Coalition Governance under Separation of Powers: Shadowing by Committee in Coalitional Presidentialism

do coalition parties in presidential systems shadow one another with an eye on reducing agency loss? Our knowledge of intra-coalitional delegation under presidentialism remains limited to a few case studies. this article addresses that question using an original data set of committee chairs in eight multiparty presidential systems around the world. it shows that shadowing—the appointment of chairs from one coalition party to committees overseeing portfolios controlled by another coalition party—is a function of the degree of agency loss the coalition expects to a specific party, operationalized in terms of ideological distance as well as the salience of portfolios. However, the data also tentatively suggest that powerful committees reduce rather than increase the incidence of shadowing, which may be attributed to the risk of intraparty agency loss under separation of powers.


Introduction
Presidents in minority situations often need the support of other parties to be able to govern effectively, which they obtain by handing them cabinet portfolios (amorim Neto 2006(amorim Neto ), pork (ames 2001;;raile, Pereira, and Power 2011), and policy influence (Magar and Moraes 2012).to the extent that parties seek policy influence, and these policy interests transcend the jurisdiction of the portfolio(s) they control, they will want to keep tabs on their partners.a nascent literature on coalition governance 1 in presidential systems has emerged over the past few years.When it comes to contract design, presidential pre-electoral coalitions seem to serve some of the same functions that coalition agreements do in parliamentary systems, as they also involve presidents and parties making public policy pledges that are costly to renege on (Kellam 2017).regarding screening and selection, presidential coalition cabinets also tend to be composed of parties that are ideologically close to the president (alemán and tsebelis 2011).Monitoring and reporting have also attracted attention, especially concerning the Brazilian case.authors have studied the employment of junior ministers (Pereira et al. 2017), committee chairs (inácio andrezende 2015), and coalition legislators filing information requests (silva and Medina 2023) for shadowing ministers.all three of these are case studies of Brazil; however, comparative research on the topic still lacking.
in this article, i seek to investigate the issue of intra-coalitional monitoring cross-nationally and cross-regionally, focusing on the use of committee chairs by parties in multiparty presidential cabinets.i set out to answer the question: do coalition partners under presidentialism strategically employ committee chairs to keep tabs on one another?as the separation of powers causes legislatures to be more relevant and powerful as policymaking actors (tsebelis 2009), the role of legislative committees in managing intra-coalitional delegation is a particularly salient topic under presidentialism.Committee chairs have important (informal) agenda-setting prerogatives and privileged access to information that makes them particularly suitable as monitors of ministerial portfolios (Fortunato, Martin, and vanberg 2019).they also have the power to amend or block legislation, enabling them to play the role of institutional checks (Fortunato 2021;Zubek 2021).While not always de jure, their election is de facto strongly influenced by party leaders (Pereira and Mueller 2000;strøm 1998), giving the latter substantial leverage to negotiate chairs of desired committees.the added benefit is that the legislators they pick for the job are loyal to the party leadership (Pearson 2015;santos and rennó 2004).
to investigate the aforementioned research question, i marshal an original database of committee chairs in eight presidential and presidentparliamentary systems around the world.these eight cases are emblematic of coalitional presidentialism.i argue that, if parties negotiate and employ committee chairs with an eye on reducing agency loss to their cabinet partners, the pattern of committee chair appointment should reflect parties' strategic considerations.in particular, i expect shadowing to be more likely in portfolios held by parties whose ideal points are further away from those of the coalition and in portfolios that are more salient.Furthermore, i test the effect of committee system powers, for which i formulate an ambiguous expectation.On the one hand, more powerful committees could have no effect on shadowing, as shadowing might be desirable regardless of committee powers.On the other hand, committees' enhanced effectiveness could encourage shadowing.a third possibility is that powerful committees may, as a result of the separation of purpose inherent in presidentialism (samuels and shugart 2010), introduce additional agency-loss risks between ministers and legislators that discourage committees' use for shadowing.employing a multilevel logistic regression model, i find support for the first and second hypotheses.Committee powers are associated with a reduction in the incidence of shadowing, lending support to the third part of the ambiguous expectation.However, this result warrants caution, as the small number of countries included in the study makes it difficult to attribute the observed variance in shadowing to committee powers.More cases would be needed to exclude effects from idiosyncratic differences between countries.
the results, thus, suggest that coalition partners under separation of powers seek to assure that collective policy pledges are faithfully implemented.Coalition Governance under Separation of Powers

