationship between outcome
uncertainty and consumer interest. Studies estimating demand for live
sport have been a cornerstone area of investigation in the field, and
have developed out of the pioneering contributions of Rottenberg (1956),
Neale (1964), and Noll (1974). While empirical studies estimating the
demand determinants of sport are common (see Borland & MacDonald,
2003; Fizel, 2001), work on NCAA college basketball is extremely
limited. To our knowledge, the only existing contribution is from
Grimshaw, Sabin, and Willes (2013), who investigate the determinants of
television viewership for the NCAA men's Final Four.
The early sport demand literature primarily used game day
attendance as a proxy of demand due to data accessibility. Although
relatively little is known about the demand for television audiences for
many sports (Mongeon & Winfree, 2012), work utilizing television
viewership data has become more common over time. Much of the early
literature utilizing viewership data focused on potential substitution
between attendance and television viewership (e.g., Allan & Roy,
2008; Baimbridge, Cameron, & Dawson, 1995; Buraimo, 2008;
Carmichael, Millington, & Simmons, 1999; Dawson & Downward,
2011; Dobson, Goddard, & Dobson, 2001; Fizel & Bennett, 1989;
Garcia & Rodriguez, 2002; Humphreys, 2002; McEvoy & Morse, 2007;
Mongeon & Winfree, 2012), with this work failing to reach a
consensus.
The utilization of television viewership data offers a number of
key advantages in comparison to attendance data. Namely, demand studies
using match attendance implicitly assume the determinants of attendance
to be similar across spectator types (Allan & Roy, 2008), likely
capturing preferences of home team supporters because of economic
factors, such as travel costs (Borland & MacDonald, 2003). Because
attendance is likely to be comprised of a high percentage of season
ticket holders, demand determinants at the time of the purchase decision
may not be the same at the time of the actual event. The use of
television viewership data avoids these issues and is more likely to
represent the preferences of a more diverse consumer sample, including
home team fans, away team supporters, and general followers of the
sport. This allows for the ability to identify consumer preferences for
not only ex ante factors, but also in-contest product quality
characteristics (Chung, Lee, & Kang, 2016).
Despite the advantages of using viewership data, work in this
context is still relatively limited, due to data availability
constraints. A large portion of this literature has estimated television
demand for European football (Alavy et al., 2010; Buraimo, 2008; Buraimo
& Simmons, 2009; Feddersen & Rott, 2011; Forrest, Simmons, &
Buraimo, 2005; Garcia & Rodriguez, 2002; Johnsen & Solvoll,
2007; Pawlowski & Budzinski, 2012; Perez, Puente, & Rodriguez,
2017), North American major league sports (e.g., Aldrich, Arcidiacono,
& Vigdor, 2005; Bruggink & Eaton, 1996; Chung et al., 2016;
Grimshaw & Burwell, 2014; Hausman & Leonard, 1997; Kanazawa
& Funk, 2001; Mongeon & Winfree, 2012; Paul & Weinbach,
2007; Tainsky, 2010; Tainsky & McEvoy, 2012; Xu, Sung, Tainsky,
& Mondello, 2015) and individual sports, such as mixed martial arts
(Tainsky, Salaga, & Santos, 2012; Watanabe, 2015), auto racing
(Berkowitz, Depken, & Wilson, 2011; Schreyer & Torgler, 2018),
road cycling (Van Reeth, 2013) and tennis (Konjer, Meier, &
Wedeking, 2017; Meier & Konjer, 2015). In terms of collegiate sport,
the literature using viewership data is limited to NCAA college football
(Brown & Salaga, 2017; Salaga & Tainsky, 2015b; Tainsky, Kerwin,
Xu, & Zhou, 2014) and the single aforementioned study on NCAA
postseason college basketball (Grimshaw et al., 2013).
Empirical work assessing Rottenberg's (1956) uncertainty of
outcome hypothesis (UOH) is intertwined with the demand literature,
given that testing the UOH is a part of almost all demand estimations
(Fort, 2005). A long line of empirical research has measured ex ante
consumer preference for anticipated outcome uncertainty on match
attendance (e.g., Buraimo & Simmons, 2008; Czarnitzki &
Stadtmann, 2002; Lee & Fort, 2008; Pawlowski & Anders, 2012) and
has demonstrated mixed results. However, one downside of testing the UOH
using attendance data is the inability to capture preference for actual
outcome uncertainty, given the advance nature of the purchase decision
(Coates, Humphreys, & Zhou, 2014).
The utilization of viewership data avoids this issue and allows for
the ability to measure both anticipated and actual outcome uncertainty.
Recent television viewership demand studies have found a positive
relationship between enhanced anticipated contest uncertainty and
viewership in professional sport (Buraimo & Simmons, 2008; Forrest
et al., 2005; Grimshaw & Burwell, 2014; Tainsky, 2010; Tainsky &
McEvoy, 2012), while Salaga and Tainsky (2015b), and Brown and Salaga
(2017) found increased viewership for games expected to be more certain
in NCAA college football.
Coates et al. (2014) further illustrate that the preferences of
consumers are likely to vary based on their respective reference points.
This also raises the question of whether preference for anticipated and
actual outcome uncertainty vary over time. To our knowledge, this is a
possibility that has not been formally investigated in the literature,
which indicates that researchers have implicitly assumed that preference
for contest uncertainty is uniform within a given consumer population
over the course of a season. We test that possibility here.
Data and Empirical Approach
Television viewership figures utilized in this study are all
nationally televised regular season and postseason conference tournament
contests from the 2014-15 NCAA Division I college basketball season. The
dependent variable is total national television viewership for each
contest, which is provided by the Nielsen Company and was collected at
SportsMediaWatch.com. In total, 948 contest-level observations exist and
all independent variables utilized are publicly available.
Given the unique nature of television viewership, it is necessary
to make an adjustment to the traditional model of demand for live sports
attendance (Tainsky et al., 2014). Specifically, price and market
related factors are not included in the modeling since the financial
expense of game viewership on television is negligible and all games
utilized in this study are broadcast at the national level. Therefore,
we specify the consumer decision to view a contest based on the
following function:
V = f (expected quality, actual quality, temporal factors,
substitues, consumer avaiability), where V is the raw total national
viewership of a given contest.
We estimate the hedonic function above using ordinary least
squares. We transform our dependent variable by taking its natural log
as raw viewership is over-dispersed with a long right tail. By making
this choice, we avoid violating the assumption of normally distributed
error terms (Long, 1997). The dependent variable is truncated from the
left at zero, since viewership for a given televised contest cannot
practically generate a value of zero. We specify White-corrected
standard errors to account for heteroscedasticity. The general
estimating equation follows.
