Resumé

Latent Variable Partial Least Squares analysis has been used to model the impact of political party organization on stable and changing patterns of party vote in Norway from 1945 to 1977. For the Norwegian Labor Party, the elaboration of a strategy of organizational encapsulation is at least as important as the economic base for maintaining the stability of the party's electoral support; it also depresses support for the opposition parties. Short-run fluctuations around the long-run stability, however, are more influence by economic changes than party organizational strategies, with the exception of the divisive 1973 election, when party organization was important for maintaining the party vote. For the non-Labor parties, party organization is less important than economic variables.

The Problem and the Setting

In Western Europe the latter half of the nineteenth century and the first years of the twentieth witnessed the gradual expansion of the national political community as new groups were extended full rights of citizenship. Between 1860 and 1920 the progressive expansion of the suffrage followed an almost inexorable logic, so that by the beginning of the Second World War, with a few major exceptions, a franchise of 'one adult, one value, one vote' was common to Western European states (Rokkan 1962, 1970b). This enlargement of the national political community in turn precipitated attempts by contending political elites to mobilize the newly enfranchised for electoral purposes (Lipset & Rokkan 1967). Efforts at political mobilization generated a new organizational form — the mass political party — aimed at ensuring the electoral solidarity of newly enfranchised citizenry and hence the predictability of electoral outcomes. Commenting on the new organizations, Max Weber (1969,105) noted their form and purpose:

---

* An earlier version of this paper was presented at the European Consortium for Political Research meetings, March 20-26, 1983, Freiburg, Germany. This paper is developed from research sponsored by the National Science Foundation under Grant SES 79-25622. Any opinions, findings and conclusions expressed in this paper are those of the author and do not necessarily reflect the views of the National Science Foundation.

The occasion for this development (the caucus system) was the democratization of franchise. In order to win the masses it became necessary to call into being a tremendous apparatus of apparently democratic associations. An electoral association had to be formed in every city district to help keep the organization incessantly in motion and to bureaucratize everything rigidly.

Where parties could successfully organize, the resulting organizational proliferation functioned to absorb the political mobilization of the period before the first world war and to reduce the tendency toward polarization by creating a myriad of cross-pressures. The long-run consequences of this process emerged after the second world war: a reduction of the importance of electoral contests because of narrowing political alternatives and fragmenting interest organizations. As a result, political participation declined as major sectors of the electorate became alienated from politics, while policy became increasingly the purview of specialized interest organizations, party elites and the bureaucracy (Rokkan 1962, 89-90). Yet a modicum of party organization remained essential for parties to win electoral contests and to maintain the appearance of party democracy.

The lethargy of the electorates proved short-lived, however, and by the 1960's the alignments of the inter-war period displayed signs of weakening as new parties or variations on old ones challenged the dominant party alignments. Scholars increasingly turned their attention to the potential for realignments (Rose & Urwin 1970, Shamir 1984).

Party Organization and the Vote

The organizational tactics elaborated by parties varied according to timing, rate and manner of suffrage expansion as well as across the social strata. In their effort to mobilize the vote, parties faced numerous obstacles conditioning their organizational responses (Lipschits 1980, Svåsand 1980). For example, the emergence of class based parties varied with industrialization, the removal of class impediments to political participation, the openness of the dominant political strata and the ability of elites to manipulate cross-cutting cleavages (Lipset 1983, Claggett et al. 1982). Where the secret ballot accompanied franchise expansion, the newly enfranchised were cut off from their social peers as well as superiors, often requiring strenuous efforts by parties to overcome this new political isolation (Rokkan 1962, 79). For parties seeking to appeal to the lower socio-economic strata the problem was compounded by a generally lower political competence (Rokkan 1962, 79, Lipset 1963, 203-4, Bendix 1969, 120-22).