Coalition Governance in Presidential Systems
in any cabinet, ministers possess hidden information on the goings on of their ministries vis-à-vis their principals, which provides them with incentives to follow their own or their parties' interests (shirking) rather than that of their principal (thies 2001). in coalition cabinets, the risk of agency loss is magnified, as these are composed of multiple parties with divergent preferences (thies 2001). in this, the coalition cabinet itself can be seen as a collective principal that delegates to the same ministers that constitute it (andeweg 2000). 2  When agents deviate from the contract agreed upon with their principals, the latter are said to suffer agency loss (Kiewiet and McCubbins 1991).
solutions for agency loss can be divided into four categories: contract design, screening and selection, monitoring and reporting, and institutional checks (Kiewiet and McCubbins 1991).Contract design is about aligning the incentives of principal and agent through a schedule of rewards and sanctions tied to the agent's behavior (Kiewiet and McCubbins 1991). in the context of coalition governance, the most important type of contract design is the coalition agreement, a public document detailing the coalition's collectively agreed-upon policy pledges (Höhmann and Krauss 2022;Moury 2011Moury , 2013;;Müller and strøm 2008;strøm and Müller 1999).screening and selection concerns choosing and screening potential agents to avoid adverse selection (Kiewiet and McCubbins 1991). in the context of coalition governance, an example of screening and selection would be parties coalescing with ideologically close partners in order to reduce conflict or agency loss (axelrod 1970; de swaan 1973).
Monitoring and reporting can take different forms, the most efficient of which is by the use of "fire alarms," that is, making it possible for third parties who are affected by the agent's actions to report undesired action to the principal (McCubbins and schwartz 1984). in the context of coalition governance, these fire alarms can be junior ministers (greene and Jensen 2016; Lipsmeyer and Pierce 2011; Pereira et al. 2017;thies 2001) or committee chairs (andré, depauw, and Martin 2016;Carroll and Cox 2012;Fortunato, Martin, and vanberg 2019;inácio and rezende 2015;Kim and Loewenberg 2005;Pukelis 2016) from parties other than that of the respective minister.Finally, institutional checks are said to exist when one agent has the power to block the activities of another agent (Kiewiet and McCubbins 1991).an example would be a minimal winning coalition, in which each member acts as a veto player (tsebelis 1995). in fact, legislative committees can also serve as institutional checks.they can often rewrite bills or even refuse to report them out of committee altogether, which Zubek (2021) aptly describes as the "gate-keeping of bills" (p.1023).
early models of coalition governance were produced with parliamentary systems in mind.However, presidential systems differ from their parliamentary counterparts in important ways, and these differences pose certain unique dilemmas for coalition governance.Most importantly, the chain of delegation is radically different under presidentialism.Presidentialism is characterized by a "grid" of delegation whereby voters simultaneously delegate to two separate agents, the president and the congress, both of which delegate to the departments (strøm 2000).Parliamentarism, on the other hand, is characterized by a single chain of delegation from voters to the parliament, from the parliament to the cabinet, and from the cabinet to the ministerial departments (strøm 2000).this difference affects executive-legislative relations in the sense that, under presidentialism, the origin and "survival" of these two branches of government are separate, whereas it is fused under parliamentarism (shugart and Carey 1992).
Legislatures and governments under presidentialism are, therefore, better able to operate autonomously from one another than under parliamentarism.under parliamentarism, the executive tends to dominate the legislative process due to the formidable agenda-setting power conferred onto it by the confidence vote and the ability to tie specific policy issues to matters of confidence (Cox 1987;Huber 1996;tsebelis 2009).this is reflected by the fact that prime ministers have higher legislative passage rates than presidents (saiegh 2011).as such, "policymaking power is concentrated to governments in parliamentary systems and to parliaments in presidential ones-exactly the opposite of what their names indicate" (tsebelis 2009, p. 17).all else equal, legislatures should, therefore, have greater relevance as loci of policymaking and contestation of executive power under presidentialism compared to parliamentarism.aside from the empowerment of the legislature, the separation of powers has an effect on the interests of legislators and presidents.the different election methods for the executive and the legislature lead to a separation of purpose between those branches.that is, while legislators are more attuned to local interests and focus on securing targeted funds for projects in their constituencies, presidents are held accountable for national-level policy (samuels and shugart 2010).
Presidentialism, however, does not only have a different pattern of delegation between the branches of government but also within the executive branch.rather than a primus inter pares, we are dealing with a primus solus (sartori 1994): a president who appoints and dismisses ministers at will, without that directly affecting her own survival in office (amorim Neto and samuels 2010).the whole concept of "coalition" is, therefore, markedly different under presidentialism.Nevertheless, the fact that presidents whose parties did not hold legislative majorities have frequently crafted majority coalition cabinets (Cheibub, Przeworski, and saiegh 2004) attests to presidents' dependence on support from parties other than their own.after all, stable legislative majorities underpinned by a distribution of coalition "goods" are an efficient means of getting one's policy program through the legislature (amorim Neto 2006).While presidential government coalitions may function in a less horizontal manner than their parliamentary counterparts, its junior partners have some leverage and influence over political outcomes (Magar and Moraes 2012).
the delegation pattern characteristic of presidentialism, both between the branches of government as well as within the executive, create agency loss First, the separation of purpose produces a misalignment of interests between the branches, making coordination across those branches, a hallmark of coalition government (Cheibub and Limongi 2010), more difficult.this also manifests itself within parties (samuels and shugart 2010).Party representatives in the executive branch might, therefore, not be 100% aligned with their parties' congressional leaders (Lupu 2016), which generates incentives for agency loss in both directions.second, the hierarchical nature of cabinets means that intra-executive interactions tend to be bilateral-between the chief executive and individual ministers-rather than multilateral (Poguntke and Webb 2005).this deprives the minister(s) of party a of information on the activities of the minister(s) of party B. Monitoring among coalition partners should, therefore, be even more important in such a context.

