‘Second-order’ versus
‘Issue-voting’ Effects in
EU Referendums
Evidence from the Irish Nice Treaty
Referendums
John Garry
Queen’s University Belfast, Northern Ireland
Michael Marsh
Trinity College Dublin, Republic of Ireland
Richard Sinnott
University College Dublin, Republic of Ireland
A B S T R A C T
Are referendums on EU treaties decided by voters’ attitudes
to Europe (the ‘issue-voting’ explanation) or by voters’ attitudes
to their national political parties and incumbent
national government (the ‘second-order election model’
explanation)? In one scenario, these referendums will
approximate to deliberative processes that will be decided
by people’s views of the merits of European integration. In
the other scenario, they will be plebiscites on the performance
of national governments. We test the two competing
explanations of the determinants of voting in EU referendums
using evidence from the two Irish referendums on the
Nice Treaty. We find that the issue-voting model outperforms
the second-order model in both referendums. However, we
also find that issue-voting was particularly important in the
more salient and more intense second referendum. Most
strikingly, attitudes to EU enlargement were much stronger
predictors of vote at Nice 2 than at Nice 1. This finding about
the rise in importance of attitudes to the EU points to the
importance of campaigning in EU referendums.
2 0 1
European Union Politics
DOI: 10.1177/1465116505051983
Volume 6 (2): 201–221
Copyright© 2005
SAGE Publications
London, Thousand Oaks CA,
New Delhi
K E Y W O R D S
᭿ effective campaigning
᭿ issue voting
᭿ Nice Treaty
᭿ referendums
᭿ second-order model
03 Garry (to/d) 12/4/05 11:26 am Page 201
Introduction
Are referendums on EU treaties decided by voters’ attitudes to Europe or by
their attitudes to their national politics and to the incumbent national government?
The political science literature is sharply divided on the matter. Given
the raft of upcoming referendums on the EU constitution, the significance of
this issue can hardly be overstated. Denmark, France, Ireland, Portugal, Spain
and the United Kingdom are due to hold referendums on the new EU constitution
in 2005 or 2006. Referendums are also very likely to occur in Belgium,
Luxembourg and the Netherlands and may occur in Estonia, Latvia, Lithuania,
Poland and Slovenia.1 In one scenario, these referendums will approximate
to deliberative processes that will be decided by people’s views of the
merits of European integration and of the new constitution. In the other
scenario, they will be plebiscites on the performance of national governments.
In this paper we assess the relative strength of these two rival interpretations
of voting in EU referendums: do people vote on the basis of their attitudes
to the EU (what we call the ‘issue-voting’ or ‘attitudes’ explanation) or on the
basis of domestic political concerns (the ‘second-order election model’ explanation)?
We also assess the impact of the salience of the referendum campaign
on the relative strength of these two explanations.
We focus on the referendums on the Nice Treaty in the Republic of
Ireland. This is a useful case for two reasons. First, Irish citizens were the
only ones who were asked what they thought about ratification of the Treaty
of Nice. Thus, Ireland is the only case we can draw on relating to the most
recent EU treaty. Secondly, there were two referendums on the Nice Treaty
in Ireland that yielded two different results. In a referendum in 2001, Irish
citizens rejected the treaty, dealing a dramatic blow to the EU’s plans. The
Nice Treaty could not be implemented unless, and until, all member states
ratified it. The Irish government then held a second referendum on the Nice
Treaty in 2002 and this time Irish voters endorsed the treaty.2 We explore the
extent to which the move from rejection to endorsement was a function of
change in the relative weight of the ‘second-order’ and ‘attitudes’ factors.
The first Nice Treaty referendum campaign was a classic case of an ineffective
campaign, with low levels of citizen interest in, and knowledge of, the
treaty. After the shock of defeat in the first Nice referendum, a much more
vigorous campaign was mounted at the second referendum, with resultant
increases in the salience of the issue and in citizens’ levels of information.
We can thus investigate the issue we are concerned with – the relative
strength of the different interpretations of voting – in two contexts that were
more or less identical except for a variable of crucial interest, namely the
intensity of campaigning.
European Union Politics 6(2)2 0 2
03 Garry (to/d) 12/4/05 11:26 am Page 202
We start by describing the two competing explanations of EU referendum
voting in more detail. We then specify how the relative merits of these
explanations can be assessed in the context of the referendum campaigns on
the Nice Treaty in Ireland. After describing the data used in the analysis, we
report our results. We conclude by discussing the implications of our findings
in the context of the impending EU referendums on the new EU constitution.
Conflicting interpretations of EU referendum voting
There are, as noted, two distinct schools of thought on why people vote the
way they do in EU referendums. One school focuses on individuals’ attitudes,
values and beliefs. People support or reject EU treaties, the argument goes,
on the basis of their underlying broad attitudes towards the EU project.
People who are generally positive towards European integration and the
development of the EU will be likely to support an EU treaty and vote ‘Yes’.
People who are generally sceptical about integration and the EU ‘project’ will
vote ‘No’. Of course, a person’s general outlook on the EU may take
somewhat different forms. Some people may worry about losing political and
economic independence; others may be concerned about the possible militarization
of Europe; still others may feel that the EU is big enough and should
not take in a large number of additional states. Furthermore, views on the
role of the government in the economy or one’s position on the social
liberal–conservative dimension may influence attitudes to the EU. Whatever
the particular nature of a person’s attitude to the EU, the issue-voting approach
suggests that it is primarily voters’ views on the development of the EU that
drive voting in a referendum on an EU treaty. (For example, on the Nordic
countries, see Aardal et al., 1998. On Denmark in particular, see Siune and
Svensson, 1993; Siune et al., 1994; Svennson, 1994, 2002. On the Baltic states
– testing the hypothesis that authoritarian values lead people to reject
membership of the EU in a referendum – see Ehin, 2001. On Norway and
Britain – focusing on the relationship between economic left–right positions
and support for membership – see Pierce et al., 1983.) Overall, the ‘attitudes’
approach assumes that views on the EU and/or on the substance of the treaty
are the main determinants of voting behaviour in EU referendums.