To overcome these obstacles, parties evolved organizational strategies to minimize cross-pressures and stimulate political participation. Only by reducing cross-pressures could parties hope to insure voter solidarity and undo the consequences of the secret ballot. The general strategy was one of organizational

encapsulation (Sartori 1968, 122), defined as an attempt to organize and incorporate within the political party as many of the everyday activities of the membership as possible. This infusion of party organization usually took the form of a proliferation of party subunits designed to meet the needs of specialized clientele groups. If the party organizational milieu 'could claim the allegiance of (the) voters in all social roles and therefore isolate them from disturbing cross-pressures' (Rokkan 1962, 79), the party's electoral support became more stable and predictable.

The political economy of organization, however, varied across the social strata. Both varying electoral administration and political competence demanded specialized organizational solutions. The secret ballot essentially appeals to the liberal urban mentality: it fits another element into the anonymous, privatized culture of the city..! (Bendix 1969,121) and thus militates against working class solidarity. Moreover, levels of political competence varied with social status, and party organizational strategies had to be adjusted accordingly. Therefore, parties seeking to mobilize the newly enfranchised lower strata confronted particularly acute obstacles to mobilization, but the creation of a homogeneous subculture could overcome the political isolation occasioned by the secret vote as well as lower political competence. In general the more individuals a party could 'encapsulate' the greater the potential for voter mobilization and electoral conformity.l By surrounding the worker with peer group influences, party organization could reduce competing loyalties and stimulate political activism. For parties appealing to the middle or upper strata, political participation was the outgrowh of pre-existing density of organizational life (Rokkan 1970a, 352-361, Barnes 1968, 111-112, Nie et al. 1971).

Organizational Encapsulation and Voter Mobilization in Norway

The following analysis models the relationship between party organizational activities and electoral success in Norway from 1945 to 1977. The development of the Norwegian party system closely approximates the preceding description. During the post-war period the system began to demonstrate signs of a thawing of the inter-war alignments with weakening of electoral support for the Labor party and the growth of the Conservative and Agrarian parties. From 1957 to 1977 vote for the Labor party declined from 48.3% to 42.3 % with a low of 35.4% in 1973 as a consequence of the EEC issue. On the other hand the Conservative party increased both its membership and vote during the period, enabling it to form a minority government in 1981 (Svåsand 1982, 7-9).

Concepts, Measures and Data

The salient concept for this analysis is organizational encapsulation defined as

the strategy of organizing as many of the daily activities of the electorate as possible or necessary to ensure electoral success. The Labor party most closely approximated the ideal of organizational encapsulation with the elaboration of constituency organizational units, trade union affiliates, consumer cooperatives and specialized units for youth, women, and social activities. The available data for this analysis consist of party organizational units, communal organizational units, party membership, trade union organizational units and membership, and cooperative society organizational membership. The data are relatively complete between 1945 and 1973; after 1973 only party membership data are available. For the Christian, Conservative, Center and Liberal parties, data consist of party membership and are relatively complete from 1965 to 1977.

Electoral data ara available for all parties during the period, while sectorial employment data are available from the censuses of 1930, 1945, 1950, 1960 and 1970. All data are at the county (fylke) level. Data at the fylke level, of course, mask a great deal of variation in both independent and dependent variables.

The variables are calculated as proportions of their relevant denominators: sectorial employment, consumer cooperative membership and trade union membership are proportions of the economically active population; party vote is proportion of votes cast; while party membership and nonvoters are proportions of the enfranchised electorate.

Model Conceptualization and Construction

This analysis seeks to model change and stability across time where change and stability are empirically defined and time is explicitly incorporated into the model. This aim is facilitated by the analytic technique available through Latent Variable Partial Least Squares (LVPLS). LVPLS speaks of latent and manifest variables. Latent variables are theoretical constructs which cannot be observed directly but which are measured by observed or manifest variables. In this sense the conceptualization is akin to factor analysis. Each manifest variable consists of a unique variance as well as common variance with other manifest variables thought to underline the theoretical or latent variable. For a fuller discussion of the analytic technique employed here, see Bentler 1980, McArdle 1982, McArdle & McDonald 1981, Lohmoller 1981, 1982a, 1982b, Reis 1982, Wold 1973, 1979, 1980.