Ln[(Viewers).sub.i] = [[beta].sub.0] +
[[beta].sub.1][AvePomeroy.sub.i] + [[beta].sub.2][Ave3yrWinPct.sub.i] +
[[beta].sub.3][AbsClosingLine.sub.i] + [[beta].sub.4][OverUnder.sub.i] +
[[beta].sub.5:35][Conference.sub.i] +
[[beta].sub.36][PowerConfOOCGm.sub.i] +
[[beta].sub.37][ConfTourney.sub.i] +
[[beta].sub.38][ConfTournChamp.sub.i] +
[[beta].sub.39][ScoreSpread.sub.i] + [[beta].sub.40][TotalDiff.sub.i] +
[[beta].sub.41:45][Month.sub.i] +
[[beta].sub.46:55][Day&StartTime.sub.i] +
[[beta].sub.56][Substitutes.sub.i] + [[beta].sub.57:65][Channel.sub.i] +
[[epsilon].sub.i]
where i represents the individual game.
The measures of anticipated contest quality capture game
characteristics prior to its start. We measure consumer preference for
anticipated absolute contest quality by utilizing statistician Ken
Pomeroy's power ratings. (3) AvePomeroy is the average pre-game
Pomeroy power rating of the two competing teams. We note the inverse
relationship between the average Pomeroy value and absolute contest
quality as the level of quality decreases as the value of the Pomeroy
average increases. Using power ratings is preferable to simple winning
percentages, given the degree of variation in competition quality across
the spectrum of college programs. This is especially true early in the
season, as some programs schedule weaker competition prior to conference
play, while other programs take the opposite approach. In addition to
the Pomeroy measure of pre-game contest quality, we also utilize the
average winning percentage of the competing teams in the three prior
seasons (Ave3yrWinPct). The inclusion of this variable accounts for the
possibility that consumer interest could be strong for teams which have
been of high quality in recent years, but may not be as high quality in
the current season. (4)
To test for anticipated outcome uncertainty, we utilize the
absolute value of the pre-game closing line point spread
(AbsClosingLine). (5) This is the most commonly used proxy for perceived
contest-level outcome uncertainty (Paul, Wachsman, & Weinbach,
2011), and captures the relative quality difference between the
competing teams. Smaller values of this variable indicate a contest
expected to be more competitive. The absolute value is used as previous
literature has suggested that it allows for the ability to distinguish
between fan preference for an outcome on a specific side from actual
preference for uncertainty in television markets (Forrest et al, 2005;
Tainsky & Jasielec, 2014). The pre-game closing line over/under
total (OverUnder) is included to measure potential consumer preference
for anticipated levels of scoring (Paul & Weinbach, 2007, 2015).
Conference affiliation may also impact viewership. Accordingly,
Conference is equal to unity when the game played was between teams in
the same conference. (6) Moreover, it may be the case that fans prefer
to watch games between teams in the traditionally powerful conferences,
or the so-called power conferences (ACC, Big East, Big 10, Big 12, SEC,
and Pac 12). Thus, PowerConfOOCGm is an indicator representing an
out-of-conference contest between teams in power conferences. The
inclusion of this variable tests consumer preference for these matchups
as they are less common in comparison to conference games. ConfTourney
and ConfTournChamp are indicators specifying whether the game played was
a conference tournament game and a conference tournament championship
game, respectively. These variables capture the potential increased
interest attributable to relative contest importance related to access
to postseason play.
Measures of actual contest quality capture within-game contest
characteristics. To measure actual outcome uncertainty, we include
ScoreSpread, which is the difference between the pre-game expectation,
set by the closing line point spread, and the final scoring margin.
Values near zero signify that the pre-game expectation set by the
betting market matched the actual contest outcome. The variable takes a
negative value if the contest was more competitive than anticipated and
includes when an underdog wins outright. Positive values occur in games
which were less competitive than expected and captures when the favorite
outperforms market expectations and covers the spread. This variable is
preferable to raw final scoring margin as the latter is collinear with
the closing line point spread for games where the actual contest outcome
matches pre-game expectations. The approach used here virtually
eliminates any collinearity between anticipated and actual outcome
uncertainty (7) and has also been used by Salaga and Tainsky (2015b),
and Brown and Salaga (2017).
We build a similar variable with respect to the level of total
scoring. TotalDiff measures consumer interest in total scoring in
reference to market expectation set by the closing line over/under
total. Values closer to zero identify contests where actual total
scoring matched market expectations. Negative values indicate that the
realized level of total scoring was lower than what was expected
according to the closing line over/under total, and vice versa for
positive values.
As in previous literature, temporal factors may also be related to
variation in television viewership. College basketball games are
scheduled by television networks in two-hour time blocks with different
standard starting times for games played during the week and on
weekends. Given this, we create a series of ten indicator variables
which simultaneously identify both start-time (in Eastern Standard Time)
and whether the game was played during the week (Monday to Friday) or on
the weekend (Saturday and Sunday). The weekday indicators include
WD7:00-8:59PM, WD9:00-10:59PM, WD11:00PM-12:59AM, and WDOther. The
weekend indicators include WE12:00-1:59PM, WE2:00-3:59PM, WE4:00-5:59PM,
WE6:00-7:59PM, WE8:00-9:59PM, and WE10:00PM&Later. WD7:00-8:59PM is
treated as the baseline in all estimations. (8)
To control for viewership being impacted by the number of direct
substitutes available to consumers, we include Substitutes. This
variable is equal to the number of nationally televised games that
overlap the two-hour time window of a given contest. For example, for a
game that begins at 7:00 PM, all other matchups which overlap the
7:00-9:00 PM time window will be treated as substitutes. It is expected
that an increase in the number of substitutes will be associated with
decreased viewership.
A number of cable and network television channels hold the rights
to broadcast NCAA college basketball games nationally. However, not all
consumers purchase access to all cable television channels. To account
for the differences in the availability of a given contest, we include a
series of indicator variables identifying the television channel on
which the game was broadcast (Channel).
Results
Descriptive Statistics
Table 1 provides summary statistics for all variables. The average
viewership in the sample is slightly under 444,000 with the regular
season contest between North Carolina and Duke on Saturday, March 7,
2015--the most watched game with 4.24 million viewers. The average
pre-game closing line point spread is approximately 7.6 points, and the
average closing line over/under total is just above 133 points. Our
summary statistics reveal that, on average, market expectations reflect
actual contest outcomes rather well. ScoreSpread and TotalDiff fall
within one point of the closing line point spread and over/under total,
respectively. Approximately 42 percent of all games were played on
weekends and two-thirds were broadcast on ESPN affiliated television
channels.
