Since the 1950s, coalition theorists have tried to explain how coalition governments make their policy proposals. To recall one of the most influential contributions, Michael McDonald and Ian Budge (Reference McDonald and Budge2005: 147–8) found that declared cabinet position better accords with the position of the median voter than with the position of the parliamentary median party or the weighted mean position of coalition parties. However, employing the same data set, Paul Warwick (Reference Warwick2001, Reference Warwick2011) obtained different results: he revealed that declared cabinet position responds to the weighted mean position of coalition parties and, where present, to the position of external support parties.
Taking its cue from the so far mixed results provided in the literature, this article investigates to what extent declared cabinet position corresponds to the policy stances staked out by governing parties at election time in their manifestos. Furthermore, drawing on established theoretical arguments, the article evaluates to what extent other political actors (that is, the parliamentary median party, the party of the prime minister (PM) and individual ministers) are able to influence declared cabinet position beyond their contribution to the weighted mean position of coalition parties. Delving into the possible answers to these questions is relevant inasmuch it tells us about what G. Bingham Powell (Reference Powell2005: 62–76) refers to as the ‘chain of responsiveness’ between voter policy preferences and government policy positions. Indeed, if the parliamentary median party proves to be the key player in determining declared cabinet position, then which government forms after the dust of the election has settled becomes less significant than scholars have so far assumed. Conversely, if declared cabinet position reflects a compromise among the coalition partners or if it is largely given by the positions of individual ministers, then the result of bargaining over government formation becomes highly relevant for explaining and predicting policy outcomes.
To date, studies dealing with the degree of ideological congruence between the electoral promises made by parties at election time and government policy stances have largely focused on the ideological right–left (RILE) dimension. This choice is consistent with Robert Dahl’s (Reference Dahl2005 [1961]: 223–328) pluralistic approach to politics, according to which the actors involved (that is, the coalition parties), their electoral promises and the resources they own (their parliamentary seat share) determine the policy content in different fields. In a sentence: politics determines policies.
However, the reverse of this often-cited sentence – Theodore Lowi’s claim that policies determine politics (1972) – also deserves our attention. Indeed, Lowi’s (Reference Lowi1972) work reminds us that the issue at stake is likely to affect the relative weight of the actors involved in the formulation of the government policy proposal.
Accordingly, this article contributes to the existing literature on how coalition governments formulate their policy statements by contrasting the degree of congruence between the declared cabinet position and the weighted mean position of governing parties as expressed in their electoral manifestos on two dimensions: the traditionally employed RILE scale and a genuinely policy-based welfare scale. This comparison checks whether the actors able to influence the policies that coalition partners jointly agree to take forward in the government declaration are the same for both dimensions.
To achieve this goal, the article draws on a largely neglected and long-abandoned part of the data collection made by the Comparative Manifesto Project (CMP; Volkens et al. Reference Volkens, Lehmann, Merz, Regel, Werner and Schultze2014): namely, the coding of government declarations in 10 European countries with the same coding scheme that has been used to analyse party manifestos over the entire post-war period (Budge et al. Reference Budge, Klingemann, Volkens, Bara and Tannenbaum2001).
Results demonstrate that the actors able to shape government declarations are not the same for the two dimensions. On the ideological RILE dimension, which includes several policy areas, the PM’s party determines the content of the declared cabinet position beyond its contribution to the weighted mean position of coalition parties. Thus, the results suggest that when the party of the PM is forced to compromise with the other coalition parties on several policy dimensions, it is able to pull the declared cabinet position towards its ideal point by pivoting and counterbalancing alternative solutions.
Instead, the parliamentary median party and competent ministers (that is, labour and social affairs ministers) are key players on the welfare dimension. Welfare policies, indeed, ask governments to distribute societal wealth to specific groups, usually subtracting resources from other societal segments. Given the distributive and redistributive conflicts entailed in this policy domain and the blame-sharing strategies pursued by governing parties forced to retrench the welfare state, the party owning the median legislator in the parliamentary arena is highly involved in decision-making processes, together with the competent ministers. Furthermore, when parties experience rising inflation while they are in office, it seems to make the declared cabinet position drift rightwards.
The article is structured as follows. The next section critically reviews the theories advanced in the literature and details the research hypotheses. Next, the data set and the model specification are described. Then, in the main analytical section, the role played by governing parties, the parliamentary median party, the PM’s party and individual ministers in shaping declared cabinet positions both on the RILE and the welfare dimensions are assessed. The last section discusses the main findings and indicates paths for future research.
THEORETICAL FRAMEWORK AND MAIN HYPOTHESES
Declared Cabinet Position on the RILE Dimension: Who is Expected to Shape it and Why?
