Resumé

In this paper elements of a theory of electoral behaviour in referendums are presented and tested on a theoretically important case. Among the main determining factors — social class, political party and ideological attitude — it is shown that on a relatively simple issue with low importance for the everyday life of the voters, ideological attitude is far more important than political party and social class in explaining referendum voting. On this basis a scheme of analysis and a number of hypotheses are suggested for further study.

Introduction

Electoral behaviour is most often studied in terms of party choice. However, the voters do not only participate in elections by choosing among parties and candidates, from time to time they are also asked to decide on a specific issue directly in a referendum. During the last decades referendums¹ have found an increasing application in Western European democracies (Butler & Ranney 1978, tables 1-3). During the last decade they have been widely applied on the issue of the enlargement of the European Economic Community; in Greece and Spain they have played a part in the process of reestablishing constitutional democracy; in Ireland they have been used on such various issues as EEC, voting age, religion, adoption, and senate representation; and in Britain they have, in addition, been used on a regional basis on the issues of Scottish and Welsh devolution (Butler & Ranney 1978, 3-21; Balsom & McAllister 1979, 394-409; Coakley 1981, 15-17; and Jones & Wilford 1982, 16-27).

Moreover, a further increase in the application of referendums in the future
can be expected, because a significant shift is taking place in the distribution
of political skills and resources. 'An increasingly large proportion of the public

---

* This is a revised version of a paper prepared for the Workshop on 'Electoral Behaviour', ECPR Joint Sessions of Workshops, Freiburg 20 March - 25 March 1983.

is beginning to have sufficient interest and understanding of national and international politics to participate in decision-making at this level' (Inglehart 1977, 3). David Butler and Austin Ranney conclude their study of referendums with the following statement: 'One thing is clear: referendums are here to stay; and they are almost certain to increase in number and importance in the years ahead' (Butler & Ranney 1978, 226).

Despite the growing number of referendums in Western European democracies and the obvious relevance of obtaining scientific knowledge of electoral behaviour in referendums, there seems to have been relatively little study of electoral behaviour in nationwide referendums (Butler & Ranney 1978, 15). Only a few studies of this kind of electoral behaviour have been published and they have mainly dealt with the EEC issue in Norway and Denmark (Petersen & Elklit 1973; Valen 1973; Elklit et al. 1974; Valen et al. 1975; Valen 1976; and Petersen 1978).²

The aims of this paper are firstly to draw attention to this area of electoral behaviour and to promote a discussion about the state of research in this area, secondly to present a hypothesis on electoral behaviour in referendums and to put this hypothesis to an empirical test, and finally to discuss briefly some theoretical implications of the findings.

Towards a Theory of Electoral Behaviour in Referendums

A theory of electoral behaviour in referendums should be based on the following elements: How can the behaviour of the voters be explained? What determines — or perhaps less rigorously formulated — what influences the electoral decisions of the voters? What factors have an influence, and how strong is the influence of one factor in relation to other factors? What are the conditions for the influence of particular factors?

These questions are, of course, of a very abstract nature. More specifically, it is of interest to ask about the influence of reference groups such as social class and political party compared with the influence of the political ideology or attitudes of the voters. To what extent do the voters act in referendums on the basis of their own judgement of the issue, and to what extent do they take cues from the social class to which they belong or from the political party which they generally support?

Socio-economic or occupational class is undoubtedly one of the main factors influencing electoral behaviour. It is a common assumption in political science that people occupied with different functions in the economic division of labour in society perceive and possess different interests; at least in the Scandinavian countries social class is undoubtedly the strongest predictor of party choice (Worre 1980, 300). Consequently, it is reasonable to expect that

the social background of the voters — at least to a certain extent — influences
their electoral behaviour in referendums.

This expectation has been confirmed in a few studies on referendum voting. Thus, an ecological analysis of the 1972 Danish EEC referendum found a significant lower degree of support for the EEC in working class constituencies than in middle class constituencies; and local voting of the total electorate interviewed immediately after the official referendum in three different polling districts confirmed that employers (including farmers) were more in favour of the Danish entry into the EEC than employees, workers etc. (Petersen & Elklit 1973, 209-210). In his study of the 1972 Norwegian EEC referendum, Henry Valen applied a nationwide voter survey; he found that differences in the electoral behaviour varied along two well-established economic cleavages: an urban-rural conflict (a conflict in the commodity market) and a left-right conflict (a conflict in the labour market). Whereas farmers and fishermen were less likely than people in urban occupations to support the EEC membership, the workers were less likely than people in middle class positions, i.e. salaried employees and independents in business, to support Norwegian entry into the EEC (Valen 1973, 217).

However, Nikolaj Petersen and Jørgen Elklit's study on the Danish EEC referendum, like Henry Valen's on the Norwegian EEC referendum, emphasize that social class was not the only, and perhaps not even the most important, factor to influence the electoral behaviour in these referendums. Moreover, it has recently been argued that socio-economic class only accounts for a small part of the variation in support for different political parties (Dahl 1982, 210). If class is of little or declining importance with regard to party choice (Worre 1980, 313-317), it is quite possible that class may be of less importance with regard to referendums than indicated by the Scandinavian EEC referendums.³ The question remains therefore to what extent and under what circumstances class can explain the electoral behaviour in referendums.

