1 Introduction
According to partisan theory, the ideological positioning of the incumbent government affects the direction of public policy (Hibbs Reference Hibbs1977; Schmidt Reference Schmidt1996). However, what are the consequences of a government which comprises the two major national parties, representing policy goals both of the left and of the right? Neo-liberals often fear a surge in state activity, as each of the two major parties try to satisfy the demands of their own electorates, thereby causing government spending to skyrocket under a grand coalition. Because they are more inclusive, grand coalitions as special forms of consensual governments are said to spend more on public policy. In contrast to majority governments, which are defined as one party majority executives as well as minimal winning coalitions, consensual governments generally comprise grand and oversized coalitions as well as minority governments (de Winter Reference Winter de2005; Lijphart Reference Lijphart1999). Their aim is thus to accommodate the demands of as many voters as possible (e.g., Crepaz Reference Crepaz1996a, Reference Crepazb; Reference Crepaz2002; Lijphart Reference Lijphart1999; Tavits Reference Tavits2004).
A grand coalition consisting of the two major German parties – the Christian Democrats (CDU/CSU) and the Social Democrats (SPD) – was formed in 2005 to head the parliament. Germany had only once before seen such a constellation of parties hold office after World War II – namely, from 1966 to 1969. For this reason, both political scientists and the politically interested public are rather unsettled by the public policy consequences of such an unlikely form of government in Germany.
This paper sheds empirical light on the impact of coalition governments by examining the influence of coalition governments on the size of the public sector in the German context by evaluating the situation in the Laender from 1992 to 2005. In doing so, we investigate whether grand coalitions, i.e., CDU/CSU and SPD coalitions, at the sub-national level are systematically related to the degree of state intervention in the fields of education policy and internal security in Germany. We also evaluate whether the effects of grand coalitions on the size of government are moderated by partisan politics. We concentrate on government expenditure for education and internal security policies because according to the Basic Constitutional Law of the Federal Republic of Germany (Basic Law, Art. 70 ff.), the Laender are granted ample room for maneuver in these policy areas. Unlike at the national level, Laender level governments have frequently been comprised of the two major parties – a further aspect which speaks to the choice of a sub-national analysis. Grand coalitions consisting of the two major parties (CDU and SPD) were in office in half of the Laender during the time period under consideration.Footnote 1 Particularly since the beginning of the 1990s, this type of government occurred much more frequently in the Laender than during the years following the end of the 1950s (Jun Reference Jun1994). From 1992 to 2005, grand coalitions as Laender executives governed in eight of the 16 states. In those states where this special kind of combination constituted the government, some of these grand coalitions were in power for over half the period of investigation. Consequently, nearly one-fifth of the governments between 1992 and 2005 were grand coalitions. The city-states of Berlin and Bremen and the Land Brandenburg experienced by far the longest era of grand coalitions during the observed time period. By contrast, grand coalitions did not constitute the executives in most of the western states – with the exception of Baden-Württemberg and recently Schleswig-Holstein (since 2005) (Schniewind Reference Schniewind, Freitag and Vatter2008).
