Strategic voting under PR: Evidence from a survey experiment in Sweden

By Annika Fredén, Lund University

What is the story?

In this blog, we often talk about strategic voting. As a reminder, we define strategic voting as voting for a party that is not one’s preferred party in a willingness to affect the electoral outcome. Typically, strategic voting concerns voters who prefer a party that has little chances of winning an election. Instead of wasting their vote in voting for this party, they cast a vote for their second (or third) preferred option that has better chances of winning.

There is another version of strategic voting that occurs in proportional representation (PR) systems. Under PR, parties usually form coalitions. When a small party risks falling below the minimal score required to have a representation in parliament (the ‘electoral threshold’), the supporters of the senior coalition partner sometimes decide to cast their vote for this small party in order to allow it to reach the threshold for the collective sake of the coalition. This is called ‘threshold insurance voting’. In previous blog posts, we show that this type of practice exists in Sweden and Germany.

In the present post, I show that a crucial condition for threshold insurance voting to occur is that the coalition lines be crystal clear. To do so, I draw upon recent survey experiment conducted during the 2014 Swedish general election’s campaign.

Coalition politics in Sweden

The Swedish party system is divided into two blocs. On the right, four parties form a coalition called the ‘Alliance’. The Alliance governed Sweden from 2006 to the 2014 general election. The coalition lines are well defined and all parties ran a common platform during the campaign.

On the opposite side, the left-wing bloc is far less established. During the campaign, Stefan Löfven, the leader of the Social Democratic Party (the largest left-wing party, which has a long tradition of governing alone in minority governments), said he was open to discussion with other parties at both sides of the political spectrum. A link with the Green Party was quickly established. The two other left-wing parties, the Left Party (far left) and the Feminist Initiative, a new formation that had just won a seat at the last European election in May declared they would support a Social Democratic government. However, the parties did not manage to coordinate as much as the right-wing coalition. At the end of the day, the left-wing parties decided to run separate campaign platforms.
The incumbent right-wing coalition lost many seats during the 2014 general election. The Social Democratic Party and the Green Party formed a minority government, while the Feminist Initiative failed to reach the 4% electoral threshold and thus gained no seat in parliament.

A survey experiment

The week before the 2014 general election, the polls released in the media revealed that the Feminist Initiative and the Christian Democratic Party (a small member of the right-wing coalition), were very close to the 4% electoral threshold. There was thus a unique opportunity to test whether the coalition signal – very clear in the case of the right-wing coalition, quite ambiguous in the case of the left-wing coalition – has an impact on the level of threshold insurance voting.

I ran a survey experiment 6 days before Election Day, through the online panel survey maintained by the laboratory LORE (University of Gothenburg). I randomly assigned different poll information to the 5,000 respondents. The reported level of support for all the parties was as released in the most recent polls. I only changed the predicted score of the Feminist Initiative and Christian Democratic Party. A third of the respondents saw these parties below the 4% electoral threshold at 2.5%, at exactly 4%, or at 5.5%. Then I asked respondents how they would vote if the polls looked like the poll they were presented.

The results show that vote intentions in favour of the Christian Democratic Party are significantly higher when the party is reported at the electoral threshold level or below (10-11% compared to 4% when it is reported as being above the electoral threshold). On the opposite side, vote intentions for the Feminist Initiative are less affected by the polls information (the level varies between 7%-13%). It seems like the party gains vote intentions when it is reported above the 4% threshold. The complete results are shown in Table 1 below. It is worth noting that highly educated people are over-represented in the sample. The vote intentions for the Feminist Initiative are higher than in the overall population.

Polls information Feminist Initiative Christian Democratic Party
2.5% 7.5% 11.7%
4.0% 13.4% 10.0%
5.5% 9.9% 4.2%
Table 1: Vote shares of the Feminist Initiative and Christian Democratic Party according to poll information


Strategic voting is a common practice in PR systems, in Sweden as in other established democracies. In the 2014 Swedish general election, I ran a survey experiment to assess how coalition signal affects strategic voting for small parties. The results show that the new Feminist Initiative, that was not part of a well-established coalition, would gain much less from threshold insurance voting than the Christian Democratic Party, a member of a well-defined right-wing coalition.

This entry was posted in Blog. Bookmark the permalink. Both comments and trackbacks are currently closed.