After all, the party loyalty of the voters is perhaps more important than class as a determinant or main influence on voting behaviour in referendums. In the most recent study on referendum voting it is concluded that 'partisan attachment is almost surely the primary force in referendum voting' (Pierce et al. 1983, 61). Comparing the Norwegian and British referendums on the membership in the European Community, they found that partisanship was the main factor determining referendum voting, whereas basic ideological attitudes were found to be of secondary importance in this respect. Party preference also seems to be the most important factor influencing the Italian voters in the 1974 referendum on the divorce law (Marradi 1976, 131). In more general terms John Coakley has — after reviewing the scattered evidence on referendum voting in a number of Western countries⁴ — argued that voters' behaviour is heavily conditioned by the standpoint of their political party (Coakley 1981, 6-7 and 25-26): 'Perhaps because of the unfamiliarity of the

situation in which people find themselves ... they tend to act in accordance with the cues from the political parties' (ibid., 7). Nikolaj Petersen and Jørgen Elklit reach the same conclusion after a study of the 1972 Danish EEC referendum: 'Voting behaviour in Danish referenda depends primarily on the "ordinary" political affiliation of the voter. Findings by Bo Särlvik in Sweden point in the same direction. If the standpoints of the political parties on the issue in question are known, voters who identify more or less with a given party will be strongly influenced by the viewpoint of the party' (Petersen & Elklit 1973, 212).

According to Nikolaj Petersen it is mainly the clarity of the party standpoint which is of importance. In a study of the impact of attitudes about federalism and integration on the outcome of the 1972 Danish EEC referendum, he found considerable support for a hypothesis, which reads as follows: 'Party is usually a powerful predictor of voting behaviour. Probably in most political issues the political party acts as the primary reference group, depending, however, on the clarity of the signal which emanates from it: the less clear and unambiguous external clues are, the more the individual will be forced to act on internal clues, in casu federalism etc. We therefore expect that the clearer the party's policy stand is, the less impact attitudes towards federalism will have on the individual's vote, and conversely the more blurred the policy stand is, the more difference will it make' (Petersen 1978, 37).

This is, in my view, a valuable contribution to a theory of electoral behaviour in referendums, because one condition for the influence of political parties as reference groups is pointed out. This point is also stressed by Roy Pierce et al., 'When the parties' national elites took a clear and uniform position, their adherents followed in impressive proportions; when the national elites were divided, so were the voters' (Pierce et al. 1983, 43). Lack of party unity may, in addition, indicate one of the conditions under which the political attitudes of the voters may have an impact on referendum voting.

John Coakley acknowledges that not all evidence supports his view of political parties as the main factor influencing referendum voting, and he raises the question: 'Is it not possible that the direction of causation is the reverse? Could not the coincidence between party policy and supporters' attitudes be explained in terms of elite responsiveness to popular pressure, parties adopting a line they know will be acceptable to their followers? Or could it not be that in some instances ... congruence between the party machine's and party voters' views is not merely the product of interaction between the two but rather is a function of a third factor, such as an underlying Weltanschauung or "world view"?' (Coakley 1981, 8). In this context he points to the 1972 Norwegian EEC referendum as an important exception. This indicates that the conditions for party influence on the electoral behaviour in referendums need further specification, if a comprehensive theory is to be developed in this field.

While John Coakley is rather sceptical about the notion of 'rational' or 'autonomous' voters and generally rejects the voters' capacity to shed party ties and weigh issues on their merits, other researchers have argued that — perhaps under specific circumstances — voters can behave 'rationally' according to a consistent value system or a political ideology. Concerning referendums, David Butler and Austin Ranney state: 'The question referred to the voters is not usually expressed as a party issue but is presented as a problem for each to judge independently. The politicians and the media may give the voters plenty of cues about the answer they should reach, but parties are at their peril if they invoke party loyalty too ostentatiously' (Butler & Ranney 1978, 19-20). The question is precisely under what circumstances a political issue is not 'a party issue', so that the voters will act according to their own political attitudes, i.e. disregarding the cues from their social class, their party or other reference groups.

Elements of a theory about the conditions under which social class, political party and individual political ideology explain the electoral behaviour in referendums have been put forward by Ole Tonsgaard in a study on the 1972 Danish EEC referendum (Tonsgaard 1974).

Ole Tonsgaard first refers to the saliency of a political issue, i.e. the position of this issue in the consciousness of the voters, as a relevant factor conditioning the influence of diverse reference groups on referendum voting. It is a rather tricky problem to measure issue saliency in a valid manner, but Tonsgaard assumes that 'we must expect that attitudes with low issue saliency are less stable than those with high issue saliency, i.e. opinion leaders and other changeagents will be more influential in the former than in the latter case' (Tonsgaard 1974, 102-103).

Whereas a low issue saliency presumably is a necessary condition for social class and political party to influence the voters, it may not be a sufficient condition. In order to determine the influence of reference groups such as class and party, Tonsgaard suggests a simple classification of political issues on the basis of two distinctions concerning the character of the referendum issue.

First, political issues may differ in respect to their complexity. The more complex an issue is, the more difficult will it be for the voters to take their own stand on the issue. This means that reference groups and opinion leaders can be supposed to have a relatively stronger influence on the electoral behaviour in a referendum on a more complex issue than in a referendum on a more simple issue. Tonsgaard thus sets up the hypothesis: 'The more complex a political issue is the more influential will the reference groups be' (ibid., 103). The second distinction offers an answer to the question about what kind of reference groups obtain an influence on a more complex issue.

