This article was downloaded by: [Universitaetbibliothek Mannheim] On: 16 September 2008 Access details: Access Details: [subscription number 789268332] Publisher Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK German Politics Publication details, including instructions for authors and subscription information: http://www.informaworld.com/smpp/title~content=t713635220 The Dimensionality of German Federal States' Policy Preferences in the Bundesrat Susumu Shikano Online Publication Date: 01 September 2008 To cite this Article Shikano, Susumu(2008)'The Dimensionality of German Federal States' Policy Preferences in the Bundesrat',German Politics,17:3,340 352 To link to this Article: DOI: 10.1080/09644000802300627 URL: http://dx.doi.org/10.1080/09644000802300627 PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf This article may be used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.
The Dimensionality of German Federal States Policy Preferences in the Bundesrat SUSUMU SHIKANO Academic discussions have increasingly attested to the fact that state governments in the German Bundesrat represent partisan interests more than state-specific interests. This view seems to be confirmed by recent developments in which opposition parties exploit the Bundesrat to block the projects of the federal government. This contrasts with the behaviour of state governments in the earlier years of the Federal Republic, which were characterised to a greater extent by more heterogeneous interests. Using roll-call vote data in the Bundesrat, this article investigates to what extent preferences of the state governments are heterogeneous. More concretely, item response models are utilised to examine the dimensionality of the policy preference constellation. The results show that, besides the partisan left right dimension, there was another dimension at work in the 1950s. The analysis of the roll-call data after German reunification shows, by contrast, the strong growth of relevance of the first partisan dimension. INTRODUCTION A series of theories have been suggested and intensively discussed concerning the empirical validity of explanations of US Congress members behaviour. It is well known that the members of Congress in the USA have more freedom from party discipline than their counterparts in Europe. Shepsle and Weingast emphasise the important role of committees in this regard. 1 Krehbiel even suggested a model by assuming that the preference of individual members of Congress is independent from party and committee. 2 In contrast, there is also a series of research which rediscovers the important role of the parties. 3 Concerning empirical validity, the contribution of Poole and Rosenthal s analysis of roll-call data cannot be over-emphasised. 4 Using roll-call votes, or recorded votes, they placed individual members of Congress in a multidimensional space and visualised main conflict lines among them. In contrast to the USA, this kind of discussion is unusual in the European context since European parties control over individual members of parliament is often assumed to be much stronger than that of US parties. 5 One exception is the debate about the Federal Council of Germany (Bundesrat). The Bundesrat is one of two chambers in Germany s legislature and represents its federal states. Differently from the US Senate, the members of the Bundesrat do not represent the electorate directly, but they represent the state government. This chamber was installed to protect the interests of federal states with smaller populations from those of larger federal states since the latter s interests can be reflected more strongly in political decisions in the German Politics, Vol.17, No.3, September 2008, pp.340 352 ISSN 0964-4008 print/1743-8993 online DOI: 10.1080/09644000802300627 # 2008 Association for the Study of German Politics
