The Contingency of Voter Learning: How Election Debates Influenced Voters’ Ability and Accuracy to Position Parties in the 2010 Dutch Election Campaign

Abstract

Election campaigns are expected to inform voters about parties’ issue positions, thereby increasing voters’ ability to influence future policy and thus enhancing the practice of democratic government. We argue that campaign learning is not only contingent on voters’ characteristics and different sources of information, but also on how parties communicate their issue positions in election debates. We combine a two-wave panel survey with content analysis data of three televised election debates. In cross-classified multilevel auto-regression models we examine the influence of these debates in the 2010 Dutch parliamentary election campaign on voters’ knowledge of the positions of eight parties on three issues. The Dutch multiparty system allows us to separate voters’ ability to position parties from their accuracy in ordering these parties. We reach three main conclusions. First, this study shows that voters become more able and accurate during the campaign. However, these campaign learning effects erode after the elections. Second, whereas voters’ attention to campaigns consistently contributes to their ability to position parties, its effect on accuracy is somewhat less consistent. Third, televised election debates contribute to what voters learn. Parties that advocate their issue positions in the debates stimulate debate viewers’ ability to position these parties on these issues. In the face of the complexity of campaigns and debates in multiparty systems, campaigns are more likely to boost voters’ subjective ability to position parties than their accuracy.

Keywords:

• campaign effects
• issue learning
• policy positions
• election debates

Notes

1. Compared to the United States, Dutch parties have a large influence on the topics to be discussed during the debates, and largely decide among themselves who partner up as rivals in one-on-one single-issue opening debates.

2. However, Arceneaux (2006) shows that learning effects about “the fundamentals” (economy and ideology) are bigger in party-list electoral countries because voters focus more on parties and less on candidates.

3. The number of viewers for the 2010 election debates ranged between 1.3 (RTL) and 2.7 million (NOS), reaching a market share of almost 40% (see Appendix B). This contrasts with the relatively minor role of televised election ads in the Netherlands (Walter & Vliegenthart, 2010).

4. Households that could not otherwise participate are provided with a computer and Internet connection.

5. Support in the second wave differs marginally from the actual results, except for the support for the Freedom Party (PVV; that was similarly underestimated in opinion polls during the 2010 campaign): Christian Democrats (CDA; –0.8ppt in LISS), Labour Party (PvdA; +0.9ppt), Liberal Conservatives (VVD; –1.1ppt), Democrats (D66; +0.1ppt), GreenLeft (GL; +1.0ppt), PVV (–4.1ppt), Socialist Party (SP; +0.9ppt), ChristianUnion (CU; +0.2ppt), Orthodox Christians (SGP; –0.3ppt), Party for the Animals (PvdD; –0.2ppt).

6. CDA, PvdA, VVD, D66, GL, PVV, SP, and CU. The two parties in parliament that were excluded did not take part in the main televised debates of this election campaign.

7. Evidently, the dummy variable Ability to Position is a crude measure of voters’ perceptual certainty of parties’ issue positions. Ideally, we would have measured issue-party positions with a certainty scale (cf. Koch, 2006). Nevertheless, this measure gives us a first indication on voters’ certainty in positioning parties: we assess within-person changes in positioning parties throughout the campaign.

8. Hansen and Pedersen (2014) also assess dyads of parties but compare all party dyads to the ”true” (i.e., mean) position in the sample. Their use of a single, overarching dimension allows this strong demand.

9. Ideally, we would have estimated a dyadic model to explain the accuracy of each ordering of two parties. However, there are too few parties to estimate such a complex model on our panel data set.

10. The growth curve model nests both waves within-persons, and includes a variable to model the learning effect (change from pre- to post-election wave). By making that variable conditional on media use and debate profile, we can assess both selection effects (if media usage and debate profile have an effect on ability and accuracy in the first wave) and socialization effects (if media usage and debate profile have a significantly stronger effect on the ability and accuracy in the second wave). Hence, the growth curve model focuses on differences rather than residuals. Downsides are that the growth curve model may be too conservative when using fewer than three waves (Singer & Willett, 2003), and that the model gets complex very quickly, due to the large number of cross-level interaction effects and resulting covariances.

11. On ability the growth curve model finds more robust evidence than the auto-regressive model for campaign learning effects; on accuracy the growth curve model finds less evidence for campaign learning effects. This may be due to ceiling effects and bottom effects, respectively.

12. A total of 1,431 respondents were able to position all relevant parties in both waves on mortgage interest tax reduction, 1,795 respondents on migrant integration, and 1,225 respondents on health care.

13. Consequently, we are not able to test Hypotheses 4 and 5 on accuracy directly.

14. The coding unit was each uninterrupted text unit of the candidates. The unit starts when the candidate begins to speak and ends when the candidate is interrupted. All the words within this unit were counted when the candidates spoke within this unit about one of the three issues. There were no units in which the candidate spoke about more than one of these three issues.

15. The post-election decline is unlikely to be caused by the main post-electoral political event, coalition negotiations. These negotiations were ongoing (lasting until October 14) and very closed procedures. Only the PVV explicitly anounced a willingness to change positions on pensioning.

16. The decline of ambivalent positions on the issue of mortgage interest tax deduction had been stimulated by the banking, housing, and government budget crises since 2008. Most notably, the two largest parties, and until 2010 coalition parties, PvdA and CDA, took up less ambivalent positions in the 2010 campaign. Whereas PvdA proposed the partial abolishment of mortgage interest tax deduction, CDA did the inverse by calling that nonnegotiable at the start of the campaign. Both changes created a clear divide and increased the salience of the issue during the campaign. These, in turn, are likely to stimulate accurate positioning of parties during the campaign.

17. Growth curve models find an effect across the board; that is more outspoken for voters who watched the debate.

Additional information

Notes on contributors

Tom W. G. van der Meer

Tom W. G. van der Meer is Associate Professor, Department of Political Science, the University of Amsterdam.

Annemarie Walter

Annemarie Walter is a Marie Curie Fellow, School of Politics and International Relations, University of Nottingham.

Peter Van Aelst

Peter Van Aelst is Associate Professor, Department of Political Science, University of Antwerp, and the Institute of Political Science of Leiden University.