Second, political issues may differ in respect to their importance to the

everyday life of the voters. If a political issue is perceived as relatively unimportant by the voters, they are likely to take their political party as their reference group. On the other hand, if the issue is perceived as highly important to the life of the voters, Tonsgaard argues that social class can be expected to be highly influential: 'Among other things class membership means having nearly the same position in the economic system as well as in general uniform living conditions; this results in homogeneous perceptions of interests and of social reality among the members of the same social class' (ibid., 104). The resulting hypothesis follows: The more a political issue is perceived as being of importance for everyday life the more homogeneous attitudes towards the proposed alternatives will be found within social classes' (ibid.).

Tonsgaard combines the two distinctions in the two-dimensional property space indicated by fig. 1 (above). Figure 1 specifies the conditions under which social class and political party are most likely to be the main reference groups, and the conditions under which the voters will act according to their own political attitudes.

In his paper Ole Tonsgaard demonstrates that the electoral behaviour on the EEC issue — which he characterises as a relatively complex and relatively unimportant issue — did follow the prediction derived from the hypothesis: that in the EEC referendum the voters would take the same stand as the party they voted for at the general election of 1971. Because of lack of survey data he was, however, unable to verify the hypothesis with regard to an issue with a high complexity and a high importance, such as the 1963 Danish Land Laws, and an issue with low complexity and low importance, such as the referendums on the voting age.

In 1979 a new referendum on the voting age took place in Denmark, and survey data from this referendum now make possible an empirical test of an issue with a low complexity and low importance for the everyday life of the voters.

An Empirical Test

Out of 14 nationwide referendums in Denmark no less than 5 have concerned the voting age, which means that Danish experiences with referendums to a large extent rest on the voting age issue (see Appendix I for a survey of Danish referendums).⁵

Universal suffrage for elections to the Danish Folketing was introduced by the Constitution of 1915, with a voting age of 25 years.6 Thereafter the voting age was unchanged until 1953. In that year a new Constitution was adopted by a referendum. At the same time the voters were given a choice between a lowering of the voting age to either 23 years or 21 years. Although the age of reaching majority was 21 years, a plurality of votes was cast in favour of 23 years.

Up to 1953 the voting age in Denmark was laid down directly in the Constitution. However, the Constitution of 1953 states that the age qualification for suffrage may be altered at any time by statute. A bill passed by the Folketing for this purpose shall receive the royal assent only when the provision for a new voting age has been put to a referendum. In the referendum, votes shall be cast for or against the bill. For the bill to be rejected, a majority of those voting in the referendum — but not less than 30 per cent of all those entitled to vote — must vote against the bill.

In 1961 a referendum passed a bill which lowered the voting age from 23
years to 21 years. The turnout was rather low at 37.3 per cent, but only 16.6 per
cent of the eligible voters cast their vote against the bill.

The age qualification was discussed again in 1968-69. With a view to preserve Scandinavian uniformity, the bourgeois coalition Government suggested that the voting age should be: lowered to 20 years. Meanwhile, when the Social Democrats agreed to lower the voting age to 18 years, a majority emerged in the Folketing for this change of the voting age. In the following referendum, however, a large majority rejected the bill: the turnout was 63.6 per cent and 49.8 per cent of the electorate voted for rejection.

In 1971 the bourgeois Government once again proposed a voting age of 20 years. This time the Folketing passed the proposal and the referendum took place together with the Folketing election in September 1971. With a high turnout at 86.2 per cent, the bill was passed, because more votes were cast in favour of the bill than against it.

Following the passing of a bill for a majority age of 18 years, the Radical Liberals in October 1976 put forward a bill that proposed lowering the voting age to 18 years. The bill was passed on 26 May 1978 by a large majority, when 131 members voted for lowering the voting age to 18 years and only 21 members voted against. The opponents were found mainly among the Liberals and the Conservatives, but they also included a few members of the Progress Party and

the Christian People's Party. Prime Minister Anker Jørgensen announced
shortly afterwards that a referendum on the issue would take place on 19
September.

When the Radical Liberals in November 1976 proposed lowering the voting age to 18 years, this proposal was by no means a popular demand. The Gallup Institute found, when asking a national sample how they would cast their vote if there was to be a referendum the next day on this matter, that only 30 per cent would support a voting age of 18 years. As shown in Table 1, there was hardly any change in public opinion during the next year, when the Folketing dealt with the bill. By the time the Folketing passed the bill in the early summer of 1978 there was still a clear majority against the proposed lowering of the voting age. The next three months, however, led to a total shift in public opinion. Information about the background for the proposal, discussion of the arguments presented for and against a lower voting age, and the deliberate campaign by the adherents all had a conclusive impact.

This remarkable change in public opinion within a few months indicates that the voters did not hold very stable opinions on the voting age issue, probably because this issue had a low saliency among most voters. On the one hand, this means that the necessary condition for social class and political party to act as reference groups seems to be fulfilled. On the other hand, the character of the issue was obviously relatively simple (whether or not the voting age should be lowered from 20 years to 18 years and some 150,000 young people should be given suffrage rights) and relatively unimportant to the everyday life of the voters. This means that we can derive from the hypothesis stated above that the low saliency of the issue was hardly sufficient for social class and political party to play a decisive role. On the contrary, our hypothesis leads to the expectation that they would play a minor role in relation to the ideological attitudes of the voters.

In the following this expectation will be tested empirically on the basis of nationwide survey data.⁷ First, the three factors under consideration, class, party and ideology, will be treated separately, then their particular influence will be disclosed.