Coalition governments have to confront a fundamental challenge in policymaking that is absent for single-party governments: government policy proposals on any specific issue have to be made jointly by parties that may express divergent policy positions and that are held separately accountable at election time. This challenge raises a crucial question: whose policy positions are ultimately reflected in government policy proposals? Thus far, the literature has provided at least four alternative accounts to answer this question.
A first account maintains that the declared cabinet position simply reflects a compromise among the governing parties. This intuitive expectation rests on the extension of William Gamson’s law (Reference Gamson1961) from the office dimension (the quantitative allocation of portfolios among coalition partners) to the policy dimension (the policy proposals ratified by coalition partners).
Taking advantage of the Government Declarations Data on 10 European countries (Budge et al. Reference Budge, Klingemann, Volkens, Bara and Tannenbaum2001), Warwick (Reference Warwick2001, Reference Warwick2011) demonstrated that declared cabinet positions on the RILE dimension correspond to the mean position of governing parties as derived from their electoral manifestos, weighted by their seat shares. Moreover, he proved that other political actors are able to bias declared cabinet positions towards their ideal points: namely, the party holding the finance portfolio, the formateur party (that is, the party that was assigned the task of putting the coalition together), parliamentary parties (Warwick Reference Warwick2001) and external support parties (Warwick Reference Warwick2011). More recently, Lanny Martin and George Vanberg (Reference Martin and Vanberg2014) found strong evidence that government-sponsored legislation adopted in Denmark, Germany and the Netherlands since the 1980s reflects a compromise among the policy positions of coalition partners. Taking our cue from these results, we formulate our first hypothesis:
Hypothesis 1: Declared cabinet position corresponds to the weighted mean position of the coalition parties, with their share of seats in the lower chamber constituting the weights.
Hypothesis 1 serves as a baseline assumption in this article. This approach is common in various studies on coalition politics and parliamentary governments (see also Grofman Reference Grofman1982). However, in addition to this extension of Gamson’s law (Reference Gamson1961), alternative accounts on the distribution of policy payoffs to the members of multiparty governments will be taken into consideration. Note that the following accounts are not considered as full alternatives to Hypothesis 1, but rather as causing deviations from the expected Gamson’s result.
A second account is the well-known median voter theorem (Black Reference Black1958; Downs Reference Downs1957: 142–63), according to which the party controlling the median legislator will have a stronger bargaining power than the other coalition parties because there are no other points in the ideological space that are preferred by a majority to its ideal point. Accordingly:
Hypothesis 2: Declared cabinet position is biased away from the weighted mean position of the coalition parties towards the position of the parliamentary median party.
Testing the median voter theorem, Ian Budge and Michael Laver (Reference Budge and Laver1992b: 409–64) found that the declared cabinet position on the RILE dimension did not mirror the positions of either the median party or the predominant party. Notwithstanding these discouraging results, McDonald and Budge (Reference McDonald and Budge2005: 147–8) found a reasonably strong relationship between declared cabinet position and that of the parliamentary median party.
These mixed results suggest that the declared cabinet position may reflect other influences (Warwick Reference Warwick2001, Reference Warwick2011). Among those factors, this article focuses on the role played by the PM’s party (third account) and by the ministers having jurisdiction in specific policy fields (fourth account).
The third account refers to the general class of proposer models (Baron Reference Baron1998; Diermeier and Feddersen Reference Diermeier and Feddersen1998), which suggests that the formateur party is able to bias the declared cabinet position towards its ideal point beyond its share of seats. These models maintain that the formateur party, being in charge of proposing an alternative to the status quo, will propose a policy that makes the median voter indifferent between accepting or rejecting its proposal (that is, the median voter ideal point will be located in the interval between the status quo and the formateur ideal point). In this way, the proposal of the formateur party will be accepted. In this article, the formateur party coincides with the party of the PM (Woldendorp et al. Reference Woldendorp, Keman and Budge2000).Footnote 1
Hypothesis 3: The declared cabinet position is biased away from the weighted mean position of the coalition parties, towards the position of the PM’s party.
Finally, the fourth account, the portfolio allocation model by Michael Laver and Kenneth Shepsle (Reference Laver and Shepsle1996: 281–5), assumes that ministers are policy dictators in their jurisdictions. This model maintains that the policy positions of the minister party will prevail in the specific policy area governed by that minister. In this regard, Warwick (Reference Warwick2001) demonstrated that the RILE position of the party of the finance minister has only a modest influence on declared cabinet positions.
Hypothesis 4: The declared cabinet position is biased away from the weighted mean position of the coalition parties towards the position of the party of the competent minister.
Going Beyond the Left–Right Continuum by Taking a Deeper Look at the Welfare Dimension
The large majority of the studies reviewed here only focused on the ideological RILE scale. By contrast, this article investigates how coalition governments build their policy proposals by developing two parallel analyses. The first tests the four hypotheses listed so far on the RILE dimension; the second extends the same analysis to the pro–anti welfare state expansion dimension. Indeed, only the comparison between these two dimensions allows us to verify whether the political actors able to pull the declared cabinet position towards their ideal points coincide.
