Resumé

The relationship between socioeconomic resources and political participation has been the subject of considerable normative-empirical debate. The present article addresses the problem on the basis of Norwegian survey data (N = 1,170) with the intent of testing both the resource-involvement relationship and the possible consequences of value over-representation among the most active citizens. The analysis of resource-involvement dependency in relation to Robert Dahl's norm-set for noncumulative resource distributions' shows only two resources to be of potential difficulty for pluralist expectations: education and gender. Other expectations as to resource-involvement inequality prove to be nonfounded for Norway. A further test of the consequences of pedagarchal and patriarchal over-representation shows the former to imply a general 'progressive' bias in political values, while the latter has no clear-cut ideological implications.

Introduction

Prior to recent developments in the field, the relationship between political participation and socioeconomic position was thought to be a well-established, and thus relatively uninteresting, phenomenon. In the first edition of Milbrath's standard overview of the literature (1965), we read (in boldface) that 'no matter how class is measured, studies consistently show that higher-class persons are more likely to participate in politics than lower-class persons' (p. 116). Further, it is stated that 'one of the most thoroughly substantiated propositions in all of social science ... [is] ... that persons near the center of society are more likely to participate in politics than persons near the periphery' (p. 113).

Both of the conclusions are retained in identical form in the revised edition of Political Participation (Milbrath & Goel 1977), but here the statement of class is followed up by data from cross-national studies which indicate that the classparticipation relationship in fact varies considerably from country to country.

---

* This article is a revised and shortened version of Report No. 16 from the project on 'Democracy in Norway'. The original report, which contains a number of highly relevant methodological subanalyses, is available on request. The author is grateful to W. Phillips Shively for his helpful comments, and to Ottar Hellevik for his challenging observations.

Whereas correlations are on a level of about .40 for the United States and India, they drop to about .30 for England and Italy; .20 for Germany and the Netherlands; and to .10 or less for Austria and Japan (the data are from Verba & Nie 1972, 340).

These differences are only briefly commented upon by Milbrath & Goel, but they actually mark a turning point for what has been termed the 'socioeconomic model' (Verba & Nie 1972, Zipp et al. 1982). Rather than a 'fact' to be taken for granted, the relationship between socioeconomic position and political activity now appears as a contextually determined condition which decades of American empirical research have shaped in its own image. The full realization of this emerged more clearly with the publication of Verba, Nie & Kim's seven-nation study in 1978, as well as with the addition of an eighth case, Norway, in the same year (Lafferty 1978). The latter study, and the book which followed it (Lafferty 1981a), have made the issue one of considerable debate in Norway (Martinussen 1977, 1983, Lafferty 1979, 1981, 1983a, 1983b, 1984, Rose & Waldahl 1982, Hellevik 1983a, 1983b), and it is in this context that the present contribution has its most immediate relevance. The goal of the study is not, however, to pursue the debate as a debate, but rather to put forth a position which builds on new data and which points toward a broader normative-empirical integration. My concern at this point is with a total assessment of the conditions promoting or retarding democratic citizenship in Norway, with an eye for more far-reaching comparative implications.

The Problem

A major goal of the project on 'Democracy in Norway' is to arrive at a general assessment of the status of democracy in Norway at the outset of the 19805.' Two preliminary conceptual steps toward this end were the choice of 'the social democratic state' as an appropriate systemic identifier, and the decision to replace the idea of 'political participation' with the broader notion of decisionmaking involvement' (Lafferty 1983b, 1984). Both decisions underline the fact that we are dealing with a special type of western polity, where social democratic governance has been the rule rather than the exception and where a central feature of the polity is widespread institutionalized decision-making access. I have both described the nature of this access and analysed its usage elsewhere, but want to reemphasize here that it is only access which is the issue in question. We will not be concerned with how decisions are actually made (the question of power or influence), nor will we be concerned with what decisions actually produce (the question of outcomes or distribution). We are, in other words, concentrating only on democratic input, with all the limitations and ambiguities which that implies. It is assumed, of course, that involvement is more successful in a self-interested direction than noninvolvement, but no attempt will be made to demonstrate the proposition here.

For many, this may appear to be an inconsequential exercise insofar as democratic assessment is concerned, and in an ultimate sense it is. An analysis of involvement alone is clearly not sufficient for a total systemic assessment, but it is of vital importance nonetheless. Not only is decision-making involvement an act of inherent worth for both individuals and communities alike, it is also an act with so many manifestations and connotations as to require piecemeal clarification before it can be adequately related to other democratic components. As we shall soon see, the delimitation of interest to but a single empirical aspect of democracy has in no way lessened the analytic-interpretive burden.

The task, therefore, is to assess the degree of equality manifest in the structure
of decision-making involvement in Norway. This will be done along two major
dimensions.

First, we will look at the relationship between involvement and socioeconomic position from the point of view of cumulative versus noncumulative resource effects. Assuming that higher and more central socioeconomic positions imply resources which promote more active and presumably more influential involvement, we will analyze the proposition in terms of singular, additive, and multiplicative resource effects. Taking our normative point of departure in the work of Robert Dahl (as set forth first in 1961 and subsequently revised in 1970, 1978, and 1982), we will try to determine the degree to which the involvement-resource relationship reflects liberal-pluralist norms. This amounts to answering three related questions: (1) To what degree is involvement a correlate of socioeconomic and center-periphery position? (2) To what degree are patterns of resource effects similar across different involvement areas? and (3) To what degree are positional effects 'cumulative' within and across the different involvement areas?

Having assessed the egalitarian nature of the involvement structure with regard to socioeconomic position, we will then go on to investigate the question of value representation. The goal here is to determine the degree of correspondence between value distributions for the population at large and for involvement 'elites' within and across the different decision-making sectors. Assuming that values which are presented by the most active participants are the values which are most likely to be realized through allocations, we will want to know how value preferences are distributed across different levels of involvement.

