Electives 2012: Google Trends as a Modern Poll System
I guess the Dutch population was quite surprised by the results of the 2012 elective campaigns. Although Emile Roemer from the socialist party (SP) and Mark Rutte from the liberal party (VVD) were the major players at the beginning of the campaign, the campaign ended up in a battle between Mark Rutte and Diederik Samsom from the Labour Party (PvdA) whereby Rutte won the elections. And what about Roemer? He just kept left behind.
Like in many other campaigns, polls seemed the biggest influencers on the results. So when september 12th approached, the public complained about the fact that the polls should for example lead to strategic voting by citizens. This strategic voting resulted in favor of Rutte and Samsom. As a result, the public asks for a prohibition of the polls in approach to election.
Although some countries already prohibit polls during elections, for me this looks like censorship in a sense that government excludes citizens from certain important information. Besides this, would this actually help anything? New media such as the internet could maybe easily serve as a substitute for the more traditional polls. In this blogpost, I’ll take a closer look at online data, which may tell us even more about voting behavior and possible results than the traditional polls such as Peil.nl owned by Maurice de Hond. The question I want to investigate shortly in this blogpost is: What could online data, captured from Google queries, tell us about the political climate?
To answer this question, I contacted Maurice de Hond to receive poll data gathered during the last campaign. De Hond is one of the mayor players as it comes to measuring public opinion. in addition, I needed data collected from one of Google’s services called Google Trends. This was necessary to make a proper comparison. For this reason I searched for two kind of queries: First on the query about the initial popular political parties (SP, PvdA and VVD). And secondly, I looked for the names of party leaders such as Roemer, Samsom and Rutte which I linked to their party’s. These entries are combined and displayed together in the first graph with Google queries (see the Infographic) whereby the number hundred represent the biggest search volume.
Looking at both graphs you can see some similarities, but there are also number of differences. Most striking are the similarities before the start of the campaign around the august 19th. Here you clearly see an equal start between the queries on Google and the poll data. One week after, around August 28th, we see some interesting changes. The VVD and the PvdA’s popularity increased in the polls, as well as it did in the Google queries. The VVD became more popular than the SP, and the gap between the SP and the PvdA was decreasing. This was the rise of the labour party during these elections. Some days later, when the PvdA became even more popular in the polls than the SP, the same PvdA-line crossed the SP-line in the popularity on the Google results. A bit before September 7th, a big shift took place. While the SP dropped down even further in the polls, the popularity of the VVD and the PvdA started to grew even more on Google. This seemed the start of the broadly discussed battle between these two political parties. Again, both graphs show a lot of similarities. Although the SP rose up in the Google results, the distance between the others stayed. This last increase in Google’s graph, is most probably caused by the upcoming elections.
Yet, there still exists some differences. Although the lines shown in the graphs, at least in relationship to each other, went pretty much in the same direction, they ended up differently. Here, in the Google graph, the PvdA appeared as a more popular party than the VVD. This while the VVD in the end won the elections as shown in the results of the second graph. The reliability of Google Trends as a poll system can be doubted in this case. At the same time, critiques on the reliability of the traditional polls exist as well. Considering this it becomes evident that Google Trends can become a substitute for traditional polls.
However there’re still some issues that needed to be solved before this claim can actually be made. Some issues aren’t fully covered yet. This analysis is firstly done quite isolated. What I didn’t mentioned in this analysis, is for example a possible relationship between polls and queries. Secondly, does this modern poll system really measure the political parties popularity? What if some bad or even negative news about a political leader appears? The amount of searches will increase, but this will tell us more about the attention rather than the popularity. A sentiment analysis needed to be done, before you actually can say about the queries. Thirty what can be said about the relationship between the results of the queries and the actual opinion of voters. It looks like these Google figures do not stand alone. These figures may be influenced by other media which are in their turn influenced by polls. Then, what if polls are really getting prohibited, this can possibly cause a completely different Google-graph, which may or may not be of any influence on the final results. Only when polls are prohibited, and new media data would be seen as a proper and trustable substitution of polls, this new media data can be seen as a displacement of the older traditional polls.
The primary question: What could online data, captured from Google queries, tell us about the political climate? It can be assumed that this question will stay unanswered for now. Though, these figures can tell us a lot, and yet they eventually shows us which party is the more popular one online. But before this question really can be answered, a deeper research, repeated and proven itself more than only one election is really needed before these queries turn themselves into a trustable and modern poll system.