Introduction
Coalition government involves cooperation and delegation and therefore poses the risk for each coalition member of being deceived by the other coalition parties (Laver and Shepsle, Reference Laver and Shepsle1990; Strøm et al., Reference Strøm2010; Zbíral et al., Reference Zbíral2023). The risk of being deceived is amplified by the delegation structure of the coalition government (Müller, Reference Müller2000; Strøm et al., Reference Strøm2010): The coalition delegates the implementation of the coalition agreement to individual cabinet ministers, who have their own policy preferences and may not adhere to the coalition agreement (Thies, Reference Thies2001; Bäck and Persson, Reference Bäck and Persson2018). From a principal–agent perspective (Müller, Reference Müller2000), the principal has a variety of control mechanisms at their disposal to mitigate this risk. This is why coalition parties closely monitor their coalition partners using ex-ante control instruments such as written coalition agreements (Klüver and Bäck, Reference Klüver and Bäck2019; Gross and Krauss, Reference Gross and Krauss2021; Höhmann and Krauss, Reference Höhmann and Krauss2022) or ex-post control instruments such as parliamentary committees, junior watchdog ministers, or parliamentary questions (PQs) (Martin and Vanberg, Reference Martin and Vanberg2004; Martin and Vanberg, Reference Martin and Vanberg2005; Martin and Vanberg, Reference Martin and Vanberg2011; Zubek and Klüver, Reference Zubek and Klüver2015; André et al., Reference André2016).
So far, studies have mainly focused on how intra-coalition control works in majority coalitions in parliamentary regimes (Martin and Whitaker, Reference Martin and Whitaker2019; Höhmann and Sieberer, Reference Höhmann and Sieberer2020), where the responsibility for governing is fully in the hands of the coalition government or in presidential systems (Mimica et al., Reference Mimica2023), where the executive is in the hands of the directly elected president. However, the same control instruments are also used in mixed regimes (see Escobar-Lemmon and Taylor-Robinson, Reference Escobar-Lemmon and Taylor-Robinson2020), where there is a dual executive structure consisting of a Head of Executive (HoE), such as a president or mayor, and a government supported by the parliament (Duverger, Reference Duverger1980; Shugart and Carey, Reference Shugart and Carey1992).Footnote 1
A dual executive makes intra-coalition control more complex and complicated for two reasons. First, the coalition parties have to monitor not only each other but also the directly elected HoE. Second, if the HoE belongs to one of the coalition parties, which is often the case in mixed regimes (Samuels and Shugart, Reference Samuels and Shugart2010; Elgie and McMenamin, Reference Elgie and McMenamin2011), the balance of power within the coalition will be affected and may be tilted in favor of that party. To compensate for the power differential, the other coalition parties may increase their intra-coalition control efforts. Thus, the focus of this article is to test these assumptions and to examine the question: What are the drivers of intra-coalition control in mixed regimes and how does the dual executive structure affect the control behavior of coalition parties?
One of the most prominent and widely studied control mechanisms used by parliamentary parties to hold the government accountable are PQs (Russo and Wiberg, Reference Russo and Wiberg2010; Martin, Reference Martin2011; Otjes and Louwerse, Reference Otjes and Louwerse2018). For intra-coalitional control, recent research has shown that PQs are a suitable control tool for monitoring individual coalition partners (Martin and Whitaker, Reference Martin and Whitaker2019; Höhmann and Sieberer, Reference Höhmann and Sieberer2020; Höhmann and Krauss, Reference Höhmann and Krauss2022). The studies show that PQs are used in parliamentary regimes to close the information gap between the coalition government and an individual minister, allowing parties to monitor their coalition partner’s ministers. The results of these studies show that Members of Parliament (MPs) ask more questions to the ministries of the coalition partner (Höhmann and Sieberer, Reference Höhmann and Sieberer2020) and that issue salience and different policy positions between the coalition partners increase the number of PQs asked to a particular ministry (Martin and Whitaker, Reference Martin and Whitaker2019).
Traditionally, the study of presidential or mixed regimes has focused primarily on constitutional analysis, while the study of parliamentary regimes has primarily revolved around the dynamics of party politics (Doyle, Reference Doyle2020). This article contributes to the literature on executive–legislative relations in mixed systems by focusing on how aspects of party politics affect intra-coalition control. To do so, I analyze a novel dataset containing PQs of 21 German city councils in cities with more than 100,000 inhabitants between 2011 and 2020. I extend the dataset by adding coalition, portfolio, issue salience, and party position data. I chose to focus on the German local level because it resembles a mixed regime (Gross and Debus, Reference Gross and Debus2018). The mixed regime at the local level in Germany consists of a directly elected mayor – the HoE – and a coalition cabinet supported by the majority of the directly elected parliamentary actors in the form of the legislative councilors. Previous research has shown that coalitions in mixed regimes at the local level are similar to their counterparts at the national level and that local actors act similarly to national politicians (Debus and Gross, Reference Debus and Gross2016; Gross and Debus, Reference Gross and Debus2018).
