Political power sharing in post-conflict democracies: investigating effects on vertical and horizontal accountability

ABSTRACT
 While it may be necessary to secure elite buy-in to peaceful competition, the literature is pessimistic about the long-term effects of a power-sharing settlement on the quality of democracy. Designing institutions to guarantee political inclusion is commonly thought to undermine vertical and horizontal accountability by incentivizing rent-seeking over responsiveness to voters. This study employs data from the Varieties of Democracy project to test arguments about the pernicious institutional effects of political power-sharing settlements in post-conflict democracies, relying on a panel dataset of 28 conflict-prone states in Sub-Saharan Africa since the onset of democracy’s Third Wave (1990–2021). The analytical technique is a time-series linear regression distinguishing between upturns and downturns across a range of continuous measures of accountability. The results show that, in line with much of the literature, political power-sharing settlements are associated with increasing executive corruption and fewer improvements in the rule of law. However, none of the other proposed mechanisms linking political power sharing to poor accountability outcomes finds consistent or significant support in the cross-national sample. Overall, these findings suggest that the relatively undemocratic institutional concessions designed to resolve conflict may not pose the serious barrier to democratic deepening and consolidation as previously assumed.


Introduction
While political scientists continue to debate whether elections in divided societies are inherently destabilizing, a consensus has emerged that institutions may need to be engineered to manage competition in countries where sub-national groups have recently gone to war. This approach traces its origins to Arend Lijphart's seminal work on consociationalism 1a unique formula of institutionalized power sharing which helped to prevent ethnic conflict in four ideal-type democracies in Western Europe. 2 Since the onset of the Third Wave, multiparty elections have reached a greater number of conflict-prone societies than ever before, and so too has the universe of power-sharing democracies expanded, with the concept now straying far from Lijphart's consociational ideal. 3 Of 48 countries in Sub-Saharan Africa, 36 held open elections for the first time between 1989 and 1995 (75%), and 17 have exhibited some form of political power sharing in the period since (35%).
In the conflict literature, political power sharing generally refers to guaranteed inclusion in central decision-making for rebel representatives or their affiliated social groups, 4 thus corresponding to the consociational features of a grand coalition and proportionality. Among scholars and international brokers of a peace process, these solutions are viewed as strategic and politically expedient concessions intended to reconcile the liberal goals of peace building and democratization. Whether implicitly or explicitly, a presumed trade-off exists between stability and democratic quality, 5 with scholars variously arguing that political power-sharing bargains entrench the cleavages of conflict in society, facilitate elite capture of the political system, and encourage rent-seeking behaviour. 6 To date, however, cross-national research into the institutional effects of power sharing often relies on large samples including more stable and consolidated democracies. 7 The limited work that exists on developing democracies suggests that power sharing may have a more positive impact than is often assumed, 8 and yet, none have investigated the institutional effects in countries emerging from conflict, specifically. Instead, quantitative research on war-to-democracy transitions focuses overwhelmingly on the effect of power sharing on the likelihood of conflict recurrence, 9 with findings often suggesting that political bargains are less stable or effective than other sub-types, such as territorial or military power sharing. 10 This article draws on recent innovations in available data and statistical techniques in order to test theories about the effects of a political power-sharing settlement on the quality of democratic institutions. Focusing on conflict-prone countries in Sub-Saharan Africa 11 since the onset of the Third Wave, the study relies on disaggregated data from the Varieties of Democracy (V-Dem) project to examine variation over time in measures of accountability. Although the democratization literature highlights improved responsiveness and accountability in the African region over this period, 12 Figure 1 suggests that this trend masks substantial variation in democratic trajectories among post-conflict countries, which merits further exploration. To what extent can cross-national variation among Africa's post-conflict democracies be explained by political power-sharing institutions, and through what precise mechanisms?
The next section draws on existing research to develop testable hypotheses regarding the potential long-term effects of political power sharing on democratic institutions, taking care to differentiate between mechanisms of vertical and horizontal accountability. The third section describes the sampling strategy, statistical technique, and indicators used to operationalize key variables. The empirical results are presented in the fourth section. While the findings support the notion that political power sharing has been detrimental to executive corruption and the rule of law, none of the other commonly proposed effects on vertical or horizontal accountability are corroborated in the cross-national analysis. Moreover, transitional forms of power sharing, though short-lived and potentially prone to conflict recurrence, appear to be associated with significant improvements across a range of accountability measures. The fifth section concludes.