FEDERAL STATES POLICY PREFERENCES IN THE BUNDESRAT 341 Federal Parliament (Bundestag). To articulate the state-specific interests, there are no formal parliamentary factions in the Bundesrat. However, academic discussions have increasingly attested to the fact that state governments in the German Bundesrat represent partisan interests rather than state-specific interests, in contrast to the original idea. 6 In contrast to the recent discussions, the dominance of partisan interests drew less attention until around 1970. Ziller, for example, concluded in 1968 that partisan politics is less visible in the Bundesrat. 7 In the later edition of the same book, by contrast, he added that after the formation of the federal-level coalition government of SPD and FDP, there were different majorities in the Bundestag and the Bundesrat. This increased the importance of partisan interests in the Bundesrat so that the policymaking of the federal government can be blocked by the Bundesrat. If this view is correct, we can expect a multi-dimensional constellation of federal states policy preferences for the earlier years of the Federal Republic, while a uni-dimensional partisan constellation between SPD- and CDU/CSU-governed states can be expected for more recent years. To test this empirically, this paper takes the same approach as Poole and Rosenthal s analysis, that is, an identification of policy preferences based on the roll-call data. This paper, however, utilises another statistical method, item response models, which has been increasingly applied in political science in recent years. 8 The remainder of this paper proceeds as follows. The next section introduces some historical background information about roll-call votes in the Bundesrat. The third section briefly introduces the data to be analysed and the statistical method utilised to identify the constellation. In the fourth section, multi-dimensional constellations of states policy preferences in the earlier years of the Federal Republic are investigated. The fifth section compares the result of the fourth section with the constellation after German reunification. The sixth section summarises and discusses the findings of this paper. ROLL-CALL VOTES IN THE BUNDESRAT As mentioned above, the German Bundesrat is composed of the delegates of the German federal states. Each member state has, independently of its attending delegates, three to six (before reunification three to five) votes according to its population. It is, however, not permitted that a state splits its votes. The states have to cast their votes unanimously. Furthermore, the Basic Law of Germany requires an absolute majority for decision-making. Abstention from voting is, therefore, equivalent to no or disagree with the proposal. At present, most of decisions are made by a show of hands. The voting behaviour of individual member states is recorded in the session protocol only upon request. This procedure of roll-calls is explicitly regulated by the rules of procedure (Geschäftsordnung) of the Bundesrat which have been changed several times since 1949. Provisional rules of procedure 1949 were passed in the fifth session of 20 October 1949. Accordingly, member states voted by a show of hands when rough unanimity existed in the plenum. Otherwise, a roll-call vote was requested by the president. The first official rules of procedure 1950, passed in the 34th session of 8 September 1950,
342 GERMAN POLITICS introduced as the regular form a roll-call vote system. Under this system, the member states were called in alphabetical order to cast their votes. After three years, the second rules of procedure were passed in the 114th session of 31 July 1953. It introduced voting by a show of hands which had also been practised despite inconsistency with the former rules of procedure. In contrast, the roll-call vote is used upon request of at least one member state. The current rules of procedure, which were passed in the 296th session of 1 July 1966, also follow this roll-call upon request. 9 The changes of the rules of procedure had an apparent impact on the number of roll-call votes (Table 1). From 1950 to 1952, there were more than 664 roll-calls, which exceeds the total sum of roll-calls in the other years in the history of the Federal Republic of Germany. At its high point, 1951, every session of the Bundesrat recorded on average nearly 10 non-unanimous roll-calls. However, shortly prior to the re-introduction of roll-call upon request in the second rules of procedure in 1953, the number of roll-calls began to decrease and the number of non-unanimous roll-call votes dropped to zero in 1959. METHOD The aim of this article is to identify the latent dimensionality of German federal states policy preferences using roll-call data. Factor analysis has been widely used in social sciences to investigate this kind of dimensionality. The