The Mechanics of Shadowing by Committee Chair
But why are legislative committees useful for tab-keeping and acting as institutional checks on ministerial drift?First, the committee system incentivizes and enables policy specialization by its members (Krehbiel 1991), especially when committee jurisdictions closely correspond to ministerial ones (Martin and vanberg 2005).Committees with more policy expertise are in a better position to monitor the corresponding minister.such exclusive correspondence also causes committee members and ministers to frequently rub shoulders.For instance, in Chile, "some of the most substantive interaction between ministers and legislators occurs during the committee deliberation phase, before a bill is reported to the floor for final passage" (alemán and Navia 2016).second, committees have powers that allow them to obtain information from the ministries they oversee, such as the ability to hold hearings of government officials, compel the attendance of witnesses, and demand evidence (Martin and vanberg 2005).third, committees can stop drift dead in its tracks by rewriting deviant bills to bring them closer to the coalition compromise (Fortunato 2021, pp. 114-15).appendix 5 shows that committees have at least some of these powers in all country cases.that does not settle the issue of committee chairs themselves, however.Chairs have substantial agenda powers that they can use to scrutinize government bills (Fortunato, Martin, and vanberg 2019). in some of the country cases studied here, such as Brazil and Malawi, committee chairs have substantial formal agenda-setting powers. in others, such as Benin and the Philippines, committee chairs, respectively, sit on a leadership body, la Conférence des Presidents, with the chamber's leadership and the party leaders (Benin); or participate in statutory monthly coordination meetings with the speaker (the Philippines).in yet others, such as indonesia and uruguay, the statutory powers of committee chairs are more modest and are restricted to determining the ordering of conflicting reports supported by an equal number of members (uruguay) or submitting draft laws to the chamber's legislation body (indonesia).appendices 3 and 5 provide an overview of formal chair powers.However, modest formal powers often camouflage more substantial informal powers (Fortunato, Martin, and vanberg 2019).For instance, in indonesia, "the rules of procedure are poorly elaborated and provide a great deal of latitude for committee chairs to interpret and apply at their discretion" (sherlock 2012, p. 561).all in all, regardless of their statutory powers, we can assume that committee chairs are influential figures both within and without their committees.Now that we have established that committees are effective monitoring instruments and their chairs have important formal and informal powers within them, the third component of the argument requires that committee chairs act as agents of their parties.according to the partisan coordination perspective of legislative organization, majority parties or coalitions as agenda cartels instrumentalize the committee system (and other legislative leadership posts) to serve their own partisan or coalitional goals (Cox and McCubbins 1993;Kiewiet and McCubbins 1991;Neto, Cox, and McCubbins 2003). in multiparty systems, committee assignments and chairmanships tend to be allocated to parties proportionally to seat share (Carroll, Cox, and Pachón 2006).Party leaders then tend to control the nominations to the committees.these leaders will tend to select loyal party members, and legislators' ambitions to obtain a coveted position as a committee member, especially as a chair with prestige and policy influence, serves as an incentive to be loyal to the party (Cox and McCubbins 1993). in all of my country cases, committee assignments follow the principle of proportionality (see appendix 5 for all committee information).Furthermore, committee assignments explicitly involve party leaders in all cases except for Chile and the Philippines.However, in Chile, congressional delegation leaders can request the transfer of their members away from the committees they were originally assigned to, subject to the approval of the Chamber's secretary.in the Philippines, the Majority (defined as the congress members who voted for the winning candidate for speaker) and the Minority (those who did not) determine committee assignments.to the extent that the Majority and Minority are made up of reasonably disciplined parties, however, party leaders are likely to have a major influence. in all of the country cases, except for the Philippines, committees elect their own chairs 3 .However, formal procedures again often mask informal ones and in european parliamentary democracies, chair selection often results from interparty negotiations (strøm 1998, p. 41).there are also examples of such informal procedures from some of my country cases.For instance, in Brazil, while committee chairs are officially elected by secret ballot within the committee, in practice they are selected by the party leaders (Pereira and Mueller 2000).Likewise, in indonesia, formally, each party may propose one candidate for committee chair (appendix 5). the chair is then elected within the committee by "consensus," only going to a nominal vote when consensus cannot be achieved.However, this is what (sherlock 2012) has to say about the role of "consensus" in indonesian legislative politics: 19399162, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/lsq.12451by Faculdade Medicina De Lisboa, Wiley Online Library on [23/02/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Coalition Governance under Separation of Powers (…) consensus does not mean that each and every DPR member has to agree individually to a proposition for it to be passed.In effect, it is only the party leaders in the committee or plenary session who deliver the votes of their party: consensus is, in reality, an agreement amongst party leaders.If agreement cannot be reached on the floor of the meeting (often because just one or two parties are dissenting and thus unanimity cannot be achieved), the leaders retire to a closed meeting and usually emerge with a decision to which all parties and their members are then said to have agreed to.(p.561) table 1 summarizes formal and informal chair selection procedures in each of my country cases.it also indicates whether committee chairs rotate during the legislative term, or are reappointed annually.the latter applies to three countries in the sample.Committee chairs that serve for the entire term should theoretically be able to accrue more expertise and consolidate more power over their committees.in sum, in legislatures across political systems and the world, party leaders can negotiate committee chairs even in spite of formal regulations that do not specify a role for them in this process.While most coalition parties are unlikely to be able to pick their top choices for committees to control (Carroll and Cox 2012, p. 227), they have at least some influence over this.

Determinants of Shadowing by Committee Chair
the main argument of this paper is that coalition parties strategically negotiate committee chairs with an eye on reducing informational asymmetries and agency loss vis-à-vis their partners.that is, shadowing by committee chair should be more likely when parties either: 1. expect a greater degree of agency loss to a specific party due to preference divergence; 2. have more to fear from agency loss as it concerns portfolios they care about or that are generally considered important.Finally, the potency of the shadowing device of interest may also play a role, but it is not clear what role, if at all.these three considerations are captured by the following three main independent variables: the ideological distance between the portfolio-holding party and its partners, the prestige of portfolios, and the powers attributed to the committee system.

Ideological Distance
the first factor that should be relevant for shadowing is ideology.ideological differences between agents and principals are a sign of preference divergence and heighten the risk of agency loss (ennser-Jedenastik 2014).studies on committee chair-based coalition monitoring have found ideological distance between the minister and the coalition to be a key predictor of shadowing in parliamentary systems (Carroll and Cox 2012) as well as in presidential systems (inácio and rezende 2015).However, the latter study finds a negative rather than a positive effect of ideological distance on intra-coalitional shadowing in Brazil.Nevertheless, the weight of the comparative evidence points in the direction of ideological distance positively affecting shadowing.in a presidential coalition government, the cabinet coalition as a whole can be seen as the informal principal of the individual ministers (the president being the constitutional principal).as the distance between the portfolio-holding party and this principal increases, we should see an increase in either type of shadowing.this brings me to my first hypothesis: Hypothesis 1 the greater the ideological distance between the portfolio-holding party and the weighted mean position of the coalition, the greater the probability that the portfolio is shadowed.

