Resumé

In re-examining May's law of curvilinear disparity, this article analyzes the structure of opinion of political parties. May suggests that voters, party leaders and party activists have different incentives to participate in politics, and therefore voters and leaders have more moderate positions on issues than party activists. This article is based on the extensions of May's law made by Kitschelt, who argues that curvilinear disparities, although not general phenomena, are bound to occur in specific circumstances. We have focused on the level of cleavage conflict in the system, the variable that Kitschelt himself found to be critical for explaining party radicalism among sub-leaders. Using the Norwegian multiparty system as a test case, we propose that radicalism among party activists is conditioned by the relative saliency of different policy dimensions. The empirical analysis, which is based on surveys of MPs, party members and voters in connection with the general election of 1993, gives very little support to our proposition.

Introduction

This article examines May's law of curvilinear disparity (May 1973). By dividing parties into three main levels, top elites, sub-elites and non-elites, May's law of curvilinear disparity may be easily summarized: Voters usually take the most moderate line on issues, sub-elites prove the most ideologically extreme, whereas top elites are located somewhere in between these two levels. The reason for this curvilinear pattern, May argues, has to do with different incentives for the actors to become involved in, and to continue to participate in, party politics.

Intuitively, May's law of curvilinear disparity seems quite plausible. Since top leaders are dependent on voter support for reelection, they will be well advised to take account of and adapt to voter attitudes and opinions. Party activists, on the other hand, are not constrained by such concerns and can base their stands on purely ideological considerations. This article evaluates the empirical relevance of May's law using data from the 1993 Norwegian

---

* Hanne Marthe Narud, Institute for Social Research, Munthesgate 31, N-0260 Oslo. Norway. E-mail: hanne.m.narud@isaf.no

parliamentary election. We do so by adding to the general formulation of May's law Kitschelt's suggestion of differential incentives. Here, Kitschelt offers several propositions about the conditions that are likely to affect ideological radicalism among party militants (Kitschelt 1989). Most important are the ideological variety among party activists and the level of cleavage mobilization in the system. By adopting and developing Kitschelt's idea, we assume that curvilinear patterns of opinions are likely to occur in issues reflecting a salient cleavage of the parties involved.

The present analysis adds a multiparty system to the line of case studies aiming at testing May's proposition, a trend which, so far, has been predominated the Anglo-American tradition. When tested in two-party systems, empirical analyses have given only weak support to the thesis as formulated by the author (see, e.g., Converse 1975; Dalton 1985; Herrera & Taylor 1994; Norris 1995). However, when Kitschelt reformulated and tested the thesis in a multiparty system, he found a distribution of opinion by and large consistent with the theoretical expectations (Kitschelt 1989). The reason for this anomaly, he suggests, has to do with the mobilization of political cleavages, i.e., the social and political polarization surrounding issues represented by political parties. The Norwegian party system offers an excellent opportunity to analyze the question of intra-party disparities. The parties are divided along several cleavage lines; hence, the proposition that party 'militants' are prone to express more radical views than top leaders on issues of particular importance to the party is easily tested. We have defined radicalism in ideological terms, since it can be measured through survey data on voter and elite attitudes to various sets of policy issues.

The article is divided into six sections. The first section briefly reviews May's law of curvilinear disparity, whereas section two offers a critical discussion of May's thesis and of the expected effect of cleavage mobilization on ideological radicalism among party activists. The third section describes the data used to test these propositions. In order to put the analysis into context, section four gives a short description of the Norwegian party system. The last two sections present and discuss the empirical findings.

May's Law

May's law of curvilinear disparity is based on two major assumptions. First, that party organization may be divided into three different strata. At the top level, we find the party elite - members of parliament, members of the executive, candidates for elected office as well as convention delegates and members of the national executive committee. Next strata is the middle elite

or what we may label the sub-leaders - regional and local party office holders, active and inactive party members as well as voting supporters. The lowest strata - the non-elite - is the inarticulate and occasional party loyalist (May 1973,135-36).

May's second assumption has to do with different incentives for the different strata to participate in politics. Most members of the top elite are elected parliamentary representatives or members of the government. Retaining office, however, depends on winning reelection which, in turn, depends on the electorate's positive evaluation of party performance. The simple formulation of May's law is built on the Downsian assumption of party competition that parties as well as voters are moved by self-interest (Downs 1957). As rational voters, well-informed about the policies of parties, the electorate will choose a party with a position close to their own views. Well aware of this, parties will formulate policies and strategies so as to maximize support from the electorate. Hence, as vote seekers and in order to retain the privileges of office, top leaders have the incentive to be as close as possible to the policy positions of their voters.

Sub-leaders, on the other hand, a great majority of them volunteer selfrecruits, no such incentives. Instead, they are motivated by ideological concerns and aspire to shape candidate selections and party programs in keeping with their own preferences (May 1973, 148-49). Consequently, principles, and not political careers, are seen to be the main motivation for sub-leaders. For this reason, the middle elite will express more extreme (or radical) opinions than the top elite and ordinary rank and file voter.