application of this method to roll-call data, however, has some problems. The general problem of the analysis of roll-call data is that it is measured in binary form (yes/no) while factor analysis assumes continuous variables. Furthermore, the rollcall data to be analysed here has many missing values. First, in 1952, three southwestern federal states (Baden, Württemberg-Baden and Württemberg-Hohenzollern) merged into Baden-Württemberg. Therefore, the roll-calls of three predecessor states after the merger and those of Baden-Württemberg before the merger do not exist. Second, there was a change of government in some member states. In Rhineland- Palatinate, for example, the grand coalition of CDU and SPD was followed by the coalition of CDU and FDP in 1951. To differentiate the two governments, we are confronted with the same problem as in Baden-Württemberg. That is, rollcalls of the CDU/FDP coalition before the change of government and those of the grand coalition thereafter are missing values. Additionally, some member states are absent in multiple roll-calls. Their roll-calls also have to be coded as missing values. A solution to the problems above is provided by item response models via a Bayesian approach. Item response models are based on an idea similar to factor analysis, while this technique is suitable for binary variables. This is also the case for roll-call votes with two options, yes or no. 10 The basic idea of item response models can be depicted as in Figure 1. The vertical axis corresponds to the probability that a member state casts a yes vote for certain bills. The horizontal axis corresponds to the latent dimension of policy preference which can be interpreted as left right ideology as well. According to the solid curve in Figure 1, the more to the right a member state is positioned on the latent dimension, the more likely it cast a yes vote for the bill. The dashed line, in contrast, exemplifies a bill in terms of what voting behaviour of
FEDERAL STATES POLICY PREFERENCES IN THE BUNDESRAT 343 TABLE 1 DEVELOPMENT OF THE NUMBER OF ROLL-CALL VOTES (1949 2005) Year Sessions Roll-Calls Non-unanimous RCs Non-unanimous RCs (per session) 1949 10 23 16 1.60 1950 33 182 167 5.06 1951 32 338 318 9.94 1952 23 144 134 5.83 1953 19 25 20 1.05 1954 17 34 29 1.71 1955 17 18 11 0.65 1956 18 17 14 0.78 1957 17 8 5 0.29 1958 14 4 1 0.07 1959 13 3 0 0.00 1960 14 5 4 0.29 1961 12 8 5 0.42 1962 67 only unanimous RCs 1968 15 3 1 0.07 1969 17 10 6 0.35 1970 14 2 1 0.07 1971 17 5 5 0.24 1972 13 3 1 0.08 1973 74 only unanimous RCs 1975 14 4 1 0.07 1976 13 5 3 0.23 1977 82 only unanimous RCs 1983 12 2 1 0.08 1984 15 3 2 0.13 1985 only unanimous RCs 1986 13 2 1 0.08 1987 89 only unanimous RCs 1990 17 5 2 0.12 1991 only unanimous RCs 1992 12 6 1 0.08 1993 14 6 5 0.36 1994 14 5 4 0.29 1995 14 10 7 0.50 1996 15 5 4 0.27 1997 13 2 1 0.08 1998 only unanimous RCs 1999 13 5 3 0.23 2000 12 5 2 0.17 2001 13 6 4 0.31 2002 13 10 8 0.62 2003 only unanimous RCs 2004 12 2 1 0.08 2005 11 2 1 0.09 member states can be less forecast by their latent policy preference. The curve can also be reversed so that the latent left position increases the likelihood of yes votes. Based on the data it is statistically determined what kind of curve is appropriate for individual bills and where the member states are on the latent dimension. 11
344 GERMAN POLITICS FIGURE 1 GRAPHICAL REPRESENTATION OF ITEM RESPONSE MODELS The model described above presumes the following formula: Logit½ Probðyes vote of federal state i to issue jþš ¼ a j þ b j1 X 1i þ b j2 X 2i where X 1i is the position of state i on the first dimension and X 2i on the second dimension; a j corresponds to the constant in regression analysis. In Figure 1, this parameter determines the horizontal position of the curve. Substantively, it corresponds to the general level of agreement for issue j; b 1 and b 2 correspond to the direction and the steepness of the curve in Figure 1. The larger its absolute value of a dimension is, the more the voting behaviour is determined by the dimension. The problem of missing values can be solved by using the Bayesian approach which can treat missing values as parameters to be estimated. 