Ministerial turnover in the German Länder (1991–2010)

A more recent development is the increase in studies on governmental elites in non-Western countries (e.g. Nikolenyi and Fettelschoss 2009; Stefan 2009).

More recent work within the research network on the selection and deselection of political elites (SEDEPE) focuses exclusively on the subnational level (e.g. a conference on subnational political elites took place in Montreal in October 2012. See www.​sedepe.​net).

For reasons of simplicity the Ministerpräsidenten of the 13 territorial Länder as well as the governing mayors (Regierende Bürgermeister) of Berlin, Bremen and Hamburg are referred to as prime ministers in this article.

The fact that this article explicitly focuses on individual ministers distinguishes it from other work that considers ministerial turnover from the perspective of the prime minister and asks under which conditions and for what reasons he or she reshuffles the cabinet, demotes or promotes ministers and induces individual ministers’ resignations (Dewan and Dowding 2005; Indridason and Kam 2008).

Party careers represent another research perspective on political elites in previous work.

Only full ministers are counted, state secretaries without a vote in cabinet are not included in the analysis.

This research focus is different from existing studies such as Indridason’s and Kam’s study on the prime minister’s ability to cope with ministerial drift. In their research, all reshuffles—transfers (a minister moving from one portfolio to another), redefinitions of a minister’s portfolio, promotions of new cabinet members as well as demotions—are defining elements of their cases (Indridason and Kam 2008, pp. 639–640). In contrast, the study at hand focuses on ministers who leave the cabinet. Transfers as well as redefinitions are included as independent variables in terms of the number of changes the ministers were faced with during their time in office.

The focus on individual ministers also speaks against a common definition of reshuffles as the simultaneous change of two or more ministers (Budge 1985). In the present analysis all ministers leaving the cabinet are taken into account, regardless of whether they exited cabinet alone or alongsideother colleagues.

The dataset as well as the stata-do files are available at the author’s website: www.sebastianjaeckle.de.

I have to thank Klaus Stolz and Jörn Fischer for drawing my attention to this simple yet helpful typology.

This category subsumes all types of scandals. The large majority are nevertheless either of a financial or political nature. Sex scandals which—according to Dowding and Kang (1998, pp. 419–425)—make up a considerable portion of British ministerial resignations particularly for conservative politicians, are apparently not a big issue in the German Länder, as Jörn Fischer and his colleagues have also pointed out for the federal level (Fischer et al. 2006).

According to the constitutions of the Länder, a vote of no confidence is possible in all Länder except Bavaria, where the prime minister is required to step downwhen the political conditions preclude a trustful collaboration with parliament. Motions of confidence are only included in the constitutions of Brandenburg, Hesse, Hamburg, Mecklenburg-West Pomerania, Saarland, Saxony-Anhalt, Schleswig-Holstein and Thuringia.

The confidence intervals that were not presented in Fig. 2b for the sake of clarity widely overlap.

Jörn Fischer and his colleagues have shown that, at least for German federal ministers, the Federal Chancellor is often decisive for the minister’s fate when the opposition or the media confront cabinet ministers with demands for resignation (Fischer et al. 2006, p. 730).

The data were provided by Marc Debus (2008). In order to determine salience-values and policy positions Debus applied the words core technique to electoral programs. His period of analysis covers 1994–2006. Consequently, it was possible to determine the policy distance for only about half of the cases in this study.

The empirical values of the ideological distances range from approximately 9–34. To account for ideological differences of factions within single party governments the ideological distance value was set to 3 for all SPG.

The classification of ministries is based on Pappi et al. (2008).

Distinguishing between ministers with and without tertiary education did not produce meaningful variance. Only three ministers in the whole sample did not attend university.

This argument resonates with a study which claims that the talent pool of potential ministers is not infinite (see Dewan and Myatt 2010). Therefore, from the perspective of a prime minister who wishes to work with the most talented personnel, recourse to ministers that have already proven their qualities in earlier governments often seems sensible.

Consequently, it is likely that the proportionality assumption does not hold for this variable and must therefore be tested.

Party membership, included in the Cox model as a time-varying covariate, is measured in days but standardized on years as this facilitates the interpretation of the coefficients.

CDU-East and the Democratic Farmers party of Germany (DBD) became part of the West CDU, while the Liberal Democratic Party of Germany (LDPD) and the National Democratic Party of Germany (NDPD) merged with the FDP. There was no SPD blocparty as the social democrats were forced to merge with the ruling SED alongside the communist party (KPD) in 1950. Consequently, unlike the CDU and the FDP, the SPD could not rely on organizational structures provided by the bloc parties. This was one of the most serious difficulties the Social Democrats were faced with in the new Länder after 1990.

For further information on this class of event history models and specific problems such as the proportionality assumption, tied events and the partial likelihood estimation procedure see (Blossfeld et al. 2007; Box-Steffensmeier and Jones 2004; Jäckle 2011, pp. 75–91; Therneau and Grambsch 2000).

For further information on how to interpret the partial likelihood coefficients of a Cox model see Box-Steffensmeier and Jones (1997, p. 1450).

Since log-log-plots as well as the Grambsch-Therneau-test indicate non-proportional hazards for the variable portfolios within spell (Grambsch and Therneau 1994; Therneau and Grambsch 2000, p. 127), an interaction with survival time was included in the model to explicitly account for this fact (see Box-Steffensmeier and Jones 2004, p. 136). The estimation of stratified models, which according to some textbooks is one way to cope with non-proportional hazards, has a number of drawbacks (see Therneau and Grambsch 2000, p. 145) and was therefore not applied in this analysis. In order to determine the appropriate form of time dependence I inspected plots of scaled Schoenfeld residuals against different functional forms of time (identity, log(t), t2) (see Box-Steffensmeier and Jones 2004, pp. 120–121). Since the identity plots provided the best results and since they are sensible from a conceptual point of view as they ease the interpretation of the interactions, the identity interactions were used in the models. Other variables show no signs of non-proportionality.

The Magdeburg model was a minority government comprising the SPD and Greens in 1994, which was tolerated by the PDS. When the Greens did not manage to re-enter parliament in 1998, prime minister Reinhard Höppner continued with a SPD single-party minority government supported by the PDS. Due to this support policy, the Magdeburg model cabinets resemble standard majority coalition governments. It is therefore not surprising that ministers in these governments tend to have longer tenure than ministers in other East German governments.

As outlined above, prime ministers exhibit significantly longer durations in cabinet. Due to these major empirical differences and because of a number of theoretical arguments (e.g. constitutional requirements and the principal-agent perspective) I decided not to include prime ministers in the statistical analysis. I believe that it is a fruitful research agenda in its own right and should therefore not be commingled with the question of turnover of regular ministers.

When interpreting these lower levels of significance, it should also be taken into account that the number of cases available for the East German subsample is much smaller.