Latent Variable Construction

Since the analysis focuses on change and variability across time, it is necessary to construct an empirical measure of change and stability. This is achieved in the following manner. First, a principal components factor analysis is done of the economic and party vote variables across time; high loaders across all time periods are considered stable, low loaders are considered unstable. This decision

rule is then used to select manifest variables as loaders for the stability and
instability of sectorial employment and vote latent variables (LVs).

Two sectorial employment LVs were constructed for each census year from the manifest variables: Stable Employment Sector LV and Variable Employment Sector LV. The following sectorial employment variables showed stability across time: Agriculture, Industry, Commerce, and Service. The following sectorial employment variables proved unstable across time: Sea Transport, Transport, Domestic Employment, Construction, Electrical, and Unknown.

Similar methods generated LVs for voting patterns: Stable Vote LVs consists of those high-loading manifest variables across time, while Unstable Vote LVs are the low loaders in each of nine electoral contests between 1945 and 1977. To incorporate across-time change the model seeks to predict the vote in one election from a previous election. If the vote is highly predicted by the preceding election, it shows stability across time. Among the party vote variables, four showed initial stability by loading highly in the principal component factor construction: Labor party, Liberal party, Center party, and the Communist party (for the years in which it participated). All other parties and non-voters show instability across time.

Party organizational variables were assigned to each time period according to appropriate time points between elections and latent variables again constructed from the manifest variables to produce a Labor and a non-Labor party organizational LVs. Models of the Labor party organizational LV indicated that some of the Labor party organizational manifest variables did not load on the LV: Labor party communal units, cooperative units and cooperative membership are not underpinnings of the theoretical concept of organizational encapsulation. The model again uses party organizational LVs at one time point to predict party organizational LVs at a later time; the strength of this coefficient similarly indicates stability. Because of missing data there are no non-Labor party organizational latent variables for the elections of 1945, 1949, 1953, 1957, and 1961.

Table 1 presents the abbreviations used in the figures; the final two digit number refers to the year of observation. The notational convention used in all figures is standard for Latent Variable Analysis; Manifest Variables are represented by squares, Latent Variables by circles, Directionality by straight arrows, Undirected Covariance by curved arrows, and Unexplained Variance by smaller circles on the manifest or latent variables.

As an estimation technique LVPLS holds advantages for the social sciences; it is statistically robust, yet less knowledge sensitive than some alternative methods (Lohmoller 1982b, 6-7). The technique allows for two methods of estimating relationships between the manifest variables and the latent variables: Inward and Outward Directed (Lohmoller 1981, P. 2.2). If the researcher believes the manifest variables are common indicators of an underlying theoretical construct, outward directedness is warranted and the coefficients are principal component factor loadings. In the following analysis, Stable and Variable Employment

Sector Latent Variables are outward directed. The assumption here was that these manifest variables share the dynamic of socio-economic change which we wish to incorporate into the analysis. Likewise, the latent variable of Labor Party Organizational Encapsulation is outward directed because again the manifest variables are assumed to represent a common dynamic of party organizational strategy.

On the other hand, if the researcher believes the manifest variables serve to measure or categorize a dimension whose relative components are unknown, inward directedness is appropriate and the coefficients are canonical correlation scores. Therefore, the Stable and Unstable Party Vote Latent Variables and the Non-Labor Party Organizational Latent Variables are inward directed.

Developing the Model

The final results presented in this research are the product of the following stages
of model development and testing which can be briefly summarized as follows:

1. Models were constructed for each election year using the closest census year to the election and the party organizational data available between the preceding election and the current election. LV construction was identical to the above described procedures.

2. Models were constructed for each successive pair of electoral contests, e.g. 1945 and 1949, 1949 and 1951, etc., following the same procedures as above. But in addition paths were directed from each LV at time 11, to its counterpart at time t 2.

3. Two models were constructed incorporating three electoral contests at years 1945, 1965, and 1973 in one model and the years 1949, 1969, and 1977 in the other. The results of the two models were compared. Again paths were directed from each LV to its counterpart in the succeding time observation.