In sum, the analysis aims at a testing of involvement norms which are largely derived from instrumentalist liberal-pluralism. This emphasis is mainly determined by the earlier debate and is in no way intended to be exhaustive of the normative aspect. Other perspectives, such as involvement-as-creative-action and involvement-as-personal-development (Lafferty 1983b), will be taken up in later reports.

The Data

The data consist of a national sample survey carried out by the Norwegian Central Bureau of Statistics in early 1981 (see Faye & Opdahl 1981). Two characteristics of the survey which must be taken into consideration for the problem at hand are its composition according to age and sex.

As a major goal of our project was to investigate the effects of occupation on democratic involvement and political values, we delimited the sample at the outset to respondents between the ages of 18 and 65. As argued elsewhere (Lafferty 1981, 33-34), this delimitation probably implies a somewhat more conservative estimate of the total amount of involvement as well as the total amount of equality, since the age group between 66 and 75, at least, tends to be more active in several of the involvement forms to be considered here. The age group over 75 most probably offsets this bias, but, as this category is smaller than the group between 66 and 75, the overall effect would nonetheless tend in a conservative direction.

As for the problem of sexual bias, this arises as a result of a higher refusal rate among women than men (Faye & Opdahl 1981, 8-9). Whereas only 44 percent of noncompleted interviews among men were due to willful refusal, the figure for women was 73 percent. These figures were approximately the same when controlled for age and region, but the difference was particularly strong for the age group between 18 and 39 in rural areas. When interpreting these differences, emphasis must be placed on the fact that it is the refusal rate which is of interest, not the overall rate of noncompleted interviews (which stands at 29 percent with no significant sexual differences). That nearly double as many women as men actively refused to be interviewed can best be understood as a reaction against the title and substantive content of the interview. The women in question apparently felt that the subject 'democracy in Norway' (of which they were informed in an advance letter) was not one which they wished to talk about. If this interpretation is correct, we can expect a slight bias in the sample to the effect that noninterested and thereby probably inactive women are somewhat underrepresented.

In addition to these two basic characteristics of the sample, there are two other operational decisions which have direct relevance for the overall analytic framework. The first is the decision to limit the present analysis only to the age group between 30 and 65. This is done for the following reasons:

(1) The population between 18 and 29 is difficult to categorize meaningfully on a wide number of socioeconomic characteristics. The basic notion of socioeconomic position in the present context is that it is somehow 'structural', or at least strongly determinative, in the effects imparted. The majority of respondents between 18 and 29 are still 'in flux', both geographically (schooling) and socially, and thus of dubious structural validity. It seems wise, therefore, to accord them their own separate analysis in terms of their own set of structural

measures.

(2) An empirical check of the relationship between age and involvement shows that age 30 is in fact an important involvement threshold. With the exception of direct-action protest, the rates of activity for all areas are considerably lower for the under-30 age groups, a finding which Nie, Verba & Kim (1974) have substantiated for five different nations.

(3) The 'established' age cohorts between 30 and 65 are the shapers and bearers
of the democratic system. It is for this group of actors and potential actors that
the discussion of democratic norms is most crucially relevant.

The second delimiting decision has to do with the scope of involvement areas to be covered. In previous reports, we have identified and operationalized four major areas of decision-making involvement: political democracy, specialinterest democracy, occupational democracy, and protest (Laf ferty 1983b, 1984). Special-interest democracy has then been further operationalized in terms of pressure-group involvement, leisure-group involvement, and service-group involvement, and occupational democracy has been operationalized in terms of occupational interest-group involvement and workplace involvement. Of the seven areas operationalized, the present analysis will focus only on four, leaving out leisure involvement, service involvement, and workplace democracy. The exclusion of workplace democracy is due mainly to methodological reasons, since the availability of this involvement opportunity is differentially distributed and the analysis requires considerable clarification and manipulation of subgroups. As workplace democracy is such an essential part of our project, we will be treating this area more fully in a separate report.

As for leisure involvement and service involvement, these have been withheld primarily for the sake of reducing the complexity of the analysis. One can also argue, however, that these are areas which are more voluntary and less influential (on a system-wide basis) than the other areas considered. Leisure activities (sports, hobbies, etc.) and service activities (health organizations, charities, community service, etc.) can be of great individual importance — even in the instrumental sense focused on here — but they are only occasionally of importance in affecting the overall state of power and value distribution. As it is this latter context which has been the central referrent in the earlier debate, I choose to concentrate on those involvement areas which are nonambiguous in this framework.

The major dependent variables of the analysis thus consist of four indexes of decision-making involvement: Political Involvement (POLIT), Occupational Interest-Group Involvement (OCK-GROUP); Nonoccupational Interest-Group Involvement (NONOCC-GROUP); and Protest Involvement (PROTEST). (All indicators are presented in the Appendix in the order of appearance.)

As measures of socioeconomic position, we have available four different types
of resource indicator: material resources in terms of personal and/or family

income (Income), ownership of own dwelling (Housing), and ownership of vacation property (Cabin); personal resources in terms of Gender, Education, Age; class and labor resources in terms of Father's occupation and own Occupation; and locational resources in terms of regional location (Periphery/center) and type of community (Rural/urban).

As shown in the Appendix, all ten resource measures have been scored in accordance with the presumed value hierarchy of the patriarchal, capitalistic, centralized welfare-state. Thus maleness, ownership, occupational prestige, and proximity to the 'center' are thought to be positive factors for involvement in the same sense as higher education, income, and age. It is this notion of a common ranked value dimension across the different resource indicators which provides the ordinal logic underlying the normative-empirical problem, and allows, where desirable, for the use of parametric statistical analysis. We are primarily interested, in other words, in testing for additive and interactive effects in relation to linear hypotheses.

As for the analysis of value representation, we will here rely on a battery of questionnaire items which probe as to preference along 13 different value dimensions. Most of the probes are forced-choice items consisting of two, presumably opposing, statements. The areas covered range across the entire breadth of current political conflicts and provide an adequate basis for a general test of representativeness.