Portfolio Salience
Finally, not all portfolios are created equal and those portfolios with the greatest potential to shape national policy should be most heavily monitored ). a case study on junior minister-based coalition shadowing in Brazil, however, did not find a significant effect for portfolio salience (Pereira et al. 2017), and neither did a case study on committee-based shadowing in that country (inácio and rezende 2015).the danger of ministerial drift to high-impact portfolios is compounded if those portfolios are controlled by parties that are ideologically distant from the coalition compromise.Hence, we can expect a positive interaction between portfolio salience and ideological distance.this brings me to my second hypothesis: Hypothesis 2a the greater the salience of the portfolio, the greater the probability that the portfolio is shadowed.

Hypothesis 2b
Higher salience portfolios are especially likely to be shadowed when they are controlled by parties whose ideal points are located further away from the coalition compromise.

Committee Powers
as we saw in the previous section, committees are more effective monitoring instruments and institutional checks when they are granted certain powers.this is supported by empirical evidence.thies (2001) finds that junior ministers are less often used for shadowing in germany compared to his other country cases, which he attributes to that country's strong committees being used instead for that purpose.this is confirmed by Kim and Loewenberg (2005) for the german case and by Lipsmeyer and Pierce (2011) in a large-n study of several parliamentary countries.Furthermore, andré, depauw, and Martin (2016) find that legislatures in (parliamentary) countries where coalition governments are the norm develop strong committees designed for keeping tabs, and Zubek (2015) finds that more heterogeneous cabinets are more likely to adopt reforms extending committee powers.More directly, Carroll and Cox (2012) find that committee shadowing is more likely in (parliamentary) countries with strong committees.
We could, therefore, expect that, when committees are more powerful, parties should be more interested in employing them for tab-keeping.Furthermore, this effect should interact with ideological distance: committee powers should matter less for shadowing when parties are close to the coalition mean compared to when they are distant from it.However, one could also argue that, within a given country, if coalition parties are interested in shadowing, they should have that interest regardless of how powerful their committee system is in comparison with other countries.Furthermore, the specificities of presidentialism warrant caution when applying hypotheses derived from parliamentarism.Following the section "Coalition governance in presidential systems", the separation of purpose generates intra-party agency loss between executive officials and congressional delegations.this means that a committee chair of party a may scrutinize a minister from party B in ways unforeseen by party a's cabinet representatives.the key point is that, compared, to, say, junior ministers, who are part of the executive and more easily controlled by their copartisan ministers or presidents, there exists a degree of uncertainty when parties rely on their legislators for shadowing, as these come with their own biases and interests.the interference of these biases in the monitoring process is likely to be more pronounced when committees are powerful, which may make cabinet parties more wary of committee-based shadowing under those circumstances.this is especially a problem where committees have gate-keeping powers that can be used to sabotage the cabinet's lawmaking ability.
as an example of such a misalignment between a party's congressional delegation and its cabinet representatives, in 2006, the indonesian PPP (united development Party), a junior coalition partner of President susilo Bambang yudhoyono, withdrew two of their ministers from the cabinet-sugiharto of state-Owned enterprises and suryadharma ali of Cooperatives and small and Medium enterprises-over disagreements with the party's congressional delegation (rachmadi 2006) 4 .
due to the difficulty of making credible ex-ante predictions about the effect of committee powers on shadowing outcomes, which may be positive, negative, or absent, i refrain from formulating a hypothesis on this and remain agnostic about the result.

Case Selection and Data
Case selection was driven by three criteria: 1. substantial experience with democracy (at least four presidential terms or 16 years), 2. an overwhelming share of multiparty governments (at least 75% of the time), 3. geographic balance, 4. a preference for emblematic cases that have featured prominently in the literature on coalitional presidentialism.the Whogov cross-sectional data set was used to shortlist cases classified as "presidential democracies" where criteria 1 and 2 held.this led to a list of 13 cases in Latin america, africa, asia, and europe.to this list, uruguay was added, where apparent singleparty governments are actually coalitions of multiple factions and, following altman (2000), are treated as such in this paper.accordingly, 100% of uruguay's governments in the period for which i have data can be considered coalitions.the only european case, switzerland, was excluded because it is too sui generis a case to be grouped together with regular presidential systems.there were three asian cases, indonesia, the Philippines, and sri Lanka, and in order not to skew the sample too much toward any specific region, a target of three cases per region was adopted.However, it turned out that sri Lanka did 19399162, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/lsq.12451by Faculdade Medicina De Lisboa, Wiley Online Library on [23/02/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Coalition Governance under Separation of Powers not have an independent committee system 5 and therefore had to be excluded.in Latin america, Brazil, Chile, Colombia, Mexico, suriname, and uruguay fit the criteria.Of these, Brazil, Chile, and uruguay feature most prominently in the literature on coalitional presidentialism 6 and were, therefore, selected.in africa, sierra Leone, Benin, Malawi, and Kenya were eligible, and the latter three were selected as they have been studied most profoundly in this context 7 .
i chose to focus on what Chaisty, Cheeseman, and Power (2018) have termed "day-one coalitions," which are those coalitions formed immediately after the inauguration of a new president.this is because coalitions and cabinets tend to change over the course of the presidential office term, and coalitions later in the term are subject to markedly different dynamics than those formed at the start of the term (altman 2000).specifically, analyzing coalition governance over the course of the electoral cycle would imply taking into consideration the possibility that shadowing may decrease over time because parties become less interested in policy as elections approach.While that is an interesting question worthy of exploration, it is simply beyond the scope of this paper.Following the "most-likely" logic, the research design has followed so far, shadowing should be most likely at the beginning of the term when time horizons are still long and there is more at stake policy-wise.as far as operationalization is concerned, i maintain a strictly cabinet-based definition of coalitions, that is, the set of parties represented in the cabinet, whether or not the government is supported on the assembly floor by parties without cabinet representation.the main reason is that access to data on legislative coalitions differs markedly among countries.
these day-1 cabinets were then matched to legislative periods, where it was important to make sure that committee chair elections were held close to the inauguration date of the cabinet.after all, the paper is built on the assumption that committee chair elections respond to cabinet makeup.this meant matching presidential inaugurations to legislative inaugurations. in some cases, this proved problematic, especially for Benin and Malawi 10 .
another challenge was matching the committees to the departments.in the case of indonesia, the correspondence between committees and departments is specified in detail by the standing orders.in the remaining cases, the correspondence had to be established by consulting both the exact jurisdictions of the committees in the standing orders as well as the exact jurisdictions of the departments on the respective websites and matching them.as the structure of the executive branch as well as the committee system change frequently over time in most countries, this process had to be 19399162, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/lsq.12451by Faculdade Medicina De Lisboa, Wiley Online Library on [23/02/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License repeated for each cabinet-legislature pair.Committees that did not match any executive department, such as housekeeping committees, or which matched only executive agencies without ministry status, which was the case for many Philippine committees, were excluded from the analysis.
For cabinet data, i largely relied on Whogov (Nyrup and Bramwell 2020), triangulated with case-specific official or media sources in case of doubts.For data on committee chairs, i heavily relied on the legislatures' websites, where necessary in combination with the Wayback Machine 11 , and also made use of Hansards and media reports for legislatures whose websites did not have this information, supplemented with secondary data in one case 12 .see appendix 4 for more details on data sources.

