Resumé

This article discusses the impact of sex and resources on political participation. The independent variables in the causal model are sex, education, organizational membership, and political involvement. Two measures of participation are used: frequency of voting and campaign participation. The results show that organization membership is the most important resource of participation. Organization is the only resource to have both indirect and direct effects. For sex and education, the effects are mediated through political involvement. The model is very weak in explaining variation in voting, but is clearly stronger in explaining variation in campaign participation.

The purpose of this paper is a rather modest one. We want to analyze some limited aspects of an established model of political participation. The starting point of our piece of work is the model proposed in the now classic article by Nie et al. (1969) and the later work of the Verba group (Verba et al. 1971; Verba and Nie 1972; Nie and Verba 1975; Verba et al. 1978). We take up only a few of the possible issues of research relating to this model, and we do not present new data. We rely on a reanalysis of data from the Norwegian Election Research Programme, especially the 1969" nationwide representative survey. Our research is hence restricted to the Norwegian case only.

There are two starting points for the reanalysis. The first is given by
Martinussen, who discusses the role of political orientations as mediating
mechanisms between resources and political participation. In the concluding

---

* The research reported in this article was initiated at the NSD/ISSC training course in political participation, Bergen, June, 1978. The article is a slightly revised version of a paper presented at the 10th Nordic Congress of Sociology, Åbo, August 21-23, 1979. The authors would like to thank Willy Martinussen who gave us access to the data. Martinussen, Ingrid Bogen, Kristen Ringdal, William Pang-yu Ting, and the participants in the group on political sociology in Åbo gave valuable comments on the original paper.

pages of Part I of Distant Democracy Martinussen (1977, 105) states that: 'The line may be taken that it is important to proceed from this point to a search for the roads leading to high political participation. The results of such an analysis would probably be roughly the same as those arrived at by Nie, Powell and Prewitt'. Our intention in this work is to follow this proposal, and it is our hope that it is possible to give a more precise answer as to the role of the direct and indirect effects of resources on political participation than is given by Martinussen (1977, 96-106). The second starting point is given in a provocative article by Lafferty (1978), who introduces a frontal attack on both the work of Nie, Powell and Prewitt and Martinussen. Lafferty's thesis is well summed up in the following formulation: 'Nie, Powell and Prewitt's conclusion on the mediating role of civic attitudes is thus of little relevance for Norway. If the mediation hypothesis is at all relevant, it is more a question of sex than class' (1978, 248). There are also other points in Lafferty's article which we do not discuss here. For an answer to a number of these, see the reply by Martinussen (1979).

On the agenda is thus the question of the importance of sex for political participation, compared with the traditional resource variables, as well as the question of the mediation of these variables through attitude variables. Is sex more important than resources, and is the effect of sex and resources direct or indirect?

1. The Model

The simplest formulation of the model of political participation in the research following the tradition from Nie et al. (1969) is given in Figure 1. The logic of this simple model is quite easy to formulate. To participate in politics the citizen needs resources and motivation. Using a wide definition,

resources are those personal and social characteristics which enable a citizen to use his political rights (Martinussen 1977, 7). This definition comprises both resources and motivational attitudes. Within our model we explicitly exclude attitudes from the concept of resources. The most used resource variables have been stratification variables and organizational membership. The idea is that political participation is easier for people in higher social strata.

Stratification is measured by commonly used indicators such as education, income, and occupational status. We shall not proceed further into the analysis of the more specific meaning of these variables in the participation context. At this point we accept the arguments for using stratification variables and organizational membership as resource indicators. Motivational attitudes are those attitudes which express the individual's psychological involvement in politics. Common variables used to measure this dimension are political interests, political information, and political efficacy. Attitudes stimulate political participation directly, but they also work to convert resources into participation. The arrows in Figure 1 can then be understood as causal relationships indicating the paths leading to high political participation.

There is, however, one problem which has been given very little attention, i.e. the direct path between resources and participation. A striking finding of Nie et al. (1969, Table 1) is that the direct effect of status on participation is not positive for four out of the five nations in their dataset. There seems, however, to be no good explanation of this finding. We shall return to this problem in the discussion of the results of the data analysis.