May's Law Challenged

Literature testing and challenging May's law is extensive, and the empirical evidence is divided (see, e.g., Valen & Katz 1964; Dalton 1985; Iversen 1994; Seyd & Whiteley 1992). Most of the empirical studies of the law have made use of American and British data on party activists. Some of these earlier attempts, however, have suffered from lack of data from the relevant strata of the party organization, e.g., the national legislative elites (see, e.g., McClosky et al. 1960). In addition, previous studies gave attention to activists' perception of rank and file voters instead of focusing on voters' self-placement, which would be a more accurate measure of voter positions. Hence, attempting to avoid the shortcomings of previous work, Herrera & Taylor (1994) took data from multiple strata of party activists, including members of Congress, and examined the structure of opinion within American political parties. However, their analysis showed that Republicans and Democrats do not display opinions that conform with the predictions of the special law of curvilinear disparity. In addition, the

analyses by Seyd & Whiteley (1992) and Norris (1995), who tested May's rule against data from the British system, gave only limited support to the hypothesis. Why are the empirical data so resistant to May's seemingly plausible hypothesis?

There may be several reasons for this. First of all, Kitschelt (1989) as well as Norris (1995) argue that May's law is built on a reductionist psychology and gives too simplified a picture of the motivational factors of the actors involved. In their opinion, too much emphasis has been put on the intrinsic vote-maximizing objectives of the parliamentary party. Second, the simple leader-follower dichotomy developed by May defines sub-leaders too broadly, failing to take account of differences between specific groups of party activists below the leadership stratum (see, e.g., Seyd & Whiteley 1992; Whiteley et al. 1994a; 1994b; Whiteley & Seyd 1996). Third, and for us most important, Kitschelt (1989) argues that the model underspecifies its dependent variable, radicalism.³ Kitschelt's main argument is that curvilinear between leaders and followers are not general phenomena, but are likely to occur only in specific circumstances. In order to test May's law, therefore, Kitschelt extends the law in three important respects.

Three Extensions of May's Law

Kitschelt's first extension of the law has to do with the ideological variety in political parties. He proposes the hypothesis that parties appeal to individuals with a wide variety of beliefs organized along a continuum of organizational, programmatic and strategic radicalism or moderation. Rather than making a distinction between ideological purists and programmatic who provide little ability to predict whether leaders will be more moderate than activists, it should be recognized that between those two ideal types there may be a number of other combinations that are unimportant to the consideration of the law of curvilinear disparity. What is important is to specify the kinds of activities ideologues and pragmatists wish to get involved in and the offices they seek.

The second extension of May's law has to do with the strength of intraparty More radical party militants, Kitschelt argues, are likely to be recruited when a) the social cleavage represented by a party is highly mobilized, b) the existing political regime gives little consideration to the demands of a party's main constituency, c) the party is in a weak competitive position, and finally when d) past moderate pragmatic party strategies have not strengthened the party organization, attracted new voters or influenced public policy. In Kitschelt's view, these four types of interaction between the party organization and its political environment affect the demand for radical policies among party militants.

The third extension concerns the vertical stratification of the party organization. In general, Kitschelt argues, curvilinearity becomes more probable the more loosely coupled a party organization is. Loose coupling means that parties impose few constraints on militants' participation in meetings and decision making processes and that political authority is not restricted to a small group of representatives elected by the militants. Constraint is high when the party leadership chooses conference delegates, medium high when the parties' basic organizations nominate conference delegates, and it is low when all party members are entitled to attend conferences and vote on party policies. Obviously, in the latter case the party leadership has little influence on which groups control the party, and consequently access to mid-level positions is fairly open. This, in turn, enhances the probability of a curvilinear disparity between more radical mid-level party activists and moderate leaders in electoral office.

In sum, we need to specify, limit and modify the law of curvilinear disparity in motivational, ideological and structural terms. Whereas Kitschelt defines these in a number of propositions structured under the various extensions, we can hardly give attention to them all here. Instead, we propose to do a more limited test which is designed to show that curvilinear disparities within parties are connected to the spatial characteristics the party system. For this purpose, we have focused on the level of cleavage conflict in the system, the variable that Kitschelt himself found to be critical when explaining radicalism among subleaders. argument is developed more thoroughly in the following section.

The Level of Cleavage Conflict in the System

According to Kitschelt (1989, 407), more radical party militants are more likely to be recruited into political parties when the social cleavage represented by a party is highly mobilized. Cleavage mobilization, as denned by Kitschelt, refers to the social and political polarization around issues represented by political parties, and it can be measured in terms of the polarization of public opinion. If cleavage mobilization is high, a large number of ideologues and volunteers for party work emerges. Hence, the greater the propensity for radicalization of intra-party groups, the greater the ideological constraints on the parliamentary party. The lower the level of cleavage mobilization, on the other hand, the greater the proportion of pragmatists who join and work in parties.

Here, we adopt the basic idea as formulated by Kitschelt and attempt to
set it in a broader theoretical context, most importantly that of the social
cleavage approach. A cleavage delineates the social base on which parties

build their support, and according to Rokkan (1970), mass politics in Western Europe have been structured around four lines of cleavages. The first two, center-periphery and church-state, emerged as a result of the national revolution in Europe. The other two cleavages derived from the industrial revolution in Europe and generated a conflict between workers and employees on the one hånd (the class cleavage) and rural (agricultural/ producers) interests versus urban (industrialists/consumers) interests on the other. The constellation of cleavages emerged in different forms across Europe and accounted for the most part for variations among the party systems which developed. Duration and strength of the various cleavages has varied, but it is the class cleavage that has dominated the political conflict for most of this century (see, e.g., Rokkan 1970; Rose 1974; Bartolini & Mair 1990).