Model and Variables
My unit of analysis is the portfolio-country year.My sample includes 417 portfolios held by partisan ministers in 31 coalition cabinets.My dependent variable, shadowing, takes a value of 1 when at least half the committees overseeing a given portfolio are chaired by members of cabinet parties other than that of the minister in charge.Otherwise, it takes a value of 0. table 2 visualizes this coding.shadowing was observed for 195 out of 417 portfolios, or 47% of the sample.While i used the terms "monitoring" and "shadowing" interchangeably up to this point, henceforth i will use the term "shadowing" strictly in the sense defined above.the reason is that "monitoring" implies a deliberate activity on the part of the committee chair that i do not directly observe in this study.Furthermore, shadowing can serve both a monitoring as well as an institutional checks function.
Portfolio-minister pairs are nested into the ministers' parties and into cabinets, both of which are, in turn, nested into countries.as these higher level units were sampled from larger populations, such a nested data structure violates a key assumption of linear regression, that is, the independence of the residuals (sommet and Morselli 2017, p. 206).Observations belonging to the same higher level unit may be more alike to one another than observations belonging to different higher level units.For instance, certain parties may be more likely to want to monopolize certain ministries and resist shadowing-let us say a farmers' party and the agriculture portfolio.alternatively, certain cabinets may have collectively decided for partners to shadow one another, and certain countries may have different norms when it comes to coalition governance, e.g., consociational norms whereby coalition cabinets are expected to govern together.
as my outcome variable is binary, the correct empirical strategy is logistic regression.However, due to my nested data structure, this logistic model must also be multilevel.Multilevel analysis "aims to disentangle the within-cluster effects (…) from the between-cluster effects" (sommet and Morselli 2017, p. 206). in this type of analysis, the number of clusters or second-level units is more important than the number of observations in each cluster, and simulations have shown that at least 50 second-level units are necessary to obtain accurate standard errors (sommet and Morselli Cabinets are more numerous but with 31 cabinets i still do not have enough to meet the 50-unit threshold.Parties in the sample, on the other hand, number 73.Not only that, but the between-cluster variance of the dependent variable is also higher for parties than for cabinets and countries.the intraclass correlation coefficients (iCC) reveal that 24% of the variance in shadowing is explained by interparty differences, but only 15% by betweencabinet differences and 20% by intercountry differences.Hence, clustering by parties is the most sensible approach.My empirical strategy is, thus, similar to the one used by Carroll and Cox (2012), who also use a multilevel logit model clustered by ministers' parties.
as such, the model specification is represented by equation ( 1) below (adapted from sommet and Morselli (2017)).…in which P Y i = 1 is the conditional probability that the outcome variable (shadowing) equals one for a portfolio i, B ∞ is the fixed intercept, B 10 is the fixed slope of level-1 predictor variable x ij , which affects a portfolio i controlled by a party j, B 0 1 is the fixed slope of level 2 predictor variable X j , which affects a party j (this variable is located at the party level and thus affects parties as a whole rather than specific portfolios), u 1j is the deviation of the cluster-specific slope of x ij from its fixed slope, which is included to control for the possibility that the effect of x ij on Y i differs between clusters, and u 0j is the random intercept variance.
My vector of independent variables includes five covariates, three of which are treated as main independent variables, and two as control variables.starting with the first group, ideological distance from the coalition mean is the distance between the ideal point of the minister's party in a onedimensional policy space and the seat share-weighted mean position of the parties represented in the cabinet.data on ideological positions were taken from the v-Party data set (düpont et al. 2021), specifically its economic leftright scale running from 0 to 6, where 0 corresponds to the far-left and 6 to the  far-right.second, in the absence of a party-specific portfolio salience measure for my country cases, i use an indicator from Whogov called "portfolio prestige," which classifies portfolios into three categories, High, Medium, and Low Prestige, based on their jurisdictions.For instance, portfolios such as defense, interior, and Finance are coded as High Prestige, portfolios such as agriculture, Housing, and transportation are coded Medium Prestige, while portfolios such as sports, tourism, and Family are coded as Low Prestige (Whogov online appendix, p. 66) 13 .third, for committee powers, i use an indicator developed by andré, depauw, and Martin (2016), which is based on nine items 14 .although the indicator was developed with parliamentary systems in mind, the items are perfectly applicable to and relevant to presidential systems.My coding of this indicator, based on constitutions, standing orders, and other laws regulating committee powers, can be found in appendix 5 for all eight country cases.Figure 1 displays committee powers over time for each country.Committee powers tend to be relatively stable over time, with most of the variation being the result of changes in the constellation of cabinet portfolios and committees that affect items 1 and 2. the major exception is Kenya, where a 2010 constitutional overhaul drastically altered committee powers.On the other end of the spectrum, committee powers remained constant over time in indonesia, with its meticulously fixed committee-portfolio "partnerships."interestingly, all of the cases with varying powers display upward trends, suggesting that committees tend to become more numerous and come to match the portfolio structure more closely over time.
the first of my two control variables is seat share, which is simply the number of seats held by the portfolio-holding party divided by the sum of the seats held by the set of cabinet parties multiplied by 100.data on the number of seats occupied by each party as well as legislature size came from a variety of case-specific official election sources.a party's seat share can affect its allotment of committee chairs, and parties with more chairs have greater leeway to appoint those to their portfolios, thus preventing those portfolios from being shadowed by their partners.seat share is also an indicator of a party's bargaining power within the coalition, which it can exploit to negotiate more favorable shadowing outcomes.second, age of democracy is operationalized as the number of years since the last time a country reached a positive Polity v score.this variable could affect shadowing for two reasons.First, a longer experience with democracy should be related to a greater degree of institutionalization of inclusive values and practices, affecting for instance the degree to which CCs are allotted to opposition parties (Carroll, Cox, and Pachón 2006, pp. 163-66).second, shadowing may be subject to a learning curve that takes time for political actors to understand and put into practice.returning now to the model specification, recall that we are dealing with two levels: the portfolio level (1) and the party level (2).Out of the independent variables, only portfolio prestige is located at level 1. ideological distance and seat share are strictly located at level 2. While committee powers and age of democracy are in reality located at the country level, in practice they are taken to lie at the party level, because the country level is not taken into account in this model.Now it also becomes clear why the random slope term u 1j is included: this is to account for potential between-party differences in the effect of portfolio prestige on shadowing.Parties may differ in the value they place on a given portfolio, and controlling a high-salience portfolio could affect their willingness to allow shadowing.
For ease of interpretation of the estimates, all continuous independent variables were grand-mean centered (sommet and Morselli 2017, p. 211).Portfolio prestige appears in the form of two dummy variables, "high prestige" and "low prestige," with "medium prestige," where most observations are concentrated, as the baseline.

