Scandinavian Political Studies, Bind 1 (New Series) (1978) 4Social Development and Political Participation: Class, Organizations and SexWilliam M. Lafferty, University of Oslo Side 233
ResuméThe relationship
between social development and political participation
has been The prevailing academic view of the relationship between social development and political participation was firmly established by Nie, Powell, and Prewitt in 1969 (1969a and 1969b). Beginning with the following 'general theory', they arrived at conclusions which have since become standard footnote fare for the burgeoning literature on political participation: "Economic
development alters the social structure of a nation. As
nations become * The author would like to thank Erik Allardt, Ole Berg, Knut Heidar, Ottar Hellevik, Lester Milbrath, Nils-Eyvind Naas, and Marvin Olsen for their helpful comments on the original paper. Side 234
concentrated in the urban areas; and (3) the density and complexity of economic and secondary organizations increase. These social changes imply political changes. Greater proportions of the population find themselves in life situations which lead to increased political information, political awareness, sense of personal political efficacy, and other relevant attitudes. These attitude changes, in turn, lead to increases in political participation" (l%9b:808). Despite the numerous citations of Nie, Powell, and Prewitt, however, very few studies have either retested or critically assessed their findings from a developmentalist perspective.1 The standard practice is rather to employ their conclusions as either foil or support for more static analyses of within-nation participatory phenomena. Thus the two most recent large-scale studies of participation - Verba and Nie's Participation in America (1972) and Martinussen's The Distant Democracy (1977) - both rely on Nie, Powell, and Prewitt without, however, touching on the developmentalist aspect of their approach. The purpose of the present study is a modest corrrection of this developmentalist neglect. The strategy is to look at the essentials of the Nie position from the perspective of participation in Norway. The goal is to show that, from a developmentalist point of view, the indicators and conclusions of the Nie-Powell-Prewitt approach are in need of serious revision if they are to have relevance for a highly developed society like Norway. In addition, however, I will briefly speculate: on the possibility that the need for revision is not only a question of Norway's advanced developmental status, but perhaps equally a question of a basic methodological bias in the original five-nation study; namely, the treatment of housewives in the construction of indexes. If the latter should prove to be the case (a test can only be carried out on the original data), we are faced with the possibility that Nie, Powell, and Prewitt's oft-cited conclusions are not so much a reflection of social development as they are a reflection of traditional sexist methodological bias. 1. Development and Participation: The Nie-Powell-Prewitt ConclusionsThe most
essential conclusions from the Nie, Powell, and Prewitt
study Urbanism: As
indicated in the 'general theory' stated above,
urbanization Side 235
was thought to be an important contributing condition for increased rates of participation. Nie, Powell, and Prewitt concluded, however, that this was not the case. The original correlations between participation and urbanization were so low (regardless of various types of control) that they decided to drop the variable from the second part of their analysis. Class: The notion that greater mobility leads to larger relative proportions of middle and upper class citizens, and that this in turn leads to higher rates of political participation, is said to be confirmed by the study. Social status correlates with participation within a range of. 18 (Germany) to .43 (U.S.), and the ranking of class structures according to low, middle, and upper proportions corresponds exactly with a ranking of proportions of'politically active' (1969a:Table 1, p. 364 and 1969b: Table 6, p. 824). Considered of even greater importance, however, is the finding (by means of path analysis) that the class-participation relationship is not a direct relationship (statistically), but rather a relationship which is mediated by specific political attitudes. When attitude variables are included in a causal model with indicators of class, organizational involvement, and participation, the original correlation between class and participation disappears entirely, leaving the major part of the explanation to causal paths involving either attitudes or organizational involvement. Nie, Powell, and Prewitt interpret this as evidence for the contention that upper and middle class positions expose the individual to both citizen norms and greater political information; factors which in turn increase feelings of both political efficacy and political interest, which then result in actual political involvement. Organizations: Finally, there is the important finding that the most significant factor of those considered is involvement in the 'organizational infrastructure'. (By the organizational infrastructure, Nie, Powell, and Prewitt mean voluntary organizations, informal group activities, and participation in the labor force.) The zero-order correlations between participation and organizational involvement vary between .48 (Germany and U.K.) and .52 (U.S.), and the path analysis shows that most of this effect is direct, i.e. not mediated by the so-called 'civic attitudes' (1969a:Table 1, p. 364 and 1969b: Table 2, p. 815). It is the conclusions on class and organizations which have been ritually appended to subsequent literature in the field, and it is these two conclusions which will be emphasized here. For the sake of the problem, however, let me first make some brief remarks on urbanism. First of all,
it is of interest to note that Martinussen's study of
participation Side 236
