Controlling Uncertainty in Coalition Governments

Abstract

Multiparty governments are based on delegation and compromises but, at the same time, coalition parties have at their disposal several legislative instruments to keep tabs on their partners. Whereas previous studies focused on policy divisiveness and issue salience as main factors able to explain parliamentary scrutiny, in this article we suggest uncertainty as a complementary factor. In particular, we theorize that the use of parliamentary questions (PQs) is a function not only of policy characteristics but also of actors involved in coalition governance. When ministers increase intra-coalition uncertainty, cabinet parties use PQs to extract information from ministers and to reduce uncertainty in policy implementation. Statistical analyses of all written and oral parliamentary questions in the Italian Chamber of Deputies between 2006 and 2018 support our main hypothesis that when intra-coalition uncertainty increases, coalition parties ask more questions of ‘hostile’ ministers.

All coalition governments face the inherent risk of agency loss. When the executive is composed by a multitude of political parties, a ‘dilemma of coalition governance’ arises (Martin and Vanberg 2011): individual parties' interests clash with collective interests of government coalition. All parties in charge of ministries might be tempted to deviate from coalition agreement in order to achieve their ideal policies at the expense of other coalition members. Political parties are first of all vote-seeking actors and to pursue their electoral goals are willing to abdicate from coalitional compromises. To avoid such risks, executives and legislative institutions provide coalition parties with an array of mechanisms to monitor ‘rival’ ministers (i.e. ministers from other parties). Previous literature has identified several ex-ante and ex-post instruments for coalition members to keep tabs on cabinet partners (Strøm et al. 2010): shadow committee chairs (Carroll and Cox 2012), legislative scrutiny (Martin and Vanberg 2011), junior ministers (Lipsmeyer and Pierce 2011; Thies 2001), coalition agreements (Klüver and Bäck 2019) and parliamentary committee systems (Andrè et al. 2016). In this article we investigate, following two recent studies on Germany (Hömann and Sieberer 2020) and the UK (Martin and Whitaker 2019), whether coalition parties make use of parliamentary questions (PQs) as additional instrument to control their partners in government. PQs are particularly attractive monitoring instruments because their main function is to extract information from the executive and they represent a very flexible and unconstrained instrument. Thus, PQs perform very well in minimizing the informational gap between ministers and coalition parties.

Moreover, our article suggests that uncertainty is an additional factor in explaining questioning activities. This differs to previous studies that have demonstrated that the engagement of cabinet parties in monitoring activities (also by using PQs) increases along with the ideological distance between the proposing minister and coalition parties, and when the specific portfolio is highly salient for coalition parties (Hömann and Sieberer 2020; Martin and Vanberg 2004; Martin and Whitaker 2019), the literature has not yet investigated the role of uncertainty in legislative scrutiny. When parties are uncertain about the decisions of their allies in cabinets, coalition parties have an incentive to adopt instruments for mitigating ministerial autonomy. In particular, we propose two potential sources of uncertainty in coalition politics: new parties in government and technocratic ministers. New parties are less predictable in their behaviours, in their policy positions and in their opportunism (Deschouwer 2008; Grotz and Weber 2016). Conversely, technocrats are less predictable because they are both experts and have professional careers outside politics. We suggest that, theoretically, the use of PQs is a matter not only of ideology but also of the actors involved in coalition governance. Moreover, we distinguish in our analyses between oral and written questions in order to unpack potential differences in the use of the two instruments. Conversely, previous research did not differentiate between the two types of questions (Hömann and Sieberer 2020) or focused only on one of them (Martin and Whitaker 2019; Tzelgov 2017).¹

In order to test our expectations empirically, we rely on a novel dataset of all written and oral questions asked in the Italian Chamber of Deputies from 2006 until 2018. Italy is a good test for our hypotheses for at least two separate reasons. First of all, in Italy legislative politics is strongly based on political parties, whereas individual efforts in asking PQs are limited (Russo 2011). Several studies approach PQs as a form of individual representation (Martin 2011), therefore to avoid individual incentives influencing questioning activities, we select a case study where parliamentarians are essentially delegates of their own parties. Moreover, the electoral system adopted in 2005 provides virtually no incentives to cultivate a personal reputation, therefore it is more likely that PQs were used for party reasons and not for personal electoral objectives during the legislative terms investigated here. Finally, the Italian case provides a sufficient comparative setting because it presents different types of government, different coalitions and majorities and alternation in cabinets. Given this institutional setting, the Italian case represents an ideal laboratory to test our theoretical arguments.