Testable hypotheses
Democratization in post-conflict settings poses unique risks, demanding careful consideration of priorities and trade-offs. 13 Except where conflict ends in secession, elections force belligerent parties to compete against each other at the ballot box, and because peace agreements often leave important issues at the heart of the conflict unresolved, the stakes of post-conflict elections are high. Meanwhile, given the continued presence of weapons and the recent memory of conflict, the opportunity cost of a recourse to violence remains relatively low for the likely losers. It is in this way that power-sharing bargains provide guarantees to contending elitesgroup leaders with the capacity to mobilize their followers for violencethat future power and security is not wholly dependent on garnering majority support. 14 Some have gone so far as to suggest that political power sharing is necessary to secure elite buy-in to peaceful forms of competition in conflict-prone societies. 15 Such is the nature of the implied trade-offnamely, between the short-term exigencies of stability and the long-term quality of democracy: Power-sharing institutions provide groups with the assurances necessary to encourage them to play by the electoral rules of the game … Power sharing is not inherently democratic. In fact, most elements of power sharing do not require democracy to function … [This] minimalist understanding of democracy stands in sharp contrast to broader definitions that highlight the importance of citizen participation and responsiveness. 16 This echoes the overwhelming consensus in the comparative democracy literature that inclusiveness and accountability are mutually exclusive goals, which vary in direct  (Coppedge et al. 2022) ranging from 0 to 1, which aggregates measures of electoral cleanliness and competitiveness, government responsiveness, and civil liberties. response to institutional choices, specifically between majoritarian and more proportional systems. 17 By prioritizing guaranteed inclusion over accountability in order to prevent conflict recurrence, power sharing is argued to "undermine the very mechanism through which elections can drive democratisation." 18 Thus, the goal of this section is to identify a clear set of expectations about the ways in which institutional guarantees for group representation may negatively impact a country's future movements along the spectrum of accountability.
At the same time, this literature increasingly acknowledges important differences between the types of institutional linkages that matter to a healthy functioning democracy by distinguishing between vertical and horizontal accountability. 19 The vertical dimension refers to electoral accountability, or the means by which citizens hold their government officials answerable for their conduct, including free and fair elections, political parties, independent media, and pluralistic civic associations. In contrast, the horizontal dimension refers to institutional checks and balances, or the degree to which state agencies impose effective constraints on the exercise of power by elected officials. This conceptualization provides a useful tool for elaborating the different theoretical mechanisms through which power sharing has been argued to undermine the quality of democracythe goal of this section.

Horizontal accountability
Relevant literature suggests that political power sharing has done little to facilitate improvement in checks and balances in African democracies. For one thing, since conflicts ending in power sharing reflect greater concessions to rebel demands than those that do not, such settlements are also more likely to include a blanket amnesty than retributive justice. 20 In Mozambique's General Peace Agreement (1992), for example, the comprehensive political bargain and the emphasis on reconciliation over justice reflected Renamo rebels' substantial bargaining power. Of the 16 final settlements included in the UCDP Peace Agreement dataset 21 involving political power sharing, 14 stipulate amnestya formal appeal to "forgive and forget"rather than the prosecution of offenses committed during conflict. Thus, with Vandeginste and Sriram, arguing that power sharing and transitional justice represent "a clash of paradigms," 22 political settlements may be unlikely to garner investment in judicial capacity and autonomyefforts that might ultimately improve the role of the courts in executive oversight over the long term. 23 Second, settlements attempting to improve inclusion of disadvantaged minorities over the long term most commonly do so in legislative institutions, such as proportional representation (PR) voting rules or seat quotas, with executive power sharing largely restricted to transitional coalitions which expire with post-conflict elections. Since real power in most African democracies "resides not with the legislature but with the president," 24 political power-sharing formulas are unlikely to improve constraints on executive abuses of power and, therefore, to mitigate the problem of super-Presidentialism which predominated in the early years of African democratization. 25 H 1A : Political power-sharing settlements are negatively associated with improvements in horizontal accountability.
