Bringing Cohesion In

Transcription

1 Bringing Cohesion In Christoph Oberst Department of Political Science and Communication Studies University of Greifswald Baderstr. 6/ Greifswald Germany Tel Paper prepared for presentation at the 5th ECPR Graduate Student Conference, University of Innsbruck, Austria, 3-5 July Abstract: Intra-party cohesion is a crucial analytical tool of veto player theory, which has long been neglected in empirical analyses explaining policy change. This paper seeks to contribute to a better understanding of the importance of party cohesion in political decision-making processes by asking does ideological party cohesion matter? In order to answer this question I will apply a newly invented index of ideological party cohesion (Jahn/Oberst 2012) for an analysis of partisan effects on labour market reforms. I will replicate the empirical study by Becher (2009) that examines the impact of veto players on labour ministers for labour market reforms, and add cohesion to the models by interaction with the ideological positions of the agenda setter. Thus, I am able to analyse the effects of a party s cohesion on a party s position. The interpretation of marginal effects of cohesion over position clearly indicates that cohesion has a relevant effect. Wordcount: 8771 without references 1

2 Introduction According to Tsebelis (Tsebelis 2002) the configurations of veto players of a political system affect the set of outcomes that can replace the status quo (the winset). These configurations depend on the number of veto players, their ideological preferences, the ideological distance of their preferences and their cohesion. Concerning cohesion Tsebelis concludes that the lower the party cohesion, the lower is policy stability (Tsebelis 2002, 84). In other words, it is expected that high tensions within a party are more likely not leading to policy change, or in a nutshell that policy outcomes are affected by tensions within a party. 1 A possible explanation of this function gives Hanna Bäck by stating that parties suffering from low intra-party cohesion are likely to lose bargaining power [ ] and as a result, heterogeneous parties should be less likely to shape policy outcomes (Bäck 2012, 73). Intra-party cohesion is a crucial analytical tool of veto player theory, but it has long been neglected in empirical analyses explaining policy change. This paper seeks to contribute to a better understanding of the importance of party cohesion in political decision-making processes and a more adequate empirical application of cohesion by asking does ideological party cohesion matter? I will answer this question by applying a newly invented index of ideological party cohesion (IPC) (Jahn and Oberst 2012) for an analysis of political institutions and partisan effects on labour market reforms. The task is to combine existing indices of partisan and veto player theory with the new measure of IPC. I will fulfil this task in a three-step manner. First I replicate the empirical study by Becher (2009) that examines the impact of veto players on labour ministers for labour market reforms. Secondly I extend the replication dataset to extend the scope of time, and therefore cases, in the paper. I will do this mainly by taking the chance to update the data of the papers dependent variable with data from the Comparative Welfare Entitlements Dataset (CWED2) project (Scruggs, Jahn, and Kuitto 2013). Additionally, I will change the ideological dimension that Becher uses ( myrl3 ), which is a parsimonious indicator with market economic, effective governance and welfare issues based on the data of the Comparative Manifestos Project (CMP), into a deductive Leftright index with country and time-specific aspects. As a third step, I add cohesion to the models by interaction with the 1 Actually Tsebelis differentiates that in decisions by majority rule policy stability increases with cohesion; in decisions by qualified majority policy stability decreases with cohesion (Tsebelis 2002, 62). I will come back to this later. 2

3 ideological positions of the agenda setting minister to be able to analyse the effects of a party s cohesion on a party s position. The rationale behind the extended replication is to test whether not only position and the range of the respective veto players of a labour minister (and its interaction) affects the outcomes of labour market reforms, but also the cohesion of the ministers party. The general result of this study is that cohesion matters. The direction and function though is not so clear and needs further elaboration. Theory There is a huge amount of literature in quantitative research that proves that the partisan hypothesis still has a considerable effect within policy making (Allan and Scruggs 2004; Avdagic 2013; Jensen 2012; Knill, Debus, and Heichel 2010; Neumayer 2003; Schmidt 1996; Schmidt 2002, just to name a view). Studies of late and since additionally prove the relevant impact of veto players either directly or via interaction with the partisan hypothesis. The study of Michael Becher (2009) is one of those studies. The aim of the Becher study is to reexamine how veto players matter for labour market reforms in [20] advanced industrial democracies (Becher 2009, 53) from with a particular focus on ministerial power. Labour market reforms, as the study s dependent variable, are measured with the CWED data on unemployment insurance entitlements (NRR; (Scruggs 2004)) by Lyle Scruggs and employment protection legislation with data from Allard (EPL; (Allard 2005)). Becher s findings are that the impact of ministerial partisanship is conditional on the ideological distance between veto players and that the capability of a minister to change the status quo towards the own preference declines, as the ideological distance between veto players increases (Becher 2009, 53). He shows that rightward 2 shifts of the minister lead to substantial cutbacks only if ideological distance is relatively small (Becher 2009, 49). He tests for different models of agenda setting power within the cabinet and concludes, that the mixed finding concerning the dependent variable is consistent with the idea that the autonomy of cabinet ministers within their jurisdiction is greater for policies that are not money intensive (Becher 2009, 54). 2 His policy dimension is called myrl3 which originally stems from the Parties Governments Legislatures (PGL) File Collection of Cusack and Engelhardt (2002). The dimension grasps f. ex. aspects of Free Entrepreneurship, Economic Orthodoxy, Governmental and Administrative Efficiency, Welfare Expansion, Market Regulation or Economic Planning. For a recent debate on the saliency and comparability of the Leftright dimension see Benoit and Däubler 2014; Budge and Meyer 2013; Dinas and Gemenis 2010; Franzmann 2013; Franzmann and Kaiser 2006; Jahn 2011; Jahn

