Click here to load reader
Upload
ioannis-andreadis
View
957
Download
1
Embed Size (px)
DESCRIPTION
Presented at VAA Research –State of the Art and PerspectivesSchloss Mickeln, Düsseldorf, March 2-3, 2012
Citation preview
Lessons Learnt from Greece -Analysis of the First Greek VAA
Andreadis I., Chadjipadelis Th. Department of Political Sciences,
Aristotle University of ThessalonikiGreece
2/3/2012 VAA Research – State of the Art and Perspectives
2
Outline of the paper
• Data quality– Data cleaning– Users’ demographic characteristics
• How do they select candidates?– Any evidence of “pure” issue voting?
• Impact of VAAs on vote choice– It depends on the confidence level of initial
selection
2/3/2012 VAA Research – State of the Art and Perspectives
3
Helpmevote popularity – sample size
• Helpmevote: First Greek VAA – It was used for the Greek Regional Elections of 2010
• In 20 days (18/10/2010 - 7/11/2010) more than 28,000 users
• 3 ways to collect information:– Position on issues (all users)– Personal information (after the issues, 21000)– Web survey after the use of VAA (5000)
2/3/2012 VAA Research – State of the Art and Perspectives
4
Data quality – Data Cleaning
• Time spent to answer each question is less than one second– The user did not have enough time to read,
understand and give a substantial answer to the question
• The pattern of answers – The user has clicked the same button (e.g. "Totally
agree" for all (or almost all) the questions• The final cleaned version of the dataset includes
28014
2/3/2012 VAA Research – State of the Art and Perspectives
5
Representative sample? Sex
0,0%
10,0%
20,0%
30,0%
40,0%
50,0%
60,0%
70,0%
Male Female
SampleCensus
2/3/2012 VAA Research – State of the Art and Perspectives
6
Representative sample? Age and Region
• Age groups 50-64 and 65+ are underrepresented – 14.1% and 1.9% instead of 21.7% and 21%
• Age group 25-49 is overrepresented– 71.4% instead of 46%
• Athens and Thessaloniki are overrepresented
2/3/2012 VAA Research – State of the Art and Perspectives
7
Representative education level?
0,0%
10,0%
20,0%
30,0%
40,0%
50,0%
60,0%
Primary Lowersecondary
Highersecondary
Tertiary Postgraduatestudies
SampleCensus
2/3/2012 VAA Research – State of the Art and Perspectives
8
Factors of vote choice
17.6%3.9%Other Candidate
39.4%39.1%Candidate backed by the
party
Other Candidate
Candidate with maximum proximity coefficient
2/3/2012 VAA Research – State of the Art and Perspectives
9
Impact on vote choice• Depends on whether the VAA user had chosen
a candidate before using the application.• Undecided > VAA can help them learn the
positions of candidates and choose the most suitable to represent them.
• Decided > two possible outcomes: – i) identical voting recommendation > strengthen the
user's intent to vote for the preselected candidate – ii) VAA does not propose the same candidate > the
VAA recommendation could undermine user's initial selection, and if the influence is strong enough, it can lead to change of voter's position.
2/3/2012 VAA Research – State of the Art and Perspectives
10
Undecided voters (38,5%)
100%467100%1902Total
94.9%44368.9%1311Have not selected a candidate after
5.1%2431.1%591Have selected a candidate after helpmevote
%N%N
Intention to abstain before
Total sample
2/3/2012 VAA Research – State of the Art and Perspectives
11
VAA proposed the same candidate (%) by confidence level of initial choice
0%
10%
20%
30%
40%
50%
60%
Not at all A little Quite Very Completely
Confidence level about the pre-selected candidate
VAA
pro
pose
d th
e sa
me
cand
idat
e (%
)
2/3/2012 VAA Research – State of the Art and Perspectives
12
Confidence level about the candidate before and after using VAA
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
50%
Not at all A little Quite Very Completely
Confidence level about the selected candidate
Use
rs in
eac
h ca
tego
ry (%
)
Before VAA After VAA
2/3/2012 VAA Research – State of the Art and Perspectives
13
Different Candidate: level of affection by confidence level
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Quite or less Very Completely
Confidence level about the pre-selected candidate
Not at all A little Quite or more
2/3/2012 VAA Research – State of the Art and Perspectives
14
Intention to change vote choice and follow VAA by conf. level
0%
5%
10%
15%
20%
25%
30%
35%
Not at all A little Quite Very Completely
Confidence level about the pre-selected candidate
Cha
nge
min
d an
d fo
llow
VA
A (%
)
2/3/2012 VAA Research – State of the Art and Perspectives
15
Conclusions (1/2)• The sample is not representative of the electorate.
– We cannot use the data collected for the estimation of the electoral outcome
• 4 out of 10 voters base their vote choice mainly on their party identification– They have selected the candidate who is supported by their
party, although he/she is not the candidate with whom they have the maximum SDC.
• 4 out of 100 voters have selected the candidate with whom they have the maximum SDC although he/she was not supported by their party “pure issue voters?”
2/3/2012 VAA Research – State of the Art and Perspectives
16
Conclusions (2/2)
• VAAs can help undecided voters to choose candidates
• May even influence decided voters by: – proposing the same candidate and affirming
their initial choice or– undermining their confidence in their initial
selection by proposing a different candidate.