12 Unlike the conventional maximum likelihood methods, the Bayesian approach does not aim to find a certain set of parameters which maximises the likelihood, but a certain area of parameter values by using simulation methods. Correspondingly, multiple sets of parameters are provided so that statistical inferences are enabled. Usage of this approach for the item response theory, however, has a drawback. Since the latent dimension has no a priori meaning in terms of directions, some sets of parameters can find a reversed picture of the latent dimension. To avoid this identification problem, the slopes of some issues have to be fixed a priori in positive or negative values. To find appropriate issues to be fixed, one needs some trial and error based on some prior knowledge from the correlation matrix. 13 CONSTELLATION OF THE FEDERAL STATES IN 1949 57 The roll-call data are divided into three sub-groups corresponding to the different rules of procedures. This is not only for the sake of estimation economy, but also to observe
FEDERAL STATES POLICY PREFERENCES IN THE BUNDESRAT 345 the selection bias. 14 As mentioned above, usage of roll-call voting was regulated differently under different rules of procedure. While the first rules applied it as the normal form of voting, the provisional and second rules provided for roll-call votes only upon request. What motivates one or more member states to request roll-call voting? Depending on the motivation, the recorded roll-call data could provide some biased information. For example, if the roll-call is requested in order to discipline the member states along the federal party s preference, the data of roll-call upon request would reflect partisan interests of member states rather than state-specific interests. 15 Since the roll-call data under the first rules of procedure do not suffer from this problem, we can also check the bias of roll-call upon request by conducting separate analysis. I begin with the analysis of roll-calls under the provisional rules of procedure. For this and other following analysis, I exclude the federal states which cast less than ten roll-call votes. Such states with less data would lead to an estimated position with a large confidence interval and bring less information. The issues with unanimous votes were excluded as well since such kind of issues bring no information concerning the underlying dimensionality. After these exclusion steps, there remained 65 non-unanimous roll-calls for the period under the provisional rules of procedure. According to Figure 2, both dimensions seem to represent to some degree the conflict between member states led by the CDU/CSU and those by the SPD. On the other hand, both dimensions contain some member states which deviate from their party line. On the horizontal dimension, Lower Saxony led by the SPD and North Rhine-Westphalia led by the CDU positioned themselves beyond the cutting line of partisan conflicts between both large parties. On the vertical dimension, three single-party governments, Baden (CDU), North Rhine-Westphalia (CDU) and Schleswig-Holstein (SPD), share very similar positions. This indicates that the vertical dimension reflects more non-partisan interests of member states. One of the issue on which the voting behaviour of member states can be better explained by the position on the vertical dimension is the cooperation between the federal and state-level governments. This indicates the more statespecific character of the vertical dimension. If one looks at the constellation as a whole, a group of four southern states, Bavaria, Baden, Württemberg-Hohenzollern and Rhineland-Palatinate, can be observed. This could indicate that a kind of north south cleavage and partisan interests are overlapped with each other for this period of time. As mentioned above, the Bundesrat revised its rules of procedure in 1950, according to which voting behaviour of member states was recorded without the explicit request by member states. The number of non-unanimous roll-call votes under the rules amounts to 583 roll-calls. According to Figure 3, which presents the identified two-dimensional constellation of federal states, again, the horizontal dimension differentiates better the cleavage between governments led by a CDU/CSU minister president and those led by SPD minister presidents. In contrast, there are some deviating states on the vertical dimension. The overlapping of the north south cleavage and partisan interests is less visible here in contrast to the constellation of Figure 2. The whole constellation of federal states is less clearly laid out. It is not easy and also beyond the scope of this article to interpret this dimensionality based on the issues