A different approach to explaining voting behaviour in EU referendums
focuses on concerns quite separate from the EU. This explanation is associated
with the theory that certain elections are best seen as ‘second-order’ elections
(Reif and Schmitt, 1980; van der Eijk et al., 1996; for a similar approach,
see Anderson, 1998). ‘Second-order’ elections are elections (or, generally, electoral
contests) such as local, presidential and European Parliament elections
Garry, Marsh and Sinnott ‘Second-order’ versus ‘Issue-voting’ Effects 2 0 3
03 Garry (to/d) 12/4/05 11:26 am Page 203
and referendums on EU and other matters that are not perceived by political
actors – including voters – to be as important as national or general elections
(which are ‘first-order’ elections). According to the theory, voting behaviour
in such second-order electoral contests is heavily influenced by first-order
considerations. Following this logic, voters might be expected to use secondorder
contests as mechanisms for signalling their support, or lack of support,
for their domestic political parties and government. (On the application of
the second-order model to EU referendums, see Franklin, Marsh and Wlezien,
1994; Franklin, Marsh and McLaren, 1994; Franklin et al., 1995; Franklin, 2002;
Svensson, 2002; Marsh, 1998; and Ferrara and Weishaupt, 2004). Voters who,
for whatever reasons, are not satisfied with the performance of the incumbent
government may take the opportunity to punish the government by
voting against the government’s wishes in a ‘second-order’ election such as
an EU referendum. Voters who are satisfied with the performance of the party
(or parties) in government may vote in line with the government’s wishes
(supporting the governing party, or parties, in the local election, voting ‘Yes’
in an EU referendum, etc.). Thus, an EU referendum may, in fact, be a general
election by another name.
A key element of the second-order election theory relates to the notion
of salience. Domestic considerations, the argument goes, will be a particularly
powerful determinant of voting in situations where the election or referendum
is perceived to be very unimportant. The more important the election
or referendum in question is seen to be, the lower the role that will be played
by domestic political (second-order) effects and the higher the role played by
citizens’ attitudes towards, or views on, the EU and the substantive content
of the treaty in question.
Several authors have also argued that the institutional context affects how
strong second-order factors are likely to be. In a development of Schneider
and Weitsman’s analysis (1996), Hug (2002) and Hug and Sciarini (2000)
distinguish between a referendum that is constitutionally necessary and one
that the government chooses to hold. Level of satisfaction with the incumbent
government and domestic party political factors are likely to play a
weaker role in the former compared with the latter. This is because much
more damage can be inflicted on the government in a non-required referendum;
losing a referendum that you did not have to hold in the first place
makes the government look particularly silly. Hug (2002) and Hug and
Sciarini (2000) also distinguish between a binding and a non-binding referendum.
In relation to the former, the result cannot be overturned, whereas
the parliament might possibly change the decision of a non-binding referendum.
Second-order effects are likely to be stronger in non-binding than in
binding referendums. This is because, in a non-binding referendum, voters
European Union Politics 6(2)2 0 4
03 Garry (to/d) 12/4/05 11:26 am Page 204
who wish to punish the incumbent government by voting against the referendum
can hope that, if the referendum is rejected, the outcome may then be
overturned by the parliament.
On balance, we would characterize both the Irish Nice referendums as
necessary and binding (for a discussion of the institutional context and, in
particular, of the initiation of referendums in Ireland, see Sinnott, 2002:
811–14). It may seem odd to view the first referendum as binding because it
clearly was not so, in the sense that the government did not accept the
decision as final and went on to hold a second referendum. However, it was
binding in the sense that the only way to overturn the decision was by having
another referendum. The first Nice referendum was also constitutionally
necessary in the sense that the government almost certainly would have been
brought to the High Court and forced to have a referendum if it had tried to
argue that the treaty did not have fundamental constitutional implications
and thus that a referendum was not necessary. A successful judicial challenge
of this sort had forced the government of the day to hold a referendum on
the Single European Act and this set a precedent that would apply to any
subsequent proposals for significant changes to EU treaties. The second Nice
referendum was politically necessary for two further reasons. The first was
that Irish political leaders have repeatedly committed themselves to holding
referendums on any changes that even approach the magnitude of those
contained in the Nice Treaty. Secondly, the government was under severe
pressure to hold a second referendum because the result of the first referendum
plunged both the accession process and Ireland’s relationship with the
EU into crisis. If Hug and Sciarini (2000) are right, then we would expect
second-order effects to be less evident in both the – necessary and binding –
Irish Nice treaty referendums. Accordingly, the Irish Nice referendums
present a tough test of the second-order election interpretation. However,
given what we said above about the change in the salience of the campaign
between Nice 1 and Nice 2, second-order effects are likely to be more evident
at Nice 1 than at Nice 2.
The ‘attitudes’ and ‘second-order’ national election interpretations of
voting behaviour in EU referendums are quite distinct from each other.3 The
attitudes approach assumes a rational and reasoned calculation by the voter
based on his or her views on EU matters. The second-order approach sees
voting in EU referendums as simply a chance for voters to express their
domestic political preferences and vote either in line with or against the
governing parties’ wishes (a vote mainly reflecting their satisfaction/dissatisfaction
with the government/governing party(ies)). The question of which
interpretation is most accurate – or has the greatest predictive power – has
important implications for how we interpret the role of referendums and the
Garry, Marsh and Sinnott ‘Second-order’ versus ‘Issue-voting’ Effects 2 0 5
03 Garry (to/d) 12/4/05 11:26 am Page 205
part that referendums play in EU democracy. As Svensson puts it: do voters
‘really address [the] issues and involve themselves actively in the policymaking
process on a vital issue or [do] they merely vote for or against the
current government?’ (2002: 733). If the latter is the case, then the upcoming
referendums on the new EU constitution will be decided by a mishmash of
extraneous factors that have little or nothing to do with European integration
or with the painstakingly negotiated EU constitution. Likewise and more
generally, if the second-order model is correct, we should beware of inferring,
in the event of a ‘Yes’ vote on EU treaty changes, that the electorate was in
favour of the substantive content of the treaty in question.
Two hypotheses follow from this discussion.
Hypothesis 1: Relative to second-order effects, issue-voting effects are stronger
predictors of voting (Yes/No) in EU referendums.
Hypothesis 2: Relative to low-salience EU referendums, high-salience EU referendums
are likely to show stronger issue-voting effects.