the contextual variable of urbanism is not at all correlated with political participation (1977:Table 3.1, p. 45). On the contrary, there is a slight negative relationship, with participation tending to increase with decreasing size of community. Within Martinussen's theoretical framework - which postulates "centrality" as an important participatory "resource" - this finding represents a clear rejection of the presumed participatory inequality inherent in the center-periphery distinction. Obviously this is a finding of direct relevance for developmentalist theory (especially in the land where 'center-periphery' was first offered (Rokkan and Valen, 1962) as an important analytical concept), but Martinussen makes no conclusions on the matter. His lack of interest is more understandable,, however, than is that of Verba and Nie (1972). Given the direct overlap in authorship, one would have thought that Verba and Nie would be as interested in the developmentalist aspect of the problem as Nie, Powell, and Prewitt originally were. Such is not the case, however. In their chapter on 'the community context of participation,' where they test out a 'mobilization model' and a 'decline-of-community model', Verba and Nie cite the original Nie-P'owell-Prewitt study as evidence of the lack of relationship between the two variables, but make no mention of the developmentalist theory, nor do they make any further reference to the original study in their analysis and conclusions. This is a shame, since their analysis is the most comprehensive carried out to date, and their conclusions that the mobilization model does not hold up is (again) of obvious importance for developmentalist theory. Space does not
allow for further comment, however, so I will simply
'Participation does indeed decline as communities grow, and, more clearly, as they begin to lose the clear boundaries that separate them from other communities. Participation in general and communal participation in particular are more widespread in more peripheral and isolated places. As one moves to the 'center' of society, such activity is inhibited' (1972:242). 2. Development and Participation in NorwayNorway's status as a highly developed nation is indisputable. As indicated in Table 1, which uses the same nations and types of indicators as those employed by Nie, Powell, and Prewitt, Norway is second only to the United States in both per capita production and service-sector employment. Side 237
![]() Table 1. Indicators of social development for six selected countries In terms of newspaper circulation per 1000 (which here replaces the 'literacy' measure used by Nie, Powell, and Prewitt), Norway is second only to the United Kingdom. When discussing participation in the Norwegian context, therefore, we are clearly viewing the problem from the leading edge of socioeconomic development (for better and for worse). The same can also be said for levels of organizational and political involvement, but here the lucidity of the case, and its relevance for the Nie-Powell-Prewitt position, are somewhat more problematical. The major difficulty - and one of prime importance for the developmentalist perspective - is the status of organizational involvement. In short, is such involvement to be considered an independent or dependent variable? For Nie, Powell, and Prewitt, involvement in the organizational substructure is the most decisive independent variable affecting political participation, with the latter defined primarily in terms of traditional, electoral-sector politics. For Willy Martinussen, however, who is forced in his study of Norway to take account of Norway's highly developed corporate-pluralist structure, organizational involvement is treated as both participation in its own right, as well as a resource/or participation. He thus employs (rather problematically as we will see) similar indicators of organizational activity on both the independent and dependent sides of the implicit causal model. So as not to
overburden the discussion at this point, let us simply
say Side 238
Whereas Nie, Powell, and Prewitt may have been justified in treating the factor as only an independent variable in 1969, it is no longer justified to do so today; at least not in countries with both high levels of organizational membership and a well-established integration of organizations in the decision-making process. The first corrective to the developmentalst perspective, therefore, is that organizational involvement is both participation and potential conditioner of participation, and that it is up to us to distinguish the two aspects operationally. How this can be done will be taken up below, but first we must return to the question of participatory levels in Norway. In comparing their five nations on the two types of involvement, Nie, Powell, and Prewitt rely on only two rankings. The first is a measure of'organizational density', which ranks nations according to the proportion of citizens belonging to organizations, and the second is a measure of the 'overall participation rate', which ranks nations according to the proportion classified as 'politically active' (1969b:Table 6, p. 824). As near as I can determine, the first indicator refers to membership in at least one voluntary association, while the second is based on having (at a minimum) discussed politics at least once a week. The range of organizational density was from 55 percent for the United States to 20 percent for Mexico, and the range of overall participation was from 46 percent for the U.S. to 25 percent for Mexico. Norway's position on the first dimension is easily established. Martinussen's data from 1969 show that at least 70 percent of the Norwegian population are members of at least one organization (1977:52), and my own survey from 1974 shows that the proportion in industrialized areas is closer to 75 percent. Victor Pestoff, in his comprehensive study of voluntary associations and Nordic party systems, provides data for 12 nations (including the five countries of the Nie-Powell-Prewitt study) which show that Norway is second only to Sweden in terms of membership in economic organizations (51 percent to 35 percent). It is not surprising, therefore, that organizational analysts in Norway refer to their topic as 'the organized society' (Hallenstvedt and Moren, 1975). The question of levels of'politically active' is, of course, more controversial, due to differences in criteria as to where to set the 'active' cutpoint. Willy Martinussen (1977:32) presents a rough political-stratification' scheme for Norway which classifies only 10 percent as 'active', 50 percent as 'passive,' and 40 percent as 'apolitical'. This image cannot be taken seriously in the present context, however, since it obviously reflects Martinussen's domestically oriented critical persuasion. A more relevant comparative perspective is provided in Lawrence Rose's recent study of Side 239