Furthermore, variance between the Laender exists concerning state activities and public expenditures (Schmidt Reference Schmidt1980; Wolf/Hildebrandt Reference Downs2008). However, one major difference in governments' spending with regard to the two policies emerges: Education and cultural affairs imply much greater costs for the states' budgets than domestic security does. The states' mean expenses from 1992 to 2005 for education (including cultural affairs) are roughly five times higher than internal security spending. With regards to expenditures for education and cultural affairs during this period, major differences between the diverse sub-national units can be observed. The expenses range from about 630 Euro per capita in Rhineland-Palatinate to almost 1000 Euro per capita in Berlin and the Free State of Thuringia. Concerning the states' expenses for internal security, the numbers vary from 115 Euro per capita in North-Rhine-Westphalia to 267 Euro per capita in Berlin.Footnote 2
Moreover, the German Laender are particularly well-suited to an evaluation of the consequences of such governments in a comparative sub-national perspective (Snyder Reference Snyder2001, Vatter/Rüefli Reference Vatter and Rüefli2003). While they do share a general structural framework, they are nevertheless idiosyncratic enough not to be seen as a single, homogenous entity (Freitag/Vatter Reference Freitag and Vatter2008; Snyder Reference Snyder2001), a characteristic that lends itself particularly well to the application of the comparative method (Lijphart Reference Lijphart1971, Reference Lijphart, Freitag and Vatter2008). Particularly in recent years, the attention paid to the German Laender has been amplified: For example, in the literature on political economy, the Laender have provided many authors with a new domain for the application of their theories and for the verification of findings of international comparative research at the sub-national level (e.g., Jochimsen/Nuscheler Reference Jochimsen and Nuscheler2007; Galli/Rossi Reference Galli and Rossi2002; Schneider Reference Schneider2007; Seitz Reference Seitz2000; Wagschal Reference Wagschal1996; Wolf Reference Wolf2006). A forerunner is Schmidt's (Reference Schmidt1980) study on the differences between the policies of the two major parties in the Laender, wherein he finds substantial distinctions between these sub-national units concerning education and domestic security policies.
The article is structured as follows: Section 2 presents relevant theories concerning public policy and the specific aspects of consensus democracy, as well as the veto-player theory. From this theoretical background, three main hypotheses are derived. Subsequently, the methodological approach, data, and important influential factors on public policy are briefly presented in section 3. In section 4, the empirical findings are presented and discussed. Cross-sectional regression models are estimated to assess the impact of various factors on the size of government in education and internal security policies. The article closes with a conclusion in section 5.
2 Grand Coalitions and Public Policy: Theories and Hypotheses
The assumption that distinct forms of coalition governments (i.e., grand coalitions between the CDU and the SPD) act as catalysts for state intervention in the public policies of the German Laender is at the core of the present analysis. This view is supported by the key insights of the concept of consensus democracies and the veto-player theory – namely, that consensual governments or collective veto points encourage government spending (Crepaz Reference Crepaz2002; Lijphart Reference Lijphart1999; Tsebelis Reference Tsebelis2002; Vatter/Freitag Reference Vatter and Freitag2007).
Lijphart's (1999) famous concept of types of democracy ranges between the two ideal points of majoritarian and consensual systems, with broad and oversized multi-party coalition governments building a cornerstone of the latter. This particular characteristic of governments is the strongest element of the ‘executives-parties dimension’ and ‘can also be seen as conceptually close to the essence of the distinction between concentration of power and the joint exercise of power’ (Lijphart Reference Lijphart1999: 245). The ideal type of a consensus democracy is most suited for heterogeneous societies: Executive power is dispersed and spread over major parts of the society. It is argued that a simple ‘majority rule is not only undemocratic but also dangerous’ if the access to the decision-making process is continually denied to large minorities of a society (Lijphart Reference Lijphart1999: 32). Consensus democracies, with their various manifestations, are ‘better at […] representing minority groups and minority interests, representing everyone more accurately, and representing people and their interests more inclusively’ (Lijphart Reference Lijphart1999: 275). In sum, the advantages of consensual political structures are that policies are more carefully deliberated, a broader spectrum of perspectives is weighed and taken into consideration, and a larger number of options are examined than in majoritarian systems (Crepaz Reference Crepaz1996a: 87–88). This amplitude of representation causes governments to be more responsive to the voters' demands. Among other things, a ‘closer proximity between government policy and voters' preferences’ is also said to be a characteristic of consensus democracies (Lijphart Reference Lijphart1999: 293). This should lead to an expansion of governments' role in society because the needs of a greater proportion of the population have to be met (Tavits Reference Tavits2004). Consequently, the size of government should increase (Tavits Reference Tavits2004). Moreover, Crepaz (Reference Crepaz1996b: 9) classifies multi-party governments with high popular support as ‘encompassing organizations’ in terms of Mancur Olson (Reference Olson1986). He states that ‘the more encompassing parties become, the more their interest and the ‘general interest’ converge; thus they tend to behave more responsibly by minimizing redistributive policies favoring particular groups’ (Crepaz Reference Crepaz1996b: 9). For the reasons stated above, Lijphart (Reference Lijphart1999: 294) again, points out that consensual democracies are associated with ‘kinder, gentler and more generous policies.’
