University of Michigan

The concept of party identification has proved useful in the analysis of political behavior in two ways. At the practical level it adds information about the forces affecting the voter's decision beyond the knowledge we obtain from ascertaining his voting preference. It contributes to predictions of a wider range of political behavior in space and time than does reported voting behavior or intended vote at a specific election.*We would expect strength of party identification to be indicative of the stability of voting patterns from election to election. Moreover, we would expect that party identifiers would play a more active role in politics than non-identifiers.

The second advantage of the concept is that it opens up theoretical possibilities for the analysis of political behavior in its assumption of long-term forces that predispose the individual to consistently favor a party and its candidates. The specific events and the particular candidates constitute short-term forces. The long- and short-term forces interact to produce the outcome of the specific election. Party identification as a major long-term factor is easy to identify and measure, but shbrt-term forces such as issues, critical events in a campaign, differential resources and their effective employment present more of a problem. Finally the interaction of the two sets of forces presents even greater difficulties. Often, however, people themselves provide the clues for how conflicting forces interact, for they have to integrate, or compromise, or aggregate these influences in reaching a decision. For example, many Democratic identifiers in the USA in 1952 and 1956 voted for Eisenhower not because they perceived him as a Republican but because they saw him as a national leader standing above the political arena. In this fashion they assimilated the short-term force of the Eisenhower personality into the long-term partisan identification and proceeded to vote for Democrats on

the rest of the ticket. By using the notion of party identification we are following Hyman's reference group theory which holds that the individual's own feeling of belonging to a group is more critical than objective group membership. In the same fashion the way he relates his partisan orientation to issues and candidates is more important than the logical weighting of an outside observer. The factual record of the way his party, when in power, has handled certain problems is not as important as how he perceives his party behaved.

The research utilizing this reference group approach in voting behavior has flourished during the past ten to twenty years and has yielded a number of fruitful generalizations. It appears to us, however, that the scope of this research has been limited especially in two respects. In the first place, the great majority of the studies have been based on data from two-party systems and especially from the United States. Secondly, the research has generally not gone into the notion of party identification as a force of attraction; it would be interesting to investigate, for example, the different motivational bases for the attitude. Our paper has the dual purpose of contributing to an extension of party identification research in these respects.

Consequently, this paper is divided into two parts,* the first of which will investigate the extent to which the familiar hypotheses concerning party identification hold true in a multiparty system. The basic model of the two-party system has already been of relevance in a number of multiparty contexts such as France, the Netherlands, Norway, and Sweden, besides what has been termed 'two-and-a-half party systems' such as Great Britain and Germany.¹ Additional evidence from another country, however, may contribute to an understanding of its assumptions. The following problems will be the subject of the data analysis and discussion:

1. Party identification as a widespread and individually stable characteristic of
mass political behavior. This is viewed in relation to the electoral records of
partisan change in the Danish political system.

2. Party identification as a predictor of voting behavior. The measures of identification obtained before the election campaign will be compared with the voting behavior reported after the campaign, and the departures from earlier orientation will be related to strength of identification.

3. The correlates of party identification in three areas. Firstly, party identification will be considered as a determinant of participation in political discussions, of stability of political interest and the use of mass media; secondly, it will be considered as a guide for the voter to opinions on current issues; and thirdly, party identification will be considered as stimulus for ingroup sympathies and outgroup hostility within the political system.

4. The sources of party identification. Despite the limitations of our data for this
purpose, we seek to make inferences from the relation of party identification

* The authors divided the labor so that Ole Borre concentrated on Section I and Daniel Katz on
Section 11.

to age, education, social class, and the interaction of these background factors
with the voter's family and personal environment.

The basic assumption underlying the analyses in the first section is that party identification can furnish a framework for comparative studies of voting behavior only if the differences in its properties, consequences, and sources between political systems are recognized and measured.

The second purpose of the paper is to proceed toward an understanding of the
different needs party identification may satisfy for the individual voter.

As a theoretical instrument party identification is still pretty blunt. The only dimension of identification which has been studied is its strength as subjectively reported by the respondent. One purpose of this paper is to move beyond this one dimension and to examine the motivational orientation of party identifiers. Attachment to a party may have different psychological meanings for different people and their allegiance may not be based upon the same reason. Such variation in motives may have different outcomes for political behavior which could not be predicted from knowing merely the strength of identification. The person who votes for his pocketbook may be strongly attached to the party supportive of his economic interests, and the same strength may be shown by the person for whom the symbols and traditions of the party carry emotional significance or the person who is ideologically committed to its programs. But the fluctuations in the economic cycle may affect the first type of identifier and have little effect upon the second two types. Accordingly we shall examine party identification as it reveals three basic motivational patterns: the pragmatic, the traditional or symbolic, and the ideological. Gabriel Almond has used a similar typology in characterizing political parties in developing nations.² But we are interested not in an overall characterization of parties but of those people who identify with a party no matter what type of party. In passing, of course, we can see whether a given political party is more of one type than another.

In the second section, party identifiers will be categorized along three dimensions
on the basis of the extent to which their reasons for party attachment were
pragmatic, symbolic, or ideological. We shall then test the following hypotheses.

1. The most stable patterns of voting behavior over tune will come from those
of traditional or symbolic orientation; the ideologically oriented will be the next
most stable, and the pragmatically oriented the least stable in voting behavior.

2. Political involvement in terms of interest and activity will be highest among

the ideological, next among the pragmatic, and lowest among the symbolic.
3, The problems people see as politically salient will interact differently ac

cording to the type of identification in affecting voting behavior. For the pragmatically oriented the ability of their party to handle economic problems will weigh heavily, whereas for the ideologically oriented the question of the major direction of party goals will be of greater significance. For the symbolically oriented, the political problems and issues of the day will be of less concern,

Finally, we expect the symbolically oriented to be rural in background, to be relatively high in religiosity and limited in education, and to be overrepresented among the self-employed. The ideological, on the other hand, should be relatively high in education, high in organizational membership and activity including political parties, from homes where parents were politically involved and disproportionately from the professions. The pragmatic should be disproportionately from those raised in urban environments, from employees or managers of fairly large enterprises, and should be low in church attendance.

The data for this study come from a panel survey held in connection with the general election in September 1971. The survey comprised two waves of interviewing, one in August before the campaign got under way, and the other in October after the election. Of the original 2114 persons over 20 years old selected for interviewing, 1499 or 71 percent were interviewed in the first wave, while 295 or 14 percent refused to be interviewed, and 320 or 15 percent were not found by the interviewers for different reasons. Of the 1499 interviewed in the first wave, 1305 or 87 percent were reinterviewed in October, reducing the basis for our analysis to 62 percent of those originally selected. Some tendency for the dropouts to be concentrated in suburbs of large cities and in intermediate age categories has been observed, though a more precise assessment of this bias awaits further analysis.

Though the response rate leaves something to be desired, it is not critical for our major objectives. Our purpose was to ascertain relationships between variables rather than to determine the precise incidence and distribution of single variables in Denmark. Even so, as the following figures indicate, the voting preference for the various parties in October followed the actual election returns fairly closely.

A limitation that we regard as more serious than a minor sampling bias is that we have only one election upon which to base generalizations. To a large extent the empirical foundation of party identification theory has been built by means of repeated observations spanning two decades. Until we can obtain a corresponding continuity in voting research in multiparty systems, generalizations of the type we are discussing below will necessarily have to be tentative in nature.

1. Party Identification in Denmark: Its Determinants, Correlates, and Consequences

Electoral Records and Extent of Party Identification

The existence of widespread and stable party identifications in a society is consistent with a pattern of electoral behavior observable in broad outline in many nations. Rather than moving in a manner that is affected by the immediate past election, as in a random walk, the basic partisan division of the vote appears to oscillate around a baseline which transcends several elections and often spans a generation or more. Such a homing or gyroscopic tendency has been commented upon by many students of electoral records and underlies much of the work done in the field of ecological analysis of voting since its inception with André Siegfried ³ V. O. Key, Jr., inferred a baseline of 'standing commitments' to support one party or the other that might exist at the level of individual voters, but he recognized at the same time that certain 'critical' elections might cause enduring alterations in these commitments for major parts of the electorate.⁴ A statistical treatment by Donald E. Stokes and Gudmund R. Iversen demonstrated that the pattern of partisan change found in US electoral history could not conceivably be described as a random walk and concluded that 'forces restoring party competition' must be assumed to operate at the aggregate level of partisan change.⁵

It is plausible to look upon the aggregate level of party identification in a country as a phenomenon affected primarily by the degree to which the same parties have dominated the political scene for a generation or more. In turn, when party identifications have become widespread and enduring in the society, this provides additional support for the long-range status quo in the mutual strength of the parties. To get into this circular process, it is likely that the party system will have to survive its first years by means of skillful leadership and satisfactory economic progress. Later on, of course, the circle may be broken for example by a severe economic crisis, a crisis of national unity, or when the political system fails to allocate power to groups with new political identities. Structural changes such as the decline of agriculture and the rise of industry tend to shift the balance between the parties, such as the increase in socialist strength and the decline in the strength of the agricultural parties in many European nations have illustrated.