The concept of political participation has been much discussed. The prevailing view which has been firmly established by the Verba group is that political participation is not a cumulative act. To replace the notion of cumulativity, they have set the focus on the theory of the modes of political participation. In particular they distinguish between voting, campaign activity, citizen-initiated contacts, and cooperative activity (Nie and Verba 1975, 9-12). We shall not explore the possibilities of using this fourfold conceptualization of participation in our analysis. The reason for this is that the data at hand are primarily confined to the electoral process. Political participation will therefore be represented only by voting and campaign activities.

In order to answer the questions stated earlier, we also need to handle the problem of sex. Lafferty's (1978) attack on much of the established research on political participation is based on the view that sex is an important but neglected variable in most existing models. Sex is relevant

in two respects. First, there appears to be a relationship between sex and resources which seems to favour men. This leaves women with fewer resources to convert into political participation. Second, the sex roles of women can be seen as inhibiting political participation. The introduction of sex in the model creates some difficulties regarding what resource-variables one should use in the causal model. Family income as an indicator of socio-economic states (SES) cannot be used since there is reason to believe this resource to be under male domination. Occupational status also creates problems. The problem is due to the difficulty of assigning occupational status to housewives. We can either give this group the status of their husbands or a separate ranking. It is obvious that both solutions have their weaknesses. The easiest thing to do is to drop the variable from the SES measure. V/e are then left with education as the only variable of status. The procedure is consonant with the choice of Verba et al. (1978, 237), but not with Lafferty (1978), who uses family income in his model. Our full model is shown in Figure 2.

The model is an extension of the simple model in Figure 1. The most important difference is that sex is introduced into the model. There are two resource indicators, education and organizational membership,. The motivation attitudes have been renamed political involvement, because this seems to be more in accordance with the data at hand. The theory clearly expects all relationships in the model to be positive. As data have already shown, in most cases the direct effect of status on participation is not positive. This is not reflected in existing theory. Thus the arrows in the model can be seen as representing the theory, not necessarily the existing

data. The applicability of the model to predict the causal structure of
political participation in Norway remains to be seen.

2. Data and the Creation of Variables

Our data are from the 1969 Norwegian election study. This study is a national representative survey of the population between the ages of 21 and 79. Details about design, sampling, and interviewing are given in Martinussen (1977, appendix II). Our work is a reanalysis of existing data, but we have not used all of the original variables in the datafile. Most of the variables presented here are created from variables in the original datafile. We shall therefore describe the procedure of creating new variables.

The variables which we could use as indicators of political participation are listed in Table 1. All the variables are assumed to be at the ordinal level. There is only one variable measuring the voting dimension, a variable with 5 values measuring frequency of vote. The rest are taken to measure campaign participation. The datafile also included variables measuring vote in the national election of 1965 and the local election of 1967. These variables could have been used to form a composite measure of voting, and could also have made a dimensional analysis meaningful. But due to problems of assigning proper values to voters not entitled to vote in the 1965 and 1967 elections, we have chosen to exclude this information from the analysis.

A dimensional analysis employing Norwegian data hjas shown that variables similar to our campaign variables belong to the same factor (Nie et al. 1978, table A-13). We especially note that the discussion variable was not included in this analysis. The variables included in the campaign factor were factorized by the use of principal component analysis. The standardized variables weighted with the factor loadings provided by the principal

component analysis, were then added in order to construct an index of
campaign activity.

The next step was to construct a composite measure for the attitudinal variables. We first created three indexes of political interest, political information, and political efficacy. The index of interest was created in the same way as the index of campaign participation - by adding weighted standardized variables. The weights were provided by principal component analysis of the variables included in the index. The indexes of political information and efficacy were created by adding the standardized items.

The index of political interest was constructed from the following items: interest in politics in general, interest in local politics, interest in national politics, interest in foreign politics, and interest in public affairs. The index of information was created from two variables: the number of correctly identified party leaders, and the knowledge of candidates on own party list. The index of efficacy was also created from two variables: the feeling of possibility to influence parliamentary decisions, and the feeling of the possibility to influence local councils. These indexes of interest, information, and efficacy were combined into a single measure of political involvement by adding the standardized items weighted with the loadings on the first principal component.