This article provides generalizable insights into how PQs are used as an intra-coalition control mechanism when the control situation is more complex due to the dual executive consisting of a directly elected HoE and a parliamentary coalition cabinet. The results show that the dual executive structure affects intra-coalition control in mixed regimes. I find that, as in pure parliamentary regimes (Martin and Whitaker, Reference Martin and Whitaker2019; Höhmann and Sieberer, Reference Höhmann and Sieberer2020), policy divisiveness and issue salience are also driving factors in explaining intra-coalition behavior in mixed regimes. The number of PQs addressed to a given portfolio increases the more the holding party and the questioning party differ in their policy positions, and the more salient the topics under the jurisdiction of a portfolio are for the party addressing the PQs. Furthermore, when one of the coalition parties is affiliated with the directly elected HoE, the other coalition partners monitor the portfolios of the HoE-affiliated party more intensively and issue significantly more PQs for these portfolios.
Principal–agent theory and its perspective on president–cabinet relations and intra-coalition control
Principal–agent theory focuses on contracting in the private and public sectors. It deals with the principal–agent problem that arises when a person or entity (the agent) acts on behalf of another entity or person (the principal) and focuses on the potential conflict of interest between the agent and the principal (Laffont, Reference Laffont2003). The potential for conflict is driven by asymmetric information and divergent interests between agents and principals. An agent is more likely to pursue their interests in circumstances where they have more information than the principal and can use this asymmetric information situation to their advantage. When an agent and a principal have divergent preferences, an agent’s pursuit of their own interests leads to an agency loss for the principal. However, the principal is not powerless and can monitor the agent to overcome asymmetric information to mitigate their disadvantage to the agent and to avoid agency loss. Thus, the principal–agent theory is useful for assessing the dynamics of political control (Laffont, Reference Laffont2003; Lane, Reference Lane2008).
Politics consists of a variety of principal–agent relationships. For example, the coalition parties are the principals of the cabinet, the cabinet is the principal of each minister, and a minister is the principal of their subordinate bureaucrats (Müller, Reference Müller2000). Thus, it is possible to identify chains of principal–agent relationships among political actors. In a democratic context, the ultimate principal is the people, and all political actors are its agents (Lane, Reference Lane2008).
To better understand intra-coalition control in mixed regimes, two branches of the principal–agent theory are particularly useful. Thus, I first focus on how principal–agent theory conceptualizes president–cabinet relations in presidential and mixed regimes. I then turn to the principal–agent theory’s conceptualization of intra-coalition control in parliamentary regimes.
In mixed and presidential regimes, the people directly legitimize two political actors: the president and the legislature. Elgie (Reference Elgie2020b) argues that, from the perspective of principal–agent theory, the basic organization of president–cabinet relations depends on whether the regime is a presidential regime or one of the two subtypes of mixed regimes, premier-presidentialism or president-parliamentarism, introduced by Shugart and Carey (Reference Shugart and Carey1992). The principal–agent perspective shows that the balance of power between the president and the cabinet differs significantly between regimes (Elgie, Reference Elgie2020b): While in premier-presidentialism, the cabinet does not depend on the president to stay or get into office, in both president-parliamentarism and premier-presidentialism, the cabinet depends at least in part on the president to be appointed and to survive.
Another branch of the principal–agent theory conceptualizes how intra-coalition control functions in parliamentary regimes. In a coalition government, the coalition delegates the implementation of the coalition agreement to individual cabinet ministers who have their own policy preferences (Müller, Reference Müller2000; Thies, Reference Thies2001; Bäck and Persson, Reference Bäck and Persson2018). This increases the risk of conflict as the individual ministers may decide to follow their own ambitions and deviate from the coalition agreement during policy implementation (Laver and Shepsle, Reference Laver and Shepsle1990). Therefore, this branch of principal–agent theory focuses on individual ministers to conceptualize intra-coalition control. From the principal–agent perspective, a minister is the agent of the coalition government and acts on its behalf. In parliamentary regimes, ministers are powerful because they are responsible for implementing and initiating policies within their portfolios (Laver and Shepsle, Reference Laver and Shepsle1994). To control its agent, the coalition government monitors the minister to ensure stability (Strøm et al., Reference Strøm2010). In terms of principal–agent theory, monitoring the agent prevents the risk of agency loss because it helps the principal overcome the informational disadvantage of not knowing precisely what the agent is doing. Thus, it is particularly important for coalition parties to monitor their partners’ ministers to ensure that individual ministers do not deviate from the negotiated coalition consensus.
PQs as a general control instrument and how they are useful for intra-coalition control
PQs are a widespread legislative tool used in various parliaments (Russo and Wiberg, Reference Russo and Wiberg2010). Opposition parties commonly use them as an oversight and control tool to hold the government accountable (Martin, Reference Martin2011; Otjes and Louwerse, Reference Otjes and Louwerse2018; Mimica et al., Reference Mimica2023). Otjes et al. (Reference Otjes2022) showed in their analysis of 53 Dutch municipalities that this pattern can also be found at the local level. In addition, PQs are also commonly used in semi-presidential regimes. Borghetto et al. (Reference Borghetto2020) used PQs to study geographic representation in Portugal. Hayward (Reference Hayward2004) focused on France and found that PQs are especially used by the French opposition as a monitoring tool. Jenny and Müller (Reference Jenny and Müller2001) studied the Austrian case, and their results show that opposition parties use PQs more often than government parties in Austria. The authors also analyzed the purpose of a PQ and found that opposition parties are more likely to use PQs to attack and criticize the government, while government parties are more likely to use PQs to highlight their successes or to emphasize election constituency interests. Kukec (Reference Kukec2022) finds that parties use PQs to give their ministers a stage for credit-taking. Their PQs contain a more positive sentiment than the PQs of other parties. The minister’s replies to the PQ are also more positive than to other parties and entail more numerical evidence to argue in favor of future or previous government actions. So, the results of Kukec (Reference Kukec2022) and Jenny and Müller (Reference Jenny and Müller2001) show that PQs are also used by government parties to give their own ministers a platform to present and explain policy decisions of the government.