To the extent that social or occupational class serves as a reference group which guides referendum voting, it is reasonable to assume that different locations in economic life encourage different perceptions of social change, and that such more general conceptions may influence the stand on a specific issue as the voting age. It could be assumed that wage-earners to a higher degree than self-employed (both employers and independents without employees) welcome changes of the stams quo when it comes to a more equal or broad distribution of rights and privileges. It is more uncertain how location in the job hierarchy among the wage-earners may be related to conceptions of social change. Individual privileges and educational background may, for instance, point in different directions.

The relationship between occupational class membership and electoral behaviour in the referendum on the voting age is shown in Table 2. It appears that blue and white collar workers to a higher degree than self-employed voted for lowering the voting age to 18 years. The difference between blue and white collar workers was primarily caused by the lower participation of the first group in the referendum. Among those who did cast their vote 62 per cent of the unskilled workers, 64 per cent of the skilled workers, 68 per cent of the lower white collar workers, and 62 per cent of the upper white collar workers or employees voted 'yes', whereas 42 per cent of the self-employed did the same. There is, in other words, a clear difference between the wage-earners (65 per cent 'yes') and the self-employed (42 per cent 'yes'). The question is, however, to what extent this difference is caused by the class background of the voters or by other factors closely related to social class, such as party preference.

Historically, the stands of the Danish political parties on the voting age issue have followed a traditional, left-right dimension: the parties on the left supporting and the parties on the right resisting a lowering of the voting age (Svensson 1978, 14-51).

The public debate in 1978 on the voting age issue was rather special in at
least two respects. First, only the supporters of lowering the voting age to

18 years were organised. Second, the campaign was by and large an effort on the part of the political elite to convince a reluctant electorate. Most of the political parties used their organisation to support the proposed lowering of the voting age, and none of the party organisations was engaged entirely in a campaign against lowering the voting age to 18 years. The few politicians who spoke against the proposal did so as individuals. Actually, resentment against lowering the voting age was mostly expressed by private citizens in letters to the editors of certain newspapers. During the first half of September, all political parties were given time to present their views on the issue on both radio and television. These broadcasts were predominantly and almost tediously in favour of lowering the voting age to 18 years. None of them was entirely against lowering, and opponents were only permitted to speak in a few of them.

During the 1978 campaign on the voting age the left and centre parties (disregarding a few national spokesmen of the Christian People's Party) were united and supported the lowering of the voting age to 18 years. The bourgeois parties on the other hand — the Liberals, the Conservatives, and the Progress Party — were split on the issue, and the signal coming from these parties was rather unclear. With this background one might expect a left-right pattern to exist in the electoral behaviour at the referendum.

The figures in Table 3, where the political parties have been ranked by decreasing percentages of 'yes' to 18 years voting age, confirm the existence of a left-right pattern. Three separate groups are easily identified. To the left the voters of the three socialist parties overwhelmingly voted for 18 years. In the centre a majority of the voters of the Justice Party, the Social Democratic Party and the Radical Liberal Party cast their votes in favour of the lower voting age. The last group is composed of the bourgeois parties whose voters rejected the bill in the referendum.

The only surprise is that the Christian People's Party and the Center Democrats — usually perceived as centre parties or center-right parties — are clearly located on the right wing. It should be noted in this context that the percentages for these parties are computed on a small number of respondents. However, a similar result was found in a local survey study (Elklit & Tonsgaard 1978, 57), which indicates that the leaders of these parties were more 'liberal' than their voters on the voting age issue. The same undoubtedly holds true for a number of Liberal and Conservative politicians. They were, to be sure, convinced by the arguments for lowering the voting age, but because of the attitudes of most of their followers they kept 'a low profile' in public. They voted for the bill in the Folketing, but they did not campaign actively on the 'yes' side up to the referendum. Thus, a reversed causation of the kind mentioned by John Coakley — from the voters towards the party leaders — did play a role in this case, at least on the right side of the party space. When the politicians of some of these parties publicly did advocate the lower voting age — as in the case of the Center Democrats and to some extent the Christian People's Party — many of their voters did not follow this party cue, but acted on their own conception of the issue; either by casting a 'no' in the referendum or by abstaining from participation.

For practical reasons the parties may in the following analysis be rearranged in the three party groups identified above. Table 4 clearly shows that support for the lower voting age declines remarkably from the left to the right in the party space. The difference in actual referendum voting was even larger than shown here, because participation in the referendum was higher among the voters on the left wing than among voters in the two other party groups. Actually, among the voters who did participate in the referendum no less than 94 per cent of the left wing voters, 69 per cent of the center voters, and only 32 per cent of the right wing voters cast a 'yes' in the referendum. The voters on the left wing supported the 18 year voting age about three times as much as the voters on the right wing!

Whereas loyalties were split on the right wing, and the bourgeois parties
gave no clear advice, there is reason to believe that in the centre and particularly

on the left both politicians and voters to a high degree shared the same democratic ideology or 'Weltanschauung'. The question to be dealt with now, is to what extent the voters held consistent attitudes on the voting age issue, and what impact such attitudes had on the actual behaviour in the referendum.

According to the hypothesis suggested by Ole Tonsgaard, the individual attitudes on the referendum issue — the political ideology of the voters — should have a decisive influence on electoral behaviour at a referendum on an issue such as the voting age. In fact, the political ideology of the voters should have an even stronger impact than social class membership or party loyalty in a referendum on a relatively simple and relatively unimportant issue as this one.