Socioeconomic Position and Involvement

The first step of the analysis is to establish the degree to which decision-making involvement varies in accordance with the individual resource distributions. The results of the analysis are shown in Table 1. (The indicator for respondent's own occupation is temporarily excluded from the analysis. It will be introduced separately below.)

The results in Table 1 show that only 15 of 36 relationships are significant at the level of p = .01.² The strength of linear association above this level varies from r = .11 to r = .24, with only three correlations above a level of .20. We are operating, in other words, within a range of explained variance between approximately 1 and 5.7 percent. Three of the relationships at or above r = .20 are for education, which is the only resource variable which shows a significant association with all four types of involvement. Most of the relationships are positive, but two of the significant relationships are negative: Age with Protest and Rural/urban with Political Involvement. The two 'ownership' variables — Housing and Cabin — are not significantly related to any of the involvement measures.

We are thus confronted with zero-order relationships of, at best, very
moderate dimensions. We need only refer to the comparative data cited above, or
to the fact that the strongest correlation among the resource variables themselves

is r = .40 (i.e., four times the explained variance for the strongest resource/involvement association), to realize that we are in a realm of relatively marginal data dependencies. On a one-to-one basis, the resource/involvement relationship is approximately what it was shown to be in the earlier studies, though there have been slight changes in the relative importance of the individual resource effects (particularly for education).

Before going into greater interpretive detail, however, let us move directly on to the question of relative effects. Do the associations shown in Table 1 reflect separate resource effects in each significant case, or are the associations reduced under conditions of mutual control? To determine which resources are ultimately most important (in an additive sense), I have carried out a multiple regression analysis with the results shown in Table 2. The models are based on bivariate relationships of r = .10 or above.

Table 2 provides us with information on the original bivariate association (r); the relative resource effect when standardized and controlled for all other variables in the equation (beta); and the relative importance of each variable in terms of its marginal contribution to the total amount of explained variance (R²). What we find is that eight of the resource variables with correlations of r = .10 or above are reduced to nonsignificance when controlled for the other variables in the analysis. Income and Father's occupation lose their significant effect in relation to all three types of involvement where they are included; Age falls out of the predictor model for Protest; and Gender — which is actually nonsignificant in relation to Occupational Interest-Group Involvement, but which was included anyway — declines even more. None of these variables contribute as much as a single additional percentage point of explained variance to their respective models.

We are thus left with four sets of predictors of varying size and strength. The

most powerful model is for Political Involvement with three significant resource predictors and a total explained variance of R2R² = .106, while the weakest model is for Occupational Interest-Group Involvement with but a single significant predictor and an explanatory potential of R2R² = .026.

There is, however, one major positional variable which is not yet in the analysis: the respondent's own occupational status. So as to get as clear a picture as possible of this highly robust 'class' indicator, as well as a more intuitive understanding of the effects of the significant variables shown in Table 2, we can conclude the initial phase of the analysis by shifting from a linear multivariate technique to a nominal technique. Tables 3 through 6 show the results of Multiple Classification Analysis (MCA) of each of the four involvement forms as predicted by the most significant variables shown in Table 2 in conjunction with Occupation. (See Andrews et al. 1973 for the parameters of the MCAtechnique). The most important results from Tables 3 through 6 can be summarized as follows:

Occupation: It should first be pointed out that the occupation indicator here employed is considerably different from the type of measure usually employed in studies of political participation. In addition to assigning housewives and 'househusbands' their own category ('Home/family'), we have distinguished between three levels of office and service employees, and two categories of selfemployed (including members of family firms and farms); one for business and the professions and one for the primary sector. The indicator is derived from a four-value standardized coding procedure (the 'Nordic Occupational Classification', 1965), and we have devoted considerable attention to a detailed recoding (Lafferty & Knutsen 1981).

At the most general level, we see that, with the exception of the model for
Occupation Interest-Group Involvement (Table 4), the controlled effects for
Occupation are all nonsignificant (F-test value at p < .01). Nonsignificant, that

is, as a categoric variable. The linear effects of a normal occupational prestige
scale are, of course, even weaker.

Secondly, we find that the effects of each occupationl category differ within
varying ranges across all four types of involvement. As the following ranges
indicate, all categories show at least one positive and one negative effect:

The ranges also show, however, that two categories in particular stand out at the higher end of the continuum: members of the crafts and trades and the selfemployed in farming and fishing. These groups show the only significant positive subcategory effects (as measured by regression analysis with dummy

variables), and the weakest and next-to-weakest negative effects. The only model where neither of these two subgroups shows the strongest positive controlled deviation from the grand mean, is for Nonoccupational Interest-Group Involvement, where the strongest effect comes from Top management.

As for the other categories, we note that the two strongest negative effects are for Business/professions (—.46),(^—.46), and Top management (—.32),(^—.32), both in relation to Occupational Interest-Group Involvement, and that the three remaining categories are all relatively stable and relatively weak. The two categories with the least positive effect are the two woman-dominated subgroups: Office/service labor and Home/family.

All in all, therefore, the relationship between occupation and decision-making is both weak and variable. The two categories of most consistent relative positive effect are the smallest in size and the most atypical in terms of the technicalindustrial division of labor. To the extent that occupational characteristics constitute involvement resources, they are of marginal — though interesting — importance.

Periphery/center: The results from the multiple regression analysis showed that the Periphery/center indicator and the Rural/urban indicator were only barely significant in relation to two types of involvement, Nonoccupational Interest-Group Involvement for the former and Political Involvement for the latter. So as to monitor these effects in the MCA-analysis, I have also included the Rural/urban indicator in the model for Nonoccupational Interest-Group Involvement. The results show that the inclusion was warranted; the rural-urban effect actually increases under conditions of mutual control.