4 The task of the following is to bring in the long neglected ideological cohesion of parties into partisan and veto player models by a model that respects the combination of these three parts with respective indices. In the empirical part of this paper I will do this by replicating and extending the above described study of Michael Becher. But first we have to define the matter of the research focus, ideological party cohesion, and then explain the way of how cohesion can have an impact on policies. What is ideological party cohesion and why should it matter? Definitions to describe the internal life of parties (Katz 2002) are manifold and often subsumed under the term factionalism (Boucek 2009). In this direction one of the oldest definitions has been provided by Raphael Zariski (1960, 33), where he defines a faction as any intra-party combination, clique or grouping whose members share a sense of common identity and common purpose and are organized to act collectively - as a distinct bloc within the party - to achieve their goals, which can be patronage [ ], the fulfilment of local, regional, or group interests, influence on party strategy, influence on party and governmental policy, and the promotion of a discrete set of values to which members of the faction subscribe (Zariski 1960, 33). Additional to this factional based definition Sartori (1976) took ideology as a driving force into account and distinguishes between factions that are ideologically oriented and factions that are office-seeking. Ergun Özbudun (1970) took a different approach and spoke conceptually of behavioural versus non behavioural approaches culminating in disciplinal versus ideological driven cohesion of a party. This could also be described as a top-down versus bottom-up approach as Bowler, Farrell and Katz (1999) put it. For my study here I will rely on this latter approach and concentrate on the bottom-up ideological driven cohesion of a party. Ideological cohesion within this approach can be seen as resulting from the interaction of individual cost-benefit calculations of the various intra-party actors with party and government rules, and with the feature of the party system (Strøm and Müller 2009, 34-35). In detail the cost-benefit calculations of the intra-party actors are determined by policy and career related preferences, institutional party and state rules, as well as by party system conditions (Strøm and Müller 2009, 34-40). The more similar the results of the costbenefit calculations of the actors are, the higher the cohesion will be. This model can be extended with the assumptions of Barry Burden (2007, 43), as cited by Baumann et al. (Baumann, Debus, and Müller 2013, 187), who sees the personal 4

5 characteristics as the driving force for the preferences of each parliamentary delegate. Burden analyses, that there are internal factors and personal characteristics like values, experiences and self-interest, and external factors like the party line, constituency preferences and lobbying that all influence the parliamentary activity either directly or filtered through preferences. A visualization of the combination of these two models can be seen in the following figure 1. Figure 1 Determinants of Party Cohesion extended Model based on Müller (2000) and Burden (2007) I will take this approach and assume that these resulting levels of party cohesion are visible on the party level and regularly manifest into the election programmes to which each party agrees upon at the party congress. Therefore ideological party cohesion is measureable as a result of cost and benefit calculations mainly driven by preferences. After having defined ideological party cohesion, I will turn to the effects cohesion can have for policy making to focus on the subject cohesion matters. Now why should ideological party cohesion matter for policy making? The answer is threefold and sets the policy making party in the focus of the analysis: first cohesion is an analytical part of the configurations of veto players of a political system that affect the set of outcomes that can replace the status quo (the winset) (Tsebelis 2002), and second it increases the agenda setting power and third the bargaining power of the respective policy making party as well. 5

6 The Effects of Cohesion within Veto Player Theory Let s start with the effects cohesion can have from the perspective of veto player theory. A veto player, according to the seminal work of George Tsebelis (2002), is defined by its potency to veto against a (labour market) reform and could be either individual (a president) or collective (a second chamber) and institutional or partisan. Parties can be defined as collective actors and when taking part in a coalition they are a relevant veto player. Tsebelis gives us three assumptions, the first is that the more veto players are involved, the higher policy stability or the more unlikely of a (labour market) reform is. The second assumption is that the wider the ideological distance of the preferences of the veto players is, the higher policy stability will be too. The third assumption is that the cohesion of the collective veto player influences the outcome as well. He concludes that the lower the party cohesion, the lower is policy stability (Tsebelis 2002) and even states that [i]t is because parties are not cohesive that policymaking becomes possible (Tsebelis 2002). His analytical tools core and winset are widely accepted, but less research has been done with the analytical concept yolk, which can be a tool to determinate cohesion in a collective veto player. The yolk can be best described as a measure how exact the ideal position of a collective veto player can be concentrated (Baltz 2009). To create the yolk we first have to identify whether the decision game is a simple majority or a qualified majority game. Let us assume first, for sake of clarity here, a simple majority game. 3 To identify the yolk now, we first have to draw median lines between the players with a simple majority on each side, now the yolk itself is the smallest circle intersecting all medians (Tsebelis 2002, 45). The centre of that circle is called (Y) and its distance to the Status Quo (SQ) is called (d), while (Y) has a measurable radius (r). With this information we can draw a wincircle around (Y) with the radius of (d + 2r). This wincircle always contains the SQ, so that outside of this circle no compromise will be settled (Tsebelis 2002, 47). Tsebelis states, that we can use this wincircle around Y as the circle for a fictitious individual veto player that can replace the collective veto player (Tsebelis 2002, 47). Thus we are able to widen the indifferences curves of the decision makers by the radii of their yolks, which then leads to the assumption that the more 3 Let allow us a small preview of the operationalization of this paper that will later lead to the answer to the question whether a minister is influenced by the cohesion of his or her party, because we should assume accordingly a labour minister of party A that wants to reform the labour market, but has to bargain that with his coalitional veto partners in cabinet of party B and party C and also with his own party A. Will the decision then be made by simple or qualified majority, or are there other mechanisms that influence the legislative process which are determined by party cohesion, like agenda setting or bargaining power? We should keep this question here in mind. 6

