1. For a summary discussion of this position, see Puhle, Hans-Jürgen, “Parlament, Parteien und Interessenverbände 1890–1914,” in Stürmer, Michael, ed., Das kaiserliche Deutschland: Politik und Gesellschaft 1870–1918 (Düsseldorf, 1970), pp. 340–77.Google Scholar

2. Rauh, Manfred, Die Parlamentarisierung des Deutschen Reiches (Düsseldorf, 1977).Google Scholar

3. See Gordon, Michael R., “Domestic Conflict and the Origins of the First World War: The British and German Cases,” Journal of Modern History 46 (1974): 191–226.CrossRefGoogle Scholar

4. On fragmentation in Germany, see Born, Karl Erich, “Structural Changes in German Social and Economic Development at the End of the Nineteenth Century,” in Sheehan, James J., ed., Imperial Germany (New York, 1976), pp. 16–38Google Scholar, and James J. Sheehan, “Conflict and Cohesion among German Elites in the Nineteenth Century,” ibid., pp. 62–92. On the integrative role of parties and parliamentary systems, see Almond, Gabriel A. and Powell, G. Bingham Jr., Comparative Politics: System, Process, and Policy, 2nd ed.(Boston, 1978), pp. 198–231.Google Scholar

5. For an outstanding party-level study, see Sheehan, James J., German Liberalism in the Nineteenth Century (Chicago, 1978)Google Scholar. Sheehan's study depends heavily on an analysis of social backgrounds.

6. Molt, Peter, Der Reichstag vor der improvisierten Revolution (Cologne and Opladen, 1963)CrossRefGoogle Scholar. See also O'Boyle, Lenore, “Liberal Political Leadership in Germany, 1867–1884,” Journal of Modern History 28 (1956): 338–52CrossRefGoogle Scholar, Sheehan, James J., “Political Leadership in the German Reichstag, 1871–1918,” American Historical Review 74 (1968): 511–28CrossRefGoogle Scholar, and Sheehan, German Liberalism. It should be noted that earlier German legislatures have been subjected to quantitative analysis of voting behavior. See Mattheisen, Donald J., “Voters and Parliaments in the German Revolution of 1848: An Analysis of the Prussian Constituent Assembly,” Central European History 5 (1972): 3–22CrossRefGoogle Scholar, and Mattheisen, Donald J., “Liberal Constitutionalism in the Frankfurt Parliament of 1848: An Inquiry Based on Roll-Call Analysis,” Central European History 12 (1979): 124–42.CrossRefGoogle Scholar

7. These Reichstags were selected for initial treatment for a number of reasons. This project grew in part out of a study of imperialist politics in Germany. During the period 1898–1906, the Reichstag played probably its most important role in German imperialism (the Naval Laws, etc.), and the data base was selected accordingly. When it was decided to extend the study to political behavior in the Reichstag as a whole, we also decided to work out our approach with a limited sample before dealing with all of the data from the post-1890 period. The period 1898–1906, for which our data were most complete, seemed an appropriate source for the sample in historical terms. These Reichstags took a number of important decisions and had come out from under the shadow of Bismarck and William II. This paper represents, not a completed analysis, but a first attempt to use various methodologies of legislative behavioral analysis to examine data from the Reichstag. We hope that once our methods become more refined and problems that appeared in the initial analysis have been overcome, we shall be able to undertake a comprehensive quantitative analysis of the post-Bismarckian Reichstag.

8. The biographical entries, based on questionnaires filled out by the members themselves, are found in Amtliches Reichstags-Handbuch: Zehnte Legislaturperiode, 1898/1903 (Berlin, 1898)Google Scholar and Amtliches Reichstags-Handbuch: Elfte Legislaturperiode, 1903/1908 (Berlin, 1903)Google Scholar. Additional biographical information is contained in Schwarz, Max, MdR: Biographisches Handbuch der Reichstage (Hanover, 1965)Google Scholar, although we relied primarily on the first two sources. Certain deputies elected in by-elections were left out of our sample because of lack of information about them at the same level as our information about the other deputies. Reports of roll-call votes are found throughout the official Stenographische Berichte of Reichstag proceedings between 1898 and 1906. We did not include a geographical factor in this initial analysis because of difficulties in selecting a suitable geographical unit to employ. Geographical factors will be included in the final, full-length study to be based on this research.