PQs are also used as an intra-coalition control tool to obtain information from a government ministry (Höhmann and Sieberer, Reference Höhmann and Sieberer2020). PQs are predestined to close the information gap between the principal (coalition government) and the agent (individual minister) (Strøm et al., Reference Strøm2010). Suppose legislators are concerned that a minister from the coalition party may deviate from the coalition agreement in proposed policy legislation or executive policy implementation. In this case, PQs signal the minister that they are monitored, and the party raising the PQ obtains information and clarification on the respective issue (Martin and Whitaker, Reference Martin and Whitaker2019).
So far, only a few studies have focused primarily on PQs used as an intra-coalition control tool. Mimica et al. (Reference Mimica2023) focused on a presidential regime and analyzed for Chilé whether party members of the president’s party generally use fewer PQs than their coalition partners. However, they find no conclusive evidence in their models that this is the case. All other studies on intra-coalition control and PQs have focused on parliamentary regimes. Martin and Whitaker (Reference Martin and Whitaker2019) show that MPs in the British House of Commons ask more PQs to a ministry held by the coalition partner, which is responsible for policy areas where the coalition partners differ in their policy positions. The greater the policy divergence between the coalition partners, the more PQs are asked. For the German Bundestag, Höhmann and Sieberer (Reference Höhmann and Sieberer2020) find that parties intensively question ministers holding a portfolio with high ideological differences between the partners and high electoral salience for the party not in charge of the respective ministry. In addition, the results of Höhmann and Krauss (Reference Höhmann and Krauss2022) show for the national level in Germany that the more detailed the coalition agreement is on a particular policy issue, the more PQs are directed at the ministry in charge of that particular policy issue.
How do parties use PQs for intra-coalition control in mixed regimes?
The empirical results of previous research and the concepts of principal–agent theory illustrate that PQs are useful intra-coalition control tools. However, so far, scholars have focused exclusively on cases within parliamentary regimes or presidential regimes. Thus, regimes with a dual executive structure are understudied. Therefore, in this section, I bring together the different branches of principal–agent theory to make it applicable to the more complex intra-coalition control situation in mixed regimes and formulate hypotheses to test whether PQs are used as an intra-coalition control instrument.
From the perspective of principal–agent theory, PQs are a useful control tool for coalition parties to counteract the risk of agency loss due to the delegation of policymaking to individual ministers (Martin and Whitaker, Reference Martin and Whitaker2019). PQs can be used to obtain information, identify agency drift, and correct agency drift (Höhmann and Sieberer, Reference Höhmann and Sieberer2020). Therefore, it is rational for a coalition party to monitor the ministers of the coalition partners. The extent to which a party controls its coalition partners depends on how costly agency loss in a particular policy area would be for the party. The greater the risk of conflict between the coalition party holding the portfolio and the controlling party, the greater the agency loss.
Although intra-coalition control is more complex in mixed regimes than in pure parliamentary systems, portfolios are also held by parties. Therefore, the principal–agent theory argument that a coalition party should monitor, especially the ministers of the other coalition parties, to ensure that those agents stick to the coalition consensus, should also be the case in mixed regimes.
The literature on mutual control in coalition governments has shown that, empirically, two factors are the main drivers for parties asking PQs to monitor the behavior of coalition partners (Martin and Whitaker, Reference Martin and Whitaker2019; Höhmann and Sieberer, Reference Höhmann and Sieberer2020). The first decisive factor is the ideological policy distance between the coalition partners, which captures the expected ministerial drift in policy terms. Since the studies of Höhmann and Sieberer (Reference Höhmann and Sieberer2020) and Martin and Whitaker (Reference Martin and Whitaker2019) empirically show that the ideological policy distance between coalition partners in parliamentary regimes plays an important role in predicting how many PQs are addressed to a particular ministry, I test whether this is also the case in a mixed regime.
Hypothesis 1: The greater the ideological policy distance between the cabinet party asking the question and the cabinet party holding the portfolio, the more PQs a party addresses to the portfolio of the respective coalition partner.
The second factor is how salient the issues within a ministry’s jurisdiction are to the controlling party. Because different policy areas are differently salient to parties, ministerial drift is particularly costly to parties when the policy areas under a portfolio’s jurisdiction are highly salient to the controlling coalition party. In these cases, ministerial drift could have high electoral costs for the controlling party or harm important policy interests. As found by Martin and Whitaker (Reference Martin and Whitaker2019), I expect parties to address more questions to portfolios when the policy areas under the portfolio’s jurisdiction are more salient to the questioning party. Thus, I test whether issue salience is a driving factor for intra-coalition control in a mixed regime.
