Primaries at the municipal level: how, how many and why

Abstract

In this article, we analyse the mayoral open primaries held from 2004 to 2015. We empirically assess their functioning and their effectiveness, especially in terms of competitiveness and turnout. We also explore the dynamics of diffusion of this instrument of intra-party democracy over time, across parties and across geographical areas. Moreover, we provide an exploratory account of the electoral consequences of mayoral primaries in terms of three dimensions of analysis: the type of municipality in which primaries are held, the features of the primary election contest, and the characteristics of the subsequent mayoral elections. Our results show that local primaries are characterized by two main features: they are fairly institutionalized, and a contagion effect across parties and geographical areas is emerging. In addition, they are quite competitive and our data show that, contrary to popular belief, external and/or more ideologically extreme candidates are less likely to win.

Keywords:

• Primary elections
• local elections
• political participation
• party politics

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Note that Table 2 illustrates the number of municipalities where open mayoral primaries are held each year, not the total number of primaries per year. We adopted this type of measure of the over-time distribution of mayoral primaries in order to avoid distortions due to cases of ‘double’ primaries. Double or ‘bilateral’ primaries are cases in which two different parties or political organizations in the same municipality organize primaries to select their mayoral candidates for the same local elections. In the 12-year time span considered in this study, we observed 56 ‘double’ primaries in 28 municipalities. There have been no cases – at least up to now – of ‘triple’ primaries.

2. The number of municipalities has thus been included among the independent variables as a proxy. The most appropriate variable for completing the specification of the model would be a measure of the size of the municipalities per province, such as the average size in terms of the number of inhabitants. This is due to the fact that, in several provinces where the number of municipalities is high, their average size is also quite small. For instance, in several provinces of Piedmont or in many other mountain areas in Italy, the average number of inhabitants per municipality is very low while the number of municipalities is quite large. Given that municipalities with less than 1000 or 500 inhabitants are of course less likely to organize local primaries, using the number of municipalities per province in the model seems to tap only one aspect of the phenomenon of the territorial diffusion of mayoral primaries and may generate biased results. If for the moment we have decided not to add the variable (average size of municipalities per province) to the independent variables of the model, its inclusion in the multivariate analyses represents the next step of our research.

3. We obtain, thus, an associated p-value of: p < 0.01 (one-tailed t-test). The critical values of t for probabilities are taken from the final tables of Bohrnstedt and Knoke (1994).

4. Turnout scores over 100% are found when the number of selectors taking part in the mayoral primary exceeds the number of votes for the mayoral nominee chosen though primaries in the subsequent local election.

5. Clearly more data are needed fully to invalidate this hypothesis empirically, and in particular a comparison would be necessary between the mayoral election results in municipalities in which primaries were held to select candidates and in municipalities where other selection methods were used. For instance, we have shown that nominees selected through primaries have won 59% of mayoral elections, but nominees selected though other methods could have won either a higher or a lower proportion of the subsequent elections. Thus, without a comparison between municipalities with and without primaries it is not possible to provide a definitive answer. We plan to gather more data shortly on election results in municipalities where other selection methods were used in order to develop our analysis further.

Additional information

Notes on contributors

Giulia Sandri

Giulia Sandri is an associate professor at the European School of Political and Social Sciences of the Catholic University of Lille, Lille, France. She was previously a research fellow at the Christ Church and at the Department of Politics and International Relations of the University of Oxford, Oxford, UK. Her main research interests are party politics, intra-party democracy and political behaviour. Her work has been recently published in Politics and Policy, Acta Politica, Comparative European Politics, Religion, State and Society, Ethnopolitics, Italian Political Science Review and Regional and Federal Studies. She has also recently got published an edited book, co-edited with Fulvio Venturino and Antonella Seddone, on Party Primaries in Comparative Perspective, Farnham, Ashgate.

Fulvio Venturino

Fulvio Venturino is an associate professor at the University of Cagliari, Cagliari, Italy. He was a member of the Scientific Board of the Italian Society for Electoral Research (SISE) between 2005 and 2010. In 2011, he launched, ‘Candidate and Leader Selection’ (C&LS), a standing group of the Italian Political Science Association (SISP). He has authored several books and articles on the issue of intra-party democracy.