The remainder of the article is organized as follows. In the next section, we present PQs as a complementary instrument for intra-coalitional control. The following section reviews previous contributions on coalition delegation and introduces the role of uncertainty in coalition governance. Afterwards, we briefly describe our main explanatory variables (new parties in government and technical ministers), and then data and methods of analysis are presented. Empirical results are discussed in the subsequent section. Finally, we discuss the main findings and provide some suggestions for future research.

Parliamentary questions as a monitoring device in coalition politics

Coalition governments, in a world of incomplete information, always create conditions for agency loss (Huber and Shipan 2000). Parties in government are constantly tempted to pursue their own preferences at the expense of coalition partners by using their authority and informational advantage. To avoid agency loss, principals (coalition parties) employ various ex-ante and ex-post mechanisms within either the executive or parliamentary arena (Strøm et al. 2010). Junior ministers, coalition agreements, informational coalition committees, legislative review and parliamentary committees are among the most important intra-coalition monitoring devices analysed in the literature (see Strøm et al. 2010 for a review). In addition, recent studies have suggested several reasons to consider parliamentary questions (PQs) as an alternative and effective instrument to scrutinize coalition parties (Hömann and Sieberer 2020; Martin and Whitaker 2019; Otjes and Louwerse 2018; Tzelgov 2017).

First of all, holding government positions may imply a loss of votes for cabinet parties (Greene et al. 2021), therefore they need to prevent large and costly policy drifts but, at the same time, parties in government need to maintain a public appearance of unity (Martin and Whitaker 2019). PQs are able to guarantee both purposes at the same time. On one side, PQs serve as position-taking instrument for political parties to distance themselves from detrimental policy decisions by the other parties (Tzelgov 2017). On the other side, PQs enable the coalition to maintain a public appearance of unity because of their less-observed nature. PQs are public and open to potentially all electors but, given their technical form, especially in respect of written questions, they ‘represent an arguably more subdued way to patrol the coalition’ (Martin and Whitaker 2019: 1471) in comparison to other monitoring tools. In other words, coalition parties can monitor ministers in a relatively hidden manner by using PQs.

Second, PQs can extract information from ministers efficiently and, therefore, reduce informational asymmetries between agents (ministers) and principals (cabinet parties). This kind of information (partially) opens the doors of ministries to coalition parties and gives them oversight of what takes place within them. Third, PQs (especially written questions) are decentralized and relatively costless instruments that give political parties in the legislature a flexible device to question any issue at any time. Monitoring entails costs in terms of time, resources (i.e. staff, information, elaboration and coordination) and opportunities (Lupia and Strøm 2008). PQs are not costless but they are associated with relatively lower costs than other controlling devices (i.e. coalition agreement or legislative review). Finally, by studying PQs researchers observe a ‘real behaviour’ instead of a ‘potential behaviour’ as in the appointment of junior ministers or committee systems (Hömann and Sieberer 2020). Politicians can use their office to keep tabs on government once selected as junior ministers or in a parliamentary committee, but researchers can only assume that this opportunity will be used. Conversely, PQs are actual instruments in the hands of parliamentarians.

There are also different reasons that incentivize the use of PQs as monitoring tool. Daniel Hömann and Ulrich Sieberer (2020) suggest two ideological factors: the size of the expected policy drift and the involved political costs. First of all, coalition parties are more incentivized to control those ministers whose policy preferences are more distant from their own. Each party in government aims to achieve decisions in line with their ideal preferences, therefore coalition parties have greater incentives to invest time and resources in monitoring ministers who take remote positions. Second, not all policy issues are relevant and potentially damaging in the same way for all parties. Some issues are particularly salient for some parties and their own constituencies (e.g. environment for green parties), and in this policy area parties have the most to lose if the implemented policies are far away from their preferences. From this perspective, questions are used to prevent potentially damaging policy drifts in those departments that are relevant for parties and their own voters. Shane Martin and Richard Whitaker (2019) and Simon Otjes and Tom Louwerse (2018) find similar results respectively in the British House of Commons and in the Netherlands. In addition, Eitan Tzelgov (2017) notes that, in Greece, (oral) PQs have been an important mechanism to minimize the ‘political blame’ associated with unpopular austerity policies during the economic crisis. In this perspective, PQs have made it possible for coalition parties to publicly assign the responsibility of unpopular policies to other parties by taking different positions from the ministers responsible for austerity policies.