Baseline Model: Determinants of College Basketball Viewership
Our baseline model (Model 1 in Table 2) estimates the determinants
of college basketball television viewership. With respect to expected
game quality, our baseline model indicates that anticipated absolute
contest quality is a significant driver of viewership. Consumers respond
to higher levels of anticipated absolute contest quality as lower
average Pomeroy ratings (i.e., higher quality teams) are associated with
significantly higher television viewership (p < 0.01). Holding all
else constant, on average, viewership is expected to increase by 0.47%
(%[DELTA]y = 100 * ([e.sup.[beta]1] - 1)) given a one-unit decrease in
the combined average Pomeroy rating of the competing teams. In other
words, a one standard deviation decrease in average Pomeroy rating is
associated with an expected viewership increase of 25.84%. Even after
controlling for real-time contest quality, the modeling illustrates that
viewership is higher when the contest features teams with higher
multi-season winning percentages. Ave3yrWinPct is statistically
significant at the one percent level, indicating that viewership is
expected to rise by 14.66% given a one standard deviation increase in
average winning percentage of the competing teams in the three prior
seasons.
In terms of anticipated outcome uncertainty, we uncover that
viewership is higher when the absolute value of the pre-game closing
line point spread (AbsClosingLine) is larger. This result demonstrates
that, on average, a one-point increase in the closing line point spread
is associated with a viewership increase of 2.17%. In opposition to the
UOH, this result indicates that consumers have preference for contests
expected to be more certain and supports existing work which analyzes
television viewership in college football (Salaga & Tainsky, 2015b).
Similar to Paul and Weinbach (2007), this study also finds that
viewership is sensitive to anticipated levels of scoring as consumer
interest is greater (p < 0.01) when the closing line over/under total
(OverUnder) is higher. On average, viewership is expected to increase by
0.62% given a one-point increase in the anticipated level of scoring.
Viewership also varies based on conference affiliation
(Conference). The majority of indicators specifying in-conference
matchups are significant in comparison to the out-of-conference (power
conference vs. non-power conference) contest baseline. Conference games
in the non-power conferences tend to display significant negative
effects. Surprisingly, we find there is no premium for conference games
in the Big East and Pac 12, but find significant positive effects for
ACC, Big 10, Big 12, and SEC conference games. Regular season
out-of-conference games between two power conference teams
(PowerConfOOCGm) generate a premium of 38.54% additional viewers.
As expected, game importance is a critical factor in consumer
interest. All else equal, the modeling indicates increased viewership
for conference tournament games (ConfTourney) with an even larger
premium of 56.36% additional viewers for conference tournament
championship games (ConfTournChamp). It may be the case that viewership
is sensitive to contest importance due to access to postseason play and
associated interest in which teams reach the NCAA Division I Men's
Basketball Tournament.
With respect to actual contest quality, ScoreSpread is negative and
significant, illustrating support for the UOH in the context of actual
contest uncertainty. This indicates preference for a final scoring
margin which is closer than the pre-game expectation set by closing line
point spread. The coefficient indicates a viewership increase of 4.17%
for every one standard deviation change in the final scoring margin
relative to the point spread. Alternatively, we find no statistically
significant impact of actual scoring levels relative to the closing line
over/under total (TotalDiff).
There is also significant variation in viewership by Month and a
gradual increase in viewership throughout the course of the season,
which may be due to the perceived importance of late season contests. In
reference to the March baseline, viewership is significantly lower in
the first two months of the season--a time when college basketball
competes directly with college football. The start time of weekday
contests also has the ability to significantly impact viewership as
games starting outside of the 7:00 PM to 8:59 PM and 9:00 PM to 10:59 PM
windows generate significantly less viewership. Alternatively,
irrespective of start time, weekend contests do not display viewership
that is significantly different in reference to the WD7:00-8:59PM
baseline.
As anticipated, an increase in the number of direct substitutes is
associated with a decrease in live viewership. The Substitutes
coefficient implies a 2.02% reduction in viewership for every Division I
broadcast that overlaps the game of interest. Channel indicators
illustrate that in reference to the ESPN2 baseline, all networks are
statistically significant, except for FOX. Games broadcast on CBS
generate the highest viewership, whereas contests on FS2 are expected to
have the lowest viewership. These two results are not overly surprising,
given CBS is a free over-the-air network, while FS2 is a relatively new
cable channel requiring purchase.
Subsample Models: Preference for Outcome Uncertainty
Results from our baseline model illustrate mixed support for the
UOH. Specifically, AbsClosingLine is positive and significant (p <
0.01), which indicates that consumers prefer to watch games expected to
be more certain--a result opposing the prediction of the UOH in terms of
anticipated contest quality. However, ScoreSpread is negative and
significant (p < 0.05), which suggests that viewership is greater
when the actual scoring margin of the game is closer than anticipated.
This finding supports the UOH in the context of actual outcome
uncertainty.
While on the surface these results appear contradictory, we note
that NCAA college basketball represents a setting which is notably
different in comparison to the major professional sports leagues--where
the majority of the empirical work on the UOH lies. By default, our
sample contains games with a wide range of contest qualities given over
340 programs play at the NCAA Division I level. Unlike other sports,
almost all programs that are unsuccessful during the regular season
still have the opportunity to play in the postseason in their respective
conference tournaments. This structure also allows for the ability of a
team to play their way into the NCAA tournament. It is possible that
consumer preference for specific contest types change over the course of
a season, along with perceived contest importance. Thus, we estimate a
series of additional models to test whether consumer viewership
preferences fluctuate throughout the course of a season.
Models 2-6 in Table 2 display coefficients for models that are
estimated on observations in specific months of the season (November to
March). The total number of contest-level observations in each subsample
is 121, 170, 250, 252, and 155, respectively. Coefficients for some
variables are omitted due to insufficient observations under a certain
condition (e.g., no Big 12 conference games are played in November or
December).
Here, we focus exclusively on the variables capturing anticipated
and actual outcome uncertainty and how they change along with the
progression of the season calendar. Our subsample models illustrate that
consumer preference for anticipated outcome uncertainty is fairly
consistent throughout a season, except for contests in March (Model 6).
As with the primary model, AbsClosingLine is positive and statistically
significant across all subsample models but becomes non-significant for
games played in March. Alternatively, with respect to actual contest
uncertainty, ScoreSpread is negative and non-significant in all models.
Given that Model 6 contains both regular season games and
post-season conference tournament contests played in the month of March,
we test whether the effect is consistent across game type and driven by
timing within the season, or whether the effect is a byproduct of game
importance. As such, we re-estimate Model 6 based on game type with
Model 7 and Model 8 in Table 3, displaying coefficients estimated from
regular season games and conference tournament matchups in March,
respectively. These models verify that consumer preference for actual
outcome uncertainty varies over time and is tied to conference
tournament contests that grant access to the NCAA tournament. While
AbsClosingLine is not statistically significant at standard levels in
Models 7 and 8, the marginal effect of ScoreSpread is negative and
statistically significant at one percent level in conference tournament
games (Model 8). Additionally, the magnitude of the marginal effect is
almost four times larger in Model 8 compared to the baseline model.