With regard to the veto player theory, it is argued that the greater the number of veto players, the more difficult it is to change the status quo, thereby limiting the capacity of state expansion (Tsebelis Reference Tsebelis2002). However, according to Crepaz (Reference Crepaz2001, Reference Crepaz2002: 173) and Wagschal (Reference Wagschal, Busch and Plümper1999a), there are different kinds of veto players with contradictory effects on policy outcomes. As far as competitive veto points – separate agencies with mutual veto powers, as in federalism and direct democracy – are concerned, one would indeed expect them to restrain government (Obinger Reference Obinger1998; Vatter/Freitag Reference Vatter and Freitag2007). Here, political power is diffused by means of institutional separation and mutual veto power leading to deadlock and restrictive effects. By contrast, collective veto points, such as parliaments, interest groups, or parties in a coalition's government favor ‘policies with expansionary consequences’, as the resistance to cutbacks ‘is certainly more expedient politically’ (Crepaz Reference Crepaz2002: 173–4). Political power is dispersed within these collective veto points where responsibility is shared within a single body. This leads to extended negotiation and logrolling, which is expected to have an expansive effect on government expenditure. In this regard, parties in a coalition's government share collective authority and interact with one another on a face-to-face-basis without the protection of separate institutions with mutual veto powers. Within partisan veto players, there is an inherent bias of all coalition partners with regards to expansionary policies through logrolling. While the parties must interact with one another on an ongoing basis, they nevertheless have distinctive constituencies with distinctive preferences. Overall, coalition governments have a reduced capacity to exercise restraint in government expenditures. Against this background, Persson et al. (Reference Persson, Roland and Tabellini2003) find that coalition governments generally spend more than one party governments. Rising government expenses are shown to be positively associated with an increasing number of governing parties (Crepaz Reference Crepaz2002) as well as with an increasing number of parties in national legislatures (Mukherjee Reference Mukherjee2003). This congruence of the concepts of collective veto players, encompassing organizations, and consensus democracy allows us to formulate the following main hypothesis:
Hypothesis 1: Multi-party coalition governments enhance government spending:
Where grand coalitions more frequently control the Laender executives, the governments in these Laender are larger.
In addition, we expect the influence of grand coalitions on the size of government to be strengthened by partisan politics reflecting the political-ideological center of the respective coalition. The hypothesis of partisan influence on public policy, the ‘parties-do-matter’ view, is a stylized empirical theory of a democratic political market (Schmidt Reference Schmidt1996). According to this approach, a major determinant of variation in policy choices and policy outputs in democracies is the strength of left, center, or liberal parties (Hibbs Reference Hibbs1977; Schmidt Reference Schmidt1996). Therefore, coalition governments do not automatically perform as transmission belts between private desires and public policy; a more active interest representation and articulation is needed. Parties are however capable of carrying out such an active articulation and representation (Tavits Reference Tavits2004: 353). In other words, the impact of consensual institutions on the size of government is contingent upon partisan influences on public policy, and vice versa. Concerning education policy, for example, leftist parties generally tend to prefer social equality and promote a stronger public sector in education (e.g., Boix Reference Boix1997; Schmidt Reference Schmidt2002; Tufte Reference Tufte1978). Social democratic or socialist parties are therefore expected to attach greater importance to the equality of educational opportunities – this should manifest itself in higher education expenditures during left-party led governments. If one expects that consensual governments execute major state expenses because ‘such systems provide better opportunities for the lower social strata to be represented in government, then this effect should be reinforced in the event of large leftist-party share of cabinet seats, provided that people from the lower social strata or interests advocating more government involvement in economy and society comprise, for a large part, the constituency of the leftist parties’ (Tavits Reference Tavits2004: 353). Furthermore, for right and center leaning parties, the field of internal security is a matter of particular importance (Norris Reference Norris and Castles2007). These parties are therefore expected to enlarge state activity in this policy field.
