Sunteți pe pagina 1din 33

Social Media in AA Baseball

By:
Zach Gerhart
Social Media in AA Baseball

Table of Contents

AA Executive Summary…………………………………………………………….2
Introduction ………………………………….…………………………………….....4
Attendance Testing ………………………………………………………………….5
Regression Estimates ………………………………………………………………6
Coefficient Interpretation ……………………………………………………….10
Model in Use ………………………………………………………………………….11
Improving Social Media …………………………………………..……………..12
Summary ………………………………………………………………………………15
Appendix A: Alternate Models …………………………………………………16
Appendix B: Social Media Tests ………………………………………………21
Appendix C: Team Data ………………………………………………………….25

1
Social Media in AA Baseball

AA Executive Summary

This study tested multiple influences of Minor League Baseball attendance, including
social media metrics such as interactions and followers, to find out which factors are
statistically significant to Minor League Baseball teams in their AA league. The findings
of this study could potentially lead to a better understanding of attendance metrics and
how to generate a higher attendance for all teams in the AA league.
Attendance was the dependent variable that this study attempted to predict. Attendance
was measured and tested two different ways. The first way was the average attendance
at each home game for each team. The second way was to measure attendance as a
percent of the market that each team was able to get to a game. Testing with market
capture percent allows us to control for market size when studying attendance. The
market capture percent was determined by taking the total attendance a team acquired
over a year and dividing it by the team’s market size, which was determined by the
Nielsen TV market size. In the event that this was not available for a city or not
representative of the true market size (e.g. Frisco is in Dallas TV Market), the population
of the city’s county was used.
The independent variables to test both measures of attendance were as follows: Win
Percent (tested three different ways, previous year’s win percent, current year win
percent, and an average of the two-year win percent), Miles from MLB Stadium, Team
Age, Stadium Age, Market Size, Twitter Interactions, Instagram Interactions, Facebook
Interactions, Instagram Followers, Twitter Followers, and Facebook Likes.
After testing both attendance variables with each of the independent variables, the
model that was able to be most accurate was the model using Average Attendance. All
three methods of win percentage proved to be insignificant in the final results when
testing for average attendance. The final model produced an R-squared of .610, meaning
that this model could predict the average attendance of an AA team in 2017 with 61.0%
confidence, making this an okay model. The model is shown below.

Figure 1: Final Model for Average Attendance for AA teams

2
Social Media in AA Baseball

This model shows that as far as social media goes, Facebook interactions and Facebook
followers were the only statistically significant social media metrics in the model. In
addition to these two social media metrics, other significant metrics were Miles from
MLB Stadium, Market Size, and Years in Current Stadium.
For a AA team to increase their average attendance, they should focus their social media
resources on Facebook. In 2017, Facebook was the only statistically significant social
media platform. In order for a AA team to grow both their Facebook interaction and
their Facebook followers, they should focus their resources on providing fans with more
engaging content, especially videos on Facebook. Videos performed exceptionally well in
regard to interactions compared to other content posted on Facebook like photos or
links. However, teams should be weary of the number of posts they have on Facebook as
the number of all types of Facebook posts had a negative relationship with the number
of followers on Facebook. One team, the Frisco RoughRiders, have all the metrics to
have one of the highest Facebook followings as they have the highest average attendance
and serve one of the larger markets in AA baseball. However, they only have the 19th
highest Facebook following. One potential reason for this is the amount of posts they
have on Facebook. Frisco posted 33% more than the next highest team, sending out
2,158 Facebook posts over the season where the second highest was only 1,404.
One non-social media result from this that was notable was that the further a team was
from a Major League Baseball stadium, the less attendance they tended to have. One
potential reason for this is that fans closer to an MLB stadium want to go to baseball
games but prefer the less expensive option of Minor League Baseball. Another reason
could be that teams further away from a Major League Baseball stadium tend to serve
much smaller markets. The AA teams that are closer to Major League Baseball teams
tend to be in the suburbs of the city of that MLB team, so they tend to have a bigger
market than those that are further away from an MLB team.
In conclusion, the model shows that the social media use of a team is significant in
predicting how much each team can capture of their market. The model was able to
explain 61.0% of the variance in average attendance for AA Minor League Baseball
teams. After testing for three social media platforms (Instagram, Twitter, and
Facebook), Facebook was the only one that was statistically significant. For teams to
increase attendance, they should try to focus on Facebook, especially videos. If a team
were to be produce more engaging videos on Facebook, they could have a higher average
attendance at their home games.