9. It should be noted that we, like practically all students of the social composition of legislatures, employ data on members' occupations as substitutes for exact information on hard-to-measure characteristics such as class and for unavailable data on incomes. Our scheme of occupational groupings is largely of our own creation, although most of the categories are similar to those used in other analyses. (See, for example, Kaelble, Hartmut, “Social Mobility in Germany, 1900–1960,” Journal of Modern History 50 (1978): 439–61CrossRefGoogle Scholar.) The scheme does not run the entire gamut of German society for the obvious reason that many occupational groups had little or no representation in the Reichstag. Many of these categories are in fact amalgamations, undertaken for convenience in presentation and calculation, of narrower categories used in the early stages of our analysis. The only one that presents much difficulty is “landholder,” which covers the range of agricultural proprietors from great landlords to small peasant farmers and includes professional agricultural managers. There are several reasons for including these in one category, the most important being that it is in many cases impossible to tell to what landholding category an individual belongs without referring to other factors such as his education, and then inferring the agrarian group to which he “must” belong. In general, it appeared to us that many of the categories into which agricultural proprietors and managers are usually divided are identified, not by economic, income, and occupational differences so much as by region, inferred social status, and political stance. These subcategories are not sufficiently independent of the other factors in the analysis for us to base the confidence in them that other researchers apparently have. In general, we have attempted to create the broadest possible categories that appeared to be consistent with our data and our understanding of the structure of Wilhelmine society.

10. Molt, , Reichstag, pp. 276–80.Google Scholar

11. Both Sheehan, , “Political Leadership,” pp. 511–28Google Scholar, and Molt, , Reichstag, pp. 274–76Google Scholar, recognize the existence and importance of this group. In 1898, 46.3 percent of deputies of major parties belonged to traditional “notable” occupational groups, 14.9 percent were apparently professional politicians, and 20.5 percent were businessmen or members of recognized professions. In 1903, the figures were 42.7 percent notables, 19.1 percent politicians, and 17.8 percent businessmen and professionals. The heavy concentration of professional political employees in the SPD was probably due to the fact that most members of the SPD could not afford the costs of belonging to the Reichstag out of private incomes and therefore needed jobs with party organs that would give them adequate free time. Until 1906, Reichstag deputies were not paid by the state.

12. In 1898, 80.6 percent of the deputies were commoners, 8.8 percent were minor noblemen entitled to use “von” before their surnames, and 10.6 percent were titled noblemen. In 1903, the percentages were 81.4 percent commoners, 10.7 percent minor noblemen, and 7.9 percent titled nobles. The percentage of noblemen among the Conservatives was 59.3 percent in 1898 and 56.9 percent in 1903.

13. Of the 1898 deputies, 51.6 percent had university educations. The figure was 50.0 percent in 1903. The small downward turn was probably due to the almost 50 percent increase in the SPD delegation in 1903, since the SPD had the lowest proportion of deputies with university educations. The Poles had the highest proportion of former university students (about two-thirds.)

14. Molt, , Reichstag, pp. 61–70Google Scholar; Rice, Stuart A., Quantitative Methods in Politics (New York, 1928), p. 297.Google Scholar

15. In 1898, at least 52.9 percent of the members either held or had held seats in a lowerlevel assembly, most often a Landtag. In 1903, the figure was 52.3 percent. These figures must be regarded as minimums, since the holding of other offices was mentioned in our biographical sources only in a nonobligatory “miscellaneous comments” section. Plural officeholding was partly a means of obtaining compensation for professional politicians, since, unlike the Reichstag, many state assemblies paid their members.

16. In 1898, 34.8 percent of Reichstag members were Catholics, 53.5 percent belonged to various Protestant denominations, 0.3 percent were Old Catholics, 1.0 percent were Jews, and 10.3 percent indicated that they had no religion. In 1903, the figures were 34.7 percent Catholics, 49.9 percent Protestant, 0.5 percent Old Catholic, 1.0 percent Jewish, and 13.7 percent no religion. In general, age differences seem to have had negligible effects on voting in all parties. See Figure 1. The “age-bulge” effect was more marked for the National Liberals and the Freisinnige Volkspartei than for the Freisinnige Vereinigung.