These results again demonstrate that consumer preference for actual
outcome uncertainty has the ability to fluctuate over the course of a
season, and in this case, is linked to contest importance as it relates
to access to the sport's premier postseason tournament.
In order to further investigate the manner in which preferences
vary between regular season and postseason games, we estimate all
regular season games separately in Model 9 of Table 3. Again, evident is
preference for anticipated contest certainty as well as preference for
actual outcome uncertainty. These results confirm that preferences
fluctuate over the course of the season calendar and appear to be tied
to contest importance.
Discussion
The results from our empirical models illustrate that consumer
preference for both anticipated and actual outcome uncertainty are not
consistent over the course of the season. Throughout the regular season,
we uncover strong consumer partiality to contests expected to be more
certain, with the effect losing magnitude at the end of the regular
season and in the postseason. The preference for a lack of anticipated
uncertainty is in opposition to the majority of work in professional
sport, but is in line with the limited work in college football (Brown
& Salaga, 2017; Salaga & Tainsky, 2015b). This divergence in
findings could be due to the heightened variation in program revenues
and subsequent team quality in college sport. It is likely that the
preference to watch high quality programs is strong and therefore the
viewership base for many games is skewed towards these teams, which are
also more likely to be favored. If a larger percentage of the viewership
base is concerned with the contest outcome relative to the higher
quality team--irrespective of whether the preference is for the favorite
to win or be upset--this could potentially explain the positive effect
of AbsClosingLine and the preference for low relative contest quality
(Coates et al., 2014).
Our variable capturing preference for actual outcome uncertainty
relative to market expectations (ScoreSpread) also varies in magnitude
and statistical significance over the course of a season as consumer
sensitivity to the uncertainty of actual contest outcomes is heightened
once postseason play begins. Though our baseline model including all
contests indicates ScoreSpread is statistically significant, the
magnitude of the effect is much stronger during postseason play.
This result also illustrates consumer sensitivity to outcomes
relative to the wagering market, providing support for similar findings
in college football (Brown & Salaga, 2017; Salaga & Tainsky,
2015a, 2015b). We find that viewership for postseason conference
tournament games is 15.30% higher than March regular season games and is
66.55% higher than regular season contests in general. Naturally, a
portion of this increase is due to higher levels of absolute contest
quality in postseason play. However, the magnitude of the effect of
ScoreSpread is approximately four times larger in the postseason,
relative to both the full sample in Model 1 and all regular season
contests in Model 9. This highlights the strength of the relationship
between viewership and interest in outcomes relative to the betting
market and that NCAA conferences and member institutions directly
benefit from betting market interest.
Empirical evidence also indicates that consumer preference for
absolute quality strongly outweighs preference for relative quality.
Both the real-time and three-year measures of absolute contest quality
are strongly associated with viewership figures. Alternatively, we find
no empirical support that viewership is negatively impacted by matchups
of low anticipated relative quality--and in fact, uncover strong
evidence that consumers actually prefer lower levels of relative
quality. From a policy standpoint, these findings fail to identify any
negative effects of power programs scheduling out-of-conference contests
against lower level Division I programs. This result also indicates that
any perceived imbalance in matchup quality is not a factor which is
negatively impacting television viewership or its revenues.
Conclusion
The television medium allows for the simultaneous evaluation of
preferences for both anticipated and actual game uncertainty. Like many
cited in Borland and Macdonald's (2003) well-regarded review of
consumer demand, the present research does not find support for audience
preference for anticipated game uncertainty. While some previous
non-significant results have been attributed to spectator partisanship
(i.e., preference for winning is difficult to disentangle from
preference for uncertainty), this does not explain the balance of the
findings. In this case, we discover that a diversified, national
television audience prefers games with a decided favorite. This is a
scenario that should be particularly familiar to college basketball fans
given the propensity of Cinderella stories such as George Mason,
Virginia Commonwealth, and Loyola-Chicago to capture the attention of
the public. Given the consistency of our findings with respect to
preference for low relative contest quality, these results could
indicate that the desire to see upsets exists throughout the course of
the year and not just in postseason play. Further work in this area is
needed.
A related question is whether television consumers of any
professional sports possess similar preferences for anticipated game
certainty. To our knowledge, neither the studies cited in Borland and
Macdonald (2003) nor others estimating demand for professional sports in
the decade and a half since that work have identified this effect. Given
that there is a small, but growing line of work which produces this
effect in college sport, this raises the question of whether this
discrepancy is a function of variation between consumer bases. Although
empirical work estimating viewership preferences across a variety of
sports has grown, future research should continue to investigate the
robustness of this finding in various contexts in intercollegiate and
professional sport.
Finally, future work should continue to consider the element of
seasonality given that many leagues play seasons spanning the better
part of a calendar year. The results here indicate that consumer
preferences are not static and have the ability to vary based on
temporal conditions and perceived contest importance. A distinct
advantage of viewership data is the ability to measure real-time
consumption choices and how elements of both perceived and actual
contest quality are associated with viewership patterns across the
season calendar. Further investigation of viewership tendencies as they
relate to seasonal timing are likely to yield additional insights into
the nature of real-time sport consumption.
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(2), 75-95.
Aldrich, E. M., Arcidiacono, P. S., & Vigdor, J. L. (2005). Do
people value racial diversity? Evidence from Nielsen ratings. Topics in
Economics Analysis and Policy, 5(1), Article 4. Retrieved from
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1515945
Allan, G., & Roy, G. (2008). Does television crowd out
spectators? New evidence from the Scottish Premier League. Journal of
Sports Economics, 9(6), 592-605.
American Gaming Association. (2016). State of the States: The AGA
Survey of the Casino Industry. Retrieved from
https://www.americangaming.org/sites/default/files/2016%20State%20of%20the%20States_FINAL.pdf
Baimbridge, M., Cameron, S., & Dawson, P. (1995). Satellite
broadcasting and match attendance: The case of rugby league. Applied
Economics Letters, 2(10), 343-346.
Berkowitz, J. P., Depken, C. A., & Wilson, D. P. (2011). When
going in circles is going backward: Outcome uncertainty in NASCAR.
Journal of Sports Economics, 12(3), 253-283.
Borland, J., & MacDonald, R. (2003). Demand for sport. Oxford
Review of Economic Policy, 19(4), 478-502.