Unfortunately, the critical first phase of this process, occurring in the age before survey research, remains obscure. By comparing age cohorts and relying on a set of plausible assumptions, however, Philip E. Converse has utilized the five-nation data by Gabriel Almond and Sidney Verba to demonstrate a positive relationship between the average strength or intensity of party identification and the age of uninterrupted democracy in the political system.⁶ If this association is applied to the Danish multiparty system, our expectations concerning the extent or average strength of party identification are directed to a comparatively high level. During the first two decades of the century, the power of the electorate in deciding government and legislature was effectively established and suffrage extended to almost all adult persons. The basic features of the Danish party system were fixed

around 1920 as a four-party system in which Social Democrats and Radical Liberals were more or less united on one side and Agrarian Liberals and Conservatives on the other side. Though many minor parties have been represented in parliament from time to time during the past fifty years, only the People's Socialist Party has secured a permanent representation, beginning with 1960. In the resulting five-party situation, Social Democrats and People's Socialists tend to form a 'socialist' group and the three remaining parties the opposite 'bourgeois' group.

What might alter our expectations concerning the extent of party identification are the changes the system has undergone since 1960. The Danish party system changed in that year from a four- to a five-party constellation. As the People's Socialist party, contrary to expectations, survived the 1964 election and proceeded to double its number of seats in 1966, the Social Democrats abandoned their traditional openness toward the center and formed a government dubbed 'the red cabinet' by the opposition because of their collaboration with the triumphant new left wing. The reaction of the Radical Liberals, the former ally of the Social Democrats, was to establish a coalition with the Conservatives and the Agrarian Liberals. This coalition took over in 1968, when the Social Democratic government was voted down by a splinter group breaking away from the People's Socialists. The Radical Liberal Party more than doubled its number of seats in 1968, and its leader became prime minister in the coalition. This government lasted until the election in September 1971, the setting of the present study. The 1971 election resulted in a tie in parliament between the two blocs, and a Social Democratic government was established by means of support from the candidates from Greenland and the Faeroes.

These recent upheavals, though modest by international standards of the 1960'5, may be the reason why our sample of Danish voters seems to fall slightly below the expectations of a very high level of party identification. Table II contains in the two upper rows the relative number of respondents who usually consider themselves adherents of a particular party, subdivided into those who considered themselves strongly convinced adherents and those not so strongly convinced. The third row represents those respondents who denied the label of adherent yet admitted that they felt closer to one party than to other parties, whereas those in the fourth row did not report such partisan leaning.

The combined count of 51 percent adherents before the election campaign and 54 percent after places Denmark slightly below the level of Norway and Sweden with respect to extent of party identification, definitely lower than the USA, and vastly below the German and English figures of around 90 percent. On the other hand, reports from France and the Netherlands indicate that still lower levels are found there. The considerable variation found between nations with a wellestablished democratic rule suggests that the Converse explanation of party identification as determined by the age of the democratic process is inadequate for developed nations.

One line of inquiry which we shall discuss relates to the other attitudinal dimensions of party identification. Besides being distributed according to intensity, as in Table 11, party identification can be distributed along other dimensions such as partisan direction, stability, and salience for the current party preference. The implications of this become apparent if we consider as a simple example that party identification might vary between societies not only with respect to its extent or average intensity but also with respect to its stability and is salience for the vote decision. From the information that party identifications are a widespread phenomenon in a particular society, one might expect to find that the partisan division of the vote was fairly stable over time. It is evident that such an inference would be premature if party identification in that society proved to have little salience for the vote decision, or if party identification in the society in question turned out to be a quite unstable phenomenon likely to be affected strongly by current political events.

Party Identification and Voting Choice

The logic of party identification as a model construct is that it distinguishes the party as a psychological group having implications for the individual voter's opinions and behavior, from the party as the terminal vote decision. The party, so to speak, may appear among both the causes and the effects in the individual decision process, an analytical distinction that does not appear to have been manifest in classical political theory. One implication that immediately follows is that the causal link between the attitude and the terminal behavior may vary in strength and be more or less direct. A strong party identification may, for example, dominate over a weak and diffuse belief system, or it may serve to integrate a belief system which then in turn produces the vote decision.

Thus, the utility of the concept of party identification by no means hinges on its salience for the vote decision in the sense that the voting choice serves as some kind of validity check on the measurement of party identification. On the contrary, its utility would lessen if it tended to convey the same information as questions on current vote intention. It would have no utility, however, if the gradation of intensity, elicited by questions on party identification, did not help in predicting voting choice in the longer perspective. But it is characteristic that the political

event that contributed most strongly in bringing the concept of party identification to the fore in voting research was precisely a situation in which party identification was demonstrated to be something different from the vote decision, namely the 1952 presidential election in the United States.⁷ Throughout the Eisenhower presidency, from 1952 to 1960, it was recognized with increasing assurance that a significant proportion of the Eisenhower vote was borrowed from the Democratic camp and that the Democratic majority in terms of party identifiers was intact and would make itself felt when the transient forces of Eisenhower's personality no longer counted. This did occur in 1960, even though other transient forces, notably John F. Kennedy's Catholicism, prevented the Democratic majority from attaining its full weight.⁸

This piece of history has a direct bearing on our present purpose because it forces us to consider what type of election we are dealing with in the present study. There are in this respect some indications that the 1971 Danish election was one in which short-term forces between the two principal party blocs were fairly well balanced. We are not so much referring to the fact that the outcome was an almost complete balance of voting strength between the former bourgeois coalition on one side and the Social Democratic and People's Socialist opposition on the other, but rather to the fact that this balance also occurred in 1960, in 1964, and in 1966. Those fairly strong fluctuations (by Danish standards) that occurred in 1960 and in 1966 were mainly confined to the voting balance within each of these blocs.

The 1968 election deviates strongly from this balance in favor of the bourgeois parties and especially in favor of the Radical Liberal Party, generally thought to be closest to the two socialist parties. The natural interpretation, according to the party identification model, is that strong short-term forces in 1968 had accumulated against the Social Democratic government and its ally, the Socialist People's Party. In contrast, the 1971 election campaign was generally referred to as a very unexciting campaign, and pessimistic forecasts were issued (erroneously, as it turned out) with respect to the voting turnout.

Had our study been initiated in an election with strong features of realignment of party identifications, such as the 1960 election, or strong features of a deviating election, such as 1968, we might have encountered atypical relationships between party identification and voting.⁹ We feel justified in regarding the 1971 election as one that exhibited normal features in this respect. This assumption becomes important as we proceed to discuss the properties of voters who defect from their party identification to vote for another party, and compare these properties with those of loyal voters. In addition to the inferential evidence we have discussed above, some empirical support for this assumption is given by Table 111 below.

It can be observed that 2 percent of the sample were socialist identifiers who defected to the non-socialist parties, whereas another 2 percent defected the opposite way. At this point, the two partisan camps were in balance. The greater proportion of non-identifiers who voted non-socialist, and the slightly larger extent of non-voting in the socialist camp, is responsible for the improved standing of the

non-socialist parties with respect to the vote division compared to the identifier division. Actually, however, the votes divided more evenly between the two blocs, so that our sample underrepresents the non-socialist voters and overrepresents the socialist voters.

The very small amount of defection across the most important cleavage line leaves the impression of a highly rigid pattern of voting behavior in the Danish political system. Such a picture corresponds to the record of close races during the period 1960-1971, with the exception of the 1968 election. Once we turn away from the bipartisan view and look at the mere fact of defection from one party to another, a much more fluid picture emerges. If the within-bloc defection is added to the above figures, the average rate of defection increases to 21 percent. That is to say, of those reporting a party identification or at least a leaning toward a particular party in the August interview, 79 point reported in the October interview that they had followed through with a vote for that party. Our faith in the functioning of our questions of party identification is enhanced by the figures in Table IV, which show that the inclination to carry through in the vote decision depends strongly on the strength of party identification, varying from 93 percent among strong identifiers to 66 percent among party 'leaners.'

What is more, these defection rates are approximately of the same size as those found in most other societies, as for example the Netherlands and the USA.⁸ But they stand in marked contrast to the ones that might be computed from Table 111. When considered as a two-bloc system, Danish political behavior exhibits a numerical balance between the two sides and a decided reluctance to defect to vote for

the other side. Considered as a multiparty system, however, defection is not an uncommon phenomenon, but evidently these defections stay within the same bloc for the most part, and thereby they greatly reduce the chance that the defection may affect government formation.

We have presented the association between party identification and voting choice over time, relating party identification in August to vote decision as reported in October, since we are interested in assessing the predictive value of party identification for the partisan vote. But this time lag naturally opens up the possibility that the correspondence may be much closer when the two terms are measured simultaneously. A more complete picture of the agreement between our various records of partisan preference is presented in Table V.

These stability figures simply denote the percentage of cases in which the two variables had the same partisan direction, excluding as missing cases those without a partisan direction on one or both variables. For example, we find that of those respondents having a party identification (or partisan leaning) in August and reporting a party vote in October, the partisan choice coincided in 85 percent of the cases - a higher figure than the 81 percent in Table IV because of the exclusion of non-voters in Table V. The main impression conveyed by Table V is that party identification and party preference tend to coincide, whereas party identification in the Danish electorate does not have an individual stability comparable to the one found, for example, in American studies. The shift of 12 percent in the direction of party identification hi the course of two months, even though this period contained an election campaign, seems rather large. It should be remembered, however, that this includes defections within the same bloc of parties.

