Voters can be influenced by voter advice websites, but they do not follow the guidance blindly

Voter advice websites, where voters are matched with parties that share similar views to their own, have started to appear in the UK after becoming popular in the Netherlands and other countries. In this post Matt Wall, André Krouwel and Thomas Vitiello discuss a new site launched for the European Parliament elections and consider how influential these services can be on the choices voters make at the ballot box. 


Deus ex machina… is this the voter of the future, as imagined by Isaac Asimov? Credit: Honda News, CC BY-NC-ND 2.0

Like all complex and contemporaneous events, the impact of the internet on political life is difficult for academics to characterise, and there is a rich and ongoing contestation about whether the communication technology enabled by the internet will improve, diminish or even affect democratic systems.

Our research, recently published in Party Politics, gives some insights into the effects one of the most widely-consumed examples of ‘digital democratic politics’ – the voter advice website (often referred to by scholars as ‘Voting Advice Applications’ or VAAs). In our study, we focused on whether users base their vote choice on the advice that they received when using a VAA site.

The premise behind these sites is a simple one – the site user is asked a number of political opinion questions and then provided with some sort of comparison of their opinions and those of the major political parties competing in a given election.

While VAAs have not yet truly taken off in the UK, they are now a regular feature of political life in democracies like the Netherlands (where such sites originated), Finland, Germany, Switzerland, Canada and Australia, where they attract millions of hits during campaign time.

Interested readers can check out a ‘live’ UK site for the upcoming 2014 European Parliament elections at:

We worked with data provided by a popular Dutch website, and focused on the 2010 Dutch general election. We reasoned that this represented a ‘most likely’ case for our hypothesis that online vote advice could influence vote choice. This is because of the high levels of partisan fragmentation and dealignment that charactertise cotemporary Dutch politics. Furthermore, VAA sites have a long history in the Netherlands, stretching back to 1997, and are more likely to be seen as a credible source of political advice.

We argued that, in order for a VAA site to be considered politically influential, it would have to push its users towards voting for the parties that it had recommended.

Our findings were nuanced – we noted a significant ‘effect’, but it was conditional. Voters were considerably more likely to vote for parties that had been ‘recommended’ to them by the website. However, this effect was only discernible when the website recommended a party that a voter was already considering strongly.

We demonstrate this effect in detail in the paper, and through the table and figure extracts here, but the take away points from our analysis were as follows.

Firstly, despite some caveats about the non-representative sampling method that we used, there is strong evidence that a causal ‘vote advice’ effect exists. This means that scholars need to think about issues such as impartiality and transparency when creating VAA sites (see this paper for detail about implementing best practice in this regard).

Secondly, it tells us something interesting about the conditional extent to which human voters take on board computer-generated voting advice.

The type of human-computer interaction that VAA websites provide was, literally, the stuff of science fiction only decades ago. In his 1955 short story The Franchise, Isaac Asimov imagined how American elections might proceed in 2008. The story centres on the interaction that takes place between Norman Muller and the super-computer ‘Multivac’. Muller must shoulder the heavy burden of being the single citizen who is deemed by Multivac to be ‘representative’ of the American population. Based solely on the opinions that it elicits from Muller, Multivac provides the results of the national election.

Multivac never directly asks Muller for his opinion on the candidates or parties running. Instead, the computer imputes logical political preferences from Muller’s opinions. Thus a theme of the story is the extent to which humans may someday be prepared to abdicate political autonomy to computers.

Amazingly, we are currently living in a world where this theme is relevant, and, indeed, researchable. According the research presented here it appears that, while citizens are influenced by the advice that they receive online, there is little evidence that many site users follow this advice blindly when it runs against their existing partisan predispositions. We may be much closer to the age of Multivac now than when Asimov penned The Franchise, but we have not (yet) arrived at a point where citizens let the computers make their decisions for them.

Note: This post represents the views of the authors, and does not give the position of LSE or Democratic Audit. Please read our comments policy before responding. Shortlink for this post:

Dr Matt Wall is a lecturer in politics at Swansea University. His research interests include Vote Advice Application (VAA) websites; online politics; electoral campaigns; and electoral system effects. Dr Wall has published research articles in journals including Electoral Studies, Party Politics, and The British Journal of Political Science.

Dr André Krouwel is an Associate Professor in Political Science at the Free University, Amsterdam and is also the academic director of Kieskompas, a company that creates VAA sites.

Thomas Vitiello is a PhD candidate at Sciences Po Paris and at Sabanci University Istanbul. His dissertation focuses on online VAAs and political campaigns. He was also involved as academic coordinator in the development of VAAs for the french presidential and european elections: La Boussole présidentielle in 2012 and La Boussole européenne in 2014.

Similar Posts