The Nice referendums in Ireland
Three issues were particularly relevant in the referendum campaigns on the
Nice Treaty in Ireland. First, citizens who were in favour of retaining as much
independence and sovereignty in Ireland as possible were likely to have voted
against the Nice Treaty. In contrast, citizens who were in favour of sharing
(or pooling) sovereignty and decision-making powers with other EU states
were likely to have voted ‘Yes’ (in favour of the referendum). Secondly,
citizens who favoured maintaining the Irish position of military ‘neutrality’
– in the sense of having nothing to do with a militarized EU – were likely to
vote ‘No’ to Nice. People who favoured moves towards a strong European
military capability were likely to vote ‘Yes’. Third, the expansion of the EU
to include a large number of (relatively poor) countries was supported by
some citizens and opposed by others. The latter perhaps regarded such an
expansion as a threat to Ireland’s ability to continue securing large amounts
of money from the EU. Ireland, in fact, was likely to become a net contributor
to, rather than beneficiary of, the EU. (For more detailed descriptions of
how divisions and debates in these areas played out in the Nice 1 and Nice
2 campaigns, see O’Mahoney, 2001; Hayward, 2002, 2003; Gilland, 2002, 2003.)
After the shock of defeat in the 2001 referendum, the pro-Nice camp
sought to address what they perceived to be the concerns of the citizenry. The
governing parties (Fianna Fáil and the Progressive Democrats), the pro-EU
opposition parties (Fine Gael and Labour) and the rest of what the anti-Nice
European Union Politics 6(2)2 0 6
03 Garry (to/d) 12/4/05 11:26 am Page 206
camp regards as the pro-EU ‘establishment’ – including the major trade
unions and business organizations – ran a very spirited and vigorous
campaign emphasizing the advantages of the European Union. The energetic
pro-Nice campaign in 2002 contrasted starkly with the lax and lethargic
campaign of the pro-Nice lobby in 2001. O’Brennan (2003), for example, states
that a key reason for the success of the ‘Yes’ side in Nice 2 was their ‘much
more vigorous and visible campaign’. At Nice 1 their campaign had been
‘lacklustre and indifferent’ and lacked ‘energy, passion, intensity, and,
crucially, a visible campaigning presence on the ground in individual
constituencies. The No campaign in contrast was charged with conviction,
well organized and gained in confidence as the campaign went on.’
One indication of the increased salience of the campaign at the second
referendum compared with the first is the fact that citizens found the media
to be more useful sources of information in relation to the referendum at Nice
2 than at Nice 1.4 At Nice 2, 64% of respondents found newspaper articles
either very or somewhat useful (compared with 44% at Nice 1), 73% found
television news and current affairs programmes very or somewhat useful
(compared with 51% at Nice 1), and 68% found radio news and current affairs
programmes very or somewhat useful (compared with 46% at Nice 1). There
was also a marked increase in the proportion of citizens who found discussion
with the family, friends and colleagues very or somewhat useful (64% at Nice
2 compared with 48% at Nice 1). One consequence of the higher-salience
campaign at Nice 2 was the increase in citizens’ level of knowledge of the
treaty and issues relating to the treaty. At Nice 1, 36% of respondents reported
that they had either a good understanding of, or understood some of, the
issues relating to the Nice Treaty. This figure had risen to 61% by the end of
the Nice 2 campaign.5 Thus, there had been a significant increase in citizens’
comprehension of the treaty and issues relating to the treaty between Nice 1
and Nice 2, suggesting that the energetic and vigorous campaign that
occurred at Nice 2 was indeed effective.
In the inter-referendum period, the pro-Nice camp sought to increase the
importance of the substantive (European) issues relative to the troublesome
issues of ‘party politics’ (aka second-order effects) and military neutrality.
Specifically in relation to the issue of neutrality, which was widely felt to have
contributed to the rejection of the treaty at Nice 1, the government agreed –
at the Seville summit in June 2002 – two Declarations with its European
partners that were then added to the Nice Treaty. The ‘National Declaration’
by Ireland states that ‘Ireland is not party to any mutual defence commitment’
and that ‘Ireland is not party to any plans to develop a European army’.
The ‘Declaration of the European Council’ states that ‘Ireland’s policy of
military neutrality is in full conformity with the Treaties, on which the
Garry, Marsh and Sinnott ‘Second-order’ versus ‘Issue-voting’ Effects 2 0 7
03 Garry (to/d) 12/4/05 11:26 am Page 207
European Union is based, including the Treaty of Nice and that there is no
obligation arising from the Treaties which would or could oblige Ireland to
depart from that policy’. Further, the government inserted a clause into the
proposed constitutional referendum text guaranteeing that Ireland would not
join any EU common defence.6 It is likely that these actions served to deemphasize
the importance of the pro- versus anti-military alliance dimension
in the run-up to Nice 2.
The pro-Nice camp also sought to de-couple issues relating to domestic
party politics from the issue of the referendum. As noted earlier, all Irish
parties, apart from the small Sinn Féin and Green parties, advocated a ‘Yes’
vote in Nice 1 and Nice 2. However, at Nice 2 the main opposition parties
(Labour and Fine Gael), in line with the government parties (Fianna Fáil and
the Progressive Democrats), emphasized the importance of detaching the
referendum issue from the issue of support for the governing parties. This is
nicely illustrated by a slogan used by the opposition Labour Party, which
sought to persuade voters not to treat Nice 2 as a means of manifesting their
opposition to the main government party, Fianna Fáil. The slogan was: ‘Hold
Your Fire. Fianna Fáil Can Wait. Europe Can’t.’ Labour thus urged voters to
save their anti-Fianna Fáil ammunition for another day. In effect, the pro-Nice
parties were implicitly subscribing to the ‘second-order election’ interpretation;
to the extent that their countermeasures were successful, they will have
served, in line with hypothesis 2 above, to weaken any potential second-order
effects in the case of Nice 2.
Data
The data that we use to test our hypotheses come from two post-referendum
nationally representative surveys commissioned by the EU and conducted by
EOS Gallup (for an extensive analysis of these two surveys, see Sinnott, 2001,
2003).7 To tap attitudes to European integration, the 2001 and 2002 surveys
asked respondents the following question:
As regards the European Union in general, which of the following comes closest to your
views: 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?
Respondents choosing the first statement are labelled ‘pro-integrationists’,
respondents choosing the second statement are labelled ‘anti-integrationists’,
and respondents indicating that they either did not know or could not choose
are labelled ‘neither pro- nor anti-integrationists’. To tap attitudes to EU
enlargement, respondents were asked:
European Union Politics 6(2)2 0 8
03 Garry (to/d) 12/4/05 11:26 am Page 208
It is envisaged that, over the coming years, there will be further enlargement of the EU.
