Voting on Europe, again and again: Stability and change in the Irish experience with EU referendums

Electoral Studies 38 (2015) 170e182

Contents lists available at ScienceDirect

Electoral Studies journal homepage: www.elsevier.com/locate/electstud

Voting on Europe, again and again: Stability and change in the Irish experience with EU referendums Michael Marsh Department of Political Science, Trinity College Dublin, Dublin 2, Ireland

a r t i c l e i n f o

a b s t r a c t

Article history: Available online 25 February 2015

This article is about comparative voting behaviour in referendums on the EU and explores variation within one country rather than variations across countries. This enables us to control for broad national context while allowing variations in the immediate referendum context, in terms of campaign intensity and incumbency. It analyses voting behaviour in the many referendums that have taken place in Ireland. The major part of the analysis deals with the five referendums since 2001, as this allows the use of the same measurement of EU support and the use of post referendum surveys. Most attention is paid to attitude to the EU, party support and satisfaction with the incumbent government, reflecting the main debates in the literature on the issues and party cues. The relative importance of each is said to depend on contextual factors such as campaign intensity and economic strength. We find both party cues and issues matter consistently, and suggestive evidence that incumbency matters to the effectiveness of cues given by the two main proEU parties but the major finding is that variations in the factors driving voting behaviour in different Irish polls on Europe are slight and barely significant. © 2015 Elsevier Ltd. All rights reserved.

Keywords: Referendums Voting European Union Second-order elections Issue voting

1. Introduction The decision to introduce direct elections to the European Parliament was justified initially as a way of addressing the 'democratic deficit' within the EU, or at least of ensuring a closer connection between voters and elites. Of course some were sceptical at the time, but in the more than thirty years since that decision was made, the experience of direct elections has not realised the ambitions people had for them and instead has tended to confirmed most of the views of the sceptics (Marsh and Mikhaylov, 2010). In this context some feel that Europe needs even more direct democracy. A recent paper argues that the experience of the Constitution Treaty, rejected in France and the Netherlands in referendums (though approved in Luxembourg and Spain) exemplified the advantages of the E-mail address: [email protected]. http://dx.doi.org/10.1016/j.electstud.2015.02.008 0261-3794/© 2015 Elsevier Ltd. All rights reserved.

referendum process and arguably its superiority relative to EP elections (Glencross and Trechsel, 2011). The wider literature on EU referendums has explored the nature of voters' decision making and considered how far decisions are made on the basis of what might be broadly termed the 'issues' and how far the domestic structure of party competition e separate from such issues e explains voting patterns. Even if there are variations in the weight of such factors across countries we might well expect referendums in a particular country to follow much the same pattern across time. This has been the broad thrust of much work on electoral behaviour in legislative elections. However, recent research has argued that the bases of referendum decision making is conditional on the broader context of the vote. This follows work on a wider group of referendums and also addresses the considerable volatility often seen in opinion during the course of campaigns. In general this means looking at the campaign itself, in terms of

M. Marsh / Electoral Studies 38 (2015) 170e182

intensity and the line-up of pro and anti forces, but also at the wider context, such as the state of the economy. The search for contextual variation e or its absence e of course requires a research design that goes beyond studies of single votes, but the need for survey data to explore voting behaviour limits the number of countries and referendums that can be included in any comparison. Important studies of the importance of context have all taken cases from different countries and so have concentrated on variation between countries rather than variation within any country. Yet a within country focus is valuable. Holding constant the broader national context allows a focus on changes between different votes and makes it a little easier to find comparable data. This paper restricts itself to just one country, Ireland, and examines aspects of the referendum experience there. There have been nine votes in all, including that on accession in 1972. (Referendums are necessitated by a Supreme Court ruling on the Single European Act: see Gallagher, 1988). Our primary interest is in exploring the variation across different referendums in terms of the weight of the different factors in the voting decision. In particular we are interested in the extent to which issues may matter more when campaigns are more intensive; and in the consequences of being in government or opposition for the ability (or willingness) of parties to persuade their supporters to follow their recommendations. The major contribution of this paper makes is to provide the most extensive study to date of variations over a single type of referendum within a single national context. It also allows us to examine the consequences of such successive votes on behaviour. It follows Svensson's (2002) study of several Danish referendums. This argued that issues were always the dominant factor but reservations have been expressed, both about the conclusions and the extent to which generations can be made from the Danish case precisely because of the number of votes on Europe (Franklin, 2002). Arguably too, as a country that has typically seen a significant anti EU party in European Parliament elections, the European issue is particularly salient in that country. 2. Issues, parties and context The positive report from Glencross and Trechsel on the European Constitution referendum experience stemmed from their findings that 'issues' were by far the strongest factor in the voting decision: 'voting on European integration without the distorting film of domestic party politics e was the primary dimension in these four votes' (Glencross and Trechsel, 2011: 768) in all four countries voting on the illfated Constitution. Another scholar with a positive view of referendums concluded similarly on the basis of his analysis of five Danish votes: 'ordinary citizens can develop beliefs and attitudes on political matters and bring them to bear on actual political decisions …. Consistent values may be developed on a salient issue and may become the basis for voting behaviour in one or more referendums' (Svensson 2002). This echoes findings in referendums in many countries on a wide range of matters, not simply EU treaties, as well as the broader study of electoral behaviour (The most comprehensive analysis is Hobolt, 2009). Of course, there is still the matter of what particular issues are relevant to an

171

assessment of the treaty. Glencross and Trechsel include assessments of the treaty itself, as well as more general views on the European Union and membership of the Union for the respondent's country. One might also consider some more specific issues related to any treaty, such as its implications for sovereignty in particular policy areas. Following McLaren (2006) some have explored the importance of perceptions of threats to national identity in EU policy generally and in provisions of treaties in particular (Lubbers, 2008). Even if 'Europe' matters most in term of its impact of the voting decision, it is worthwhile to dig a little more deeply, if possible, and see what elements of 'Europe' are critical. Campaigns around Irish votes on EU treaties have raised concerns about issues as various as abortion, neutrality, corporation tax, workers' rights and conscription (see e.g. Garry et al., 2005; Garry 2012; Hayward 2002; Sinnott 2002a, 2002b). Such issues are sometimes described as 'first order' in contrast to factors that stem from domestic political competition, the 'second order' factors (Franklin et al.,1994a; Franklin et al., 1994b; Franklin et al., 1995; see also Pierce et al., 1983). In its simplest version, this would see referendums as no more than an expression of support for the government of the time the vote is taken, although this is arguably a straw-man construction of the second order thesis (Franklin, 2002) which has found theoretical and empirical support elsewhere (e.g. Schneider and Weitsman, 1996; Ray, 2003; most extensively Hobolt, 2009). More seriously it can be argued that voters will take their cues from parties they support, particularly where they have no wellformed attitudes towards the subject of the referendum. However, some parties may give stronger cues than others, with opposition parties in favour of a treaty perhaps less inclined to encourage supporters to vote yes than they would do if they were in government. In addition, some parties may well be divided on Europe, which is typically not a foundational issue in most party systems, and so opt to stay on the sidelines. Petithomme outlines these dynamics and provides some support for them on the basis of EU referendum campaigns in eight countries. In particular he finds that mainstream opposition parties can be unreliable campaigners for the 'yes' side (Petithomme, 2011: 106). How well formed attitudes on the EU actually are seems likely to vary, but even where attitudes to the EU are relatively stable, the particular constructions placed on a treaty may vary and with it attitudes to the treaty. In the absence of good information about what is in, or not in the treaty, party cues can obviously be relevant even if a more general orientation to Europe is stable. Franklin (2002) suggests that where there are several votes on Europe in a country this may contribute to a greater stability in terms of attitudes to the EU and to that extent weaken the impact of second order factors, suggesting that the Danish case on which Svensson (2002) relies may be untypical. The Irish experience that is explored here, with nine votes to Denmark's five, may show whether this is the case. Certainly most studies on EU referendums do show a significant impact of these second order considerations, whether this is simply rejecting a proposal put by an unpopular government or following party cues. The primary focus of this paper is to explore the stability of referendum voting behaviour in the Irish case, but to the