Brady, E. (2016). NCAA extends tournament deal with CBS, Turner
through 2032 for $8.8 billion. USA Today. Retrieved from
https://www.usatoday.com/story/sports/ncaab/2016/04/12/ncaa-contract-extension-cbs-turner-ncaa-tournament-march-madness/82939124/
Brown, K. M., & Salaga, S. (2017). NCAA football television
viewership: Product quality and consumer preference relative to market
expectations. Sport Management Review,
http://dx.doi.org/10.1016/j.smr.2017.08.008
Bruggink, T. H., & Eaton, J. W. (1996). Rebuilding attendance
in Major League Baseball: The demand for individual games. In J. Fizel,
E. Gustafsen, & L. Hadley (Eds.), Baseball Economics. Current
Research (pp. 9-31). Westport: CT: Praeger.
Buraimo, B. (2008). Stadium attendance and television audience
demand in English league football. Managerial and Decision Economics,
29(6), 513-523.
Buraimo, B., & Simmons, R. (2008). Do sports fans really value
uncertainty of outcome? Evidence from the English Premier League.
International Journal of Sport Finance, 3(3), 146-155.
Buraimo, B., & Simmons, R. (2009). A tale of two audiences:
Spectators, television viewers and outcome uncertainty in Spanish
football. Journal of Economics and Business, 61(4), 326-338.
Buraimo, B., & Simmons, R. (2015). Uncertainty of outcome or
star quality? Television audience demand for English Premier League
football. International Journal of the Economics of Business, 22(3),
449-469.
Carmichael, F., Millington, J., & Simmons, R. (1999).
Elasticity of demand for Rugby League attendance and the impact of
BskyB. Applied Economics Letters, 6(12), 797-800.
Carmin, M. (2017). Men's Basketball Revenue, Expenses: Where
Purdue ranks in Big Ten. Journal & Courier. Retrieved from
http://www.jconline.com/story/sports/blogs/mike-carmin/2017/04/11/mens-basketball-revenue-expenses-where-purdue-ranks-big-ten/100331366/
Chung, J., Lee, Y. H., & Kang, J. H. (2016). Ex ante and ex
post expectations of outcome uncertainty and baseball television
viewership. Journal of Sports Economics, 17(8), 790-812.
Coates, D., Humphreys, B. R., & Zhou, L. (2014).
Reference-dependent preferences, loss aversion, and live game
attendance. Economic Inquiry, 52(3), 959-973.
Czarnitzki, D., & Stadtmann, G. (2002). Uncertainty of outcome
versus reputation: Empirical evidence for the First German Football
Division. Empirical Economics, 27(1), 101-112.
Dawson, P., & Downward, P. (2011). Participation, spectatorship
and media coverage in sport: Some initial insights. In W. Andreff (Ed.),
Contemporary issues in sports economics: Participation and professional
team sports (pp. 15-42). Cheltenham: Edward Elgar.
Dobson, S., Goddard, J. A., & Dobson, S. (2001). The economics
of football. Cambridge: Cambridge University Press.
Feddersen, A., & Rott, A. (2011). Determinants of demand for
televised live football: Features of the German national football team.
Journal of Sports Economics, 12(3), 352-369.
Fizel, J. L. (2001). Handbook of sports economics research.
Amorick, New York: ME Sharpe Inc.
Fizel, J. L., & Bennett, R. W. (1989). The impact of college
football telecasts on college football attendance. Social Science
Quarterly, 70(4), 980-988.
Forrest, D., Simmons, R., & Buraimo, B. (2005). Outcome
uncertainty and the couch potato audience. Scottish Journal of Political
Economy, 52(4), 641-661.
Fort, R. (2005). The golden anniversary of "The baseball
players' labor market." Journal of Sports Economics, 6(4),
347-358.
Garcia, J., & Rodriguez, P. (2002). The determinants of
football match attendance revisited: Empirical evidence from the Spanish
football league. Journal of Sports Economics, 3(1), 18-38.
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(3), 329-343.
Grimshaw, S. D., Sabin, R. P., & Willes, K. M. (2013). Analysis
of the NCAA men's Final Four TV audience. Journal of Quantitative
Analysis in Sports, 9(2), 115-126.
Hausman, J. A., & Leonard, G. K. (1997). Superstars in the
National Basketball Association: Economic value and policy. Journal of
Labor Economics, 15(4), 586-624.
Humphreys, B. R. (2002). Alternative measures of competitive
balance in sports leagues. Journal of Sports Economics, 3(2), 133-148.
Johnsen, H., & Solvoll, M. (2007). The demand for televised
football. European Sport Management Quarterly, 7(4), 311-335.
Kanazawa, M. T., & Funk, J. P. (2001). Racial discrimination in
professional basketball: Evidence from Nielsen ratings. Economic
Inquiry, 39(4), 599-608.
Konjer, M., Meier, H. E., & Wedeking, K. (2017). Consumer
demand for telecasts of tennis matches in Germany. Journal of Sports
Economics, 18(4), 351-375.
Kutner, M. H., Nachtsheim, C., & Neter, J. (2004). Applied
linear regression models (4th ed.). New York, NY: McGraw-Hill Irwin.
Lee, Y. H., & Fort, R. (2008). Attendance and the
uncertainty-of-outcome hypothesis in baseball. Review of Industrial
Organization, 33(4), 281-295.
Long, J. S. (1997). Regression models for categorical and limited
dependent variables. Thousand Oaks, CA. Sage Publications.
McEvoy, C. D., & Morse, A. L. (2007). An investigation of the
relationship between television broadcasting and game attendance.
International Journal of Sport Management and Marketing, 2(3), 222-235.
Meier, H. E., & Konjer, M. (2015). Is there a premium for
beauty in sport consumption? Evidence from German TV ratings for tennis
matches. European Journal for Sport and Society, 12(3), 309-340.
Mills, B., & Fort, R. (2014). League-level attendance and
outcome uncertainty in US pro sports leagues. Economic Inquiry, 52(1),
205-218.
Mongeon, K., & Winfree, J. (2012). Comparison of television and
gate demand in the National Basketball Association. Sport Management
Review, 15(1), 72-79.
Neale, W. C. (1964). The peculiar economics of professional sports.
The Quarterly Journal of Economics, 78(1), 1-14.
Noll, R. G. (1974). Government and the sports business. Washington,
DC: Brookings institution.
Paul, R. J., Wachsman, Y., & Weinbach, A. P. (2011). The role
of uncertainty of outcome and scoring in the determination of fan
satisfaction in the NFL. Journal of Sports Economics, 12(2), 213-221.
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(3), 199-211.
Paul, R. J., & Weinbach, A. P. (2015). The betting market as a
forecast of television ratings for primetime NFL football. International
Journal of Sport Finance, 10(3), 284-292.
Pawlowski, T., & Anders, C. (2012). Stadium attendance in
German professional football--The (un) importance of uncertainty of
outcome reconsidered. Applied Economics Letters, 19(16), 1553-1556.