It is possible, of course, that the 1971 election campaign caused an unusual amount of realignment in party identification. Our data do not permit us to check this possibility. But the finding about the strength of identification and rate of defection does not seem to be specific to the 1971 election. Strong identifiers reported a much greater loyalty to their party for the period 1960-1968. Of those old enough to vote in 1960, 81 percent of the strong identifiers said they had voted for their party at all four general elections, whereas only 64 percent of the weak identifiers and only 38 percent of the party leaners reported to have done so. The differential between the three groups is even greater for the voting history of the individual than for defections at the 1971 election.

Correlates of Party Identification

The importance of party identification has been recognized many times in discussions about the functioning of the political system. For the good or the bad of representative democracy, it offers a view of the party's empirical role, which cuts deeply into political theory generally. This is not the place to attempt to do justice to controversies about the impact of parties in organizing demands and mobilizing support for the political system, or the dialectics of tangible interests versus belief systems and ideology, let alone the discussion of voter rationality. Instead, we have chosen to present data on the correlation of party identification with indicators in three suggestive areas. First, we shall follow up on the hypothesis that party identifiers participate more in politics, expose themselves more to the political content of mass media, and report greater political interest. Secondly, we examine the notion that strong identifiers of a party take more specific stands on issues than weakly associated partisans, and that they consequently are chiefly responsible for partisan differences in public opinion. And thirdly, we explore our data for the possible affective overtones party identifications may carry, with the perspective that party identification may generate a climate of conflict in the political culture. All three areas are discussed more fully in section two, but at the moment we lay down the foundation for this discussion by relating some indicators of strength of party identification in the three areas: 1) political participation, 2) stands on issues, and 3) ingroup/outgroup feelings.

Party identification and political participation. Our evidence in this area is a variation on a theme first sounded in the Erie County study, published in 1944, which has lingered in the debate ever since.¹¹ The finding that party changers during the campaign were on the average less interested in politics than were nonchangers topped the list of a series of findings all of which ran more or less counter to traditional views on the voter. The finding was repeated in British studies during the 1950's and extended in these studies to cover turnover between parties from one election to the next.¹² The criticism by H. Daudt called attention to certain methodological weaknesses and hasty generalizations in the community studies, and made clear that 'floating voters' were often not below average in political information and interest when non-voters were taken into account.¹³ On the whole, the diffuse and vague nature of the findings has been a comfort to those who believe that a sort of Brahmin caste exists in the electorate, combining high involvement with low partisan commitment. Philip E. Converse called attention to the opposite caste of obstinate partisans who were practically incommunicado during the election campaign. Especially in elections with a low general activity level, these voters were seen to confound the floating voter hypothesis markedly.¹⁴

In a multiparty system, new opportunities arise for confusion. The floating voter may be stranded on the reef of an intermediate party, or long-distance floaters may obey the hypothesis while short-distance floaters may not, as demonstrated by Bo Sårlvik for the Swedish electorate.¹⁵

Having access only to recall data from earlier elections, we are not in a position to deal with the issue of long-range electoral change. We propose instead that different kinds of participation may be differently related to strength of party identification and thereby to partisan change. In Table VI, a wide variety of indicators of participation is correlated with strength of party identification as elicited in the August interview. Since length of school education is generally thought of as an alternative determinant of political participation, we have added the correlations between education and the various indicators to the Table.

The indicators have been classified into three groups. The upper group contains such items as political discussion, interest, and party-oriented activities and has been derived from answers from both interviews. The middle group deals with media exposure before the election campaign and is naturally collected from the August interview. The lower group of items represents mass media exposure during the campaign and accordingly is taken from the October interview.

A glance at the upper group discloses that general political interest, as measured in the two interviews, is positively related to both education (.17) and to strength of party identification (.24). Participation in political discussions, either before or during the campaign, shows a very slight relationship with identification. Education is positively related to discussion before and during the campaign, but unrelated to opinion leadership. Again a different pattern emerges in the partyoriented activities, which are positively related to strength of party identification as expected, but unrelated or negatively related to education. The overall activity index, which sums up the scores on all the above indicators, shows a positive relationship to both strength of party identification and education. Evidently, what causes this relationship is different for the two factors: interest and partisan activity are productive of high overall scores among strong party identifiers, while interest and discussion make for high scores among the well educated. It seems that party identification and school education generate different kinds of activity but are similar in promoting general political interest.

With respect to media use before the campaign, we find that the strongly identified partisan is attentive to radio and TV programs with political content, whereas the well educated concentrate on finding political information in the printed media. Both factors make for political reading in newspapers, though we would anticipate that the kind of newspaper articles the two active categories read tend to be different.

This pattern of media use appears by and large to continue during the campaign, with party identifiers being disproportionately represented in the TV audience and partly also in the radio audience. In our summary indices at the bottom of the Table, education appears to lose its importance during the campaign, the correlation dropping from 0.17 to 0.04. This is apparently due to differences in the questions on the two indices, with the questions in October dealing more with TV and radio during the campaign. This emphasis can be defended, however, on the ground that it is especially in TV and radio that political information is intensified during the campaign. It is likely, though we cannot support this directly, that in total exposure to political information, the well educated tend to lose their lead over the rest of the population during the election campaign.

Party identification and stands on issues. A common notion behind a number of studies has been that persons with higher political involvement typically differ from those with less involvement in their political attitudes. In the first place, the difference is one of the frequency of the 'Don't know' answers on questions of ideology and current political issues. The involved should of course be those with opinions. The matter becomes more complicated when we deal with a neutral,

depends, or neither agree nor disagree category. The sophisticated, highly involved voter may tend to favor such a response. In the second place, the strong partisan should consistently favor the stand of his own party whereas the weak partisan or the independent should not show support for any single party.

In the two-party system it is sometimes assumed that issues play a minor role compared to a multiparty system, in that each of the two parties represents a coalition of many interest and factional groups. Hence the two-party platforms are compromises, which are wide umbrellas to cover divergent groups. In contrast, in a multiparty system the many factions and interests can find a real home, with farmers flocking to the Agrarian Party and wage workers to the socialist parties. On the basis of this assumption one would expect to find a stronger relationship between party identification and attitudinal position on issues in the multiparty than in the two-party system. The results of our survey, however, show that the two systems are much more similar than one would assume. It is the exception rather than the rule to find an issue that really separated the Danish parties in 1971. The consensus is remarkable and raises questions about the Almond theory that an important function of political parties is to articulate and aggregate the interests of different groups. Kent Jennings, it should be noted, reports similar findings from Holland.

Thus when we examine the relation of party identification to stand on issues we are limited to the few issues on which there were some differences among the parties. Four of our questions did separate the parties to some degree. One of them concerns the taxing of private houses and reads:

1. The taxing advantages of the owners of one-family houses should be removed.

Another was in the domain of public investment control:

2. Credits for industry should be directed by a public board.

The two remaining statements deal with the major issue of the campaign, Denmark's entry into the European Community; in the eyes of many observers, this discussion hinged on the British entry, which at that time was not a certainty. Among our statements on this issue, we selected the following two:

3. We should under no circumstances join the EEC.
4. Denmark should join the EEC together with England.

The respondents voting for the five major parties exhibited marked partisan differences on all four issues. In Table VII, we have calculated the average conservatism' of the voters of these five parties on each of the four issues by assigning scores from one to five to the respondent according to his choice of response from 'Agree completely' to 'Disagree completely.' Agreement is considered the socialist position on the first three statements but the conservative position on the last.

The order of the parties from socialist to conservative is similar on the three statements except that some uncertainty exists on whether the Conservative Party or the Agrarian Liberal Party is the more conservative. If strength of party identification is related to position on political issues, then the correlations in Table VIII

between identification and conservatism should be on the negative side for socialist voters and on the positive side for bourgeois voters. In other words, the stronger the identifier with a socialist party the more likely he should be to reject the conservative stand on an issue; the stronger the identifier with a bourgeois party the more likely he should be to accept a conservative position on a given issue. (In this analysis party identifiers and party leaners who defected to vote for a different party are excluded from the correlations.) There is some slight support for the notion that party identification is related to the individual's stand on issues. All

of the ten correlations for the two socialist parties are negative and seven of the correlations for the two most conservative parties are positive. The evidence is fragile, however, in that not a single correlation for three of the parties is significant. This is additional confirmation of the hypothesis that the Danish political parties in 1971 did not represent different positions on issues. What was at stake was who should run the country.

There is one exception to the lack of findings on identification and stand on issues, and this occurs for the center party, the Radical Liberals. Strong identifiers in this party were more likely to be opposed to Denmark's entry into the European Common Market than party leaners. It appears that the strongly partisan Radical Liberal, in spite of the recent cooperation of his party with the conservative bloc, is close to the socialists in some issues. Historically his party has in fact cooperated with the Social Democrats. Thus Radical Liberals as a center group can still move in either a right or a left direction depending upon the situation.

