2. Scope of Project
■ To determine whether social media, mainlyTwitter has a
significant effect on who receives the most votes for the
ASG
■ Prior Research 40 year study
– 1980s – Advanced Stats
– 1990s- Advanced andTraditional
– 2000s- Popularity
– 2010s-Twitter with Stats
3. Scope of Project
■ 2015 ASG
– Ballot released on May 3 and voting closed on July 3
– Throughout voting MLB released the names of the top 1/3
of players receiving votes at each position
■ Collected data for each voting period
4. Variables
■ Dependent variable- Place on Ballot
– 1st through 5th or off
■ IndependentVariables (all indexes)
– Traditional Statistics
– Advanced Statistics*
– Player Popularity
– Team Popularity*
– TeamTweet Efficiency (ratio engagement to tweets)
– PlayerTweet Efficiency
5. Model
■ Applied an Ordered Logit
– Non-linear procedure that looks at which variables
influence where the player finished in the voting at each
ballot update
■ Odds of being in a higher place on the ballot rather than being
lower
6. Significance is based on Standard Errors which are not reported on the table
Variable Period 1 Period 2 Period 3 Period 4 Period 5 Period 6
Est. Odds Ratio Est. Odds Ratio Est. Odds Ratio Est. Odds Ratio Est. Odds Ratio Est. Odds Ratio
RPPALL 0.015 1.015 0.014 1.014 0.014 1.014 0.014 1.014 0.013 1.013 0.014 1.015
TI((P1-P6) 0.002 1.002 -0.117 0.890 -0.102 0.903 -0.089 0.914 -0.080 0.923 -0.080 0.923
TTE(P1-P6) 0.120 1.128 0.566 1.762 0.041 1.042 0.161 1.175 0.190 1.209 0.133 1.142
PTE(P1-P6) 0.034 1.035 0.119 1.127 0.019 1.019 0.006 1.094 0.062 1.064 -0.010 0.990
R-SQUARE 0.2387 0.4267 0.3157 0.3381 0.3183 0.3050
Coefficients in RED are statistically significant at the .01 level; those in GREEN at the .05 level; those in BLUE at the 0.10 level. All tests are one-tailed.
Cumulative Logit Results forVoting for Major League All-Stars, 2015
7. HowTeams are usingTwitter
Focus on performance statistics
Teams can useTwitter to respond to what fans are voting on in a particular period
11. Moving forward
Age and relationships to voters
0
1
2
3
4
5
6
7
8
9
10
AuthorizedTwitter Accounts ByTeam
Yes
No
12. Moving Forward -Multinomial logic
■ Relax the ordered logit’s assumption that the parameter
estimates do not vary by position on the ballot
■ Allows me to look at odds of being in 1st vs. 2nd, etc.
■ Should reduce correlation between fandom and tweets
13. Moving Forward- Nationality/Identification and
Twitter Engagement
1
1
6
12
2
26
1
1
1
5
1
175
23
Aruba
Brazil
Canada
Cuba
Curacao
Domican Republic
Japan
Netherlands
Panama
Puerto Rico
South Korea
United States
Venezuela
0 20 40 60 80 100 120 140 160 180 200
Birth Country of Players on 2015 ASG Ballot
Editor's Notes
Limitation- top 1/3- where they finished. Weak Twitter historical data available.
Larger goal- understand how social media can be used by MLB, teams and players successfully
Long term- understand WHY tweets are liked, RT, responded to
Larger goal- understand how social media can be used by MLB, teams and players successfully
Long term- understand WHY tweets are liked, RT, responded to
Every tweet sent from team and players mentioning ASG
Correlation between popularity so high- only used Player Pop
Correlation between stats so high only used traditional
In every period, a one-point increase in the player popularity index increases the odds of being in a higher place on the ballot rather than being lower in the voting by about 1.014 times
The odds of being in a higher category are about 90% of the odds of being lower on the voting scale for a one-point increase in the index of traditional statistics
Team tweet efficiency seems to make a statistically significant difference in the player’s place of the ballot in periods 2, 4, 5 and 6.
Player tweet efficiency has a statistically significant effect on the outcomes in the ordered logit model estimated for periods 2, 4, and 5, with the largest effect found in Period 2
Point out vote reduction at end of 4- why both pop
Miggy
Jersey Boys- todd Fraizer
Even when normalize still widely different
Clear that there are diminishing returns
Often same tweet- different results
Average age all-star viewer mid-50s
how many on twitter- how many vote
AL- 9
NL-8
Twitter to be effective at getting votes as to generate engament and fandom-- clients of Jay-Z, friends of famous people, videos of celebrities, russell wilson
Expect biggest impact to happen at 3-5—smaller markets