Against this background, we hypothesize that the claimed expansionary effect of grand coalitions varies with the degree of participation by leftist or rightist parties in the decision making process. Concerning security policy, we expect that growing strength of rightist parties results in even higher spending in the field of internal security policy; whereas strong left parties should lead to increases in education spending. In terms of the German Laender, Wolf (Reference Wolf2006: 123ff.) argues in favor of a more explicit partisan effect on government expenditure in the sub-national units; the maneuverability is greater in the Laender because there are fewer veto players than at the national level. The evidence, however, is ambiguous; some authors find affirmation of the partisan hypothesis in certain policies: Schmidt (Reference Schmidt1980) reports higher expenditures in education and internal security policy during social democratic governments; some report higher spending by left-oriented executives in labor policy (Schmid/Blancke Reference Schmid, Blancke, Schneider and Wehling2006; Schmid/Hedrich Reference Schmid, Hedrich, Wolf and Hildebrandt2008). Other authors do not however find support for the partisan influence on state expenditures in general (Seitz Reference Seitz2000) or on education expenditures in particular (Wolf Reference Wolf2005; Zohlnhöfer Reference Zohlnhöfer, Schmidt, Ostheim, Siegel and Zohlnhöfer2007). Galli and Rossi (Reference Galli and Rossi2002: 298) find that rightist governments spend more, as measured by total expenditures. To some extent, their results support the hypothesis that leftist governments ‘are inclined to spend more on education and health; rightist governments on administration.’Footnote 3 These findings lead us to the next working hypotheses:
Hypothesis 2a: Grand coalitions tend to spend more on internal security policies when rightist parties are stronger.
Hypothesis 2b: Grand coalitions tend to spend more on education policies when leftist parties are stronger.
3 Research Design, Method, and Data
For the remainder of this paper, our primary focus will be on testing the hypotheses presented above. The influence of grand coalitions on government spending will be analyzed for the period from 1992 to 2005. We chose 1992 as our starting point because this was the first year with complete data available for all Laender. There was a certain time period at the beginning of the 1990s where not all indicators were homogenized for all German Laender. Moreover, more recent data have not yet been made available by the German Federal Statistical Office. For several reasons, the limits inherent to this time period do not seem to be problematic: First, with a period covering 14 years, three to five governments per state are included in the analysis. Second, the ‘old’ Federal Republic of Germany must be differentiated from the ‘new’ Federal Republic of Germany. Reunification has brought about countless changes: Not only has the number of states grown from 11 to 16, but the relationships between the states and to the federal government have also been altered to some extent.Footnote 4 Third, Laender-level grand coalitions have been on the rise since 1990. Between 1972 and 1991, for example, no grand coalitions between the CDU and the SPD occurred in any state. Finally, the analysis of a time period that is close to the present one allows one to draw conclusions about the current state politics.
Our investigation is based on an analysis of the influence of grand coalitions on the size of government in the 16 German Laender. To test the assumed relationships, multiple cross-sectional OLS regressions are conducted by incorporating several control variables into the models. The rationale for including these factors is as follows: ‘[E]ven where institutional arrangements do contribute to overall differences in specific capabilities, moreover, these effects usually are strongly mediated by other institutional and non-institutional factors’ (Weaver/ Rockman Reference Weaver, Rockman, Weaver and Rockman1993: 446). For each variable, the mean for the years 1992–2005 was calculated.Footnote 5 In other words, the multivariate cross-sectional analyses are obtained by averaging the values of the individual variables.Footnote 6
Our dependent variable is the size of government and is measured by two indicators: total public expenditure of the Laender for internal security and for education (and cultural affairs). Both are measured in euros per capita and adjusted for price. The main independent variable is the mean proportion of time grand coalitions governed in the Laender executives during the years 1992 to 2005. Based on established theories of state expenditure in cross-national comparative policy research, a number of other control variables are included in the model.Footnote 7 Subsequently, the resulting control hypotheses derived from the relevant literature on policy analysis and political economy used to assess the impact of grand coalitions on government spending are discussed. One major and well-established hypothesis in the comparative policy research is the partisan approach, which states that left-wing parties generally tend to spend more than right-wing parties (i.e., Blais et al. Reference Blais, Donald and Dion1993; Boix Reference Boix1997; Bräuninger 2002, Reference Bräuninger2005; Cusack Reference Cusack1997; Cusack/Fuchs Reference Cusack, Fuchs, Obinger, Wagschal and Kittel2003; Hibbs Reference Hibbs1977, Reference Hibbs1992; Hicks/Swank Reference Hicks and Swank1992; Jochimsen/Nuscheler Reference Jochimsen and Nuscheler2007; Mukherjee Reference Mukherjee2003; Schmidt Reference Schmidt1996; Zohlnhöfer Reference Zohlnhöfer2001).Footnote 8 The strength of both right and left parties is measured as the respective proportion of seats these parties hold in the Laender parliaments.