In order to test this prediction it is necessary to identify and measure the ideological attitudes of the voters on this issue. This could be done by combining the voters' opinions about the main arguments for and against lowering the voting age to 18 years. In the survey conducted immediately after the referendum the respondents were confronted with a number of items about these arguments. Four of these items are, in particular, suitable for the identification and measurement of the specific ideological attitudes on the voting age:

1) 'Do you think that the voting age and the majority age should be the
same in this country?'

2) 'Do you think that the 18-19 year-olds are politically less mature than the
electorate as a whole?'

3) 'Do you think that a turn to the left in Danish politics is a likely result of
giving the vote to the 18-19 year-olds?'

4) 'Do you think that in deciding the voting age in this country the voting
age in countries such as Sweden, Britain, and USA should be taken into
consideration?'

The responses on these four items have been combined into an additive index by taking an affirmative response to the first and the last items and a negative response to the second and third items as an expression of the most positive attitude towards lowering the voting age to 18 years.9

In Table 5 it is demonstrated how the responses have been regrouped in three main attitudes towards lowering the voting age to 18 years. The table indicates that many voters held fairly consistent attitudes on this subject. The index values 3 or 4 express three or four responses in the same direction, and the index value 2 expresses either three consistent and a single inconsistent response or two consistent and two 'empty' responses ('don't know' or 'no answer'). All these combinations revealing reasonably consistent conceptions and values comprise 63 per cent of the respondents. Only 37 per cent gave inconsistent and/or indifferent responses resulting in the index values 1 or 0.

As predicted from the hypothesis outlined above, Table 6 proves a strong relationship between ideological attitude and electoral behaviour in the referendum on the voting age. The 'yes' percentages increase from 6 per cent among-the voters with a negative attitude to 73 per cent among voters with a positive attitude. The participation in the referendum was lowest among the voters without a clear attitude. Among those who actually participated in the referendum only 8 per cent of the voters with a negative attitude, 58 per cent of the voters without a clear attitude, and no less than 95 per cent of the voters with a positive attitude voted 'yes' at the referendum. This is, indeed, a very strong indication of the voters' ability to act 'rationally' on their own judgement of the issue at hand. It is evident that the relationship between the specific ideological attitude of the voters and their referendum voting is much stronger than the relationship between the voters' party preference and their referendum voting.

It might be argued that the close relationship between electoral behaviour and attitude towards lowering the voting age is no valid explanation of the referendum behaviour. It might be argued that the ideological attitude does not explain the referendum behaviour because the attitudes identified are subsequent rationalisations of the previous behaviour. To this objection it should be said that a high degree of stability has been shown to exist between January and September 1978 in the public attitudes towards the central

arguments about lowering the voting age (Svensson 1981, pp. 30-35). Even if panel data, which would be necessary to elucidate this point fully, are not available, the absence of any substantial change of public attitudes contradicts the objection mentioned.

Even if this result may be taken as a verification of the hypothesis suggested by Ole Tonsgaard, a real verification demands a closer analysis. It could be argued that despite the close relationship between ideological attitude and referendum voting found in Table 6, social class and political party have nevertheless served as reference groups. It could be argued that social class and political party influence referendum voting in two ways: indirectly through an influence on the political attitudes of the voters and directly as reference groups. To what extent and under what conditions does this argument hold true?

If the social class membership of the voters served as a reference group in the referendum on the voting age, it should be possible to disclose substantial class differences among voters with the same ideological attitude towards lowering the voting age.

The figures in Table 7 show how many of the voters participating in the referendum cast a 'yes' vote f°r lowering the voting age to 18 years. Reading the figures horizontally the table shows that among voters with a consistent negative or positive attitude towards lowering the voting age, there were only small class differences in referendum voting between wage-earners and selfemployed. On the other hand, among the voters without such attitudes, among the inconsistent, neutral or indifferent voters who did bother to participate in the referendum, membership of a social class to some extent served as a reference group.lo This indicates that in this referendum social class first of all influenced the electoral behaviour indirectly by furthering specific ideological attitudes. To the extent that consistent attitudes were developed, membership of a social class only played a minor role as a reference group. The condition

for a direct influence on the electoral behaviour in a referendum on an issue
as the voting age seems to be the absence of consistent attitudes on the
referendum issue.

The next question is to what extent and under what conditions political parties directly influenced referendum voting. It is reasonable to assume that a strong relationship exists between the more general and specific ideological attitudes of the voters on the one hand and the party preference of the voters on the other. Whether it is the political parties which have an impact on the ideological attitudes of the voters, or whether the voters ('rationally') choose among the parties on the basis of their own perceptions and values, is of minor importance in this context. The political parties and their leaders undoubtedly influence the attitudes of their supporters, and in this way they may — indirectly — also influence their electoral behaviour in referendums. The question here is to what extent and under what conditions the parties have a direct influence, because the voters take cue from the party on the specific issue.

Reading the figures in Table 8 horizontally, the table shows that among voters with a consistent negative or positive attitude towards lowering the voting age there are virtually no party differences.¹¹ This means that on this issue voters with consistent attitudes were only to a limited extent influenced directly by their party preference. In fact, in this case the political parties did not serve as reference groups. The voters were certainly able to act on their own without having to take a cue from the political party they usually support. Among the voters without consistent attitudes on this issue, the political parties apparently did serve as reference groups. The party differences among these voters even tend to be larger than among all voters: voters on the left supporting the bill to a higher degree (96 per cent compared to 94 per cent) and voters on the right to a smaller degree (30 per cent compared to 32 per cent). The differences between the whole electorate and the voters without consistent attitudes may be small, but they emphasize the conditional nature of the relationship found earlier between political party and electoral behaviour in the referendum on the voting age. The ideological attitude of the voters on

the referendum issue specifies the conditions under which the original relationships
found are strengthened or weakened (Rosenberg 1965, 106).