172

M. Marsh / Electoral Studies 38 (2015) 170e182

extent that we find variation over time in what matters, there are several explanations available for any instability that we might find. Several studies of EU referendums have explored not simply what factors are important, but how and when they become important, highlighting the contingency inherent e but often not acknowledged e in many models. Just as 'the campaign' is accepted as a factor in elections, so it is in referendums (De Vreese and Semetko, 2004). Arguably, given the greater unfamiliarity of referendum proposals relative to party choice in elections, the campaign might be expected to be more important. Certainly, opinion change during a campaign in frequently striking, and decisions are arguably made later than they would be in a general election (LeDuc, 2003). What is most interesting here is the impact that the campaign will have on the weight of the factors described above. There are several possibilities suggested in the literature. Hobolt (2007; 2009) advanced the hypothesis that the intensity of the campaign is a key variable in determining the relative importance of attitudes towards the EU on the vote. This hypothesis is based on a much wider work on elections of all kinds as well as referendums and plebiscites. As Hobolt notes: 'political information enables individuals to link attitudes with specific policies and in turn act on the basis of these attitudes (Hobolt, 2007: 88; see also Zaller 1992). Measuring intensity through a content analysis of the print media, she found that across a sample of eleven referendums in three countries (Denmark, Ireland and Norway) there was a linear relationship between intensity and the impact of EU attitudes on vote choice (Hobolt, 2007: 98e100). A more extensive analysis, using a threefold measure of intensity and more cases confirmed that relationship (Hobolt, 2009: chapter 4). Moreover, at individual level, the impact of attitudes on vote was stronger amongst those more exposed to campaigns (2007: 102e104). Attitude to the EU attitudes here was measured on a summated scale comprising several indicators, not always the same ones for each country. The strong relationship shown owes much to a few high intensity campaigns; there is a weaker pattern across the Irish cases included. That said, Marsh (2007) found when comparing the two referendums on the Nice Treaty in Ireland that the impact of attitudes to EU integration was marginally stronger in the second vote e which followed a much more intense campaign, that generated substantially higher turnout. Even so, Garry et al. (2005), using a wider set of questions, found that while attitudes to enlargement were much more important, the impact of general support for integration was no greater in Nice 2 than in Nice 1. Sattler and Urpelainen (2012) have offered another explanation, arguing that economic conditions will influence the factors swaying referendum voting behaviour. Poor economic performance will amplify the effects of treaty knowledge, as better-informed citizens will value international integration as insurance. They also outline a factor that will reduce the importance to the vote of trust in government, arguing that where concessions have been made to a treaty to help a government win a yes vote, trust will be set aside because the concessions themselves will mitigate concerns. The latter assumes, of course, that the issue is trust in the incumbents rather than dislike of them,

or a partisan wish to give them a bloody nose. With respect to economic conditions making international integration more appealing, we would argue that this should apply regardless of information levels, even if it is more striking amongst those with more information. While campaigns may provide information that enables voters to link attitudes and politics, they may also sway which attitudes are most salient by their role in framing the debate, indicating e in the case of EU treaties e what are the most important features of the treaty in question (DeVreese and Semetko, 2004). LeDuc argues that 'a referendum on the EU Constitution need not necessarily be about 'Europe' … the way in which the issue is framed, together with the dynamics of the campaign is just as likely to determine what voters believe they are voting on' (LeDuc, 2005: 26). One campaign may highlight economic advantage, another increased immigration; in the Nice campaigns in Ireland, as has been said, abortion and neutrality were emphasised by some groups on the 'no' side as against 'enlargement' on the yes side. This leads to some simple general expectations about the variation in effects over time. H1. the factors underlying voting behaviour in referendums shows little change over time However, if we do find variation: H2a. the impact of attitudes to Europe will be higher in more intense campaigns H2b. the impact of satisfaction with the government will be lower in more intense campaigns H2c. the effectiveness of 'yes' parties to mobilise support will be greater when they are in government than when they are in opposition H2d. the impact of attitudes to Europe will be higher when the economy is weak than when it is strong H2e. the impact of government satisfaction will be weaker in 'second' referendums when real or apparent concessions have been made Most studies also take account of demographic factors, and many show some variations in support according to education, occupational class, gender and age. These may be expected to influence referendum voting through their influence on attitudes to the EU or through variable partisanship. The theoretical underpinnings of such relationships with attitudes are less clear, and so any reasons why there might be variation over time in the effects of such variables is also obscure. Attitudinal research has focused on 'hard', utilitarian (e.g. Gabel, 1998) and 'soft', identity, factors (McLaren, 2002; Hooghe and Marks, 2005). In the analysis that follows controls are used for demographic factors but the main focus of our analysis will be on the attitudinal and party related factors. 3. Data and analytic strategy Fig. 1 shows the varying patterns of support in each referendum, and the turnout at each. Support has varied, for a high of 83% in the Accession vote to a low of 46% in the

M. Marsh / Electoral Studies 38 (2015) 170e182

Fig. 1. Results of nine referendums on the EU in Ireland.

first vote on Nice, and 47% in the first vote on Lisbon. One might see a downward trend in the willingness of the Irish public to support further integration. Certainly the highest 'yes' vote is found in the first three referendums, but there is no trend since the 1998 vote on Amsterdam. Nice I coincided with a very low turnout, seen by some as critical to the 'no' vote (Sinnott, 2003), but turnout was higher in Lisbon I (which was also lost) than in the most recent referendum which passed quite comfortably. The Irish case has its own idiosyncrasies. In particular, the party system lacks the typical left-right character of most European party systems, dominated as it has been by two centre right parties, each of them pro European. Nor is the EU itself a salient issue in national party competition. The three largest parties in this period can all be characterised as pro European. Not surprisingly, voters see little difference between the parties (Marsh et al., 2008). The issue has really only been salient in the context of referendums. The economic crisis, and the role of the EU in that crisis, and in constraining policy responses to it, has been the subject of a lot more debate in the period since the 'bailout' in 2010. But even in the 2011 election, fought in the immediate aftermath of the 'bailout', attitudes to Europe did not seem to be much of a factor in voting behaviour (Cunningham and Marsh, 2011). In general, since 1992 opposition to the EU has been voiced by the smaller parties on the fringes of the system, and by civil society groups concerned primarily with the threats that the EU is seen to offer to Ireland's neutrality in foreign policy and its anti in abortion regime. Only in 2012, when the anti-EU Sinn Fe was the third placed party in opinion polls was this situation very different. That being so, it is arguably particularly important to see how effectively voters can bring to bear their attitudes to the EU on the voting decision, and also how government office can alter the effectiveness of the main parties in their campaigns for 'yes' votes. The most recent referendum, with a changed party system, and arguably a more salient EU for many people, offers a potential contrast with earlier votes. 3.1. Survey data Surveys were carried out before or after all of these votes, but in some cases although poll findings were published the data itself is not available. Opinion poll data for newspapers is usually based on fieldwork carried out