We chose the welfare dimension for several reasons. It is the most debated policy area in the literature on the degree of congruence between parties’ long-term ideological positions, their contextual electoral pledges and subsequent policy choices (e.g. Barnes Reference Barnes2013; Häusermann et al. Reference Häusermann, Picot and Geering2013).
The stream of literature known as Partisan Theory (e.g. Hibbs Reference Hibbs1992), for example, expects governing parties to maximize the interests of their core electoral constituencies. Left-wing governments, being the advocates of the working class, are expected to fight for welfare state enlargement or at least to minimize welfare state retrenchment; right-wing governments, being supported by wealthier groups, are expected to favour a reduction of state intervention into the economic sphere.
In the same vein, McDonald and Budge (Reference McDonald and Budge2005: 149–50) applied the median voter theorem (Black Reference Black1958; Downs Reference Downs1957: 142–63) to the welfare domain with encouraging results: the welfare position of the median legislator positively correlates with declared cabinet position on the same topic. However, other scholars reply that parties’ ability to shape welfare policies has been reduced, if not completely erased, by institutional configurations (e.g. Laver and Shepsle Reference Laver and Shepsle1996: 61–124), divided government and shared policy control (e.g. Tsebelis Reference Tsebelis2002: 187–206), economic constraints or policy legacies (e.g. Pierson Reference Pierson1994: 27–50).
Studies in this field search for a connection between governing parties’ positions and social expenditure or aggregate indexes of welfare state generosity. These studies, however, tend to operationalize government partisanship through dummy or categorical variables indicating whether the cabinet is more left- or right-leaning. Even when more sophisticated measures based on party manifestos or expert surveys are employed, governing parties’ positions are put in direct relationships with social policy outcomes without investigating the intermediate passage between the positions expressed by governing parties in their manifestos at election time and the declaration pronounced by each newly formed government at the beginning of the mandate. This article intends to shed light on this missing link.
Moreover, we chose the welfare dimension because it is related to the most conflict-ridden cleavage in industrial democracies, the capital–labour cleavage (Pierson Reference Pierson1994: 27–50). Thus, governments are expected to make considerable efforts because they know they will be largely evaluated according to their achievements in this field. Furthermore, the role of the state in providing social services is the most debated topic in government declarations (Martin and Vanberg Reference Martin and Vanberg2014). Finally, providing welfare support is a large part of what industrial democracies do. Thus, the focus on this domain, which implies heavy budgetary consequences, allows elucidation of the so-called politics of constrained choice.
As explained by Lowi (Reference Lowi1972), the policy domain at stake affects the actors taking part in decision-making processes and their relative weights. He applied his well-known policy taxonomy (that is, distributive, redistributive, regulatory and constituent policies) to US presidential politics from Roosevelt to Johnson, showing that Congress played a major role in distributive and regulatory policies, while the executive power was stronger in shaping redistributive policies.
Adapting Lowi’s theoretical claim to this analysis, the actors able to shape declared cabinet positions on the RILE and the welfare dimensions are expected to differ. Indeed, on the genuinely policy-based welfare dimension, which typically involves distributive and redistributive conflicts among segments of society, the legislative assembly (for distributive policies) and the competent ministers (for redistributive ones) are likely to exert significant influence on the formulation of the declared cabinet position.
Conversely, on the ideological RILE dimension, which is a super-issue involving heterogeneous policy areas (as well as the macroeconomic dimension, it includes military policies, human rights, internationalism, constitutionalism, political authority, traditional morality, law and order, etc.), the PM’s party is expected to play the leading role. The PM, indeed, is likely to pull the declared cabinet position towards his/her ideal point by balancing and counterbalancing the policy proposals formulated by the other ministers and coalition parties in several policy areas.