The EU is at present negotiating with 12 candidate countries. Ten of these countries are
in eastern and central Europe. The other two are Cyprus and Malta. We are interested in
how people feel about further enlargement of the EU and what people see as the possible
advantages and disadvantages of such enlargement. First of all, in general terms, are you
in favour or against such enlargement of the EU?
Respondents choosing the response option ‘in favour’ are labelled ‘proenlargement’,
respondents choosing the option ‘against’ are labelled ‘antienlargement’,
and respondents indicating that they did not know are labelled
‘neither pro- nor anti-enlargement’. To tap attitudes to military neutrality,
respondents were asked which of the following two statements they agreed
with:
Ireland should do everything it can to strengthen its neutrality even if this means being
less involved in EU co-operation on foreign and defence policy OR Ireland should be
willing to accept limitations on its neutrality so that it can be more fully involved in EU
co-operation on foreign and defence policy.
Respondents were asked to place themselves on a nine-point scale (1 = agree
fully with the first statement and 9 = agree fully with the second statement).
To retain the same metric as the measures of integration and enlargement, we
trichotomize responses: 1–4 = ‘pro-neutrality’; 6–9 = ‘anti-neutrality’; 5 (‘no
opinion’ and ‘don’t know’) = ‘neither pro- nor anti-neutrality’.
In order to operationalize ‘second-order’ effects, the party preferences of
respondents were tapped in both surveys and respondents were assigned one
or the other of the following labels: supporter of a government party (Fianna
Fáil or Progressive Democrats), supporter of a pro-Nice Treaty opposition
party (Fine Gael or Labour), supporter of one of the small anti-Nice opposition
parties (Sinn Féin or the Green party), or a supporter of either no party
or a very minor party (other). Ideally, a question concerning how satisfied
each respondent was with the government would also have been asked in
both surveys. However, a ‘satisfaction’ question was asked only at Nice 2 and
so we used ‘satisfaction’ only in our analysis of that particular referendum.
(Age, social class and sex were also measured and are used in the analysis as
demographic control variables.)8 The Appendix, which can be found on the
EUP homepage, reports the frequencies for our main independent variables.
Results
Table 1 reports the results of three multivariate logistic regressions. Each
regression includes issue-voting and second-order variables as predictors of
Garry, Marsh and Sinnott ‘Second-order’ versus ‘Issue-voting’ Effects 2 0 9
03 Garry (to/d) 12/4/05 11:26 am Page 209
‘Yes’ voting in the referendum in question (controlling for demographic
effects). Our issue-voting variables relate to attitudes to EU integration, EU
enlargement and military neutrality. In columns 1 and 2 of Table 1, secondorder
effects are operationalized in terms of whether or not the respondent
is a supporter of a governing party or of one of the opposition parties. In
column 3 we re-run the analysis for Nice 2 but this time we also include a
‘satisfaction with government’ variable (which we have only for Nice 2) as
part of the operationalization of the second-order model. All three regressions
reported in Table 1 show that the issue effect and second-order effect variables
are related to voting in the predicted direction. Respondents who are in
favour of EU integration tend to vote ‘Yes’, as do respondents who favour
enlargement and respondents who are in favour of limiting military neutrality.
Also, those who support parties that are not in government are significantly
less likely than respondents who support the governing parties to vote
‘Yes’. A similar pattern emerges when the explicit government satisfaction
variable is included: respondents who are very dissatisfied with the incumbent
government (the reference category) are less likely than those who are
either less dissatisfied or those who are satisfied to vote ‘Yes’.
Although these findings are interesting in themselves, their real value is
that they enable us to test our two fundamental hypotheses regarding the
relative impact of second-order factors and issues on how people vote in EU
referendums. We can test the relative strength of the two interpretations of
voting by assessing how much each approach contributes to an explanation
of voting in the referendums. We begin by focusing on Nice 1. We run a model
(see top row of Table 2) using only demographics to predict voting behaviour
(log likelihood = –339.9). Adding second-order effects to this model –
operationalized in terms of party support – improves the fit of the model by
6.0% (log likelihood declines to –319.7). However, the addition of ‘issue
effects’ to this ‘demographics + second-order’ model improves the model fit
more substantially. The log likelihood declines from –319.7 to –267.6, representing
an improvement of fit of 16.3%.9 When the same series of analyses is
conducted for Nice 2 we also find that adding ‘issues’ to a ‘demographics +
second-order’ model increases the model fit. However, this time the increase
in model fit is much larger, 26.2%.
The key importance of issue effects also holds when we use a more
comprehensive operationalization of second-order effects (see Table 3). For
Nice 2, we run a model in which second-order effects are operationalized in
terms of party support and level of satisfaction with the government. As
reported in Table 3, this more comprehensive operationalization of secondorder
effects has little impact. The model improvement that results from the
addition of ‘issue effects’ to a ‘demographics + “comprehensive” second-order
model’ (99.5, or 23.7%) is still much larger than the model improvement that
European Union Politics 6(2)2 1 0
03 Garry (to/d) 12/4/05 11:26 am Page 210
results from the addition of a comprehensive operationalization of secondorder
effects to a ‘demographics only’ model (64.6, or 13.3%).
These results confirm both our hypotheses. First, in both of the Irish referendums
on the Nice treaty, issues are stronger predictors of vote choice than are
second-order effects. Secondly, when we compare the predictive strength of
Garry, Marsh and Sinnott ‘Second-order’ versus ‘Issue-voting’ Effects 2 1 1
Table 1 Summary of logistic regression analyses: ‘Yes’ voting in the 2001 and 2002
Nice Treaty referendums in the Republic of Ireland
2001 Nice Treaty 2002 Nice Treaty
————————— —————————————————–——
Coeff. P >|z| Coeff. P >|z| Coeff. P >|z|
Second-order effects
(Fianna Fáil/ Progressive
Democrats = reference
category)
Fine Gael/Labour –0.83 0.003 –0.72 0.008 –0.51 0.070
Sinn Féin/Green Party –2.20 0.000 –1.56 0.000 –1.50 0.000
Others –0.91 0.001 –1.13 0.000 –1.16 0.000
(Very unsatisfied with
govt = reference category)
Quite unsatisfied with govt 0.84 0.002
Quite/very satisfied with govt 0.81 0.003
Issue-voting effects
(Anti-integration =
reference category)
Centrist 0.41 0.143 0.03 0.931 –0.01 0.948
Pro-integration 1.37 0.000 1.36 0.000 1.36 0.000
(Anti-enlargement =
reference category)
Centrist 0.32 0.382 1.83 0.000 1.81 0.000
Pro-enlargement 1.06 0.001 2.27 0.000 2.16 0.000
(Strengthen neutrality =
reference category)
Centrist 0.62 0.016 0.71 0.008 0.61 0.032
Limit neutrality 1.22 0.000 0.75 0.012 0.70 0.023
N 504 768 768
McFadden’s pseudo R2 (adj.) 0.18 0.30 0.31
Log likelihood –267.6 (18) 0.000 –338.0 (18) 0.000 –319.99 (20) 0.000
Notes: Sex, age and social class controlled (co-efficients for these or for the intercept not
reported); data are weighted to reflect the actual referendum results.