173

during the campaigns. Sometimes public opinion seemed to change significantly between the poll and the vote itself, so such polls are often less than ideal. One exception was an exit poll carried out for the national radio/television network RTE after the Amsterdam vote. This vote took place at the same time as another referendum that attracted much more attention, and the exit poll asked no questions on the Amsterdam Treaty beyond one on voting behaviour. The data are richer from Nice 1 onwards. There were more opinion polls and extensive post-referendum studies were carried out by IMS for the government or European Commission. There are also post-referendum academic studies commissioned from RED C Research for the last three votes. Most of the more detailed analysis here concerns just the last five votes, but for some variables we also extend our analysis back to 1992. Four sets of explanatory variables are used. Full details are given in Appendix 1 but a summary is given here. Firstly, we use attitudes to integration. The main item used here is the item: As regards the European Union in general, which of the following statements comes closest to your view? Ireland should do all it can to unite fully with the European Union OR Ireland should do all it can to protect its independence from the European Union In all cases responses are dichotomised into pro-closer integration/not pro-closer integration. The second set comprises support for parties and satisfaction with the government. The latter is sometimes a four-point scale, and sometimes a dichotomy. We dichotomised all instances of four point scales giving us a simple satisfied/ not satisfied distinction. Party support is usually vote intention, with one exception, where it is the party the respondent usually supports. Of course stronger measures of support would be preferable, since vote intention is a weak indication that someone would trust in the cues given by that party. That is so, but this is the only data available, and is still sufficient to give us some clear patterns. We work with just the four largest parties as of il (FF) Labour and Sinn Fe in 2011: Fine Gael (FG), Fianna Fa (SF), the first three the largest three at the time of all referendums. Other parties are grouped with all sorts of undecided voters. The last constitutes the reference category. While such a heterogeneous category is not ideal, our interest is more in differences between the four parties defined earlier as we wanted to keep all of these explicitly in each model.1 The third set of variables consists of demographics: age, gender and social class. Age is typically grouped and not always in the same way, so we have dichotomised that at (as close to) above or below 50 as is possible. Class is grouped into three: essentially manual workers, non-manual and farmers, who are plentiful enough to merit separate measurement. Finally, we make use of a variety of other variables, some of which can be used across most of the data sets. These

1 The partisan composition of 'others' varies but largest category is Independents in the 2001e12 period. However, the pro EU PDs were a significant element in the 1990s.

174

M. Marsh / Electoral Studies 38 (2015) 170e182

include subjective and objectives measures of knowledge, both of the Treaty at issue and the European Union, and items assessing views of real or imagined elements of the treaties, such as provisions on neutrality or abortion. Distributions of all the standard set of variables are provided in Appendix 2. The context for each referendum in terms of economic conditions and control of government is straightforward. The intensity of the campaign is less so, but will be measured here using the proxy variable of turnout (see Hobolt, 2009: 96-7). Amsterdam is a special case as turnout was relatively high, but for reasons probably unconnected to any campaign on the Treaty. Turnout since Maastricht has ranged only from 50 to 59% (excluding Nice I, which saw only 35% turn out to vote). Another measure of campaign intensity is the level of understanding voters claim to have of the proposal and its implications. Surveys have used different wording when assessing this, but comparable post-vote data (see Sinnott, 2002b; 2003) for Nice 1 and Nice 2, shows the percentage who say they were “vaguely aware” or “don't know” when asked about their understanding of the issues was 77% for Nice 1 and 42% for Nice 2 respectively. Post-vote surveys carried out in the weeks after voting for the Referendum Commission and detailed in individual reports found the percentages saying they understood the issues “not at all” or “not particularly well” was 57% after Lisbon 1, 27% after Lisbon 2 and 39% after the Fiscal Treaty. This suggests an ordering of campaign effectiveness going from least at Nice 1 through Lisbon 1, Nice 2, Fiscal Treaty to Lisbon 2. This differs from a turnout ranking only in reversing the order of the Fiscal Treaty and Nice 2. 3.2. Analytic strategy Our analytic strategy is a simple one. First of all, we run the same model across as many of the referendums as is possible, treating each one separately. This is done using logistic regression given the dichotomous dependent variable (voting yes or no). We can then look at the coefficients (shown as odds ratios) for each variable so as to see how much they vary across time, and if so in what manner. This is not unproblematic. While OLS coefficients can be compared across models, whether nested or separate, logistic regression coefficients either in their natural state or when transformed into odds ratios, are not strictly comparable. As a recent review of this problem explains, this is essentially because of bias from unobserved variables, regardless of whether or not such variables are related to the observed variables (Mood, 2010, 67). While there is no perfect solution there are several ways to mitigate the difficulty. Here we will compare the average mean effects of the observed variables over time. These are reasonably comparable (Mood, 2010, 78) and so allow us to explore how far the impact of issues as

2 The Average Mean Effect (AME) is not the same as commonly employed marginal effect (MFX). The latter is the impact of a factor, x, on the Y variable conditional on a particular set of characteristics, typically described by the mean values of the other variables. AME is the average effect of x on Y. see Mood 2010, 75. In essence, it treats those with and without the x characteristic as similar in every respect but that. Calculations on here were made by the Stata 11 margins command.

opposed to domestic loyalties vary from vote to vote.2 These initial models are simple ones. Arguably, if we could estimate a richer model, we might find more (or less) variation. Secondly, therefore we add in other variables as available to see if that alters the conclusions drawn from our initial analysis. Following this somewhat exploratory strategy, we will test the variability between the different votes more formally, but seeing how well one model fits all referendums, contrasting a full model on the pooled data with alternative models allowing effects of different variables to vary across referendums, and models which assume no such variation to assess how different are these several referendums in terms of voting behaviour. 4. Results and discussion Table 1 provides details of the estimates from separate logit estimations on data sets from the Maastricht Treaty onwards.3 The estimates are expressed here as odds ratios which are somewhat more intuitive. As is apparent, there are key variables missing in some of these. The EU integration measure is not available in the pre Nice surveys, and government satisfaction is missing in the post election survey by Lansdowne on Nice I. The pro-integration measure is always highly significant with those favouring more integration typically much likely to vote yes as those not in favour, being most pronounced in the second vote on Lisbon and lowest in the first vote on Nice (in both of the two surveys). This appears to be the strongest effect, particularly since we know that the distribution of opinion on this item is reasonably balanced (see Appendix 2). Satisfaction with government is also typically significant, with the second vote on Lisbon, when government satisfaction was particularly low e the only exception. The effect here though is typically much smaller, around one third of the size of the EU attitude effect. The Fiscal Treaty appears to show far the most pronounced satisfaction effect. As far as party cues are concerned, there are consistently significant effects for FF: only in the pre-election survey before the first vote in Nice was the effect not significant. The effect of FG support is less striking, but the range is bigger, reaching its maximum to 3.8 in the Fiscal Treaty referendum. This was the first referendum on the EU when FG had been in government and FF in opposition. The Labour effect is always much smaller and typically not significant. SF supporters, in contrast, tend to vote no to a significant degree. The effects for the demographic variables are of similar size to the others, except for gender, which was very significant only for the Amsterdam Treaty. (Government satisfaction and attitudes to integration are missing here.) In general non-manual workers and farmers are more likely to vote yes than manual workers, although the non-manual effect is significant in the full models only in the Fiscal Treaty, Lisbon 1 and Nice 2 votes. One final notable feature of Table 1 is the increasing size of McFadden's (adjusted) pseudo R2. This is

3 We have used data weighted to confirm to the actual distribution of 'yes' and 'non' votes each time although the results from unweighted data are in this case no different in any important respect.