Party identification and ingroup/'outgroup feelings. In addition to the mainly cognitive functions of party identification we have examined above, we want to scan our data for evidence that parties as psychological groups nurture affective values for their congregation of voters. Such a notion is consistent with the apprehensive views on partisan commitments that have been expressed by classical writers of political science. It also dovetails with some of the findings in political socialization that point to formation of partisan sympathies fairly early in the childhood of many voters, before an understanding of partisan differences in politics and ideology has developed.¹⁶

The component of attitudes most directly and readily measured is the affective component. Accordingly we asked people to tell us how much they liked or disliked the various parties - using the sympathy thermometer for this purpose. The interviewer presented the thermometer in the following way:

We would like to hear how you feel about the various political parties. I have here a card showing a kind of thermometer. It is called the sympathy thermometer, and we ask you to give the parties a temperature according to how well you like them. All parties you like will be given plus degrees, and the better you like the party the higher temperature you give give it. Parties you do not like are given minus degrees. If you neither like nor dislike a party you should give zero degrees.

Whether we used the simple scoring of own party or took the difference between
score of own party and the average score of other parties, the correlation with
strength of party identification was positive, as witnessed by Table IX.

Questions less heavily affective but still reflecting some emotional tone with respect to ingroup/outgroup reactions dealt with a) wanting own party to form the government alone, b) choosing a prime minister from own party, c) not conceiving of voting for another party, and d) naming specific parties outside one's possible choice.

Table IX shows significant relations between strength of party identification and five of the six indicators of ingroup/outgroup feeling. Party identification does involve an affective component. It is of interest, however, that these correlations are not higher, in that partisan stand on issues is only slightly related to strength of identification. If neither beliefs about issues nor affect account for very much of the variable of party identification, then what does identification consist of? Contributing factors can be found in attachment to leaders and in differential sources of motivational attraction, the latter of which will concern us in the last section of this paper.

Sources of Party Identification

Party identification can be conceived of as an intervening variable between the voter's social experience and his political behavior. Until now we have concentrated on showing that the strength of this kind of reference group attachment is relevant for predicting behavior in a variety of fields, including of course partisan choice. Our experience has been that party identification functions in Denmark much as in other political systems, though some findings do not generalize to Danish voting behavior in quite the same form. Our next move will be to discuss its origin and social conditions for change. Previous studies have focused on age and education as related to the learning process, and on norms residing in the family, social class, and personal environment of the voter in an attempt to describe the transmission mechanism and agents. These studies have been the point of departure for our analysis.

Age and transmission of party identification. A general finding in research on party identification is that older voters identify more strongly than younger voters, and in this respect Denmark is no exception. As shown by Table X, the proportion of our respondents who classified themselves as adherents of one party or another rose systematically from 34 percent in the ages under 30 years to 70 percent among persons in their sixties or seventies.

This pattern conforms to the predictions from a learning hypothesis on the
micro level, but we should attend to the fact that the pattern might also arise from

a very different process, namely a downward secular trend in the frequency of party identification. If young people nowadays were less inclined than young people a generation ago to identify with a party, and remained at this lower level throughout their career as voters, a similar pattern would be observed. The two hypotheses naturally lead to quite different predictions about the future level of aggregate party identification, the learning hypothesis predicting an almost unchanged level as against the hypothesis of a secular decline which predicts a longrun decrease in aggregate partisanship as the electorate's older and more partisan cohorts die. We should therefore be anxious to consider whatever faint evidence might support one or the other of the hypotheses. For lack of more direct types of evidence we turn to inferences on the basis of variables that may be thought of as intervening in the transmission of partisanship from one generation to the next. Converse and Dupeux have employed the respondents' recall of parental partisanship in an attempt to account for the contrast between the high level of partisanship in the USA and the moderate level in the French political culture.¹⁷ Even though we are here working with differential rates of party identification between subcultures within the same political system, the insights that are gained by focusing on this transmission link in the politization process give us every reason to explore it also in a national setting.

Following the learning hypothesis we would imagine a kind of steady-state process in which every group reported their fathers' partisanship with similar frequency. In the course of the generation span of roughly 30 years, the new cohorts who entered the electorate would have built up party identification to an extent that would bring them on a par with their fathers.

By contrast, the hypothesis of a secular decline would, in its purest form, hold that the proportion who recalled their father's partisanship were highest in the older generations because they were socialized at a time when partisanship was more widespread than now. Indeed, the recall frequency would be expected to increase with increasing age parallel with the increase in respondents' own partisanship, since for example our 50-59-year-old respondents have been drawn from roughly the same cohorts as those fathers whom our 20-29-year-old respondents have reported on.

Table XI shows that arguments for both sides are to be found in our data. On
one hand, party identification become more frequent with increasing age, both

among those who recall their parental partisanship and among those who do not recall, with the latter group lagging behind the former throughout the age groups. On the other hand, the recalls are more frequent among middle-aged and older voters than among the young. Before taking this as evidence of a downward trend in partisan attachments, we should, however, consider another possibility. Middleaged voters may be better able to reconstruct which party their fathers used to vote for, in part because they have known their father for a longer period of their adult life, and in part because they have become more partisan themselves and thereby more sensitive to their father's partisan choice, whether or not it agrees with their own partisanship. The causal direction of the associations between the three variables is far from clear, and the only thing that stands out is the verification of the learning process.

Education and party identification. In contrast to most other societies, party identification in Denmark is associated with low education. Among respondents with only seven-year primary school education, three-fifths identified with a party, while among those with 12-year gymnasium schooling less than one in five were willing to call themselves adherents of a party. Educational standards have naturally been rising in Denmark as in other societies, and we may therefore wonder whether the age relationship is a reflection of the educational relationship or vice versa. Table XII shows that even though one relationship is slightly weakened when the other is controlled for, both age and education continue to affect party identification substantially, the frequency of such attachments reaching 73 percent among older respondents with low education and dwindling off to 38 percent among young, well-educated respondents.

This remarkable difference indicates that the functions of party identification we have been discussing, such as generating partisan stability in voting and motivating certain kinds of political participation, are anything but preserves of the welleducated strata of the Danish electorate. If we inquire into the origin of the education differential in party identification, we should consider the possiblity of generation transmission. The well-educated sector may constitute a political subculture with weak party identification and children born into this subculture perpetuate it.

Our explorations in this direction are of necessity inconclusive. Since mass education beyond seven years is a comparatively modern phenomenon, only a few fathers with secondary education are reported on in the sample, and these do not seem particularly nonpartisan. The majority of our respondents with secondary education came from homes where the father had a primary education. It is especially in this group of respondents that we find a reluctance to form enduring partisan ties. It appears that the expanding educational system has weakened party affiliations especially where it has removed people from homes without a tradition of secondary education.

Table XIII describes the combined effect of age, education, and recall of father's partisanship on the formation of party identification. The right-hand column shows that the frequency of recalling the parental partisanship is not affected much by respondent's education. The other figures in the Table suggest an intricate interplay perhaps better recognized by the graphic representation in Figure 1.

The learning process we infer from the positive slope of the lines appears to
work for all four groups but more forcefully so in the two groups with primary
education than in the two groups with secondary education. The effect of a parental

partisanship, as reported by the respondents, is almost negligible for voters with a secondary education. For those with primary education, it is very distinct and acts to give children of partisan fathers a headstart in party identification in front of the other three groups.

Social class and general partisan orientation. In accounting for the partisan direction of party identification, rather than its very existence or strength, the research tradition has generally focused on societal cleavage lines and in particular on social class. The übiquitous impact of social class caused S. M. Lipset to refer to elections as 'the democratic translation of the class struggle.'¹⁸ In line with this view, class voting has been a prominent topic of comparative research, and, especially in British studies, the cornerstone of voting theory and research.¹⁹

In our exploration of class voting we have deliberately suppressed the nuances of the multiparty system in order to assess the basic validity of a dichotomous conception of Danish voting behavior. The crude distinction between socialist and bourgeois (or non-socialist) party identification matches in a remarkable way another dichotomy, that of working-class versus middle-class identification. Table XIV shows that 86 percent of those who identify with the working class also

identify with one of the socialist parties. Of those who are middle-class identifiers only 37 percent identified with one of the socialist parties. This class difference is as great as is found in any other country. (The question on class identification was open and did not provide response categories. Nevertheless some 85 percent of the sample perceived themselves as either working or middle class.)

The figures in Table XIV also suggest that family has at least as great an impact on party identification as has social class. The social inheritance of the general political orientation of the father occurred in 85 percent of the cases, where the father had a socialist orientation, and in 72 percent of the cases in which he had a bourgeois orientation. This asymmetry in favor of the socialist side of the political spectrum more than offsets the asymmetry in the opposite direction, which stems from a larger number identifying with the middle class than with the working class. As the cell entries make clear, socialist families have succeeded to a large extent in transmitting a socialist party identification even to children who have acquired a middle-class orientation. On the other hand, working class identifiers who have been reared in bourgeois families have tended to defect from their parental partisanship and to become socialist party identifiers.