DATA DESCRIPTION AND MODEL SPECIFICATION
Dependent Variables
The key ingredient to build the two dependent variables employed in the analysis is provided by a long-abandoned part of the data collection made by the CMP, the Government Declarations Data (Budge et al. Reference Budge, Klingemann, Volkens, Bara and Tannenbaum2001). It contains the coding of government declarations in 10 European countries (Belgium, Denmark, France, Ireland, Italy, Germany, Luxembourg, the Netherlands, Norway and Sweden) from the end of the Second World War to the end of the 1990s (Table A1 in the Appendix reports the time covered for individual countries). Government declarations are public speeches pronounced in institutional settings (usually the parliament) by a head of state on behalf of a recently formed government or by a PM at the time of his or her investiture debate (Laver and Budge Reference Laver and Budge1992: 19). These statements initiate the governing process by detailing the official programme to which the government publicly commits itself. Accordingly, they are an intent rather than a fulfilment with regard to policies (McDonald and Budge Reference McDonald and Budge2005: 141).Footnote 2
Because of bargaining, anticipated administrative and practical constraints and the need to obtain parliamentary approval, government declarations differ from the ‘weighted wish list’ derived from the electoral manifestos of the coalition parties (Budge et al. Reference Budge, Klingemann, Volkens, Bara and Tannenbaum2001: 172–4). Indeed, party manifestos are static documents written by parties just before election time (Dolezal et al. Reference Dolezal, Ennser-Jedenastik and Müller2012). Leonard Ray (Reference Ray2007) defined these documents as contracts between parties and voters containing a realistic assessment of the policies that they would implement if elected. However, together with policy pledges, party manifestos contain party advertising and rather abstract statements of the party identity and philosophy. Notably, Martin Dolezal et al. (Reference Dolezal, Ennser-Jedenastik and Müller2012) underlined that since party manifestos are written to be used during the electoral campaign, they embed positive references to past party records and attacks on their competitors (negative campaigning).Footnote 3
The difference between government declarations and the electoral manifestos of coalition parties also emerges from the content analysis of these documents conducted by the CMP. Indeed, government declarations are usually shorter than party manifestos (in each country, the mean number of quasi-sentences in government declarations is lower than the mean number of quasi-sentences in party manifestos). Moreover, by comparing the policy content of these two types of documents, it emerges that the electoral dynamics that dominate party manifestos are substituted by more credible policy pledges in government declarations. Indeed, in its investiture speech, the government is supposed to publicly engage and involve the members of parliament (MPs), shifting the dimension of conflict from purely ideological positions to issues more related to the actual agenda of the cabinet. Accordingly, policy fields such as the need to keep military treaty obligations, to support international organizations and the European Union, to enforce law and order, to improve government and administrative efficiency and to reduce budget deficits are more frequent in government declarations than in party manifestos.
Having clarified the difference between government declarations and the electoral manifestos of the coalition parties, we detail the operationalization of our dependent variables. The first dependent variable, declared cabinet position on the RILE dimension, is measured through the well-known method proposed by Budge and Laver (Reference Budge and Laver1992a: 1–64). Specifically, the position held by each government on the RILE dimension is equal to the difference between the proportion of each government declaration devoted to 13 categories identified as right-wing and the proportion devoted to 13 left-wing ones.Footnote 4 The theoretical range of the RILE scale is −100 (extreme left) to +100 (extreme right), although in practice this variable is between −45.8 and +55.1 (see Table A2 in the Appendix).
The second dependent variable, declared cabinet position on the welfare dimension, is calculated by subtracting the proportion of each government declaration devoted to welfare state expansion (CMP variables per503 and per504) from the proportion devoted to welfare state retrenchment (CMP variable per505). The theoretical range of the welfare scale is −100 (welfare state expansion) to +100 (welfare state retrenchment), but in the sample the variable ranges from −40.9 to +3.2 (see Table A2).
The CMP variables employed to build the welfare scale are included in the formula used to estimate the RILE scale. To what extent do government welfare positions resemble those on the RILE dimension? To answer this question, Figure 1 plots the pairwise correlation coefficient between the two.
Pairwise Correlation between Declared Cabinet RILE and Welfare Positions
The variable declared cabinet position on the welfare dimension is positively correlated with that on the RILE dimension (i.e. 0.565; see Table A3 in the Appendix). However, this coefficient is far from being equal to 1, suggesting that these two dependent variables, even if similar, do not describe exactly the same phenomenon. This result gives reason to test whether the same political actors affect declared cabinet positions on these two dimensions.
Main Independent and Control Variables
For independent variables, we used the CMP coding of party electoral manifestos (Volkens et al. Reference Volkens, Lehmann, Merz, Regel, Werner and Schultze2014) to measure the weighted mean position of governing parties, the position of the median party in parliament, that of the PM’s party and that of the parties holding the finance, the labour and the social affairs portfolios.
Using Katsunori Seki and Laron Williams (Reference Seki and Williams2014), we merged the CMP data set (Volkens et al. Reference Volkens, Lehmann, Merz, Regel, Werner and Schultze2014) with the information on government compositions provided by Jaap Woldendorp et al. (Reference Woldendorp, Keman and Budge2000). The cabinet weighted mean on the RILE dimension corresponds to the mean of the RILE positions of cabinet parties, weighted by their share of seats in the lower chamber (Powell Reference Powell2009). Table A3 shows that the cabinet weighted mean on the RILE dimension is positively correlated with the declared cabinet position on the same dimension (0.446). This coefficient, however, is far from being equal to 1, making the case for other actors affecting the first dependent variable.