M. Marsh / Electoral Studies 38 (2015) 170e182

175

Table 1 Model estimates for EU referendums 1992e2012: odds ratios and standard errors. Maastricht

Amsterdam

Nice1

Nice1

Nice2

Nice2

Lisbon 1

Lisbon 2

Fiscal Treaty

IMS 1.61*
0.41 1.4
0.47 1.16
0.44

Exit poll 2.05***
0.22 1.90***
0.25 1.41**
0.22

0.87
0.2 3.03***
0.85 1.76*
0.55 1.07
0.23 1.91***
0.43

1.34***
0.08 1.22**
0.11 1.78***
0.35 0.65***
0.06

IMS 2.76***
0.64 1.15
0.36 1.12
0.48 0.27*
0.21 0.92
0.19 1.2
0.29 1.2
0.32 0.76
0.15

3.59***
0.77 0.39*
0.21

MRBI 2.19***
0.52 1.85**
0.52 0.84
0.26 0.46**
0.17 1.77***
0.36 1.3
0.26 1.96**
0.62 0.72*
0.13 2.96***
0.63 7.56***
1.49 0.20***
0.09

IMS 2.24***
0.53 3.09***
1.01 0.73
0.26 0.29**
0.14 1.19
0.24 1.51**
0.31 2.01**
0.71 0.89
0.17 2.04***
0.44 5.35***
1.05 0.26***
0.12

RED C 2.34***
0.66 1.84**
0.58 0.78
0.3 0.30*
0.19 2.06***
0.49 2.04***
0.48 2.74*
1.44 1.01
0.23 2.50***
0.6 7.91***
2.03 0.02***
0.02

RED C 3.25***
1.21 1.49
0.46 1.5
0.44 0.26***
0.13 1.83**
0.45 1.33
0.31 2.78
2 0.9
0.23 1.76
0.64 11.71***
2.93 0.14***
0.09

RED C 2.34***
0.62 3.79***
1.11 1.3
0.4 0.38**
0.15 1.18***
0.05 1.73***
0.35 2.88**
1.4 1.09
0.22 5.55***
1.26 8.51***
1.8 0.11***
0.04

504 0.13

713 0.26

744 0.22

578 0.26

769 0.32

976 0.45

Constant

1.07
0.48

0.9
0.18

MRBI 1.39
0.31 1.02
0.24 0.99
0.28 0.84
0.37 1.02
0.18 1.09
0.2 0.97
0.26 1.08
0.18 1.73***
0.32 4.31***
0.73 0.22***
0.09

Observations McFadden's Pseudo R2

740 0.06

2457 0.04

727 0.11

FF FG Labour SF Over 50 Non manual Farmer Male Satisfied with government Pro integration

Notes: details of each survey in Appendix 1. * sig at .05, ** sig at .01, *** sig at .001.

0.45 in the most recent vote, having been tiny in earlier votes and even at Lisbon 1 only 0.26. We shall return to this last point, but first of all we need to look more closely at the variations between the coefficients over time. Do they show any of the expected patterns? To do this as explained above we calculate the average mean effect of each factor at each time. We can start with the effects of attitudes to integration and satisfaction with government. These are shown graphically in Fig. 2 taking just those surveys that include both measures. The graph shows the effects and standard errors, showing the range at the 95% level of probability. Several features of Fig. 2 are important. The first is that most of the differences between the effects of attitudes to

Fig. 2. First and second order impact on yes vote 2001e2012: average marginal effects and 95% confidence intervals.

integration in each referendum are not significant. Only the weakest effect (Fiscal Treaty) is (just) significantly smaller than the strongest (Lisbon 1). Moreover, there is no sign that higher effects for this measure come in the more salient, higher turnout campaigns. The second important feature is that the variations in the effects of government satisfaction are also not so great as to be significant. In most cases they are almost the same. The size of the government satisfaction effect is particularly large in the vote on the Fiscal Treaty but not (quite) significantly larger even that than insignificant effect on the Lisbon 2 vote e when government satisfaction was particularly low. . There is little support here for any expectations regarding the variations in either satisfaction with government or support for integration. The Fiscal Treaty vote saw a turnout close to average; subjective levels of information were similar to those it Lisbon 2 and Nice 2, and while this was not obviously a renegotiated treaty, nor was Nice 1 or Lisbon 1. Fig. 3 shows the variations in the effects of party support (relative to others and no party) again in graphic terms. Only at the time of the last vote was FF in opposition and FG in government. The similarities across referendums are more apparent than any differences. FF and FG are typically pro, and SF anti, while the Labour effect is never significant. FF support was usually the biggest party effect on 'yes' voting, bigger than FG support. This was reversed in the most recent referendum when FG was for the first time in government (for an EU treaty vote). FG supporters were slightly more inclined to vote 'yes' than FF supporters in Nice 2. FG support is only a significant factor in Nice 2 and

176

M. Marsh / Electoral Studies 38 (2015) 170e182

Fig. 3. Impact of party support on yes vote 2001e2012: average marginal effects and 95% confidence intervals.

the Fiscal Treaty; FF support is always significant e but differences across years are again not significant. The Nice 1 effects were estimated using a pre-election survey, the only one in this set. Adding in evidence from the 1992 and 1998 referendums e where either attitudes to integration, or government satisfaction and attitudes to integration are missing e shows FF supporters always more inclined to vote yes than those of FG. Available tabulations from surveys on the Single European Act also show FF voters more supportive (see Gallagher 1988). Hence, there is some support for the expectation that the main party of government would mobilise supporters more effectively when in government than it would in opposition, but the evidence is not overwhelming. Note too that Labour supporters tend to favour neither side; even when in government as they were in 2012. The party was also in government in 1992, at supporters were also not inclined to vote yes to a significant degree in the Maastricht vote either. Finally, The Nice 2 poll used here is the only one that uses 'party you usually support' rather than vote intention. Using an MRBI pre-vote poll instead, which measures party by voting intention, shows FF support as a stronger effect on 'yes' voting than FG support. One aspect of this party analysis that is arguably problematic is the inclusion of both party e as vote inclination e and government satisfaction in the same model. Party effects are then shown holding satisfaction constant, but arguably satisfaction is bound up with party support, as both cause and consequence. Another set of models was run excluding government satisfaction. Results, showing just FF and FG effects, are in Fig. 4. These show a slightly clearer pattern than those in Fig. 3. Now, the impact of the largest opposition party is usually always non-significant, while the impact of the main government party is always significant. The difference between FF and FG is not large enough to be significant except in 2012 - but it is fairly consistent, and reverses when FG and not FF is in government. The next stage of our analysis involves extending the model estimates to include broader measures of attitudes and some indicators of interest and information. The