A clear awareness of the very different intensity of particular cleavages made Lipset & Rokkan (1967) direct their attention to the way in which cleavages had been politicized and depoliticized at certain points in time. The saliency of cleavages is the important factor in understanding the mobilization of party followers. Here, we base our predictions on the assumption that a great number of the policy controversies - as they appear today - can be traced back to conflicts that formed a part of the 'old' cleavage system. That is, we may - with some exceptions - predict the parties' stand on modern issues through insight into their placement in the traditional cleavage structure. The level of polarization between parties and, hence, the extent to which we will expect curvilinear disparities between leaders and followers, is then conditioned by the dimensionality of the policy space.

The Policy Space of Parties

In multiparty systems, parties appeal to individuals with a variety of beliefs organized along multiple conflict dimensions. It is implicit that these parties' (and individuals') preferences may be described in terms of a position in these conflict dimensions. Here, we use the term conflict dimension as synonymous with ideological dimension and even issue dimension, even though the latter is most commonly used as a way of operationalizing the first two. We assume then that the traditional cleavages are, to some extent, reflected in the policy positions of the various parties as they appear on some issue dimension. Moreover, the weight - or the relative importance - that different parties attach to different dimensions may be interpreted as the saliency of these dimensions.

The most common way of operationalizing the idea of a relevant
dimension in Western Europe is to use the left-right scale. It represents a
simple and convenient way to simplify a complex reality, and, consistently

with the predominance of the 'class cleavage,' we often think of the leftrightdimension social and economic policy as marking the birth of the typical and modem West European party system (see, e.g., Bartolini & Mair 1990). However, if we take account of the variation in the strength and presence of the pre-industrial and non-class cleavages, it is easy to recognize other ideological dimensions that are quite independent of the left-right dimension (see, e.g., Pedersen et al. 1971; Converse & Valen 1971; Daalder & Mair 1983; Daalder 1984). For instance, there is no reason to believe that rural interests, per se, should coincide with leftist interests, or that they should automatically be correlated to rightist piacement on the left-right scale. Lijphart (1984:128), listing the various issue dimensions of partisan conflicts, explicitly relates them to the lasting party system cleavages in Western Europe. Among the most important dimensions (in addition to the socioeconomic class dimension), he identifles religious and urban-rural controversies. For the purpose of empirical analysis, which dimensions should be constructed is clearly a question conditioned by country-specific characteristics. From an a priori knowledge of the cleavage structure of specific countries, the researcher attempts to construct issue dimensions that best capture the dimensionality of the policy space. Hence, in the context of Norwegian politics, dimensions that take account of territorial and religious cleavages need to be constructed.

For us, the important point is that parties build up their support on the basis of issue types which they have made their own, e.g., morality, welfare, agriculture, green issues, public spending, and their preferences relative to these issues determine their policy positions in the policy space (Budge & Farlie 1983; see also Petrocik 1996). For the analytical framework of this article, the term saliency is crucial, because it accounts for the expected differences of opinion between different levels of the party. Returning to Kitschelt (1989), where radicalization among party activists was determined by the level of cleavage mobilization in the party system, we would expect that radicalization among sub-leaders is conditioned by the relative saliency of different dimensions.⁴ Therefore, we would anticipate party activists to be more extreme than party leaders on issues that belong to the 'heartland' of the parties concerned. That is, for religious parties we would expect curvilinear disparities between leaders and followers on issues that are highly salient for the party; e.g., church matters, abortion, euthanasia etc. For socialist or conservative parties, we would expect curvilinear patterns to occur on issues related to the left-right dimension; e.g., privatization of public enterprises, social issues, taxation etc.

The following sections set out to test the above proposition empirically.
First, however, a brief account of the cleavages and policy conflicts that
form the basis of the party system in Norway is required.

The Norwegian Party System

The Norwegian party system has been politically defined around six cleavage dimensions, determined by economic, geographical and cultural circumstances.⁵ The class cleavage and sectoral cleavage were determined by economic conflicts in the labor market and the commodity market respectively. A territorial cleavage between center and periphery was partly overlapped by three cultural cleavages: a socio-cultural conflict between two different versions of the Norwegian language, a moral conflict articulated the teetotalist movement, and a religious conflict over control of the Lutheran State Church. More recently, the socioeconomic class cleavage has been expressed through the left-right dimension, the territorial and sectoral cleavages are reflected through the urban-rural dimension, and the cultural cleavages can be characterized as a moral-religious dimension.

A number of cleavages have thus influenced Norwegian politics and contributed to the development of the party system. The class cleavage today accounts for, from left to right, the Socialist Left Party, Labor, the Liberals, the Conservatives and the Progress Party. The two parties at the center of the left-right dimension are based specifically on other cleavages: The Christian People's Party is based on a religious program, whereas the agrarian Center Party has its electoral stronghold in rural and peripheral districts (see Rokkan & Valen 1962; 1964; Rokkan 1967; Valen & Urwin 1985). The major division, however, not only in shaping electoral preferences, also in terms of government alternatives, has been along the left-right axis.