3
Social Media in AA Baseball

Introduction
When trying to grow attendance for AA Minor League Baseball teams, it is
important to know what metrics are significant in the attendance equation. The initial
question that was attempted to solve with this study was if social media was an
important factor in getting fans to the ballpark or not. In addition to social media
metrics, the study tests other metrics that could potentially influence a team’s
attendance such as how long the team has been in their current city, their proximity to a
superior substitute (Major League Baseball), Market Size, and others. In addition to
testing if social media is a factor, this study attempts to see what kind of social media
posts influence the amount of followers a team has on each social media platform.

4
Social Media in AA Baseball

Attendance Testing
To test attendance, this study looked at two different attendance metrics to try
and provide a model that would be as confident as possible. The two dependent
variables that the study tested for attendance was Average Attendance for a Game and
Market Capture Percentage, which was a metric derived from taking each team’s total
attendance over a season and dividing it by the market size. The market size was
determined by Nielsen TV Market for each team. For some teams, the market was either
not reflective of the team’s true market or it wasn’t available (e.g. Frisco is in Dallas TV
market). For these teams, the market size was determined by using the most recent
population data for the county in which the team is located. Each model started with 11
independent variables. The independent variables are listed below:
• Win Percentage, tested three different ways
• Current year win percentage
• Last year’s win percentage
• Average win percentage between current and last year
• Miles from nearest Major League Baseball Team
• How long the team has been established in the city
• How long the team has been in its current stadium
• Market Size
• Twitter Interactions (Retweets + Likes)
• Facebook Interactions (Likes + Comments + Shares)
• Instagram Interactions (Likes + Comments)
• Twitter Followers
• Instagram Followers
• Facebook Likes
The data used in these models was obtained from a variety of sources. The
attendance and social media data was obtained from Minor League Baseball
Headquarters. Each team’s win percentage was obtained from Baseball Reference. Miles
from nearest Major League Baseball team was acquired from Google Maps by using the
city the current team is located in and getting the driving distance to the nearest MLB
stadium. The length a team has been located in a city and how long each team has been
in its city was obtained from each team’s Wikipedia page. Market size was obtained by
Nielsen TV market size or county population when applicable.

5
Social Media in AA Baseball

Regression Estimates
After testing each attendance variable with each set of independent variables, including
each different Win Percent option, the best model was one measuring attendance using
Average Attendance per Game. When testing for Win Percent, all three iterations of Win
Percent were statistically insignificant. This final model produced the highest adjusted
R-squared of any of the six different models tested for AA baseball. For reference, you
will be able to find the other five models in Appendix A: Alternate Models
. Below is the initial model and the process to reduce it to the final model for the model
best fitting model.

Initial Model
The initial regression is as follows:
Average Attendance = 5,384 - 2,236(Current Year Win Percent) - 6.046 (Miles from
MLB Stadium) - 4.628(Year’s Team in City) + .001578 (Market Size) - 34.60 (Stadium
Age) -.007605 (Twitter Interactions) +.01645 (Facebook Interactions) - .01137
(Instagram Interactions) + .06888 (Instagram Followers) -.002498 (Twitter Followers)
+ .02008 (Facebook Followers)

Figure 2: The initial model of average attendance of AA teams using current year’s win
percentage as variable for measuring team’s win percentage

In the initial model, the F-statistic p-value (.0020) is less than the level of
significance of 0.10 so at least one of the independent variables probably affects the

6
Social Media in AA Baseball

dependent variable, Average Attendance. The Adjusted R-Squared is 58.2% which


means that the independent variables explains 58.2% of the variance in the dependent
variable. The t-statistic p-value for two of the variables is initially below the level of
significance of 0.10 (Miles from MLB Stadium and Market Size). Most of the variables
were above the level of significance of 0.10 (Win Percent, Team Age, Stadium Age,
Twitter Impressions, Facebook Interactions, Instagram Interactions, Instagram
Followers, Twitter Followers, and Facebook Follwers). With the initial model passing
the F-test, having an Adjusted R-Square of .5821, and most variables failing their t-test,
the initial model was poor and had some work to do.