Hypothesis 2: The more salient the portfolio’s area of jurisdiction is for a cabinet party not holding the portfolio, the more PQs a party addresses to the portfolio.
In a mixed regime, all legislators must focus their control efforts on both parts of the executive. They must control the directly elected HoE and the cabinet. Alemán and Tsebelis (Reference Alemán and Tsebelis2011) find that strong and independent legislators use oversight tools to monitor the directly elected HoE. Thus, as long as the HoE is independent of all cabinet parties, intra-coalition control in mixed regimes would be fully covered by Hypothesis I. In reality, however, the directly elected HoE is often aligned with a political party through party membership in mixed regimes (Samuels and Shugart, Reference Samuels and Shugart2010; Elgie and McMenamin, Reference Elgie and McMenamin2011).Footnote 2 Thus, it is possible for the HoE to be affiliated with one of the coalition partners.Footnote 3 In such cases, the dual executive is partially intermingled.
As modern government becomes more complex, executive agents must gather and analyze information on a large scale (Lane, Reference Lane2008). Having more or better information gives political actors an advantage over their competitors. Thus, the ability to combine information with the HoE gives the coalition party affiliated with the HoE an advantage over the other coalition parties and leads to information asymmetry between the coalition partners, as one of the partners now has more information due to its affiliation with the executive head. Thus, the asymmetric information game gets skewed in favor of the affiliated coalition partner. From a principal–agent perspective, the remaining coalition partners are now forced to try to close the information gap to minimize their disadvantage.
To do so, they must increase their control efforts over the coalition partner affiliated with the HoE. This increase in control is also in the interest of the ultimate principal, the people, for two reasons. First, the collaboration between the HoE and their affiliated party also increases the information asymmetry between the ultimate principal and those agents in favor of the latter. Thus, reducing the information asymmetry is also in the interest of the people. Second, a concentration of power in a certain group of agents due to affiliation could lead to collusion among these agents and tempt them to pursue their own interests rather than the interests of their ultimate principal. From the perspective of principal–agent theory, it can, therefore, be assumed that the other coalition partners, as rational actors, will monitor the portfolios of the affiliated party more closely to compensate for the difference in power – irrespective of the respective ideological policy distance between the parties. Thus, my third hypothesis is:
Hypothesis 3: If one of the coalition parties is affiliated with the directly elected HoE, the other coalition parties address more PQs to all portfolios held by the aligned party.
Data and methods
As a starting point for empirically testing my hypotheses and studying mixed regimes and intra-coalition control, I focus on the mixed regime subtype of premier-presidentialism. The German local level is a mixed regime (Gross and Debus, Reference Gross and Debus2018) with a dual executive consisting of a directly elected mayor (HoE) who is either an independent or a member of one of the political parties represented in the city’s council and a cabinet supported by the council majority.Footnote 4 Below, I provide more information on the mayor as the HoE of the mixed regime at the German local level and on the cabinet and illustrate why they are the local equivalent of a president and a cabinet in state politics.
The mayor is the most powerful and prominent individual political actor at the local level as they are the head of the local administration (Egner, Reference Egner2015) and, in most German states, the chair of the municipal council and various committees. As the head of the administration, the municipal council, and various committees, the mayor also has access to all kinds of information, which makes the mayor a particularly well-informed political actor at the local level. Thus, as the head of the city, the mayor is comparable to other HoEs, such as presidents who serve as the head of state (Gross and Debus, Reference Gross and Debus2018).
At the German local level, ministries are called departments, and the head of a department is a deputy mayor. Department heads are the functional equivalent of ministers at the state and national levels (Egner, Reference Egner2015; Debus and Gross, Reference Debus and Gross2016; Gross and Debus, Reference Gross and Debus2018). Together with the mayor, the department heads are referred to as the city government (Egner, Reference Egner2015). According to Debus and Gross (Reference Debus and Gross2016), these city governments have extensive rights of self-government and can thus shape local policy. In the German multilevel system, administrative functions are largely reassigned to the state and local levels (Gross and Krauss, Reference Gross and Krauss2021). Thus, according to Gross and Krauss (Reference Gross and Krauss2021), politicians at the local level are de facto responsible for dealing with issues that de jure fall within the realm of federal decision-making. Moreover, the results of Egner’s (Reference Egner2015) survey of 900 German city councilors show that 92% of the councilors consider the mayor to be an important or very important political actor at the local level, and the portfolio heads were considered to be important or very important actors by 68% of all councilors. In addition, the surveyed councilors stated that monitoring the executive is an important task for a city councilor. Overall, these results reflect the strong position of the mayor at the local level in Germany and show that portfolio holders are important political actors who should be monitored to avoid the risk of policy drift.
Portfolio allocation in German cities is based on partisan considerations, despite the legal requirement for public advertising (Ecker and Gross, Reference Ecker and Gross2023). The analysis of 69 German cities by Ecker and Gross (Reference Ecker and Gross2023) shows that the distribution of seats in the local council influences the selection of department heads, and parties tend to choose allocations that reflect their electoral strength. Coalition agreements also play a crucial role in the allocation process (Ecker and Gross, Reference Ecker and Gross2023): Despite the substantial share of non-partisan department heads, coalition agreements have a significant impact on allocation decisions, and coalition parties are much more likely to receive portfolio allocations. Overall, portfolio allocation at the local level in Germany functions similarly to portfolio allocation at higher political levels, following a partisan logic and prioritizing proportionality.