In this article, we theorize that policy reasons are not the only motivations for monitoring activities, but uncertainty also plays a relevant role in coalition governance. Knowing ideal points of parties' ministers or having similar positions is not enough for coalition partners to feel reliable and confident about their actual behaviour. Ministers and their parties may be incentivized to deviate from policy agreements even if policy preferences are similar or well known. As suggested by Kaare Strøm et al. (2010: 520), ‘uncertainty in coalition-government situations makes cooperation and delegation risky and therefore drives parties to seek ways to control their coalition parties’.

Problems in delegation: the role of uncertainty

Being in government implies a double electoral competition for parties in a coalition: against opposition parties, but also against coalition partners. Each government component tries to maximize their informational advantage in terms of votes (also and often) at the expense of coalition partners. In other words, they are allies in the cabinet, but rivals in the electoral arena, and electoral pressures prevail in many circumstances. There are two consequences for coalition parties. First, they compete in government to obtain the adoption of policies closest to their preferences, and each party applies this strategy in its areas of jurisdiction. Second, coalition governments need mechanisms to control and constrain coalition partners' behaviours and strategies.

Strøm et al. (2010) suggest that delegation failures arise from three factors: preference divergence, uncertainty and opportunism. Whereas previous research has primarily focused on preference divergence (Hömann and Sieberer 2020; Martin and Vanberg 2004, 2011; Martin and Whitaker 2019) as the main explanatory condition that produce monitoring activities of coalition parties, here we concentrate on uncertainty.² Our main argument is that particular characteristics of political parties and ministers (i.e. new parties in government and technical ministers) increase the uncertainty within the coalition. When intra-coalition uncertainty increases, political parties can predict less accurately which policies will be implemented by their partners, and therefore they need to police more closely those departments managed by rival parties.

The term ‘uncertainty’ has been used with a number of meanings in the political science literature. The most prevalent conceptualization, derived from game-theory models of strategic interaction, is incomplete information. More specifically, uncertainty implies the imprecision with which political actors are able to predict future interactions because of lack of information (Przeworski 1991). In this perspective, coalition governments are intrinsically based on uncertainty. In multiparty governments, knowing what will happen in the future implies (among the other things) knowing what the other parties will do. Each party in coalition can enjoy the benefit of office as long as cooperation lasts, but it is possible that other (electoral) interests invite parties to drift from the coalition agreement. There are two main endogenous sources of uncertainty in coalitional politics (Lupia and Strøm 2008): lack of information about each other's strategies, and lack of information about the relationship between implemented policies and the political outcomes political parties seek.³ For coalition parties, being uncertain both about the behaviour of their allies and about the potential effects of their decisions increases the opportunity costs of not controlling the other members. Not controlling potential strategies and decisions by allies in government could mean a loss of votes in the next election.

For these reasons, we suggest two potential sources of uncertainty in multiparty government: the entry of a new party in government coalition and delegation to technocratic ministers. In both cases, what drives coalition parties to scrutinize their partner is the absence or the different degree of access to information. In the first case, coalition parties have to increase their efforts to monitor activities because new parties in government do not have a ‘past’ that illustrates their propensity to deviate from or to maintain policy agreements. In the second case, given their technical advantage in the jurisdiction and their different policy goals, coalition parties must gather information on the content of draft bills to evaluate potential (electoral) consequences. Greater information asymmetries means greater uncertainty over coalition parties' behaviours and, therefore, a rise in intra-coalition control in order to avoid potential policy drifts.

Uncertainty becomes a challenge in coalition governments when a new party enters the coalition. First of all, parties without past government experience are less predictable in their (potential) opportunistic behaviours than those who have already spent time in such positions (Grotz and Weber 2016). Their reliability is still to be built and, therefore, coalition parties must invest more resources to check potential policy drift. Second, since new parties in government lack expertise and routine, they need to make and implement decisions for new problems within new conditions. Moreover, relations with the other parties in government need to be entirely created and organized. They do not have a ‘government history’ on which they can anchor their behaviours and strategies and, more important, this history is not available for coalition partners, which cannot anticipate how the other parties will act.⁴ At the same time, new parties are less identifiable in ideological terms, in policy goals, and they are more likely to shift their positions unexpectedly (Grotz and Weber 2016).⁵ Lack of information, in this perspective, increases the need for mechanisms to control ministers' behaviour and decisions. On one side, a party's first time in government means that it has to create almost everything from scratch. On the other side, when a new party enters government for the first time, the uncertainty for coalition governments increases, and therefore coalition partners need to devote more time and resources to legislative oversight.

