Analysis of the Candidate.ie project by Dr Ciar叩n Mc Mahon of Dublin Business School - Social media usage by candidates in the 2011 Irish General Election
1 of 64
Downloaded 19 times
More Related Content
Mc Mahon, C. (2011). Social media usage by candidates in the 2011 Irish General Election
1. Candidate.ie
Social media usage by candidates in the
2011 Republic of Ireland General Election
Dr Ciar叩n Mc Mahon,
Department of Psychology,
Dublin Business School.
2. Background
2010 US mid-term elections Facebook.com (3 November,
2010)
98 House races
74% of candidates with the most Facebook fans won
Senate 19 races
81% of candidates with the most Facebook fans won
Donegal South-West by-election (24th November, 2010)
high correlation observed between 1st preference votes
and Facebook friends (r = 0.889, N = 5, p < 0.05).
4. Data collection
Follower/fan data collected from 21st to 24th February 2011.
T-tests comparing data collected at either extreme of time period
revealed no differences.
1st preference vote share was collected from the RT News election
Candidates gender was inferred from their names (!)
Age was sourced and cross-referenced from various reputable
online sources, and later politely requested via email.
Population density data was calculated from constituency area
measurements gratefully supplied by Richard Cantillion of GAMMA
Ltd. using D叩il Boundaries provided by Ben Raue,
www.tallyroom.com.au and population statistics from the Central
Statistics Office.
5. Data collection
Election
566 candidates
481 male, 85 female.
Reliable date of birth sourced for 372
average 48.22 (range 22 75, st. dev. 11.451).
125 incumbents stood for election
6. Data collection
Twitter
325 accounts,
average 467.52 followers (2-11465, st. dev. 1039.81,
sum 151,945)
hence 57.4% of all candidates were on Twitter
7. Data collection
Facebook
432 accounts
average of 730.65 friends/fans/members (1 to 5000,
st.dev. 875.10, sum 315,640)
78.8% of all candidates had Facebook accounts of one
kind or another
8. Data collection
Facebook
432 accounts
316 Friend accounts, 112 Fanpages, and 3 were groups.
14 private Facebook accounts were also identified, but as their
Friend totals were not visible, had to be removed from the final
analysis.
When a candidate had both a Friend and Fanpage, the page with
the highest level of support, as a better indicator of popularity,
was used (unless the candidate requested otherwise).
NB no difference between candidates with Friend or Fanpages in
terms of likelihood of getting elected (U = 17578, N1 = 317, N2 =
112, p = .850 ) nor more likely to get votes (t = .356, df =
182.019, p = .722)
10. Research Questions
What factors are related to
having a presence on social media?
being popular on social media?
How does social media presence and
popularity relate to votes received?
Ultimately, how does social media relate to
getting elected....
11. U have no chance if ur not on
facebook and twitter. If u can't do
social media then can u do anything
at all?
Male, Labour, Louth
12. 1. Gender
Presence:
Facebook
Female candidates more likely to have an account 2(1, N = 566) =
4.085, p < 0.5. However weak strength association = 0.085.
Twitter
No difference between genders in having accounts 2(1, N = 566) =
2.929 , p = 0.087.
Popularity
Facebook
no difference in friends/fans (t = 1.060, df = 430, p = .290)
Twitter
no difference in followers (t = -0.367, df = 323, p = 0.714)
13. 2. Age
Younger candidates expected to be more popular
Facebook
no negative correlation observed between a candidates age
and their number of friends/fans (r = - .059, N = 310, p = >
.05).
Twitter
no negative correlation observed between a candidates age
and their number of followers (r = .073, N = 254, p = > .05).
Further analyses, as regards age predicting presence
on social media, did not reveal models of any
significance
14. 3. Party affiliation
Facebook
Presence
differences observed across parties in terms of their
candidates having a Facebook account
2(10, N = 566) = 81.280, p < 0.001, with moderate strength
observed = 0.379
16. 3. Party affiliation
Facebook
Popularity
differences between the parties in terms of their
candidates popularity on Facebook (F(9,422) = 6.040, p
< 0.01)
18. 3. Party affiliation
Twitter
Presence
differences between parties and the likelihood of their
candidates having a Twitter account 2(10, N = 566) =
86.268, p < 0.001, with moderate strength observed
= 0.390
20. 3. Party affiliation
Twitter
Popularity
differences between the parties in terms of their
candidates popularity on Twitter (F(9,315) = 2.454, p <
0.05)
22. 4. Constituencies
Facebook
Presence
no differences across constituencies 2(1, N = 42) =
52.129, p = 0.136.
Popularity
no differences across constituencies in terms of
candidates popularity on Facebook (F(42, 431) = 1.399,
p = 0.056).
23. 4. Constituencies
Twitter
Presence
no differences across constituencies 2(1, N = 42) =
43.743 , p = 0.397.
Popularity
differences across constituencies in terms of popularity
on Twitter (F(42, 324) = 1.508, p < 0.05).
25. 5. Urban/rural
Facebook
Presence
no differences between urban and rural candidates
and their having a Facebook account 2(1, N = 566) =
0.278, p = .598.
Popularity
differences between urban and rural candidates in
terms of their candidates popularity on Facebook (t =
2.297, df = 366.277, p < 0.05) with a confidence interval
of CI95 (27.876, 359.798).
27. 5. Urban/rural
Twitter
Presence
a difference between urban and rural candidates and
their having a Twitter account 2(1, N = 566) = 3.986, p
< 0.05, though with only weak strength = - 0.084.
29. 5. Urban/rural
Twitter
Popularity
statistically differences between the urban and rural
candidates in terms of their candidates popularity on
Twitter (t = - 2.267, df = 323, p < 0.05) with a
confidence interval of CI95 (- 496.006, -35.083).