In 1990, the two-block party structure broke down over the dispute about Norway's entry into the European Union (EU). Previous research shows that the conflict over EU activates all underlying cleavages in the Norwegian system and thus creates tensions, not only between parties, but also within the political parties (Rokkan & Valen 1964; Gleditsch & Hellevik 1977; Valen 1976; 1994; Bjørklund & Hellevik 1993; Narud 1995a; 1995b). Most important in this regard is the mobilizing effect on territorial cleavages, the center-periphery and the urban-rural dimensions. But even moral-religious interests are affected, most notably the question of free trade and import of alcohol, a matter of concern to the Christian People's Party. Hence, the very complex question of EU membership affects several party-specific issues that cut across established party lines and is crucial for the understanding of the circumstances surrounding the general election of 1993 (see, e.g., Valen 1994; Aardal & Valen 1995; Narud 1996; Narud & Valen 1996).

In a Norwegian context, we expect curvilinear disparities between leaders
and followers to be conditioned by party-specific ideologies. For the Center

Party, we expect a curvilinear pattern of opinion on issues related to the urban-rural dimension, for the Christian People's Party on issues related to the moral-religious dimension and for the other parties, the Socialist Left, Labor, the Conservatives and the Progress Party, on issues related to the left-right dimension.

Data

To register the opinions held by different party strata, we used three sources of interview data.⁷ In the spring session of 1993, questionnaires were sent to representatives of Stortinget (the Norwegian parliament) and to party members who were taking part in the candidate nomination process.⁸ In addition, a voter survey was conducted after the election.⁹ The respondents were asked to place themselves on various sets of ten point issue scales. The questions were identical for all levels of the party. Unfortunately, the 1993 voter survey lacked three of the relevant items. For this reason, and in order to make comparisons, we have supplemented the data with three of the scales from the 1989 voter survey, which are identical to those we apply for the leadership level. Thus, the data give information about the level of polarization between the various parties as well as the level of concordance in the party organization. The identical wording of the questions as well as the close time interval between the data sets (for the elite data in particular) give us a rare opportunity to make comparisons between the various levels of the party. We have defined the elected party representatives as the top elite, the rank and file voter as the non-elite, and the party activists participating as delegates at the party conventions as the sub-elites.

Like May (1973), we assume that party activists operating in the internal arena are mainly motivated by ideological concerns and that they are not constrained by the same tactical considerations as the top leaders, who operate in the parliamentary arena as well as the electoral arena. Hence, we will expect the disparity between voters and leaders to be at its most extreme point with the party activists. Top leaders, on the other hand, will be somewhere in between the party activists and the rank and file voter.

Analysis

In the surveys, the respondents were asked to place themselves on four ten-point issue scales in addition to the left-right scale. The relevant issues concerned private versus public health care, alcohol policy, agricultural policy and environmental policy. All these issues have preoccupied parties as well as voters with varying intensity for the last two or three decades, and

their relative importance among the parties varies. Alcohol policy involves the question of public control of production and sale of liquor and is particularly salient to the Christian People's Party and the Progress Party, which hold opposite views on the issue. Agricultural policy concerns the level of public support to Norwegian farming and has traditionally been salient to the Center Party and to a certain extent to the Progress Party, the latter opposing public subsidies in general and promoting free market competition. Environmental policy is about so-called green issues, i.e., fighting pollution and the protection of natural resources and has been of great concern to the Socialist Left Party. The question of public versus private health care is of principal interest to all the parties competing along the left-right axis, i.e., the Socialists, Labor, the Conservatives and the Progress Party.

Ideally, in order to assess the emphasis that the parties attach to these issues, we would like to have had information about the relative amount of space attributed to the relevant issues in the party manifestos (see, e.g., Harmel, Janda & Tan 1995), in the electoral campaign or simply openended and elite reports of important problems (see, e.g., Petrocik 1996). Lacking such information, however, we will base our observations on the results of previous research, arguing that these issues are of great principal interest to the parties concerned and therefore salient to them. Consequently, we assume that they reflect the underlying cleavages (as well as potential future conflicts) in the party system (see, e.g., Aardal & Valen 1989; Macdonald et al. 1991).¹⁰ Moreover, they are highly affected by the EU membership issue, since membership would provoke changes in Norwegian alcohol and agricultural policies that do not conform to the objectives of, e.g., the Center Party or the Christian People's Party. In addition, the question of economic liberalism connected with the policy of the European Union generates conflicts that clearly provoke polarization among the parties along the left-right dimension (Narud 1995a).'!

Figure 1 provides perceptions of party position on the left-right scale of all three strata of respondents. Observe that the Liberal Party was not represented in Stortinget at the time and is therefore left out of the analysis. The question was worded as follows: 'There is a lot of talk about the conflict between left and right in politics. Here is a scale that reads from 1 on the left, that is, those who are placed politically furthest to the left, to 10 on the right, that is, those who are politically furthest to the right. Where would you place yourself on such a scale?'