Hypothesis 1: Coalition parties are more likely to ask questions of ministers from parties that are participating in government for the first time.

Newness, and therefore uncertainty, is also a condition for those ministers who cover a specific portfolio for the first time even if they are members of parties with past experience in government. For this kind of minister, cabinet office is a new experience with new responsibilities, new incentives and new opportunities, which all together (might) encourage behaviours that deviate from the expected political line. At the same time, we expect new ministers who belong to parties with past government experience to create a lower level of uncertainty than their counterpart from new parties in government, because party structure (might) constrain their opportunism and their behaviours once appointed as ministers.

Hypothesis 2a: Coalition parties submit more PQs to those ministers who are participating in government for the first time.

Conversely, we can suppose that the need for coalition parties to monitor their new partners decreases when ministers have past experience in government:

Hypothesis 2b: The effect of party newness on the number of questions asked by coalition parties decreases when ministers have past experience in the specific portfolio.

A second source of uncertainty in coalition government derives from the presence of technocrats. They can contribute to increase intra-coalition uncertainty because they are expert and independent from partisan control (Alexiadou et al. 2022). They are different from partisan ministers for (at least) two main reasons. First, non-partisan ministers are motivated by different aims (Alesina and Tabellini 2005): politicians want to be re-elected, whereas technocrats want to be perceived as competent in performing assigned tasks.⁶ The former are vote-oriented actors, whereas the latter are policy-oriented. Furthermore, non-partisan ministers' decisions are based on technical and procedural foundations and, unlike politicians, they are not interested in distributive outcomes of implemented policies, but rather on policy efficiency and effectiveness. In other words, they are policy-oriented, whatever it takes and whatever it means in electoral terms.⁷ As a consequence, they are only marginally concerned about the political implications of their tasks for political parties in government.

Second, technocrats’ behaviour is less constrained by political structures because they are generally selected for their autonomy from partisan logics and opportunisms. This condition also makes them independent of the other ministers and coalition parties. Partisan ministers are involved in intra- and interparty politics, whereas technocrats are totally free and disconnected from that competition. The political autonomy of this type of minister makes them a sort of ‘unidentified political object’, which needs a greater degree of scrutiny to avoid electoral damage resulting from their technical and unexpected decisions. After all, once their government experience is concluded, technicians (probably) leave politics and return to their professions, while politicians have to face the (electoral) consequences of decisions taken by technocrats during the government experience. Although the presence of technocratic ministers in cabinets, on the one hand, has several advantages and positive opportunities for the government, on the other hand, it (potentially) increases uncertainty within the coalition. As a consequence, technocrats are (or, at least, they are considered as) free agents that are harder to control than their partisan colleagues (Alexiadou and Gunaydin 2019).

Finally, uncertainty might arise from technical advantages that those in office have at their disposal (Lupia and Strøm 2008; Martin and Vanberg 2004). All ministers enjoy an informational advantage, but technocrats, because of their specific knowledge and competence in the policy issue, increase this informational gap. They have the opportunity to use asymmetrical expertise and technical knowledge to advance their own convenience at the expense of partisan interests. Their technical preferences, strategies and policy outcomes are (in many circumstances) unclear to coalition parties, making technocratic ministers worthy of attention. More specifically, since non-partisan ministers have more information, skills and technical competence and pursue different goals, the political consequences of their decisions are obscure for politicians and, as such, they need more scrutiny in order to avoid ‘policy uncertainty’ and potentially unsafe results (Huber and Shipan 2000):

Hypothesis 3: Coalition parties ask more questions of technocratic ministers than of partisan ministers.

