Proxy Politics, Economic Protest, or Traditionalist Backlash: Croatia’s Referendum on the Constitutional Definition of Marriage

Abstract

This article sheds light on the popular sources of opposition to the extension of gay rights in Eastern Europe by analysing the results of the 2013 Croatian referendum on the constitutional definition of marriage—the first referendum of this kind in Europe. Contrary to popular interpretations, our aggregate-level analysis reveals that the referendum results primarily reflected the pattern of support for the two principal electoral blocs, rather than communities’ traditionalist characteristics or grievances stemming from economic adversity. The article thereby stresses the importance of embedding the issue of contention over gay rights in Eastern Europe into the context of conventional political competition.

This work was supported by the Leverhulme Trust and the Isaac Newton Trust (ECF-2012-399\7). The authors express their gratitude to Irena Kravos (Croatian Electoral Commission), Ivanka Purić (Croatian Bureau of Statistics), Mirna Valinger (Croatian Tax Administration), and in particular to Dr Maruška Vizek (the Institute of Economics, Zagreb) for invaluable help with data collection. They also thank Tomislav Kaniški of The Miroslav Krleža Institute of Lexicography for creating the maps in Figure 1.

Notes

¹ Three referenda close in their subject matter to Croatia’s took place in June 2005 in Switzerland, June 2011 in Liechtenstein, and March 2012 in Slovenia. The Switzerland and Liechtenstein referenda were concerning registered partnership laws. They were both approved: by 58% of the voters in Switzerland and 68.7% of the voters in Liechtenstein. The Slovenian referendum was on the new Family Law that was to give registered same-sex partnerships all rights of married heterosexual couples except adoption. The law was rejected, with 54.5% voting against on a 30.3% turnout. Ireland was the second European country to hold a referendum on the constitutional definition of marriage on 22 May 2015. The referendum question defined marriage as a union between two persons regardless of sex. The measure passed with 62% of the votes.

² In the immediate run-up to the referendum, the Jewish community withdrew its support for the campaign of ‘In the Name of the Family’ (Kovačević Barišić 2013).

³ The coalition was named after the Kukuriku Hotel near Rijeka where the constituent parties (Social Democratic Party (SDP), Croatian People’s Party (Hrvatska narodna stranka—HNS), Istrian Democratic Assembly (Istarski demokratski sabor—IDS), and the Croatian Pensioners’ Party (Hrvatska stranka umirovljenika—HSU)) signed their coalition agreement.

⁴ The final confirmed number of valid signatures was 683,948 (Hrvatski sabor 2013).

⁵ ‘Referendum izaziva nelagodu, glasujte protiv’, Tportal.hr, 30 November 2013, available at: http://www.tportal.hr/vijesti/hrvatska/301853/Referendum-izaziva-nelagodu-glasujte-protiv.html, accessed 13 May 2014.

⁶ ‘Opačić: Iza referenduma i nereda u Vukovaru stoji HDZ!’, Dnevnik.hr, 16 November 2013, available at: http://dnevnik.hr/vijesti/hrvatska/milanka-opacic-iza-referenduma-i-nereda-u-vukovaru-stoji-hdz---311460.html, accessed 13 May 2014.

⁷ The proposed referendum question was proclaimed unconstitutional by the Constitutional Court in August 2014.

⁸ ‘Croatians Vote Against Gay Marriage’, Economist, 5 December 2013, available at: http://www.economist.com/blogs/easternapproaches/2013/12/croatia, accessed 13 May 2014.

⁹ Of the post-communist East European states, only the Czech Republic, Estonia, Hungary, and Slovenia—in addition to Croatia—have provisions for same-sex partnerships; their legislation is, however, more limited than the Croatian Law on Same-Sex Life Partnerships.

¹⁰ See Gaines and Garand (2010) for conflicting individual-level evidence.

¹¹ Looking at the referendum results in this more encompassing way raised the issue of Croatia’s problematic voter rolls, which have been in administrative disarray for years. Considering the proximity of the 2011 census to the referendum, and the fact that the census definition of adult residents was virtually identical to the legal definition of a citizen with the right to vote, we used the census figures for adult residents of a municipality as a superior measure of the number of voters than the electoral register figures. Furthermore, in our calculations we considered invalid ballots (0.6% of those who voted) as abstentions.

¹² The variable ‘% Married’ measures the proportion of families in a given municipality with couples in a heterosexual marriage; ‘% Divorced’ measures the proportion of the adult population which is divorced; ‘% Workforce female’ represents the proportion of the municipality’s workforce which is female; and finally ‘% Non-believers’ represents the proportion of the municipality’s population not belonging to any organised religion. Considering Hypothesis 3, therefore, this would mean that we expect ‘% Married’ to have a positive relationship with ‘Marriage Yes’ and negative with ‘Marriage No’, whereas the opposite should hold for ‘% Divorced’, ‘% Workforce female’, and ‘% Non-believers’.

¹³ ‘Years of education’; ‘Average age’; ‘Log settlement size’; ‘% Local born’; ‘% Croats’, and ‘War disabled per ‘000’. ‘Average age’ represents the average years of age of the municipality’s population, whereas ‘Years of education’ represents the average years of education of the municipality’s population older than 15 years of age—both based on the 2011 census.

¹⁴ ‘Log settlement size’ in our case actually represents the log of the weighted average settlement size in a given municipality:

¹⁵ For further discussion, see Angrist and Pischke (2009, pp. 103–6).

¹⁶ Defined as: , where is the sum of squares of the effect of interest, while is the total sum of squares for all effects, interactions, and errors in the ANOVA. The formula implies we needed to use the standard OLS regression to calculate this effect. We felt confident doing that because our OLS results were nearly identical to the results of our beta maximum likelihood estimation model.

¹⁷ , where is the error variance attributed to the effect of interest.