The defection from family norms visible in these two cell entries, where such norms conflict with the social class of the respondent, has commonly been viewed as a consequence of social mobility. Thus one would suggest that in the Danish society, the socialist parties have succeeded in winning the balance of the voters' loyalties among both upwardly and downwardly mobile persons. Thereby these parties have averted the threat of losing ground in the face of the white-collar expansion that has taken place during the last generation. But as pointed out by Arthur S. Goldberg, the argument of social mobility as a source of defection from family norms is logically valid only in the cases where these family norms are typical or normal for the social class of the parent family.²⁰ If the family norm is atypical or deviant, such as in a working class family with bourgeois orientation, the defection might instead be expected among those voters who remained in the working class, that is, among stationary rather than socially mobile respondents. In Goldberg's data the deviant family norms came through with surprising frequency in accounting for defections. This modification of the mobility hypothesis into a more rational set of expectations appears even more necessary after a look at the Danish data, which are presented in Table XV. We have placed the workingclass

identifiers with a middle-class background together with the middle-class identifiers with a working-class background in the upper row, and contrasted these voters with the respondents who identified with the class corresponding to their background in the lower row. The 'normal' family norms have been identified as respondents who grew up in working-class homes with socialist fathers, and those who grew up in middle-class homes with bourgeois fathers. These have been placed in the left-hand column and contrasted with those with a 'deviant' family norm. The cell entries are the percentages who defected from the family norm in these four cases. As the Table shows, such defection was very frequent, around 40 percent, in the two cells representing a strain between the family norm and the present social environment of the voter. A much lower rate of defection, around 15 percent, was found for the small number of respondents who moved away from a deviant family orientation and into a social class that was more congenial with the family norm. The lowest rate of defection, around 10 percent, was found among that majority of the voters who came from normal families and remained in the social class of the family. This is similar to the findings of Valen and Katz in Norway where strength of party identification was not as great among the occupationally mobile as among the occupationally stable.²¹

The right-hand marginals show that our data support the moblity hypothesis in its crude form: socially mobile persons are more likely to defect from parental political norms than are stationary persons. The bottom row marginals would support the hypothesis that a deviant family norm results in defections. But it is especially the interaction between these two variables that is related to defection from family norms, signifying that defections are related to the strain between family and class norms. In our data, this relationship comes through much more distinctly than in Goldberg's original data, even though Goldberg combines six different variables to predict which party his respondents could be expected to vote for, whereas we have relied on a crude class dichotomy.

Homogeneity of personal environment. Social class identification denotes the voter's psychological position in the broad cleavage which the majority of our respondents appear to perceive and to associate with the chief governmental alternatives. The more narrow environment of the voter may be expected to mediate class consciousness or its translation into party politics in ways that differ as to

their intensity and direction. Table XV can from this point be viewed as assessing the degree to which past primary ties have succeeded in deflecting the voter's inclination toward class voting. Conversely the low frequency of defections that are observed in the Table where social class norms and parental family norms coincide may be taken as an indication of the reinforcing effect of primary groups on class norms. The emphasis the authors of the early community studies in Erie County and Elmira put on this reinforcing effect has sometimes been interpreted to mean that the differential partisan division in the broader social structure is conditioned by primary group attachments.²²

We intend to assess the effects of the primary environment by using the responses to questions on primary group homogeneity similar to the ones used in the community studies. We asked our respondent to judge whether it was his impression that his closest friends have about the same or a different partisan orientation as himself, followed by a similar question about respondent's colleagues at work. Admittedly such questions may elicit non-veridical perceptions from the respondent, and national surveys are not designed to check the validity of the answers by means of followup interviews with the friends and colleagues. Apart from this danger, the homogeneity of individuals' personal environment may reflect both cause and effect. Party identification may determine choice of friends, and conversely one's circle of friends may determine his political orientation; the process is circular.

We shall assume that a socialist party identifier perceives a socialist norm when he reports that it is his 'impression that all or nearly all of (his) friends have about the same partisan orientation as (himself)' but that he perceives a bourgeois norm when 'all or nearly all have a different orientation than (himself).' Indeed the strength of the socialist norm has been differentiated by steps into five grades according to the choice between the five response categories, the above two responses representing the two extremes. For a voter who identifies with a bourgeois party we have naturally reversed the scale so that the strongest socialist norm perceived is represented by the response that all or nearly all of the voter's friends have a different orientation than himself. In constructing this scale we have assumed that our respondents have the same idea of the notion of general partisan direction as we have. Some disturbances naturally enter when this assumption is not true. A communist party identifier, for example, may perceive that his friends have a different general partisan orientation than himself if these friends arc mostly Social Democrats, and an Agrarian Liberal may find that conservative friends have a different orientation than himself. Conversely, an old-style Radical Liberal may not feel that his Social Democratic friends differ from himself in general partisan orientation, and this will cause errors in the opposite direction. In view of the fairly clearcut bipartisan picture emerging from the previous tables, we do not think, however, that such disturbances will be serious in our assessments of the environmental norms.

Table XVI shows the correlations between the voter's own party identification
and his perception of environmental orientation so defined.23

The simple correlations in the upper row of Table XVI are of roughly the same size as the correlations between socialist/bourgeois party identification and reported vote of respondent's father (0.57) or between party identification and social class identification (0.45). Controlling for these factors means a decrease in the correlation with environmental norms but by no means a tendency for them to vanish into insignificance. Environmental norms as perceived by the respondent seem to rise as a third factor side by side with the two already investigated in determining the general direction of the voter's party identification.

In conclusion our findings about background variables affecting strength and general direction of party identification might to a considerable extent be anticipated from earlier research in other political systems. Two phenomena may, however, be fairly novel or more outstanding in Danish voting behavior compared to previous research. One is the disinclination among well-educated voters to form enduring partisan ties, and the other is the tendency for the seemingly multifaceted party system to gravitate into a bipartisan set of voter perceptions. The wider perspectives of these results for political ideology and participation, and their relationship to social structure past and present, would in themselves warrant a closer analysis which, however, lies beyond our present purpose.

2. The Psychological Nature of Party Identification

The psychological basis for party identification should give us information about political behavior that goes beyond that provided by mere strength of identification. Strength of identification, as usually measured, calls for the respondent's categorizing himself as a strong party adherent, a weak adherent, a leaner (feeling closer to one party than to another), or an independent. The motivation for such identification is not inquired into, and this omission has two weaknesses. In the first place, strength of identification through such self-characterization is not the equivalent of strength of motivation. People may identify with a party through conformity with the perceived opinions of others rather than from intrinsic needs of their own and may shift allegiance accordingly as they perceive changes in the

climate of opinion. In the second place, forms of intrinsic motivation may have variable effects upon different types of political behavior, whether it be stability of voting behavior, active involvement in the campaign, or feelings toward rival parties. We have distinguished three types or forms of motivational orientation: 1) the pragmatic, 2) the ideological, and 3) the symbolic.²⁴ People attracted to a party because they see it as protecting or enhancing their economic interests would exemplify the pragmatic orientation. People who are attached to the general philosophy of their party - to its long-range program - can be said to be ideologically oriented. Finally, the symbolically oriented are those emotionally attached to the traditional labels and signs of the party.

All three types of motives can be found in the same person, but in general many
people will emphasize one type of motive over another.

Our measures of these three patterns are admittedly not adequate because the interview schedule provided data for a variety of interests, and any single research concern was severely limited in the number of questions permitted. Nevertheless, we believe that this more analytic approach to party identification has sufficient theoretical and practical possibilities to justify exploration. It can open the door to more discriminating analysis even though in this one study we do not move much beyond the threshold.

The major question utilized to give the basis for an index of type and level of motivational attraction was repeated in the August and October surveys and called for first, second, and last reasons for party identification. The reasons were precoded alternatives following two of the three motive types. The index also included answers to a related question: 'What was the main reason why you voted the way you did?' This was an open question, and people's answers were coded in categories some of which could be scored for our three types. These two questions constituted the major part of the index. Some general statements of belief, strongly endorsed, from a long battery of items were also given scores in the ideological and symbolic indices.²⁵

The usefulness of ascertaining the motivational basis for attachment to party is shown in Table XVII, which breaks down strength of identification and type of attachment as they relate to voting behavior. Predictions about voting behavior can be improved beyond those made by strength of identification if we also know the type and strength of motivational attraction. Those who are low in motivational attraction are less likely than those who are high to vote for the party they identify with. The differences on this score are slight for symbolic and ideological involvement, but appreciable for pragmatic attraction. When the level of pragmatic motivation is high some 93 percent of the identifiers vote for the party of their choice; when it is low only 72 percent vote for their party.

The importance of looking at the kind of motivational attraction to the party is much more pronounced when we look at weak identifiers and leaners. Strong identifiers vote so heavily for their party that it is difficult to improve the prediction. Nonetheless, level of pragmatic involvement does make some difference even among strong identifiers. Of those low in pragmatic attraction some 93 percent

vote for their party; this jumps to 97 percent for those high in pragmatic involvement. Among party leaners those low in pragmatic attraction support their party at the polls to the tune of only 60 percent, compared to 88 percent of those high in pragmatism. Similarly, the weak identifiers high in pragmatic attraction are 15 percent higher in voting for their party than weak identifiers low in pragmatic involvement.

Ideological attraction helps the prediction of voting behavior only among the
party leaners, where 72 percent of those high in ideological involvement vote for
their party as compared to only 61 percent low in such involvement.

The level or degree of symbolic attraction to one's party does nothing to help predictions of voting behavior among differing degrees of strength of party identification. Theoretically, one would have expected as much or more differential influence of levels of symbolic commitment as for pragmatic or ideological involvement. It would appear that our admittedly meager measures of level and type of motivation failed to tap the symbolic dimension adequately. The distribution of scores on the symbolic level in Table XVII also suggests this in that they seem randomly distributed around the means for the group varying in strength of identification.