An important variable comes from the socio-economic determination hypothesis (e.g., Schmidt Reference Schmidt1996): It is assumed that economic prosperity leads to new challenges for the state as well as a new set of societal problems to tackle, thereby leading to higher government spending. Therefore, the gross domestic product per capita (adjusted for price) is included in the analysis. Finally, a dummy variable is included in the analysis to control for the considerable differences that continue to persist between the old and the new Laender (e.g., Schmid/Blancke Reference Schmid, Blancke, Schneider and Wehling2006; Schmid/Hedrich Reference Schmid, Hedrich, Wolf and Hildebrandt2008; Wolf Reference Wolf2006). A brief overview of all variables, their measurements, and the expected relationships can be found in the appendix.
4 Empirical Findings
This section analyzes the impact of grand coalitions on the size of government in the German Laender between 1992 and 2005. Multiple regression models were estimated to evaluate the influence of the different factors discussed. For each of the four models, the means of the different variables were used for the regression equation.Footnote 9 Table 1 reports the empirical findings.
Table 1: Determinants of Public Sector Size in German Laender, means 1992–2005
Notes: OLS regressions, standardized coefficients. Values of the t-statistic in parentheses. * = significant at the 10 per cent level (two-tailed test), ** = significant at the 5 per cent level (two-tailed test), *** = significant at the 1 per cent level (two-tailed test); see text and appendix for the operationalizations and data sources. All key financial data (domestic security expenditure per capita, education expenditure per capita, and gross domestic product per capita) are price-adjusted values (source: Federal Statistical Office).
With regard to models 1 and 2, which display the results of the determinants of public spending on internal security, up to 50 per cent of the variance in the dependent variable expenditure can be explained. Additionally, the hypothesized positive effect of grand coalitions on domestic security expenses is confirmed: The more frequently grand coalitions as special forms of consensual governments represent the Laender executives during the period from 1992 to 2005, the higher the public expenditures are for internal security. However, the effect of grand coalitions on the size of government is not moderated by partisan politics. The empirical analyses do not reveal any partisan influence of rightist parties on this policy. All other factors appear to be of minor importance: Neither economic prosperity nor the difference between the old and the new Laender are systematically related to the public expenditure for internal security. The results do however warrant further testing. As we are dealing with a very small number of cases, individual units (in this case, Laender) can quickly exert a large influence on the estimation of the parameters. Given these circumstances, we have performed a variety of regression diagnostics in order to test for robustness of the above findings.
The estimated regression models do not suffer from heteroscedasticity and therefore the standard errors did not need to be corrected. Furthermore, tests for multicollinearity revealed no critical results. A number of additional regression methods have been applied to assess the robustness of the results. An analysis of outliers and influential cases was also performed. The value of the Breusch-Pagan test for heteroscedasticity is not significant and well below the critical value; all values of the variance inflation factor are well below the critical value of 10 (Schnell Reference Schnell1994). Lastly, robust regression methods (such as median regression, robust regression, and regression with robust standard errors) were also conducted.