participation in two Norwegian municipalities, where, in comparing Norway with five other countries (Austria, Japan, Netherlands, the United States, and India), he concludes that Norwegian participatory levels are equal to or higher than the other cases for nearly all types of activity (1976:175). Without further ado, these comparisons should serve to establish Norway's status as a participatory polity, at the same time that Jhey confirm the general validity of the basic Nie-Powell-Prewitt premise. If the Norwegian indicators of development, organizational involvement, and participation were today added to the rankings of the original five nations, the impression of a strong correlation between development and participation would be clearly strengthened. The next question is to see if this also holds true for the posited reasons for this relationship, i.e. the conclusions on class and organizations. 3. Class and OrganizationsIn analyzing the relevance of Norwegian conditions for the Nie-Powell position, I will begin with a brief critical presentation of the results from Martinussen's Distant Democracy, and then go over to my own data from three industrialized Norwegian communities. The aim of Martinussen's study is to show that ideals of democratic participation are not being realized in Norway. In trying to make a case for widespread participatory inequality, he gives the impression that both social status and organizational involvement are important determinants of political participation, and that both types of resource are unevenly distributed in a systematic and cumulative manner (pp. 60-63). This would seem to support Nie, Powell, and Prewitt, and indicate that Norway represents but another case of the oft-cited conclusions stated above. My position, however, is that the reality of the situation is quite different. As I see it, the variables in question are related as follows: (1) Class is not an important determinant of political participation in Norway. (2) Organizational involvement in interest groups, which in Norway must be considered as instrumentally effective political participation, is not seriously affected by class either, nor is it seriously affected by involvement in voluntary associations which are not specifically interest-oriented. In the following, I will try to support these views by, first, arguing that Martinussen's data either actually support them or are inconclusive due to methodological difficulties, and, second, showing that they are verifiable with my own survey data. Space does not allow for a detailed critique of Side 240
Martinussen's
analysis, so I will rely only on a series of summary
points. (1) Martinussen uses three indicators which directly reflect social class: education, 'economic well-being' (a composite index of income and ownership), and a measure of 'control over work situation' (a ranking of occupations according to presumed free-time). The latter is particularly interesting, since Martinussen here breaks with standard methodological practice and assigns 'housewives with children at home' their own (lowest) category on the index. Other housewives, however, are totally left out of the index, and thereby also excluded from the multivariate analysis. (2) As already mentioned, Martinussen covers organizational involvement with both dependent-variable and independent-variable indicators. The former is a measure of 'organizational activity' which taps degrees of actual participation in parties and other voluntary associations (excluding occupational interest groups), while the latter is a measure of organizational membership' which taps simple membership in voluntary associations (including occupational interest groups). Party activity and membership are weighted highest by both indicators. (3) In addition to the index of organizational activity, Martinussen also measures participation by indexes of'representational activity' and 'direct political action'. These two measures actually split the participatory dimension measured by Nie, Powell, and Prewitt, with the latter corresponding directly to items 2 and 3 of the Nie-Powell-Prewitt composite index, while the former covers the same types of activity (party involvement and campaigning) which are covered by Nie, Powell, and Prewitt's item 5. Given the fact that Martinussen's index of representational activity is most similar to the type of'mass democratic participation' implied by Nie, Powell, and Prewitt's theory (if not their indicators), and considering further that Martinussen's index of direct action is excessively skewed (a two-value measure which is distributed 88 to 12) and thus highly problematical from a methodological point of view, I will concentrate here only on the findings for organizational and representational activity. Furthermore, I will concentrate only on the results shown by the eta coefficients of the MCA analysis, as these provide (in my opinion) the most accurate indicators of correlation.2 (4) The findings, which are here reproduced in Table 2, show that there are only two relationships of major importance, both of them containing the indicator of organizational membership. At first glance, it appears as though the Nie-Powell-Prewitt results are duplicated on this dimension. But, as Martinussen himself is forced to admit, the relationship between organizational membership and organizational activity is largely tautological Side 241
![]() Table 2. Relationship between class, organizational involvement, and political participation (p. 62). Just how tautological can be illustrated by a simple experiment with my own data, where the eta coefficient between two measures which are logically identical with Martinussen's is exactly .59, but where the removal of party membership alone from the membership index is enough to halve the coefficient to .29. Similar controls for the other organizational types would obviously reduce the correlation even more. I would also maintain, however, that the same type of effect is also present in the relationship between organizational membership and representational activity, although not quite so blatantly. Here party membership is inclusively covered by the high categories of the membership index, at the same time that party-dependent activities (nominating party candidates and holding party office) dominate the highest category of the representational index. Nie, Powell, and Prewitt are careful to exclude party membership from their measure of organizational involvement so as to avoid this type of autocorrelation, but Martinussen has not been so discriminating. All in all, therefore, I feel that Martinussen's findings on organizational involvement are too strongly affected by conceptual and methodological overlap to be clearly interpreted. As we will see below, a closer attention to the distinctness of different participatory-organizational sectors produces considerably different results. (5) As for the relationship between social status and participation, Martinussen's results show that only a single coefficient is stronger than .20 (i.e. approximately 4 percent explained variance). Considering the fact that the eta coefficient expresses both linear and nonlinear relationships, these are nonimpressive findings indeed. Furthermore, as we will see below, the effect from control over work situation is less a result of Side 242
occupational
status and/or temporal privilege than it is a reflection