Stability of Party Identification and Motivational Attraction

The measure of party identification will show shifts over time even from pre- to
post-election surveys, especially in countries with multiparty systems. These shifts

can be a) in degree of adherence to the same party and b) in actual defection from one party to another. A move from one party to another is often less of a jump in the multiparty system than in the two-party system. Our August and October measures of the same people do in fact show considerable shifting (see pp. 75-78).

It is interesting, however, that such changes are predicted much more adequately from knowing degree and type of motivational involvement than from sheer strength of identification. We measured shifts from our August-October surveys by using a step scale as follows:

A shift of one step in the same party counted 1 and from weak identification to independence 2. For example, a change from moderate identification with one party to weak identification with another scored 5. Strength of identification predicted positively to number of changes, whereas all three types of motivational involvement were associated with stability, as the following figures indicate:

That shifting is positively related to strength of identification is due in considerable measure to a regression effect. People who make an extreme score on a measure the first time frequently do not sustain their high score a second time. Nonetheless, the overall stability of the measure of strength of identification leaves something to be desired. It is greatly strengthened by the information about motivational involvement. The greater such involvement, the more likely it is that people will show the same strength of party identification. This is true for all three types of involvement, but it is especially true for pragmatic orientation.

Party Loyalty

Our hypothesis that stability of voting over time would be greater among the
identifiers with symbolic attachment to their party than other types of identifiers
was not confirmed (Table XVIII). High symbolic attraction is associated with

greater party loyalty than low traditional attachment (67 percent of those high in such attraction were loyal or fairly loyal as compared to 53 percent of those low in symbolic attachment). The correlation is not high, however, and is not quite as sizable as it is for pragmatic involvement and party loyalty where 74 percent of those high in pragmatic attraction were highly loyal or fairly loyal compared to 51 percent of the low pragmatic group. Ideological involvement showed the weakest relation to party loyalty, though we had expected it to be intermediate between symbolic and pragmatic. The failure of the data to support our predictions may be due to the weakness of our measures of motivational orientation and to our theoretical underestimation of the stability of pragmatic attachment to a party. The practical reasons for supporting a given party may have more of an enduring quality in a stable society than we had anticipated. This could account for the higher association between loyalty and pragmatism than between loyalty and ideology. It could not account, however, for the failure of traditionalism to relate strongly to party loyalty. It looks very much as if our symbolic measure was the poorest of the three in tapping the appropriate dimension of motivational involvement.

Political Interest and Activity

Our second general hypothesis was that political activism would be greatest among the ideologically oriented, least among the symbolically oriented, with the pragmatists in the middle. By political activism we included various levels of involvement from political interest, through following the campaign in the media, to activities ranging from participation in discussions to attending party meetings and party rallies. The results do support the prediction, as Tables XIX and XX attest. At the activity level ideological orientation is consistently and significantly related to political interest and participation. Pragmatic orientation shows little relationship and traditional orientation a significant negative relationship. Ideological orientation in itself does every bit as well as strength of identification in predicting to overall political activity. It does better, in fact, on some aspects of participation - specifically, discussing political matters with friends and family - than does strength of identification. It does not do as well as party identification in predicting party activities such as belonging to the party and attending its meetings (Table XIX).

If we look at another aspect of political involvement, namely following the campaign in the media, including newspaper, radio, and television, we again find support for our hypothesis. The index for use of the media for political matters in the August survey showed a significant positive relationship between ideological orientation

and media use, no relationship between pragmatism and media use, and a negative relationship between traditionalism and media use - just the order predicted. Moreover, ideological orientation tells us more about media use than does strength of party identification, according to the August figures. The October index, based upon the post-election survey, gives additional confirmatory evidence for the hypothesis, although now there is not as great a difference between the pragmatically and the symbolically oriented. Nor, for that matter, is there as much difference in the October index between strength of identification and ideological orientation (Table XX).

Since ideological orientation is related to education, the question immediately
arises as to whether our findings are simply a function of differences in amount of
schooling. The correlations in Table XIX answer this question directly. In the firsl

place, the correlation between education and political activity is .20 and between ideological orientation and political activity is .26. When education is partialed out, moreover, the correlation between activity and ideological orientation drops only three points to .23. Nor does this partialing affect the order predicted on the activity index. The media indices show the same pattern. The August index shows a correlation of .22 for ideology and media use and only .17 for education and media use. When education is partialed out the correlation between media use and ideological orientation is still .19. The predicted order is in no way affected by holding education constant. The October index for media use gives highly similar results. The correlation between media use and education is only .04, whereas media use correlates .17 with ideological orientation. Partialing out education reduces this latter correlation to .15, but again does not affect the pattern of relationships predicted and actually observed.

Finally, it is of interest that partialing out strength of party identification does not destroy either the positive relationship between ideological orientation and political activity or the negative relationship between political activity and symbolic orientation. This is also true of these relationships with respect to the two media indices. Thus type of motivational orientation clearly gives us additional information about political involvement and activism beyond the knowledge furnished by strength of identification.

Stand on Issues

We lack the data for an adequate test of the hypothesis that motivational orientation can predict differential support for parties through the mediating variable of perception of party's stand on political issues. If the party position in a given election emphasizes its basic philosophy and program, it should maximize its appeal to its ideologically oriented adherents, whereas an attempt to handle immediate economic problems would maximize its appeal to the pragmatically inclined. The issues of the day will have less meaning for the symbolically oriented. Though our pre- and post-surveys did have questions on political issues, there was no systematic attempt to get at perceptions of the positions of the parties. All that we can put to test is the relation of motivational orientation to respondent's own stand among adherents to the various parties on four issues on which some of the parties disagreed.

The first question asked about removing the tax advantages of owners of onefamily houses. Presumably the bourgeois parties would oppose such a move and the radical parties would favor it. In Table XXI motivational orientation appears to be more important than adherence to any given party with respect to the position people take in moving toward greater economic equality. In general the symbolically oriented are for the tax on single homes whether they are in the Conservative Party or its polar opposite, the Socialist People's Party. If they are pragmatists, however, they oppose this measure whether followers of the Conservative or Socialist People's Party. On control of public investments, the support is

again from the symbolically involved in three of the parties, and again from the two extreme parties as well as from followers of the Agrarian Party. Opposition to this issue is from the ideologically oriented and once more type of motivational involvement is more revealing than party preference. On the two questions of entrance to the European Common Market, the Social Democrats show the greatest consistency in opposing entrance no matter what the nature of their motivational orientation. Not so the Socialist People's Party, for in spite of the strong position of the party leaders against joining the Common Market, the adherents of the party of pragmatic inclination wanted to see Denmark enter the Common Market. Similarly, among the Conservatives the traditionalists were against joining, while the pragmatists were more inclined to favor Denmark's entry.

In summary, issue position on questions supposedly dividing the left and right parties gave no clear lines of demarcation. Motivational orientation cut across party lines and some adherents of the Socialist People's Party seemed to have more in common with traditionalists in the Conservative Party than with the pragmatists in their own party.

Sharpening of the Ingroup/Outgroup Dichotomy

It follows from our theoretical position that the mobilization of ingroup feeling, the drawing of sharp boundary lines, and the erection of barriers against other parties would appear most frequently among those who are symbolically oriented. The data in Table XXII simply do not confirm this hypothesis. In all, seven measures plus an index are used to test this prediction. Three of them are based upon a graphic rating scale of feeling - the so-called sympathy thermometer. Respondents indicated on a thermometer the degrees above and below zero which expressed their liking or disliking for the various parties. The first measure was the sympathy expressed for own party relative to other parties; the second was the number of parties given a positive sympathy rating; and third, the absolute sympathy score for own party. Sympathy for own party, whether absolute or relative to other parties, showed significant positive correlations with ideological and pragmatic involvement but not with symbolic orientation.

The other four measures were based upon questions of a) choosing a prime minister from own party, b) choosing own party to govern alone rather than by coalition, c) conceiving of other party as possible recipient of one's vote, and d) naming specific parties which one could not possibly vote for. The last two items did not prove to be sensitive for our purposes. Choosing a prime minister from one's own party correlated significantly with pragmatic orientation and this relationship held up when strength of identification and education were held constant. Again traditionalism showed no relation to insisting upon a prime minister from own party. Wanting one's party to go it alone was also significantly correlated with pragmatic orientation and showed no relation to traditional orientation. For these two questions pragmatism was as good, and even a slightly superior predictor as strength of party identification.

The overall index for all seven items showed pragmatism clearly related to ingroup solidarity, with ideology slightly but positively related, and traditionalism not at all related. The best overall predictor was strength of identification (.37), but pragmatic involvement added to predictive power in that it correlated .19 with strength of identification held constant. Why pragmatism rather than traditionalism should have been associated with narrow and restrictive boundaries for the ingroup is not clear. It is very much like the findings on party loyalty and party voting. Again, this may be due in part to the weakness of the symbolic measure and in part to the lack of political knowledgeability associated with emotional attraction of a symbolic character.

Motivational Orientation and Background Factors

We anticipated that the ideologically oriented would be drawn more from the better educated and higher income groups in the urban centers, and the symbolically oriented from rural backgrounds with less education and with more religious involvement. The pragmatists in general should be intermediate between the ideologists and the traditionalists on most of these dimensions. Because education was seen as a critical variable, we did partial it out in tables, presenting data on differences in political behavior associated with motivational orientation.