346 GERMAN POLITICS FIGURE 2 CONSTELLATION OF FEDERAL STATES UNDER THE PROVISIONAL RULES OF PROCEDURE Each label shows a member state and the composition of its government. NRW: North Rhine-Westphalia; Rhineland P.: Rhineland Palatinate; Schleswig H.: Schleswig Holstein; W.-Baden: Württemberg-Baden; W.-Hohenzollern: Württemberg-Hohenzollern. c: CDU/CSU; s: SPD; f: FDP; z: Zentrum. Circles are the member states led by a CDU/CSU minister president. Triangles are the states led by a SPD minister president. Crosses are the other states. since multiple roll-call votes were recorded for many issues due to the rules of procedure. It is, however, an important task for future work. In 1953, the Bundesrat re-introduced roll-call upon request in its second rules of procedure. Thereafter, 63 non-unanimous roll-calls were recorded until the 1957 Bundestag election. Figure 4 gives the constellation of federal states based on the 63 roll-calls. In contrast to the two constellations above, the groups of federal states are easier to find. In particular, the horizontal dimension clearly differentiates governments with CDU/CSU minister president from those with SPD minister presidents. During this period of time, there were some changes of minister presidents in terms of party affiliation. The governments in Berlin, Bavaria and North Rhine-Westphalia
FEDERAL STATES POLICY PREFERENCES IN THE BUNDESRAT 347 FIGURE 3 CONSTELLATION OF FEDERAL STATES UNDER THE FIRST RULES OF PROCEDURE Each label shows a member state and the composition of its government. Baden-W.: Baden-Württemberg; NRW: North Rhine-Westphalia; Rhineland P.: Rhineland Palatinate; Schleswig H.: Schleswig Holstein; W.-Baden: Württemberg-Baden; W.-Hohenzollern: Württemberg-Hohenzollern. c: CDU/CSU; s: SPD; f: FDP; z: Zentrum; b: BHE; d: DP. Circles are the member states led by a CDU/CSU minister president. Triangles are the states led by a SPD minister president. Crosses are the other states. changed the party affiliation of their minister presidents from CDU/CSU to SPD. In one case, the office of minister president changed in the other direction, that is, from SPD to DP, a conservative party. 16 If one looks at the movement of the positions of these states, its direction is uniform along the horizontal dimension while the movement on the vertical dimension is different among individual states (see dashed arrows in Figure 4). This, again, indicates that the vertical dimension reflects more state-specific interest than the horizontal ones. Thus far, we have confirmed that the horizontal dimension reflects partisan interests rather than state-specific interests, and vice versa concerning the vertical dimension. The most interesting question is to what degree these dimensions determine the voting behaviour of member states in the Bundesrat. This can be observed by
348 GERMAN POLITICS FIGURE 4 CONSTELLATION OF FEDERAL STATES UNDER THE SECOND RULES OF PROCEDURE Each label shows a member state and the composition of its government. Baden-W.: Baden-Württemberg; NRW: North Rhine-Westphalia; Rhineland P.: Rhineland Palatinate; Schleswig H.: Schleswig Holstein. c: CDU/CSU; s: SPD; f: FDP; z: Zentrum; b: BHE; d: DP; BP: Bavaria Party. Circles are the member states led by a CDU/CSU minister president. Triangles are the states led by a SPD minister president. Crosses are the other states. comparing correct predictions based on individual dimensions. In this regard, one has to note that simple correct prediction rates would be misleading. As demonstrated above, item response models decompose the vote probability into issue-specific constants and ideological terms. For example, if an issue was agreed by 60 per cent of member states, the constant term can automatically provide correct predictions of 60 per cent of voting behaviour. Therefore, one needs predictions of three different kinds of model specification: only constants, constants and first dimension, and constants and both dimensions. The improvement in correct prediction rates should be attributed to the predictive power of individual dimensions. Table 2 gives the corresponding results. Accordingly, the first dimension of partisan interests proves to always have better prediction rates than the second dimension. However, the ratio of