Defining and measuring new parties in government and technical ministers

New parties in government

We consider as new parties in government all those parties that hold a government portfolio for the first time (prime minister or ministry) (Deschouwer 2008). In some cases, parties gain government status after a long time in opposition (e.g. the Communist Party in Italy), whereas others might simply be brand-new parties. A few more words are needed on the latter category. While, at first glance, defining what a ‘new party’ is might be considered a straightforward issue, in the literature there are two main approaches (Deschouwer 2008; Grotz and Weber 2016). On one side, a more restrictive approach links newness to parties that have a novel name and structure, and do not have any important political figure from past parties among their major members (Grotz and Weber 2016). Conversely, a more inclusive definition considers as new all those parties with a novel name and structure even if they are composed of politicians from established parties and/or they are the successor of previously active parties. The restrictive approach does not consider those new party structures that enter in the government without a past experience to be independent party entities. From our perspective, new parties that derive from mergers of or splits from other parties present brand-new structures and procedures unbound from their old parties. As a consequence, new ‘dominant coalitions’ (Panebianco 1988), with different and new policy preferences and strategies, lead political parties and their policy choices in government. As new parties, they face important pressure from the electoral arena, and defection from government compromise is a possible strategy to obtain (maintain) votes at elections. For example, the New Democratic Centre (NCD) was a new party formed by a split from the People of Freedom (a party with a history of government experience), but its novelty status enhanced uncertainty during its first experience in government, in Renzi's cabinet, because of the new power structures within the party. In other words, this party is something different from its predecessor, and therefore its behaviour is not completely predictable by coalition partners. For these reasons, we argue that those parties that are not ‘totally new’ are also able to add uncertainty in coalition politics.

Empirically, we code our main independent variable (New party in government) as a dummy with value 1 for all those (old or new) parties at their first experience in government, and value 0 otherwise. At the same time, uncertainty in day-by-day governance can derive from partisan ministers at their first experience in a specific office, even if they belong to an established party with a government history. The first time is special also for ministers, and therefore we include in our models a variable that measures First experience as minister and it is given a value of 1 when it is the minister's first experience, and 0 otherwise. Finally, we adopt an interaction term between new party in government and first experience as minister in order to test Hypothesis 2b.

Technocratic ministers

Technocrats are defined as professionals who have policy expertise within their departments and have never held office at national, subnational, local or supranational level before and during their ministerial appointment (Alexiadou and Gunaydin 2019; McDonnell and Valbruzzi 2014). First, technicians are (with few exceptions) experts and specialists chosen because of their specific knowledge in a policy area or their managerial capacities in the assigned policy portfolio (Costa Pinto et al. 2018). When their professional expertise is not aligned with the ministerial jurisdiction, they have been appointed for partisan and not for technical reasons, therefore they are actually political ministers. For example, an economist who is minister of health is not an expert because his/her professional specialization differs from the ministerial jurisdiction. His/her appointment derives from political and not technical reasons.⁸ Second, technocrats are not interested in a political career and their position as minister stems from their professional reputation. Ideally, technocrats have no political affiliation (they are not party members) and they are non-elected. This qualification as non-partisan makes it possible to implement reforms with high political costs, to signal a certain policy direction or to increase government credibility at the international level (Alexiadou et al. 2022). Technicians owe their positions to their specific competences and they are concerned about their reputation in the professional community. They are not prone to electoral cycles or to partisan needs in terms of votes. In other words, technicians are free from political games because they are non-elected. Of course, when a technician decides to take up a political career by running in an election, he/she turns into a politician and loses his/her status of technician. For all these reasons, we include in our statistical analyses the variable Technocrat, which is coded 1 when a minister presents all three of the following dimensions at the time of his/her appointment: (1) he/she has never been elected to a public office; (2) he/she is not a formal member of any party; (3) he/she has professional expertise in the policy jurisdiction he/she occupies in government.⁹

Data, method and control variables

To test our argument, we use written and oral questions asked by Italian coalition governments between 2006 and 2018 in the Chamber of Deputies. Italy is a good test because six different coalition governments have been in charge of cabinet positions during this period, and this gives us sufficient variations in intra-cabinet composition, new parties in government and technocratic ministers.¹⁰ Moreover, the choice of the Italian case gives us the opportunity to analyse an institutional setting where personal electoral incentives are absent. In fact, the electoral system adopted in the three rounds of elections (2006, 2008 and 2013) was a proportional system with blocked lists.