An even more pessimistic perspective exists, suggesting that these kinds of bargains represent a politically-driven distribution of the spoils of state power and, therefore, provide "a tacit agreement on corruption." 26 A wealth of academic studies and policy reports have concluded that incentives and opportunities for corruption tend to increase during a post-conflict transition, particularly where international aid agencies create a "state of exception" in order to stabilize a liberal peace-building paradigm. 27 Power sharing may amplify this problem by guaranteeing that the leaders of contending factions have direct access to the levers of the state. Sierra Leone's Lomé Agreement (1999) granted the Chairmanship of the state's mining ministry to rebel RUF leader Foday Sankoh, making him "answerable only to the President of Sierra Leone." 28 After a political power-sharing agreement is reached, elites vie to establish sovereignty over the new institutional or territorial arenas allocated to them, providing new opportunities for rent seeking through the control or regulation of lucrative economic sectors, access to public funds and tax revenues, or the solicitation of campaign contributions. 29 H 1B : Political power-sharing settlements are associated with a decline in horizontal accountability.

Vertical accountability
Relevant literature suggests that political power sharing may constrain the development of electoral accountability, with such institutional formulas often referred to as a "vertically exclusive elitist equilibrium," "government by elite cartel," and "democracy on stilts." 30 By design, power-sharing democracies mitigate competition between the elites of belligerent groups in order to prevent conflict and as a result, the sanctioning mechanism of electionsthe risk that elected officials will be punished for poor performancemay be undermined if vote share has a negligible impact on a pre-determined formula. 31 Under such conditions, elected officials are dis-incentivized from building linkages with voters or investing in governing capacity and public good provision. 32 Various case studies of Mozambique have shown that politics continue to be dominated by elite-level bargaining outside of formal democratic processes decades after a power-sharing settlement ended the civil war. Despite substantial international investment in its "donor darling," 33 the country's two main political parties have failed to develop a meaningful presence outside of the capital, Maputo, while former rebel Renamo leader Afonso Dhlakama continued to represent the opposition in negotiations around major political decisions, despite not holding elected office, long after the end of the conflict. 34 H 2A : Political power-sharing settlements are negatively associated with improvements in vertical accountability.
Critics have proposed three further mechanisms through which the effects of power sharing may be more deleterious than merely impeding institutional development. First, unlike in centripetal electoral systems designed to encourage cross-ethnic voting, power-sharing democracies are deliberately designed to preserve intra-ethnic voting blocs, 35 which may incentivise candidates to replace performance-based policy platforms with parochial appeals to identity and investment in patronage networks. 36 Second where power-sharing bargains are used to overcome periods of electoral uncertainty and potential crisis, such as Kenya's 2007 episode of post-election violence, they are often perceived as betrayal or "brand dilution" by voters. 37 The result may be voter apathy and declining participationas attributed in the media's reporting of historically low levels of voter registration and turnout in Kenya's 2022 general electionor even the "cascading abandonment [of major parties] by their core constituents" and the rise of anti-democratic parties. 38 Third, an emerging literature associates post-conflict power sharing with an "exclusion-amid-inclusion problem," highlighting the increasing marginalization of social groups such as women, migrants, or other non-violent minorities not party to the settlement. 39 For example, a 60-40 proportionality rule is used to allocate political power between Hutus and Tutsis in Burundi but fails to provide explicit space for the country's small Twa minority. This section has drawn on relevant literature to develop testable expectations regarding the effects of post-conflict power sharing on democratic institutions. While political power sharing may be necessary to convince belligerent parties to compete peacefully in elections, a wider lens suggests that guaranteed representation may undermine vertical and horizontal accountability beyond the transition period. A range of theoretical propositions have been considered, suggesting that not only do such settlements have the potential to impede democratic deepening, but even more worryingly, that there are reasons to expect some indicators to have deteriorated among post-conflict power-sharing democracies. The next section describes the data and method used to test these expectations.

Data and method
The dataset includes all countries in Sub-Saharan Africa that experienced an episode of armed conflict since independence, 40 as identified using the UCDP Armed Conflict and Non-State Conflict datasets. 41 The threshold for inclusion is 25 battle-related deaths in a single conflict-year, either against the state or between sub-national groups. Starting the selection criteria period at independence excludes anti-colonial wars, and as the theoretical focus is on the institutions that mediate competition between groups sharing the same political space, I also exclude border disputes, violent attacks attributed to hardliner terrorist groups, 42 and brief coup-related violence. 43 Of 48 countries in the region, this strategy yields a sample of 28 conflictprone cases.