31. 6. Incumbency
Facebook
Presence
no differences between incumbent and non-sitting with
regard to their having a Facebook account 2(1, N =
566) = 3.459, p = .063.
Popularity
statistically differences between the incumbent and
non-sitting candidates in terms of their candidates
popularity on Facebook (t = 3.532, df = 430, p < 0.001)
with a confidence interval of CI95 (156.620, 549.525).
33. 6. Incumbency
Twitter
Presence
difference between incumbent and non-sitting
candidates in terms having a Twitter account 2(1, N =
566) = 13.948, p < 0.001, though with only weak
strength = .157.
35. 6. Incumbency
Twitter
Popularity
statistically differences between the incumbent and
non-sitting candidates in terms of their candidates
popularity on Twitter (t = - 4.126, df = 323, p < 0.001)
with a confidence interval of CI95 (271.585, 766.660).
37. My advice would be don't set up an
account unless you are going to use
it. I followed several candidates who
rarely posted. And I certainly felt that
many of the candidates had nothing
to do with their accounts, that it was
a team member. I would advise that
if it is a team member, that they are
upfront about it.
Female, Greens, Dublin South
38. 7. Votes
Correlation
Facebook
positive correlation observed between a candidates
popularity on Facebook and the number of first
preference votes received (r = .450, N = 432, p < 0.01,
one-tailed)
Twitter
positive correlation observed between a candidates
popularity on Twitter and the number of first
preference votes received (r = .164, N = 325, p < 0.01,
one-tailed)
39. 7. Votes
First preferences
Facebook
difference between the number of first preference votes
received by candidates who had a Facebook account and
those who did not F(1, 562) = 6.019, p < .05.
Twitter
difference between the number of first preference votes
received by candidates who had a Twitter account and those
who did not F(1, 562) = 19.404, p < .001.
but no interaction effect for having both F(1, 562) =
1.098, p = .295.
i.e. an effect for one or the other, but no bonus
40. 7. Votes
Twitter Facebook Mean Std. Dev. N
Yes Yes 4885.26 3663.938 306
No 4241.58 3486.822 19
Total 4847.63 3651.784 325
No Yes 3347.78 3032.523 140
No 1744.45 2952.871 101
Total 2675.84 3096.398 241
Total Yes 4402.65 3547.363 446
No 2139.82 3162.951 120
Total 3922.90 3588.197 566
42. 7. Votes
Multiple regression analysis (stepwise method)
predictor variables:
number of followers
number of friends or fans
incumbency
population density
constituency
urbanity
affiliation.
model emerged F(7, 287) = 35.702, p < 0.005 which
predicted 45.2% of the variance.
Facebook support, incumbency and party were
predictors, though the others were not.
43. 7. Votes
B Std. Error Beta
Number of followers .029 .160 .009
Number of friends or fans 1.188 .185 .306*
If they had a seat -1943.192 404.823 -.233*
Constituency -.299 13.533 -.001
Urbanity 117.151 442.337 .016
Population density -.293 .145 -.120
Affiliation -380.579 45.418 -.401*
44. Pre -Social media there was no way
Mick Wallace could have amassed
13,000 first prefernce votes in the
space of 3 weeks in Wexford.
The lads running his page deserve a
medal.
Male, Independent, Wexford
45. 8. Success
Facebook
those with accounts were more likely to get
elected than those without 2(1, N = 566) =
14.767, p < 0.0005 . = 0.162.
49. 8. Success
Incumbents
no differences among incumbents between those
who had Facebook accounts as to whether or not
they won a seat 2(1, N = 125) = .775, p =.397.
no differences among incumbents between those
who had Twitter accounts as to whether or not
they won a seat 2(1, N = 125) = .080, p = .777.
50. 8. Success
Challengers
Facebook
differences between those who had Facebook
accounts as to whether or not they won a seat 2(1, N =
441) = 19.339, p = .0005. = .209
52. 8. Success
Challengers
Twitter
differences between those who had Twitter accounts
as to whether or not they won a seat 2(1, N = 441) =
24.198, p = .0005. = .234
54. 8. Success
(Continuing to ignore incumbents, no differences)
Urban
Facebook
Challengers who had accounts were not more likely to
get elected than those who did not 2(1, N = 155) =
2.845, p = .092.
Twitter
Challengers who had accounts were more likely to get
elected than those who did not 2(1, N = 155) = 8.114, p
= .005. = .229.
56. 8. Success
(Continuing to ignore incumbents, no differences)
Rural
Facebook
Challengers who had accounts were more likely to get
elected than those who did not 2(1, N = 286) = 17.577,
p = .0005. = .248
Twitter
Challengers who had accounts were more likely to get
elected than those who did not 2(1, N = 286) = 15.972,
p = .0005. = .236
61. A curious effect...
Facebook
comparing incumbents/challengers with successful/failed
candidates
a significant effect of success on number of friends/fans F(1,
428) = 34.323, p < .0005
but no effect of success
Twitter
comparing incumbents/challengers with successful/failed
candidates
a significant effect of success on number of followers F(1,
562) = 14.422, p < .05
but no effect of incumbency
62. 8. Success
Logistic regression model
success in election
predicted using urbanity, population density, party
affiliation, number of Twitter followers, number of
Facebook friends/fans, incumbency, age and gender as
variables
233 cases examined
predicted election (omnibus chi-square 146.37, df =
16, p < 0.0005)
accounted for between 46.6 and 62.3% of the
variance, with 81.8% of successful and 84.6% of
unsuccessful candidates correctly predicted
64. Success
Interestingly, while the number of Twitter
followers a candidate had does not seem to
had a impact on their success, each
Facebook fan or friend increased their
chance of getting elected by a factor of
1.001...