Connecting Niche Party Vote Change in First- and Second-Order Elections

Abstract

Niche parties in European countries have struggled to win seats in national legislatures. Accounts of niche party development describe how attempts to win these seats often begin with second-order election campaigns for the European Parliament (EP) or a regional assembly. Strong second-order campaigns can signal that a party is locally competitive, which will help niche parties by reducing defections due to strategic voting in later first-order elections. In this paper, I argue that according to such accounts, improvements in second-order election results should be correlated with subsequent improvements in first-order election results in any given constituency. I also argue that the magnitude of this correlation can be compared across different types of second-order elections, to gauge how credible voters perceive these second-order signals of local viability to be. I find that only regional assembly election results, not EP election results, are consistently and statistically significantly correlated with national election results. This suggests that niche parties can only build their support through bottom-up rather than top-down means, and that EP election results cannot be used to predict how niche parties will perform at national elections.

Acknowledgements

The author gratefully acknowledges the very helpful comments of the editor and anonymous reviewers, as well as comments on earlier versions of the paper from Jens Olav Dahlgaard, Michael McDonald, Bonnie Meguid, Antti Pajala, Carlos Scartascini, Olga Shvetsova, and Josh Zingher. Any remaining errors are the responsibility of the author.

Disclosure Statement

No potential conflict of interest was reported by the author.

Notes

1. This is implemented by using the Comparative Manifestos Project (CMP) data (Klingemann et al. 2006) to find parties whose manifestos mentioned the economy at a rate at least one standard deviation below the mean (weighted by vote share) rate amongst parties competing in the previous election. Then, the party must emphasize some non-economic issue at a rate at least standard deviation above the rate at which that issue is mentioned in the manifestos of the other parties. More details on this measurement strategy can be found in Wagner (2012).

2. See Van der Eijk and Franklin (2004) for an important alternative thesis, however. They argue that EP elections are a “sleeping giant” and so in the future could increasingly become their own first-order sphere.

3. Several authors have highlighted how this process can work in practice. Many accounts of the electoral history of European Green Parties argue that the 1989 European elections were the start of a “green tide” (Burchill 2002; Meguid 2008; Müller-Rommel 1989; O'Neill 1997; Richardson and Rootes 1996).

4. 26,618 interviewees were contacted face-to-face in all 27 EU member states, in the autumn of 2008 (European Commission, Directorate-General for Communication 2009).

5. The same pattern emerges in similar surveys about whether citizens feel attached to their region, to their nation, or to Europe. In 2004, across the EU, 51% felt very attached to their region, and 56% felt very attached to their nation, but only 20% felt very attached to Europe. Similar patterns emerge in other years.

6. As detailed further on the author's website at https://sites.google.com/a/binghamton.edu/benfarrer/home, the data were collected from a variety of sources. The EED was the source for most of the European elections data and most of the National elections data, and then Regional elections data were added from country-specific sources. However, when these country-specific sources also had National or European elections data, those sources replaced the EED results. Country-specific sources were the statistics department of the Landtag in Austria, the Danish statistical handbook, the Statistics Finland website (http://www.stat.fi/tk/he/vaalit/index_en.html), the French Interior Ministry, the Ergebnisse fürherer Landtagswahlen publication of the German Federal Returning Officer (http://www.bundeswahlleiter.de/de/landtagswahlen/ergebnisse/downloads/ltw_erg_gesamt.pdf), the http://www.nlverkiezingen.com/index_en.html website for the Netherlands, and the ARGOS electoral archive website for Spain (http://www.argos.gva.es/ahe/).

7. Excluding these countries introduces the possibility of bias, so I examined national-level correlations in the excluded countries. In the countries where this was possible, there were no discernible differences between countries that are included in this paper and those which are included, in terms of the magnitude of the correlations between EP and national – and regional and national – elections.

8. This was the system for all European elections in the countries and years considered here, although EP elections have been conducted under different rules, for example in the UK.

9. Using single regional party lists, the “winning” list (either >50% at the first ballot or the majority at the second) got 25% of the seats, and the rest of the seats were allocated proportionally to any party with over 3% of the vote (Bess 2003; Bruno and Jerôme-Speziari 2000; Jérôme and Lewis-Beck 1999; Lewis-Beck 2000; Nay 2001).

10. Since the Bavarian election returns are highly distinct (for a long period Bavaria was the only two-vote system, and throughout the period in question was the only state with open-list voting), they are not included in the analysis.

11. The party ballot (zweitestimmer) from the German national elections was used rather than the district-specific ballot, and the first-round vote totals from the French elections were used.

12. The reference category here is France.

13. All other variables are held to their means, except second-order turnout which is set to a minimal value to simulate typical low-salience elections. When calculating marginal effects, standard errors are not clustered by country, but this does not significantly change the magnitude of these standard errors.

14. Empirical results are also robust to the inclusion of this index as a control variable.