Similarly, we assessed the cabinet weighted mean on the welfare dimension as the mean of the welfare positions of cabinet parties, weighted by their share of seats in the lower chamber. As before, the correlation coefficient is far from being equal to 1 (0.401; see Table A3).
The positions of the median legislator on the RILE and welfare dimensions have been calculated by assuming that legislator positions can be represented by the positions of the parties to which they belong. To identify the party containing the median legislator on the RILE dimension, parliamentary parties have to be arrayed in a left–right order. To do the same on the welfare dimension, parties have to be ranked according to their scores on the welfare issue.Footnote 5
Hypothesis 2 suggests that the positions of the median legislator influence declared cabinet positions on the RILE and the welfare dimensions beyond their contribution to the weighted mean. As in Warwick (Reference Warwick2011), these types of effects are captured by two variables. The first variable, median party – cabinet distance on the RILE dimension, corresponds to the RILE position of the median party minus the cabinet weighted mean on the same dimension. The second variable, median party – cabinet distance on the welfare dimension, measures the difference between the welfare position of the median party and the cabinet weighted mean on the same dimension.
Hypothesis 3 maintains that the position of the PM’s party influences declared cabinet positions on the RILE and welfare dimensions beyond its contribution to the weighted mean. This party has been identified through Woldendorp et al.’s (2000) data set. Thus, the variable PM’s party – cabinet distance on the RILE dimension is equal to the RILE position of the PM’s party minus the cabinet weighted mean on the same dimension. Similarly, the variable PM’s party – cabinet distance on the welfare dimension records the deviation of the PM’s party from the cabinet weighted mean on the welfare dimension.
The last political actors that are potentially able to affect declared cabinet positions are ministers (Hypothesis 4). Laver and Shepsle’s portfolio allocation model (1996: 281–5) is less easily tested with the data at hand because the RILE and the welfare scales appear too encompassing to be under the exclusive control of any one minister. However, following Warwick (Reference Warwick2011), it is possible to assert whether the declared cabinet position on the RILE dimension is heavily influenced by the position of the party holding the finance minister and whether the declared cabinet position on the welfare dimension is heavily influenced by those of the parties holding the labour and social affairs portfolios. These types of effects can be verified by determining whether the deviations of the positions of these ministers from the cabinet weighted mean (i.e. finance minister – cabinet distance on the RILE dimension and labour/social affairs ministers – cabinet distance on the welfare dimension) exert any independent influence on declared cabinet positions.
All the hypotheses formulated in the first section of this article focus on the political actors able to affect the final formulation of declared cabinet positions on the RILE and welfare dimensions. However, external influences are possible too. In particular, the declared cabinet position is likely to deviate from the coalition parties’ electoral promises towards a more rightist attitude in response to adverse economic conditions. Even if reasonable, this expectation has not found empirical support so far (Warwick Reference Warwick2001, Reference Warwick2011). This article controls for adverse economic conditions through three variables. Change in inflation rate records the difference between the inflation rate registered in the month in which the government was formed and the inflation rate in the last month of the mandate of the preceding government. Similarly, the variable change in unemployment rate registers the corresponding difference for the unemployment rate.Footnote 6 The same occurs for the variable change in GDP (Strøm et al. Reference Strøm, Müller and Bergman2008). Table A2 in the Appendix provides descriptive statistics for all the variables listed so far, while Tables A3, A4 and A5 display pairwise correlations.
Concerning the choice of the functional form, it may be that the declared cabinet position will be influenced by the position of its predecessor (serial correlation) and that the declared cabinet position in one country may be systematically different from that in another country (panel heteroscedasticity). Thus, all the models include a first-order lag of the dependent variable and employ panel-corrected standard errors (Beck and Katz Reference Beck and Katz1995). The lagged dependent variables in this analysis are the declared cabinet positions on the RILE and welfare dimensions of the immediately preceding government. The inclusion of the lagged dependent variables captures the idea that declared cabinet positions may be influenced by the ‘dead weight’ of past policies: whatever policy position a coalition government agrees on, it may be necessary or expedient to move policies to that ideal point only gradually.Footnote 7 In McDonald and Budge’s words, ‘policies change from one government to the next, but not much and not rapidly’ (2005: 171).
Finally, all the models embed a set of country dummies to control for time-invariant cross-country differences (Wilson and Butler Reference Wilson and Butler2007).
RESULTS
The testing procedure will be guided by the notion that, in the absence of any other influences, the declared cabinet position should reflect the weighted mean position of governing parties (Hypothesis 1). The hypotheses regarding the other political actors (Hypotheses 2, 3 and 4), indeed, involve Hypothesis 1 as a reference point.