7 concentrated the yolk is, the higher policy stability will be! The following figure gives a good impression for that function, only with the extended circles there ll be a chance to agree on a point within the plane named WABC. Figure 2 Difference between individual and collective decision makers (Tsebelis 1995, 300) The above leads Tsebelis to two conjectures. The first is that Policy stability increases as the m-cohesion of a collective veto player increases (as the radius of the yolk decreases) and the second that an increase in size of [ ] a collective veto player [ ] increases its m- cohesion (decreases the size of its yolk), and consequently increases policy stability (Tsebelis 2002, 48). In oversimplified words, we can summarize for simple majority decisions that the more cohesive a collective veto player is, the less likely a reform will be. As we kept in mind, the above holds true for simple majorities, turning to qualified majorities the result of the model summarized in conjectures is quite diverse. Tsebelis starts again with the three step approach, where majority lines between the actors have to be drawn first. The difference to the median lines, as for simple majorities, is that now only relevant dividers count, called q-dividers, which are relevant when they leave the SQ on the not mayoral side of the divider. Then quite similarly a q-yolk and q-wincircle are to be created. The conjecture then is quite the other direction. Tsebelis shows that in cases of qualified majority policy stability decreases as the q-cohesion of a collective veto player 7

8 increases (Tsebelis 2002, 54) and secondly, but less astonishing, he proves with his model, that the higher the threshold for the majority in a collective veto player is, policy stability increases (or remains at least the same) (Tsebelis 2002, 54). Again in oversimplified words, we can summarize that with qualified majority the more cohesive a collective veto player is, the more likely reforms will be made. It is a hard task to empirically test the above described effects cohesion can have within veto player theory, due to the fact, that Tsebelis comes to two adversary conjectures. We simply do not know whether decisions in ministerial or cabinet decision making, which involves for example coordination among coalition partners, are arrived at with simple or qualified majority. I will overcome this problem by taking advantage of the concepts of Agenda Setting Power and Bargaining Power. The Effects of Cohesion via Agenda Setting and Bargaining Power There are several studies stating that the concepts of Agenda Setting and Bargaining Power either have strong implications for or are implicated by the cohesion of parties (Bäck 2009; Bäck 2012; Blumenau 2012; Cox and McCubbins 2005; Debus and Bräuninger 2009; Giannetti and Laver 2009; Laver and Shepsle 1996; Meyer 2012; Mitchell 1999; Pedersen 2010). These studies will be introduced here shortly to link ideological party cohesion to policy making via Agenda Setting Power and Bargaining Power. I will start with Agenda Setting Power, the bedrock of party government (Cox and McCubbins 2005, 204). Agenda Setting Power Agenda Setting Power as described in the seminal work of Cox and McCubbins (2005) is the power to push or to slow down the legislation of bills. Whereas tools of the former are subsumed under the term positive agenda power, tools of the later are called negative agenda power. From Cox and McCubbins (2005) we do know that party cohesion, in terms of homogeneity of preferences, increases positive agenda power, through favouring the distribution of positive instead of negative power. The mechanism for this is, that becoming more homogenous in preferences, trust between the members of the group with proposal and veto powers will be increased (Cox and McCubbins 2005, 204). Ways to accomplish this, however, are changes in the rules, or [ ] changes in the fiduciary standard expected of officeholders, or both (Cox and McCubbins 2005, 208), which could be for example, a removal of committee chairmen by the party. 8