17. See, for example, Molt, , Reichstag, pp. 18–37, 332Google Scholar; Kehr, Eckart, Battleship Building and Party Politics in Germany 1894–1901, trans. Anderson, Pauline R. and Anderson, Eugene N. (Chicago and London, 1973), pp. 3–27, 306–58Google Scholar; Holborn, Hajo, A History of Modern Germany 1840–1945 (New York, 1969), pp. 320–26, 349–50.Google Scholar

18. For a brief explanation of eta, see Nie, Norman H., et al. , Statistical Package for the Social Sciences, 2nd ed.(New York, 1976), p. 230Google Scholar. For more detailed treatments of lambda, gamma, and the other measures used here, see Blalock, Hubert M. Jr., Social Statistics, 2nd ed. (New York, 1972)Google Scholar. On roll-call vote analysis in general, see Anderson, Lee F. et al. , Legislative Roll-Call Analysis (Evanston, III., 1966).Google Scholar

19. Summary of t-tests (independent-sample, two-tailed) of differences between means performed on eta-scores using electoral periods as treatment groups (eta-scores are measures of association between social characteristics and voting behavior on specific roll-call votes):

20. Summary of Pearson's product-moment correlations performed on eta-scores between social variables:

It will be noted that not all product-moment correlations maintain themselves at the same levels between electoral periods. Since we have already determined that age is an unimportant factor in predicting voting outcomes, the lack of continuity when age is paired with other variables is also probably unimportant. In the case of the substantial drop in the value of r for the “party-occupation” pair, it can be hypothesized that between electoral periods, parties on the whole may have becomes less characterized by particular occupational concentrations.

21. Lambda is a measure of association between nominal variables in crosstabulations, measured on a scale from 0.0 to 1.0. The higher the value of lambda, the stronger the association. Lambda measures the extent to which knowledge of the value of an independent variable increases ability to predict the value of the dependent variable—in this case, for example, how much our knowledge of a member's occupation increases our ability to predict his party. In the calculation of lambda, it is possible to regard either variable as the independent one (thereby producing two separate measures) or to take a kind of average (the symmetric value). See Blalock, , Social Statistics, pp. 302–3.Google Scholar The values obtained for lambda were:

22. Three-way crosstabulations which were run between party and individual roll-calls, controlling for occupation, showed no significant reduction in the association be tween party membership and voting, thus confirming the dominance of the party factor.

23. The smaller parties would be worth a study in themselves. One of them, the Polish fraction, was not officially recognized as a party but was actually larger than the Freisinnige Vereinigung in our period. But to have left out the Vereinigung would have given a mistaken impression of the strength of the liberal left. An examination of individual votes indicates that the minor parties had little impact on voting in our period.

24. The most difficult question in cohesion analysis is how absences and abstentions should be counted. Any method involves distortion. We have worked, therefore, with three different indices of our own construction (defined in Table 2) which seem to cover most possibilities.

25. See Clausen, Aage R. and Holmberg, Sören, “Legislative Voting Analysis in Disciplined Multi-Party Systems: The Swedish Case,” in Aydelotte, William O., ed., The History of Parliamentary Behavior (Princeton, 1977), pp. 165–66.Google Scholar

26. See Nipperdey, Thomas, Die Organisation der deutschen Parteien vor 1918 (Düsseldorf, 1961).Google Scholar

27. The relative social cohesiveness of the parties was measured with the occupational information contained in Table 1, after it was determined that the amount of variance in other social categories (education and religion) within parties was so uniform as to make ranking on those factors meaningless. The parties were ranked on social cohesiveness according to the minimum number of occupational categories in Table 1 that were required to account for 75 percent of the parties' Reichstag members. Parties accounted for by the same number of categories were ranked on the basis of the difference between the percentage amount of the membership covered by the next lower number of categories and 75 percent. The rankings thus obtained were tested for correlations with the rankings of voting cohesion summarized in Table 2 using Spearman's rho, a simple measure of ranklist correlation. In 1898–1903 session, for example, the ranking of social cohesion among the parties, from highest to lowest, was Conservatives, Free Conservatives, Social Democrats, National Liberals, Freisinnige Volkspartei, Freisinnige Vereinigung, and Center. This ranking was correlated with the “A” ranking of voting cohesion at a factor of r_s = −0.11, and with the “B” ranking at a factor of r_s = 0.39, both of which indicate quite low levels of correlation.

28. It is possible to cite cases of this sort of process in the resolution of a great many issues before the Reichstag. See, for example, the decision of the Center on the Naval Law of 1898, described in, among other places, Gottwald, Herbert, “Der Umfall des Zentrums: Die Stellung der Zentrumspartei zur Flottenvorlage von 1897,” in Klein, Fritz, ed., Studien zum deutschen Imperialisms vor 1914 (Berlin, 1976), pp. 181–223.Google Scholar Here, as in other cases, we can see the compromising and consensus-building functions of a party at work. Sometimes, of course, they did not work, as when the Conservatives split over the 1902 tariff reform because the intransigence of the Bund der Landwirte was proof against the workings of party loyalty and discipline.