Pawlowski, T., & Budzinski, O. (2012). The (monetary) value of
competitive balance for sport consumers: A stated preferences approach
to European professional football. International Journal of Sport
Finance, 8(2), 112-123.
Perez, L., Puente, V., & Rodriguez, P. (2017). Factors
determining TV soccer viewing: Does uncertainty of outcome really
matter? International Journal of Sport Finance, 12(2), 124-139.
Rottenberg, S. (1956). The baseball players' labor market.
Journal of political economy, 64(3), 242-258.
Salaga, S., & Tainsky, S. (2015a). Betting lines and college
football television ratings. Economics Letters, 132, 112-116.
Salaga, S., & Tainsky, S. (2015b). The effects of outcome
uncertainty, scoring, and pregame expectations on Nielsen ratings for
bowl championship series games. Journal of Sports Economics, 16(5),
439-459.
Schreyer, D., Schmidt, S. L., & Torgler, B. (2018). Game
outcome uncertainty in the English Premier League: Do German fans care?
Journal of Sports Economics, 19(5), 625-644.
Schreyer, D., & Torgler, B. (2018). On the role of race outcome
uncertainty in the TV demand for Formula 1 Grands Prix. Journal of
Sports Economics, 19(2), 211-229.
Soebbing, B. P. (2008). Competitive balance and attendance in major
league baseball: An empirical test of the uncertainty of outcome
hypothesis. International Journal of Sport Finance, 3(2), 119-126.
Statista. (2017). Sports on TV in the U.S. Retrieved from
https://www.statista.com/study/23358/sports-on-tv-in-the-us-statista-dossier/
Sung, H., & Mills, B. M. (2017). Estimation of game-level
attendance in major league soccer: Outcome uncertainty and absolute
quality considerations. Sport Management Review,
https://doi.org/10.1016/j.smr.2017.12.002
Tainsky, S. (2010). Television broadcast demand for National
Football League contests. Journal of Sports Economics, 11(6), 629-640.
Tainsky, S., & Jasielec, M. (2014). Television viewership of
out-of-market games in league markets: Traditional demand shifters and
local team influence. Journal of Sport Management, 28(1), 94-108.
Tainsky, S., Kerwin, S., Xu, J., & Zhou, Y. (2014). Will the
real fans please remain seated? Gender and television ratings for
pre-game and game broadcasts. Sport Management Review, 17(2), 190-204.
Tainsky, S., & McEvoy, C. D. (2012). Television broadcast
demand in markets without local teams. Journal of Sports Economics,
13(3), 250-265.
Tainsky, S., Salaga, S., & Santos, C. A. (2012). Determinants
of pay-per-view broadcast viewership in sports: The case of the Ultimate
Fighting Championship. Journal of Sport Management, 27(1), 43-58.
U.S. Department of Education Office of Postsecondary Education.
(2018). The equity in athletics data analysis cutting tool. Retrieved
from https://ope.ed.gov/athletics/#/
Van Reeth, D. (2013). TV demand for the Tour de France: The
importance of stage characteristics versus outcome uncertainty,
patriotism, and doping. International Journal of Sport Finance, 8(1),
39-60.
Watanabe, N. M. (2015). Sources of direct demand: An examination of
demand for the Ultimate Fighting Championship. International Jounal of
Sport Finance, 10(1), 26-41.
Xu, J., Sung, H., Tainsky, S., & Mondello, M. (2015). A tale of
three cities: Intra-game ratings in winning, losing, and neutral
markets. International Journal of Sport Finance, 10(2), 122-137.
Byungju Kang, (1) Steven Salaga, (1) Scott Tainsky, (2) Matthew
Juravich (3)
(1) University of Georgia
(2) Wayne State University
(3) University of Akron
Byungju Kang is a doctoral student in the Sport Management and
Policy Program at the University of Georgia. His current research
focuses on sports demand and sports betting markets.
Steven Salaga is an assistant professor in the program in Sport
Management and Policy at the University of Georgia. His research focuses
on the economics of professional and collegiate sports.
Scott Tainsky is an associate professor and the director of sport
& entertainment management in the Department of Management and
Information Systems at Wayne State University. His research examines
sport economics and sport management.
Matthew Juravich, is an associate professor in the School of Sport
Science and Wellness Education at the University of Akron. His research
examines issues related to strategic management and managerial
performance in sport.
(1) The term "baby boomer" refers to the 76 million
people born in the U.S. between 1946 and 1964 following the end of World
War II. "Generation X" refers to the segment of the U.S.
population following the baby boomers and includes those born between
the early 1960s and the early 1980s.
(2) To illustrate, the Indiana University men's basketball
program generated $23.47 million in 2016 revenues. Big Ten conference
partner Rutgers University generated $5.48 million in revenues over the
same time period (Carmin, 2017).
(3) Ken Pomeroy has been producing and disseminating college
basketball power ratings since 2003. These ratings are commonly
referenced in popular media and can be found at http://kenpom.com.
(4) The correlation between AvePomeroy and Ave3YrWinPct is -0.5312,
which is at an acceptable level with respect to concern regarding
collinearity (Kutner, Nachtsheim, & Neter, 2004). Correlation
coefficients indicate no other concerns regarding potential collinearity
between variables.
(5) All point spread and over/under point total data was collected
from https://www.sportsinsights.com/.
(7) The correlation between AbsClosingLine and ScoreSpread is
0.0286.
(8) We tested other specifications for both contest start time and
day of the week and found the joint indicator variables utilized to
provide superior explanatory power.
Table 1. Summary Statistics.