We have constructed an original dataset containing information on the number of questions asked by each party in government to each rival minister. Our observations are at party–minister level and include all combinations of coalition parties and government portfolios for each cabinet.¹¹ Because we are interested in the usage of PQs as an intra-coalition monitoring device, our dataset is composed of parties in government, excluding opposition parties. In total we have 377 dyadic observations.¹²

The dependent variable is the total number of oral and written questions asked by a coalition party of a minister belonging to another party in government.¹³ For example, during the People of Freedom–Northern League cabinet (the Berlusconi IV government, 2008–2011), we have collected the total number of questions that Northern League deputies asked of ministries held by the People of Freedom and vice versa. Data on questions were obtained from the official website of the Chamber of Deputies,¹⁴ whereas data on the composition of governments and portfolios assigned to each party are based respectively on the Parliaments and Governments (ParlGov) dataset (Döring et al. 2022) and the official website of the Italian government.¹⁵

As suggested by the literature, oral and written questions present two key differences (Rozenberg and Martin 2011). Oral questions, given their level of publicity (especially during the question time), deal with more general policy issues. Conversely, written questions tend to be more specific and technical in nature. Written procedures can be used to extract detailed information from ministers, whereas oral questions appear to be more appropriate for political theatre and as part of a position-taking strategy. Because of these differences, we decided to implement separate statistical models in order to check potential divergences between oral and written questions.

To further increase confidence in our results, we control for other factors that might influence levels of questions asked by coalition parties. Junior ministers may be strategically appointed to exert control over ministers, reducing the need for coalition parties to scrutinize ministers (Lipsmeyer and Pierce 2011). Therefore, we introduce in our models a dummy variable (Junior minister) with a value of 1 if the coalition party has a junior minister in the jurisdiction and 0 otherwise. The variable Same coalition identifies when a minister's party and coalition party have been together in previous cabinets (value 1; otherwise 0). Parties that have shared government experience in the past should have greater knowledge about each other and this is expected to reduce information uncertainty (Martin and Stevenson 2010). The number of PQs can also be influenced by the time a minister spends in office. The variable Tenure measures the total number of days a minister has spent in that specific portfolio.

Moreover, we control for Party size and Share of ministries. The first variable aims to capture the straightforward effect of coalition party size: the more seats the party has, the greater the number of questions asked (Otjes and Louwerse 2018). The variable Party size is the logged number of seats in parliament because we expect larger parties to ask more questions than smaller ones, but the marginal effect of additional members is expected to decrease. Conversely, Share of ministries reflects the proportion of ministries for each party and assumes that coalition parties with more ministries in cabinet are able to better control policy dynamics within the government and, therefore, they have less need to control, throughout the parliament, their allies (Thies 2001).

At the same time, we can expect that ministers from smaller parties in government receive fewer questions. Therefore, we have introduced two variables that mirror those used to measure the size of the coalition parties (Government party size and Share of government party ministries). Another potential variable able to confound the hypothesized effects is the nature of the coalition. In particular, oversized coalitions should reduce incentives to check for other coalition parties. The variable Surplus majority is coded 1 when the coalition government is oversized.

As discussed above, previous studies have shown that policy and ideological factors have a large impact on parliamentary scrutiny. In particular, Hömann and Sieberer (2020) propose two distinct conditions: the size of potential policy drift and the political costs associated with drift. In our model, to test their effects we measure two separate variables: Policy distance and Issue relevance, and in both cases we apply the logarithmic transformation of the Manifesto Project (MARPOR) data proposed by Will Lowe et al. (2011). These data possess two notable advantages: first, they allow to capture policy distance and issue relevance across the time and over multiple dimensions; second, data provide a more accurate measure of parties' positions because the scale considers either the relative weight of each sentence in electoral manifestos or the decreasing marginal effect of each extra-sentence.

In particular, to measure policy distance we create a specific policy dimension for each ministerial jurisdiction and then we calculate the absolute difference between the position of each minister's party and coalition parties. Details on the construction of policy dimensions and the corresponding ministry are provided in the Online Appendix (Table 1A). As concerns the operationalization of issue relevance we use the policy importance index proposed by Will Lowe et al. (2011), calculated as follows:

$$\theta = \log \displaystyle{{R + L + 1} \over N}$$

where R and L are respectively the total number of sentences for right and left scale, while N is the total number of sentences in the manifesto.

The saliency of each jurisdiction can also derive from the relevance of the government department. We adopt the index proposed by James Druckman and Paul Warwick (2005) to measure Ministerial relevance, where 1 means a ‘normal portfolio’ and a score above 1 reflects how much more important that department is in comparison to a ‘normal’ ministry (e.g. a score of 1.5 indicates that the office is 50% more important than average ministries). Similarly, any below-normal portfolio receives a score less than 1.