The measures of resources are simpler. Education is a variable measuring number of years of completed formal education after primary school. Organizational membership is the number of memberships in voluntary organizations (excluding political parties). Sex is a dichotomous variable, women scoring 0 and men 1.

3. Findings

An appropriate method of testing a recursive causal model is to use path analysis (Asher 1976; Duncan 1975, chap. 3). In Figure 3 the results are presented for the voting dimension of participation. We first notice the rather weak paths from sex to education (.08) and from sex to organizational membership (.07) (all coefficients reported in the text are standardized regression coefficients). The effect of education on organization Is also moderate (.18). The direct paths from sex and the resource variables to political involvement are stronger (.23 for sex, .26 for education, and .25 for organization). The indirect paths from sex to political involvement are rather weak due to the weakness of the paths between sex and the resource variables.

The latter are in turn partly a result of our operationalization of these variables. Obviously, the length of formal education does not fully reflect the level of education. There may be a tendency that women have many years of education at a low level, while men have a greater part of their educational years on a higher level. It is also likely that the operationalization of organizational membership suffers from weaknesses of the same kind. Women's membership may be in organizations at a greater distance from the political sphere than men's organizations.

The most interesting relationships in Figure 3 are those which lead to the ultimate dependent variable, voting. According to the model set out in Figure 2, we would expect the direct paths from sex, education, and organization to be positive, but there are no direct paths which are statistically significant. What remains for sex, education, and organization are the indirect effects of these variables on voting. Table 2 gives a summary of the decomposition of the effects.

In short, indirect effects of the variables are very close to one another. The only single indirect effect which is of some importance is the mediation through political involvement. The model explains an extremely low proportion of the variance in the ultimate dependent variable. This can indicate that sex and resources are rather irrelevant to the voting dimension of political participation, it is also a possibility that the small beta's may be due to weak epistemic correlations between the indicators and the

theoretical variables. But it is, however, not surprising that voting in
Norway is an act of participation which shows a rather egalitarian picture.

We shall now turn to the other dependent variable, participation in campaigns. The result of the test is presented in Figure 4. The relationships between the variables on the left side of the model are the same as those for the model in Figure 3, and will therefore not be commented upon. The main difference between the model of voting and the model of campaig'ning is that the path between political involvement and the dependent variable is stronger in the latter (.43). There is only one statistically significant direct path from the resource variables and sex to participation, that from organization. As a consequence of the increasing strength of the path from political involvement to campaign, the indirect effects of the

variables are also stronger in this model. The decomposition of effects is
shown in Table 3.

The indirect effect from sex is .12, from organization .10, and from education .15. The single most important indirect effect is again the mediation through political involvement. The model is still weak (R² = .22), but it is evident that sex and resources are more relevant to campaigning than to voting.

4. Conclusion

Our purpose was to analyze the relative importance of sex as compared to resources in a causal model of political participation. More specifically, we also wanted to see if the effects of these variables on participation were direct or indirect. The answers to our initial questions seem to be the following. Both sex and resources are quite irrelevant to voting, although there is a tendency to weak indirect effects. For campaign participation the model must still be considered weak, but it is clearly relevant. Organizational membership is the most important resource for campaign participation. This variable is the only one which has both direct and indirect effects. Sex is the number two predictor, but the effect is mediated through the other variables in the model, especially through political involvement. Education has a positive total impact on campaign participation. The direct effect of education is, however, negative; the indirect effect clearly positive. Education is thus a two-faced resource for political participation.

Comparison with the findings from the original Nie et al. (1969) study is
difficult due to different operationalizations of the variables. Still, it seems
that our findings confirm two of the main points. The first is that the effect

of social status (of which education is an item) is mediated through political
involvement. The second is the importance of organization, both as a
direct and as an indirect source of participation.