After completing the initial model, the next thing that the study did was to check for
multicollinearity. The test for multicollinearity is used to see if any of the variables are
highly correlated with each other. If two variables are highly correlated with each other
or an insignificant variable highly influences a significant one, the decision whether to
keep or drop a variable from the model could change.

Figure 3: The correlation matrix for each of the independent variables used in testing

Since no variables were highly correlated (Beta of 0.8-0.9), their correlation to each
other does not affect whether to drop or keep any of the variables.

7
Social Media in AA Baseball

Final Model
After dropping out statistically insignificant variables one by one, the study was able to
develop the following final model:

Average Attendance = 3,866 -5.407 (Miles from MLB Stadium) +.001492 (Market Size)
- 32.05 (Stadium Age) +.01461 (Facebook Interactions) +.02364 (Facebook Followers)

Figure 4: The final model of average attendance of AA teams

In the final model, the F-statistic p-value (.0000) is less than the level of
significance of 0.10 so at least one of the independent variables probably affects the
dependent variable, Average Attendance. The Adjusted R-Squared is 61.0% which
means that the independent variables explains 61.0% of the variance in the dependent
variable. The t-statistic p-value for all of the variables is below the level of significance of
.10. With the final model passing the F-test, having an Adjusted R-Square of .6101, and
all of the variables passing their t-test, the final model is okay for predicting the average
attendance of AA teams.
In looking at the final model, there are a few things that stand out. The first thing that
stands out is that Facebook is the only statistically significant social media platform.
After testing for both interactions and followers on Instagram, Twitter, and Facebook,
Facebook interactions and Facebook followers were the only social media metrics to be
statistically significant. For every 100 followers that a team has, they will get 2.37 people
in attendance per game. For 100 interactions that a team has, they will get 1.46 people in
attendance per game. This shows that teams that are able to gain Facebook as a
marketing channel to individuals (followers), then they will be more likely to get them to
a game. Also, if the team is able to put out content that fans want to interact with (likes,
comments, or shares), then they will get even more attendance. As fans see the message
that teams put out and it’s one they want to interact with, they are more likely to come
to the stadium and watch the team.

8
Social Media in AA Baseball

Another interesting result of the model was that teams that had newer stadiums would
attract a higher attendance. This could either be the result of newer stadiums having a
higher capacity, or that fans want to go to a newer stadium. On average, for every year
old that the stadium is, a team can expect 32 less fans per game.
The final interesting result from this set of data was that as teams were located farther
from a Major League Baseball stadium, their average attendance went down. For every
100 miles a team was from a MLB stadium, their attendance would drop 541 fans per
game. This could be due to a number of reasons. One reason that it could be is that fans
closer to a Major League Baseball stadium are more likely to be baseball fans and attend
more. However, since the number is so high, an alternative reason is probably more
likely. One alternative reason would be that the teams that are in a smaller market
happen to be the furthest from a MLB stadium. Looking at the data for 2017, the teams
with the three lowest attendance (Jackson Generals, Biloxi Shuckers, and Mobile Bay
Bears) were all 250+ miles from an MLB Stadium. There were three teams with an
average attendance over 6,000 (Frisco RoughRiders, Reading Fightins, and Richmond
Flying Squirrels) and the furthest one of these was from an MLB stadium was the Flying
Squirrels at 109 miles. These teams also have a higher than average market size for the
MLB stadium. While the distance from a MLB stadium does have a negative correlation,
the study believes that it could be due to alternate reasons such as market size, not that
there are less baseball fans.

9
Social Media in AA Baseball

Coefficient Interpretation

Figure 5: Final model’s coefficients for the independent variables showing their influence
on the dependent variable, average attendance.