FEDERAL STATES POLICY PREFERENCES IN THE BUNDESRAT 349 TABLE 2 CORRECT PREDICTION OF VOTING BEHAVIOUR IN THE BUNDESRAT Only Constants Improvement by Adding 1st Dimension to Constants (a) Improvement by Adding 2nd Dimensions to Constants and 1st Dimension (b) Ratio of 1st to 2nd Dimension (a/b) Provisional rules of procedure 62.9% þ8.5pp þ5.0pp 1.7 1st rules of procedure 64.9% þ6.3pp þ4.3pp 1.5 2nd rules of procedure 64.3% þ10.5pp þ5.8pp 1.8 After reunification 64.5% þ21.1pp þ4.6pp 4.6 first dimension to second dimension shows no major dominance of the first dimension. The ratio is slightly lower under the first rules of procedure without roll-calls upon request. This may indicate that the bias of roll-calls upon request in favour of partisan interest. The difference is, however, not a significant one. COMPARISON WITH THE CONSTELLATION AFTER THE REUNIFICATION As mentioned in the introduction, academic discussions have attested to the fact that state governments in the German Bundesrat represent partisan interests more than state-specific interests, in particular, since the formation of the federal-level SPD/ FDP coalition government in 1969. To test this view empirically, the same method was applied to 42 non-unanimous roll-calls after German reunification. It would be preferable to analyse roll-call data also in the 1970s and 1980s. Unfortunately, the number of roll-call votes in this period of time is extremely low (see Table 1). Correspondingly, each state government participated in only a small number of roll-calls. Therefore, this analysis would be less conclusive. Figure 5 shows a constellation similar to that of Figure 4 along the horizontal dimension. The governments with CDU/CSU minister presidents and those with SPD minister presidents are relatively clearly divided along the horizontal dimension. Furthermore, grand coalitions ( cs and sc in the figure) and the social liberal coalition in Rhineland-Palatinate are positioned in the middle of the horizontal dimension while left-wing ( sg sp and s ) and right-wing coalitions ( cf and c ) take more extreme positions. In this sense, the order of coalitions on the horizontal dimension is more consistent than earlier. In contrast, governments in each group take different positions on the vertical dimension. It is, however, not clear which substantive meaning this vertical dimension possesses. This dimension shows neither a west east nor a north south nor a small large states cleavage. If one observes the prediction rate of each dimension after reunification, it is evident that the first dimension s predictive power has significantly increased while that of the second dimension remains at the same level (Table 2). This establishes the dominance of the first dimension of partisan interests against the second dimension. These results, at least, confirm the proposition of the increasing role of partisan interests in the Bundesrat.
350 GERMAN POLITICS FIGURE 5 CONSTELLATION OF FEDERAL STATES AFTER GERMAN REUNIFICATION Each label shows a member state and the composition of its government. Baden-W.: Baden-Württemberg; NRW: North Rhine-Westphalia; Rhineland P.: Rhineland Palatinate; Schleswig H.: Schleswig Holstein; Mecklenburg-P.: Mecklenburg-Pommerania; Saxony-A.: Saxony-Anhalt. c: CDU/CSU; s: SPD; f: FDP; g: the Greens; p: PDS; statt: STATT Party. Circles are the member states led by a CDU/CSU minister president. Triangles are the states led by a SPD minister president. Crosses are the other states. SUMMARY AND DISCUSSION Using roll-call vote data in the Bundesrat, this paper investigates to what extent preferences of the state governments are determined by partisan interests and further factors. To this end, item response models are utilised to examine the dimensionality of the policy preference constellation. The results show that, besides the partisan left right dimension, there was another dimension at work in the 1950s. The analysis of the roll-call data after German reunification shows, by contrast, the strong growth of relevance of the first partisan dimension. This increasing relevance of partisan interests in the Bundesrat has been pointed out by various researchers for a long time. However, this paper contributed to the widely shared view by providing empirical results based on a statistical method.