Source: Commission/EOS Gallup post-referendum surveys on Irish Nice referendums 2001 and
2002.
03 Garry (to/d) 12/4/05 11:26 am Page 211
EuropeanUnionPolitics6(2)212
Table 2 Contribution of ‘second-order’ and ‘issue-effect’ variables to the explanation of voting in the Nice referendums
Nice 1 Nice 2
Log likelihood (df) Log likelihood (df)
(A) Intercept + demographics –339.9 (9) –484.0 (9)
(B) Intercept + demographics + second-order effects (party support) –319.7 (12) –457.1 (12)
(C) Intercept + demographics + second-order effects (party support)
+ issues (three issues) –267.6 (18) –337.2 (18)
Nice 1 Nice 2
—————————————————————————— ———————————————————————————
Change Change
in log Statistical % improvement in log Statistical % improvement
likelihood significance in log likelihood significance in log
(df) of change likelihood (df) of change likelihood
(B) compared with (A) 20.3 (3) .000 6.0 26.9 (3) .000 5.6
(C) compared with (B) 52.0 (6) .000 16.3 119.9 (6) .000 26.2
03Garry(to/d)12/4/0511:26amPage212
issues across the two referendums, issues turn out to be stronger predictors in
the second, more salient referendum.
We now focus our attention on the individual issues – EU integration,
EU enlargement and military neutrality – and assess the relative impact of
these issues at each referendum. To do this we display again (in Table 4) the
information presented in Table 1, but this time we report ‘conditional
maximum effects’ instead of the raw coefficients (which do not lend themselves
to easy substantive interpretation). In terms of the impact of issues at
Nice 1, the sizes of the effects of the three different subjects were quite similar
to each other. Pro-integrationists were 33 percentage points more likely than
anti-integrationists to vote ‘Yes’, pro-enlargement voters were 25 percentage
points more likely than anti-enlargement voters to vote ‘Yes’, and voters who
favoured limiting neutrality were 29 percentage points more likely than voters
favouring a strengthening of neutrality to vote ‘Yes’. However, the relative
size of these attitude effects changed quite dramatically between Nice 1 and
Nice 2. The impact of attitudes to integration declined slightly and the impact
of attitudes to neutrality also declined somewhat, yet the impact of these
Garry, Marsh and Sinnott ‘Second-order’ versus ‘Issue-voting’ Effects 2 1 3
Table 3 Contribution of ‘second-order’ (including government satisfaction) and
‘issue-effect’ variables to the explanation of voting in the second Nice referendum
Log likelihood (df)
(A) Intercept + demographics –484.0 (9)
(B) Intercept + demographics + second-order effects
(party support) –457.1 (12)
(C) Intercept + demographics + second-order effects
(government satisfaction) –439.5 (11)
(D) Intercept + demographics + second-order effects
(party support and government satisfaction) –419.4 (14)
(E) Intercept + demographics + second-order effects
(party support and government satisfaction) + issue
effects (three issues) –319.9 (20)
Change Statistical % improvement
in log significance in log
likelihood (df) of change likelihood
(B) compared with (A) 26.9 (3) .000 5.6
(C) compared with (A) 44.5 (2) .000 9.2
(D) compared with (A) 64.6 (5) .000 13.3
(E) compared with (D) 99.5 (6) .000 23.7
03 Garry (to/d) 12/4/05 11:26 am Page 213
reductions was more than offset by the rise in importance of attitudes to
enlargement. In 2002, pro-enlargement voters were 51 percentage points more
likely than anti-enlargement voters to vote ‘Yes’ (twice the size of the 2001
effect). Similarly, ‘centrists’ on enlargement were 43 percentage points more
likely than anti-enlargement voters to vote ‘Yes’ in 2002 (compared with 7
percentage points in 2001). Overall, anti-enlargement voters were 50 percentage
points less likely than other voters to vote ‘Yes’ (whereas they had been
only 19 percentage points less likely at Nice 1). In the third column of Table
4, we see that a different operationalization of second-order effects – using
level of satisfaction with the incumbent government in addition to party
support – alters the size of each of the attitude effects only very marginally.
Thus, there was a change in the relative strength of second-order and
attitude factors between Nice 1 and Nice 2, driven mainly by an increase in
European Union Politics 6(2)2 1 4
Table 4 Conditional maximum effects of each variable with remaining dummies at
zero and other variables at their mean
2001 2002(a) 2002(b)
Second-order effects
(Fianna Fáil/ Progressive Democrats) (+25) (+20) (+19)
Fine Gael/Labour –21 –13 –9
Sinn Féin/Green Party –44 –33 –32
Others –22 –23 –24
(Very unsatisfied with govt) (–18)
Quite unsatisfied with govt +18
Quite/very satisfied with govt +17
Attitude effects
(Anti-integration) (–23) (–20) (–20)
Centrist +9 +1 0
Pro-integration +33 +26 +26
(Anti-enlargement) (–19) (–50) (–47)
Centrist +7 +43 +42
Pro-enlargement +25 +51 +48
(Strengthen neutrality) (–20) (–15) (–13)
Centrist +14 +14 +12
Limit neutrality +29 +15 +14
Note: Table entries indicate that in 2001, for example, Fine Gael/Labour voters were 21 percentage
points less likely to vote ‘Yes’ than were Fianna Fáil (FF) voters, other things being equal. In
contrast, FF voters were 25 percentage points more likely to vote ‘Yes’ than were non-FF voters.
Percentages in italics are not significant at the .05 level.
03 Garry (to/d) 12/4/05 11:26 am Page 214
the significance of the issue of EU enlargement. Does this increase in the
impact of attitude factors explain why the result of the referendum was
different at Nice 2, or do we also need to take into account changes in the
distribution of opinions on the individual second-order and attitude factors?
