Vote on removal of EU justice opt-out – can we predict the result?
With the exam project in Social Data Science we have chosen to focus on the vote on Danish EU justice opt-out. Previously, exit polls and surveys has tried to predict outcomes of votes similar to this and other votes in Danish politics. This analysis is new in to aspects:
It is the first time that there is a vote on the removal of an opt-out
Data used for the model is taken from the social media
Therefore we want to explore the following:
_Can we predict the outcome of the Danish referendum in December based on Facebook data? If time allows: How do these results differ from Twitter-data and Google Trends?_
Social media as a source
As previously mentioned we wish to use data from the social media as input in our model to predict the outcome of the vote.
We wish to use the social media to get an unbiased look into the Danes attitudes towards the subject and use this to predict the outcome of the vote the 3. of December 2015.
The choice of Facebook as a source, is to be understood in the light that a large and broad part of the Danish population uses Facebook. According to DR’s media development report from 2015, 3,5 mio. people used Facebook monthly in 2014 (Danmarks Radio, 2015). The same report shows that 73% uses social media monthly and 59% uses the social media daily . The choice of data from the social media thus gives us the opportunity to get a broad base of the Danes attitudes towards the subject. Afterwards our results will be compared to surveys’ prediction of the election result, the final result and, if time allows, the predictions that would have been made with data from Twitter and Google Trends.
On a more theoretically note we hope that these data will give us an advantage by showing voters’ true and revealed preferences compared to surveys where people might feel pressured to lie or not answer truthfully.
Prediction
The model we wish to use for prediction is from a Greek prediction experiment in which the national vote was predicted (Askitas, 2015). The vote was held on 5. of July and was also a YES/NO vote. As in the Greek example, we want to have precise and time-determined data before the vote. Our model will be built on data from the month up to the election and the model is going to be a ratio between the expected no and yes votes that we estimate from the collected social media data.
Project Description – Social Data Science
Vote on removal of EU justice opt-out – can we predict the result? With the exam project in Social Data Science we have chosen to focus on the vote on Danish EU justice opt-out. Previously, exit polls and surveys has tried to predict outcomes of votes similar to this and other votes in Danish politics. This analysis is new in to aspects:
Therefore we want to explore the following:
_Can we predict the outcome of the Danish referendum in December based on Facebook data? If time allows: How do these results differ from Twitter-data and Google Trends?_
Social media as a source As previously mentioned we wish to use data from the social media as input in our model to predict the outcome of the vote.
We wish to use the social media to get an unbiased look into the Danes attitudes towards the subject and use this to predict the outcome of the vote the 3. of December 2015.
The choice of Facebook as a source, is to be understood in the light that a large and broad part of the Danish population uses Facebook. According to DR’s media development report from 2015, 3,5 mio. people used Facebook monthly in 2014 (Danmarks Radio, 2015). The same report shows that 73% uses social media monthly and 59% uses the social media daily . The choice of data from the social media thus gives us the opportunity to get a broad base of the Danes attitudes towards the subject. Afterwards our results will be compared to surveys’ prediction of the election result, the final result and, if time allows, the predictions that would have been made with data from Twitter and Google Trends.
On a more theoretically note we hope that these data will give us an advantage by showing voters’ true and revealed preferences compared to surveys where people might feel pressured to lie or not answer truthfully.
Prediction The model we wish to use for prediction is from a Greek prediction experiment in which the national vote was predicted (Askitas, 2015). The vote was held on 5. of July and was also a YES/NO vote. As in the Greek example, we want to have precise and time-determined data before the vote. Our model will be built on data from the month up to the election and the model is going to be a ratio between the expected no and yes votes that we estimate from the collected social media data.
Tentative literature