Results and Discussion
the regression coefficients and the odds ratios are reported in table 3. Model 2 adds the interaction effects.i begin by examining the results associated with Hypothesis 1, which expects monitoring to increase as a party's distance to the coalition's weighted mean position increases.table 3 shows that the effect is significant at the 0.1 level in model 1 and comes with the right sign. the odds ratio of 1.90 in model 1 suggests that a unit increase in a portfolioholding party's distance from the weighted coalition mean is associated with a 90% increase in the probability of the portfolio being shadowed.Figure 2 shows the predicted probabilities of shadowing as a function of ideological distance in model 1. it reveals that every unit increase in ideological distance is associated with a 20% point increase in the probability of shadowing.Of course, the relative increase going from −0.5 to 0.5 is greater than that going from 0.5 to 1.5 (recall that the variable is centered on the sample mean).the former is associated with a jump from around 40% probability to around 56% (a 40% increase), whereas the latter is accompanied by a jump from around 56% to about 71% (a 27% increase).these effects are more modest than those suggested by the odds ratios.
When adding the interactions in model 2; however, the impact of ideological distance diminishes and loses its significance at the 0.1 level, while the interactions themselves are not significant.this could point to suppression: apparently, the interactions of distance with portfolio prestige and committee powers explain part of the variance in shadowing, which was previously explained only by distance.Hypothesis 1 holds, but with the caveat that some of the effect of distance is explained by its interaction with prestige and powers.
Moving on to Hypothesis 2a, which predicts shadowing to increase as a function of portfolio prestige, the log odds for high prestige come with the wrong sign and fail to reach significance.the odds ratio of 0.72 indicates that a high-prestige portfolio is, on average, 38% less likely to be shadowed than a medium prestige counterpart.On the other hand, low prestige comes with the expected sign and is significant at the 0.05 level.the odds ratio of 0.36 suggests that a low prestige portfolio is 2.78 times less likely to be shadowed than a medium prestige one.Figure 3 shows the predicted probabilities of shadowing for portfolio prestige.it suggests that the probability of shadowing increases from around 25% at low prestige to around 48% at medium prestige, to then come down to 40% at high prestige.the drop in shadowing probability from medium to low prestige amounts to a percentage decrease of about 48%, again more modest than the odds ratio suggests.this result leads me to accept Hypothesis 2a, with the caveat that the negative effect of prestige on shadowing is primarily driven by low prestige portfolios being less likely to be shadowed.the effect of prestige also appears to be curvilinear, but since the effects of medium and high prestige cannot be statistically distinguished from one another, this cannot be confidently affirmed.
Hypothesis 2b expects high salience portfolios to be more likely to be shadowed, especially when they are held by parties that are distant from the Figure 2 Predicted Probabilities of shadowing as a Function of ideological distance ideological coalition mean.this is borne out by the log odds of the interaction between ideological distance and high prestige, which comes with a positive sign.the odds ratio suggests that, for high prestige portfolios, every unit increase in ideological distance makes the portfolio 69% more likely to be shadowed, which is greater than the main effect of distance.this effect is not significant, however.For low-prestige portfolios, on the other hand, one unit increase in ideological distance reduces the probability of shadowing by a factor of 7. this result is counter-intuitive.even when prestige is low, we would still expect ideological distance to positively affect shadowing.a possible explanation could be that ideologically distant parties resist shadowing across the board, which they get away with in low prestige portfolios but not in medium or high-prestige ones. in any case, the lack of significance leads me to reject Hypothesis 2b.Finally, the effect of committee powers is highly significant at the 0.01 level and comes with a negative sign. the odds ratio of 0.68 suggests that a unit increase in committee powers makes shadowing 1.47 times less likely to occur.Figure 4 shows the predicted probabilities of shadowing as a function of committee powers and suggests every one-point increase in committee powers is associated with a decrease in the probability of shadowing of 10 percentage points, corresponding to a percentage change of around −18%.Furthermore, in line with the initial expectations, the interaction between committee powers and ideological distance comes with a positive sign, and the odds ratio implies that, when either variable is held constant, a one-unit increase in the other is associated with a 29% increase in the odds of shadowing.this suggests that the interaction with committee powers reinforces the effect of ideological distance, while the interaction with distance mitigates the negative impact of powers.this effect is insignificant, however.i will analyze these results in more detail after discussing the controls.seat share comes with the right sign and is highly significant, although the effect size is quite small: a unit increase in the seat of the party in charge of a given portfolio leads to an average decrease in the probability of the portfolio being shadowed of only 3%.Finally, age of democracy fails to reach significance and has an odds ratio very close to 1, meaning that democratic experience is largely irrelevant to the observed shadowing patterns.
the committee powers result lends support to the final prediction proffered in the "Committee powers" subsection, namely, that relying on strong committees for monitoring and reporting and institutional checks comes with its own set of agency loss risks as a result of the separation of powers.However, caution is warranted as there are alternative explanations that must be examined as well: 1. stronger committees may also allocate more influence to the minority or individual dissenting members, and 2. stronger committees may have relatively weaker chairs, meaning that control over the chair is not as useful.these explanations are examined in appendix 3, but no evidence is encountered in support of either of them.this does not mean that the prediction stands confirmed; however, due to the small number of countries it covers, the data are not very well suited to test it.Nevertheless, the committee powers result is interesting and warrants further investigation in a larger sample of countries, preferably one including both parliamentary and presidential systems.