What we find for the dimension as a whole is that there is a clear linear tendency for both types of involvement to decrease with increasing urbanization. For Nonoccupational Interest-Group Involvement, there is a slight tendency in the direction of lower activity levels for the two most northern regions, but the

linear effect for the country as a whole is clearly non-significant.

Gender: Gender was included in three models even though its relationship with Occupational Interest-Group Involvement was just below the significance level, Interestingly enough, however, it is only in relation to this area that the variable retains its independent effect. In relation to Political Involvement, the association declines from an eta equal to .22 to a beta equal to .14, and in relation to Protest, the change is from .13 to .05. The reductions are in both cases attributable to the control for occupation which includes the predominantly women's category of 'Home/family'.

Education: Finally, there is the variable of most consistent and persistent importance: Education. The MCA-analysis shows that the additional control for occupation does little to affect the dominant linear effect of schooling. The effect is attenuated for Nonoccupational Interest-Group Involvement (where it is only barely significant), but it is strengthened in relation to Occupational Group Involvement. The strongest effect is for Protest (beta = .25), where the difference between the second lowest and highest categories is equal to a predicted increase of 1.02 (—.23(^—.23 to + .79) in relation to the grand mean. This is also, it should be noted, the strongest predicted intra-variable difference of the entire analysis.

Having concluded the analysis of zero-order and relative additative effects, we
can now make a first normative assessment. What do the results reported imply
for the state of democratic values in Norway?

As mentioned above — and as argued for previously (Lafferty 1981,1983b) — the values of most direct relevance are taken to be those derivable from Robert Dahl's position on the difference between a system of 'cumulative inequalities in political resources' and a system of 'noncumulative' or 'dispersed' inequalities. As originally stated by Dahl (1961, 228), a system of dispersed inequalities is marked by the following six characteristics:

1. Many different kinds of resources for influencing officials are available to
different citizens.

2. With few exceptions, these resources are unequally distributed.

3. Individuals best off in their access to one kind of resource are often badly off
with respect to many other resources.

4. No one influence resource dominates all the others in all or even in most key
decisions.

5. With some exceptions, an influence resource is effective in some issue areas
or in some specific decisions but not in all.

6. Virtually no one, and certainly no group of more than a few individuals, is
entirely lacking in some influence resources.

It must be remembered that these standards were originally formulated from
the perspective of actual influence analysis at the peak level of decision-making

in a large American town. They were not meant to apply to the population as a whole in relation to decision-making involvement alone. As they still represent, however, the only attempt to formulate norms for the assessment of political resources and equality (that I am aware of), they serve as an adequate and interesting point of departure.

The first thing to notice with the set is that it presupposes inequality in individual resource distribution. The position makes no attempt, in other words, to deny that resource inequality exists. On the contrary, the opening sentence of Who Governs? is: 'In a political system where nearly every adult may vote but where knowledge, wealth, social position, access to officials, and other resources are unequally distributed, who actually governs?'

What the position does seem to presuppose, however, is a tendency for resource possession to be weakly interrelated. Those best-off in access to one kind of resource are seen as being 'often' badly off in relation to other kinds of resources. Given the conventional SES wisdom in this area, this might seem to be a highly unreasonable assumption. Two points are thus in order.

First, Dahl was not referring to resource relationships for the population as a whole, but for political voters at the level of broad-ranging policy decisions. At this level, it is not unreasonable to stipulate that voters with (e.g.) wealth are not particularly well endowed with either organizational resources, publicadministration resources, or electoral resources.

Of greater importance for our problem, however, is the fact that even at the level of the population as a whole, the presupposition is not that unreasonable for Norway. A brief check of the intercorrelations among the nine resource variables used in the bivariate analysis (Table 1) shows that less than half of the associations are significant (17 out of 36 at p = .01), and that more than a third of these are negative! Furthermore, it most be remembered that correlations reflect no more than tendencies. There will, even in the strongest relationships, exist a considerable amount of cross-trend variation. If we look, for example, at the relationship between education and income, which shows the strongest resource-resource association in the data (r = .40), and if we check for crosscategory congruity on the basis of high/low dichotomization of the variables, we find that 33 percent of those with the two highest education scores rank 'low' (scores 1-4) on income.

DahPs initial standards for a system of dispersed inequalities are thus not that problematical for the Norwegian situation. The essence of the norm-set lies, however, in the fourth and fifth characteristics, where the connection is made between resources and decisions. In 'translating' these standards to the problem of resources and involvement (i.e., without the benefit of actual decisional outcomes), I have interpreted Dahl to imply: (1) that no political resource should be clearly or consistently dominant in any one involvement area, nor (2) that any one resource should be dominant in its involvement effect across involvement areas. Different resources should, in other words, exert different effects on different

types of involvement at different times.

Before checking these norms against our data, it is perhaps necessary to clearly state that the standards in question are willfully negligent of simple bivariate relationships. They imply nothing at all for the assessment of one-toone resource-involvement associations. This is because there is nothing of a nonproblematical nature to be said about such associations from the point of view of either multivariate causality or pluralist democratic structure. The point might be thought to be obvious, but, as several contributions to the Norwegian debate have attempted to establish democratic inequality on the basis of bivariate relationships alone, it must be made again. Simple correlations may be impressive or nonimpressive in terms of explained variance, or they may be larger or smaller than similar correlations in other systems or at other times in the same system, but as indications of democratic inequality they are clearly nondecisive. Dahl's norm-set is stipulated for a system of dispersed inequalifres, and it is a stipulation of clear and necessary relevance for the highly democratized system at hand.

Returning, therefore, to an assessment of the multivariate results in light of the standards reformulated from Dahl, we have what appears to be a clear negative indication as to the existence of a system of dispersed inequalities. The effect of education, both within and across the four areas investigated, is clearly dominant as an explanatory factor. Under the conditions of the automatic stepwise procedure, it alone accounts for between 42 and 74 percent of the total explained variance in each involvement model (Table 2). Without it, there would be very little to point to in the way of systematic inequality, the effect of gender not withstanding.3

Summing up the analysis thus far, the following conclusions seem justified:

(1) In relation to the systemic perspective under consideration, there are at
least three highly interesting negative findings.

