From coalition theory to coalition puzzles

Political competition is often described and indeed typically structured as a choice between governments (Laver 1998: 1). In political systems where the executive is responsible to the legislature, such as parliamentary democracies, the process of government formation constitutes a crucial phase and arena in democratic governance: it concerns the translation of electoral and parliamentary power into executive power, and the possibility to implement policies that have been democratically endorsed and legitimated by the electorate. The question of ‘which parties get in’ government is especially of interest where election outcomes are not decisive, that is when no political party controls a majority of seats in parliament. In these ‘minority situations’, either a single party fills all cabinet positions and aims at retaining executive power by seeking majority support (sometimes on an ad hoc basis) in parliament, or a coalition of parties forms, generally in order for the cabinet to rely on a stable majority in the legislature. In most countries elections usually do not automatically decide which party will be in power by awarding it a majority of seats in the relevant parliamentary assembly. Therefore, ‘understanding how a given election result leads to a given government is, when all is said and done, simply one of the most important substantive projects in political science’ (Laver and Schofield 1990: 89), and this is the task of what is usually referred to as ‘coalition formation’ studies. Of course, the outcome that emerges from the coalition formation process has various aspects, such as the allocation of ministerial portfolios or the contents of the agreement on the new government’s policies, but the latter are themselves the product of negotiations among partners who agreed to try to form a coalition, making the partisan composition of the coalition the crucial aspect of the outcome. Given this substantial importance, it is no surprise that coalition theory applied to government formation is one of the most flourishing areas of research in political science. After about five decades of research by game theorists, country experts and comparative politics scholars on why some coalition governments form while others do not, this field has indeed become highly mature. We can see this maturation in terms of the diversity of theoretical approaches and models that are competing, their degree of formalization, the sophistication of statistical methods applied, and the richness of data sets used for testing hypotheses derived from theory. Unfortunately, the research field still suffers from major shortcomings. First, progress made on the theoretical side has hardly resulted in significant systematic improvements in the explanation and prediction of real-world governments. Note that recent work has used appropriate statistical devices to model the selection of a particular government over all possible ‘coalitions’ (given a distribution of seats in parliament) and performed multivariate analyses pitting competing hypotheses against each other. But even the most comprehensive statistical model combining two dozen variables drawn from various theoretical schools, and therefore lacking parsimony and internal consistency, only explains less than half of real-world government formation outcomes in post-1945 Western Europe (Martin and Stevenson 2001: 47). The results of these empirical studies suggest that scholars should further investigate possible theoretical innovations to boost coalition theories’ predictive performance and seek to enhance the internal consistency of the field’s theoretical body. Second, traditional coalition theories, based on cooperative games, 1 make predictions about the government composition outcome without making any reference to the process of government formation. As a result, most empirical studies of coalition formation have only evaluated the predictive capacity of theories, or the causal effects of potential independent variables in large-n studies. What is yet largely lacking to come to full causal explanations is the identification of the causal mechanisms posited by these theories, that is the processes and intervening variables through which an independent variable exerts an effect on a dependent variable (George and Bennett 2005; Mahoney 2000). Much in line with what Milton Friedman (1953) advocated, the quality of rational choice theories has also been mainly judged upon the accuracy of its predicted outcomes without evaluating whether their core assumptions are accurate representations of reality. 2 On the other hand, more recent theories based on non-cooperative games concentrate on the bargaining process, and are therefore more promising in terms of illuminating potential causal mechanisms, by taking into account the constraining role of institutions and specifying a sequence of events leading to the outcome. But the standard predicted implication of these neo-institutionalist theories - that the party designated to make the first move (the ‘formateur party’) should be advantaged in the outcomes of government formation 3 - has been rarely put to empirical tests (see however Warwick 1996; Martin and Stevenson 2001; Bäck and Dumont 2008). This lack of empirical evaluation is partly explained by the fact that scholars refrain from forcing theories into making predictions beyond their scope, knowing that the main models drawn from noncooperative game theory have been, for reasons of analytical tractability, typically developed for simple three-party legislature settings that are seldom found in real-world parliaments. 4 Another reason for the lack of empirical tests of these non-cooperative games is that the complex chains of causation envisioned by strategic interaction models are quite difficult to reduce to a set of independent variables. Hence, the usual quantitative template for empirical study, regression analysis, is not seen as an adequate method to evaluate such theories (Bates et al. 1998; Hall 2003).