We address this question by running two counterfactual simulations. First,
we can look at Nice 1 under the conditions that (the distributions of the variables
relating to) party support and attitudes were as they became at Nice 2
but keeping the Nice 1 model estimates of the impact of those conditions. The
predicted vote using that model would not have been much different from
the actual outcome at Nice 1: 48% ‘Yes’ instead of 46%. Second, we can look
at Nice 2 under the conditions that party support and attitudes were as they
were at Nice 1 but keeping the Nice 2 model estimates of the impact of those
conditions. The predicted vote is, again, only slightly different: 61% ‘Yes’
instead of 63%.10 We can conclude from this that the different referendum
result the second time around was not a function of changing marginals – i.e.
a change in the distributions on party support and the issue variables. Rather,
the result of the referendum changed because of the greater impact of the attitudinal
variables. Even so, what was vital here was the changing impact of
the different attitude factors, most notably the increase in the strength of the
relationship between attitudes to EU enlargement and voting behaviour (and
also the somewhat diminished impact of the issue of neutrality).
Discussion and conclusion
Referendums on the issue of the new EU constitution are pending in a
substantial number of EU member states. Is voting behaviour in these referendums
likely to reflect citizens’ thoughts on the future of the EU and on the
new constitution, or is it likely simply to reflect citizens’ concerns about
domestic party politics and their views on incumbent national governments?
Our analysis of voting in the Irish Nice referendums suggests that, although
the effect of satisfaction/dissatisfaction with the incumbent government (i.e.
the second-order effect) is detectable, it played a much smaller role in determining
the outcome compared with the effect of attitudes to a range of
European issues. In short, both Irish referendums on the Nice Treaty were
closer to being processes of deliberation on EU issues than to being plebiscites
on the incumbent government.
The Irish experience with referendums on the Nice Treaty also suggests
that the more vigorous the campaign, the greater the effect of the key
substantive issue relating to the referendum – in this case attitudes to EU
enlargement – and the less the effect of second-order considerations. This is
Garry, Marsh and Sinnott ‘Second-order’ versus ‘Issue-voting’ Effects 2 1 5
03 Garry (to/d) 12/4/05 11:26 am Page 215
quite an unusual conclusion in that, in relation to general elections,
campaigns are typically viewed as much less important than ‘long-term’
determinants of voting such as social structure, party identification and
ideology. It may be, however, that there is much more scope for a vigorous
campaign to have an impact in referendums. All the main Irish parties – the
pro-EU governing Fianna Fáil and Progressive Democrats and the pro-EU
opposition Labour and Fine Gael parties – were shaken and stirred by the
‘No’ at Nice 1. Only the small anti-EU Sinn Féin and Green parties were
pleased with the result. Rejection of the Nice Treaty in the first referendum
was a serious and unexpected blow to the Irish political establishment.
Instead of the brief and limp campaign at Nice 1, the main parties, and
particularly their civil society allies, delivered a vigorous, spirited and
lengthy campaign at Nice 2. Strategically, they sought to ‘de-couple’ two
dangerous issues from the issue overtly at stake in the referendum. First,
they sought – through the ‘Declarations’ and the constitutional provision
mentioned above – to convince ‘anti-military alliance’ voters that the Nice
Treaty was not in fact anything to be afraid of. The pro-camp also sought to
defuse and neutralize the issue of domestic party politics. Thus they
campaigned vigorously to persuade citizens that the treaty was about the
future of Europe – and enlargement in particular – and not a referendum on
the popularity of the Irish government. The pro-European opposition parties
(Labour and Fine Gael) were particularly important in attempting to limit
the impact of second-order effects on voting behaviour in the referendum.
The fact that satisfaction with the government was significantly lower at Nice
2 than at Nice 1 does not seem to have negatively impacted on the result.
Crucially, the effective campaigning of the ‘Yes’ side made the issue of EU
enlargement the key issue in the Nice 2 campaign.
The implications for member state governments facing referendums on
the EU constitution can be briefly stated. Government satisfaction levels and
support for the domestic political parties are likely to play some role in determining
the outcome, but this role is presumably much smaller than the role
played by ‘issue effects’, that is by attitudes to European integration and to
issues arising from the new constitution. However, the extent of the impact
of such European attitudes on the outcome depends on the vigour and the
effectiveness of the referendum campaign.
Notes
1 Referendums are unlikely to be conducted in the Slovak Republic, Hungary
or Finland and will almost certainly not be conducted in Austria, Germany,
European Union Politics 6(2)2 1 6
03 Garry (to/d) 12/4/05 11:26 am Page 216
Greece, Italy, Sweden, Cyprus, the Czech Republic or Malta. For an
overview of the probable incidence of referendums, see URL: http://www.
euractiv.com/cgi-bin/cgint.exe?204&OIDN=500751&-tt=fucr.
2 The main effect of the treaty was to make changes in the institutional structures
and procedures of the EU with a view to the impending enlargement
of its membership. The voting proportions were 53.9% ‘No’ to 46.1% ‘Yes’ in
the 7 June 2001 referendum on a turnout of 34.8% and 62.9% ‘Yes’ to 37.1%
‘No’ on a turnout of 49.5% in the 18 October 2002 referendum. In our 2001
survey, ‘Yes’ voters are underrepresented (40.5%) and, in our 2002 survey,
‘Yes’ voters are overrepresented (74.2%). We thus weight our survey data so
that they are representative of the actual results.
3 What we call the ‘issues’ and ‘second-order’ interpretations of voting in EU
referendums cover two of the main schools of thought on the topic of voting
in EU referendums. Ehin (2001) categorizes existing explanations into a threefold
typology: values (or ‘attitudes’), ‘domestic politics’ (or ‘second-order’
national election effects) and ‘utilitarian expectations’. The rational economic
actor model (‘utilitarian expectations’ model) holds that EU integration differentially
benefits certain segments of a national population and that individuals
who believe they will benefit economically will vote ‘Yes’ in an EU
referendum and people who believe they will not benefit economically will
vote ‘No’ (Gabel and Palmer, 1995; Anderson and Reichert, 1996; Gabel,
1998a, 1998b). These studies use class, sex and education as proxies for individual
competitiveness, which they hypothesize is related to support for the
EU (richer, more educated men are better able to exploit economic opportunities
in the liberal EU marketplace and thus support a pro-EU position in
referendums).