29. This technique is somewhat complicated. For a full discussion, see Clausen, and Holmberg, , “Legislative Voting Analysis,” pp. 159–85.Google Scholar Briefly, one lists, for each roll-call in the sample, the parties that voted “yes” and those that voted “no.” One then makes a list of all possible orderings of parties in sequence (party A, party B, party C; party A, party C, party B; etc.) Whichever of these orderings is consistent with the lineup of parties on the greatest number of roll-calls is taken to reveal the dominant dimension of attitudes at the party level.

30. Gamma coefficients were calculated between individual members' voting patterns and the dominant party-level dimension. These indicated a high degree of correlation between rank-ordering and individual voting behavior (

in 1898–1903 and 0.89 in 1903–6). Gamma is a measure of association between variables that are at least at the ordinal level. It measures concordance and discordance between all pairs of items in a table. Values range from −1.0 (highly discordant) to 1.0 (highly concordant). See Blalock, , Social Statistics, pp. 424–26, 440–42.Google Scholar

31. On the tariff question in general, see Barkin, Kenneth D., The Controversy over German Industrialization, 1890–1902 (Chicago, 1970)Google Scholar, and Puhle, Hans-Jurgen, Agrarische Interessenpolitik und preussischer Konservatismus (Hanover, 1966).Google Scholar On Erzberger and the Center, see Epstein, Klaus W., Matthias Erzberger and the Dilemma of German Democracy (Princeton, 1959), pp. 38–60Google Scholar, and Zeender, John K., The German Center Party, 1890–1906 (Philadelphia, 1976), pp. 99–101.Google Scholar

32. The tariff issue was of course one of the major questions of Wilhelmine politics. Bülow's general raising of tariffs in 1902, bitterly fought by liberals in the Reichstag, is usually explained as the result of an uneasy realliance of big Prussian agriculture and heavy industry and of Bülow's attempt to turn the alliance to his own advantage. Proposals for changing the criminal law always brought out a dichotomy between supporters of “law and order” and libertarians, as well as between those who favored outlawing the SPD and those who did not. This split seems, on the basis of our analysis, to have been very similar to that on the tariff question. The full significance of the party-level analysis and the individual-level factor analysis that follows depends, of course, on correlating the results with a detailed examination of issues before the Reichstag. Lack of space precludes our doing this here.

33. Roll-calls selected for factoring were those with a gamma value of less than 0.95 in the dominant party-level dimension. A principal-components factoring method, with a criterion minimum eigenvalue of 1.0 for retention of a factor, was used to extract the initial-factor matrix. Terminal solution was obtained by direct oblimin oblique rotation, δ=0. “No” votes were coded as 1, “not voting” as 2, and “yes” as 3. The creation of an interval scale in this manner involves assumptions, the validity of which is frequently debated. For this as for other reasons given in the text, our factor analysis must not be regarded as a definitive treatment of the factors present in the data. We selected factor analysis rather than other forms of dimensional analysis (cluster analysis, nonmetric multidimensional scaling, etc.) solely because programs for factor analysis were most readily available for use with the computer system with which we worked. Factor analysis is an adequate technique for the exploratory purposes outlined here. For more detailed and systematic analysis and testing of hypotheses, other forms of dimensional analysis are probably more appropriate. See Horst, Paul, Factor Analysis of Data Matrices (New York, 1965), esp. pp. 492–595.Google Scholar

34. See Williamson, John G., Karl Helfferich 1872–1924: Economist, Financier, Politician (Princeton, 1971), pp. 39–42.Google Scholar

35. On “liberal imperialism,” see Mommsen, Wolfgang J., Max Weber und die deutsche Politik, 2nd ed. (Tübingen, 1974), pp. 97–146.Google Scholar

36. On these and related issues, see Witt, Peter-Christian, Die Finanzpolitik des deutschen Reiches von 1903 bis 1913: Eine Studie zur Innenpolitik des wilhelmischen Deutschlands (Lübeck and Hamburg, 1970), esp. pp. 80–110.Google Scholar

37. The politics of the December 1906 dissolution and the ensuing election are discussed in Crothers, George Dunlop, The German Elections of 1907 (New York, 1941)Google Scholar, and Smith, Woodruff D., The German Colonial Empire (Chapel Hill, 1978), pp. 183–91.Google Scholar

38. On the question of parties and interest organizations in general, see Stegmann, Dirk, Die Erben Bismarcks: Parteien und Verbände in der Spätphase des wilhelmischen Deutschlands (Cologne, 1970).Google Scholar