In general, these expectations are fulfilled when we examine the data about background correlates (Table XXIII). Education is positively related to ideology, significantly but slightly (.12); whereas it is negatively related to traditionalism (-.35), with pragmatism negatively related and in an intermediate position (-.19). Father's education shows the same pattern as does personal income and social status. So too does rural-urban background, with the traditionalists found most frequently in the countryside and the ideologists more frequently in the cities. Church-going is also highest among the traditionalists, but the poorest attenders are the pragmatists. The pragmatists are also the lowest group in parental social status and social class. Finally, the class identification of the individual shows significant differences related to motivational orientation, with the pragmatists tending to identify with the working class and the ideologists with the middle class.

An interesting finding anticipated by our theoretical approach was the correlation between age and symbolic attraction to the party (.30). From a theoretical point of view, one would expect traditional loyalty to a party to be based primarily on early childhood conditioning and on early experiences in a political party. We would also expect to find such emotional attachment remaining with the person as he gets older, even though there are changes in pragmatic and ideological considerations. This would make for a correlation between traditionalism and age. The facts are that neither pragmatic nor ideological orientation shows any significant correlation with age, though strength of party identification does. It may be, of course, that the relation between traditionalism and age reflects the more common concern with symbolic values in an older period. Our older people may not have become more traditionally oriented with the passing years, but their generation may have had this orientation at an early age.

The same threefold pattern of motivation we have employed for individual identification with parties has been used by G. Almond to describe different types of parties. He distinguishes the pragmatic party characterized by secular bargaining with great potential for aggregating interests, the ideological with a Weltanschauung

usually revolutionary, reactionary, or nationalistic, and the traditional concerned with interests of a particular ethnic or religious group and performing more of an articulation than an aggregation function. To our knowledge no one has tested party differences of this character through ascertaining the motivational orientations of party followers.

Our data for Denmark (Table XXIV) show that for the most part the political parties represent a mix of motive patterns. Nevertheless there are some interesting differences among the parties. The Social Democrats are conspicuously higher in attracting people for pragmatic reasons than any of the other parties, and considerably lower than the others in ideological appeal. The Socialist People's, the Radical Liberal, and the Conservative Parties are all high on ideology and low on pragmatism. The Agrarian Party is intermediate in holding people on the basis of ideology, but high in its traditional appeal. Symbolic involvement is also strong for the Radical Liberals but less important for the Conservatives and Socialist People's Party. One could, for shorthand purposes, say that the Agrarian Party is the party of tradition, the Social Democrats the pragmatic party, the Socialist People's Party and Conservatives ideological parties, and the Radical Liberals a combination of ideology and traditionalism. This characterization, however, could be misleading in that we are dealing with relative tendencies and not distinctive attributes. There is some confirmation of Almond's theorizing in that the extreme parties on the left-right dimension are more ideological, and the Agrarian Party representing a limited grouping is more traditional.

Summary of Section 2

In two general areas, our specific predictions were confirmed in distinguishing types of motivational orientation and their consequences and antecedents. The ideologically involved showed the highest degree of political activity in discussing politics among friends and relatives and attending political rallies and party meetings, with the symbolically oriented much more passive and the pragmatically oriented intermediate between the other two groups. Moreover, the same pattern held with respect to following the campaign in the media. Thus ideologists showed more political activism than the traditionalists. Both of these sets of findings (activity and media use) were not affected by holding level of education constant.

Our expectations about the background determinants of the three types of motivational orientations received clear support. The ideologically oriented came more from upper income, higher educated urban populations, and the symbolically inclined more from lower income, lower educated rural sections. The symbolically inclined were also higher in religious activity. The pragmatists tended to fall between these two groups, and were higher in working class identification. The symbolically oriented, as anticipated, were found more heavily among older people.

Two hypotheses received no support. We had anticipated the greatest stability in voting behavior to come from the traditionalists. It did not. It came from the pragmatists. We also expected that ingroup solidarity would be found most frequently among the traditionalists. It was not. It was found most frequently among the pragmatists.

We had little data to test hypotheses about political issues. What evidence we
did have suggested that the people reacted to political issues less on the basis of
allegiance to party and more on the basis of motivational orientation.

Finally, our most general hypothesis was that type and level of motivational orientation would yield significant information beyond that furnished by strength of identification. Encouraging support for this expectation appeared in a number of ways: 1) for some types of political behavior like political activity, media use, and attitudes toward other parties, the correlations with a given type of motivation were as high and sometimes higher than with strength of identification. 2) Even where, as in the majority of instances, strength of identification was a better predictor than type of motivation, correlations between the dependent variables and type of motivation did not wash out when strength of party identificiation was held constant. In other words, even the prediction of voting behavior based on strength of identification would be improved by including measures of motivational level and type. 3) Party identification is not of one piece. For example, party identifiers who are highly supportive of their party at the polls may not be active in supporting the party during the campaign. Motivational orientations are helpful in predicting different patterns of political behavior.

Some speculation about our findings may be helpful for future studies. All three
indices of motivational involvement need to be strengthened, especially the symbolic.
We found our pragmatic people behaving the way we expected traditionalists

to. It could be that pragmatism plays much the same role for blue-collar workers in the cities as traditionalism plays for the poorly educated in rural areas. In neither case is there a great deal of political activism, nor is there an extensive knowledge of political issues. In fact, both of these types of involvement show negative correlations with amount of political activity, with traditionalism being more negative than pragmatism. In seeking information about the campaign, there is a slight tendency for symbolic orientation to result in less media use, and no relationship for pragmatic involvement. Since the pragmatists were the most stable in voting patterns, electoral loyalty was apparently not based upon knowledge of issues and programs. This is reinforced by our finding that people did not follow their parties on issues as much as they did their basic motivational orientation. What is needed in further studies is not only more adequate measures of motivational patterns, but more theoretical elaboration of the linkages between these patterns and various forms of political activity.

3. Conclusions

The picture that emerges from our survey findings is one of a stable political system in which the parties reflect little in the way of sharp differences of interest or ideology. Parties were not differentiated by their position on issues in 1971. The one critical issue facing the nation on which there was some division of opinion was Denmark's entry into the Common Market. This was not, however, a campaign issue in that only the small minority party of the Socialist People's Party opposed entrance. The bourgeois parties all favored entrance, and the Social Democrats swept the issue under the rug.

If issues are relatively remote for mobilizing partisan followings, what is the major function of political parties? In Denmark they do afford the people some choice in the selection of leaders. But is the consensus among the Danish people so complete that this is the only function political parties perform? There is another purpose they may serve, namely they may provide a general goal orientation for the Danish electorate. People may not be knowledgeable about specific issues and the parties may not make salient differences with respect to particular measures. But the general direction in which the country can move is a factor in people's support of their party. Though none of the major parties would reject the social welfare state, the degree to which the country will move toward a socialist democracy does divide the parties. The division, however, is not so much between individual parties as between socialist and bourgeois parties. This is reflected in the fact that party identification 'does not predict voting behavior for single parties as well as in the United States. If we group the parties into socialist and bourgeois blocs, then party identification predicts voting behavior better than in the United States.

The basis of this division, moreover, is clearer in its determining context than

it is for party difference in the United States. Social class provides the framework
for political differences and is reinforced by primary group background in a society
where social mobility has not been extensive.

Since partisan identification is related to class membership and class interests, its strongest form appears in pragmatic involvement. It is the pragmatists who show the greatest loyalty in their voting behavior. Ideological attraction to a party is related to political activism but not to stability of voting behavior. The ideologists can be strong identifiers, but their very knowledge and concern about issues can result in defections to another party, though generally to one very close to their own in the political spectrum. This is related to the fact that the changing vote in Denmark lacks the non-involvement, know-nothing character of the changing vote in some other countries (the so-called 'floating voter' hypothesis). Finally, the symbolic orientation to political parties is less heavily affective than one would expect in a two-party system. The symbolically oriented are the identifiers who resemble the poorly informed and uninvolved American voters. Symbolic attachment is highest among the poorly educated and among rural church goers. It is not accompanied by interest in and attention to the campaign or by political activism. The symbolically identified are the passive followers whose loyalty over the years seems less dependable than that of the pragmatically identified, but who can be mobilized to support their party by the appropriate appeals.

Party identification in Denmark has many of the same characteristics as party identification in other political systems. It predicts fairly well the voting behavior, party loyalty, and political activism. It is a much poorer predictor of the positions people take on political issues. Its antecedents are in the socialization in the family, the primary group and social class, and in present membership in groupings congruent with those of the period of socialization. Identification, however, cloaks differences in motivational attractions and predictions about forms of poltical activity are improved if we know whether the party's attraction to the individual is ideological, pragmatic, or traditional.

---

NOTES

1. In particular the applications to the Norwegian and the Swedish party systems are encouraging for our purpose. See Angus Campbell and Henry Valen, 'Party Identification in Norway and the United States,' Public Opinion Quarterly 25 (Winter 1961); Henry Valen and Daniel Katz, Political Parties in Norway, Oslo: Universitetsforlaget, 1967, especially Chapter 7; Henry Valen and Willy Martinussen, Velgere og politiske frontlinjer, Oslo: Gyldendal, 1972; Bo Sårlvik, 'Political Stability and Change in the Swedish Electorate; Scandinavian Political Studies, Volume 1, Oslo: Universitetsforlaget, 1966; Bo Sårlvik, 'Party Politics and Electoral Opinion Formation: A Study of Issues in Swedish Politics 1956—1960,' Scandinavian Political Studies, Volume 2, Oslo: Universitetsforlaget, 1967; Bo Sårlvik, 'Voting Behavior in Shifting Election Winds: An Overview of the Swedish General Elections 1964—1968,' Scandinavian Political Studies, Volume 5, Oslo: Universitetsforlaget, 1970.