This analytical section is structured into three parts. A first subsection investigates the role played by the additional political actors (the parliamentary median party, the PM’s party and the party holding the finance or the labour/social affairs ministers) on the RILE dimension. Next, a second subsection replicates the same analysis on the welfare dimension.
Note that the effects of each political actor on declared cabinet positions are investigated one by one. This choice is imposed by the shortage of data; the Government Declarations Data (Budge et al. Reference Budge, Klingemann, Volkens, Bara and Tannenbaum2001) provides complete information on just 157 governments (see Table A1), while there are four hypotheses to be tested on each dimension, without taking into account economic control variables and country fixed effects. However, acknowledging the importance of identifying the net effect of each independent variable when all other influences are controlled for, a third subsection displays two additional model specifications in which declared cabinet positions on the RILE and welfare dimensions are regressed on the most important independent variables referring to both political actors and economic conditions.Footnote 8 The results displayed in this third subsection clarify the previous ones.
Influences on Declared Cabinet Position on the RILE Dimension
The models displayed in Table 1 prove that declared cabinet position on the RILE dimension is strongly affected by the position of its predecessor and by the weighted mean position of coalition parties (Hypothesis 1). Model M1 proves that if the weighted mean position of the coalition parties moves one point to the right, the position staked out by the government in its declaration is likely to increase by approximately 0.3 points.
Influences on Declared Cabinet Position on the RILE Dimension
Notes: OLS estimations with panel corrected standard errors in parentheses. The dependent variable is declared cabinet position on the RILE dimension. Model M3 has 145 observations rather than 147 because the variable PM’s party – cabinet distance is missing for the governments Barre I and II in France. Indeed, Raymond Barre has never been affiliated to any party.
† p<0.10, * p<0.05, ** p<0.01, *** p<0.001.
However, the beta referring to the weighted mean position of coalition parties is positive and significant but far from being equal to 1. Thus, it is worth looking for additional political actors and external forces able to shape the declared cabinet position on the RILE dimension in addition to the coalition partners.
Model M2 does not support Hypothesis 2, according to which governments respond to the position of the parliamentary median party; indeed, the variable median party – cabinet distance does not reach statistical significance. Models M3 and M4 lead to the same conclusions for Hypotheses 3 and 4, according to which the PM’s party and the party holding the finance minister are likely to bias the declared cabinet position towards their ideal points. Indeed, the betas referring to the variables PM’s party – cabinet and finance minister – cabinet distances are not statistically significant.Footnote 9 However, the variable PM’s party – cabinet distance is quite near to conventional levels of statistical significance (p=0.105), suggesting that Hypothesis 4 merits deeper investigation.Footnote 10
The results discussed so far demonstrate that, even taking into account the RILE positions of the parliamentary median party, the PM’s party and the finance minister, substantial differences remain between the declared cabinet position and the weighted mean position of governing parties. These two variables, indeed, systematically differ from each other in their long-term country-level equilibria, corresponding to the country means. Government declarations contain stances that are substantially more right-wing (the average across the pooled data set is −0.22) than those that cabinet parties would have preferred, judging from their own electoral manifestos (the average is −7.97). This tendency holds within every country but two (France and Ireland). This empirical regularity may reflect the preoccupation of governments with ‘administrative concerns and ongoing matters of government which have not necessarily entered into the election campaign’ (Laver and Budge Reference Laver and Budge1992: 410–12). Might the economic conditions experienced by governments help explain this rightward shift? All the models displayed in Table 1 are clear in rejecting this explanation: declared cabinet position on the RILE dimension is insensitive to changes in inflation rate, unemployment rate and GDP since the end of the preceding government mandate.
Influences on Declared Cabinet Position on the Welfare Dimension
Do the political actors and external forces affecting the declared cabinet position on the RILE dimension behave in the same way on the welfare dimension? Table 2 answers this question.
Influences on Declared Cabinet Position on the Welfare Dimension
Notes: OLS estimations with panel corrected standard errors in parentheses. The dependent variable is declared cabinet position on the welfare dimension. Model M7 has 145 observations rather than 147 because the variable PM’s party – cabinet distance is missing for the governments Barre I and II in France. Indeed, Raymond Barre has never been affiliated to any party.
† p<0.10, * p<0.05, ** p<0.01, *** p<0.001.
The weighted mean position of governing parties determines the declared cabinet position on the welfare scale as well (Hypothesis 1). Indeed, Model M5 demonstrates that if the weighted mean position of coalition parties moves one point towards a more contractionary attitude, the position staked out by the government is likely to move in the same direction by 0.12 points. However, as for the RILE dimension, the coefficient referring to the weighted mean position of governing parties is far from 1. Moreover, once we look for political actors and external forces able to bias the declared cabinet position on the welfare dimension, the estimates tell a different story from the one discussed above.