First, there is strong confirmation of the earlier results as to the importance of center/periphery placement. With the sole exception of a slight tendency toward lower levels of involvement in Nonoccupational Interest Groups in the country's two northermost regions (Table 5), there is no involvement disadvantage associated with regionalism at all. As for the rural/urban dimension, the findings are in a direction opposite to that of the general hypothesis. In every instance where the variable is at all significant, involvement increases with increasing rurality. Milbrath and Goel's center/periphery generalization (referred to above) is quite simply invalid for Norway. (These results from 1981 are the same as Martinussen's from 1969. See Martinussen 1977, Table 3.1, p. 45.)

A second area of negative findings is in relation to material resources. Neither degree of home ownership/control nor possession of vacation property show meaningful associations with involvement, and even the relatively weak bivariate associations for income prove to be spurious. Also here the results are in line with the earlier studies (Martinussen 1977, 61, 'Economic well-being', and Lafferty

1981, 69) and out of line with the other Milbrath-Goel generalizations. As a
general dimension, material well-being is of marginal importance for decisionmaking
involvement in Norway.4

Third, there is the question of occupation. Though it is more difficult to find exact comparisons or clear expectations as to inequality, it seems reasonable to assume that, of the occupational categories employed here, the social-position thesis would predict that higher management functionaries and independent businessmen and professionals should top the involvement hierarchy. The results clearly show that this is not the case, however. In only a single instance — Nonoccupational Interest-Group Involvement — do we find leading functionaries slightly more active (controlled effect) than the other groupings, while businessmen and professionals are never better than number three, and in two cases (Occupational Involvement and Protest) they are actually at the very bottom of the averages. If the notion of an 'upper-class' bias in involvement is to be 'saved' on the basis of occupation, it must be in terms of farmers, fishermen, tradesmen and craftsmen.

(2) The findings of positive importance for the problem of involvement inequality are thus limited to the area of personal resources. Even here, however, two of the four measures turn up insignificant. Neither age nor the personal characteristics which might attach to one's father's occupation show meaningful independent effects on levels of involvement. As to age, however, it must be recalled that our sample is limited to those between 30 and 65, and that the excluded oldest and youngest subsamples can be expected to have particular characteristics. From the point of view of democracy and personal (as opposed to group) involvement, however, the sample at hand provides sufficient grounds for a valid negative conclusion.

We are thus left with two variables of decisive importance: education and gender. We have also observed that the effect of education is both strong enough (relatively) and consistent enough to constitute a falsification of the dispersedinequality norm derived from Dahl.

Resource Representation and Values

Having narrowed the search for normatively problematical resource indicators down to education and gender, we can now turn to the question of consequence. What do the unequal effects of the two resource variables seem to imply for decision-making outcomes?

We have, of course, no data on actual outcomes, but we do have data on political values and we are willing to assume in the present case that values do influence decisions. Let us begin, however, by first depicting the combined effects of the resources in question in a clearer and more intuitive manner.

Table 7 provides us with subgroup averages for the five involvement indexes

broken down according to education and gender. (The reader need merely consult the Appendix to determine what an average involvement score means in terms of the different activity levels for each index.) So as to more clearly identify the profiles of the different effects, the relevant subgroups have been tested for significant differences of means.

Although the Table merely presents information which we have already analysed by other means, it does so in a new and interesting way. We discover, for example, that even though education is a significant linear predictor for all types of involvement, there are two types — Occupational Interest-Group Involvement and Protest — where the effect is not significant at the lower end of the: continuum. We see also that gender differences are significant in only four subgroups and that all four are in relation to either 'low' or 'medium' levels of education. Not only are gender differences nonsignificant for all high-education subgroups, but — much more dramatically — highly educated women are, on the average, more active than highly educated men in three of the five areas. The effects of 'maleness' are, in other words, mainly concentrated within Political Involvement among those with middle and low education. Among those with an education corresponding to gymnasium or higher (i.e., approximately 12 or more years of schooling), the involvement advantage shows signs of tipping toward women.

As for the effect of education itself, we see that, of the four areas of involvement, it is only Protest where the average level of activity increases more than a single index point from the lowest to the highest level. Of particular interest is the fact that the entire variation attributable to education in relation to Political Involvement is concentrated in the range between scores 2 and 3 on the index. Considering that the standard deviation for the Political Involvement index is equal to 1.30, we have a general idea of the dimensionality of the strongest resource predictor. (The greater range between 'low' and 'high' education in relation to the Cumulative Index must be seen, of course, in relation to an index range between 0 and 11. The standard deviation for the Cumulative Index is 2.69.)

The effects of the variables are nonetheless present, so let us conclude by trying to derive a picture of what the differences involved may mean. To do so, I have structured an analysis which aims at a test of value representation. Having shown that education technically violates the standards of a non-cumulative resource system, we are now interested in how the inequality in question is related to value congruity across the relevant subpopulations. First of all, what are the relevant subpopulations?

As nearly as I. can determine from the normative perspective presented by (e.g.) Hellevik (1983) or Verba, Nie & Kim (1978), we should be concerned with how 'high-resource' subgroups who are overrepresented among the most active participants may contribute to decision-making outcomes which are biased in the direction of that subgroup's resource interests. In the present case, this should

mean that highly educated male activists can be expected to promote values which are to the advantage of highly educated men. (One already begins to understand, I believe, that 'high education' is a more difficult resource to problematize in relation to equal representation than maleness, but let us return to that later.) So as to investigate this possibility, we can approach the data from two perspectives: First, we want to establish the actual degree of misrepresentation attaching to the resource-involvement relationships, and, second, we want to test for any value incongruity associated with the differences in question.

