Scottish Referendum behind Big Data: An Innovative Information-Demand-Based Analysis with Google Trends
This article is part of theGIST Science for Society 2014 Article Competition. The top three articles will present their work at the conference. Vote for your favourite article here. Voting closes on Friday 31st October 2014.
It was hard to analyse the Scottish referendum featured with many twists and turns. Through an unique approach with big data from Google, besides a regression on final referendum results, our research answers the following questions: Did Scots get enough clear information on the referendum? Was there ‘Yes’ momentum after the rejection of a currency union this February? What was the role of emotion and rationality in Scots’ voting decision process?
2. The Referendum-Related Issues
Many Scots thought they got no enough information before the referendum. It might not be the case as referendum-related information was flooding through various media. However, surrounded by conflicting news reports and distinctive polling results, Scottish voters actually were lack of clear information for their decision.
Undoubtedly, the most debated referendum issue was currency. The polling results from several companies did show some gains on the ‘Yes’ side after George Osborne rejected the idea of a currency union this February. Some reports claimed that the ‘Yes’ campaign gained momentum from the currency debates. However, the claim remains to be verified.
For many Scots, their decision to vote ‘Yes’ or ‘No’ depended on both rational factors related to their long-term interests, and emotional ones influenced by the on-going events in Scotland and the rest of UK. It was difficult to distinguish between emotion and rationality, and their effects on the decisions of the Scottish voters.
3 An Innovative Approach Based on Big Data Set from Google
Traditional political researches are based on polling results, news reports, etc., which are essentially information supplied to voters. Instead, our approach is based on search volume of the Google Trends, a big data set measuring the active information demand of the Scottish voters.
Google Trends is a service provided by Google to allow people get access to the relative online search volume for any keyword within any region of the world. The Google Trends data, presented in a [0, 100] interval, is the weekly index analysing the percentage of the number of searches that have been conducted for a particular term relative to the total number of searches conducted over time. The larger the index is, the higher the information demand and searches are.
In order to measure the information demand of potential ‘Yes’ voters, we use the Google search volume within Scotland on the key word ‘Alex Almond’ because the key word was the most popular choice of referendum-related searches and highly correlated with the results of previous Scottish Parliament elections. For robustness check, we adopt search volume of the key word ‘SNP’ within Scotland as the control variable of our model.
With the available information demand data, we set testing criteria before conducting further empirical analysis.
(1) Testing Criteria of Clear Information
We assume that if potential ‘Yes’ voters found clear information from Google, the search volume of ‘Alex Almond’ should have significant and positive effects on the polling results of ‘Yes’ votes.
(2) Testing Criteria on Information-Related ‘Yes’ Momentum
If the currency debates created ‘Yes’ momentum, potential ‘Yes’ votes should increase and the increasing speed should sustain.
(3) Testing Criteria on Emotion and Rationality
To distinctively measure the votes out of emotion and rationality, we propose that the emotion-related effects last at most one week and the rationality-related ones exist after four weeks.
4. Empirical Findings
Our empirical analysis covers the time period between 25 August 2013 and 16 August 2014, when the polls from most mainstream pollsters were available. The variable on potential ‘Yes’ voters, is constructed by the average opinion polling results from six polling companies, i.e., ICM, Dipso Maori, Panel base, Curvation, TNS BMRB and You Gov. Other variables are constructed from the Google Trends data within Scotland on the keyword ‘Alex Almond’ and ‘SNP’.
4.1 Increase in ‘Yes’ votes, Less Clear Information and No ‘Yes’ Momentum
We find that 4.3% swing voters became ‘Yes’ voters four weeks after the rejection of the sterling zone. However, the information effects became smaller and insignificant after 15th March 2014, which means there was no enough clear information nor ‘Yes’ momentum until 16th August 2014.
4.2 Emotion-Driven ‘Yes’ Votes after the Rejection of Currency Union
This section discusses whether emotion or rationality mattered in the decision-making process of the Scottish voters. We find that between 16th March and 16th August the effects on the potential ‘Yes’ voters disappeared within one week, which suggests the information-related ‘Yes’ votes were driven by short-term emotion.
4.3 Regression of Referendum Results
We run a regression on the referendum results based on the available information until 16th August 2014 and find that the ‘Yes’ side would get 42%-44% votes. If unplanned events were considered, the votes for independence would be in the area of 38%-47%. In either case, our regression showed that Scotland would stay in UK.
This paper shows a distinctive picture of the Scottish referendum from the ones provided by the traditional approaches. We demonstrate that currency was an important issue related to a 4.3% increase in ‘Yes’ votes. However, between 16th March and 16th August 2014, there was no enough clear information for potential ‘Yes’ voters who casted their votes driven by short-term emotion. Furthermore, there was no significant information-related ‘Yes’ momentum and the ‘No’ side indeed prevailed.