10
Social Media in AA Baseball

Model in Use

Figure 6: An example of the model in use with two different teams, one with a lower
average attendance and one with a higher average attendance

11
Social Media in AA Baseball

Improving Social Media

Facebook
Since Facebook interactions and followers were both statistically significant to
predicting average attendance of a AA team, this study wanted to look at what metrics
from Facebook could influence these two. If teams were to have a better understanding
of what could increase their Facebook interactions and followers, they could also
increase the amount of people they have at the stadium.
Using both Facebook interactions and Facebook followers as the dependent
variable, the study looked at four different Facebook metrics to see their independent
relation with each of the dependent variables. The four metrics were as follows:
• Number of Total Posts
• Number of Photo Posts
• Number of Video Posts
• Number of Link Posts
Unlike predicting attendance where all of the independent variables were used in
a total model where each independent variable could potentially influence each other,
this study just looked at each independent variable independently. This method was
used due to their being many factors outside of social media use that would affect the
number of Facebook interactions and Facebook followers such as the general size of fan
base. This method allows us to look at just the social media metrics and their influence
on these two Facebook metrics.

Figure 7: Results from independent testing of Facebook metrics showing their influence on
Facebook interactions

12
Social Media in AA Baseball

Figure 8: Results from independent testing of Facebook metrics showing their influence on
Facebook followers

The testing from looking at Facebook interactions shows that content that is
more engaging is more likely to be highly interacted with. Videos performed extremely
well as video posts on average had five times the interactions compared to any other
type of posts. From previous studies, videos have been shown to perform exceptionally
well on Facebook and it is the same for AA Minor League Baseball teams. If teams want
to grow their Facebook interactions to potentially increase their average attendance,
they should focus on posting engaging videos.
However, as we see in the Facebook followers study, teams may want to not focus
on the quantity of posts but rather the quality. The study showed that no matter the type
of posts, as number of posts went up, followers went down. This could be for a number
of reasons. One of these reasons is that Facebook followers is not representative of the
posts that a team has. This could be that teams do a good job of getting their fans at the
ballpark to follow them on Facebook. Since Facebook followers positively correlates with
average attendance, there could also be a reverse relationship where teams with higher
attendance have more fans, negating the relation of posts to followers (also could
explain the lower confidence).
A second reason that every post type has a negative relation is that teams are
posting too much, driving off potential followers. Looking at the individual team data,
one team stands out in particular. The Frisco RoughRiders had only the 19th most
Facebook followers out of 30 teams despite being in a larger market and having the
highest average attendance of all Minor League Teams. These two factors alone should
be enough to have a higher fan base and have a top 5 Facebook following. The Frisco
RoughRiders had 2,158 Facebook posts during the regular season, over 700 more than
the next highest team. They also had 1,596 photo posts, over 700 more than the second
highest team again. The Frisco RoughRiders could be a prime example of a team that is
focusing too much on quantity, when they should be focused on quality. Other teams
with high post count are also on the lower side of Facebook followers, so having too
much quantity could be the reason for the negative coefficients.

13
Social Media in AA Baseball

In conclusion, for teams to capitalize on their social media to attract more fans to
the ballpark, they should focus on producing engaging content. Of Instagram, Twitter,
and Facebook social media platforms, Facebook was the only one that was statistically
significant to predicating average attendance in 2017. With Facebook, videos perform
exceptionally well compared to other content such as photos and links. However, teams
should be cautious of the amount a team was posted. As the number of posts of all type
of posts went up, the average number of followers went down.