I chose to focus on German major cities with more than 100,000 inhabitants for multiple reasons. First, studies show that in Germany, mayors of populous municipalities, such as German cities, are often members of one of the parties in the city council (Pollex et al., Reference Pollex2021; Nyhuis et al., Reference Nyhuis2022). Thus, the German local level is an ideal case to test my third hypothesis. Second, major German cities perform both municipal (Gemeindeaufgaben) and district functions (Kreisaufgaben) because they are so-called independent cities (kreisfreie Städte), which are equivalent to a district. Third, PQs are an available legislative instrument in all councils, and PQs are commonly used by local parties in Germany. Fourth, the case selection allows me to compare several coalitions influenced by the same political cultures, under similar institutional settings, and during nearly the same time period. Fifth, coalitions at the local level in mixed regimes function similarly to their counterparts at the national level (Debus and Gross, Reference Debus and Gross2016; Gross and Debus, Reference Gross and Debus2018).
I use a novel dataset from the German local level, which covers all PQs of one legislative period in German city councils of major cities between 2011 and 2020 (for a detailed description, see Gross et al. (Reference Gross2023)). PQs were collected using web scraping and manually edited where necessary. The dataset includes the dominant policy issue for each PQ according to the Comparative Agenda Project (CAP) coding scheme (Baumgartner et al., Reference Baumgartner, Weible and Sabatier2018). Labeling was performed using a semi-automated supervised classification approach (for details, see Gross et al. (Reference Gross2023) and Velimsky et al. (Reference Velimsky2023)).
I use a subset of this dataset of 21 cities consisting of all councils with a coalition cabinet (see online Appendix B-1 for details). I focus on all parties that were coalition members – including five of the six largest parties in Germany that are represented in the German Bundestag and three smaller parties.Footnote 5 The dataset consists of 3,342 written PQs.Footnote 6 I use this data to create the dataset used in the main analysis of this article.
I created a new dataset consisting of all portfolios that were in place during the legislative period of the 21 cities covered in the PQ dataset. During a legislative period, it is possible that a portfolio is newly created or that a portfolio’s area of responsibility is rearranged. It is also possible that the head of the portfolio changed during a legislative period. I consider each newly created portfolio, each rearranged portfolio, or each portfolio where the head has changed as an individual case. Since, at the local level, PQs are not addressed directly to a portfolio, I determined, based on the portfolio’s area of responsibility, which CAP policy issues are under the jurisdiction of the respective portfolio. Overall, the dataset entails the name of the portfolio, the head of the portfolio, the party affiliation of the head, the date the portfolio started working, the date the portfolio stopped being in place, the duration of the portfolio being active, and its areas of responsibility based on the CAP policy coding scheme. The portfolio dataset consists of 112 unique portfolios (see online Appendix B-2 for details).
The unit of interest – as it is common in intra-coalition control research (Martin and Whitaker, Reference Martin and Whitaker2019; Höhmann and Sieberer, Reference Höhmann and Sieberer2020) – are all party-portfolio combinations where the party raising PQs is not in control of the respective portfolio. The main dataset consists of 165 such party-portfolio combinations. Focusing only on party portfolio combinations where the party addressing a portfolio is not the holder of the portfolio ensures that the models capture control behavior and not credit-taking since the latter occurs more than when a governing party is addressing their own portfolios (Kukec, Reference Kukec2022).
The covered 21 cities entail a variety of different coalition constellations and, therefore, show sufficient intra-cabinet policy conflict variation. Focusing on German major cities has the advantage that the city councils are highly professional and work similarly to the parliaments at higher levels (Egner, Reference Egner2015; Gross, Reference Gross2023). Moreover, while parties at the local level in Germany are embedded in national party organizations, they are free to draft their own electoral manifestos (Gross and Jankowski, Reference Gross and Jankowski2020) and to directly influence policy outcomes in their cities with their decisions (Goerres and Tepe, Reference Goerres and Tepe2013) without the intervention of their national party organizations.
To test my hypotheses, I use a multilevel negative-binomial model. I opted for a negative-binomial model since the dependent variable is overdispersive. I use a two-level setup with hierarchically nested random intercepts (parties nested in cities). Thus, the intercept of the model is based on each city-party combination. I do this to control for city differences in PQ use and the possibility that certain parties in some cities generally ask more PQs than others.
Dependent variable and explanatory variables
My dependent variable is the number of written PQs submitted during a legislative period by a party that addresses a policy issue of a portfolio controlled by a coalition partner. I retrieved this data by combining the PQ dataset and the portfolio dataset. I created a count variable capturing the sum of PQs a party addressed on a policy issue under the jurisdiction of a portfolio.
My first explanatory variable captures whether the portfolio holding party is affiliated with the directly elected mayor using the dummy variable Mayor from portfolio party (0 = portfolio not held by the party affiliated with the mayor; 1 = portfolio held by the party affiliated with the mayor). I use data from Debus and Gross (Reference Debus and Gross2016) for this variable. I have added to this data any mayoral changes during the covered legislative periods, based on information from the respective city websites. Since the data structure is based on the council terms rather than mayoral terms, in the few cases where the mayor’s party changed during the legislative period, I coded this variable based on the party that was affiliated with the mayor for the longest period of time during the legislative period.