There are a number of advantages to restricting the scope of the analysis to Sub-Saharan Africa, such as regional similarities in colonial experience, time since independence, and the onset of democracy, which improves comparability. Prior to 1990, none of Africa's conflict-prone states met the minimum procedural definition of electoral democracy, yet nearly all had held their first multiparty elections by 1995. This mitigates some of the potential for omitted variables to bias the results, while also suggesting that early institutional choices may have had a greater impact on mechanisms of accountability than in other parts of the world. Meanwhile, a recent literature has emphasized the greater degree of variation that exists, for example, in the strength of legislatures, party institutionalization and composition, and electoral system type than was recognized in early work on African democratization. 44 All of this suggests that the region provides a fertile testing ground for theories of the effects of power sharing on post-conflict democratic trajectories. The panel dataset provides coverage of these 28 countries since 1990, the onset of the Third Wave, for a total of 875 country-years.

Statistical technique
The big debates about democratization focus on explaining regime transitions, with cross-national studies using categorical measures of the dependent variable. 45 In light of the goal of this study, I avoid a dichotomous or trichotomous operationalization of democracy, which would rely on arbitrary thresholds and risk separating only marginally dissimilar cases into discrete categories. Consistent with the competing various logics elaborated in the previous section, I acknowledge that movements towards and away from democracy are likely to have different causes. The analysis therefore relies on a regression technique elaborated by Jan Toerell, 46 which distinguishes between positive and negative movements along a continuous dependent variable.
The basic regression model is as follows, where i = country and t = year: On the right-hand side of the equation, X represents a vector of explanatory variables measuring political power sharing as well as a range of controls that may have an independent effect on both the level of accountability and the signing of a power-sharing settlement. Each time-variant variable in X is lagged one year in order to avoid endogeneity. As is standard, the model also includes a lagged value of the dependent variable, Y i,t-1 , on the right-hand side of the equation, which provides a further check against endogeneity bias by controlling for the possibility that Y i had an effect on the variables in X i prior to the year t-1. 47 It also provides a proxy control for any determinants of Y that are unintentionally omitted from the model, thereby reducing residual error. 48 Next, in order to distinguish between the causes of improvements and depreciations -"upturns" versus "downturns" in accountabilitythe model is analysed separately against cases in which ΔY is positive and those in which it is negative. In other words, a year-on-year improvement in accountability means that Y i,t -Y i,t-1 > 0, while a decline means that Y i,t -Y i,t-1 < 0. When the model is testing upturns, all cases where ΔY < 0 are set to 0, and vice versa when testing downturns. This yields the following amended regression equation for testing upturns in accountability: And the following for testing downturns: Holding the explanatory model specified by vector X constant, the equation is run against a range of disaggregated measures of accountability (Y ) in order to investigate which mechanisms might best explain the impact of power sharing, if any.

Dependent variables
The data employed to operationalize outcomes of interest come from the Varieties of Democracy (V-Dem) dataset v12. 49 The primary measure of horizontal accountability is the multivariate Horizontal Accountability index, a continuous indicator ranging from 0 (low) to 1 (high), which aggregates nine discrete variables quantifying "the extent to which state institutions hold the executive branch of the government accoun-  50 As the theoretical logic underlying H 1A suggests two mechanisms through which power sharing may impede upturns in horizontal accountability, the individual indices for Judicial and Legislative constraints on the executive are also included. 51 Theories connecting power sharing to downturns in horizontal accountability (H 1B ) emphasize the perverse incentives that encourage rent-seeking behaviour among elites. The dataset includes three additional indicators to test this logic in different ways. Most broadly, V-Dem's Rule of law index captures the extent to which government officials comply with the law in the exercise of authority, combining 14 discrete variables that measure the extent to which laws are enforced consistently and transparently across various branches of government without bribery or corruption. In order to zoom in closer on the rent-seeking mechanism, the Regime corruption index measures the extent to which political actors in all government branches abuse their positions for private gain, 52 while the Executive corruption index measures the extent to which members of the executive branch accept bribes or embezzle or misappropriate funds. 53 Vertical accountability is measured using the multivariate Vertical accountability indexagain, a continuous indicator ranging from 0 (low) to 1 (high)which quantifies "the ability of the population to hold government officials accountable through elections and political parties." 54 The Party institutionalisation index is used to test the primary mechanism behind H 2A , combining six variables which capture systemic coherence, organizational linkages with the electorate, and programmatic policy platforms. 55 Scholars propose three different mechanisms through which power sharing may produce a downturn in vertical accountability. First is the notion that power sharing, as an uncompetitive form of government by elite cartel, causes disengagement among voters. To test this, I include the Participatory component index, which measures the degree of citizens' active participation in political processes and civil society, as opposed to delegation. 56 Second, to test the argument that power sharing entrenches ethnicity in politics by encouraging patronage and parochial campaign messaging, the dataset includes the index of Clientelism. 57 Finally, the "exclusionamid-inclusion" mechanism is tested using the Women's political empowerment index and the Equal access index, 58 which capture the degree to which various social groups have equal opportunity to participate and affect policy decisions.