4 The surveys used are those described in the ‘data’ section below. The question
wording in the surveys was as follows: ‘There are many different ways in
which people get information in relation to referendums. I have a list here of
several possible sources of information. Please say how useful, if at all, you
found each of them in the lead up to the [. . .] referendum on the Nice Treaty
[. . .] Using this card, would you say you found each of the sources mentioned
very valuable, somewhat valuable, of little or no value or did you not notice
or come across the source in question at all?’
5 The wording of the question was: ‘By the date of the referendum [state date]
how good was your understanding of the issues involved?’ The response
options were: I had a good understanding of what the Treaty was all about;
I understood some of the issues but not all that was involved; I was only
vaguely aware of the issues involved; I did not know what the Treaty was
about at all.
6 See ‘The Nice Treaty: Explaining the Issues – Irish Neutrality and European
Security’, Irish Institute of European Affairs, URL: http://www.iiea.com/
files/nice/nice8.pdf (pp. 3–4).
7 The surveys were carried out under the framework contract (Flash Eurobarometer)
with EOS Gallup Europe on behalf of the Eurobarometer Unit in
the Directorate General for Press and Communications of the European
Commission. Fieldwork for the 2001 survey was conducted between 20 August
and 10 September 2001 among a quota sample of 1245 adults. Fieldwork for
Garry, Marsh and Sinnott ‘Second-order’ versus ‘Issue-voting’ Effects 2 1 7
03 Garry (to/d) 12/4/05 11:26 am Page 217
the second survey was conducted between 15 November and 9 December 2002
among a quota sample of 1203 adults. For replication purposes, the survey data
on which the analysis presented in this paper is based are available at URL:
http://www.ucd.ie/issda/dataset-info/nicedata.htm
8 There were some slight differences in the question wordings between the Nice
1 and Nice 2 surveys. Because of the ongoing nature of the enlargement
negotiations, it was necessary to change the wording of the question in 2002:
‘The European Union is at present finishing negotiations with 10 countries
about joining the European Union in 2004. We are interested in how people
feel about this enlargement of the EU. In general terms, are you in favour or
against this enlargement of the EU?’ There is no reason to believe that this
shortened introductory wording rendered the results of the enlargement
questions in 2001 and 2002 non-comparable. The over-time trends from the
2001 and 2002 surveys are very similar to those found by Eurobarometer 57
and Eurobarometer 58 from a standard enlargement question (see Sinnott, 2003:
8–9). The party preference question was also slightly different in the two
surveys. In the Nice 1 survey, respondents were asked which party they
would vote for if there was a general election tomorrow, and in Nice 2 respondents
were asked to indicate the party that they ‘usually support’. The breakdown
of responses to both questions is similar at both time points (see online
appendix). Note that the wording of the ‘satisfaction’ question was: ‘Overall
are you generally satisfied or dissatisfied with the way the government is
running the country?’ The response options were: very satisfied; quite
satisfied; quite dissatisfied; very dissatisfied. Finally, in both surveys, the age
of the respondent was assigned to one or the other of five categories, and
social class was measured using the usual market research AB, C1, C2, DE
and F categories. The gender of the respondent was also categorized.
9 There is much discussion in the literature about the possible direction of the
causal relationship between our two sets of independent variables: support
for the incumbent government/ governing parties and attitudes to the EU.
Anderson (1998), Gabel (1998c) and Ray (2003) argue that attitudes to the EU
to some extent mediate the effect of attitudes to the incumbent government;
in other words, whether or not you support the incumbent government determines
whether you have positive attitudes to the EU, which in turn determines
whether you support an EU treaty in a referendum. Thus, in the
analysis reported in Tables 2 and 3, we include our variables in that order –
‘second-order’ effects first followed by ‘issue’ effects. (However, when we ran
a model regressing EU support on party and demographic variables, we
found no significant association between party support and EU support. Also,
when we ran a model regressing support for enlargement on party and demographic
variables, we found no significant association between party support
and enlargement support. Thus, there appear to be no significant endogeneity
effects and it does not seem that attitude effects are mediating party
support effects. This is in line with the general finding in Irish data that voters
tend not to distinguish between the main parties on EU issues.)
10 We do not include the ‘government satisfaction’ variable in our operationalization
of second-order effects in either of the two simulations described in
this paragraph. This is because, as noted, the question concerning
European Union Politics 6(2)2 1 8
03 Garry (to/d) 12/4/05 11:26 am Page 218
‘government satisfaction’ was asked only in the 2002 survey. However, if we
were to include this (2002) variable in our simulation of what the result would
have been in 2001 if the conditions of 2002 had pertained at 2001, then almost
certainly our simulated/predicted result would have been lower than 48%.
This is because contemporary opinion poll data suggest that government
satisfaction levels were much lower in 2002 than they had been in 2001:
satisfaction with the government dropped from 59% in May 2001 to 33% in
October 2002 (see Table A at URL: http://www.tnsmrbi.ie/MSWord/
IRISH%20TIMES%20TNS%20MRBI%20POLL%20POLITICAL%20ISSUES.d
oc). Thus, the anti-government side of the Nice debate would have benefited
if the (low) satisfaction levels of 2002 had existed in 2001.
References
Aardal, Bernt, Anders Todal Jenssen, Henrik Oscarsson, Risto Sankiaho and Erika
Saynassalo (1998) ‘Can Ideology Explain the EU Vote?’, in Anders Todal
Jenssen, Perti Pesonen and Mikael Gilljam (eds) To Join or Not to Join. Oslo:
Scandinavian University Press.
Anderson, Christopher (1998) ‘When in Doubt, Use Proxies: Attitudes toward
Domestic Politics and Support for the EU’, Comparative Political Studies 31(5):
569–601.
Anderson Christopher and M. Shawn Reichert (1996) ‘Economic Benefits and
Support for Membership in the European Union: A Cross-National Analysis’,
Journal of Public Policy 15(3): 231–49.
Ehin, Piret (2001) ‘Determinants of Public Support for EU Membership: Data from
the Baltic Countries’, European Journal of Political Research 40: 31–56.
Ferrara, Federico and J. Timo Weishaupt (2004) ‘Get Your Act Together: Party
Performance in European Parliament Elections’, European Union Politics 5(3):
283–306.
Franklin, Mark (2002) ‘Learning from the Danish Case: A Comment on Palle
Svensson’s Critique of the Franklin Thesis’, European Journal of Political Research
41(6): 751–7.
Franklin, Mark, Michael Marsh and Lauren McLaren (1994) ‘Uncorking the Bottle:
Popular Opposition to European Unification in the Wake of Maastricht’, Journal
of Common Market Studies 32(4): 455–72.
