We are delighted to welcome the following guest post from David Fortunato (UC-Merced) on how voters in the Netherlands form expectations about who will govern in a complex multiparty system.
*********************************************************
In multiparty parliamentary systems, where coalition governance is the norm, the election is just the first step in deciding who will govern. After the votes are counted and parliamentary seats are awarded, two or more parties must come together to form a legislative majority and an executive cabinet that will run the country. Political scientists have argued that if voters in these countries want to maximize the impact of their ballot, they need to have some understanding of how these coalition bargaining episodes may play out. But political scientists know very little about how voters may form these expectations, or if they do at all.
As part of an ongoing series of projects investigating how these voters think about coalition formation and coalition policy-making, Randy Stevenson and I commissioned a pre-electoral survey in the Netherlands in the days leading up to last week’s election. We asked voters which party they thought would win the Prime Ministry and which parties, if any, they believed would join the PM in coalition. Looking at the voters’ predictions of which coalition will form can tell us if they are thinking about coalition formation in a systematic way (which is interesting for people concerned with predicting voting behavior) and it can also give us an idea of how adept they are at predicting political outcomes.
First the distribution of prime minister predictions. There seems to be general consensus that the PM will come from either the VVD (liberals) or the PVDA (labor). Given that the polls leading up to the election had these two parties neck and neck and clearly ahead of the third party, this makes sense. Note that voters overwhelming chose the party they believed would win a plurality as the prime minister. This is something we typically assume voters would do, but, to our knowledge, have never tested.
Looking at which coalitions voters thought would form, given their choice of prime minister, we were surprised at the variation in predicted coalitions. Many of us would predict that voters would key in only a handful of likely coalitions — clearly this is not the case. We could speculate on where this variation comes from. It may be that most Dutch voters simply don’t think about these things and are choosing at random. It may be that the variation in the mental models voters use to predict cabinet formation is quite large. It may also be that voters use very few mental models of formation, but that there is tremendous variation in the inputs that they are feeding into their models (this idea is foundational to our research).
To get a different look at the data, we plotted the dyad frequencies in the predicted coalitions — this will show us if voters are using some set of reliable anchor points in building their predicted coalitions. We calculated the proportion of voters who believed that the coalition would include, for example, the CDA and PVDA, regardless if they chose just those two parties, or those two with one or several others. We then split the data according to political knowledge and found something interesting: high knowledge voters keyed in on a handful of party pairs when assembling their predicted coalitions, whereas less knowledgeable voters had more diffuse expectations. For example, 68% high knowledge voters included the VVD and PVDA in their predicted coalition, but only 41% of less knowledgeable voters did the same.
The plurality of predictions from both groups involve a coalition of the VVD and PVDA with one or more bridge parties between them. However, more knowledgeable voters largely converged on a smaller set of alternatives in more systematic fashion. Both subgroups, however, have the same top 4 predictions: disregarding status (prime minister v. junior partner) and in order, 1) VVD, PVDA, CDA, D66; 2) VVD, PVDA, CDA; 3) VVD, PVDA, D66; 4) VVD, PVDA. Hopefully we will see how accurate these expectations are sooner rather than later.