«Der Open-Access-Publikationsserver der ZBW – Leibniz-Informationszentrum Wirtschaft The Open Access Publication Server of the ZBW – Leibniz ...»
The structural characteristics shown in table 4 are arranged in terms of the dimensions of a polycratic (liberal) democracy as depicted in table 2. This classification of characteristics makes it easier to determine what the specific indices measure. One distinction shown in table 2 has not yet been discussed, namely that between the primary and secondary characteristics of the governmental system. This distinction directly affects the question of the importance and weighting of institutional characteristics. With reference to Tsebelis (1995), institutional characteristics can be classified as peripheral if they are relevant only for decision-making on certain subjects (e.g., independent central bank in financial policy and economic decisions), or if – under certain circumstances – they can merely amend decisions that have been already taken (e.g., constitutional courts). Primary structural characteristics are those that directly, permanently, and comprehensively structure or limit policy decision making processes.
The index that differs most from all the others with respect to the selection of structural characteristics is Lijphart’s ‘executive-parties dimension’ (see table 4).1 It includes only empirical structural characteristics that relate to actor constellations. The other indices are based largely on formal (constitutional) structural characteristics, and mainly on those of the governmental system. Only Huber et al.’s index also includes electoral law, and that of 1 The attribute “interest group pluralism“ listed by Lijphart has not been included in Table 4. According to our conceptualization it is no structural attribut of democratic regimes.
Colomer takes an empirical structural characteristic into account as well (effective number of parliamentary parties).
1 ‘Federal-unitary dimension’ 2 ‘Executive-parties dimension’ In the context of our theoretical distinctions, this extension is problematic. If the governmental and the electoral systems are the two fundamental dimensions of a democratic regime (see table 2), then they must also be equal in weight and systematically linked.
Merely additively appending a structural characteristic of the electoral system to several structural characteristics of the governmental system is thus insufficient. A similar argument can be advanced for the combination of formal and empirical structural characteristics. On the other hand, taking account of only one additional structural characteristic that does not belong to the governmental system in the case of Huber et al. and Colomer is of little quantitative consequence. Both construct additive indices of five (Huber et al.) and four (Colomer) structural characteristics.2 With the exception of Lijphart’s ‘executiveWith the exception of Lijphart’s (1999) two indices, all other indices are additive. Lijphart’s indices are averages of standardized variables.
parties dimension,’ the indices can thus essentially be considered as measurements of the governmental system.
The five indices of the governmental system – Lijphart (‘federal-unitary dimension’), Huber et al., Tsebelis, Colomer, and Schmidt – differ in two aspects. First, whether or not they take presidentialism as a primary structural characteristic, and, second, whether they take secondary structural characteristics into account alongside primary ones. The indices of Lijphart and Schmidt take no account of presidentialism. This is somewhat surprising, because presidentialism is generally regarded as a prototype of a system of separation of powers (see especially Weaver and Rockman 1993, but also Tsebelis 1995). Lijphart (1984) offers an empirical argument for excluding this characteristic. In his factor analysis, presidentialism weighs only very lightly on the ‘federal-unitary dimension.’ In our opinion, this empirical finding is, however, not a convincing reason to exclude a structural characteristic. If we take distribution of power as a central point of reference for the development of a typology – as Lijphart does – presidentialism is without any doubt an indicator of the distribution of power. Schmidt does not explicitly justify exclusion of presidentialism.
However, justification for so doing can be formulated. The need to include presidentialism in a veto-player index is reduced in proportion to how far it is assumed that the effects of presidentialism can be attributed to characteristics other than the mere fact that the president is an institutional veto player. Other characteristics that come in question are, for example, the personalisation of politics in presidentialism and the strong dependence of the president on public opinion. This in turn would imply that the veto-player approach alone does not suffice to identify the different forms of democratic regime.
Still more fraught with consequences than the question of including presidentialism in the typology is the second aspect of additionally including secondary structural characteristics of the governmental system. This is done in both Lijphart’s ‘federal-unitary dimension’ and in the index developed by Schmidt. In both cases the number of secondary structural characteristics even exceeds that of primary characteristics (see table 4). This is especially important for Schmidt’s additive index, which assumes individual structural characteristics to be equal in weight. This extension of institutional veto players has its advantages and disadvantages. The disadvantages lie in their labelling as secondary structural characteristics. These characteristics have to do with institutions of the governmental system that play a role in political decision-making processes either on specific topics only or after the fact. This state of affairs can either invite excluding the secondary characteristics from index construction, as Tsebelis (1995) does, or placing less weight on secondary than on primary characteristics. But there is no theoretical criterion available for such weighting, so that this solution cannot be put into practice. A third alternative is to include secondary structural characteristics in index construction only with respect to specific topics. If, for example, economic performance is to be explained, the characteristic ‘independent central bank’ would be included. But this would mean that different indices would need to be constructed depending on what policies are to be explained. However, if we want to judge the quality of different institutional arrangements, it does not make much sense constantly to vary the indices for recording these institutional arrangements.