of male In short, we have a situation where the effects of organizational involvement on participation are unclear, and where the effects of class are seemingly nonimportant. In trying to clarify the first; and establish the second, let us now turn to data from my own study of three industrialized communities. Surveys in the three areas, which lie in three different regions of southern Norway, were carried out by the Gallup organization in 1974.3 Although the results are not directly representative of Norway as a whole, I have as yet found no reason to question their general relevance. As near as I can
determine, the interrelationships among the variables in
The first task was to operationally clarify the question of organizational involvement. I have done so by means of three different organizational indicators: (1) Membership and activity in political parties is treated as a key item in an indicator of electoral activity.s By 'electoral activity' is meant those activities, primarily party-related, which aim at influencing arenas of electoral-representative decision-making. (2) Membership and activity in interest groups is treated as a separate dependent-variable indicator of interest-group activity.6 By 'interest group' is meant any organization which can reasonably be said to be constituted primarily for the purpose of promoting and maintaining a specific type of politically dependent group interest. (3) Membership in voluntary associations which are not political parties or interest groups, is treated as an independentvariable indicator of organizational membership.1 The most common types of this mode of involvement are sport and leisure organizations, service and health organizations, and cultural, educational, and religious organizations. As measures of class, I have employed the three standard SES indicators: education, income, and occupational status.8 The measures for education and income are straightforward rank-ordered scales, but the indicator of occupational status is nonconventional in its treatment of housewives. Instead of assigning housewives to their husband's occupations (as is done by nearly all studies in sociology and political science, including Nie, Powell, and Prewitt and Verba and Nie), I have assigned housewives to their own categories at the low end of the index, distinguishing between those with no wage-earning employment at all (score of 0) and those with a part-time job (score of 1). The normative grounds for Side 243
such a
procedure are obvious, but my theoretical reasoning is
that the Table 3 provides
the results of a multiple-regression analysis for two
![]() Table 3. Multiple regression analysis of the relationship between political participation, class, and organizational membership. Community sample and subsample of the 'actively employed'. relationship: that between occupational status and interest-group activity (r = .37). All other relationships are relatively weak, with none of them reaching a level of 4 percent explained variance. Organizational membership is the strongest predictor of electoral activity, but its effect is modest to say the least. Education and income are marginal in both their absolute and relative effects. We are left,
therefore, with the single relationship between
occupational Side 244
measure, we are
immediately interested in knowing how much of the
An answer is quickly provided by running an analysis on the same set of indicators for only the 'actively employed'. (For the sake of comparison, I have also run a subsample analysis for electoral activity.) We see that, when housewives are removed from the analysis, the effect of job status all but disappears entirely for both types of political engagement. We are left with two very weak models, a fact which, in my opinion, clearly strengthens the critical interpretation of Martinussen made above, and serves to establish the following general conclusions for Norway: (1) organizational involvement must be distinguished according to interest-group and noninterest-group activity; (2) the latter is not strongly related to the former, nor is it strongly related to electoral activity; and (3) class is not an important determinant of political participation in either of the two major channels of influence. 4. SexBut we can hardly stop there. In trying to establish the relevance of the Nie-Powell-Prewitt position for Norway, we have revealed the vital importance of sex for at least one key aspect of class: occupational status. The results of the analysis in Table 3 show that the most important aspect of the division of labor is not the job hierarchy, but having a 'job' in the first place. If 'nonworking' housewives are left out of the analysis, job prestige is shown to be meaningless. What we want to know now, therefore, ![]() Table 4. Multiple regression analysis of the relationship between political participation, class, organizational membership, and sex. Community sample (N = 753) Side 245
is the direct
effect of sex in relation to both occupational
membership The results are shown in Table 4. It has been necessary, of course, to exclude the occupational measure from the analysis so as to avoid the obvious autocorrelation between job, status and sex (r = .60). Aside from this technical necessity, however, the exchange actually results in a threefold gain in information: (1) conceptually, the variable of sex is both more easily interpre table and (apparently) of greater theoretical relevance for the problem at hand; (2) methodologically, we regain those respondents who proved impossible to rank on occupation (students, pensioners, and invalids), thus increasing the sample size for the multivariate analysis; and (3) empirically, sex proves to be a better overall predictor. The latter advantage is made clear by the increase in R2R2 for both models; from .05 to .09 for electoral activity, and from . 15 to .18 for interest-group activity. Sex not only exerts an effect exactly equal to occupation in the latter model, but actually serves to improve the relative effects of the other indicators. The beta for income is particularly strengthened (though still not of major importance), indicating that the measure picks up those occupational effects which are not covered by sex. In relation to electoral activity, the relative effect of sex is double that of occupational status, at the same time that the effect of organizational membership is totally unchanged. The two remaining class indicators are only slightly altered. In sum, sex is the most important 'structural' determinant of political participation in Norway. The effect is stronger for interest-group activity than for electoral activity, but nonetheless dominant in both areas. Furthermore, the beta's of Table 4 show that the effect is nonspurious, i.e. not dependent upon either socioeconomic status or organizational involvement. There is, in other words, something about gender which influences participation regardless of social position or nonpolitical organizational activity.9 5. Civic AttitudesWe have now shown that the direct relationship between class and political participation is negligible in Norway, and we have demonstrated the necessity (and consequences) of a more rigid operationalization of organizational involvement. The remaining point to be discussed is the question of civic attitudes. As stated above, Nie, Powell, and Prewitt concluded that such attitudes decisively mediate the relationship between status and participation, but are relatively nonimportant for the relationship between Side 246