Conclusion
a burgeoning literature on coalition governance in parliamentary systems has established that coalition partners govern together and manage delegation to one another's ministers through a variety of methods.the one that has attracted most attention is monitoring by junior ministers, committee chairs, and a host of legislative instruments.in presidential systems, however, where multiparty cabinets are now the norm, this issue has gained some traction but has so far been limited to case studies on Brazil.this is unfortunate because the greater centrality of legislatures in presidential systems means that the role of committees in shadowing is not only more salient but is likely to be characterized by unique dynamics, and therefore, deserves to be studied comparatively.My contribution to the literature is threefold.First, i approach committeebased shadowing in multiparty presidential systems comparatively, including emblematic country cases from three continents in my sample.second, i present original data on committee chairs in these systems and employ an existing measure of committee powers (andré, depauw, and Martin 2016) that i coded by hand based on my collection of data from legal documents governing legislative organization.third, i find that shadowing is positively associated with the ideological distance between a minister and the coalition, as well as with the salience of portfolios.
More concretely, i ask whether cabinet parties in presidential systems keep tabs on one another using committee chairs.i formulate three hypothetical effects we should see if committee chairs truly serve intra-coalitional monitoring and institutional checks functions.Based on a multilevel logistic regression analysis, i present three main findings related to these hypotheses.First, in line with the literature on coalition governance, i find that the ideological distance between a minister and the coalition compromise increases the probability that that minister is shadowed by a committee chair from another coalition party.second, like parliamentary studies, i find a significant effect of portfolio salience on shadowing.specifically, low salience portfolios are less likely to be shadowed than medium and high salience ones.third, i do not find a significant interaction effect of salience and distance, but including that interaction does reduce the significance of the main effect of distance to below conventional levels.additionally, in contrast to the parliamentary literature, i find that committee powers reduce rather than increase the incidence of shadowing.this result can potentially be attributed to the greater risk of agency loss between parties' cabinet representatives and congressional delegations under separation of powers, making cabinet parties more wary of committee-based shadowing when committees are stronger.However, this result must be interpreted with caution, as the small number of cases does not allow me to disentangle the effects of committee powers from those of idiosyncratic country-level differences.in sum, the results tentatively suggest that parties in presidential coalition governments employ committee chairs to shadow portfolios held by their partners.
One of the main implications of these results is that, under presidentialism, parties join coalitions (partially) for policy reasons and have an interest in assuring that the coalitions' agreed-upon policy objectives are implemented by their partners.
However, this study also suffers from a number of limitations.First, in the absence of qualitative evidence of shadowing, it must be kept in mind that the observed patterns of committee chair and portfolio allocation do not in itself constitute definitive evidence of shadowing by said committee chairs (Pukelis 2016).second, the study does not control for alternative shadowing devices that may be used, such as junior ministers, parliamentary questions, or interministerial committees, which could affect committee-based shadowing  (Carroll and Cox 2012).third, asymmetries in data access between countries and regions mean that the share of asian and especially african cabinets in the sample is smaller than what would have been ideal, with the potential to affect the generalizability of the findings.
Future research should, first, invest in data collection and aim to replicate a study of the kind at hand on a larger sample of countries and cabinets, moving beyond paradigmatic cases after my initial foray into the subject.second, it must investigate not only which portfolios and parties are more likely to be shadowed, but also which parties are more likely to do the shadowing.Finally, fine-grained case studies that employ interviews with legislators in one or more relevant countries must be conducted to shed light on how committee-based shadowing actually works in practice.
Data Availability Statement.the data that support the findings of this study are available on request from the corresponding author.the data are not publicly available due to privacy or ethical restrictions.
Joris Alberdingk Thijm <joris.thijm@ics.ulisboa.pt> is a PhD candidate at the Institute of Social Sciences, University of Lisbon.