Fig. 4. Impact of FF and FG support (excluding government satisfaction) on yes vote 2001e2012: average marginal effects and 95% confidence intervals.

objective here is to see whether omitted variables bias the results obtained so far. None of these additional variables are available in all surveys. Table 2 shows the analysis from Table 1, this time with additional variables as available. We have restricted this to just the last 5 votes, and the surveys used for Figs. 2 and 3, with the exception of Nice 1 where the post vote survey is used, as only that one has any additional variables. As already explained, comparing estimates across nested models is also problematic but the discussion that follows is confirmed by comparisons of average marginal effects, although these are not shown here. In essence, the patterns described already are not altered by this additional information. The particularly strong effect of the integration variable in Lisbon 2 is reduced somewhat, but remains strong; elsewhere it is effectively unaltered. Government satisfaction is marginally reduced, but remains significant in all cases with the exception of Lisbon 2. Party differences between FF and FG are slightly larger for Lisbon, and smaller for Nice 2 - where FG supporters were more supportive than those of FF. The added variables were in some cases highly significant, and while the same variables did not always have exactly the same effects, differences across years were typically not significant. So far we have little evidence to suggest that there is significant variation in the effects of the political and attitudinal variable explored here. Our initial hypothesis certainly cannot be rejected; nor is there much of a pattern in the (non significant) variations that we do find. The final element in this analysis is a more formal assessment of how different the referendums are from one another in terms of the impact of each of the variables in our basic model. We will do this by contrasting the performance of a model run on the pooled data sets used in Figs. 2 and 3, with more limited models that assume that the coefficients are the same in each referendum. There are several such models. One assumes all coefficients are the same, a second that all coefficients are the same but that dummies are needed to allow for different levels of 'yes' at each vote; a third that there are also variations in the party estimates, a fourth assumes also variations in government satisfaction, a fifth

M. Marsh / Electoral Studies 38 (2015) 170e182

177

Table 2 Extended model estimates for EU referendums 2001e2012: odds ratios and standard errors.

FF FG Labour SF Age ABC1 Farmer Male Gov. satisfaction Pro integration

Fiscal

Fiscal

Lisbon 2

Lisbon 2

Lisbon 1

Lisbon 1

Nice 2 IMS

Nice 2 IMS

Nice 1 IMS

Nice 1 IMS

2.34***
0.62 3.79***
1.11 1.3
0.4 0.38**
0.15 1.18***
0.05 1.73***
0.35 2.88**
1.4 1.09
0.22 5.55***
1.26 8.51***
1.8

2.27***
0.61 3.68***
1.08 1.36
0.42 0.38**
0.15 1.17***
0.05 1.74***
0.36 2.93**
1.47 1.17
0.24 5.66***
1.32 8.73***
1.85

3.25***
1.21 1.49
0.46 1.5
0.44 0.26***
0.13 1.83**
0.45 1.33
0.31 2.78
2 0.9
0.23 1.76
0.64 11.71***
2.93

3.79***
1.5 1.24
0.4 1.28
0.41 0.25***
0.12 1.70**
0.41 1.12
0.3 2.07
1.19 1.17
0.29 1.35
0.51 8.72***
2.17

2.34***
0.66 1.84**
0.58 0.78
0.3 0.30*
0.19 2.06***
0.49 2.04***
0.48 2.74*
1.44 1.01
0.23 2.50***
0.6 7.91***
2.03

2.48***
0.72 1.62
0.53 0.74
0.3 0.30*
0.21 1.91***
0.47 2.08***
0.52 3.23**
1.8 1.2
0.29 2.26***
0.56 7.77***
2.1

2.24***
0.53 3.09***
1.01 0.73
0.26 0.29**
0.14 1.19
0.24 1.51**
0.31 2.01**
0.71 0.89
0.17 2.04***
0.44 5.35***
1.05

2.15***
0.54 2.59***
0.88 0.67
0.24 0.34**
0.18 1.07
0.23 1.22
0.27 1.96*
0.76 1.01
0.22 1.81***
0.41 4.55***
0.94

2.76***
0.64 1.15
0.36 1.12
0.48 0.27*
0.21 0.92
0.19 1.2
0.29 1.2
0.32 0.76
0.15

2.95***
0.7 1.18
0.38 1.02
0.49 0.20**
0.16 1
0.22 1.01
0.26 1.21
0.33 0.88
0.19

3.59***
0.77

3.28***
0.73

Additional variables Non-singular identity Knowledge of EU

1.74***
0.37

Attitude to abortion (Never justified) Attitude to neutrality (Abandon) Subjective knowledge Of Treaty Moral liberalism Constant Observations McFadden's Pseudo R2

3.77***
0.97 1.23
0.32 0.71
0.17 2.25***
0.59

1.12
0.3

1.54**
0.33 1.58
0.44 1.31
0.28 2.16***
0.63 1.46
0.35

0.86
0.2 2.82***
0.67

0.76
0.17

0.11***
0.04 976 0.45

0.08***
0.03 976 0.45

0.14***
0.09 769 0.32

0.06***
0.04 767 0.38

0.02***
0.02 578 0.26

0.01***
0.01 572 0.30

0.26***
0.12 744 0.22

0.13***
0.07 731 0.25

1.19
0.34

0.39*
0.21 504 0.13

3.27***
0.86 1.45
0.34 1.01
0.21 0.18***
0.11 504 0.17

Notes: details of each survey in Appendix 1. * sig at .05, ** sig at .01, *** sig at .001.

also variations in the impact of attitudes to integration and a sixth also variations in the demographic effects as well as referendum dummies. Table 3 shows the performance of each of these models as measured by a variety of summary statistics: McFadden's adjusted pseudo R2, AIC and BIC. Taking the R2 measure first, we see that there is a clear gain from including year dummies. The outcomes of the various referendums were different, and the inclusion of

Table 3 Comparing performance of more and less inclusive models of voting in 5 EU referendums. Model

AIC

BIC

Adj McFadden R2

All interactions þ YD Party and Gov sat interactions þ YD Gov sat interactions þ YD Party interactions þ YD EU integration interactions þ YD Demographics interactions þ YD Year Dummies (YD) only No interactions:

3520.0 3521.1 3528.9 3528.7 3543.6 3544.7 3551.1 3713.5

3861.2 3739.5 3647.8 3735.2 3662.2 3714.2 3645.3 3782.2

0.269 0.269 0.267 0.266 0.263 0.264 0.262 0.229

Note: unweighted data.

the dummies simply indicates that these differences were not merely due to changes in the distributions of the independent variables. Inclusion of the dummies increases R2 from 0.228 to 0.262, an improvement of 15% even allowing for the additional variables. After that gains are pretty small. Even with all interactions R2 rises only to 0.269, and improvement of 18%. The most effective addition seems to be government satisfaction, followed by parties and then EU attitudes. However, none of these improves the model by much. The AIC criterion tells the same story: the big gain is the inclusion of referendum dummies. Thereafter the gains are marginal. The BIC measure tells a slightly different story. This is because BIC has a bias towards more parsimonious models and exacts higher penalties for adding more variables. BIC tells us the best model is the one with just year dummies, and worst is that with all interactions. However, a model with just government satisfaction (in addition to year dummies) is not significantly worse than the best model, and those with EU attitudes and with parties and government satisfaction also only marginally less attractive. We do not have to choose a model here, but this analysis underlines the argument made above. When it comes to the determinants of voting behaviour in Irish referendums