Although secondary structural characteristics can become players only on specific topics, temporarily, and at a subsequent stage, it is quite possible for a plurality of such veto players together to constitute a democratic regime that considerable restricts decisionmaking processes as a whole. This possibility cannot be covered by a type and index construction confined to primary structural characteristics. In constructing an index, it must therefore be decided what weight is to be given to this possibility and whether the problem of implicitly overweighting secondary structural characteristics in the construction of the additive index is to be accepted.
The theoretically less equivocal solution is certainly to construct types and indices only on the basis of central governmental-system structural characteristics. Tsebelis appears to have adopted this procedure most fully (see table 4). However, this is only the case for his selection of the institutional veto players to be considered in principle. Whether they actually do count as veto players is made to depend on another criterion. This can most easily be demonstrated in the case of bicameralism. Only if the majority relations between the parties represented in the upper house differ from those in the lower house does the upper house count as an additional veto player (Tsebelis 1995). If the same parties have a majority in both houses, the two houses count as only one veto player. For Tsebelis, the status of institutional veto players thus also depends on contingent actor constellations. This means that the index for a given country can change over time even though the constitutional structure does not change. Tsebelis’ procedure can be useful if one is primarily interested in explaining specific policy outcomes, as Tsebelis explicitly states (cf. introduction).
However, it is not an appropriate procedure if the prime concern is to evaluate constitutional structure. In the following section we therefore construct an index that takes up Tsebelis’ selection of institutional veto players but leaves their dependence on actor constellations out of account.
6. Dimensions of the Indices
We have assigned the indices we have analysed to the presidentialism-parliamentarism and veto-player typological approaches. This has been done on the basis of theoretical considerations and of self assessment by the authors who developed the indices in question. In this section we examine the extent to which this a priori assignment is empirically justified.
For this purpose we performed an exploratory factor analysis. In accordance with our allocation we expect two factors. The two presidentialism-parliamentarism indices (Shugart/Carey, Sartori) ought to constitute the one, and the veto-player indices (Lijphart’s ‘federal-unitary dimension,’ Huber et al., Colomer, and Schmidt) the other.
Besides the indices already described, two others were included in the factor analysis: a ‘minimal governmental system index A’ and a ‘minimal governmental system index B.’ Both are additive indices. They are termed ‘minimal’ because they are based exclusively on formal (constitutional) and primary structural characteristics. These are bicameralism and federalism in index A, and in index B presidentialism is added to these two characteristics. The ‘minimal governmental index B’ is thus a new construction, which is based on Tsebelis’ selection of institutional characteristics (see table 4). Of the eight indices included in the factor analysis, these two – together with Sartori’s presidentialismparliamentarism index – are by far the most parsimonious.
1 Principal component analysis: varimax rotation; explained variance = 92%; N = 23 2 Formal and primary structural characteristics: bicameralism and federalism 3 Formal and primary structural characteristics: bicameralism and presidentialism.
Source: Lijphart (Lijphart 1999); Huber et al, Colomer, Schmidt (Schmidt 1996); Shugart/Carey (1992), Sartori (1994a) The results of the factor analysis are not very surprising (see table 5). Precisely the two expected factors or components are extracted: a component of the veto-player indices and a component of the presidentialism indices. However, there are some remarkable detailed results. The factor loading of all veto-player indices on the first component is extraordinarily high. The indices that exhibit low loading are those that include presidentialism as a structural characteristic and therefore exhibit incidental loading on the second component, which refers to presidentialism. For the presidentialism components, the factor loading of the more complex indices of Shugart/Carey and the more simple index of Sartori are at almost the same level. These results permit conclusions about the two problems of index construction we have already mentioned. The distinction between simple and complex indices and between indices that are based only on primary structural characteristics of the governmental system or that also take secondary structural characteristics into account play no role in factor structure. All veto-player indices record the latent construct equally well, and the same is true for the two presidentialism indices. This implies that the more parsimonious indices, too, adequately measure the given latent construct. The practical consequences for research will be discussed in the concluding remarks.