9 One step further, based on the chain of delegation by Strøm (2000; 2003; Strøm and Müller 2009), Thomas Meyer (2012, 485) examines in his study the consequences of intra-party dissent for coalition governments. He asks how intra-party dissent affect[s] a minister s agenda setting-power (Meyer 2012, 485). He builds simulation models on the assumption of the chain of delegation where: Voters delegate to the members of parliament who, in turn, delegate power to the cabinet. Within the cabinet, jurisdictions are delegated to individual ministers, and these ministers delegate many crucial tasks to their civil servants (Meyer 2012, 485). Meyer combines this with the assumption of Laver and Shepsle (1994; 1996) that ministers are agenda setters within their jurisdiction (Becher 2009, 37). He further individualizes the decision making process and his simulations suggest: that parties may benefit from deviant behavior within their own ranks: whereas the party in charge of a respective portfolio suffers from an increase in the number of veto players, the coalition partner can move the policy output closer to its ideal position than it would be the case in the unitary actor model. (Meyer 2012, 501) He concludes that dissent-shirking and (potential) sabotage lead to a decrease of the minister s agenda setting power (Meyer 2012, 501), which fosters the party that is not holding the ministerial post. The resulting functional direction of the effects ideological party cohesion has on Agenda Setting Power is therefore: the more homogenous parties are, the more positive agenda power they have. Bargaining Power Concerning the effect cohesion has on Bargaining Power the results of several studies can be summarized as: homogenous parties do have more Bargaining Power. Bargaining Power can be simply defined as having a better bargaining position over a second actor in a bargaining situation due to different aspects of which information is one of the most important. For example, if party B knows that party A is incoherent on a policy move, the bargaining position of the leader of party A is weakened, since he or she could be replaced by the dissent faction within party A. In her study Helene Pedersen examines the impact of intra-party politics on the coalition behaviour of political parties (Pedersen 2010). She tests the impact of the formal internal power distribution within Danish political parties on their ability to participate in winning legislative accommodations by setting up a model with external and internal bargaining 9

10 resources. Her results favour the hypothesis of Strøm (1990) that parties with decentralized decision procedures are less effective in coalition negotiations, which is contrary to Maor s (1998) proposition that decentralized parties are stronger in coalition negotiations (Pedersen 2010, 750), and suggest that parties with strong national party organs are more inclined to stick to policy ideals, making them less attractive and flexible in inter-party negotiations. (Pedersen 2010, 750). Drawing on the works of Laver and Schofield (1990), Luebbert (1986) and Mitchell (1999) Hanna Bäck (2008, 74; 2009) points into the same direction and suggests that intra-party tensions can have systematic effects on bargaining [ ] more specifically, tensions should negatively affect parties ability to enter government. Mitchell (1999, 281) states that it is unlikely that dissent backbenchers of a party seek to break down the own coalition after it is once set up, but they can rebel on particular policy items, punishing the party s ministers for drifting too far away from party policies and reminding their leaders that they cannot always be taken for granted. It is clear, that the more diverse the preferences of all actors within the party are, the higher are the chances for such a rebellion against drifting away from the actors preferences. Mitchell concludes that coalition maintenance primarily involves continuous bargaining in which delicate trade-off has to be struck between negotiations among and within parties (Mitchell 1999, 283). Concerning the effect of cohesion on policy outcomes itself, Bäck states that (2012, 73): Intra-party cohesion may also affect policy outcomes due to that parties suffering from low intra-party cohesion are likely to lose bargaining power. This in turn may lead to that their impact on legislative output and the implementation stage decreases, and as a result, heterogeneous parties should be less likely to shape policy outcomes. In a similar vein Debus and Bräuninger (2009) see effects of cohesion in the negotiations of portfolio allocations during coalition bargaining. They show that the appointment of individual ministers who belong to party factions is used to increase bargaining leverage inside the coalition government (Debus and Bräuninger 2009, 141). Here the conjecture can be made that a minister is then dependent or accountable to its faction. Cohesion thus certainly plays a role for her or him too. 10

11 A Model of Constraints of Ministerial Policy Preferences I will take the approaches described above and use them to link the assumptions that Tsebelis made for cohesion and the assumptions by the above cited scholars for agenda setting and bargaining power to model answers to the question whether policy outcomes are affected by tensions within a party or in simple words: whether cohesion matters. But instead of modelling intra-party dissent with forms of not joining (shirking) and sabotage, as for example Meyer did, I will take ideological party cohesion as the driving force and subsume the above cited mechanisms of Agenda Setting Power and Bargaining Power as ministerial bargaining power to set up the following hypothesis: H1: A minister backed by a homogenous party is more likely able to set the agenda in her or his preferred direction due to ministerial bargaining power (see also Bäck 2012, 37). Therefore cohesion constraints the room for manoeuvre available to the agenda setter (Bäck 2012; Becher 2009; Tsebelis 2002) and so I assume that policy output is not only constrained by the policy preference of the ministry, but also by the ideological party cohesion of the minister s party. A visualization of this, extended with a veto player function as an additional constraining factor, which is the veto player preference constellation of the ministry (f.ex. coalitional partners, presidents, second chambers), can be seen in the following figure. Figure 3 Constraints of Ministerial Policy Preferences In the above model the directional mechanism for policy outputs is the ministerial policy preference. This is in line with partisan theory and draws also on assumptions from Laver 11