Our dependent variable is composed of count data and it is over-dispersed (the variance is larger than the mean), and therefore a simple regression model cannot be used and the most appropriate statistical model is the negative binomial regression (Cameron and Trivedi 2013). All models are estimated with robust standard errors clustered by cabinets in order to control for potential correlation within each government. Descriptive statistics for all variables included in our models can be found in the Online Appendix (Table 5A).

Results

Table 1 shows results for negative binomial regressions.¹⁶ Since our regressions are log-linear models, we report in Table 2 the expected values of PQs that are easier to interpret and clearly show the substantive effect of each main independent variable.¹⁷ We should note, however, that the variable First experience as minister is in the predicted direction but it is not statistically significant (except for oral PQs). This can be explained by the fact that, in Italy, political parties maintain a pivotal role in legislative and executive politics (Giannetti and Laver 2005). Party leadership exerts formal control over ministers' behaviour, and therefore their (lack of) experience in government is countered by the party's constraints and reliability. In other words, ministers without past government experience benefit from the reliability of parties they belong to.

Table 1. The Role of Uncertainty in Explaining the Number of PQs

Notes: Negative binomial regression; dependent variable: Number of PQs; robust standard errors in parentheses are clustered by governments; significance level: *** p < 0.01; ** p < 0.05; * p < 0.1.

Table 2. Expected Numbers of PQs for Different Levels of Independent Variables

Notes: Expected values are based on Model 1 (Table 1). For party size, tenure minister, ministerial relevance and policy distance value 1 and 2 are set to 1 SD below and above the mean.

At the same time, it is possible to observe differences in oral and written questions (Models 3 and 5). Data show that written questions are used to scrutinize new parties and technocrat ministers, whereas oral questions maintain the statistical significance only for the variable New party in government. This is not surprising given the different nature of the two types of questions. As suggested by the literature, oral questions are more general and oriented towards political competition, and therefore they are more often used openly to control a new party in government. Conversely, technical ministers are controlled by a technical instrument, such as written questions, where more detailed and specific information can be easily asked from ministers.

Turning our attention to control variables, Ministerial relevance has a positive and substantive effect on PQ activities. In line with the findings of Christine Lipsmeyer and Heather Pierce (2011), we find that the number of questions rises with the importance of the portfolio. Data in Table 2 show more specifically how and to what extent the number of PQs increases with the relevance of ministries. Moving the index from 1 standard deviation below its mean to 1 standard deviation above the mean, the predicted number of PQs rises from 5 questions to 10, which means a growth of 100%. Prestigious and prominent ministries, such as the Ministry of Economy or the Ministry of Foreign Affairs, represent an important electoral resource for the party in charge and, consequently, a potential electoral threat for coalitional parties. Furthermore, relevant ministries are also those that guarantee greater opportunities for distributive policies. For these reasons, controlling relevant departments is an additional strategy in intra-coalition politics. Moreover, we find evidence that parties that have governed together in the past (Same coalition) have less incentives to monitor their rivals in government because they have more familiarity with them.

In line with Michael Thies's suggestions (2001), our findings provide evidence of the role of junior minsters in downsizing policy drift by reducing the ministers' informational advantage. In particular, our data suggest that junior ministers are not substitutes for other monitoring instruments, but rather complements. Legislators, alerted by junior ministers, are incentivized to collect more information (by PQs) from ministers about potential deviations from policy agreements. In this perspective, junior ministers perform a ‘fire alarm’ function, alerting coalition parties about executive decisions with potential adverse (electoral) consequences. This role as fire alarm is consistent with findings from Andrea Pedrazzani and Francesco Zucchini (2013), who suggest that junior ministers in Italy induce legislators to amend broadly bills proposed by hostile ministers.

Surplus majority has different directions for oral and written questions. In oversized coalitions, oral questions are used to scrutinize ministers from rival parties. Indeed, the number of oral PQs increases by 200% (from 2 to 6). This can be explained by the nature of oversized governments: in a surplus majority all parties need strong instruments to keep the line in government and to take positions against rivals, and oral PQs are a possible tool in this way. Conversely, given the relatively hidden nature of written questions and the weak impact on government parties, parties in surplus majorities have less incentive to monitor their allies by using these technical instruments.