The second explanatory variable is the Issue salience of a portfolio’s jurisdictional policy areas to the party asking the question. To capture issue salience, I use data from Gross et al. (Reference Gross2023), who automatically determined the 200 most important keywords per policy issue based on their labeled PQ dataset for the 15 policy issues based on the CAP coding scheme.Footnote 7 They used the keywords per topic as separate dictionaries and determined how often these keywords were used in a manifesto. They calculated a party’s salience of a particular policy issue in its manifestos by dividing the number of keyword occurrences by the total number of words in a manifesto and z-transformed the variable. I determine a party’s issue salience for a given portfolio by taking the average of all policy areas under the portfolio’s control. For example, if a portfolio is responsible for transportation and community development, I add the issue salience values of a party for both topics and divide the sum by two. This measure is particularly appropriate for this article because it is a relational measure and, therefore, captures well the extent to which the salience differs between parties across policy issues.
My third explanatory variable is the Ideological distance between the asking party and the party holding the portfolio. To determine the positions of each party, I use an expanded dataset from Gross and Jankowski (Reference Gross and Jankowski2020) covering various dimensions of political conflict and party positions at the German local level. The authors created those measures by determining via the Chapel Hill Expert Survey the position of a party for a specific policy/ideological dimension. They used party manifestos from the national level as reference texts for wordscore calibration. They then used the local manifestos to determine via wordscores (Laver et al., Reference Laver2003) the concrete value of a local party for each policy/ideological dimension in their dataset. To measure the position of a coalition party in a specific policy area as accurately as possible, I assigned to each portfolio the dimension that captures the portfolio’s area of responsibility the best (see Appendix B-2). The ideological distance between the asking party and the portfolio holding party is then calculated as the absolute value of the difference between the position score of the holder party and the party asking the PQ. For example, suppose the Green party is a coalition member in charge of the environment portfolio, and the Left party is a coalition partner. In that case, I use the absolute difference between the position score of the Greens and the Left party for the environmental dimension of the positional dataset. Moreover, by including the Mayor from portfolio variable in the model, I ensure that the effect captured by the variable Ideological distance is measured accurately and is not biased by the effect of PQs being used to monitor the head of the executive (Otjes and Louwerse, Reference Otjes and Louwerse2018).
Control variables
I control for several factors that previous research has shown may play a role in explaining the number of written PQs a party addresses to a coalition partner’s portfolio. Since party size affects how many resources a party has to ask questions (Otjes and Louwerse, Reference Otjes and Louwerse2018), I use the variable Seats in council (log) to control for that. The variable is measured as the log of the number of local council seats held. The data used for this variable comes from Gross and Debus (Reference Gross and Debus2018). I also control for the total number of questions a portfolio received to capture issue salience differences between portfolios (Green-Pedersen and Mortensen, Reference Green-Pedersen and Mortensen2010; Vliegenthart and Walgrave, Reference Vliegenthart and Walgrave2011; Otjes and Louwerse, Reference Otjes and Louwerse2018).
The dummy variable Multiparty captures whether the coalition of the asking party consists of two (0 = two-party coalition) or more than two parties (1 = multiparty coalition). The variable is used to check if there is a substantial difference between multiparty coalitions and two-party coalitions in their intra-coalition control behavior using PQs. The dummy variable Same coalition accounts for cabinets with parties from the previous cabinet. Research shows that in such coalitions, partners have built up some trust during their previous cooperation and, therefore, control their partner to a lesser extent (Thies, Reference Thies2001; Höhmann and Sieberer, Reference Höhmann and Sieberer2020). Finally, since portfolio heads and portfolio policy allocation have changed for some portfolios, I control for the duration a portfolio has been in place.
Empirical results
In this section, I present the analysis of my hypotheses. First, I show descriptively how many written PQs coalition parties use on average to a portfolio and to whom they address the PQs. I then test my hypotheses using multivariate analysis.
Descriptive analysis
Since PQs on the local level have not been studied extensively, I first show descriptively to whom coalition parties direct their questions. In the mixed regime at the local level in Germany, portfolios can be held by one of the coalition parties, by an independent head of the portfolio, or by a member of a party that is not a member of the coalition. Therefore, the use of PQs at the local level in Germany is more complex than at the national level.
Independent portfolio heads are quite common at the local level in Germany (Ecker and Gross, Reference Ecker and Gross2023).Footnote 8 Of the 180 total portfolios, the majority of portfolios are headed by a member of one of the coalition parties (112; 62%), 26 (14%) are headed by a party member of a party that is not a member of the coalition, and 42 (23%) are headed by an independent.