Model M6 supports Hypothesis 2, demonstrating that the median party plays a crucial role. Indeed, a unitary increase in the variable median party – cabinet distance moves the declared cabinet position approximately 0.3 points towards a more contractionary attitude.
Model M7 shows that the welfare position of the PM’s party is unable to affect the dependent variable. Accordingly, Hypothesis 3 is rejected.
Things are different for the position of the party holding the labour and the social affairs ministers. Indeed, as displayed by model M8, this party is likely to bias the declared cabinet position towards its ideal point by approximately 0.007 points. However, the magnitude of this effect and the level of statistical significance, which is above conventional standards, cast some doubt on the acceptance of Hypothesis 4, which needs deeper investigation.
The systematic rightward shift in declared cabinet positions registered on the RILE dimension appears on the welfare dimension, as well: across the pooled data set, the country mean for declared cabinet position is equal to −7.146, while that for the cabinet weighted mean is equal to −11.719.
The models displayed in Table 2 test whether this shift can be explained by economic circumstances. The declared cabinet position on the welfare dimension seems to be affected by changes in the inflation rate since the end of the preceding government: a unitary increase in this variable makes the declared cabinet position move towards a welfare state retrenchment attitude by approximately 0.5 point. However, changes in the unemployment rate and GDP do not affect the dependent variable.
Simultaneous Influences on Declared Cabinet Position on the RILE and the Welfare Dimensions
The two previous subsections estimated the potential effects of each political actor on declared cabinet positions on the RILE and welfare dimensions one by one because the data set provides complete information on just 157 governments in 10 countries (see Table A1). However, to identify the net effect of each independent variable when all other influences are controlled for, Table 3 displays two additional models in which declared cabinet positions on the RILE and the welfare dimensions are regressed on the most important independent variables at once.
Simultaneous Influences on Declared Cabinet RILE and Welfare Positions
Notes: OLS estimations with panel corrected standard errors in parentheses. Models M9 and M10 have 145 observations rather than 147 because the variable PM’s party – cabinet distance is missing for the governments Barre I and II in France. Indeed, Raymond Barre has never been affiliated to any party.
† p<0.10, * p<0.05, ** p<0.01, *** p<0.001.
Overall, the results are consistent with the previous ones. First, when all the other influences are controlled for, the parliamentary median party confirms its inability to bias the declared cabinet position on the RILE dimension (M9), while it proves to be crucial on the welfare dimension (M10). Second, the PM’s party demonstrates the ability to bias declared cabinet position on the RILE dimension (M9), but not on the welfare dimension (M10). These findings demonstrate that the PM’s party exerts extra leverage on declared cabinet position on the RILE dimension, which is absent for the parliamentary median party on the same dimension. Accordingly, when these two key positions are occupied by the same party (see endnote 10), it is the fact of being the party of the PM, rather than the median legislator status, that grants this party its additional effect on declared cabinet position on the RILE dimension. On the welfare dimension, the mechanism is reversed: in cases of overlap between the party of the PM and the median legislator, it is the median legislator status that grants this party extra leverage on the declared cabinet position on the welfare dimension.
Third, these additional models shed light on Hypothesis 4. The position of the party holding the finance portfolio is unable to affect the declared cabinet position on the RILE dimension, while that of the party holding the labour and the social affairs ministers has a positive and statistically significant effect on the declared cabinet position on the welfare dimension.
Finally, a rising inflation rate from the end of the previous cabinet does not affect the declared cabinet position on the RILE dimension, but seems to have a statistically significant effect on the welfare dimension, shifting the declared cabinet position towards a more contractionary attitude.Footnote 11
DISCUSSION AND CONCLUSION
This article investigated several hypotheses concerning the main determinants of the positions adopted by coalition governments in 10 Western European parliamentary democracies since the end of the Second World War. As highlighted by the literature review, previous influential studies based only on the left–right dimension provided by the CMP coding of coalition government declarations and party manifestos provided mixed results.
This article contributes to the existing literature by adding a genuinely policy-based pro–anti welfare scale to the ideological RILE scale traditionally employed. This choice allows us to check whether the same actors are able to influence declared cabinet position on these two dimensions.
The results demonstrate that there is a connection between declared cabinet policy stances and the weighted mean positions of coalition parties as derived from their electoral manifestos, both on the RILE and on the welfare dimensions. Indeed, this connection appears in both Models M1 and M5, making the lack of fit between declared cabinet positions and coalition parties’ electoral promises lamented by Budge and Laver (Reference Budge and Laver1992b: 409–64) and McDonald and Budge (Reference McDonald and Budge2005: 145–8) appear overly pessimistic. Moreover, this first finding is consistent with the results obtained by Martin and Vanberg (Reference Martin and Vanberg2014) in the cases of Denmark, Germany and the Netherlands. However, the relationship between the declared cabinet position and the weighted mean position of coalition parties is far from the one-to-one correspondence needed to affirm that Gamson’s law (Reference Gamson1961) applies to policy payoffs as well as to portfolio allocations, making Warwick’s claims appear overly optimistic too (2001, 2011).