12 and Shepsle s Ministerial Discretion Model (Laver and Shepsle 1990; 1994; 1996; 1999; 1999). In their model a coalition is a kind of agreement about the allocation of ministerial portfolios and coalitional behaviour is then constrained by the credibility of the ministers in cabinet, their parties preferences and their autonomous like policymaking. Tsebelis summarizes the model as that [M]inisterial discretion results from the minister s ability to shape the agenda of collective cabinet decisions rather than to determine cabinet decision once the agenda had been set (Laver and Shepsle 1996, 33 cited in Tsebelis 2002, 107). However, indeed Laver and Shepsle see their approach as easily expendable to take account of intraparty politics precisely because they are not dealing with political parties, themselves, but important party politicians cabinet ministers (Laver and Shepsle 1996, 247). These cabinet ministers seek to establish credible policy positions requested by their parties (Laver and Shepsle 1996, 248). The authors show that in certain circumstances at least, intraparty politics can have a profound impact on the making and breaking of governments through three mechanisms (Laver and Shepsle 1996, 250), of which the first is having more than one credible policy position available (Laver and Shepsle 1996, 250), the second is the impact of changes of factional leadership and the third the impact of party splitting (Laver and Shepsle 1996, ). Concerning the first mechanism the authors state that a party can gain useful strategic options from having senior politicians with diverse policy positions (Laver and Shepsle 1996, 259) by enabling the party leadership to switch between ideal points and so move policies. Although being aware of the recently renewed criticism of the model by Tsebelis (2002, 108), who disagrees and denies that ministers do have enough agenda setting power to control the agenda 4, I nevertheless stick to the model of Laver and Shepsle and take three theoretical assumptions from it: First, a minister is able to decide on policies, second, he is dependent from the support of his party, third, although being strategically an advantage for the party leadership to have a diversity of ideal points among its senior politicians (Laver and Shepsle 1996, 256), being ideological cohesive equips a party s minister with power to dictate policy in his jurisdiction and this accumulates the minister s credibility. In the next chapter I will empirically analyse the assumptions made above to answer the question whether ideological party cohesion matters, or not. 4 Tsebelis simultaneously points to the importance and respective impact of the role of prime ministers, the negotiated government program and government meetings that all diminish the power of the portfolio minister (Tsebelis 2002, 108; Tsebelis and Ha 2013). 12

13 Empiric Part Operationalization Thus far, most studies with party cohesion in focus were interested in explaining coalition building or portfolio allocation and only some with other outcomes of interparty negotiations. Additionally, most of these studies deal with roll-call data as a proxy for party cohesion or use data such as the Rice Index or the Agreement Index (Hix, Noury, and Roland 2005; Powell 2000; Rice 1925). That kind of data is clearly linked to a disciplinal concept of cohesion as defined above. Due the ability of the US congressional roll-call data, these studies are often narrowed to this very special area, which again is rather based on a disciplinal than a preferential mode of cohesion up to now. An inspiring attempt with the focus to explain government formation as determined also by ideological party cohesion has been conducted by Paul Warwick (2000; Warwick 2006). Warwick conducted a policy horizons approach where limits of compromise result in lower and upper bounds for each party on each policy dimension, which can serve as a proxy for ideological party cohesion. Although he also tests his approach empirically with data received from expert judgments and the Comparative Manifestos Project, I will take a different track and use the newly invented index of Jahn and Oberst (2012) for several reasons. Jahn and Oberst were the first to develop an index of ideological party cohesion (IPC) that compares party cohesion over time and across various political parties (2012, 222). Therefore this index is suitable for time-series-cross-section models. The authors define parties as collective actors with the implication that every actor within the party has its own set of positions and preferences. In their study they limit themselves to a single policy dimension, although they state that intraparty conflict often occurs when two or more ideological dimensions diverge (Jahn and Oberst 2012, 223). The index could be criticized on the occurrence of low cohesion values stating a cohesive party with fewer tensions, but hiding the fact that the analysed policy dimension is simply not important for the parties actors. The index therefore accounts for saliency via the analytical tool importance and so controls for the saliency of the conflict for each party (and therefore the party system too). Additionally, the index can be combined to two- or more-dimensional indices via this tool. On the other hand, the authors used a deductive country- and time-specific Leftright dimension (LR) conducted by Jahn (2011) based on data of the CMP Project (Budge et al. 2001; Klingemann et al. 2006; Volkens et al. 2013). This LR dimension should be heavily 13

14 relevant for most of the parties. 5 By focusing on ideological party cohesion as manifested in party programmes conducted for elections I pin myself to a non-behavioural or ideological driven definition of cohesion where the unit of research is a party at one point in time. The index itself is conducted on the weighted average deviation of the Leftright statements a party makes in an election manifesto. Additionally the degrees of radicalism of the statements are taken into account, which have been derived from stimulus scores of the multi-dimensional scaling analysis conducted for the LR index. Because the data has many outliers we used the mean absolute deviation (ad). We weight the ad with the frequency a statement has been made. This rawdata is finally multiplied with the importance (saliency score) of the dimension for the specific party. The formula for the weighted ad is: In the equation, ad w is the weighted average deviation. The term w i is the weight for the ith observation, which is the frequency of use of a specific statement. The stimulus score in the ith statement is x i, while b calculates the weighted arithmetic mean of all stimulus scores. (Jahn and Oberst 2012, 231). The index I use for veto player data has been conducted by Jahn (2010) and is also based on the time- and country specific Leftright index of Jahn (2011). It takes beneath positional data of coalition partners also positional data of second chambers and presidents into account and departs partially from the index Tsebelis conducted (see Tsebelis and Chang 2004). Like within Tsebelis index the ideological range between the veto players that are farthest away counts. Veto players that fall between the poles of the range will be absorbed. The index differs from Tsebelis index in questions of bicameral and presidential strength, and it uses time variant positional data of the parties, which is not the least merit of this index. The higher the values of this index the more diverse the ideology of the veto players of the index is. 5 For an interesting debate about the assets and drawbacks of this Leftright index see Benoit and Däubler 2014; Budge and Meyer 2013; Dinas and Gemenis 2010; Franzmann 2013; Franzmann and Kaiser 2006; Jahn 2011; Jahn