Like the primary indices of Horizontal and Vertical accountability, all outcome variables are scaled from less (0) to more democratic (1), with the exception of measures of corruption and clientelism, which range from normatively better (0) to worse (1). The results are interpreted accordingly.

Independent variables
As the analytical focus of this study is the effect of political power sharing on postconflict institutions, I relied on the University of Notre Dame's Peace Accords Matrix (PAM) to identify settlements for several reasons. First, unlike other datasets, the PAM disaggregates between provisional sub-types in a settlement, allowing me to distinguish political bargains from those involving territorial, military, or economic power sharing. Second, unlike UN Peacemaker and the University of Edinburgh's PA-X database, the PAM focuses only on comprehensive, final settlements involving major parties to the conflict. 59 This excludes less substantive accordssuch as a series of partial agreements in Central African Republic (2011Republic ( -2019 and South Sudan (2016-2019), for examplewhich are unlikely to have a long-term impact on constitutional design. Finally, the PAM provides extensive information about the implementation of key provisions on an annual basis. The MPLA government in Angola reached an agreement with UNITA rebels to decentralize power in 1994 and again in 2002, yet failed to pass legislation on decentralization until 2008. I can therefore ensure that the coding captures institutionalized reforms rather than merely an agreement to share power, increasing confidence that any correlation between a political bargain and accountability is not spurious. Table 1 provides a complete list of countries and settlements in the sample.
Recent research has demonstrated the important variation that exists among political power-sharing settlements, from transitional pacting arrangements during the implementation period, to a more robust and permanent overhaul of the political system exhibiting features of consociationalism. Although existing arguments about power sharing's institutional effects are not explicit about the type of political bargain, there is reason to believe that the two sub-types generate different structures of incentives and opportunities and, therefore, have different effects on the behaviour of group elites. The transitory nature of inclusive pacts may heighten the potential for conflict recurrence, 60 for example, but have little direct impact on the long-term trajectory of democratic accountabilityand yet these kinds of ad hoc, short-lived formulas are often the focus of power-sharing's critics.
I therefore include two discrete measures of political power sharing. Consociational refers to constitutionalized reforms guaranteeing representation in institutions for central government decision-making over the long term, such as PR electoral rules, seat quotas, rotating presidencies (e.g. Comoros), or proportionality in allocating cabinet ministries. In contrast, transitional refers to pacting arrangements designed to expire with post-conflict elections, such as unity governments or inclusive governing coalitions. Relying on PAM, both variables are coded as "1" for all years in which the institutions are in place and "0" otherwise. As a concrete illustration, South Africa is coded "1" for transitional power sharing from 1993 to 1995, while an interim constitution provided the terms for an inclusive Government of National Unity, and consociational power sharing in all years after the 1996 constitution established an electoral system based on party-list PR at all levels. 61 By design, the consociational label employed here is more lenient than that typically implied in comparative politics, as it is intended merely to capture constitutional changes where the IR conceptualization of power sharing overlaps with a Lijphartian approach to conflict prevention. While this limits the degree to which the current study speaks to the literature on consociationalism, specifically, this operationalization improves internal validity by (a) differentiating between an overhaul of the rules of political competition and the much lower threshold of transitional stabilizing measures, 62 while (b) capturing wide variation in the nature and design of constitutional reforms in the African sample. 63 A number of additional variables are included in vector X to control for factors that may have an independent effect on either the quality of democracy or the likelihood of reaching a power-sharing settlement, or both. First, the model includes three proxies of modernization, 64 chosen both for their prevalence in the literature as well as optimal coverage of the relevant sample: (a) GDP per capita, measured in constant 2015 US$, Since decentralization mitigates the lucrative nature of power sharing at the national level, 65 V-Dem's federal-unitary index is included to capture the degree to which power is territorially devolved. While quantitative findings have been mixed, it is also common to control for macro-structural variables which impede governance. 66 The vector X includes measures of total country Area in square kilometres and Population size (WDI), as well as Fearon & Laitin's measure of Mountainous terrain. 67 Moreover, as resource dependence reduces the government's reliance on its population for tax revenues, thus dis-incentivising responsiveness and providing rents for corruption and patronage, 68 I include a measure of natural resource production as a percentage of GDP.