Overall, the dataset consists of roughly the same number of portfolios per policy area between 33 and 40 portfolios (see Appendix B-4). In 13 of the 15 policy areas, coalition parties hold the majority of portfolios. In these policy areas, coalition parties hold between 57% and 73% of all portfolios in a policy area. For the policy areas of transportation and community development, coalition parties hold 43% of the respective portfolios and independents hold 49% of all transportation portfolios and 48% of all community development portfolios. Since both of these policy areas are at the local level policy areas that require specific expertise, it makes sense that independents are more common here than in other policy areas, given their technocratic background and the fact that independents leading these portfolios at the local level are often experts in urban development. Independents hold between 12 and 32% of all portfolios in the other 13 policy areas. Non-coalition parties hold between 6 and 24% of all portfolios in a policy area.Footnote 9
The dataset consists of 165 party-portfolio combinations in which a coalition party addresses PQs to a portfolio of a coalition partner, 112 party-portfolio combinations in which a coalition party addresses PQs to one of its own portfolios, 75 party-portfolio combinations in which a party addresses PQs to a portfolio held by a non-coalition party, and 107 party-portfolio combinations in which a coalition party addresses PQs to an independent portfolio. So, in the majority of party-portfolio combinations, coalition parties use PQs to control portfolios held by the coalition cabinet.
Figure 1 shows the average number of PQs used by coalition parties to address a particular portfolio type. Coalition parties use the most PQs to address portfolios held by an independent or a member of another coalition party. Coalition parties also use PQs to address one of their own portfolios, however, to a lesser extent than they focus on the portfolios of their coalition partners. The results indicate that PQs play an important role as an intra-coalition control instrument in the mixed regime of the German local level. Since the focus of this article is on how PQs are used as an intra-coalition tool, I focus from here on all PQs that address a portfolio held by a coalition partner or held by one’s own party and do not focus on independent portfolios or portfolios held by a non-coalition party.
The average number of PQs per portfolio by all cabinets.
Notes: Bars indicate the average number of PQs that all coalition parties combined on average addressed to a portfolio led by their coalition partner, a head of their own party, a portfolio holder of a party that is not in the coalition, or to a portfolio holder without party alignment. Error bars show the confidence interval (P < 0.05).
Figure 2 shows that, on average, coalition parties not affiliated with the mayor’s party address slightly more PQs to a portfolio held by the coalition partner than do parties affiliated with the mayor. I find the same pattern for PQs addressed to their own portfolios. Here, the difference between parties that are affiliated with the mayor and those unaffiliated is even larger. The results suggest that the use of PQs to control coalition partners may seem to be more important for parties that are not aligned with the mayor. Thus, these findings provide descriptive tentative support for the third hypothesis.
Usage of PQs by parties that are affiliated with the mayor vs. parties that are not affiliated with the mayor.
Note: Bars indicate the average number of PQs of either all parties aligned or vice versa not aligned with the mayor addressed to a portfolio led by their coalition partner or a head of their own party. Error bars show the confidence interval (P < 0.05).
Multivariate analysis
To test my hypotheses about how ideological conflict and issue salience affect the extent to which governing parties use PQs to monitor their coalition partners, I turn to multivariate analysis.Footnote 10
In Table 1, Model 1 shows the results focusing only on the coalition government and not taking into account the dual executive structure between the coalition government and the directly elected mayor. I find empirical support for the first and second hypotheses. The results show that the ideological distance between the party asking a PQ and the portfolio holding the respective portfolio significantly impacts how many PQs a party asks toward a particular portfolio. The higher the ideological difference, the more PQs are asked. The respective incidence rate ratio (IRR) shows that an increase of 1 of the ideological difference between the party asking a PQ and the coalition partner holding the respective portfolio leads to an increase in the number of PQs this party addresses to this particular portfolio by the factor of 1.66. Thus, an increase of 1 in the ideological difference between the party asking a PQ and the coalition partner holding the respective portfolio leads to a 66% increase in the number of PQs. In addition, the variable issue salience is also significantly positive. The higher the controlling party’s issue salience of the portfolio’s area of jurisdiction is, the more PQs are addressed. Based on the IRR, an increase of 1 in the issue salience increases the number of PQs a party addresses to a portfolio by 22%.
Explaining the number and share of PQs used to control the coalition partner
Notes: Multi-level negative binomial regression models with random intercepts (parties are nested in cities). Dependent variable: total number of PQs per portfolio asked by a coalition party in local city councils. Standard errors of Log-Mean coefficients are in parentheses. IRR = incidence rate ratio. Total no. of PQs to a portfolio and the duration of a portfolio are measured in increments of 100. Significance levels: *P < 0.05; **P < 0.01; ***P < 0.001.
Model 2 shows the results accounting for the institutional situation of a mixed regime due to the additional executive actor in the form of the directly elected mayor. Model 2 shows a positive significant effect for the variable capturing whether parties not affiliated with the mayor address more PQs to portfolios held by the coalition partner who is affiliated with the mayor. These findings are in line with the third hypothesis and show that parties control all portfolios of the coalition partner who is affiliated with the mayor more intensively.
Furthermore, the model shows that the effects of the variables ideological distance and issue salience hold even if the effect of shared governing responsibilities is accounted for by adding the mayor variable. I also find support in Model 2 for the first and second hypotheses. The ideological difference between the party asking a question and the party in charge of the receiving portfolio significantly affects the number of PQs a party addresses to a portfolio held by a coalition partner. The respective IRR displays that an increase of 1 in the ideological difference between the party asking a PQ and the coalition partner holding the respective portfolio leads to an increase in the number of PQs the party addresses to this particular portfolio by the factor of around 1.5. In addition, the variable issue salience is also significantly positive. The higher the controlling party’s issue salience of the portfolio’s area of jurisdiction is, the more PQs are addressed. Based on the IRR, an increase of 1 in the issue salience increases the number of PQs a party addresses to a portfolio by 18%.