15 Replication and extension of the Becher 2009 study What follows is a replication of a part of the study of Michael Becher (2009) and its step by step extension. I will use his transparent and reluctant replication files, but focus exclusively on his models with net unemployment insurance entitlements as dependent variable. Afterwards I interchange his political dimension and extend the cases of Becher s models with additional data to broaden the scope of time. Finally I will bring cohesion into the models. In the empiric report here I will focus on the results of the regressions and the marginal effects exclusively. Descriptive data for the variables is to be found in the appendix. I will not report on the control variables, as long as they are in line with the results of the original study. As a starting point I replicated Becher s findings for the original time frame of the years 1972 until 2000 with his original dataset and came to exactly the same results (see regression model 1). As a second step I changed the policy dimension from myrl3 to the Leftright dimension conducted by Jahn (2011) for the partisanship of the labour minister and the adjusted veto player index ( jahn2 in orig.) also conducted by Jahn (2010) for the veto player data. I got the same, slightly better in terms of R² results for the original time frame (see regression model 2). Worth mentioning is that the veto player distance variable gained significance. At the controllers, the insignificant controller budget deficit changed its sign and the lagged controller percentage of population over 64 lost significance. Besides that, every sign and level of significance showed in the same direction. Therefore, I assume using the LR Dimension of Jahn is suitable and adds slightly explanatory power. As a third step (see regression model 3), I extended all variables 6 to the time frame of until Doing this, the model lost half of its explanatory power. However, since the impact of the partisan data is much more in my research focus, than explaining the reforms of unemployment insurance rates, I will rely on the models nevertheless. The outcome of the partisan variables and most part of the controllers is in the same direction and levels of significance. 6 This extension is possible by thankfully drawing on the datasets of Armingeon, Knöpfel, Weisstanner, Engler, Potolidis, and Gerber 2013 and for the dependent variable on data from the CWED2 project of Scruggs, Jahn, and Kuitto Becher (2009, 38) sets the 1973 oil crisis as starting point for his study, the robustness of the models when extending the time frame to the year of 1971 is to be checked accordingly. There is some missing data producing gaps, most variables start

16 Subsequently, I set up the same model but bring cohesion in via twopartite interaction with the positional variable without (see regression model 4) and with veto players as a main effect (see regression model 5). Again all signs show the expected direction and the interaction term is significant. Finally, I set up a model (see regression model 6) that includes positional, veto player and cohesional aspects as a tripartite interaction. To interpret tripartite interactions is not an easy task and although the tripartite interaction term is not significant, I stick to the advice of Mitchell (2012, 163), that in the presence of a three-way interaction, interpretations of lower interactions are not fruitful and marginal effects should be considered to interpret the impact of the interaction. The following tables show the regression estimates and a short explanation of the time and dimensional frame used for each regression. # Model Time period Cases Ideol. Dep. dimension Var. 1 Becher 2009 original data Myrl3 Orig. 2 Becher 2009 original data; same cases; LR dimension 3 Extended data; extended cases; LR dimension 4 Extended data; extended cases; LR dimension; Cohesion included 5 Extended data; extended cases; LR dimension; Cohesion and Veto Players included 6 Extended data; extended cases; LR dimension; Cohesion and Veto Players included as 3 term interaction LR Orig LR Ext LR Ext LR Ext LR Ext. 16

17 Table 2: Panel regression estimates (1) (2) (3) (4) (5) (6) Labourminister Position * (0.005) * (0.034) * (0.016) * (0.034) * (0.035) * (0.058) Vetoplayer Distance (0.005) * (0.025) (0.014) * (0.014) (0.030) Labourminister Cohesion * (0.009) * (0.010) * (0.013) Position # Vetoplayer * (0.006) * (0.163) * (0.096) (0.441) Position # Cohesion * (0.082) * (0.085) (0.139) Vetoplayer # Cohesion * (0.081) Position # Vetoplayer # Cohesion (0.908) Lagged dependent variable * (0.026) * (0.024) * (0.015) * (0.015) * (0.015) * (0.015) L.Corporatism * (0.012) * (0.009) * (0.005) * (0.005) * (0.005) (0.005) L.Unemployment rate * * * * * * D.Unemployment rate * (0.002) * (0.002) * * L.Trade openness (%GDP) * * * D.Trade openness (%GDP) * * * * L.Log of GDP per capita * (0.041) * (0.040) (0.007) (0.008) (0.008) (0.007) L.GDP growth * * * * * * L.Population over 64 (%) * (0.003) (0.003) D.Population over 64 (%) (0.019) (0.014) (0.007) (0.007) (0.007) (0.007) L.De-industrialization * (0.002) * (0.002) * * * * D.De-industrialization * (0.002) * (0.002) L.Budget deficit (%GDP) Countries r adjusted_r Observations Panel-corrected standard errors in parentheses; * p<0.05, all models with random and fixed effects, not reported here; L= one year lag; D= first difference; Excluding CAN and GBR because of assuming not to have veto players does not alter the results of the marginal effects. 17