The figure shows that four of the parties, the Socialist Left Party, Labor, the Conservatives and the Progress Party, display a pattern of a small fan. That is, the political elites are situated in a more extreme position than their voters. There is, however, no real tendency towards a curvilinear pattern of opinions; the activists hold almost identical positions to those of the top

elites. The two centrist parties have a different opinion structure. For the Christian People's Party there are practically no differences between the various party levels. This is also the case for the Center Party, but the top leaders place themselves a little more to the right than the sub-elites and the non-elites.

Let us examine the issue that we assumed to structure the vote along the left-right continuum, the issue of whether or not to allow private health services on a commercial basis. The question was worded as follows: 'Value 1 expresses the preference that private health care should play a greater role on a commercial basis, while value 10 expresses the preference that health care should be a public responsibility only. Where would you place yourself on this scale?'¹²

Figure 2 gives some support to the hypothesis that parties competing along the left-right dimension are likely to display a curvilinear pattern of opinions on issues related to left-right controversies. On the health issue, there is a clear tendency of extremism among the party activists concerning the Socialist Left Party, the Conservatives and the Progress Party. The only exception is the Labor Party where top leaders prove to be the most leftist, non-leaders take the most moderate line, and sub-leaders in between. Observe also that the Center Party displays a curvilinear pattern of opinions, whereas the various strata in the Christian People's Party are very close to each other. All in all, the difference between top leaders and subleaders not very big.

Let us move to the two issues of great relevance to the centrist parties

and the Progress Party. We start with agricultural policies, a question typically related to the sectoral cleavage. Previous findings indicate that the Center Party faces the Progress Party as its main antagonist on this particular issue (Aardal & Valen 1989; Valen 1990, Narud 1995b). Consistentwith main hypothesis, our expectation then is that these two parties display a curvilinear pattern of opinions. The respondents were asked the following question: 'Some say that Norwegian agriculture should manage without government support and tariff protection against foreign competition. Let us assume that those who hold that position are placed at 1 on this scale. Others think that the present system of government support for agriculture should be maintained. Let us assume that they are given the value 10 on the scale. Of course, there are some who are between these extremes. Where would you place yourself on such a scale?' Figure 3 shows how the respondents placed themselves on this particular issue.

Figure 3 confirms that the extreme parties are now the agrarian Center Party on the one hand, giving whole-hearted support to agricultural subsidies, and the Progress Party on the other, opposing such subsidies. However, the results run contrary to our expectations; there is no tendency whatsoever towards a curvilinear pattern for these two parties. The top elites of the Progress Party are by far the most extreme, whereas the party voters are located in a centrist position concerning this issue. The party's sub-leaders are positioned between the top elites and the voters. The Center

Party reveals quite a different pattern. In this case almost no differences are evident between the various levels of the party, even though there is a slight tendency for the top elite to be more extreme than the sub-leaders and the voters. Three of the parties, the Conservatives, Labor and the Christian People's Party, show a curvilinear pattern, whereas the top leader and the sub-leaders of the Socialist Left Party place themselves very close to each other, but in a more extreme position than their voters.

The next question concerns alcohol policy, one of the traditional moralreligious in Norwegian politics, and of particular relevance to the Christian People's Party and the Progress Party, the latter being the prime advocate for a more liberal legislation on the sale of wine and liquor. The respondents were given the following options: 'Value 1 denotes the position that alcohol should be sold without restrictions and at greatly reduced prices. Value 10 means that the sale and production of alcohol should be more severely regulated than it is today. Where would you place yourself on such a scale?' Figure 4 shows how the respondents placed themselves on this dimension.

Again, the Progress Party elites are in a far more extreme position than their voters, but on this issue the sub-leaders are very close to the party representatives. For the Christian People's Party, on the other hand, there is a slight curvilinear tendency. This is also the case for the Conservatives. There is almost no difference between the position of the various strata of

the Center Party, whereas Labor and the Socialist Left Party display a
fan-like pattern of opinions.

Finally, on the question of environmental protection (Figure 5), the respondents were presented with the following wording: 'Value 1 denotes the desire to give much higher priority to the protection of the environment, even if it leads to a considerably lower standard of living for everyone, including yourself. Value 10 denotes the position that protection of the environment should not go so far as to affect our standard of living.'

On this issue the parties are positioned in accordance with their distribution from left to right, with the Socialist Left Party and the Progress Party as the main opponents. However, the patterns of opinion are not curvilinear. The top elites are in the most extreme position, the voters are the most moderate on this issue, whereas the sub-elites place themselves in a middle position. The opinion structure of the Progress Party mirrors that of the Socialist Left Party. In the Labor Party the voters are the most moderate, the elites are the most extreme, and the sub-elites are in between. Results for the other three parties suggest very small differences between the various levels of the party hierarchy.

Summary

In sum, four different categories of configurations can be observed (cf. Figure 6). First, we observe a pattern similar to a fan, in which the top elites hold the most radical opinions, the non-elites are the most moderate, and the sub-elites are in between. Second, on some occasions a small fan appears, indicating that the voters hold the most moderate views on the issues concerned, but where no differences between the various elite levels are evident. Third, we observe an opinion pattern compatible to a line, in which all strata hold close to identical positions on the scale. Finally, in some cases and consistent with May's law, there is a curvilinear pattern of

opinions, in which the sub-elites hold the most extreme views, the non-elites
hold the least extreme views, and the top elites are situated somewhere in
between the middle and bottom level.