Table 8 provides figures on the överrepresentation of 'high-resource' citizens among the 'most-active' decision-making participants. So as to provide a maximum impact in regard to inequality, I have purposefully delimited the category of 'actives' to those who are truly most active. The analysis thus concentrates only on those who have attained the highest score on the three indexes for institutionalized involvement. Protest is left out of the analysis as there are, to my knowledge, no reasonable norms for equal representation among ad-hoc activists.

The results in Table 8 show that, whereas those with 'high' education (gymnasium or above) constitute 13 percent of the sample between 30 and 65, they constitute respectively 18 percent, 20 percent, and 19 percent of the 'most active' participants in the three areas of decision-making. Thus, in the 'worst case' of överrepresentation (occupational interest-group), there is an excess of 7 percent highly-educated in relation to the sample as a whole. In relation to political involvement, the difference is 5 percent.

The situation for gender is considerably more variable. Whereas the male disproportion in relation to occupational interest groups is fully 21 percentage points (71 to 50), there is no difference at all in respect to nonoccupational interest groups. Technically speaking, therefore, the latter area could be dropped from further analysis — seeing that representivity here is exactly what it 'should' be — but the Dahlsian perspective on cross-sector pluralism warrants its maintenance nonetheless.

But what about all the other disproportionate relationships? What do they
'mean' in terms of democratic norms?

First of all, it must be understood that their meaning is already operationally proscribed by an instrumentalist presupposition, i.e., we assume it is the highest degree of involvement which is of most decisive importance in relation to personal interests. Had we set the criterion for 'most active' only one score lower on the involvement indexes, we would have had much different (lower) representation ratios.

Second, it must be pointed out that any normative assessment of the overrepresentation involved must be viewed in relation to the actual numbers involved, and, further, that the numbers are different for the two resource dimensions. Whereas 'high-education' citizens are outnumbered by 5 to 1 among the 'most active', males have built-in majorities in two of the three areas. The stronger statistical relationship for education is thus actually less problematical for democracy in a procedural sense than is the weaker but numerically more consequent relationship for gender. (Another clear example of the importance of norm specification.)

Finally, there is the very general problem of determining whether all forms of
resource misrepresentation are equally problematical, or whether some forms

are more democratically reprehensible than others. Is any background or positional factor which proves (empirically) to be a 'resource' of equal merit in a representative normative context? Is underrepresentation of the more poorly educated of equal negative valence with underrepresentation of women? Are we equally unhappy with a pedagarchical bias as we are with patriarchal dominance?

The questions themselves indicate that the answers — wherever else they may be found — are unlikely to emerge from consensual democratic norms. Given the greater procedural implications of male dominance, I assume that most democrats would react more negatively on this dimension, but I also know that this is an area where personal assumptions are sooner confounded than confirmed. I leave the issue to the representation theorists, therefore, and move on to the final question: What does the överrepresentation imply for value preference and possible decision-making outcomes? Lacking consensual guidelines in procedural norms, we can at least see if there are differences in relation to political values, so that each may then struggle (or not struggle, as the case may be) with the clear trade-offs between democratization and ideology involved.

The procedure used to investigate this aspect of the problem is quite simple. Our survey includes 13 probes on values of central political interest. Ten of these probes are of the 'opposed-value' type, where the respondent is asked to choose between two statements believed to be of opposite valence and consequence, and the other three are of the 'graded-response' type, where respondents express degrees of agreement or preference in relation to a single value statement. The issues covered by the probes include the degree of statist intervention, economic regulation, income equality, the welfare state, worker-control, the role of referendums, farm subsidies, support for rural culture and traditions, economic growth, governmental decentralization, abortion, gender quotas, and the role of religion in politics (see Appendix).

To test for the degree of value representation on all of these issues, I have divided the sample into three different groups and tested for value differences by means of a simple T-test for difference of means (SPSS 'TTEST' procedure, SPS, 1975, pp. 267-274). The groups tested for are: (1) the overrepresented groups of 'high-resource elites'; (2) the rest of the 'most-active group' ('low-resource elites'); and (3) the rest of the sample ('the populace*). It will be noted that the test applied is an overly strict one as it does not measure differences between the overrepresented and the total population (where the mean for the latter includes the preferences of the former), but instead divides the sample into three separate and mutually exclusive groups. My reasons for doing this are both procedural (the demands of the 'TTEST' program) and substantive (an interest in 'worst-case' testing). The procedure has been applied to the two resource dimensions separately, and the results of the positive means-difference tests are shown in Tables 9 and 10. The results in these two tables are only for those instances where

the difference between the overrepresented 'elite⁵ group (the term 'elite' being used here solely as a stylistic convention) and the rest of the non-elite 'populace' is significant (with the significance level here lowered to p = .05) for at least one of the three decision-making areas. For the sake of comprehensiveness, the figures for all three areas are included when a single comparison is significant. We can thus determine — in line with the earlier discussion of pluralist norms — whether value differences are cumulative and reinforcing across the different involvement areas, or whether they are instead sector specific and possibly incongruous.

The first datum of most general interest in this regard is the large proportion of negative results. The analysis of both resource dimensions provides us at the outset with 13x3x3 = 78 problem-relevant comparisons. Of these, only 20 (26%) turn up significant at the generous .05 level. There are, in other words, no significant value differences between the overrepresented elites and the lessactive citizenry in nearly three-fourths of the cases tested.

A second general feature of the data is a clear difference in the way the significant findings are distributed for the two resource dimensions. The significant cases are equally distributed between the two (11 for education and 9 for gender), but whereas the important differences for education are concentrated as to issue and cumulative as to effect (only four values with nearly total cross-sector similarity), those for gender are both more dispersed and more sector-specific (eight values with no cases of total cross-sector significance, and only one of twosector significance). Education thus retains the clear cumulative profile which we saw in relation to involvement, but its effect is more limited in relation to issuerelevant values.