ENDNOTES
the author would like to thank all individuals who provided invaluable comments and suggestions, including, but not limited to: the organizers and participants of the workshop "Presidents, governments, and Policymaking" at the 2023 eCPr Joint sessions in toulouse; Jorge Fernandes, andrés Malamud, and Octavio amorim Neto; and his colleagues at the institute of social sciences that provided feedback on a presentation of an earlier version of this article.Furthermore, the author thanks the Political science department at texas a&M university, and especially Michelle taylor-robinson, for hosting him while he wrote the revised version of the paper, and for the opportunity to present it to faculty and students (whose feedback was greatly appreciated).all remaining mistakes and omissions remain his sole responsibility.this research received funding from the Fundação para a Ciência e a Tecnologia, scholarship reference 2020.04539.Bd.
1. i use the term 'coalition governance' in the sense of Bergman, Bäck, and Hellström (2021), who define it as the middle stage of the coalition lifecycle and associate it with delegation among coalition partners.Coalition governance in this usage is not to be confused with 'coalition management', a term adopted by many students of coalitional presidentialism that refers primarily to the president's use of 'coalition tools' to attract and retain the (overt) support of coalition partners, rather than being concerned with reducing (hidden) agency loss to coalition partners (Bertholini and Pereira 2017;Chaisty, Cheeseman, andPower 2014, 2018;Pereira, Bertholini, and raile 2016;Praça, Freitas, and Hoepers 2011).
2. Of course, under presidentialism, the formal principal of the cabinet is the president.However, a president who chooses to govern through a coalition cabinet is per definition not governing alone.informally, we can then also treat the presidential coalition cabinet as a collective principal.
3. in the Philippines, chair selection happens simultaneously with the committee assignments on the proposal of the Majority, subject to confirmation by the plenary (appendix 5).
4.Not much later, the party canceled the withdrawal of the ministers, on the grounds that the problems between them and the party had been resolved, attributing the initial conflict to miscommunication between the ministers and the congressional delegation (rachmadi 2006).
5. 'sectoral Oversight Committees' (sOCs) were only established in 2016, with the only policy-related committees existing up to that point being the 'Ministerial Consultative Committees', which were chaired by the minister in charge of the corresponding portfolio, and therefore, hardly independent of the executive (tennakoon and Jayathilake 2021, pp.157, 160). it would have been possible to include the 2019 day-1 cabinet, were it not for the fact that v-Party lacked data on the then-presidential party, the sLPP.
7. see Chaisty, Cheeseman, and Power (2014, 2018). 8. this is what Levy (2008) calls the "inverse sinatra inference-if i cannot make it there, i cannot make it anywhere" (p.12).9. see appendix 1 for a list of the cabinets included in the study.10. in Benin, out of the two day-1 legislatures for which i had data access, 2015 and 2019, i had to exceptionally use non-initial cabinets, and ultimately had to drop 2019 as that year's elections saw the prohibition of all but two parties (Benin: Freedom in the World 2021 Country Report 2021), including all parties to which the sitting ministers were affiliated, without it being clear to which of the two legal parties they had switched if at all. a similar problem existed in Malawi, where the electoral court annulled the results of the 2019 presidential election, which had been held concurrently with the legislative election, but which now had to be redone (Phiri 2020).the president elected in the second election was only inaugurated in 2020, such that there is no link between committee chair allocation (which had taken place in 2019) and cabinet composition.
14. these are: (1) the number of committees, (2) the degree of correspondence of committee jurisdictions to ministerial portfolios, (3) the ability to compel ministerial attendance and testimony, and (4) civil servants' attendance and testimony, (5) the existence of permanent staff for administration, drafting, and research, (6) the ability to initiate legislation, and (7) rewrite bills whereby the committee version of the bill takes precedence over the original version in plenary voting, (8) the committee stage preceding the plenary stage, and, finally, (9) timetable control (andré, depauw, and Martin 2016, p. 111).

Figure
Figure 1 Committee Powers Over time, grouped by Country

Figure 3
Figure 3 Predicted Probabilities of shadowing as a Function of Portfolio Prestige

Figure 4
Figure 4 Predicted Probabilities of shadowing as a Function of Committee Powers 19399162, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/lsq.12451by Faculdade Medicina De Lisboa, Wiley Online Library on [23/02/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Coalition Governance under Separation of Powers risks.

taBLe 1
Chair selection Procedures Pereira and Mueller (2000)tate during the legislature.inthosecases, the chair serves for 1 year.Otherwise, the chair serves for the entire legislative period.bNotexplicitly stated in the standing orders, but empirically confirmed.cPereiraandMueller (2000).d alday (2022).e sherlock (2012).f Kasambara (2019).g Chasquetti (2016, p. 223).19399162, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/lsq.12451by Faculdade Medicina De Lisboa, Wiley Online Library on [23/02/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 19399162, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/lsq.12451by Faculdade Medicina De Lisboa, Wiley Online Library on [23/02/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Coalition Governance under Separation of Powers by coalition partners.research on parliamentary systems has shown that parties are more likely to employ junior ministers to shadow portfolios that are most salient to them (greene and Jensen 2016; Lipsmeyer and Pierce 2011; thies 2001 19399162, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/lsq.12451by Faculdade Medicina De Lisboa, Wiley Online Library on [23/02/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License 19399162, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/lsq.12451by Faculdade Medicina De Lisboa, Wiley Online Library on [23/02/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Coalition Governance under Separation of Powers 2017, p. 206).the question then becomes what higher order level to choose for this study.Countries are too few in number, as there are only eight.

taBLe 2
Coding of the Outcome variable, shadowingNote: parties a and B are cabinet parties, while party C is an opposition party.For simplicity, the table shows a situation in which one committee oversees the portfolio.

taBLe 3
Note: the second model adds the interactions, standard errors are reported in round brackets below the estimates, and odds ratios in square brackets below the standard errors.standard deviations are reported in round brackets below the random effects.Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/lsq.12451by Faculdade Medicina De Lisboa, Wiley Online Library on [23/02/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Coalition Governance under Separation of Powers 19399162, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/lsq.12451by Faculdade Medicina De Lisboa, Wiley Online Library on [23/02/2024].See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions)on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Coalition Governance under Separation of Powers