Table 1 sums up the various patterns of opinion reported in Figures 1 to
5 for each of the six parties on all five issues, whereas Table 2 shows the
patterns of opinions related to the saliency of issues.

The tables clearly indicate that curvilinear patterns of opinion are not conditioned by the saliency of certain issues. Rather, in a majority of cases the findings confirm the well-known observation that political elites adopt more extreme policy positions than their electorates. In most cases, the top leaders seem to hold the most extreme position on the scales, whereas the sub-leaders are somewhere in between the top-leaders and the voters. Most frequently, the parties holding a wing position on the scales seem to display such a structure of opinion. Finally, in those cases where curvilinear patterns can actually be observed, we find only limited support for the hypothesis that this pattern is conditioned by party-specific ideologies. In less than one third of the cases does radicalization among sub-elites occur on issues that are highly salient to the parties.

Conclusion

To be sure, curvilinear patterns do exist. In fact, such a pattern emerges in almost one third of our cases. What we have failed to prove, however, is why they occur on some issues and not others. Related to our initial theoretical framework, the results are rather surprising. Had this been a two-party system, or had the test of May's law been done on the basis of unidimensional models, the results would by no means have been unique. In a multiparty system, however, where parties appeal to individuals with a variety of beliefs organized along multiple conflict dimensions, we would expect the social distinctiveness of parties to matter in differences of opinions within the party elites. Moreover, compared to many other tests of May's proposition, our test has been more comprehensive in the sense that multiple issue scales that reflect the underlying cleavages of the system have been applied. Yet, we have been as unsuccessful in proving the hypothesis right as a number of other scholars in this field. Two questions then come to mind: Why are the empirical findings so consistently resistant to May's hypothesis? And why do political scientists - the present authors are no exception - bother to go on testing it?

The answer to the last question, it seems, lies in the theoretical refinement of the model and the credible arguments that are presented concerning the motivational factors of the political elites. The simple formulation of May's law is built on the Downsian interpretation of party competition. The proximity logic of this theory states that, in order to satisfy the demands of the electorate, the optimal vote seeking strategy for top elites is to adopt a position in the policy space as close as possible to that of the majority of the voters. In a unidimensional system with voters distributed along the center of the policy dimension, this implies that every candidate will act as though he were the mean voter (Downs 1957; Enelow & Hinich 1984). Moreover, the assumption that middle elites are more ideologically oriented seems plausible from simple observations of party behavior. Party activists are highly visible actors, either taking part in factional activity at party conventions or operating in news media, and they are also prime actors in the process of candidate selection. Moreover, they are not constrained by demands for reelection. From this we would expect them, as indeed May states, to hold more extreme views than the top elites. In a majority of cases, however, they do not. Instead, the most common pattern is that of a fan.

The lack of empirical support for May's hypothesis may be due to several reasons. Whereas a number of scholars have pointed to factors as differential elite incentives, properties of the party organization, or information discrepancies (e.g., Norris 1995, Rose 1974; Valen & Katz 1964), little attention has been given to attributes of the individual voter and the dynamics of mass-elite relationships. Recent theories of issue voting emphasize the role

of parties as advocates of conflict (see, e.g., Rabinowitz & Macdonald 1989; Macdonald et al. 1991; Listhaug et al. 1990) or as opinion leaders (see, e.g., Przeworski & Sprague 1986) in the process of electoral competition. The directional theory of electoral competition elaborated by Rabinowitz & Macdonald (1989), for instance, suggests that party competition takes place on the basis of centrifugal forces, and that voters prefer parties to take strong stands on issues that are important to them. Most voters, directional theory claims, are not well informed, sophisticated individuals with clear preferences who try to maximize their benefits by calculating which party is closest to them on certain policy issues. They picture politics in fairly simple terms and perceive issues as divided into opposite sides. Voters are attracted to parties that are clearly on their preferred side and prefer parties to mark ideological direction. Consequently, the best strategy for top elites to maximize their votes (and hence, to be reelected) is to adopt an active role as opinion makers rather than adapting policy positions to accommodate public sentiments. Hence, the lack of support for May's hypothesis may have to do with the nature of party competition, indicating that the model has misinterpreted the dynamics of mass-elite linkages.

The idea of parties as advocates of conflict, which indeed is suggested by the alternative theories of issue voting, is of great empirical and theoretical interest for at least two reasons. First, if voters vote on the basis of affect and prefer parties to guide them ideologically, there is no reason for top elites to take a stand on policy issues identical to those of their voters. Quite the contrary, in order to attract voters, on salient issues they ought to be in a more extreme position than that of their own electorates. Second, since ideological purity is good vote-maximizing strategy, the interests of the top elites do not run contrary to those of the party activists. Consequently, the strategic dilemma of the top elite, as formulated by May, does not seem to hold true.

---

NOTES

1. This article was first presented for the workshop on Parliamentary Party Groups, ECPR Joint Sessions of Workshops, Oslo, March 27-April 3, 1996. The discussion at the workshop was helpful for the revision of the paper. In addition to the workshop participants, we want to thank Wolfgang Müller, Jo Saglie and Henry Valen for comments.

2. Audun Skare passed away in January 1998.

3. In his article, Kitschelt (1989, 402-6) lists six arguments that cast doubt on the validity of the simple leader-follower dichotomy. In the text, for the purpose of this article, we have listed only three of them.