Finally, we can inquire as to what the significant value differences actually
consist of. In assessing these results, we wish also to direct our attention to the

other intergroup relationships involved: that between /ow-resource elites and the populace, as well as that between high-resource and low-resource elites. The different combinations of value preference among the three groups will have clear implications for an overall normative assessment.

All of the values in Tables 9 and 10 are scored in a 'progressive' direction. That is, each set of percentages expresses the proportion 'for' values which are viewed (in the eyes of the author) as being change-oriented in relation to more established or traditional positions. As the directionality imposed will hardly be consensual, the reader is advised to become acquainted with the actual probes in the Appendix.s

With this admonition in mind, the results in Table 9 indicate that higheducation elites are more 'progressive' than either the populace or their co-elites on all four of the significant value dimensions and across all three of the involvement areas. The differences are strongest in relation to support for the welfare state and weakest in relation to a greater preference for governmental decentralization. We also note, however, that the 'bias' in question is in the same direction as differences between the rest of the elite and the populace in 9 of the 12 areas covered, and that in three of those latter cases the difference is strong enough to also reach significance. We are talking, in other words, about an överrepresentation bias which: (a) is associated with a minority of the most-active citizens; (b) is clearly 'progressive' in nature; and (c) for the most part merely serves to strengthen an existing progressive trend among the most-active in general. In only two instances (both in connection with statist intervention in the economy) does the elite group which is underrepresented in terms of educational competence constitute less than an absolute value majority.

The situation for gender misrepresentation is much more complex. There is neither a general ideological bias across values nor a consistent gender effect across involvement areas. (It must be remembered that nonoccupational interestgroup involvement is not characterized by misrepresentation. The data are included in the table only for the purpose of an overall normative assessment.) There are in fact only two value dimensions where the trend across all three decision-making areas is clearly consistent. Male elites are less in favor of a delimitation of economic growth and more in favor of a delimitation of the role of religion in politics. In every other value area, there are either directly contradictory gender profiles across the involvement types or relations of no significance whatsoever. In the two value areas of most direct significance for gender politics — gender-related quotas and abortion — there were no significant differences at all in the first instance (though men prove to be consistently more positive to quotas), while the differences in the second instance are, as shown in Table 10, mixed. Women elites are slightly more pro-abortion within the occupational-interest-group sector, but the difference here is much weaker than either the clearly reversed preference in the nonoccupational-interest-group sector or the weaker difference among political elites. All in all, therefore, the results

shown in Table 10 indicate that any prediction as to the value-allocating effects of reduced inequality in decision-making involvement must be made with considerable circumspection. Gender is the involvement resource of most decisive consequence for procedural democratic norms, but its relationship with political values is typically pluralist.

Conclusions

As the separate empirical conclusions have already been summarized in detail,
I will conclude only with a discussion of the normative-empirical implications.

On a comparative basis, the results of the present analysis confirm earlier findings as to greater resource-involvement equality in Norway. Whereas Milbrath & Goel (1977) report correlations in the .30 to .40 range for other Western democracies, the strongest Norwegian correlation is .24, and no indicators other than education and gender are above .20. The relationship between socioeconomic position, as measured particularly by wealth, socioregional centrality, and occupation, is much less of a problem for democratic equality in Norway than it is in most other Western democracies.

What does remain of a problematic nature in relation to normative presuppositions, are the equality profiles of education and gender. Both of these characteristics exhibit what we might call 'necessary-minimum' signs of inequality in the present context. The effects of education on involvement are

'cumulative' in several of the senses here applied, while the effects of gender are
both weaker and more involvement-specific.

Beyond the question of variable effect, however, there is the problem of normative import. We have posed the question as to the meaning of educational inequality as opposed to sexual inequality. On an abstract level, are there democratic ideals which stress the desirability of equality across the educational hierarchy? Is it not instead a question of hoping that all can be both welleducated and active? In the absence of other economic or class-structural associations, is not education a quality which should be maximized among the most actively involved?

Gender, on the other hand, is clearly a 'structural' variable: a characteristic which most of us can do little with, and which we would like to see reduced to total equality. Here again, therefore, it appears to the present author that the involvement inequalities associated with gender are the most problematical from a normative standpoint, even though these inequalities seem to be somewhat reduced in relation to the previous findings. Norway thus remains, in this view, more deficient as a patriarchy than as a class society — at least in relation to decision-making involvement.

Even this generalization becomes blurred, however, when we take into consideration the final aspect of our analysis: the question of value representation. Responding to the critical prods of colleagues, an attempt was made to assess the possible consequences of the demonstrated resource inequality. Given the fact that education and gender seem to bias the probability of involvement, what does the bias actually imply in terms of value-issue preferences?

The answer proved to be of great but varied interest for the two resource dimensions. An överrepresentation of highly educated activists has no significance whatsoever in 9 of 13 value areas, and in the 4 areas where differences are significant, they are all in the direction of a 'progressive⁵ or 'leftist' bias. Highly educated activists ('elites') are more progressive than either their lesser educated elitist colleagures or the population at large in the areas of welfare-state support, statist intervention, farm subsidies, and governmental decentralization. The effect is, moreover, present across all three involvement areas studied. The implication is clear: Insofar as advocates of representational equality are also 'progressive' in a traditional (Norwegian) left-right sense, they are in a bind.

In somewhat less of a bind, but with no less confusion, are also equalityminded feminists. Överrepresentation of men is associated with value differences in 8 of the 13 value areas covered, but there can be no talk here of a unimodal bias. For most value areas, there exist contradictory tendencies across the different decision-making areas, such that there is no way to draw general implications for democratic norms. Of special interest in this regard, however, is the fact that the single significant women's value area — 'free abortion' — is characterized by pfofeminist male bias in two of the three areas. Overrepresented male elites are more liberal on abortion than both female elites and the populace in

both the political and nonoccupational interest-group areas. It is only in relation to occupational interest groups that highly active women are markedly more in favor of self-determined abortion vis-å-vis the populace (but not vis-å-vis male elites).