Franklin, Mark, Michael Marsh and Christopher Wlezien (1994) ‘Attitudes
towards Europe and Referendum Votes: A Response to Siune and Svensson’,
Electoral Studies 13: 117–21.
Franklin, Mark, Cees van der Eijk and Michael Marsh (1995) ‘Referendum
Outcomes and Trust in Government: Public Support for Europe in the Wake
of Maastricht’, West European Politics 18(3): 101–7.
Gabel, Mathew (1998a) Interests and Integration: Market Liberalization, Public
Opinion, and European Union. Ann Arbor: University of Michigan Press.
Gabel, Mathew (1998b) ‘Economic Integration and Mass Politics: Market Liberalisation
and Public Attitudes in the European Union’, American Journal of Political
Science 42(3): 936–53.
Garry, Marsh and Sinnott ‘Second-order’ versus ‘Issue-voting’ Effects 2 1 9
03 Garry (to/d) 12/4/05 11:26 am Page 219
Gabel, Mathew (1998c) ‘Public Support for European Integration: An Empirical
Test of Five Theories’, Journal of Politics 60(2): 333–54.
Gabel, Mathew and Henry Palmer (1995) ‘Understanding Variation in Public
Support for European Integration’, European Journal of Political Research 27(1):
3–19.
Gilland, Karin (2002) ‘Ireland’s (First) Referendum on the Treaty of Nice’, Journal
of Common Market Studies 40: 527–35.
Gilland, Karin (2003) ‘Ireland at the Polls: Thirty Years of European-Related Referendums,
1972–2002’, presented at the York Referendum Challenge Forum,
York, 15 February 2003; URL: http://www.iri-europe.org/reports/
York-KarinGilland.ppt.
Hayward, Katy (2002) ‘Not a Nice Surprise: An analysis of the Debate on the Nice
Treaty in the Republic of Ireland’, Irish Studies in International Affairs 13: 167–86.
Hayward, Katy (2003) ‘ “If at First You Don’t Succeed. . .”: The Second Referendum
on the Treaty of Nice, 2002’, Irish Political Studies 18(1): 120–32.
Hug, Simon (2002) Voices of Europe: Citizens, Referendums, and European Integration.
Lanham, MD: Rowman & Littlefield.
Hug, Simon and Pascal Sciarini (2000) ‘Referendums on European Integration: Do
Institutions Matter in the Voter’s Decision?’, Comparative Political Studies 33(1):
3–36.
Marsh, Michael (1998) ‘Testing the Second-Order Election Model after Four
European Elections’, British Journal of Political Science 28: 591–607.
O’Brennan, John (2003) ‘Ireland’s Return to “Normal” Voting Patterns on EU
Issues: The 2002 Nice Treaty Referendum’, European Political Science
12(2); URL: http://www.essex.ac.uk/ECPR/publications/eps/onlineissues/
spring2003/feature.htm.
O’Mahoney, Jane (2001) ‘Not So Nice’, Irish Political Studies 16: 201–15.
Pierce, Roy, Henry Valen and Ola Listhaug (1983) ‘Referendum Voting Behaviour:
The Norwegian and British Referenda on Membership in the European
Community’, American Journal of Political Science 27(1): 43–63.
Ray, Leonard (2003) ‘Reconsidering the Link between Incumbent Support and ProEU
Opinion’, European Union Politics 4(3): 259–79.
Reif, Karlheinz and Hermann Schmitt (1980) ‘Nine Second Order National
Elections: A Conceptual Framework for the Analysis of European Election
Results’, European Journal of Political Research 8(1): 3–44.
Schneider, Gerald and Patricia Weitsman (1996) ‘The Punishment Trap: Integration
Referendums as Popularity Contests’, Comparative Political Studies 28(4):
582–607.
Sinnott, Richard (2001) Attitudes and Behaviour of the Irish Electorate in the
Referendum on the Treaty of Nice; URL: http://www.ucd.ie/dempart/
workingpapers/nice1.pdf.
Sinnott, Richard (2002) ‘Cleavages, Parties and Referendums: Relationships
between Representative and Direct Democracy in the Republic of Ireland’,
European Journal of Political Research 41: 811–26.
Sinnott, Richard (2003) Attitudes and Behaviour of the Irish Electorate in the
Second Referendum on the Treaty of Nice, URL: http://www.ucd.ie/dempart/
workingpapers/nice2.pdf.
Siune, Karen and Palle Svensson (1993) ‘The Danes and the Maastricht Treaty: The
Danish EC Referendum of June 1992’, Electoral Studies 12(2): 99–111.
European Union Politics 6(2)2 2 0
03 Garry (to/d) 12/4/05 11:26 am Page 220
Siune, Karen, Palle Svensson and Ola Tonsgaard (1994) ‘The European Union: The
Danes Said “No” in 1992 but “Yes” in 1993 – How and Why?’, Electoral Studies
13(2): 107–16.
Svensson, Palle (1994) ‘The Danish Yes to Maastricht and Edinburgh: The EC
Referendum of May 1993’, Scandinavian Political Studies 17(1): 69–82.
Svensson, Palle (2002) ‘Five Danish Referendums on the European Community
and Union: A Critical Assessment of the Franklin Thesis’, European Journal of
Political Research 41(6): 733–50.
Van der Eijk, Cees, Mark Franklin and Michael Marsh (1996) ‘What Voters Teach
Us about Europe-Wide Elections: What Europe-Wide Elections Teach Us about
Voters’, Electoral Studies 15(2): 149–66.
About the authors
John Garry is Lecturer in Comparative Political Science, Department of
Politics and International Studies, Queens’ University Belfast, 19–21
University Square, Belfast BT7 1PA, Northern Ireland, UK.
Fax: +44 28 9023 5373
E-mail: j.garry@qub.ac.uk
Michael Marsh is Associate Professor and Head of the Department of
Political Science, Trinity College Dublin, 1 College Green, Dublin 2,
Republic of Ireland.
Fax: +353 1 677 0546
E-mail: mmarsh@tcd.ie
Richard Sinnott is Associate Professor in Politics, Department of
Politics, University College Dublin, Belfield, Dublin 4, Republic of
Ireland.
Fax: +353 1 716 1171
E-mail: Richard.Sinnott@ucd.ie
Garry, Marsh and Sinnott ‘Second-order’ versus ‘Issue-voting’ Effects 2 2 1
03 Garry (to/d) 12/4/05 11:26 am Page 221