Turning our attention to policy dimensions, we find contrasting results for the two variables. On one side, Policy distance indicates that coalition parties ask more questions of those ministers who are more distant in terms of policy positions. The substantive impact of this variable is clear if we look at predicted values (Table 2): shifting policy distance by one standard deviation below the mean to 1 standard deviation above the mean, the predicted value of total PQs is expected to increase from 26 to 34 questions (a growth of 31%). Not surprisingly, the impact of policy distance on oral questions is even more marked, reaching 50%. Also in this case, oral questions appear as a political instrument, used by MPs in the intra-coalitional theatre. On the other side, and in contrast to previous findings (Hömann and Sieberer 2020; Martin and Whitaker 2019), our results demonstrate that, in the Italian case, issue relevance is not significant for coalition parties. A possible explanation stems from the office-seeking character of political parties in Italy (Giannetti and Laver 2005). Coalition parties have more incentive to monitor important ministries in terms of prestige and distributional benefits because these are the areas where parties have the most to lose from abdication, even if these issues are not relevant in their electoral platforms. Ministerial prestige overshadows policy issues, even those most salient for the party, and induces parties to monitor those departments that guarantee the greatest redistributive payoffs.

At the same time and not surprisingly, all models show that coalition party size has a significant and large effect on their questioning activities. Although larger parties have more resources to control government decisions from within (being a large party means having more (junior) ministers), they exploit parliamentarians at their disposal to scrutinize government actions. Also in this case, our results suggest that PQs are used as complementary instruments by parliamentarians in order to keep tabs on the executive.

Finally, in order to assess Hypothesis 2b, we turn our attention to Models 2, 4 and 6. The multiplicative variables are all statistically significant, but to better understand their substantive effects we show in Figure 1 the marginal effects (Brambor et al. 2006). The figures clearly illustrate that new parties in government are more questioned by coalition parties than parties with past experience, but when an expert minister is appointed to the ministry the impact of newness decreases. Regarding oral questions, the conditional effect of past experience in government is slightly different and, in particular, the marginal effect is higher when ministers have previously spent time in the same portfolio.

Figure 1. The Interaction Effect of ‘New party in government’ and ‘First experience as minister’ on PQs

Note: Solid line = new party in government; dashed line = parties with past experience in government; vertical bars show 95% confidence interval.

Conclusions

Coalition government is intrinsically based on delegation and compromise. Both conditions make it necessary for all parties in government to control if and how other members are inclined to abdicate from the coalition agreement. Previous literature has provided empirical evidence about what factors incentivized parliamentary control in coalition government. In particular, policy divisiveness and issue relevance have been proposed as the main reasons for monitoring executive coalition members. In this article we approach the study of coalition government from a new perspective. First of all, we suggest that uncertainty is an additional factor that can increase the need to monitor and find more information in intra-coalition governance. Second, we show that PQs are an additional intra-coalition monitoring device. By using a new dataset for Italian multiparty governments (2006–2018), we find support for our central hypotheses. Uncertainty, as measured by new parties in government and non-partisan ministers, increases the use of oral and written PQs to check on allies in Italian governments. New parties are less predictable in their behaviour and, therefore, they are more subjected to control by their coalition partners. At the same time, technocratic ministers, because of their technical competence and independence from partisan logics and resources, have more opportunities to implement policies with uncertain (electoral) outcomes for political parties. In both cases, uncertainty leads to greater scrutiny of ministers.

Our study presents an important step forward in coalition governments studies. It is the first attempt to study uncertainty in intra-coalition governance and suggests the necessity to look beyond ideological division to understand fully the political game within multiparty governments. We also contribute to the more general research on PQs by investigating the different roles of written and oral questions in coalition governments, whereas previous studies do not distinguish between these types or are based on only one. In particular, our findings demonstrate how written questions are used to obtain detailed and technical information, especially from technocrats. Oral questions, on the other hand, are devoted to publicly take positions against rival parties in government. At the same time, we are conscious that much further research is required to better understand the role of uncertainty.

Specifically, we identify three potential paths of research. First, further studies should test whether our results can be generalized to other countries or settings. The Italian case is a good starting point to be relatively confident about our results because all hypotheses are theory-driven, based on past research and are far from being country-specific. However, at the same time we recognize that much more investigation in comparative perspectives is needed to generalize with more confidence our arguments. Second, electoral pressures are not the same throughout the electoral cycle. The pressure increases when elections are coming, therefore future research should consider the timing of questioning activities. Finally, in this article we focus on uncertainty within coalition government but uncertainty also depends on external shocks. Wars, pandemics and economic crises are all examples of events that potentially increase uncertainty for governments and therefore the need to scrutinize ministers.