4. Two models of four electoral contest were constructed: one model — basedon the 1945, 1957, 1965, and 1973 elections and incorporating the employment census data for 1946, 1960, and 1970; the second model basedon the 1945, 1961, 1969, and 1977 elections and incorporating the census data for 1946., 1960, and 1970. Again, the two models were tested and compared to each other. Measurement construction and path direction were identical to the above.

5. The models constructed in Step No. 4 above were introduced with two changes. First, the control variable of the total number of votes cast was dropped, and second, the LVs for Stable Employment Sector and Variable Employment Sector for each census period were combined to form only two LVs: one in which all stable employment MVs were the loadings and one in which all unstable employment MVs were the loadings. This model was then tested against its counterpart for the other four electoral contests. Figs. 2 to 5 present the results of this analysis which are consistent with the results of each previous step outlined above.

Models of Stability and Change

Fig. 1 presents the factor pattern for the LVs of Stable Employment Sector and Variable Employment Sector. The arrows are directed outward in the LVPLS algorithm and therefore the coefficients of the manifest variables are predicted from the LVs at the same time the LVs are fitted to the inner model.

The entire model is constructed in a piece-meal, but iterative, process and thus
should be conceptualized as a whole. However, for simplicity of representation
it is better to analyze the model's predictions for each election.

Fig. 2 represents the model's predictions for the 1945 election; only paths of an absolute value of .2 or greater are shown. The overall fit of the model is good, predicting about 70% of the variance in the vote LVs. The loadings of the manifest variables on the Labor Party Organization LV are high and approximately equal, generated by principal components analysis and constrained to fit the parameters specified in the inner model. The vote LVs are inward directed and the coefficients of the manifest variables should be interpreted as scores where the algorithm seeks the best standardized linear combination of the MVs to fit the relationships specified in the inner or path model.

The most interesting feature of Fig. 2 is that Labor Party Organization has its greatest impact through negative effects on Liberal party vote. Contrary to expectations, strong party organization depresses the opposition vote and participation rather than increasing its own party vote. The most likely explanation for the latter is that in strongly organized districts some opposition parties have little chance of success, hence their potential supporters abstain. This pattern also appeared in earlier models. Other paths are more expected; Stable Economic Sector LV has a positive effect and Variable Economic Sector LV has a negative effect on Labor Party Organization, suggesting that stable employment patterns facilitate party organization. The Stable Economic Sector LV also depresses participation directly, while the Variable Economic Sector LV shows a positive impact on Liberal vote.

Fig. 3 presents the second part of the model with the results for the 1957 election. The stability of the model is evident by the strong paths from the vote and party organization LVs in the 1945 election to their counterparts in the 1957 election and the resulting increase in the predictive success of the model; consequently the highest scores for the MVs are again achieved by the Liberal vote for the Stable Vote LV and Non-voters for the Unstable Vote LV. After removal of the time-dependent processes, only Labor Party Organization LV has a positive impact on Unstable Vote LV. These results are again consistent with those of earlier models.

Fig. 4 presents the third part of the model for the 1965 election and again generates results similar to those of the first two parts; paths indicating stability are high, while Labor Party Organization LV slightly depresses the Liberal vote and reduces participation. Changes in the Labor Party Organization LV also have a slight stimulative effect on the Liberal's party organization while

depressing that of the Center party, but this is less important than the effects of the sectorial employment changes on the Non-Labor Party Organizational LV. On the other hand, Non-Labor Party Organization LV shows no impact on party vote, a finding again consistent across all models.