The identification of influential cases (or outliers) was carried out graphically via the inspection of the leverage vs. squared-residuals plot, as well as formally by inspecting the values of Cook's Distance (values higher than 1 (4/n) constitute influential cases; Kohler/Kreuter Reference Kohler and Kreuter2001). These tests lead us to question the robustness of the estimations. For example, model 2, which examines the impact of grand coalitions and rightist parties on governments' expenditure on internal security, is very sensitive to the inclusion of one specific Land. Grand coalition thus loses its significance when Berlin is excluded from the analysis. Additionally, the coefficient of the interaction effect changes its sign and shows a negative influence. It appears that the relationship between the independent and the dependent variable is produced solely by the influence of the Land Berlin. Compared to the other Laender, Berlin constitutes a special and exceptional case for several reasons. Not only is Berlin the federal capital of Germany and is faced with major liabilities in terms of protecting foreign embassies and most of the federal state authorities, it is also a melting pot of cultures, as it is the only federal state that was itself reunited with unification. While we can control for East or West German history in the other Laender, we cannot do so for Berlin. It therefore should not come as much of a surprise that Berlin exerts such a strong influence on the analysis of the determinants of Laender security spending. Therefore, further interpretation of the model and an investigation of the (insignificant) marginal effect of grand coalitions will not be carried out.
When looking at models 3 and 4, the results of the multivariate analyses can be summarized as follows: First, and most notably, grand coalitions have a positive influence on education expenditure: Where grand coalitions as special forms of consensual governments were more frequent in the Laender executives between 1992 and 2005, education expenditures were also higher. Second, with regard to the partisan variable, it should be noted that the strength of left parties is significantly related to the size of government in education policy. The effect however contradicts the findings of international comparative studies and supports the results of Wolf (Reference Wolf2006) in his analyses of educational policy in Germany: The stronger left-oriented parties are, the less a government spends for education.Footnote 10 How is this finding to be explained? In general, social democrats in Germany attach great importance to the field of vocational training, which is mainly financed by private companies (Wolf Reference Wolf2006).Footnote 11 This should benefit their electorate more than investments in the very costly sector of higher education. Furthermore, Alber (Reference Alber and Flora1986) argues that, in general, the German labor movement has historically regarded education as subordinate topic. This disregard is explained by the fact that the (Prussian) state attended to the matter of education relatively early (Alber Reference Alber and Flora1986; Wolf Reference Wolf2006). These circumstances may be responsible for the unexpected negative influence of stronger left parties on education expenditure. Third, the coefficient of the interaction effect has a negative sign and is significant at the five per cent level (Table 1). There is thus a multiplicative effect between grand coalitions and left parties on education expenditure. When left parties within grand coalitions are weaker, higher government spending for education and cultural affairs can be observed. If left parties possess medium or low strength (i.e., less than about 59 per cent), a significant marginal effect of grand coalitions on education expenditure arises. Of the 16 states, roughly 80 per cent are marked by a significant variation in education expenditure that is contingent upon the strength of left parties. Fourth, the coefficients of the control variables are both significant: Economic prosperity leads to increased expenditure for education policy and the new Laender also spend more in this field. Finally, it can be reported that model 4 is capable of explaining roughly 64 per cent of the variance in government spending on education.