178

M. Marsh / Electoral Studies 38 (2015) 170e182

on the EU, the similarities far outweigh the differences. Essentially it seems as if the Irish public voted on Europe again and again in much the same way. Of course the outcomes were different: two of the votes constituted rejections of treaties and the rest acceptance. The models themselves do not provide us with an explanation of this but the different outcomes were not simply because the voters could not translate views on the EU into their vote, or because on one or two occasions there was a very powerful rejection of the government. 5. Conclusions This paper has taken the unique experience of the Irish Republic whose people have voted in more referendums on Europe than have those in any other EU country. Where the modal number is no more than 1, there have been 9 votes in Ireland; leaving aside accession referendums most states have had no such votes at all. That being so, it is more than worthwhile to assess the Irish experience in the light of arguments about what factors matter in EU referendums. This is an argument that is not simply an academic one, but one with implications for the value of referendums in reducing the 'democratic deficit' that continues to afflict the European Union. A key feature in this debate is whether people decide on 'the issues' or whether the dynamics of national politics are the primary drivers of people's vote. Recent research has suggested that the weight of each is conditioned by the quality of campaign, with more intense campaigns tending to provide the information necessary to allow voters to bring their views on the EU to bear on their decision, while weak campaigns leave voters reliant on cues given by their parties, or on their own attitude towards the government in power, as classic 'second order' process of decision making. While surveys have been carried out on behaviour in all referendums on the EU in Ireland, not all of this is available for multivariate analysis, and not all of that contains sufficient items for analysis. Here we looked at referendums since 1992, and looked most closely at those from 2001 onwards. Our initial focus was on what factors seemed to explain voting. The answer was that attitudes to integration seemed to be the strongest single factor, but that party cues also mattered, as did a measure of satisfaction with government, which even controlling for party and EU attitudes was a significant factor in explaining support at all referendums looked at apart from the second vote on the Lisbon Treaty. So far, this is in accord with most of the research that finds party cues and government cues matter, as do attitudes to Europe. We also find demographic patterns in behaviour, with 'yes' voters more middle class and older. Our second focus was on the variations e or the lack of them e in these patterns over time. While we expected no variation, there are several explanations ready made for such variations should they be found. However, our analysis seems to have rendered most of these redundant. It suggests that there is little variation to be found. Adding interaction terms to a model including the basic variables and individual referendum dummies provided

very little improvement in fit. We do see changes in effects but most of these are not significant, and may be no more than sampling error. This is true of both 'first' and 'second' order effects. The switch in the FF and FG effects when government satisfaction was excluded is the probably the strongest result, but even in that instance the variations in the FF effect over time are not significant. The FG effect in 2012 is not significantly greater than that in 2002 but it is significantly larger than that for the pooled 2001-9 set. One reason for this general stability may be that the differences between the qualities of the different campaigns may be too small to generate the sort of differences we might expect to find. Most Irish political observers would say that the campaign over Nice 2 was much more intensive than was the campaign over Nice 1. (e.g. Hayward, 2003; Garry et al., 2005; O'Brennan, 2003) Certainly turnout was much higher and voters claimed to be much better informed, but Hobolt found media coverage e while more intensive the second time around e was still much less intensive that in comparable referendums in Norway and Denmark (Hobolt, 2007) and all four cases assessed by Hobolt using a wider definition of intensity are in the lower half of the 18 cases she examines (Hobolt, 2009: 95). It is harder to suggest there is little real variation in economic confidence as Ireland went from boom in 2001-7 to bust in 2008 onwards. We do see some signs of the expected variation in the ability (or willingness) of the two largest parties to persuade supporters to vote yes, as FG supporters seem to have been most pro European when their party was in government, but party differences are relatively small and between FG and FF e typically non significant. The basic finding here then is that the determinants of voting behaviour in Irish referendums on the EU have varied little over the last five votes, and what earlier data is available suggests that it may have varied little over the last seven or eight at least. That said we do find one clear pattern. This is the improving performance of the model over time. Voting has become more predicable. Voting behaviour on the Fiscal Treaty was particularly predictable, with pseudo R2 close to 0.5. Bigger class differences, sharper political differences and a higher than average weight of EU attitudes all combined to provide this result. These results have a number of implications for our understanding of voting behaviour in referendums and our assessment of the quality of referendum decisions. Even in a country with many votes on Europe, party cues and assessments of government performance can play a role in vote choice. It is certainly not all about 'Europe'. This may be disappointing for some referendum enthusiasts, and certainly is in contrast within Svensson's (2002) conclusions. Of course our search for issue effects is a minimalist one. Models containing a wider range of items on features of each treaty are more successful. But it remains the case that party cues and government satisfaction continue to play a significant role even in enhanced models. The similarity in the views of FF and FG on Europe makes Ireland a good site for looking at the impact of incumbency. The argument that incumbency plays a role in the ability or willingness of parties to campaign for a 'yes'

M. Marsh / Electoral Studies 38 (2015) 170e182

vote also does get some support, but the effects are not entirely unambiguous. Despite several arguments in the literature that lead us to expect attitudes to matter more and government satisfaction less when campaigns are more intense, we found no such clear patterns in the data. Rather we concluded that the effects of each of the variables were substantially the same from referendum to referendum. In the case of campaigns it is possible that this is because variation between the campaigns was relatively small. Given differences between these referendums in turnout and the level of subjective understanding expressed by voters, this suggests that campaign intensity needs to vary much more if it is to matter, probably more than we have seen across EU votes in Ireland, and limits the power of such explanations. This suggests that we cannot simply say that issues would matter more if only parties and civil society groups tried harder, and campaigned more intensively (Lupia and Johnston, 2001). In two of the votes, Nice 2 and Lisbon 2, campaigns went on for more than 12 months and the polls suggested views were well set long before the second votes e in as much as the size of the 'yes' and 'no' votes did not change much in the last few months before each vote. Yet 'issue' voting was not significantly higher in these two referendums than in the other three. The stability of the vote function in these referendums is in contrast to some recent literature. More unexpected was the growing effectiveness of the vote function in accounting for the variation in behaviour within each referendum. We can offer no firm explanation for this, particularly since the improvement is continuous and not simply a feature of the Fiscal Treaty vote, beyond Franklin's (2002) suggestion that repeated referendums might have an 'educative' effect. This does not necessarily translate into stronger and more stable attitudes on Europe, but is does seem to have brought about less random, or unexplained variations in voting behaviour. That is perhaps encouraging for proponents of the referendum device.

179

Variables: Age 18e49/50þ Class ABC1/C2DE/Farmers Party: vote intention MRBI 2001. This was a face-to-face survey of 1001 electors conducted by Market Research Bureau of Ireland for The Irish Times 8 days before the Nice 1 vote. Variables: Age 18e49/50þ Class ABC1/C2DE/Farmers Government satisfaction: satisfied, dissatisfied, no opinion As regards the European Union in general, which of the following comes closest to your own views? _ Ireland should do all it can to unite fully with the European Union _ Ireland should do all it can to protect its independence from the European Union _ Can't choose, don't know. Party: vote intention Subjective understanding of Nice treaty: good understanding, some understanding, vaguely aware, no understanding, no opinion IMS 2001 This was a face-to-face survey of 1378 electors conducted by Irish Marketing Surveys for the EC Office in Dublin. Fieldwork for the survey was conducted between 20th August and 10th September 2001 among a quota sample of 1245 adults. For a report and questionnaire see Sinnott 2002b. Variables: Age 18e49/50þ Class ABC1/C2DE/Farmers Government satisfaction: satisfied, dissatisfied, no opinion Attitudes to EU: As regards the European Union in general, which of the following comes closest to your own views?