Although the field has largely moved on from its pre-occupation with ethnic fractionalization, 69 highly polarized societies may be more prone to conflict and more likely to reach consociational settlements, so the model includes Montalvo and Reynal-Querol's measure of Ethnic polarization. 70 As there is some evidence that former British colonies score higher on measures related to governing capacity and civil society, 71 a dummy variable for British colony is coded as "1" if the country was under British administrative rule at the end of the Second World War and "0" otherwise. To control for the effects that more intense and protracted conflicts have on state institutions, the vector includes a dummy for Civil war, coded as "1" if conflict reached the threshold of 1,000 total battle deaths. 72 Finally, I include a dummy variable coded as "1" if any other peace agreements were signed in the previous five years, relying on the quality of coverage of Edinburgh's PA-X database. The model specification of vector X is held constant throughout the analysis, while varying the outcome measures of accountability. Certain variables which are common in explanatory models of democratization, such as diffusion effects and regime age, have been excluded here due to lack of variation in the African sample. The Online appendix provides a full description of variables, their sources, and summary statistics.

Addressing sources of bias
It is possible that political power-sharing settlements to conflict are more likely in countries facing greater barriers to governance, those with social cleavages which independently undermine accountability, or even relatively more democratic countries which have less room for improvement on the outcome indicators. 73 In order to address these potential sources of bias, a simple difference-in-means test compares the baseline subset of countries which do not get power sharing to two alternative categoriesthose exhibiting any form of political power sharing in the time period, or those which adopt consociational reformsusing a t-test of the mean values at the beginning of the panel (1990) across a range of variables. The results indicate that there are no statistically significant differences across these subsets in terms of accountability scores, population density, terrain, income level, or ethnic salience prior to the onset of power sharing (see the Online appendix). The only exception is that countries which adopt consociational solutions appear less dependent on natural resources, although the difference barely meets the lowest level of significance (p < 0.01). Unfortunately, it is not possible to eliminate sources of endogeneity that emerge over time within the sample. Incumbents who sense their authority and governance capacity slipping may be more likely to offer power-sharing concessions in an attempt to forestall instability, for example, and declining accountability in subsequent years may reflect this antecedent trend. While careful attention to measurement of proposed causal mechanisms somewhat helps to address this problem, should the statistical findings support the hypotheses elaborated in the second section, then it may be necessary to investigate the possibility of reverse causation.

Discussion
The results of the linear regression analysis of the effects of a political power-sharing settlement on horizontal accountability are provided in Table 2, with panel corrected standard errors in parentheses (full results in the Online appendix). Where consociational settlements have been implemented, the results lend some support to the hypotheses elaborated in the second section, however the negative effect of power sharing is not strong or consistent across all outcome indicators. In fact, there is no clear relationship between consociational power sharing and year-on-year changes on the overall HA index, positive or negative. Rather, the effect is limited to just two mechanisms of horizontal accountabilityall other variables held constant, consociational power sharing has a significant, negative association with improvements in the rule of law in Sub-Saharan Africa (H1 a ), as well as a significant, positive association with increasing levels of executive corruption (H1 b ).