The estimated regression models 3 and 4 suffer neither from the problem of heteroscedasticity nor multicollinearity. The value of the Breusch-Pagan test for heteroscedasticity is not significant and below the critical value. The values of the variance inflation factor are well below the critical value of 10 (Schnell Reference Schnell1994: 247). Additional regression methods have been applied to assess the robustness of the results.Footnote 12 Further analysis yielded outliers and influential cases: The exclusion of the three most problematic cases (Baden-Württemberg and Berlin, identified via a leverage vs. squared-residuals plot, and Hamburg, additionally identified via the inspection of the values of Cook's Distance) does not change the sign of the estimated coefficients and the coefficients of the three main variables of interest remain significant. It should also be noted that the reported findings are not sensitive to the random inclusion of other controlling factors such as the proportion of Catholics as a socio-cultural factor (Castles Reference Castles1989, Reference Castles1994; Hega Reference Hega1999; Schmidt Reference Schmidt2002, Reference Schmidt2004), voter turnout (Cusack/Fuchs Reference Cusack, Fuchs, Obinger, Wagschal and Kittel2003; Tavits Reference Tavits2004), level of debt (Wolf Reference Wolf2006), the degree of dependence on international trade (Cameron Reference Cameron1978; Cusack/Fuchs Reference Cusack, Fuchs, Obinger, Wagschal and Kittel2003; Garrett/Mitchell Reference Garrett and Mitchell2001; Scharpf Reference Scharpf1999; Tavits Reference Tavits2004), and – specific to the German Laender – the balance of the Fiscal Equalisation System (Wolf Reference Wolf2006).Footnote 13
5 Conclusion
This article finds that grand coalitions differ in their influence on public expenditure according to policy domains and partisan politics. However, in contrast to our theoretical expectations, stronger left parties in the Laender lead to lower expenses for education policy and cultural affairs (see also, Wolf Reference Wolf2006). In contrast to their international counterparts, the German left parties do not invest significantly more in education than the Christian Democrats or the Liberals. Social democratic parties and labour unions are advocates of in-plant training (Thelen Reference Thelen2004: 41) as part of Germany's dual system of vocational training. Their constituencies include youths in low social strata who benefit particularly from this popular alternative to university studies. As regards vocational training, a large part of spending is done by private companies. This is supported by the finding of Stern (Reference Stern2000) that expenses for schools (per pupil) are higher in SPD-governed Laender but expenses for universities (per student) are considerably higher in CDU-governed Laender (Hetmeier/Weiß 2001: 39; Klemm Reference Klemm, Cortina, Baumert, Leschinsky, Mayer and Trommer2003).
What are the implications of our empirical findings for other federal systems, where education is typically devolved? In the United States, the phenomenon of divided government may be considered as an alternative form of grand coalition,. Empirical research on divided government (Alesina/Rosenthal Reference Alesina and Rosenthal1995; Alt/Lowry Reference Alt and Lowry1994; Fiorina Reference Fiorina1996) has found evidence that in divided governments with competing parties whose core constituents have different preferences which can be addressed through public spending there tends to be more spending on public policy in general and on education in particular than in unified governments. The distinctive objectives of the different parties who control the executive and legislative branches of government, coupled with the uncertainty about the outcome of the next election, seem to foster public growth. Our findings are also in line with recent research on Switzerland, another federal country with a very decentralised education system. Freitag/Bühlmann's (Reference Freitag and Bühlmann2003) study of public education expenditure in Swiss cantons shows that educational expenditure is systematically related to the cabinet type and the number of parties in cantonal government: the more encompassing the grand coalition government in a Swiss canton, the higher the spending for public education.
Some authors argue that minority governments resemble grand coalitions very closely (Lijphart Reference Lijphart1999; Strøm Reference Strøm1984, Reference Strøm1990). Usually, minority coalitions need to find additional legislative support through bargaining as the case arises. Therefore, those minority governments which are very common for example in Scandinavia should perform similarly to the grand coalitions analyzed in this paper.
Can linkages between the sub-national and the national levels be made based on the present findings? Downs (Reference Downs1998: 244), for example, suggests that national elites have the opportunity to learn from the experimentation at sub-national levels. According to our findings, it remains unclear whether the current grand federal coalition should be more conducive to the growth of the public sector. There are two obstacles to the expansion of the public sector at the federal level. First, with regard to education policy, the strength of the SPD within the coalition should not encourage government expenditure. Second, compared to the Laender, the room for maneuver at the federal level is constrained because there are more veto players than at the sub-national level (Wolf Reference Wolf2006: 123ff.), such as an independent central bank, bicameralism, a powerful constitutional court and the diffusion of federal political power imposing mutual vetoes. Thus, as Schmidt (Reference Schmidt1996: 174) has argued the Federal Republic of Germany can be regarded as a grand coalition governed by a formal or informal grand coalition of the major established parties and a formal or hidden grand coalition of federal government and state governments.
Acknowledgement
This article was written as part of a research project on patterns of democracy in the German Laender which was funded by the German Research Foundation. We would like to thank Birgit Jacob, Jennifer Shore, the anonymous reviewers and the editor for their valuable comments on previous drafts.
Appendix: Variables, Hypotheses and Measurement