Finally, in order to give an over-all picture of the causal relationships between social class, political party and ideological attitudes as independent variables, and electoral behaviour in the referendum as the dependent variable, and in order to determine the magnitude of the direct and indirect influence of the independent variables, multiple regression and path analysis will be applied.l2

The main results of the path analysis are presented in Figure 2. The numbers in this figure are standardized regression coefficients which estimate the strength of the relationships between the variables. The arrows represent assumed causal linkages or paths of causal influence. The direction of causation is rather obvious. Logically there could only be a reversed relationship between political party and ideological attitudes, which, however, seems quite unlikely because of the specific meaning of the ideological attitude towards lowering the voting age. The arrows pointing to 'electoral behaviour' in the figure indicate the direct influence of the independent variables controlling for the influence of the other variables.

The path analysis combines the results of the bivariate and trivariate analysis ₁₃ Figure 2 and Table 9 show that social class membership has almost no direct influence on the electoral behaviour in the referendum. The impact of social class is in this referendum mainly indirect in its importance for the party preference of the voters. Political party has an impact on the electoral behaviour in the referendum which is evenly divided between a direct influence where the party of the voters serves as a reference group, and an indirect influence through the ideological attitudes of the voters. Most important, Figure 2 and Table 9 strongly demonstrate that the electoral behaviour in the referendum was predominantly determined directly by the voters' own ideological attitudes on the issue.

In summary, it can be said that the hypothesis suggested by Ole Tonsgaard has in general been confirmed. At the referendum on the voting age, i.e. a relatively simple and relatively unimportant issue, the voters first of all acted on their own opinions about the arguments for and against lowering the voting

age. Political parties served as reference groups for the voters to some extent, whereas social class was without any function of this kind. To this general confirmation of the hypothesis one important modification or specification should be added: a conditional relationship has been disclosed between social class and political party on the one hand and referendum voting on the other. When the voters had consistent attitudes on the specific issue of the voting age, social class and political party had virtually no impact on the electoral behaviour at the referendum. It was only when the voters lacked such consistent attitudes that, in particular, political parties served as reference groups for the voters.

Conclusion

In this paper a few groping steps have been taken towards the development of a theory of electoral behaviour in referendums. A hypothesis has been stated and empirically tested on a particular or perhaps even peculiar, but nevertheless theoretically important case. Whereas it seems to be too early to talk seriously about building a 'theory' or even a 'model', it is claimed that it is possible and of theoretical relevance to suggest a scheme of analysis by pointing out a number of dimensions to study, and to formulate some crude hypotheses for further empirical test and theoretical refinement. At least five dimensions have been indicated as deserving a closer examination if we want to explain or perhaps even to predict the electoral behaviour in a referendum (and consequently the outcome of that referendum):

1) How salient is the referendum issue to the voters? If the issue is of low
saliency, a necessary condition for social class and political party to serve
as reference groups is fulfilled. Conversely, if the issue is very salient, the

voters may rely on their own attitudes which presumably are more stable in
this case.

2) How complex is the referendum issue? If the issue is relatively difficult and complex, only a few voters could be expected to act on their own opinions and attitudes. In this case reference groups such as social class and political party will probably play an important role in influencing the referendum voting. If the issue, on the other hand, is of a more simple nature, a greater number of voters are likely to hold consistent attitudes on the subject and may need no help from reference groups.

3) How important is the referendum issue to the every day life of the voters? If the issue is relatively important and with potential consequences close to the voters, in particular if the issue concerns the immediate economic interests of the voters, social class rather than political party will be taken as a reference group. In this case the standpoints of interest organisations may play a decisive role for the electoral behaviour in the referendum. If the issue is, however, perceived as more remote by the voters, their party loyalty is more likely to guide their behaviour at the referendum.

4) How clear is the standpoint or the cue of the reference group in question? If the advice or the 'signal' coming from reference groups such as social class and political party is unclear, a large number of voters — everything else being equal — are likely to act on their own judgement of the referendum issue.

5) How consistent are the attitudes of the voters on the referendum issue? If the voters have consistent attitudes, their membership of a social class or their support for a political party is less likely to have considerable impact on their behaviour in the referendum, at least when the issue is fairly simple and unimportant. It is reasonable to assume that the number of voters with consistent attitudes varies with the degree of complexity of the issue (cf. point 2 above).

It is obvious that a number of difficult problems remain unsolved by this listing of relevant dimensions. Let me, finally, point to three problems that need further attention. First, the determinants of electoral behaviour treated in this paper by no means explain the whole variation in referendum voting (in the voting age case social class, political party, and ideological attitude explain a little more than half the variation, 57 per cent). This means that a search for important independent variables has to be continued. Second, there is the problem of identification and measurement of variables. To what extent is, for instance, the relatively large number of respondents with consistent attitudes on lowering the voting age purely a result of the operationalisation applied? More generally, it could be asked what criteria are valid in the measurement of the consistency of ideological attitudes. Finally, national

variations could be expected. The case study in this paper concerns a country without profound racial, ethnic, religious, linguistic or perhaps even without important regional cleavages. The political culture of Denmark is generally taken to be comparatively homogeneous. It is conceivable that the scheme of analysis and the hypotheses suggested in this paper are biased by this background. Only comparative research may clarify this question and bring us closer to a theory of electoral behaviour in referendums.