4. Observe the difference between the present approach and that of Kitschelt. Whereas Kitschelt predicts the saliency of conflicts to vary between parties in different party systems (the Belgian ecology parties, Agalev, located in Flanders, and Ecolo, located in Wallonia), we expect the saliency of conflicts, and hence the probability of curvilinear disparities, to vary between parties within the same party system.

---

5. For a detailed description of the Norwegian party system, see Rokkan (1967); Rokkan (1970); Valen & Rokkan (1974); Valen & Urwin (1985).

6. The moral-religious dimension does not include the socio-cultural question of languages, which in recent times has been more or less absent from the political debate.

7. Several data sources are available for the construction of the policy space of parties. The European Party Manifesto Project (see Budge et al. 1987) covers the post-war election programs of parties in 19 West European States, whereas Harmel & Janda's more recent Party Change Project has developed judgmental data on actual issue positions for the period 1950-1990 for the nine most significant parties of the United States, United Kingdom and Germany (see, e.g., Harmel et al. 1995; Harmel, Janda & Tan 1995). In addition, Castles & Mair (1984) have produced left-right political scales based on expert judgment, whereas Laver & Hunt (1992), also on the basis of expert judgment, have constructed multidimensional policy scales. For our purpose, however, since our main interest is variation within (not between) the political parties, interview data from different levels of the parties are the most useful.

8. The candidate survey was sent to three categories of party members in three Norwegian counties (fylker). The respondents consisted of party delegates, members who were suggested as list candidates and the actual list candidates. The total number of respondents was 2,338, which gave a response rate of 76 percent. For the top elites, the response rate was 77 percent (of a total of 165).

9. The Norwegian Program of Electoral Research was founded in 1957, and electoral survey s have been conducted every fourth year, except in 1961, in connection with general elections (see Valen & Aardal 1994). The average number of respondents is around 2000.

10. Observe that, here, we make no distinction between 'sleeping' issues of importance to the parties and issues that are intensified, e.g., for strategic reasons during the electoral campaign. We simply argue that certain issues (the core issues) are of principal and ideological importance to the parties, and that their stand on those issues is not likely to be neutral. For instance, abortion and euthanasia are of principal relevance to the Norwegian Christian People's Party, and the party will take a strong anti position on these issues whether or not they are on the public agenda. For an intuitive discussion about the problem of emphasis versus position regarding electoral platforms, see Janda et al. 1995 and Harmel, Janda & Tan 1995.

11. Analyses of the 1993 party nominations indicate that the EU issue affected factional activity during the process of candidate selection (Skare 1994). Of no consequence to the theoretical argument of this paper, it could nevertheless be argued that this situation was likely to mobilize a large number of party militants. increasing the likelihood for mid-level extremism.

12. Observe that, in order to be consistent with the direction of left-right preferences (as they appear in Figure 1), the value numbers on this scale are reversed in relation to the wording of the question, so that 'leftist' views get the lowest value numbers (close to 1), and 'rightist' views get the highest value numbers (close to 10).

REFERENCES

Aardal, B. 1993. Energi og Miljø. Doctoral dissertation. Oslo: Institutt for samfunnsforskning.

Aardal, B. & Valen, H. 1989. Velgere, partier og politisk anstand. Oslo: Statistisk
Sentralbyrå.

Aardal, B. & Valen, H. 1995. Konflikt og Opinion. Oslo: NKS-forlaget.

Bartolini, S. & Mair, P. 1990. Identity, Competition and Electoral Availability: The Stability
of European Electorates 1885-1985. Cambridge: Cambridge University Press.

Bjørklund, T. & Hellevik, O. 1993. 'Et Valg i EF-stridens Tegn,' Tidsskrift for Samfunnsforskning
433-58.

Budge, I. & Farlie, D. 1983. Voting and Party Competition. London: Wiley.

Budge, 1., Robertson, D. & Hearl, D., eds. 1987. Ideology, Strategy and Party Change: Spatial
Analyses of Post-War Election Programmes in 19 Democracies. Cambridge: Cambridge
University Press.

Castles, F. & Mair, P. 1984. 'Left-Right Political Scales: Some "Expert" Judgments,'
European Journal of Political Research 12, 73-88.

Converse, P. 1975. 'Public Opinion and Voting Behavior,' in Greenstein & Polsby, eds.,
Handbook of Political Science, vol. IV. Mass.: Addison-Wesley.

Converse, P. & Valen, H. 1971. 'Dimensions of Cleavage and Perceived Party Distances in
Norwegian Voting,' Scandinavian Political Studies 6, 107-52.

Daalder, H. 1984. 'In Search for the Center of European Party Systems,' American Political
Science Review7&,92-\O9.

Daalder, H. & Mair, P., eds. 1983. Western European Party Systems. London: Sage.

de Swaan, A. 1973. Coalition Theories and Cabinet Formations. Amsterdam: Elsevier.

Gleditsch, N. P. & Hellevik, O. 1977. Kampen om EF. Oslo: Pax Forlag.

Harmel, R., Janda, K. & Tan, A. 1995. 'Substance versus Packaging: An Empirical Analysis
of Parties' Issue Profiles,' Paper, APSA Annual Meeting, Chicago.