The results for the 1973 election in Fig. 5 are among the most interesting as well as offering some insights into the setback for the Labor party in that election occasioned by the party's stance on the EEC question. The model shows evidence of instability in electoral patterns by the reduction of the path coefficients from the Stable and Unstable Vote LV's in 1965 to their counterparts in 1973. Labor Party Organization shows evidence of continued stability, but its impact on voting patterns is markedly different. Whereas in previous elections Labor Party Organization depressed opposition vote, in the 1973 election the presence of Labor Party Organization stimulated Labor party vote, but also shows an association with Liberal, Conservative, Center and 'other' party vote as well as non-voters. On the other hand, Labor Party Organization has a negative impact on Christian and New People's party vote. While these coefficients are not strong, they do receive support from previous analysis of the 1973 election as well as offer some additional insights into the impact of the EEC issue on that election. The stimulative effect of Labor Party Organization on Labor vote as well as that of other parties is best explained by the analysis of Labor losses between 1969 and 1973; Labor losses in 1973 occurred heavily in areas where voter-member ratio was above the party's national average and restrained in areas where its member-voter ratio was below the national average in 1969 (Svåsand 1982, 11). In other words, higher density of party organization stabilizes and moderately stimulates party vote in the context of a divisive election. The stimulative effect of Labor Party Organization on the Liberal vote and its depressive effect on Christian vote and paticipation has been noted by Rokkan & Valen (1974, 212-15). The authors state that some Labor supporters opposed to the party's position on the EEC chose the 'No' parties or abstained in the 1973 election. The model's results also conform to the more general interpretation that the 1973 vote held overtones of the center-periphery conflict long noted in Norwegian politics (Rokkan & Valen 1974, Hellevik et al. 1975). The Labor party's losses were greatest in peripheral areas which are also those where its voter-member ratio was above its national average. However, the model does not offer support for the hypothesis of Rokkan & Valen (1974, 213) that most of Labor's losses were to the left. The most predictable MVs in the LVPLS model from the party organization LV are the right parties, not the left parties.

Most of the explained variance in the 1973 election is accounted for by the vote LVs of the previous election and the sectorial employment LVs. The relative importance of Labor party organization is reduced from the previous elections. In this and all models constructed for this analysis non-Labor party organization is weakly, if at all, related to party vote, supporting the argument that party organizational activity is more important to parties appealing to the lower strata of society.

Conclusions

Conclusions drawn from this research must be tempered by the statistical problems remaining in the data as well as the limited number of units of observation. Within these limitations the findings provide an important role for party organization in the study of electoral behavior. The general findings of the models confirm the stability of Norwegian voting patterns paralleled by the stability of parties' organizational activities. The only exception to this is the 1973 election where a more complex relationship develops.

Models of change across time provide an unexpected function for the Norwegian Labor party organizational activity: the depressing of opposition party vote, with the 1973 election again providing the exception to the rule. The decay of party organization and its anticipated consequences on the stability of the inter-war electoral alliances does not appear in the models — rather stability of organization and vote is greater than the changes across time.

Non-Labor Party Organization is only weakly related to either stability or change in party vote, confirming the hypothesis that for middle and upper social strata political activity is a natural outgrowth of social and economic life needing little support or guidance from party organizational activities.

Economic variables are at least as important as party organizational activities for the Labor party, but organizational activities augment the effects of the economic variables. For non-Labor parties economic variables are more important than organizational efforts. Where economic variables impact on non party organizational activities, these impacts are not translated into party electoral support.

---

NOTE

1. The importance of the party activist as a mobilizer of electoral action follows from numerous studies: Katz & Lazarsfeld 1955, Katz 1957, and Lazarsfeld et al. 1944, who formulated the process as two-step flow of information in which the party activist through everyday activities plays the role of the opinion leader and information source. Thus, the more individuals a party could 'encapsulate" the greater the effect on its electoral base. A number of empirical studies argue in favor of this thesis: In the US. see Cutright & Rossi 1958a, 1958b, and 1961. In Britain, see Butler & Stokes 1974, 218ff, Janosik 1968, 62-84, Rose 1967, McKitterick 1960, Berry 1969, Potter 1951. In Norway, see Torgersen 1964, 1967, Valen & Katz 1961, 228ff, 1967, 129-39, Svåsand 1974, Rokkan 1970a, 1970b, 352-431, and Rokkan & Valen 1964.

REFERENCES

Barnes, S.H. 1968. 'Party Democracy and the Logic of Collective Action', pp. 105-38 in Crotty. W.J.
ed. Approaches to the Study of Party Organization. Boston: Allyn and Bacon.

Bendix, R. 1969. Nation-Building and Citizenship. New York: Doubleday, Anchor Books.

Bentler, P.M. 1980. 'Multivariate Analysis with Latent Variables: Causal Modeling', Annual Review
of Psychology 34, 419-56.

