Voters in a referendum obtain information and derive voting cues from a variety of sources. Some of these, such as political parties or ideological orientations, are similar to those also found to be influential in elections. Others can be quite different. In some referendums, the issue may be entirely new and unfamiliar to many voters, initiating a ‘learning’ or ‘cue–taking’ process specific to the campaign itself. In referendum campaigns, parties may be internally divided and sometimes send conflicting signals to their electorates. As a result, voting behaviour in referendums often exhibits greater volatility than is found in elections. In the ten papers included in this Special Issue of EJPR, we focus on the process of opinion formation and change which occurred in a number of European, North American and Australia/New Zealand referendums held under a variety of different institutional and political conditions. In this essay, I argue that there are three distinctive patterns of opinion formation and reversal that tend to occur in referendum campaigns, each of which has significant consequences both for voting choice and for referendum outcomes.

Denmark had five referendums in the period from 1972 to 1998 dealing with Danish membership in the European Community, the Single European Act, the Maastricht Treaty, the Edinburgh Agreement and the Amsterdam Treaty. Did the Danes really address these issues and involve themselves actively in the policy–making process on a vital issue or did they merely vote for or against the current government? The latter option represents the ‘second order’ elections argument advanced by Mark Franklin and others (see Franklin's article in this issue). If correct in this instance, it may have important and negative consequences for the potential of referendums to involve citizens more directly in the way they are governed. In this article, the Franklin thesis is assessed on the basis of data on voting behaviour in five Danish referendums on Europe and the democratic implications of these findings are discussed.

Palle Svensson in this issue of EJPR has objected to the characterisation of Danish voters made by Franklin and others who, in various publications, expounded the thesis that on issues of low salience, referendum votes tend to follow party lines. Svensson finds evidence that the Maastricht Treaty was not an issue of low salience to Danish voters in the ratification referendums conducted there, and gives other details of the evolution of public opinion regarding Europe that clarify the circumstances in which our thesis should apply. In the light of his arguments, this Comment presents a more nuanced version of the thesis that learns from the Danish case, and should be of greater utility than our earlier version in helping to interpret the role of government standing in referendum outcomes.

Referendums impose considerable informational demands on voters. Recent theoretical and empirical research has emphasized the different shortcuts and heuristics they may employ in deciding how to vote. Relying on a substantial series of votes at the national level in Switzerland, we provide empirical tests on how Swiss voters cope with the informational demands in referendum voting. We combine simple heuristics, like partisan cues and endorsements, with indicators of instrumental interests to explain citizens’ choices in a series of votes.

In this article we examine the scope for campaign effects in citizen-initiated referendum (CIR) elections. Given the context of CIR elections, television effects can be seen to be even bigger and more important than in candidate elections. We use survey data on information demands made upon voters in CIR campaigns, and the information sources they use, in order to gauge the relative importance of various sources that voters rely upon in making voting decisions. We then examine the relative importance of television advertising as a source of information in CIR campaigns. We find that voters report using many sources of information, with few voters relying exclusively upon television advertisements. Rather than telling voters which way to vote, television campaigns may simply raise awareness of CIRs and so encourage voters to seek cues elsewhere, in particular from ballot guides where cues are more readily discerned.

In a two–stage referendum process, in 1992 and 1993, New Zealand voters authorised a radical change from a plurality electoral system to proportional representation. Although ad hoc groups, not the major political parties, led the debate for and against the change, the nature of the issue ensured that partisanship had a strong effect on voters’ choices. Yet many voters were also guided by their political principles and values, including fairness of representation on one side, and preference for single–party, decisive governments on the other. Experience of proportional representation since 1996 has resulted in a fairly high level of uncertainty toward electoral change, independent of partisanship. Values and partisanship, however, remain the two stabilising influences.

In the Irish case, a combination of quite restrictive constitutional provisions for consulting the people and extensive social and political change has produced a variety of forms of initiation of referendums and a surprisingly large number of them (25 since the late 1960s). Referendums in Ireland have frequently been related to underlying political cleavages while only very infrequently evoking major inter–party conflict. This suggests that referendums may have insulated the party system against the impact of potentially contentious issues. However, analysis of voting behaviour in the 1992 general election and abortion referendums indicates some linkage between referendum voting and party choice at the aggregate level. Additionally, evidence from the Nice Treaty referendum of 1901 suggests that the lack of major party involvement in the campaign contributed to a demonstrable gap in people's understanding of the issues. The low level of understanding in turn certainly contributed to the massive level of abstention and to a NO vote that was contrary to the preferences of almost all the parties. Thus, there are limits to the extent to which the organs of representative democracy can, or can afford to, distance themselves from the process of direct democracy.

The rarity of referendums in the United Kingdom makes them unusual events for voters, parties and students of voting behaviour. The 1997 devolution referendums in Scotland and Wales were highly significant since they led to major constitutional changes. In the novel circumstances of a referendum, political parties and the media played an important role in providing voters with information to guide their voting decisions. Longstanding identities and opinions relating to the pros and cons of the issue involved and attitudes towards the government influenced voting. Reassuringly, opinions about the effects of devolution were highly influential.

Although a majority of Australian voters favour the introduction of a republic, in the November 1999 referendum a majority of them nevertheless voted to retain the monarchy. This article explains the background to this apparently perverse outcome. The central problem was asking the Australian electorate to make a complex, technical choice about the system of government, in the absence of clear partisan cues. As a result, republicans were divided on the method of election for the head of state, effectively resulting in three separate groups of voters favouring change. Using survey data collected just after the referendum, four hypotheses are tested to explain the result. The most important influence on voting was views about whether or not to sever the link with Britain, followed by the positive and negative aspects of the proposed change, and the cues presented by the leaders of the respective YES and NO campaigns. The interaction between these factors enabled political elites opposed to change to manipulate the result in their favour.

National referendums progressively became a part of Italian political life in the last three decades of the twentieth century. The increased use made of them came as a genuine surprise. Some 53 issues were put before the electorate between 1970 and 2000, the majority following citizen petitions, with a smaller number promoted by regional councils. The so–called ‘abrogative referendum,’ conceived as a sort of modern–day veto, proved to be a multi–use political tool wielded predominantly by the Radical Party, a political movement largely outside the Italian party system. Parties were the main actors in referendums as far as the mobilisation of voters was concerned. Indeed, the parties gave birth to a variety of alignments, some supporting mobilisation and competition, while others encouraged abstension and effective demobilisation. By virtue mainly of the constitutional rule that requires 50 per cent turnout for the result to be valid, the vote on 18 issues was declared void. Many lessons can be learned from the Italian experience, and one in particular is of special interest to all who value the survival of liberal democracies: all electors are equal, but in referendums non–voters are more equal than the voters.