Harmel, R., Heo, U., Tan, A. & Janda, K. 1995. 'Performance, Leadership, Factions and
Party Change: An Empirical Analysis,' West European Politics 18, 1-33.

Heidar, K. 1988. Partidemokrati på Prøve. Oslo: Universitetsforlaget.

Herrera, R. & Taylor, M. 1994. 'The Structure of Opinions in American Political Parties,'
Political Studies 42, 676-89.

Iversen, T. 1994. 'Political Leadership and Representation in West European
Democracies: A Test of Three Models of Voting,' American Journal of Political Science
38, 45-74.

Janda, K.., Harmel, R., Edens, C. & Goff, P. 1995. 'Changes in Party Identity: Evidence from
Party Manifestos,' Party Politics 1, 171-96.

Kitschelt, H. 1989. 'The Internal Politics of Parties: The Law of Curvilinear Disparity
Revisited,' Political Studies 37, 400-21.

Laver, M. & B. Hunt 1992. Policy and Party Competition. New York: Routledge.

Listhaug, 0., Macdonald, S. E. & Rabinowitz, G. 1990. 'A Comparative Spatial Analysis of
European Party Systems,' Scandinavian Political Studies 13, 227-54.

Macdonald, S. E., Listhaug, O. & Rabinowitz, G. 1991. 'Issues and Party Support in
Multiparty Systems,' American Political Science Review 85, 1107-31.

May, J. 1973. 'Opinion Structure of Political Parties: The Special Law of Curvilinear
Disparity,' Political Studies 21, 135-51.

McClosky, H., Hoffmann, P. & O'Hara, R., 'Issue Conflict and Consensus among Party

Narud, H. M. 1995a. 'Issue Saliency, Policy Distances and Coalition Bargaining,' West
European Politics 18, 361-88.

Narud, H. M. 1995b. 'Coalition Termination in Norway: Models and Cases,' Scandinavian
Political Studies 18, 1-24.

Narud, H. M. 1996. 'Electoral Competition and Coalition Bargaining in MultiParty Systems,'
Journal of Theoretical Politics 8, 499-525.

Narud, H. M. & Valen, H. 1996. 'Decline of Electoral Turnout: The Case of Norway,'
European Journal of Political Research 29, 235-56.

Norris, P. 1995. 'May's Law of Curvilinear Disparity Revisited,' Party Politics 1, 29^47.

Petrocik, J. R. 1996. 'Issue Ownership in Presidential Elections, with a 1980 Case Study,'
American Journal ofPolitical Science 40, 825-50.

Przeworski, A. & J. Sprague 1986. Paper S'.ones: A History of Electoral Socialism. Chicago:
Chicago University Press.

Rabinowitz, G. & Macdonald, S. E. 1989. 'A Directional Theory of Issue Voting,' American
Political Science Review 83, 93-121.

Rokkan, S. 1967. 'Geography, Religion and Social Class. Crosscutting Cleavages in
Norwegian Politics,' in Lipset, S. M. & Rokkan, S., eds., Party Systems and Voter

Rokkan, S. 1970. Citizens, Elections, Parties. Bergen: Universitetsforlaget.

Rokkan, S. & Valen, H. 1962. 'The Mobilization of the Periphery.' in Rokkan, S., ed..
Approaches to the Study of Political Participation. Bergen: Christian Michelsens Institutt.

Rose, R., ed. 1974, Electoral Behaviour: A Comparative Handbook. New York: Free Press.

Skare, A. 1994. Nominasjoner. Kandidalutvelgingens Politiske Innhold. Politisk Utvelging og
Politiske Endringer. Master's thesis. Oslo: Institutt for statsvitenskap.

Valen, H. 1976. 'National Conflict Structure and Foreign Politics: The Impact of the EC Issue
on Perceived Cleavages in Norwegian Politics,' European Journal of Political Research 4,
47-82.

Valen, H. 1990. 'Coalitions and Political Distances,' in Sånkiaho, R., ed., People and Their
Polities. The Finnish Political Science Association, 198-213.

Valen, H. 1994. 'Norway: A Storting Election in the Shadow of the EU." Electoral Studies
13, 169-79.

Valen, H. & Katz, D. 1964. Political Parties in Norway. Oslo: Universitetsforlaget.

Valen, H. & Urwin, D. 1985. 'De Politiske Partiene,' in Nordby. T., ed.. Storting og Regjering
1945-1985. Oslo: Kunnskapsforlaget.

Valen, H. & Aardal, B. 1994. 'The Norwegian Program of Electoral Research," European
Journal ofPolitical Research 25, 287-309.

Whiteley, P. & Seyd, P. 1996. 'Exit, Voice and Loyalty - the Dynamics of Participation in
the British Labour and Conservative Parties.' Paper. Panel on Party Activists in
Comparative Perspective, APSA Annual Meeting. San Francisco.

Whiteley, P., Seyd, P. & Richardson, J. 1994a. True Blues: The Politics of Conservative Party
Membership. Oxford: Clarendon Press.

Whiteley, P., Seyd, P., Richardson, J. & Bissell, P. 1994b. 'Explaining Party Activism: The
Case of the British Conservative Party,' British Journal ofPolitical Science 24, 79-94.