Berry, D. 1969. 'Party Membership and Social Participation', Political Studies 1, 196-207.
Butler, D. & Stokes, D. 1974. Political Change in Britain. New York: St. Martin's Press.

Claggett, W., Loesch, J., Shively, W.P. & Snell, R. 1982. 'Political Leadership and the Development of
Political Cleavages in Imperial Germany, 1871-1912', American Journal of Political Science 26,
643-63.

Cutright, P. & Rossi, R.H. 1961. 'The Impact of Party Organization in an Industrial Setting', pp. 81-116
in Janowitz, M., ed., Community Power Systems. Glencoe: Free Press.

Cutright, P. 1958a. 'Party Organization in Primary Elections.' American Journal of Sociology 64,
262-69.

Cutright, P. 1958b. 'Grass Roots Politicians and the Vote.' American Sociological Review 23,171-79.

Hellevik, 0., Gleditsch, N.P. & Ringdal, K. 1975. 'The Common Market Issue in Norway: A Conflict
Between Center and Periphery', Journal of Peace Research 12, 37-53.

Janosik, E.G. 1968. Constituency Labor Parties in Britain. New York: Praeger.

Katz, E. 1957. 'The Two-Step Flow of Communication: An Up-to-date Report on a Hypothesis.'
Public Opinion Quarterly 21, 67-68.

Katz, E. & Lazarsfeld, P.F. 1955. Personal Influence. Glencoe: Free Press.

Lazarsfeld, P.F., Berelson, P. & Gaudet, H. 1944. The Peoples' Choice. New York: Macmillan.

Lipschits, I. 1980. 'The Formation of Parties in National Political Systems', pp. 303-38 in Buesch, 0.,
ed., Waehlerbewegung in der Europaeischen Geschichte. Berlin: Colloquium Verlag.

Lipset, S.M. 1983. 'Radicalism or Reformism: The Sources of Working-class Polities', American
Political Science Review 11, 1-18.

Lipset, S.M. 1963. Political Man. New York: Doubleday Anchor.

Lipset, S.M. & Rokkan, S., eds. 1967. Party Systems and Voter Alignments: Cross-National Perspectives.
New York: Free Press.

LohmoUer, J. 1982a. 'Path Models for Latent Variables Indirectly Observed by Categorical and
Interval-Scaled Variables: An Extension of Partial Least Squares (PLS) Soft Modeling.' Paper,
Symposium on Cultural Indicators for Comparative Study of Culture, Vienna.

Lohmoller, J. 1982b. An Overview of Latent Variable Path Analysis'. Paper. American Educational
Research Association. New York.

Lohmoller, J. 1981. LVPLS 1.6 Program Manual: Latent Variable Path Analysis with Partial Least
Estimation. Neubiberg.

McArdle, J.J. 1982. 'Modeling Hierarchical Latent Variable Systems: Comparative Results from
LVPLS and LISREC Paper. American Educational Research Association. New York.

McArdle, J.J. & McDonald, R.P. 1981. 'A Simple Algebraic Representation of Structural Equation
Models'. Unpublished Paper. Department of Psychology, University of Denver.

McKitterick, T.E.M. 1960. 'The Membership and the Party', The Political Quarterly 31, 312-23.

Nie, N.H., Powell, G.8., & Prewett, K. 1971. 'Social Structure and Political Participation: Developmental
Relationships', pp. 406-430 in Finkle, J.L. & Gable, R.W., eds., Political Development and
Social Change. New York: Wiley.

Nonnan, R. & Wold, H. 1977. 'NIPALS Path Modeling with Latent Variables', Scandinavian Journal
of Educational Research 21, 33-61.

Potter, A.M. 1951. 'British Party Organization', Political Science Quarterly 66, 65-86.

Rokkan, S. 1970a. 'Electoral Activity, Party Membership and Organizational Influence1, pp. 352-61 in
Rokkan, S., Campbell, A., Torsvik, P. & Valen, H., eds., Citizens, Elections, Parties. New York:
David McKay.