organizational involvement and participation. Our point of departure for testing out this aspect of the problem must, of empirical necessity, be different from that of Nie, Powell, and Prewitt, but the results prove to be relevant and revealing nonetheless. The analytical strategy chosen is to construct recursive causal models of the type used by Nie, Powell, and Prewitt for both the electoral and interest-group modes of activity (Table 5). In addition to the participatory and structural indicators used above, I have included three attitude measures which are labelled perceived government impact, political efficacy, and political attention, in line with three of the five measures used by Nie, Powell, and Prewitt.10 Data for the other two measures ('citizen duty' and 'political information') are not available in the community survey. ![]() Table 5. Path coefficients (beta's) for eight-variable recursive model. Community sample (N = 739) The presumed causal logic of the models is implicit in the ordering of the indicators, which represent successive regression equations. Sex and education are given primary causal importance and first tested for their effect on income. These three are then tested for their effect on organizational involvement, and so on. The attitude variables are included in the same order as that employed by Nie, Powell, and Prewitt (1969b: Table 2, Side 247
p. 815). The coefficients are standardized regression coefficients (beta's), which is to say that they indicate the amount of change (expressed in units of standard deviation) produced in the dependent variable by the independent variable when all the other independent variables are controlled for. For ease of interpretation, I have left out coefficients smaller than .08 (F-ratio nonsignificant at .01). The three most
important results of Table 5 can be summarized as
(1) Of the three attitude measures, only one - political attention - has any direct effect on participation. The effect is very strong, however, for electoral activity, where it becomes the dominant predictor in the model. This is in line with the general pattern in the Nie-Powell-Prewitt data, but it is a finding which I personally find difficult to interpret. Political interest and the reading of political news can just as well reflect participation as be a cause of it. Either way, how ever, the effect is shown to be three times as strong for electoral activity as for interest-group activity, and this indirectly confirms Nie, Powell, and Prewitt on the point that organizational involvement is less related to civic attitudes than electoral-sector involvement. What we gain on their perspective, however, is the understanding that: (a) the involvement in (question is interest-group involvement; (b) that such involvement can be considered both politically effective and politically sufficient for those involved; and (c) that it is an involvement dominated by men. To reformulate the conclusion: Interest-group involvement is a male-dominated mode of realizing personal interests in relation to the polity which is not dependent on civic attitudes. (2) In line with Nie, Powell, and Prewitt's finding, social class has no direct effect on electoral activity, but does have an indirect effect which goes from education through the other attitude measures to political attention. What Table 5 also shows, however, is that this indirect effect is both relatively weak in absolute terms, and far from decisive in relation to political attention, which is also affected by sex and organizational involvement. Furthermore, income exerts no effect at all on attitudes, other than a very slight indirect effect on political attention through organizational membership. In short, with sex in the model, the weak zero-order relationship between class and electoral activity is shown to be composed of several equally weak indirect effects. (3) It is, therefore, sex which retains the dominant structural position in the electoral-activity model, in terms of both direct and indirect effects. Directly, the effect is weakened from what it was in the four-variable model, due mainly to the strong indirect effect through political attention. Side 248
A mathematical summary of the direct and indirect effects for electoral activity, shows that the total effects for all factors are nearly identical with what they are for the four-variable model.11 In other words, the introduction of the attitude dimension increases our understanding of how structural factors and attitudes are interrelated, but it does not alter the relcitive importance of the structural factors in explaining the activity in question. Nie, Powell, and Prewitt's conclusion on the mediating role of civic attitudes is thus of little importance for Norway. If the mediation hypothesis is at all relevant, it is more a question of sex than class. 6. ConclusionsIn trying to determine the relevance of Norway for Nie, Powell, and Prewitt's developmentalist perspective, it has been necessary to alter the conceptual-operational basis of the perspective itself. Following Martinussen's practice, which reflects the 'two-tier' nature of Norway's decision-making system, I have distinguished between two participatory sectors: the electoral-representational and interest-group channels of influence. This change of perspective would seem to be mandatory for any system which has developed beyond a certain (as yet unspecified) threshold of corporate-group integration in the official governing process. Norway is clearly above this threshold, and, presumably, so were the United States, England, and Germany (at least) at the time of the Nie study. Be that as it may, Nie, Powell, and Prewitt have not conceptualized the 'organizational infrastructure' in these terms, so we are left with a basic incompatibility of views from the outset. Even where the distinction is made, however - as with Martinussen - we find that conceptual clarity is not the same as operational preciseness. By failing to demarcate adequately the dependent and independent aspects of organizational involvement in his empirical indicators, Martinussen has produced results which are strongly affected by either direct or indirect autocorrelation. The resulting impression is one of strong independent organizational effects; a conclusion which seems to coincide with that of Nie, Powell, and Prewitt, but which a more rigid operationalization shows to be a misleading coincidence. When the party-political, interest-group, and avowedly-nonpolitical elements of organizational involvement are clearly separated, we find that the latter - which is logically the only element qualified for independent-variable status - is but marginally related to either of the former. As for the second