18 The marginal effects of cohesion What follows are the interpretation of the influence of the marginal effects of the respective interacted partisan variables for four of the six models above. Figure 4 Marginal effects for model 1, 2, 3 and 5 Model 1 Model 2 Becher original Model Becher Model within LR Dimension vrange_myrl3 Observations: a_jahn2100 Observations: 382 Model 3 Model 5 Oberst Model, LR Dimension, extended time frame, POS#VPL Oberst Model, LR Dim., extended time frame, POS#COH; VPL a_jahn2100 Observations: COHESION Observations: 570 The interpretation of the marginal effects for the first model, the original model of Becher, is that veto players have a relevant effect. The effect is significant and negative for veto players until (.75), which can be interpreted as until that range of veto player, a significant effect shows that the more veto player, the less the capability of the labour minister to implement the policy preferred (Becher 2009, 49). At the end of the scale the effect becomes significant again but turns to the positive side of the scale. Becher (2009, 51) interprets this rather rare event as a possible outcome of logrolling among parties. In the second model the direction of the effect is the same, although the turning effect starts at lower levels of veto player influence and is more distinct. In my interpretation 18

19 the capability of the minister to implement his preferred policy is so small due to high levels of veto player influence, that the minister even has to go against his preferred partisan line. Nevertheless we can result from this two models that veto players do have an impact on policy outcomes. Extending the time frame as in the third model weakens the relevancy of this effect and the room to manoeuver in the preferred direction is even narrower. Additionally, the turning effect starts already with lower levels of veto player. Bringing cohesion in as in the fourth (not reported here) and fifth model leads to the assumption that also cohesion has a relevant and significant effect on policy making. The less tensions within the minister s party, the more capable of implementing the preferred policy the minister is. This result is significant until a level of cohesion of (.45). The result shows that, when the minister is backed by an internal unified party, the minister can act in line of the partisan hypothesis. With increasing levels of cohesion this effect becomes insignificant, but also positive, which implies a similar trend as for veto player effects. The interpretation of the average marginal effect of the tripartite interaction underlines the above assumed interpretations. The result of the model with the tripartite interaction of position, cohesion and veto players is very interesting. First, as can be seen from the average marginal effects of position as a function of cohesion and veto players (figure 5), it can be said, that cohesion and veto players do have relevant impact on the positional impact of the labour ministries. The positional impact is negative and significant, when there are almost none and less veto players (dotted and diamonded line) and this impact shrinks by increasing levels of cohesion, stating that the more incoherent the party of the labour minister is and the more veto players act, the less partisan impact on NRR will there be. By that we can interpret that a left labour minister can raise NRR better when s/he has no internal dispute and less veto players, whereas a right minister can reduce NRR better within the same constellation of low internal dispute and veto player. This effect is significant for the constellation of zero veto players through all levels of cohesion and also for small veto players (.1) until a high level of cohesion (.45). 19

20 Figure 5 Average Marginal Effect of the three-way Interaction Avrg Marg Effects of Position as a function of Cohesion and VPL POS#COH#VPL Cohesion L.a_jahn2100=0 L.a_jahn2100=.2 L.a_jahn2100=.4 L.a_jahn2100=.1 L.a_jahn2100=.3 Vetoplayer: From.0 to.4; no confid. interv. shown Observations: 570 This result is generally in line with partisan hypothesis. But with higher levels of veto players and more internal dispute the partisan effect is turned around, stating a positive effect of ministerial position on the NRR rate, which can be interpreted as follows: when many veto players are to be played and internal division is high, the partisan effect is then that left ministers reduce and right ministers expand NRR. This effect is significant for a (.35) level of cohesion with high veto players (.3) and also significant for a (.25) level of cohesion with very high veto player (.4) constellations. The following graph shows the above levels of veto players in detail with confidence intervals on a 95% level to be able to analyse the significance of the different veto player levels. 20

21 Figure 6 Average marginal effects of position as a function of cohesion and veto players by different levels of veto player Avrg Marg Effects of Position as a function of Cohesion and VPL POS#COH#VPL Avrg Marg Effects of Position as a function of Cohesion and VPL POS#COH#VPL Cohesion Vetoplayer:.0 Observations: Cohesion Vetoplayer:.1 Observations: 570 Avrg Marg Effects of Position as a function of Cohesion and VPL POS#COH#VPL Avrg Marg Effects of Position as a function of Cohesion and VPL POS#COH#VPL Cohesion Vetoplayer:.2 Observations: Cohesion Vetoplayer:.3 Observations: 570 Avrg Marg Effects of Position as a function of Cohesion and VPL POS#COH#VPL Cohesion Vetoplayer:.4 Observations: 570 The Graphs show in ascending order the average marginal effects of position as a function of cohesion and veto players by different levels of veto players. A negative effect is in line with the partisan hypothesis, zero cohesion is associated with less ideological tensions within a party and zero veto players are associated with less adverse veto players. 21