2. Gabriel A. Almond and James S. Coleman (eds.), The Politics of the Developing Areas, Princeton, N.J.: Princeton University Press, 1948, p. 47.

3. André Siegfried, Tableau politique de la France tie I'ouest sous la troisiéme république, Paris: Colin, 1913.

---

4. Vincent O. Key, Jr., 'A theory of critical elections,' Journal of Politics (1956).

5. Donald E. Stokes and Gudmund R. Iversen, 'On the existence of forces restoring party competiton," Public Opinion Quarterly 26 (Summer 1956).

6. Philip E. Converse, 'Of time and partisan stability,' Comparative Political Studies 11, No. 2 (July 1969).

7. See Angus Campbell, Gerald Gurin, and Warren E. Miller, The Voter Decides, Evanston, 111.: Row, Peterson, 1954, and Angus Campbell, Philip E. Converse, Warren E. Miller, and Donald E. Stokes, The American Voter, New York: John Wiley & Sons, 1960.

8. See Philip E. Converse, Angus Campbell, Warren E. Miller, and Donald E. Stokes, 'Stability and change in 1960: A reinstating election,' American Political Science Review 55 (June 1961).

9. See Angus Campbell, 'A classification of the presidential elections,' Chapter 4 of Angus Campbell, Philip E. Converse, Warren E. Miller, and Donald E. Stokes, Elections and the Political Order, New York: John Wiley & Sons, 1966.

10. Ibid., and M. Kent Jennings, Partisan Commitment and Electoral Behavior in The Netherlands, University of Michigan, June 1972 (mimeo).

11. Paul F. Lazarsfeld, Bernard Berelsen, and Hazel Gaudet, The People's Choice, 3rd ed., New York: Columbia University Press, 1968.

12. See for example R. S. Milne and H. C. Mackenzie, Straight Fight, London: Hansard Society, 1954, and Marginal Seat, 1955, London: Hansard Society, 1958.

13. H. Daudt, Floating Voters and the Floating Vote, Leiden 1961.

14. Philip E. Converse, 'Information flow and the stability of partisan attitudes,' Public Opinion Quarterly 26 (Winter 1962).

15. Bo Sarlvik, Electoral Behavior in the Swedish Multiparty System, University of Gothenburg, 1971 (mimeo), p. 38, and Chapter IV.

16. See for example Fred I. Greenstein, Children and Politics, rev. ed., New Haven: Yale University Press, 1969, Chapter 4.

17. Philip E. Converse and Georges Dupeux, 'Politicization of the electorate in France and the United States,' Public Opinion Quarterly 26 (Spring 1962).

18. Seymour M. Lipset, Political Man, London: Heinemann, 1963, p. 220.

19. Cf. the prominence of social class in David Butler and Donald Stokes, Political Change in Britain, London: Macmillan, 1969.

20. Arthur S. Goldberg, 'Social Determinism and Rationality as Bases of Party Identification.' American Political Science Review 63, No. 1 (March 1969).

21. Valen and Katz, op. cit., p. 214.

22. See Lazarsfeld, Berelson, and Gaudet, op. cit., Chapter 15; Bernard R. Berelson, Paul F. Lazarsfeld, and William McPhee, Voting, Chicago: The University of Chicago Press, 1954, Chapter 6; and H. Daudt, op. cit., Chapter 8.

23. In these controls, parental socialist norm has simply been assigned 0 or 1 depending on whether the father was reported as usually voting for a bourgeois party or a socialist party, respectively. The strength of the class norm, however, has been graded into five steps: 0 for strong middle-class identifiers, 1 for middle-class leaners, 2 for those without class identification, 3 for working-class leaners, and 4 for working-class identifiers.

24. For a fuller account of these motivational dimensions, see Daniel Katz, Herbert Kelman, and Vasso Vassiliou, 'A Comparative Approach to the Study of Nationalism,' Peace Research Society Papers XIV (The Ann Arbor Conference, 1969).

25. The following questions were used in scoring the degree of ideological, pragmatic, and symbolic/traditional orientation: 1. Why do you prefer this party? Here is a card with the most common reasons we hear from people. Which of these fits you best? Which fits you second best? These questions occurred in the August questionnaire in connection with the vote intention and in the October questionnaire in connection with the questions on party identification. The response card indicated five response categories. Respondents who chose the response 'I like the ideas which the party stands for' as the best reason were scored 2 on the index of ideology, and those who chose it as the second best reason were scored 1. Respondents who chose the response 'The politics of the party is to the advantage of me and my family¹ as the best reason were scored 2 on the index of pragmatism, and those who chose it as their second best reason were scored 1. This scoring was done in both questionnaires. 2. What was the main reason why you voted the way you did? This question was used in the October questionnaire in connection with the report of the actual vote. No response categories were indicated. Responses were coded and classified into some thirty categories, of which the following types were scored on the index of ideology: References to conviction or idealistic reasons (scored 2) Ref. to ideas of the party, or what the party stands for (scored 1) Ref. to socialist, liberal, conservative, moderate, progressive, bourgeois party or similar labels (scored 1) Ref. to the party's good policy or its stand on overriding issues such as economic goals and the EEC (scored 1) The following types were scored 1 on the index of pragmatism: References to taxes and social expenses (mainly pensions). Ref. to the party being for the workers, for the trade unions, or simply, for 'us. The present government has failed, R. wants a new government. The following types of responses were scored 1 on the index of traditionalism: It's a good party, it's the best party It has governed well R. has always voted that way; R. has grown up with it. 3. Everybody should have the same wage. In politics one ought to strive to give everybody the same opportunities and the same treatment, no matter what their education and their occupation. Those who save should not be punished by property taxes. The government has too little control over private investments. One must prevent people from getting rich merely by inheriting. The index of ideology was scored 1 for each of these items on which R. chose the Completely Agree or Completely Disagree response, and -1 each time R. chose the Neither Agree nor Disagree response. Convicts should lose their right to vote, except for people who have been fined. The authorities should keep an eye on which tourists are let into the country. Foreign workers should not be allowed to squeeze Danes out of their jobs. The index of traditionalism was scored 2 for Completely Agree, 1 for Partly Agree, -1 for Partly Disagree, and —2 for Completely Disagree responses to each of these items. For each respondent, the scores were added up on the three indices separately. The ranges were collapsed so that each respondent was assigned an integer score from 1 to 6 on each index. In this form the three indices entered the correlations.

---

25. The following questions were used in scoring the degree of ideological, pragmatic, and symbolic/traditional orientation: 1. Why do you prefer this party? Here is a card with the most common reasons we hear from people. Which of these fits you best? Which fits you second best? These questions occurred in the August questionnaire in connection with the vote intention and in the October questionnaire in connection with the questions on party identification. The response card indicated five response categories. Respondents who chose the response 'I like the ideas which the party stands for' as the best reason were scored 2 on the index of ideology, and those who chose it as the second best reason were scored 1. Respondents who chose the response 'The politics of the party is to the advantage of me and my family¹ as the best reason were scored 2 on the index of pragmatism, and those who chose it as their second best reason were scored 1. This scoring was done in both questionnaires. 2. What was the main reason why you voted the way you did? This question was used in the October questionnaire in connection with the report of the actual vote. No response categories were indicated. Responses were coded and classified into some thirty categories, of which the following types were scored on the index of ideology: References to conviction or idealistic reasons (scored 2) Ref. to ideas of the party, or what the party stands for (scored 1) Ref. to socialist, liberal, conservative, moderate, progressive, bourgeois party or similar labels (scored 1) Ref. to the party's good policy or its stand on overriding issues such as economic goals and the EEC (scored 1) The following types were scored 1 on the index of pragmatism: References to taxes and social expenses (mainly pensions). Ref. to the party being for the workers, for the trade unions, or simply, for 'us. The present government has failed, R. wants a new government. The following types of responses were scored 1 on the index of traditionalism: It's a good party, it's the best party It has governed well R. has always voted that way; R. has grown up with it. 3. Everybody should have the same wage. In politics one ought to strive to give everybody the same opportunities and the same treatment, no matter what their education and their occupation. Those who save should not be punished by property taxes. The government has too little control over private investments. One must prevent people from getting rich merely by inheriting. The index of ideology was scored 1 for each of these items on which R. chose the Completely Agree or Completely Disagree response, and -1 each time R. chose the Neither Agree nor Disagree response. Convicts should lose their right to vote, except for people who have been fined. The authorities should keep an eye on which tourists are let into the country. Foreign workers should not be allowed to squeeze Danes out of their jobs. The index of traditionalism was scored 2 for Completely Agree, 1 for Partly Agree, -1 for Partly Disagree, and —2 for Completely Disagree responses to each of these items. For each respondent, the scores were added up on the three indices separately. The ranges were collapsed so that each respondent was assigned an integer score from 1 to 6 on each index. In this form the three indices entered the correlations.

26. See note 25.