major factor in the Nie-Powell-Prewitt perspective -
Side 249
social class -1 have argued that, here too, Martinus sen's view of Norway must be critically assessed. The keystone of Martinussen's distant-democracy' image is that social stratification is a major cause of participatory inequality. I have disputed this image on the basis of both Martinussen's data and my own, and feel justified in concluding that class is not an important determinant of participation in Norway. In the process of arriving at this conclusion, however, we 'stumbled upon' a basic social characteristic which does mean something for participation: sex. Starting with the firm conviction that housewives can no longer be considered as status appendages of their husbands, we quickly discovered that, in terms of explanatory potential, sex is equal to occupational status - and then some. Rather than either class position or organizational membership, sex is that characteristic which is most likely to increase the probability of political involvement, particularly in interest groups. Having established this, it was then necessary to include sex in the analysis of the final aspect of the developmentalist perspective: civic attitudes. This diminished the comparability of the findings even more, but such is the price of progress. What we found was that the so-called civic attitudes are not strongly related to interest-group activity at all, but that at least one of them - political attention - is very strongly related to electoral activity. In an indirect way, these findings seem to correspond with those of Nie, Powell, and Prewitt, but apparently for the wrong reasons. For Nie, Powell, and Prewitt, nearly the entire effect of attitudes on participation was traceable to social status. They were somewhat confused (and a mite worried) over the fact that organizational involvement seemed to affect participation without positive attitudes, but they could only speculate on why this was so. The important thing was that the mediation hypothesis explained why status had no effect at all when measured in relation to attitudes and organizational involvement. My own data show, however, that the direct effect of attitudes on electoral activity is nearly exactly what it was for Nie, Powell, and Prewitt, without a strong zero-order relationship for status, and with clear evidence that status is but a partial and weak source of positive attitudes. Both sex and organizational membership are, in fact, more important direct determinants of political attention, which alone accounts for nearly all of the attitude effect. We have, in other words, a situation where attitudes are strongly related to the same type of activity measured by Nie, Powell, and Prewitt, but where attitudes themselves are a reflection of several different factors, only one of which is social position. Side 250
My interpretation of this for Norway is that an attitude like political attention is just as much a result of political engagement as it is a cause of it. That the same result does not accrue to interest-group activity can be understood from the simple fact that this type of engagement is instrumentally effective (and civically legitimate) in its own right, and does not require an interest in, or monitoring of, the alternative channel of partydominated politics. We are dealing, in other words, with a vital distinction between expressive and instrumental politics; a distinction which becomes of major importance in the corporatized welfare state.12 It is here, I believe, that the revision of the developmentalist position must begin, and Norway - as the proclaimed prototype of the 'new polity' (Heisler and Kvavik, 1974) - is the proper point of departure. But what, in the end, does all this mean for Nie, Powell, and Prewitt? Do the findings on class and attitudes seriously weaken their position? Or does the analysis of interest-group activity perhaps strengthen it, though on somewhat different grounds? Unfortunately, it's impossible to say in either case. Not because the present analysis has been indecisive, but because it has given rise to problems and perspectives which reflect back on the five-nation study and make its results indecisive - whether similar or different from those presented here. Two problems are of particular importance. First, there is a problem connected with the operationalization of political participation which is similar to that discussed for Martinussen. Nie, Powell, and Prewitt include an item in the participation index which is derived from, and thus dependent on, organizational membership. Since such membership is also tapped by the index of organizational involvement, there is bound to be a certain amount of autocorrelation between the two measures. Second, Nie, Powell, and Prewitt follow the standard practice of assigning housewives to their husbands' occupational status, at the same time that housewives are systematically scored lower on the index of organizational involvement, because they do not partake in the 'economically active' work force (?). I have myself constructed a similar index of organizational involvement and find that it correlates with sex at r = .45, while a composite status index (education and family income) correlates at r = .01! Considering the fact that Norwegian women are obviously more 'organizationally involved' today than the women of the five-nation study were in 1962, it is clear that the effect in the Nie, Powell, Prewitt data must be considerably greater. And what might the
effect of these two procedures be? It is pure Side 251