---

NOTES

1. Concerning terminology, I follow the highest authority, using the awkward 'referendums' instead of the more common 'referenda'. According to Butler & Ranney 1978, note 2, 4-5, the editor of the Oxford English Dictionary writes: 'Usage varies, even in high piaces, and both referendums and referenda are found in print. My own view is that referendum is logically preferable as a plural form meaning ballots on one issue (as a Latin gerund, referendum has no plural). The Latin plural gerundive referenda, meaning 'things to be referred', necessarily connotes a plurality of issues. A note to this effect is being added to our entry for the word in the forthcoming third volume of A Supplement to the Oxford English Dictionary. By preferring Referendums as your title you have the angels of Rome and of the O.E.D. on your side'.

2. The most notable exceptions, at least to my knowledge, are Coakley 1981 and Pierce et al. 1983. I have seen no study of the electoral behaviour in the 1980 Swedish Nuclear Power Referendum. The extensive British literature on the 1975 Common Market referendum contains very little on individual electoral behaviour; see, however, Butler & Kitzinger 1976, 252 ff; Särlvik et al. 1976, 103-112, and King 1977, 132 ff.

3. It has, however, been argued that it is misleading to speak of a decline in class voting, at least in Denmark. It is perhaps more correct to speak of a change in class voting, because some class subcultures are weakened, whereas others may be strengthened; see for this argument, Andersen 1983.

4. John Coakley refers, in addition to the Scandinavian countries, to Australian referendums in the 19705, French referendums from 1945 to the 19705, the 1974 Italian Divorce Referendum, and two Rhodesian referendums from 1969.

5. It has been claimed that since 1953 referendums in Denmark have been nearly as frequent as general elections: In the period 1953-79 11 referendums and 12 general elections took place (Miller 1982, 56). This is a truth which needs modification. Of the 11 referendums four on the land laws took place on the same day (25 June, 1963), and two other referendums on the Constitution and the voting age were also managed on the same day (28 May, 1953). As mentioned in the text, the 1971 referendum took place together with the general election on 21 September, 1971. This means that the referendums 'fill up' much less in the public consciousness than the frequent elections of the 1970s where general elections were held every second year. During the last decade only two referendums — the 1972 EEC referendum and the 1978 Voting Age referendum — stand out as single events.

6. To be correct, according to a temporary provision to the 1915 Constitution, the voting age was to be gradually lowered to 25 years over a period of 16 years. This provision was, however, revoked in 1920 along with the constitutional changes following the reunification of Denmark and North Schleswig, i.e. the voting age was at once lowered to 25 years in 1920, Svensson 1978, 22-23.

7. The data were collected by the Danish Gallup Institute. Due to a grant from the Danish Social Science Research Council, the author of this paper was able to have a number of items on the voting age issue included in the first Gallup omnibus after the referendum. The interviews took place between 23 September and 1 October 1978 with 1,141 respondents making up a representative sample of Danes above 15 years. The figures reported in this paper only comprise the voters in September 1978, i.e. adults above 20 years (N = 1.054). The sample hit the 'no' percentage exactly, whereas voting participation and the 'yes' percentage were slightly overestimated. Among the respondents, 41 per cent reported a 'yes' vote (34 per cent in the referendum), 29 per cent reported a 'no' vote (29 per cent in the referendum) and 70 reported to have voted (63 per cent in the referendum). The data file is available at the Danish Data Archives as DDA22S.

8. The differences between the percentages in this table and the corresponding percentages in Table 2 in Svensson 1979, 70, is due to the weighing of the data which was done by the Gallup Institute with the information applied in the 1979-article. In this paper the data have not been weighed.

---

9. Respondents were given +1 point for every positive answer to an item favouring the lower voting age and every negative answer to an item rejecting the lower voting age. Conversely, respondents were given —1 point for every negative answer to an item favouring the lower voting age and every positive answer to an item rejecting the lower voting age. Answering 'don't know' or giving no answer obtained 0 point. Application of the SPSS Guttman scale program proved the resulting index to be a reasonable one-dimensional, cumulative scale, the coefficient of reproducibility being .91 and the coefficient of scalability being .56.

10. Reading the figures in Table 7 vertically is tantamount to a control of the relationship between attitude towards lowering the voting age and referendum voting by social class. The table shows that the strong relationship between attitudes and electoral behaviour (see Tabel 6) is found within each social class, i.e. the relationship between this ideological attitude and the electoral behaviour in the referendum is not a spurious one; social class is not a common background variable determining both the attitude and the electoral behaviour.

11. Reading the figures in Table 8 vertically is tantamount to a control of the relationship between attitude towards lowering the voting age and referendum voting by political party. The table shows that the original relationship (see Table 6) is found within all party groups, i.e. the relationship is not a spurious one.

12. Multiple regression analysis requires that all variables are measured on interval scale and that the relationships are linear and additive. Whereas the second requirement seems fulfilled by the results of the bivariate and trivariate analysis, there is, admittedly, a problem with respect to the first requirement of measurement level. None of the variables applied in this analysis are measured on the interval scale level. However, by using one dichotomous, nominal scale variable (social class recoded as a 'dummy' variable with the categories 'wage-earners' and 'self-employed') and two ordinal scale level variables (party preference and ideological attitude) as independent variables and a dichotomous, nominal scale variable ('yes' and 'no' to lowering the voting age to 18 years) as dependent variable, the first requirement is approached to an extent which is now conventionally accepted.