Variable Mean SD Min Max
Viewers 417222.6 590059.8 5000 4241000
AvePomeroy 78.4995 55.7753 3 321.5
Ave3yrWinPct 0.6139 0.0960 0.315 0.829
AbsClosingLine 7.6229 5.7794 0 41
OverUnder 133.4182 9.9448 106 208
Conference
ACC 0.0865 0.2812 0 1
Big12 0.0981 0.2976 0 1
BigEast 0.0812 0.2733 0 1
Big10 0.0654 0.2474 0 1
Pac12 0.0643 0.2455 0 1
SEC 0.0612 0.2398 0 1
AmericaEast 0.0021 0.0459 0 1
AmericanAthletic 0.0812 0.2733 0 1
Atlantic10 0.0443 0.2059 0 1
AtlanticSun 0.0011 0.0325 0 1
BigSky 0.0011 0.0325 0 1
BigSouth 0.0053 0.0725 0 1
BigWest 0.0063 0.0793 0 1
Colonial 0.0105 0.1022 0 1
ConferenceUSA 0.0021 0.0459 0 1
Horizon 0.0084 0.0915 0 1
Ivy 0.0011 0.0325 0 1
MetroAtlantic 0.0084 0.0915 0 1
MidAmerican 0.0063 0.0793 0 1
MidEastern 0.0053 0.0725 0 1
MissouriValley 0.0116 0.1071 0 1
MountainWest 0.0116 0.1071 0 1
Northeast 0.0032 0.0562 0 1
OhioValley 0.0074 0.0857 0 1
Southern 0.0011 0.0325 0 1
Southland 0.0011 0.0325 0 1
Southwestern 0.0042 0.0649 0 1
SunBelt 0.0032 0.0562 0 1
Summit 0.0011 0.0325 0 1
WestCoast 0.0200 0.1402 0 1
WesternAthletic 0.0011 0.0325 0 1
PowerConfOOCGm 0.1034 0.3046 0 1
ConfTourney 0.0949 0.2933 0 1
ConfTournChamp 0.0316 0.1751 0 1
ScoreSpread -0.1324 10.5422 -33.5 37.5
TotalDiff 0.3761 16.5565 -56.5 67
Month
November 0.1276 0.3339 0 1
December 0.1793 0.3838 0 1
January 0.2637 0.4409 0 1
February 0.2658 0.4420 0 1
March (baseline) 0.1635 0.3700 0 1
Day&StartTime
WD7:00-8:59PM 0.2321 0.4224 0 1
(baseline)
WD9:00-10:59PM 0.1973 0.3981 0 1
WD11:00PM-12:59AM 0.0485 0.2150 0 1
WDOther 0.1002 0.3004 0 1
WE12:00-1:59PM 0.0992 0.2990 0 1
WE2:00-3:59PM 0.1034 0.3046 0 1
WE4:00-5:59PM 0.0802 0.2717 0 1
WE6:00-7:59PM 0.0580 0.2339 0 1
WE8:00-9:59PM 0.0496 0.2172 0 1
WE10:00PM&Later 0.0316 0.1751 0 1
Substitutes 2.9757 2.2274 0 12
Channel
CBS 0.0359 0.1861 0 1
ESPN 0.1508 0.3581 0 1
ESPNEWS 0.0496 0.2172 0 1
ESPNU 0.3196 0.4666 0 1
FOX 0.0127 0.1119 0 1
FS1 0.1392 0.3464 0 1
FS2 0.0053 0.0725 0 1
NBCSN 0.0496 0.2172 0 1
ESPN2 (baseline) 0.2373 0.4257 0 1
Note. N = 948.
Table 2. Determinants of Television Viewership in NCAA College
Basketball.
Model 1 Model 2
Baseline November
Coef. z Coef. z
AvePomeroy -0.005 (***) -8.45 -0.010 (***) -4.09
Ave3yrWinPct 1.527 (***) 6.95 -0.139 -0.16
AbsClosingLine 0.021 (***) 5.77 0.050 (***) 5.29
OverUnder 0.006 (***) 3.35 -0.000 -0.03
Conference
ACC 0.491 (***) 4.66
Big12 0.257 (**) 2.38
BigEast 0.189 1.58 0.229 1.16
Big10 0.476 (***) 4.46
Pac12 0.124 1.09
SEC 0.322 (***) 2.77
AmericaEast -0.258 -1.55
American Athletic -0.248 (**) -2.19
Atlantic10 -0.413 (***) -3.01
AtlanticSun -0.550 (***) -3.90
BigSky -0.614 (***) -4.06
BigSouth 0.132 0.36
BigWest -0.143 -0.61
Colonial -0.129 -0.54
ConferenceUSA -0.304 -0.82
Horizon -0.606 (***) -3.50
Ivy -0.031 -0.20
MetroAtlantic -0.605 (***) -3.40
MidAmerican -0.369 (*) -1.74
MidEastern -0.347 -1.46
MissouriValley -0.267 -1.63
MountainWest -0.160 -1.02
Northeast -0.529 (**) -2.21
OhioValley -0.839 (***) -4.22
Southern 0.277 (**) 2.02
Southland -1.561 (***) -11.64
Southwestern -0.251 -1.18
SunBelt -0.747 (***) -5.31
Summit -0.382 (***) -2.82
WestCoast -0.261 -1.63
WesternAthletic -0.477 (***) -2.67
PowerConfOOCGm 0.326 (***) 4.25 0.266 (**) 2.16
Conf Tourney 0.216 (**) 2.48
ConfTournChamp 0.447 (***) 5.09
ScoreSpread -0.004 (**) -2.53 -0.006 -1.16
TotalDiff 0.001 0.54 -0.001 -0.40
Month
November -0.454 (***) -3.41
December -0.186 (*) -1.71
January -0.037 -0.58
February -0.022 -0.34
Day&StartTime
WD9:00-10:59PM -0.030 -0.59 -0.189 -1.00
WD11:00PM-12:59AM -0.148 (*) -1.85 -0.246 -1.31
WDOther -0.188 (**) -2.52 -0.229 -1.60
WE12:00-1:59PM 0.026 0.44 -0.186 -1.01
WE2:00-3:59PM 0.074 1.03 -1.248 (***) -6.31
WE4:00-5:59PM 0.067 0.91 -0.525 (*) -1.69
WE6:00-7:59PM 0.046 0.73 -0.220 -0.91
WE8:00-9:59PM 0.048 0.65 -0.891 (***) -4.13
WE10:00PM&Later -0.015 -0.12 -0.700 (***) -3.70
Substitutes -0.020 (**) -2.18 0.008 0.19
Channel
CBS 0.871 (***) 11.56
ESPN 0.673 (***) 13.87 0.580 (***) 3.34
ESPNEWS -1.116 (***) -10.68 -0.806 (**) -2.52
ESPNU -1.037 (***) -22.27 -1.129 (***) -8.82
FOX 0.133 0.97
FS1 -1.367 (***) -15.42 -1.488 (***) -8.86
FS2 -2.299 (***) -9.42 -2.059 (***) -4.68
NBCSN -1.405 (***) -12.78 -1.452 (***) -7.33
Constant 11.226 (***) 37.31 13.014 (***) 12.46
N 948 121
R-sq 0.874 0.850
adj. R-sq 0.865 0.812
Model 3 Model 4
December January
Coef. z Coef. z
AvePomeroy -0.003 (**) -2.55 -0.004 (***) -3.98
Ave3yrWinPct 1.764 (**) 2.60 1.757 (***) 4.56