In countries such as South Africa, Burundi, and Mozambique, political reforms designed to reduce the competitiveness of national elections has combined with demographic realities to produce an entrenched ruling majority largely unthreatened by the possibility of electoral defeat. As a result, a culture of "systemic impunity" has flourished in executive ministries responsible for public procurement contracts and forestry and mining concessions, for example. 74 Not only are post-conflict power-sharing democracies "on average associated with higher aggregate levels … of corruption" than post-conflict countries without power sharing, 75 but the cross-national evidence presented here suggests that constitutionalized power sharing may actively enable corruption and act as a barrier against development of the rule of the law, thus corroborating theories of political power sharing as a "tacit agreement on corruption." 76 On the other hand, not all types of political bargains appear to have the same institutional effects. During years in which transitional power-sharing institutions are in place, countries appear to exhibit significant improvements across measures of horizontal accountability, especially judicial constraints on the executive and the rule of law, as well as the overall HA index ( Table 2). This contradicts recent scholarship suggesting that the provisional nature of pacting arrangements creates short-term time horizons, thereby increasing incentives for rent-seeking during the transitional period. 77 The positive association with constraints on the executive is especially noteworthy in light of the academic community's growing awareness of the role of the courts in preventing authoritarian backsliding. 78 Thus, even ad hoc agreements to co-opt rebel leaders into the levers of state powersuch as the widely-criticized awarding of lucrative cabinet positions to rebel leaders in Liberia and Sierra Leone 79may increase the likelihood of democratic consolidation after transitional arrangements are dissolved. Table 3 shows the results of the linear regression analysis of the relationship between political power sharing and vertical accountability (VA), with panel corrected standard   Note: Panel corrected standard errors are in parentheses. The dependent variable in each model is the annual change in the accountability measure, distinguishing between upturns (+) and downturns (−), and including one lagged measure of the dependent variable on the right-hand side of the equation. The controls in all models are the following variables: territorial decentralization (V-Dem), country size (square kilometres, log), population (log), ethnic polarization, percent mountainous range (log), GDP per capita in constant 2010 US$ (log), infant mortality rate (log), GDP per capita annual growth rate, natural resource production as percentage of GDP, former British colony (0/1), civil war threshold met (0/1), and other peace agreements signed in previous five years (0/1). All time-variant independent variables are lagged one year. *** p < 0.01, ** p < 0.05, * p < 0.1. errors in parentheses (full results reported in the Online appendix). Initial results do not lend strong support to either H2 a or H2 b . In fact, consociational reforms have a significant, positive association with improvements in the overall VA index, although the magnitude of the effect is small, meeting only the lowest level of statistical significance. As the effect does not align with any of the expectations derived from existing literature (section II), the mechanism through which vertical accountability improves after the implementation of consociational reforms remains unclear (Table 3). Upon review of the components of the multivariate VA index, it is possible that this subset of cases exhibits greater improvements in the quality and cleanliness of elections, perhaps due to the fact that manipulation is unnecessary where power-sharing reforms reduce the uncertainty of electoral outcomes. As the current analysis relies on year-on-year changes in the outcome, while indicators of the quality of elections are measured and vary only in election years, it is not possible to test this expectation with the current data, but this is a worthy area for further exploration.
Notably, the cross-national evidence presented here does not confirm the exclusionamid-inclusion problem often attributed to post-conflict power sharing, although again there are important limitations to the data. On the one hand, the results demonstrate that the representation of women has not been impeded, at least in the African sample, and moreover that settlements involving transitional power sharing are associated with significant improvements in women's political empowerment. This contradicts the hypothesis derived from an existing literature which focuses on the potentially unique case of Northern Ireland, while corroborating a small body of recent work suggesting that the "logic of accommodation on which power sharing is based" may contribute to higher levels of participation and engagement among groups that have previously been marginalized, such as women. 80 Tripp notes that the active involvement of women's NGOs in substantively comprehensive peace processessuch as in Liberia and South Africaallowed female candidates to build networks and political capital, leading to an increase in national-level representation of female politicians in the post-conflict period. 81 On the other hand, the result on the equal access measure is more difficult to interpretwithout data measuring inclusion at the group level, it is impossible to determine whether increased representation for parties to the settlement may be "cancelled out" by increased marginalization of other minorities. Again, this merits further investigation.