Rokkan, S. 1970b. 'The Mobilization of the Periphery: Data on Turnout, Party Membership and
Candidate Recruitment in Norway', pp. 181-225 in Rokkan, S., Campbell, A., Torsvik, P. & Valen,
H., eds., Citizens, Elections, Parties. New York: David McKay.

Rokkan, S. 1962. 'The Comparative Study of Participation: Notes Toward a Perspective on Current:
Research', pp. 47-90 in Ranney, A., ed., Essays on the Behavioral Study of Politics. Urbana:
University of Illinois Press.

Rokkan, S. & Valen, H. 1974. 'Norway: The Election of the Storting in September, 1973'. Scandinavian
Political Studies 9, 205-17.

Rokkan, S. & Valen, H. 1964. 'Regional Contrasts in Norwegian Politics: A Review of Data from Official Statistics and from Sample Surveys', pp. 162-240 in AUardt, E. & Littunen, Y., eds., Cleavages, Ideologies and Party Systems: Contributions to Comparative Political Sociology. Helsinki: Transactions of the Westermarck Society, volume 10.

Reis, H. 1982. 'An Introduction to the Use of Structural Equations: Prospects and Problems', pp.
255-87 in Wheller, L., ed., Review of Personality and Social Psychology. Beverly Hills, CA: Sage.

Rose, R. 1967. Influencing Voters: A Study in Campaign Rationality. New York: St. Martin's.

Rose, K. & Urwin, D. 1970. 'Persistence and Change in Western Party Systems Since 1945', Political
Studies 18, 287-319.

Sartori, G. 1968. 'European Political Parties: The Case of Polarized Pluralism', pp. 115-49 in Dahl,
R.A. & Neubauer, D.E., eds., Readings in Modern Political Analysis. New York: Prentice-Hall.

Shamir, M. 1984. Are Western Party Systems "Frozen"? A Dynamic Analysis', Comparative Political
Studies 17, 35-80.

Svåsand, L. 1982. 'Party Organization and Partisan Mobilization in Norway'. Paper, American
Political Science Association meetings, Denver.

Svåsand, L. 1980. 'Democratization and Party Formation in Scandinavia', pp. 398-411 in Buesch, 0.,
ed., Waehlerbewegung in der Europaeischen Geschichte. Berlin: Colloquium Verlag.

Svåsand, L. 1974. 'Regional Variations in the Process of Party Formation in Norway'. Paper, ECPR
workshop on Regional Variation in Europe, Strassbourg.

Torgersen, U. 1967. 'The Formation of Parties in Norway; The Problem of Right-Left Differences',
Scandinavian Political Studies 2, 43-68.

Torgersen, U. 1964. 'The Structure of Urban Political Parties in Norway During the Period of Extended Suffrage. 1884-1898', pp. 377-99 in Allardt, E. & Littunen, Y., eds., Cleavages, Ideologies and Party Systems: Contributions to Political Sociology. Helsinki: Transactions of the Westermarck Society, volume 10.

Valen, H. & Katz, D. 1961. 'An Electoral Contest in a Norwegian Province1, pp. 207-36 in Janowit;:,
M., ed., Community Power Systems. Glencoe: Free Press.

Weber, M. 1969. 'Politics as a Vocation', pp. 77-128 in Gerth, H.H. & Mills, C.W., eds., From Max
Weber: Essays in Sociology. New York: Oxford University Press.

Wold, H. 1980. 'Model Construction and Evaluation When Theoretical Knowledge is Scarce', pp.
47-74 in Kamenta, J. & Ramsey, J., eds., Evaluation of Economic Models. New York: Academic
Press.

Wold, H. 1979. 'Soft Modeling: The Basic Design and Some Extensions', in Joereskog K.G. and Wold,
H., eds., Proceedings of the Conference 'Systems Under Indirect Observation. Causality,
Structure, Prediction.' Cartigny near Geneva.

Wold, H. 1973. 'Nonlinear Iterative Partial Least Squares (NIPALS) Modeling: Some Current
Developments', pp. 383-407 in Krishnaiah, P.R., ed., Multivariate Analysis 111. New York:
Academic Press.