AbsClosingLine 0.027 (***) 3.71 0.015 (**) 2.38
OverUnder 0.010 (**) 1.98 0.004 1.04
Conference
ACC 0.410 (*) 1.82 0.536 (*) 1.74
Big12 0.198 0.67
BigEast 0.500 (**) 2.47 -0.168 -0.46
Big10 0.483 (*) 1.90 0.349 1.15
Pac12 -0.144 -0.47
SEC 0.192 0.63
AmericaEast
American Athletic -0.483 (*) -1.79 -0.505 (*) -1.67
Atlantic10 -0.411 -1.16
AtlanticSun
BigSky
BigSouth -0.493 -1.17
BigWest -0.126 -0.33
Colonial -0.416 -0.82
ConferenceUSA
Horizon -0.815 -1.49
Ivy 0.251 0.61
MetroAtlantic -0.868 -1.44
MidAmerican -0.026 -0.06
MidEastern -0.610 (*) -1.83
MissouriValley -0.375 -0.87
MountainWest -0.208 -0.65
Northeast -0.565 -1.61
OhioValley -1.449 (***) -3.53
Southern
Southland
Southwestern -0.017 -0.05
SunBelt
Summit
WestCoast -0.343 -0.67 -0.349 -0.96
WesternAthletic
PowerConfOOCGm 0.287 (**) 2.04 0.704 1.50
Conf Tourney
ConfTournChamp
ScoreSpread -0.006 -1.58 -0.001 -0.20
TotalDiff 0.003 1.31 0.001 0.67
Month
November
December
January
February
Day&StartTime
WD9:00-10:59PM -0.128 -0.98 0.028 0.29
WD11:00PM-12:59AM 0.185 1.15 -0.104 -0.49
WDOther -0.048 -0.25 -0.630 (**) -2.37
WE12:00-1:59PM 0.062 0.33 0.142 1.40
WE2:00-3:59PM 0.261 1.43 0.101 1.09
WE4:00-5:59PM 0.045 0.28 -0.028 -0.18
WE6:00-7:59PM -0.019 -0.12 0.062 0.55
WE8:00-9:59PM -0.040 -0.26 0.040 0.33
WE10:00PM&Later -0.655 -1.57 0.312 (**) 1.99
Substitutes -0.019 -0.48 0.020 1.16
Channel
CBS 1.014 (***) 4.78 0.907 (***) 5.27
ESPN 0.661 (***) 4.32 0.741 (***) 8.27
ESPNEWS -1.564 (***) -4.33 -1.110 (***) -7.56
ESPNU -1.357 (***) -11.58 -0.886 (***) -10.10
FOX -0.169 -0.81 0.491 1.31
FS1 -1.707 (***) -8.87 -0.989 (***) -4.61
FS2 -2.412 (***) -10.73
NBCSN -1.556 (***) -3.51 -1.570 (***) -6.49
Constant 10.480 (***) 13.63 11.158 (***) 18.36
N 170 250
R-sq 0.829 0.913
adj. R-sq 0.793 0.892
Model 5 Model 6
February March
Coef. z Coef. z
AvePomeroy -0.005 (***) -4.59 -0.006 (***) -4.37
Ave3yrWinPct 1.430 (***) 3.62 0.957 (*) 1.88
AbsClosingLine 0.017 (**) 2.33 0.001 0.09
OverUnder 0.007 (**) 2.22 0.003 0.60
Conference
ACC 0.326 (***) 3.19 0.150 0.41
Big12 0.073 0.67 -0.070 -0.18
BigEast -0.284 -1.64 0.177 0.58
Big10 0.174 1.46 0.369 0.98
Pac12 -0.186 -1.23 0.118 0.43
SEC 0.124 0.77 0.470 1.00
AmericaEast -0.319 (*) -1.85 -0.501 (*) -1.83
American Athletic -0.457 (***) -2.87 -0.256 -0.74
Atlantic10 -0.440 (***) -2.80 -0.767 (*) -1.71
AtlanticSun -0.600 -1.32
BigSky -1.195 (***) -3.15
BigSouth 0.122 0.15 0.324 0.74
BigWest -1.021 (**) -2.43 -0.376 (*) -1.78
Colonial -0.485 -1.24 -0.350 -0.61
ConferenceUSA -1.103 (***) -6.69 0.492 1.58
Horizon -0.754 (***) -3.67 -1.158 (***) -2.69
Ivy
MetroAtlantic -0.817 (***) -3.83 -0.758 (*) -1.89
MidAmerican -0.807 (***) -4.54 -0.596 (*) -1.66
MidEastern -0.582 -1.37 -0.336 -0.66
MissouriValley -0.379 (**) -2.16 -0.309 -0.88
MountainWest -0.759 (***) -4.03 -0.516 -1.25
Northeast -1.198 (***) -6.28 -0.451 -1.13
OhioValley -1.210 (***) -7.91 -0.681 (*) -1.69
Southern 0.211 0.60
Southland -2.014 (***) -5.17
Southwestern -0.755 (***) -3.38 -0.442 -1.30
SunBelt -0.757 (***) -3.48 -1.253 (**) -2.46
Summit -0.716 (*) -1.87
WestCoast -0.524 (***) -2.63 -0.365 -0.96
WesternAthletic
PowerConfOOCGm
Conf Tourney 0.393 (***) 3.91
ConfTournChamp 0.307 (**) 2.40
ScoreSpread -0.003 -1.14 -0.006 -1.60
TotalDiff -0.000 -0.12 -0.006 (**) -2.11
Month
November
December
January
February
Day&StartTime
WD9:00-10:59PM 0.045 0.43 -0.003 -0.02
WD11:00PM-12:59AM -0.125 -0.65 -0.438 (**) -2.23
WDOther -0.305 (**) -2.05 -0.517 (***) -3.51
WE12:00-1:59PM 0.126 1.04 -0.164 -0.77
WE2:00-3:59PM 0.228 (*) 1.77 -0.100 -0.68
WE4:00-5:59PM 0.332 (***) 3.09 0.041 0.22
WE6:00-7:59PM 0.179 1.40 0.109 0.71
WE8:00-9:59PM 0.217 1.26 0.448 (*) 1.92
WE10:00PM&Later 0.423 (**) 2.13 -0.395 -1.58
Substitutes -0.051 (***) -3.68 -0.073 (***) -3.02
Channel
CBS 0.976 (***) 8.73 1.005 (***) 5.08
ESPN 0.824 (***) 8.36 0.594 (***) 4.44
ESPNEWS -1.106 (***) -5.17 -0.447 (*) -1.72
ESPNU -0.906 (***) -10.32 -1.042 (***) -6.31
FOX 0.498 (***) 3.16 0.268 0.84
FS1 -1.007 (***) -7.10 -1.578 (***) -6.08
FS2 -1.373 (***) -7.46 -2.362 (***) -5.98
NBCSN -1.229 (***) -5.80 -1.235 (***) -3.37
Constant 11.271 (***) 25.73 12.501 (***) 16.65
N 252 155
R-sq 0.922 0.948
adj. R-sq 0.903 0.919
Note. (*) p < 0.10, (**) p < 0.05, (***) p < 0.01
COPYRIGHT 2018 Fitness Information Technology Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2018 Gale, Cengage Learning. All rights reserved.