speculation but, based on the relationships in the Norwegian data, the following possibilities seem reasonable: The zero-order correlation between organizational involve ment and political participation is inflated due to autocorrelation, at the same time that the zero-order correlation between status and participation is deflated due to nonparticipating highhusband-status housewives. When attitudes are introduced into the model, the status correlation vanishes entirely because attitudes separate high-participating males and low-participating females better than maledominated status does. Attitudes thus take over from status as an invisible mediator of sex rather than as a direct mediator of status. At the same time, when the final step of the recursive model is reached and participation is entered as the dependent variable, the effect of organizational involvement jumps, partly due to the lack of attitude mediation, but well assisted by the autocorrelation attaching to common scores for organizational membership. In short, the Nie-Powell-Prewitt conclusions may reflect the effects of sex and autocorrelation as much as they reflect the effects actually described. All this is only educated guesswork but until a retest is carried out (or published) there is no way of determining the exact meaning of the Norwegian findings for the Nie-Powell-Prewitt perspective. Meanwhile, the present study has pointed up three major problem areas for future research: (1) the relationship between attitudes and participation: the problem of symmetric effects; (2) the contrast between electoral and organizational involvement: the expressive-instrumental dimension; and (3) the vital role of sex: the nature of its resilient residual effect. The latter point has also provided us with an important object lesson: the social indicator of 'sex' is neglected only at the cost of descriptive-theoretical veracity. NOTES 1 The only study I know of which lias taken up the developmentalist aspect of the Nie problem directly is Langton and Karns (1974). Presumably, many of the issues involved will be analysed in Verba, Nie, and Kim's forthcoming Participation and Political Equality: A Seven Nation Comparison (reference in Verba and Shabad, 1978). 2 Martinussen does not agree here. He feels that the best impression of his data is provided by the gamma coefficients (1977:61-65, especially footnote 33). He does not seem to be aware, however, of the tendency for gamma to be seriously inflated when indicator categories are few and skewed, as they are with several of his measures. Furthermore, he is apparently misinformed on the elacoefficient which does not, as he maintains, measure only linear correlation. To the contrary, eta measures all types of association on the basis of analysis of variance for each category of the independent variable. Rather than presenting a narrow indication of correlation, the eta picks up interactions which - in Martinussen's case - may actually go against his hypotheses on hierarchical dependence. The only way to determine the exact nature of the effect is to inspect the partial coefficients for each indicator category - something which Martinussen has either failed to do, or at least failed to report. Hence we are not really sure what the eta's mean for Martinussen's clearly directional hypotheses. The characteristics of gamma are discussed in Mueller, Schuessler, and Costner (1970:279-290), and of eta (as used in MCA analysis) in Andrews, Morgan, and Sonquist (1969). Side 252
2 Martinussen does not agree here. He feels that the best impression of his data is provided by the gamma coefficients (1977:61-65, especially footnote 33). He does not seem to be aware, however, of the tendency for gamma to be seriously inflated when indicator categories are few and skewed, as they are with several of his measures. Furthermore, he is apparently misinformed on the elacoefficient which does not, as he maintains, measure only linear correlation. To the contrary, eta measures all types of association on the basis of analysis of variance for each category of the independent variable. Rather than presenting a narrow indication of correlation, the eta picks up interactions which - in Martinussen's case - may actually go against his hypotheses on hierarchical dependence. The only way to determine the exact nature of the effect is to inspect the partial coefficients for each indicator category - something which Martinussen has either failed to do, or at least failed to report. Hence we are not really sure what the eta's mean for Martinussen's clearly directional hypotheses. The characteristics of gamma are discussed in Mueller, Schuessler, and Costner (1970:279-290), and of eta (as used in MCA analysis) in Andrews, Morgan, and Sonquist (1969). 3 The communities are Tinn, Odda, and Jevnaker municipalities. The project has been financed by the Norwegian Research Council for Science and the Humanities (NAVF) and carried out at the Institute for Social Research in Oslo. Previous reports on the survey data are available in Lafferty (1976) and Halsaa Albrektsen (1977). A comprehensive report on Participation and Democracy in Norway is forthcoming. 4 This has been tested by running an MCA analysis on the relationship between organizational membership and organizational activity, using my data and - as nearly as possible - Martinussen's index criteria. Despite several differences in the items making up the indexes, the beta's vary no more thano/7^ one-hundredth from those shown in Martinussen's Table 3.4. Obviously, both samples are tapping common types of variable dependency. 5 The index of electoral activity is scored as follows: 4 = either (a) currently holds office in a political party or is an active party member, or (b) campaigns 'often, or (c) attends meetings of the local Municipal Council or other public bodies 'often' (10.1%); 3 = either (d) donates money to a political party 'often', or (e) agitates for a party 'often', or (f) does either (b) or (c) above 'occasionally', or (g) is a nonactive member of a political party (16.6%); 2 = either does (b) or (c) above 'seldom', or does (d) or (e) above either 'occasionally' or 'seldom' (26.4%); / = does none of the above, but has at least voted in each of the past three elections (33.9%); 0 = does none of the above (13.0%). 6 The index of interest-group activity is scored as follows: 3 = 'very active' in at least one interest group (10.6%); 2= 'quite active'in at least one interest group (16.4%); /= 'not very active' in at least one interest group (25.6%); 0 = not a member of any interest group (46.7%). 7 The index of organizational membership is scored as follows: 3 = three or more memberships in nonpolitical voluntary organizations (5.3%); 2— two memberships (12.0%); / = one membership (31.9%); 0 = no memberships (50.8%). 8 The three class indicators are scored as follows: Education: 4 = completed high school (gymnasium) or higher professional school (10.7%); 3 =- completed middle school (realskole) or middle-level technical or health school (18.9%); 2 = has not completed middle school, but has some schooling after primary school (7 years), including lower technical, commercial, or administrative schools (26.9%); / = primary school plus diverse courses or special 'schools for housewives' (22.0%); 0 = seven years of primary school (folkeskole) only (21.5%). Income: ranked in five categories of family income, from iess than 40,000 kroner" (11.6%) to 'over 70,000 kroner' (15.7%). Occupational status: 6 = higher-level functionaries and higher-level independents. both business and professional (7.7%); 5= middle-level functionaries (7.8%);4 = technicians and lower-level self-employed (5.9%); i = foremen and lower business, administrative, and service personnel (9.8%); 2 = nonskilled and skilied workers (28.8%); / = housewives with part-time employment (20.5%); 0 = housewives with no nondomestic employment (20.5%). In addition there were 5.4% who proved difficult to categorize: students, pensioners, and invalids. 