In addition to this single women's-related issue, there are two other trends where the value bias is consistent. Women elites are more in favor of both economic regulation and limited economic growth than either their male colleagues or the population at large, at the same time that they are equally consistent in preferring a greater degree of religiosity in politics. The situation appears promising, in other words, for religious-ecological feminists, but otherwise the push for greater gender equality would seem to demand serious progressive/feminist compromises.

Taken as a whole, the analysis would thus seem to confirm two of the basic conclusions of the previous studies. First, Norway clearly exhibits signs of exceptional equality in regard to the relationship between socioeconomic position and decision-milking involvement, whether one views the problem from a comparative or normative perspective. Second, it is the problem of gender bias which poses the most serious structural challenge to democratic norms in Norway, even though — as here illustrated — it is both a problem undergoing change and one of dubious ideological valence.

---

NOTES

1. The project has been funded by the Norwegian Research Council for Science and the Humanities, the Norwegian Ministry of Local Government and Labor, and the Institute of Political Science;, University of Oslo.

2. Both the linear product-moment approach and the level of significance (p = .01) are specifically chosen as appropriate for the normative-empirical problem at hand. The goal is to be liberally inclusive without being trivially cluttered, basedon guidelines from the normative dialogue.

3. The notion of cumulative inequality tested for here is but one of the aspects of the general phenomenon discussed in the Norwegian debate. Three other notions — identified as the 'cumulative resource effect', the 'cumulative involvement effect', and the 'interactive resource effect' — have been tested for in the original report. The findings for all three tests failed to change the general conclusion here stated. Regardless of how cumulative inequality is measured or tested, it is primarily education and secondarily gender which provide problems of any magnitude for the equality norms here specified. Interested readers should consult the original report as cited above, pp. 21-31.

4. For the sake of normative clarity, let me once again state that 'marginal importance" in this connection has to do with the overall degree of association for the Norwegian system as a whole This is definitely not to say that (in this case) material well-being cannot be of considerable importance for involvement at the margin, i.e., for that very small proportion of Norwegian citizens who are materially destitute. There are those who apparently feel that so long as any citizens are adversely affected by resource deficiencies at the margins, there is a problem with democratic equality. It is the opinion expressed here, however, that such a feeling builds on a premise of absolute distributional equality which is neither normatively nor empirically defensible.

5. For a better understanding of the contextual nature of the probes and the ideological directionality employed here, see Lafferty & Knutsen 1984.

REFERENCES

Andrews, F., Morgan, J. & Sonquist, J. 1973. Multiple Classification Analysis. Institute for Social
Research, University of Michigan.

Dahl, A. Who Governs? 1961. New Haven: Yale University Press.

Dahl, A. 1978. 'On Removing Certain Impediments to Democracy in the United States', Dissent,
Summer, 310-324.

Dahl, A. 1982. Dilemmas of Pluralist Democracy: Autonony vs. Control. New Haven: Yale
University Press.

Faye, A. & Opdahl, S. 1981. 'Rapport om gjennomføringen av undersøkelse om demokrati i Norge,
Interne Notater. Oslo: Statistisk Sentralbyrå.

Hellevik, O. 1983a. 'Politisk deltakelse i Norge — Begrenset og skjev?' Tidsskrift for Samfunnsforskning
24, 3-30. (a)

Hellevik, O. 1983b. 'Svar til W.M. Lafferty', Tidsskrift for Samfunnsforskning 24, 397-398.

Lafferty, W.M. 1978. 'Social development and political participation', Scandinavian Political Studies
1, 233-254.

Lafferty, W.M. 1981. Participation and Democracy in Norway: The 'Distant Democracy' Revisited.
Oslo: Norwegian University Press.

Lafferty, W.M. 1983a. 'Politisk deltagelse i Norge", Tidsskrift for Samfunnsforskning, 392-396.

Lafferty, W.M. 1983b. 'Political participation in the social-democratic state: A normative empirical
framework for the analysis of decision-making involvement in Norway', Scandinavian Political
Studies 6 (New series), 281-308.

Lafferty, W.M. 1984. 'Decision-making involvement in Norway: The nature and scope of citizen
access in a social democratic polity', European Journal of Political Research 12, 43-58.

Lafferty, W.N. & Knutsen, O. 1981. Yrke som variable: Noen erfaringer med å bruke 'Nordiske Yrkesklassifisering' som grunnlag for oppbygging av yrkesvariabel og et foslag til løsning. Report No. 6, Project on 'Democracy in Norway: Participation and Basic Values', Institute of Political Science, University of Oslo.

Lafferty, W.M. & Knutsen, O. 1984. 'Leftist and rightist ideology in a social democratic state: An
analysis of Norway in the midst of the conservative resurgence", British Journal of Political
Science 14, 345-367.

Martinussen, W. 1977. The Distant Democracy. London: John Wiley & Sons.

Martinussen, W. 1983. 'Demokratiets sosiale forutsetninger'. Statsvetenskaplig Tidskrift 3, 239-242.

Milbrath, L.W. & Goel, M.L. 1977. Political Participation (2nd edition). Chicago: Rand McNally.

Nordic Occupational Classification 1965. (Nordisk Yrkesklassifisering. Standard for yrkesgruppering
i offentlig norsk statistikk). Oslo: Central Bureau of Statistics.

Rose, L.E. & Waldahl, R. 1982. 'The distribution of political participation in Norway: Alternative
perspectives on a problem of democratic theory', Scandinavian Political Studies 5 (New series),
285-314.

Verba, S. & Nie, N.H. 1972. Participation in America: Political Democracy and Social Equality. New
York: Harper & Row.

Verba, S.; Nie, N.H. & Kim, J. 1978. Participation and Political Equality: A Seven-Nation Comparison.
Cambridge: Cambridge University Press.

Zipp, J.F., Landerman, R. & Luebke, P. 1982. 'Political parties and political participation: A re-