22 Discussion By replicating and extending the study of Becher (2009) with cases I could show that his findings that veto players constrain the room to manoeuvre still hold true within a different political dimension. The dimension I have analysed here is a deductive Leftright index with additional time- and country specific aspects conducted by Jahn (2011). In a next step I brought ideological party cohesion in via a newly invented index by Jahn and Oberst (2012). I interacted this index in different models with position and in a three-way interaction with position and veto player aspects. Doing this I not only confirmed the result of Becher, that it is premature to dismiss the importance of veto players (Becher 2009, 34), I have also shown that a minister backed by an ideological homogenous party is more likely able to set the agenda in her or his preferred direction. Therefore cohesion matters for policy outputs. Since this is a relevant effect, cohesion should no longer be neglected in policy research too. The analysis brought also to light that the higher internal dispute of the ministerial party and the broader the ideological range of veto players is, the capability of a minister to implement policies in the preferred direction shrinks. Additionally with high levels of internal dispute and veto players ministerial effects are even associated with outputs against the line of partisan hypothesis. Further research has to be done whether this could be linked to concepts like blame avoidance or is just a result of partisan logrolling, like Becher (2009, 51) puts it. The model above should also be tested in other policy areas. By proving the applicability of the new index of ideological party cohesion the door has been opened for until now neglected research concerning cohesion. References Allan, James and Lyle Scruggs "Political Partisanship and Welfare State Reform in Advanced Industrial Societies." American Journal of Political Science 48 (3): Allard, Gayle J. "Measuring job security over time: In search of a historical indicator for EPL (employment protection legislation)." Armingeon, Klaus, Laura Knöpfel, David Weisstanner, Sarah Engler, Panajotis Potolidis, and Marlène Gerber Comparative Political Data Set I Bern. Institute of Political Science, University of Bern. Avdagic, Sabina "Partisanship, political constraints, and employment protection reforms in an era of austerity." European Political Science Review 5 (03): (Accessed March 31, 2014). 22

23 Bäck, Hanna "Intra-Party Politics and Coalition Formation. Evidence from Swedish Local Government." Party Politics 14 (1):71 89 (Accessed March 30, 2014). Bäck, Hanna "Intra-party politics and local coalition formation." In Intra-party Politics and Coalition Governments, eds. Daniela Giannetti, and Kenneth Benoit. London: Routledge, Bäck, Hanna "The ideological cohesion of parliamentary parties and its implications for decision-making in modern democracies." Statsvetenskaplig tidskrifts 114 (1): (Accessed December 11, 2013). Baltz, Konstantin "Spieltheoretische Modellierung in den international vergleichenden Beziehungen." In Methoden der vergleichenden Politik- und Sozialwissenschaft. Neue Entwicklungen und Anwendungen, eds. Susanne Pickel, Gert Pickel, Hans-Joachim Lauth, and Detlef Jahn. Wiesbaden: VS Verlag für Sozialwissenschaften, Baumann, Markus, Marc Debus, and Jochen Müller "Das legislative Verhalten von Bundestagsabgeordneten zwischen persönlichen Charakteristika, Wahlkreisinteressen und Parteilinie. Eine Untersuchung am Beispiel der Auseinandersetzung um die Präimplantationsdiagnostik." Zeitschrift für Politikwissenschaft 23 (2): (Accessed March 11, 2014). Becher, M "Constraining Ministerial Power: The Impact of Veto Players on Labor Market Reforms in Industrial Democracies, " Comparative Political Studies 43 (1):33 60 (Accessed June 10, 2013). Benoit, Kenneth and Thomas Däubler Putting Text in Context: How to Estimate Better Left- Right Positions by Scaling Party Manifesto Data using Item Response Theory. Prepared for the Mapping Policy Preferences from Texts Conference, May 15 16, 2014, Berlin. Blumenau, Jack "Agenda Control and Party Cohesion in the European Parliament. Draft paper prepared for presentation at the European Political Science Association General Conference, Berlin, June 2012." (Accessed March 31, 2014). Boucek, Françoise "Rethinking Factionalism: Typologies, Intra-Party Dynamics and Three Faces of Factionalism." Party Politics 15 (4): Bowler, Shaun, David M. Farrell, and Richard S. Katz "Party Cohesion, Party Dsicipline, and Parliaments." In Party Discipline and Parliamentary Government, eds. Shaun Bowler, David M. Farrell, and Richard S. Katz. Columbus: Ohio State University Press, Budge, Ian, Hans-Dieter Klingemann, Andrea Volkens, Judith L. Bara, and Eric Tanenbaum, eds Mapping Policy Preferences. Estimates for Parties, Electors, and Governments, Oxford: Oxford University Press. Budge, Ian and Thomas Meyer "Understanding and Validating the Left-Right Scale (RILE)." In Mapping Policy Preferences from Texts. Statistical Solutions for Manifesto Analysts, eds. Andrea Volkens, Judith L. Bara, Ian Budge, Michael D. McDonald, and Hans-Dieter Klingemann. Oxford: Oxford University Press, Burden, Barry C Personal roots of representation. Princeton: Princeton University Press. Cox, Gary W. and Mathew D. McCubbins Setting the Agenda. Responsible Party Government in the U. S. House of Representatives. Cambridge: Cambridge University Press. Debus, Marc and Thomas Bräuninger "Intra-party factions and coalitional bargaining in Germany." In Intra-party Politics and Coalition Governments, eds. Daniela Giannetti, and Kenneth Benoit. London: Routledge,