The negative direct effect of education on campaign participation is also partly consonant with the work of Nie et al. (1969, 812), where the USA is the only country that shows a positive direct link between status and participation. For the other countries (UK, Germany, Italy, and Mexico) the direct path from status leans to the negative side, with Germany as the strongest case. In the later work of Verba et al. (1971,75-79), the negative direct impact of education is presented in a way more comparable to our research. In this case the direct path from education to voting was shown to be negative for Austria, India, Japan, and Nigeria. The only nation coming out on the positive side was again the USA. The pattern for campaign activities is more unclear with Austria and Japan showing direct negative paths and India and the USA positive paths. For Nigeria, no measure of the campaign factor was ascertained. Our results on this point are somewhat unclear, but it should be about time to consider education as. a resource which does not necessarily lead to participation. For the non-involved the effect of education would seem to be the opposite, away from participation. There may be different answers to this problem. One question which could be asked is how rational it is for a person to participate in politics when he lacks interest, information, and efficacy? The answer to this could be that then it is not rational to participate. And hence education leads to rationality, which is in accordance with general theory. Another answer is to give the institutional forces a more central position in the model and analyze whether the constraints of institutions affect the high- and low-educated groups in different ways. The strengthening of the institutional analysis is perhaps the most important feature of the comparative analysis of Verba et al. (1978).

Our analysis of sex has shown that the effect of this variable on participation is only indirect. It seems theoretically very difficult to accept direct effects from sex on political participation. We can assume women to be less active than men in politics due to lack of resources, lack of political involvement (which is especially important in our model), or because women have their own culture (which Albrektsen (1977) argues, but does not empirically verify). A direct path between sex and participation must be interpreted as mediating variables not specified in the model, or as biology. The latter factor is very difficult to give a meaningful interpretation in a model of political participation. It is also obvious that sex does not have the unique importance argued by Lafferty (1978) in his analysis of

participation based on data from three industrialized communities in
Norway. But sex is nonelheless a variable which gives a better understanding
of participation.

Our analysis has been modest in scope. We willingly accept that we have only touched upon a few of the issues of the rich theoretical and empirical analysis of political participation provided by the earlier work on the Norwegian 1969 data (Martinussen 1977). In particular we will point to the fact that Martinussen (1977, 35-65) gives a much more elaborate picture of barriers and resources than we have been able to include in our causal model. Further reanalysis of participation data should also be extended to more variables and to a longer time period. This could, for example, be done in the Norwegian case by utilizing the valuable time series data from the Norwegian Election Research Programme. Analysis of trends in participation could then also be carried out (see Andersen 1975).

REFERENCES

Albrektsen, B.H. 1977: Kvinner og politisk deltakelse. Oslo: Pax.

Andersen, K. 1975: Working Women and Political Participation, 1952-1972, American
Journal of Political Science 19, 439-453.

Asher, H. B. 1976: Causal Modeling. Beverly Hills: Sage.

Duncan, O. D. 1975: Introduction to Structural Equation Models. New York: Academic
Press.

Lafferty, W. M. 1978: Social Development and Political Participation: Class, Organizations
and Sex, Scandinavian Political Studies, Vol. 1, new series, 233-254.

Martinussen, W. 1977: The Distant Democracy. London: Wiley.

Martinussen, W. 1979: On the Inability of Seeing the Forests for the Trees, Review Article,
Scandinavian Political Studies, Vol. 2, new series, 179-182.

Nie, N.H., Powell, G. B. Jr. and Prewitt, K. 1969: Social Structure and Political Participation:
Developmental Relationships. Parts 1 and 2, American Political Science Review 63,
361-378, 808-832.

Nie, N. H. and Verba, S. 1975: Political Participation, pp. 1-74in Greenstein, F. I. and Plsby,
N. W. (eds.): Handbook of Political Science, Vol. 4 Reading, Mass.: Addison-Wesley.

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

Verba, S., Nie, N.H. and Kim, J. 1971: The Modes of Participation: A Cross-National
Comparison. Beverly Hills: Sage.

Verba, S., Nie, N.H. and Kim, J. 1978: Participation and Political Equality. Cambridge:
Cambridge University Press.