Acknowledgements A version of this analysis was presented at Nuffield College in November 2013 and I should like to thank Steve Fisher and James Tilley in particular for their helpful comments. I am also grateful to Sara Hobolt for her response on another draft. Appendix 1. Surveys used in Tables 1 and 2 IMS 1992. This was a face-to-face pre-vote survey of 1293 electors conducted by Irish Marketing Surveys for the Irish Independent on May 2, 1992. Variables: Age 18e49/50þ Class ABC1/C2DE/Farmers How satisfied are you with the way the government is running the country: satisfied, dissatisfied, no opinion Party: vote intention Lansdowne/RTE Exit poll 1998. This was a face-to-face exit poll of 2513 electors conducted by Lansdowne Market research for RTE.

_ Ireland should do all it can to unite fully with the European Union _ Ireland should do all it can to protect its independence from the European Union _ Can't choose, don't know. Attitudes to Neutrality, and to moral liberalism I have a number of statements here that people sometimes make about Europe. I would like you to indicate on this scale which of each pair of opposing statements comes closest to your view. A score of one would indicate that you agree fully with the statement on the left. A score of nine would indicate that you agree fully with the statement on the right. Of course your view could be somewhere in between. Also of course there may be issues that you have no particular view on. If so, please just say this and we will move on to the next item. The current proposals for the development of the European Union will make things like divorce and abortion more easily available in Ireland

180

M. Marsh / Electoral Studies 38 (2015) 170e182

The current proposals for the development of the European Union will have no effect whatsoever on whether things like divorce and abortion will become more easily available in Ireland. No Opinion/Don't Know Ireland should do everything it can to strengthen its neutrality even if this means being less involved in EU cooperation on foreign and defence policy Ireland should be willing to accept limitations on its neutrality so that it can be more fully involved in EU cooperation on foreign and defence policy No Opinion/Don't Know Party: vote intention. Subjective understanding: By the date of the referendum (June 7th), how good was your understanding of the issues involved? Please use this card to choose the phrase that applies best to you:

Ireland should do everything it can to strengthen its neutrality even if this means being less involved in EU cooperation on foreign and defence policy Ireland should be willing to accept limitations on its neutrality so that it can be more fully involved in EU cooperation on foreign and defence policy No Opinion/Don't Know Abortion is never justified Abortion is always justified No Opinion/Don't Know Party: Which, if any, of the political parties do you usually support? Subjective understanding: By the date of the referendum (19th October), how good was your understanding of the issues involved? Please use this card to choose the phrase that applies best to you:

e I had a good understanding of what the Treaty was all about e I understood some of the issues but not all that was involved. e I was only vaguely aware of the issue involved e I did not know what the Treaty was about at all

e I had a good understanding of what the Treaty was all about e I understood some of the issues but not all that was involved. e I was only vaguely aware of the issue involved e I did not know what the Treaty was about at all

Objective knowledge: 5 item quiz IMS 2002 This was a face-to-face survey of 1203 electors conducted between 15th November, 2002 and 9th December, 2002 by Millward Brown IMS/EOS Gallup Europe for the EC representation in Dublin. See Sinnott 2003 for full details Variables: Age 18e49/50þ Class ABC1/C2DE/Farmers Government satisfaction: Overall are you generally satisfied or dissatisfied with the way the government is running the country? very satisfied, quite satisfied, quite dissatisfied, very dissatisfied, no opinion Attitudes to EU: As regards the European Union in general, which of the following comes closest to your own views?

Objective knowledge: 2 item quiz. RED C 2008. This was a post-referendum telephone poll of 1002 electors conducted in the week after the referendum by RED C for an academic team coordinated by Michael Marsh Age 18e44/45þ Class ABC1/C2DE/Farmers Government satisfaction: Thinking about the current Fianna Fail, Progressive Democrat coalition government, how satisfied are you overall with the performance of the government? Very satisfied Quite satisfied Quite unsatisfied Very unsatisfied. Attitudes to EU: As regards the European Union in general, which of the following statements comes closest to your view? “Ireland should do all it can to unite fully with the European Union” OR “Ireland should do all it can to protect its independence from the European Union” Identity Generally speaking, do you see yourself as Irish only, Irish and European, or European only? Irish only; Both Irish and European; European only Knowledge: 4 item quiz (same questions as in 2009 study) Attitudes to Neutrality and Abortion Now I am going to read out some statements that other people have made about political and social issues. For each can you tell me how much you agree with them or not? [Agree Strongly Agree slightly Disagree slightly Disagree strongly]

_ Ireland should do all it can to unite fully with the European Union _ Ireland should do all it can to protect its independence from the European Union _ Can't choose, don't know. Attitudes to Neutrality, and to Abortion I have a number of statements here that people sometimes make. I would like you to indicate on this scale which of each pair of opposing statements comes closest to your view. A score of one would indicate that you agree fully with the statement on the left. A score of nine would indicate that you agree fully with the statement on the right. Of course your view could be somewhere in between. Also of course there may be issues that you have no particular view on. If so, please just say this and we will move on to the next item.

- Ireland should accept limitations on its neutrality so that it can be more fully involved in EU co-operation on foreign and defence policy - Abortion should never be allowed in Ireland

M. Marsh / Electoral Studies 38 (2015) 170e182

RED C 2009. This was a post-referendum telephone poll of 1011 electors conducted in the week after the referendum by RED C Research for an academic team coordinated by Michael Marsh

181

Knowledge: 6 item quiz Appendix 2

Table 2 Distribution of main variables. Voters only. Weighted data.

Yes Govt satisf ProEU Non-manual Farmer FF FG Labour SF

1992 MRBI

1998 Lansdowne

2001 MRBI

2001 IMS

2002 MRBI

2002 IMS

2008 RED C

2009 RED C

2012 RED C

0.71 0.45 na 0.30 0.17 0.47 0.18 0.08 na

0.62 na na 0.45 0.06 0.48 0.19 0.10 na

0.46 0.60 0.49 0.37 0.13 0.38 0.20 0.09 0.04

0.46 na 0.60 0.43 0.27 0.41 0.15 0.07 0.04

0.63 0.35 0.59 0.39 0.11 0.36 0.15 0.10 0.07

0.63 0.39 0.58 0.42 0.12 0.43 0.15 0.09 0.05

0.47 0.44 0.63 0.39 0.10 0.33 0.21 0.10 0.08

0.67 0.18 0.63 0.41 0.11 0.21 0.29 0.13 0.08

0.60 0.47 0.52 0.42 0.08 0.14 0.33 0.13 0.15

Identity: Generally speaking, do you see yourself as Irish only, Irish and European, or European only? Irish only; Both Irish and European; European only Knowledge: 4 item quiz (same questions as in 2008 study) Attitudes to Neutrality and Abortion: Now I am going to read out some statements that other people have made about political and social issues. For each can you tell me how much you agree with them or not? [Agree Strongly Agree slightly Disagree slightly Disagree strongly] - Ireland should accept limitations on its neutrality so that it can be more fully involved in EU co-operation on foreign and defence policy - Abortion should never be allowed in Ireland RED C 2012. This was a telephone poll of 1000 electors conducted in the week after the referendum. The sample was taken largely from those contacted in previous RED C polls over the previous 4 weeks. The study was carried out by RED C Research for an academic team coordinated by Michael Marsh. Variables: Age: 18e44/45þ Class: ABC1/C2DE/Farmers Government satisfaction: How satisfied are you with the performance of the Fine Gael/Labour coalition government since it came into office after the last election e very satisfied, fairly satisfied, not very satisfied, not at all satisfied? Attitudes to EU: As regards the European Union in general, which of the following statements comes closest to your view? “Ireland should do all it can to unite fully with the European Union” OR “Ireland should do all it can to protect its independence from the European Union” Subjective knowledge: How would you describe your knowledge of the Fiscal Treaty? I don't know anything about it at all; I don't know very much about it; I know quite a lot about it; I'm extremely well informed about it