Conclusion and discussion
Previous research focusing on intra-coalition control has shown that cabinet parties use PQs to monitor their coalition partners (Martin and Whitaker, Reference Martin and Whitaker2019; Höhmann and Sieberer, Reference Höhmann and Sieberer2020; Mimica et al., Reference Mimica2023). However, these studies focused only on parliamentary and presidential regimes and not on mixed regimes. In this article, I provide one of the first analyses of intra-coalition control in a mixed regime with a dual executive structure between a coalition cabinet and a directly elected head of the executive. Due to the dual executive, the intra-coalition control situation is more complex than in pure parliamentarism or pure presidentialism.
Focusing on the mixed regime at the local level in Germany, I find that the ideological policy distance and the issue salience drive intra-coalition control. Coalition parties address more PQs to the portfolios of their coalition partners when the policy distance between the two on the respective policy issue is greater and when the portfolio’s area of jurisdiction is more salient to the controlling party – as is also the case in parliamentary regimes (Martin and Whitaker, Reference Martin and Whitaker2019; Höhmann and Sieberer, Reference Höhmann and Sieberer2020). In addition, my results show that intra-coalition control increases when the balance of power between the coalition parties is affected by the directly elected mayor being a member of one of those parties. In these cases, the coalition partners that are not aligned with the mayor address more PQs to all portfolios held by the party with the mayor.
Traditionally, scholars of presidential or mixed regimes have focused more on a constitutional approach, while scholars of parliamentary regimes have used approaches that focus more on the dynamics of party politics (Doyle, Reference Doyle2020). On the one hand, the results of this article show that coalition parties in mixed regimes use legislative monitoring tools similar to their counterparts in parliamentary regimes to keep tabs on their coalition partners. It shows that focusing on aspects of party politics in mixed regimes can lead to new insights and deepen our understanding of the functioning of parliaments in general. On the other hand, this article contributes to our knowledge of executive–legislative relations in mixed systems by showing that the dual executive structure in a mixed system affects intra-coalition control behavior and that parties take the directly elected HoE into account.
This study represents an important step in understanding intra-coalition control mechanisms in mixed regimes. Since previous research has shown that local-level coalitions in mixed regimes function similarly to their national-level counterparts and that coalition actors act similarly across political levels (Debus and Gross, Reference Debus and Gross2016; Gross and Debus, Reference Gross and Debus2018), these findings also have broader implications. However, more research is needed to fully understand the dynamics of intra-coalition control in mixed regimes and its implications for executive–legislative relations. This study focused on local data. Although the mixed regime at the local level in Germany functions similarly to mixed regimes at higher political levels, the local level differs from higher levels in terms of media attention and the issue areas in which politicians have the decision-making competencies. Compared to national politics, the content of PQs is less likely to be picked up by the media, so parties have limited opportunities to use PQs to signal to voters that they are acting in the public’s interest. Thus, PQs may be even more of a primary control instrument at the local level than at higher political levels. Moreover, the issue areas where local politicians have broad competence differ from those at the national level. Politicians at the local level focus much more on community development, transportation, and domestic commerce issues, and to a lesser extent on issues that are important at higher levels, such as macroeconomics, the environment, or education. Thus, to ensure that the results found in this article hold, future research should test whether the same drivers of intra-coalition control found in this article also apply to mixed regimes at higher political levels.
Focusing on the national level would also provide an opportunity to study the heterogeneity of mixed regimes. Different cases of mixed regimes differ in whether the HoE is also a principal of the coalition cabinet or whether the cabinet depends only on the support of the majority of the legislature to survive. In particular, it would be interesting to see whether mixed regimes of the president-parliamentarism subtype, such as Peru, Iceland, or Niger (Elgie and McMenamin, Reference Elgie and McMenamin2011), where the cabinet is never entirely independent of the directly elected president, affect the dynamics of intra-coalition control differently than in the mixed regimes of the premier-presidential subtype studied in this article.
Since mixed regimes use the same legislative control instruments as parliamentary and presidential regimes, new insights gained for mixed regimes could also be useful for research on parliamentary or presidential regimes. For example, research has shown that the prime minister is becoming stronger in some parliamentary regimes and thus increasingly resembles a HoE from mixed regimes (Bäck et al., Reference Bäck2009). Thus, it is possible that the more complex control logic due to the HoE is also present in these parliamentary systems since the prime minister’s party is also in a privileged position of power and information, as is a party affiliated with the HoE in a mixed regime. Overall, such research offers an opportunity to connect different branches of the literature better and to understand how different regime types are similar or different and could help deepen our general understanding of how parliaments and executive–legislative relations work.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/S1755773923000322.
Acknowledgements
I would like to thank Martin Gross, Carsten Schwemmer, Dominic Nyhuis, Philipp Köker, and Morten Harmening for their valuable suggestions and helpful feedback. I would also like to thank Louis Drexler, Kolja Hiebl, Hanna Hieronymus, Aaron Reudenbach, Annica Missy, Janina Schindler, Sinéad Thielen, and Katia Werkmeister for excellent research assistance.
Funding statement
Funding for this project was provided by the German Research Foundation through grant numbers GR 5526/1-1 and NY 123/1-1.
Competing interests
The author declares no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Open access