In sum, the results of the cross-national analysis suggest that much of the pessimism in the literature on political power sharing is unfounded. Compared to the rest of the sample of post-conflict democracies in Sub-Saharan Africa, consociational reforms appear to worsen executive corruption and undermine improvements the rule of law, as expected. However, none of the other proposed mechanisms regarding horizontal accountability, such as the incompatibility between political power sharing and judicial reform, 82 are corroborated here. Regarding the somewhat surprising positive association with improvements in vertical accountability, the findings align with other quantitative studies that have shown power-sharing institutions to be beneficial to processes of democratization in the developing world more broadly. 83 Finally, where negotiated settlements involve the inclusion of rebel representatives in transitional coalitions, although these transactional bargains have been the focus of some of power sharing's most vocal critics, the data point to significant improvements across various measures of accountability in the years leading up to post-conflict elections, including constraints on the executive, social inclusion, and participation.   Note: Panel corrected standard errors are in parentheses. The dependent variable in each model is the annual change in the accountability measure, distinguishing between upturns (+) and downturns (−), and including one lagged measure of the dependent variable on the right-hand side of the equation. The controls in all models are the following variables: territorial decentralization (V-Dem), country size (square kilometres, log), population (log), ethnic polarization, percent mountainous range (log), GDP per capita in constant 2010 US$ (log), infant mortality rate (log), GDP per capita annual growth rate, natural resource production as percentage of GDP, former British colony (0/1), civil war threshold met (0/1), and other peace agreements signed in previous five years (0/1). All time-variant independent variables are lagged one year. *** p < 0.01, ** p < 0.05, * p < 0.1.

Conclusion
Although international actors promote a liberal peace-building paradigm which views multiparty democracy and conflict resolution as mutually reinforcing, electoral risks may be too high for opposition elites, especially those with opportunities to resist using violence. To achieve elite buy-in to elections, it may be necessary to sacrifice some democratic purity for institutional guarantees which mitigate the need to mobilize majority support. In this way, political power sharing is often viewed as a necessary evil, sacrificing democratic accountability and responsiveness for guaranteed representation and inclusion. 84 However, the findings reported here suggest that such political concessions may be less detrimental to democratic institutions than previously thought. Among Africa's post-conflict democracies, political power sharing has been associated with increasing executive corruption and fewer improvements in the rule of law, as is consistent with existing literature. On the other hand, none of the other mechanisms of horizontal or vertical accountability appear to be negatively impacted by the implementation of a political power-sharing settlement over the long term. In contrast to arguments that sunset clauses are necessary to combat the harmful institutional effects of political power-sharing settlements, 85 therefore, the results of this study suggest that such effects may be case-specific and limited to places where anti-democratic elites have made the successful transition into politics.
This article reinforces calls for increasing conceptual disaggregation of postconflict power sharing and for greater attention to the mechanisms linking cause and effect. First, the results speak to the importance of distinguishing between institutions of power sharing beyond the common four-fold categorization in IR, building especially on recent work which sheds light on the different institutional functions of power sharing between adoption and implementation and between the transitional period and constitutionalized reforms. 86 Second, despite a general consensus about the pernicious effects of political power sharing, a larger sample points to variation in the degree to which different mechanisms of accountability are likely to be affected over the long term by the strategies used to convince rebels to demobilize in the short term.
It is important to address some remaining limitations to the present research agenda. First, the positive result for transitional power sharing is perhaps not as surprising as it would appear, since country-years with unity governments are also those associated with major regime change and investment from international actors. It is not possible to analyse here whether the significant improvements in accountability observed in these years have more sustainable, long-term effects after post-conflict elections are held and transitional power-sharing formulas expire, as this would introduce multicollinearity into the models. Second, the analysis relies on a measure of key independent variables combining settlement provisions and implementation. Some previous research suggests that political power sharing is more prone to conflict recurrence than other subtypes of power sharing, especially transitional bargains, and this may ultimately undermine the implementation of settlement provisions. 87 In other words, political power sharing may negatively impact democracy indirectly, because the recurrence of conflict derails elections, and not through the damaging effect of the institutions themselves. Both of these are worthy questions for future research.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Data availability statement
The author confirms that the data and .do files are available in the Harvard Dataverse at https://doi. org/10.7910/DVN/ZBBMJX. The Online appendix provides descriptions and summary statistics for all variables. These data were primarily derived from the following resources available in the public domain: the Varieties of Democracy (V-Dem) Project (www.v-dem.net, see Coppedge et al. 2022) and World Development Indicators (databank.worldbank.org).