References Cunningham, K., Marsh, M., 2011. A Positive Choice, or Anyone but Fianna il? In: Gallagher, M., Marsh, M. (Eds.), How Ireland voted 2011. Fa Palgrave Macmillan, Basingstoke, pp. 172e202. De Vreese, C.H., Semetko, H.A., 2004. Political Campaigning in Referendums: Framing the Referendum Issue. Routledge, Abingdon. Franklin, M., 2002. Learning from the Danish case: a comment on Palle Svensson's critique of the Franklin thesis. Eur. J. Polit. Res. 41, 751e757. Franklin, M., van der Eijk, C., Marsh, M., 1995. Referendum outcomes and trust in government: public support for Europe in the wake of Maastricht. West Eur. Polit. 18 (3), 101e107. Franklin, M., Marsh, M., McLaren, L., 1994a. Uncorking the bottle: popular opposition to European unification in the wake of Maastricht. J. Common Mark. Stud. 32 (4), 455e472. Franklin, M., Marsh, M., Wlezien, C., 1994b. Attitudes towards Europe and referendum votes: a response to Siune and Svensson. Elect. Stud. 13 (2), 117e121. Gabel, M., 1998. Interests and Integration: Market Liberalization, Public Opinion, and European Integration. University of Michigan Press, Ann Arbor. Gallagher, M., 1988. The Single European Act referendum. Ir. Polit. Stud. 3, 77e82. Garry, J., Marsh, M., Sinnott, R., 2005. “Second-order” versus “Issue voting” effects in EU referendums: evidence from the Irish Nice Treaty referendums. Eur. Union Polit. 6 (2), 201e221. Glencross, A., Trechsel, A., 2011. First or second order referendums? understanding the votes on the EU constitutional treaty in four EU Member states. West Eur. Polit. 34 (4), 755e772. Hooghe, L., Marks, G., 2005. Calculation, community and cues. Eur. Union Polit. 6 (4), 419e443. Hayward, K., 2002. Not a nice Surprise: an analysis of the debate on the Nice Treaty in the Republic of Ireland. Ir. Stud. Int. Aff. 13, 167e186. Hayward, K., 2003. “If at first you don't succeed…”: the second referendum on the Treaty of Nice, 2002. Ir. Polit. Stud. 18 (1), 120e132. Hobolt, S.B., 2009. Europe in Question: the Role of Political Information in Referendums on European Integration. Oxford University Press, Oxford. Hobolt, S.B., 2007. Campaign information and voting behaviour. In: de Vreese, C.H. (Ed.), The Dynamics of Referendum Campaigns: an International Perspective. Palgrave Macmillan, Basingstoke, pp. 84e116. Lubbers, M., 2008. Regarding the Dutch ‘Nee’ to the European Constitution: a test of the identity, utilitarian and political approaches to voting ‘No’. Eur. Union Polit. 9 (1), 59e86. LeDuc, L., 2005. Opinion formation and change in referendum campaigns. In: de Vreese, C.H. (Ed.), The Dynamics of Referendum Campaigns: an International Perspective. Palgrave Macmillan, Basingstoke, pp. 21e46. LeDuc, L., 2003. The Politics of Direct Democracy: Referendums in Global Perspective. University of Toronto Press, Toronto. Lupia, A., Johnston, R., 2001. Are voters to blame? Voter competence and elite maneuvers in referendums. In: Mendelsohn, M., Parkin, A. (Eds.), Referendum Democracy: Citizens, Elites and Deliberation in Referendum Campaigns. Palgrave Macmillar, New York, pp. 191e210.

182

M. Marsh / Electoral Studies 38 (2015) 170e182

Marsh, M., 2007. Referendum campaigns: changing what people think or changing what they think about? In: de Vreese, C. (Ed.), The Dynamics of Referendum Campaigns: an International Perspective. Palgrave Macmillan, Basingstoke, pp. 63e83. Marsh, M., Sinnott, R., Kennedy, F., Garry, J., 2008. The Irish voter. Manchester University Press, Manchester. Marsh, M., Mikhaylov, S., 2010. European parliament elections and EU Governance. Living Rev. Eur. Gov. 4 (1), 41. McLaren, L., 2002. Public Support for the EU: Cost/Benefit Analysis or Perceived Cultural Threat. J. Polit. 64 (2), 551e556. McLaren, L., 2006. Identity, Interests and Attitudes to European Integration. Palgrave Macmillan, Basingstoke. Mood, C., 2010. Logistic regression: why we cannot do what we think we can do, and what we can do about it. Eur. Sociol. Rev. 10 (1), 67e82. O'Brennan, J., 2003. Ireland's return to “normal” voting patterns on EU issues: the 2002 Nice treaty referendum. Eur. Polit. Sci. 2 (2), 5e13. Petithomme, M., 2011. Awakening the sleeping giant? the displacement of the partisan cleavage and change in government-opposition dynamics in EU referendums. Perspect. Eur. Polit. Soc. 12 (1), 89e110. Pierce, R., Valen, H., Listhaug, O., 1983. Referendum voting behaviour: the Norwegian and British Referenda in membership in the European Community. Am. J. Polit. Sci. 27 (1), 43e63.

Ray, L., 2003. When parties matter. J. Polit. 65 (4), 978e994. Sattler, T., Urpelainen, J., 2012. Explaining public support for international integration: how do national conditions and treaty characteristics interact with individual beliefs? J. Polit. 74 (4), 1108e1124. Schneider, G., Weitsman, P., 1996. The punishment trap: integration referendums as popularity contests. Comp. Polit. Stud. 28 (4), 582e607. Sinnott, R., 2003. Attitudes and Behaviour of the Irish Electorate in the Second Referendum on the Treaty of Nice. ISSC Discussion Paper Series; WP2003/18. Sinnott, R., 2002a. Cleavages, parties and referendums: relationships between representative and direct democracy in the Republic of Ireland. Eur. J. Polit. Res. 41, 811e826. Sinnott, R., 2002b. Attitudes and Behaviour of the Irish Electorate in the Referendum on the Treaty of Nice. http://www.ucd.ie/dempart/ workingpapers/nice1.pdf. Svensson, P., 2002. Five Danish referendums on the European Community and European Union: a critical assessment of the Franklin thesis. Eur. J. Polit. Res. 41, 733e750. Zaller, J.R., 1992. The nature and origins of mass opinion. Cambridge University Press, Cambridge.