9 The pursuit of the 'something' in question is already under way in Norway. Using the same data employed here, Beatrice Halsaa Albrektsen (1977) has presented a comprehensive analysis of the zero-order relationships for a large number of variables. Her conclusion is that none of the variables tested satisfactorily explain the differences in participation between women and men, and she suggests the importance of a more basic 'women's culture' which structu es fernale experience away from political involvement. Ottar Hellevik( 1978) has given HalsaaAlbrektsen's workconsiderable attention, and has taken the first steps toward developing and testing a more logically rigid causal model. The present analysis goes beyond Halsaa Albrektsen's work by distinguishing more closely between the different types of participation, and by applying multivariate controls and path analysis. Side 253
9 The pursuit of the 'something' in question is already under way in Norway. Using the same data employed here, Beatrice Halsaa Albrektsen (1977) has presented a comprehensive analysis of the zero-order relationships for a large number of variables. Her conclusion is that none of the variables tested satisfactorily explain the differences in participation between women and men, and she suggests the importance of a more basic 'women's culture' which structu es fernale experience away from political involvement. Ottar Hellevik( 1978) has given HalsaaAlbrektsen's workconsiderable attention, and has taken the first steps toward developing and testing a more logically rigid causal model. The present analysis goes beyond Halsaa Albrektsen's work by distinguishing more closely between the different types of participation, and by applying multivariate controls and path analysis. 10 The questions used to construct tlie attitude measures are completely different from those used by Nie, Powell, and Prewiit, but the dimensions tapped are, I believe, the same. The questions and indicators art:: Perceived government impacf. 'What happens in politics has little meaning for my day-to-day life.' Score: 3 = disa,;;ree completely (32.9%); 2 = disagree partly (25.3%); / = agree partly (25.1%); 0 ~ aigree completely (16.7%). Political efficiacy: Question one: 'Choose one of the following two: (A) As far as world affairs are concerned, most of us are victims of forces we can neither understand nor control. (B) By taking an active part in political and social affairs, the common man can participate in controlling world development.' Question two: 'People like me can vote, but there's nothing else we can do to influence politics.' Score: 2 = choice of (B) on Ql and disagrees completely or partly on Q2 (22.2%); / = mixed choice on both questions (48.5%); 0 = choice of(A) onQl and eitheragrees completely or partly onQ2 (29.3%). Political attention: Question one: 'When it comes to politics, would you say that you are generally very interested, somewhat interested, or not very interestedT Question two: 'How thoroughly do you read political news in the papers? Would you say quite thoroughly, rather superficially, or not at allT Score: 2 — either very interested or somewhat interested in politics and ireads political news thoroughly (25.0%); / = mixed responses on both questions (40.1%); 0 = not very interested in politics and reads political news only superficially or not at all (35.0%). 11 The summary is basedon an algorithm developed by Vernon L. Greene (1977) which gives an exact estimate of the direct, indirect, and total effects in hierarchical causal models. The total effects (in terms of path coefficients) of each predictor in the model shown in Table 5 are: Sex (.245), education (.063), income (.029), organizational membership (.159), perceived government impact (. 100), political efficacy (.135), and political attention (.440). The technique has been made available for computer use in Oslo by Nils-Eyvind Naas and Hans J. Bakke. 12 The most relevant discussions of the instrumentalist-expressive dichotomy in this context are Himmelstrand (1969), MiJbrath (1972:12-13), and Di Palma (1970:199-215). REFERENCESAndrews, F.,
Morgan J. and Songquist, T. 1969. Multiple
Classification Analysis: A Report Di Palma, G.
1970. Apathy and Participation: Mass Politics in Western
Societies. New York Greene, V. L.
1977. 'An algorithm far total and indirect causal
effects.' Political Methology Hallenstvedt, A.,
Moren, T. 1975. 'Det organiserte samfunn'. In Natalie
Rogoff Ramsøy and Halsaa
Albrektsen, B. 1977. Kvinnei og Politisk Deltakelse.
Oslo: Pax Forlag. Hellevik, O.
1978. Om Beatrice Halsaa Albrektsens: 'Kvinner og
politisk deltakelse'. Tidsskrift Heisler, M. O. and
Kvavik, R. 1974. Politics in Europe. New York: David
McKay Co., Inc. Himmelstrand, U.
1969. 'A theoretical and empirical approach to
depolitization and political ILO, 1977. Year
Book of Labor Statistics. Geneva: ILO. Lafferty, W. M.
1976. Basic needs and political values: Some
perspectives from Norway on Langton, K. P.
and Karns D. A. 1974. 'Political socialization and
national development: Martinussen, W.
1977. The Distant Democracy: Social Inequality,
Political Resources, and Milbrath, L. W.
1972. Political Participation: How and Why Do People Get
Involved in Mueller, J. H.,
Schuessler, K. F. and Costner, H. L. 1970. Statistical
Reasoning in Sociology. Nie, N. H.,
Bingham Powell, G. Jr. and Prewitt, K. 1969a. 'Social
structure and political Nie, N. H.,
Bingham Powell, G. Jr. and Prewitt, K. 1969b 'Social
Structure and political Pestoff, V. A.
1977. Voluntary Associations and Nordic Party Systems.
Stockholm: University Rokkan, S., and Valen, H. 1962. 'The mobilization of the periphery'. In Stein Rokkan (ed.) Approaches to the Study of Political Participation. Bergen: Christian Michelsen Institute. Reprinted in Stein Rokkan. Citizens, Elections Parties. Oslo: Universitetsforlaget, pp. 181-225. Rose, L. E. 1976.
Political Participation in Norway: Patterns of Citizen
Behavior in Two UNESCO, 1977.
Statistical Yearbook. Paris: UNESCO. United Nations 1977,
Yearbook of National Accounts Statistics. New York:
United Nations. Verba, S. and
Nie, N. H. 1972: Participation in America: Political
Democracy and Social Verba, S. and
Shabad G. 1978. 'Workers' councils and political
stratification: The Yugoslav |