14
Social Media in AA Baseball

Summary
Overall, the study was able to answer the questions it set out to answer. Those
questions were does social media have an influence on attendance and how can teams
grow their social media. The study was able to produce a model that had five significant
independent variables and predict the dependent variable, average attendance, with
61.0% confidence. Two of those independent variables were social media measurers,
both coming from Facebook: Facebook interactions and Facebook followers. Facebook
interactions and Facebook followers were able to have a positive influence on the
average attendance of a team. For every 100 Facebook followers a team had, they
averaged 2.37 fans per game above expectations. This shows that if teams are able to
increase their social media presence, they are able to make an impact of their attendance
over a season.
After the study answered that social media does have an influence on attendance,
the study set out to see how teams can grow their followers through their social media
use. The study showed that teams that use more engaging content such as photo, videos,
and links on Facebook were able to have more interactions. As seen in previous studies,
videos performed exceptionally well, averaging five times the number of engagements
on video posts compared to that of photo or link posts. However, teams should be weary
of the amount of posts they have. As the number of posts a team has increased, the
number of followers a team had tended to decrease. One example of this is the Frisco
RoughRiders. The Frisco RoughRiders have the highest average attendance and serve in
a relatively large market, but only have the 19th highest Facebook follower amount out of
30 teams. One reason that the fan base doesn’t translate over to Facebook is that they
have the highest number of Facebook posts by any team by 33%. Teams should focus on
creating engaging, quality content over high quantity content as potential followers
which turn in to potential fans could be turned off at the number of posts on a team’s
feed.
In conclusion, the study was able to show that social media does have a positive
influence on attendance. Teams that had more Facebook interactions and Facebook
followers were more likely to have a higher attendance. Teams can grow their
interactions Facebook by using more engaging content like photos and links but should
especially focus on videos for Facebook. However, teams need to manage the quantity of
their content as well as the quality of the content. If teams are able to gain more
followers on Facebook and have more interactions with fans, they should be able to
increase their average attendance.

15
Social Media in AA Baseball

Appendix A: Alternate Models

Figure A-1: Initial model for market capture percent of AA teams using current year win
percent as variable for winning percentage

Figure A-2: Final model for market capture percent of AA teams using current year win
percent as variable for winning percentage

16
Social Media in AA Baseball

Figure A-3: Initial model for market capture percent of AA teams using last year’s win
percent as variable for winning percentage

Figure A-4: Final model for market capture percent of AA teams using last year’s win
percent as variable for winning percentage

17
Social Media in AA Baseball

Figure A-5: Initial model for market capture percent of AA teams using an average of last
year’s win percent and current year’s win percent as variable for winning percentage

Figure A-6: Final model for market capture percent of AA teams using an average of last
year’s win percent and current year’s win percent as variable for winning percentage

18
Social Media in AA Baseball

Figure A-7: Initial model for average attendance of AA teams using last year’s win percent
as variable for winning percentage

Figure A-8: Final model for average attendance of AA teams using last year’s win percent
as variable for winning percentage

19
Social Media in AA Baseball

Figure A-9: Initial model for average attendance of AA teams using an average of last
year’s win percent and current year’s win percent as variable for winning percentage

Figure A-10: Initial model for average attendance of AA teams using an average of last
year’s win percent and current year’s win percent as variable for winning percentage

20
Social Media in AA Baseball

Appendix B: Social Media Tests

Figure B-1: Results for the number of Facebook posts a AA team posts to the number of
Facebook interactions that the team has

Figure B-2: Results for the number of Facebook photo posts a AA team posts to the
number of Facebook interactions that the team has

21
Social Media in AA Baseball

Figure B-3: Results for the number of Facebook video posts a AA team posts to the number
of Facebook interactions that the team has

Figure B-4: Results for the number of Facebook link posts a AA team posts to the number
of Facebook interactions that the team has

22
Social Media in AA Baseball

Figure B-5: Results for the number of Facebook posts a AA team posts to the number of
Facebook followers that the team has

Figure B-6: Results for the number of Facebook photo posts a AA team posts to the
number of Facebook followers that the team has

23
Social Media in AA Baseball

Figure B-7: Results for the number of Facebook video posts a AA team posts to the number
of Facebook followers that the team has

Figure B-8: Results for the number of Facebook link posts a AAA team posts to the number
of Facebook followers that the team has

24
Social Media in AA Baseball

Appendix C: Team Data

25
Social Media in AA Baseball

26
Social Media in AA Baseball

27
Social Media in AA Baseball

28
Social Media in AA Baseball

29
Social Media in AA Baseball

30
Social Media in AA Baseball

31
Social Media in AA Baseball

32

S-ar putea să vă placă și