FEDERAL STATES POLICY PREFERENCES IN THE BUNDESRAT 351 We have to note that this article s results do not exclude another possible explanation: that the interests of federal parties and their state-level parties became more congruent. 17 Unfortunately, the analysis of roll-call data above cannot give any answer in this regard since it provides only policy preferences of individual federal states in the Bundesrat, but nothing about the federal parties preferences. One possible test could be enabled by a systematic analysis of state-level and federal-level party manifestos. 18 This would also provide external validation of the results above. A further interesting by-product of this article s results concerns the selection bias problem in using roll-call vote data. The above analysis of roll-calls under different rules of procedure showed only minor difference in terms of dimensionality. This is good news for researchers who are afraid of a partisan bias of roll-call votes. However, we also have to note that the results are based on the data from the earlier years of the Federal Republic of Germany. It is also possible that the motivation to request roll-call voting can change in the course of time. For this problem, more detailed analysis of protocols concerning the substantive content of agendas will be needed in future. NOTES 1. Kenneth A. Shepsle and Barry R. Weingast, The Institutional Foundations of Committee Power, American Political Science Review 81 (1987), pp.85 127. 2. Keith Krehbiel, Pivotal Politics: A Theory of U.S. Lawmaking (Chicago: University of Chicago Press, 1998). 3. See, e.g. Gary W. Cox and Frances Rosenbluth, The Electoral Fortunes of Legislative Factions in Japan, American Political Science Review 87/3 (1993), pp.577 89. 4. Keith T. Poole and Howard Rosenthal, Congress: A Political Economic History of Roll Call Voting (New York: Oxford University Press, 1997). 5. See, however, an exceptional study: Ulrich Sieberer, Party Unity in Parliamentary Democracies: A Comparative Analysis, Journal of Legislative Studies 12/2 (2006), pp.150 78. 6. Gerhard Lehmbruch, Parteienwettbewerb im Bundesstaat. Regelsysteme und Spannungslagen im Institutionengefüge der Bundesrepublik Deutschland, 2. erweiterte Auflage (Opladen: Westdeutscher Verlag, 1998). 7. Gebhard Ziller, Der Bundesrat, 2. überarbeitete Auflage (Frankfurt am Main: Athenäum Verlag, 1968). 8. For example, Joshua Clinton, Simon Jackman and Douglas Rivers, The Statistical Analysis of Roll Call Data, American Political Science Review, 98/2 (2004), pp.355 70; Andrew D. Martin and Kevin M. Quinn, Dynamic Ideal Point Estimation via Markov Chain Monte Carlo for the U.S. Supreme Court, 1953 1999, Political Analysis 10 (2002), pp.134 53. 9. For the historical development of rules of procedure see Konrad Reuter, Verfassungsrechtliche Grundlagen, Kommentar zur Geschäftsordnung, Praxis des Bundesrates, Praxishandbuch Bundesrat (Heidelberg CF: Müller Juristischer Verlag, 1991), pp.313 17. 10. Member states can also abstain from voting. However, as mentioned above, the Basic Law of Germany requires an absolute majority for decision-making in the Bundesrat. Therefore, abstention is treated as equivalent to no or disagree with the proposal in the following analysis. 11. For more detail see e.g. Clinton et al., The Statistical Analysis of Roll Call Data. 12. Jeff Gill, Bayesian Methods: A Social and Behavioral Sciences Approach (Boca Raton, FL: Chapman and Hall/CRC, 2002). 13. The list of issues whose direction is fixed is available from the author upon request. Furthermore, to facilitate intuitive interpretation of results, I give in the following the graphical constellation of individual federal states based on the mean estimate of X (see above). More detailed estimation results are also available upon request. 14. Gary King, Robert O. Keohane and Sidney Verba, Designing Social Inquiry: Scientific Inference in Qualitative Research (Princeton, NJ: Princeton University Press, 1994). 15. See for this and further kinds of discussion in terms of European Parliament Janina Thiem, Explaining Roll Call Vote request in the European Parliament, MZES Working Paper, 2006, p.90.
352 GERMAN POLITICS 16. The DP minister president, Heinrich Hollwege, later changed his party affiliation to the CDU. 17. Sven Leunig, Länder- versus Parteiinteressen im Bundesrat. Realer Dualismus oder fiktive Differenzierung?, Aus Politik und Zeitgeschichte B50 51 (2004), pp.33 8. 18. For the analysis of party manifestos in recent years see Marc Debus, Party Competition and Government Formation in Multi-level Settings: Evidence from Germany, Government and Opposition (2008: forthcoming).