Notably, the results suggest that other political actors and external forces are able to bias declared cabinet positions. Moreover, such political actors and external forces do not behave in the same way on the two dimensions studied here. Starting with the RILE scale, the findings seem to support, at least in part, the central requirement of proposer models because the PM’s party exploits its position by extracting more than its proportional share of the policy payoff. Notice that the agenda-setting power of the PM’s party may also originate from the PM’s personal traits, such as charismatic leadership and communication style (see endnote 1). Moving to the welfare dimension, our results suggest that Lowi’s (Reference Lowi1972) motto, according to which ‘policies determine politics’, is partially true: the political actors shaping the declared cabinet position on the welfare dimension do not coincide with the ones on the RILE scale.
In particular, in addition to the weighted mean position of coalition parties, which continues to be a determinant of the declared cabinet position, the parliamentary median party and the competent ministers appear to be crucial. As expected, the distributive and redistributive nature of the programmes included in the welfare dimension, which see economic resources to be provided to specific societal segments, usually at the expense of other groups, grants additional leverage in decision-making processes to the party of the median legislator and the party of the labour and social affairs ministers.
The finding of a significant effect emanating from the parliamentary median party on the welfare dimension is intriguing. Indeed, this party is that of the PM in 43 per cent of the governments in our sample. Moreover, 51 per cent of the parties that occupy the median parliamentary position on the RILE dimension are also that of the median legislator on the welfare dimension. However, once we control for the influences emanating from the key political actors simultaneously (see Table 3), the median legislator status proves to be relevant in shaping declared cabinet position only on the welfare dimension (see also Table A6 in the Appendix). This result suggests that scholars focusing on the degree of congruence between government partisanship and policy outcomes in the welfare domain should seriously consider the policy positions of the party of the median legislator when dealing with the measurement of their independent variable. Indeed, this party is likely to employ its legislative median status to move the declared cabinet position towards its ideal point, thus potentially affecting subsequent policy outcomes in the welfare domain.
The finding of a significant effect emanating from the parties holding the labour and social affairs ministers are consistent with the ministerial autonomy assumption that underpins Laver and Shepsle’s (Reference Laver and Shepsle1996: 281–5) portfolio allocation model. Indeed, even if the results highlight that the declared cabinet position on the welfare dimension more closely reflects the policy positions of the entire cabinet, net of other factors, the parties of the labour and social affairs ministers do wield a degree of influence over the declared cabinet position on the welfare dimension. This finding is consistent with the evidence provided by Lucy Barnes (Reference Barnes2013), according to which the policy positions of the ministers responsible for welfare state programmes are strong determinants of the level of welfare generosity. Concerning external forces, the declared cabinet position on the welfare dimension tends to respond to a rising inflation rate experienced by the coalition parties when in office.
Finally, it is worth recognizing the most important limitation affecting the analysis performed here, namely, data scarcity on the explanandum (declared cabinet positions). This article assessed the degree of congruence between declared cabinet positions and the weighted mean positions of governing parties on the RILE and the welfare dimensions and proved that these two measures of government partisanship do not describe exactly the same phenomenon. As famously underlined by McDonald and Budge (Reference McDonald and Budge2005: 141), government declarations ‘should certainly be taken as better indicators of eventual policy than simple party electoral program intentions. They have the advantage of outlining a whole range of plans, for legislation and administration as well as spending.’
Accordingly, it seems extremely important and promising that future research carry on the coding procedure of government declarations conducted by the Manifesto Research Group from the aftermath of the Second World War to the mid-1990s. This choice will allow a comparison of government positions as derived from their declarations with those obtained by computing the weighted mean of coalition parties’ positions as derived from their electoral manifestos, thus exploiting a large amount of data already available online.
Appendix
The Coverage of the CMP Government Declarations Data
Descriptive Statistics
Pairwise Correlations among Declared Cabinet Position and Cabinet Weighted Mean on the RILE and Welfare Dimensions
Pairwise Correlations among Variables Referring to Political Actors on the RILE Dimension
Pairwise Correlations among Variables Referring to Political Actors on the Welfare Dimension
Influences on Declared Cabinet RILE and Welfare Positions Without Overlapping Cases
Acknowledgements
An early version of this article was discussed by Fabio Franchino, Robert Franzese and Vincenzo Galasso. I am grateful for their helpful suggestions. Moreover, Luigi Curini and Andrea Ceron were by my side while I was working on this article; their help has been invaluable. Finally, I sincerely thank the two anonymous reviewers: their detailed comments improved the quality of the article. The usual disclaimer applies.