13. For a more elaborate bivariate and trivariate analysis comprising not only social class, political party and ideological attitude, but also age, education and political interest as independent variables, see Svensson 1981, 45-103. It is shown, for instance, that there is a close relationship between age and electoral behaviour in the referendum on the voting age (cf. Nielsen 1970 and Svensson 1979). It is also shown, however, that this relationship disappears when it is controlled by political parly and, in particular ideological attitude.

REFERENCES

Andersen, J.G. 1983. Class Voting in Disguise? The Concept of Social Class, the Political Role of Women, and the Decline of Class Voting in Denmark, Paper prepared for presentation at the workshop on 'Electoral Behaviour' at the ECPR Joint Sessions, Freiburg, March 20-25, 1983.

Balsom, D. & McAllister, I. 1979. 'The Scottish and Welsh Devolution Referenda of 1979:
Constitutional Change and Popular Choice', Parliamentary Affairs, vol. 32, pp. 394-409.

Butler, D. & Kitzinger, U. 1976. The 1975 Referendum, London: Macmillan.

Butler, D. & Ranney, A. (eds.) 1978. Referendum. A Comparative Study of Practice and Theory,
Washington, D.C.: American Enterprise Institute for Public Policy Research.

Coakley, J. 1981. The Referendum and Popular Participation in the Irish Political System.
Paper prepared for presentation at the ECPR workship on Referendums, Lancaster, 27
March to 4 April 1981.

Dahl, R.A. 1982. Dilemmas of Pluralist Democracy, New Haven: Yale University Press.

Elklit, J. etal. 1974. The Danish EC Referendum. Some Preliminary Results, Institute of Political
Science, University of Aarhus, March 1974.

Elklit, J. & Tonsgaard, O. 1978. 'Folkeafstemningen i september 1978 om 18 års valgret', Politico,
vol. 10, no. 3, pp. 34-60.

Inglehart, R. 1977. The Silent Revolution, Princeton, New Jersey: Princeton University Press.

Jones, B. & Wilford, R. 1982. 'Further Considerations on the Referendum: The Evidence of the
Welsh Vote on Devolution', Political Studies, vol. 30, no. 1, pp. 16-27.

King, A. 1977. Britain Says Yes. The 1975 Referendum on the Common Market, Washington,
D.C.: American Enterprise Institute for Public Policy Research.

Marradi, A. 1976. 'Italy's Referendum on Divorce: Survey and Ecological Evidence Analysed',
European Journal of Political Research, vol. 4, no. 1, pp. 115-139.

Miller, K.E. 1982. 'Policy-Making by Referendum: The Danish Experience', West European
Politics, vol. 5, no. 1, pp. 54-67.

Nielsen, H.J. 1970. 'Voting Age of 18 years. Adopted by the Danish Folketing, rejected by the
Popular Referendum', Scandinavian Political Studies, vol. 5, pp. 301-305.

Petersen, N. 1978. 'Attitudes towards European Integration and the Danish Common Market
Referendum', Scandinavian Political Studies, vol. 1 (New Series), no. 1, pp. 23-42.

Petersen, N. & Elklit, J. 1973. 'Denmark Enters the European Communities', Scandinavian
Political Studies, vol. 8, pp. 198-213.

Pierce, R., Valen, H. & Listhaug, O. 1983. 'Referendum Voting: The Norwegian and British
Referenda on Membership in the European Community', American Journal of Political
Science, vol. 27, no. 1, pp. 43-63.

Rosenberg, M. 1968. The Logic of Survey Analysis, New York: Basic Books.

Särlvik, B. 1959. Opinionsbildningen vid folkomröstningen 1957, Stockholm: Statens Offentliga
Utredninger, 1959:10.

Särlvik, B. et al. 1976. 'Britain's Membership of the EEC: A Profile of the Electoral Opinions
in the Spring of 1974 — with a Postscript on the Referendum', European Journal of Political
Research, vol. 4, no. 1, pp. 83-113.

Svensson, P. 1978. De unges valgret, Århus: Forlaget Politica.

Svensson, P. 1979. 'The Lowering of the Voting Age in Denmark: The Referendum of September
1978', Scandinavian Political Studies, vol. 2, (New Series), no. 1, pp. 65-72.

Svensson, P. 1981. Folkeafstemningen om 18 års valgret, Institut for Statskundskab, Aarhus
Universitet.

Statistisk Årbog (Statistical Yearbook) 1981. Copenhagen: Danmarks Statistik.

Tonsgaard, 0.1974. 'Party Voting and Referendum Voting', in Elklit, Jørgenetal. 1974, pp. 100-111.

Valen, H. 1976. 'National Conflict Structures and Foreign Politics: The Impact of the EEC
Issue on Perceived Cleavages in Norwegian Polities', European Journal of Political Research,
vol. 4, no. 1, pp. 47-82.

Valen, H., Elklit, J. & Tonsgaard, O. 1975. Economic Conflicts and Orientations towards Europe
in Denmark and Norway, Paper prepared for the ECPR workshop on 'Mass/Elite Attitudes
towards Europe', 7-12 April 1975.

Worre, T. 1980. 'Class Parties and Class Voting in the Scandinavian Countries', Scandinavian
Political Studies, vol. 3 (New Series), no. 4, pp. 299-320.