## Abstract

Why does fighting recur following some civil conflict peace settlements, but not others? What kind of agreements are associated with more enduring peace? Post-conflict elections can often complicate and even undermine peace agreements. Agreements that contain “electoral participation provisions,” however, may help stabilize settlements and produce more enduring peace. Electoral participation provisions mandate that rebel groups be allowed to compete alongside the government in post-conflict elections. Such provisions encourage external actors, such as intergovernmental organizations and foreign donors, to become engaged in post-conflict elections. As part of this engagement, they can provide incentives to the parties to adhere to the terms of the settlement, as well as detect and sanction instances of noncompliance. New cross-national data suggest that conflict after peace settlements recurs less often when electoral participation provisions are included than when they are not. The data also suggest that this pacifying relationship holds when combatants expect international engagement.

## Introduction

Should elections be held when civil conflicts end? Does allowing combatants to compete as political parties, an increasingly common aspect of negotiated settlements,1 reduce the chances that conflict will recur? This article seeks to assess the effect of these post-conflict elections on the likelihood of enduring peace.

Negotiated settlements that end civil conflict and bring enduring peace are difficult to design—and the stakes are high. Fighting recurred in 40 percent of civil conflicts from 1975 to 2005 that ended with a peace agreement. Up to 90 percent of civil conflicts since 2000 have been linked to the recurrence of earlier conflicts.2 Moreover, civil conflicts have produced millions of deaths since 1945, five times as many as have interstate wars.3

Policymakers and scholars are divided on the usefulness of post-conflict elections. Those on the ground in many cases support such elections as an important part of peace processes. Combatants and democracy activists frequently push for elections in the final stages of negotiating settlements, while United Nations (UN) and other intergovernmental officials often supervise such contests in the countries in which they operate.4 One of the earliest examples of such support is South Africa, when founders of the Umkhonto We Sizwe, the armed wing of the African National Congress, demanded in 1985 that the government “treat us like a political party” and provide in the constitution for open participation.5 Not long after, the United Nations helped conduct elections in nearby Namibia following the signing of a negotiated settlement.6 Some analysts also support holding post-conflict elections as a means to promote democracy.7

Many scholars who study post-conflict states, on the other hand, posit that elections may lead to renewed fighting, particularly when they are part of rapid democratization processes in states with weak institutions.8 The uncertainty inherent in many elections about how power will be distributed can destabilize efforts to incorporate formerly warring parties.9 Holding post-conflict elections, therefore, could be riskier than distributing power by more certain means, such as establishing territorial strongholds for warring factions or institutionalizing fixed-formula representation of groups in the new government.

In this article, I argue that the existing debate has overlooked an important set of post-conflict elections: those that enable former rebel parties to participate alongside government parties. “Electoral participation provisions”—clauses mandating that rebel parties compete alongside government parties in post-conflict elections—are a common feature of peace agreements. After the end of the Cold War in 1989, combatants negotiated the inclusion of electoral participation provisions into a number of settlements in seemingly intractable conflicts, including those in El Salvador, Mozambique, and Bosnia and Herzegovina. Since then, such provisions have appeared in almost half of the more than 100 settlements that have been signed by combatants. This article presents and tests a theory of these provisions' consequences, showing that, although post-conflict elections as a whole do not produce peace,10 electoral participation provisions are associated with an 80 percent increase in the odds that peace endures.11

I argue that electoral participation provisions facilitate external engagement to enforce compliance with negotiated settlements; in doing so, they help overcome commitment problems and contribute to enduring peace. Commitment problems reflect each side's fear that its opponent will renege on a mutually beneficial agreement to increase its benefits during a phase in the peace process when that opponent is relatively strong. Such commitment problems are ubiquitous and may cause settlements to fail.12 Long-term armed intervention that guarantees the terms of the agreement through force, considered common wisdom for how to overcome commitment problems, is often costly and therefore not thought to be credible to combatants. A more credible mechanism for external engagement, however, involves the expansion of democracy-promotion programs since the Cold War,13 which has increased the ability of international actors to engage in elections in many states, including post-conflict states. Electoral participation provisions aid in the enforcement of negotiated settlements because they establish cycles coinciding with electoral processes—for example, political party registration and campaigns—that coordinate actors' efforts at detecting and sanctioning noncompliance with settlements. Increasing external engagement in these processes, then, offers a mechanism for international actors to monitor and enforce combatant compliance. This theory of external engagement implies that electoral participation provisions, together with democracy-promotion programs, produce less precarious settlements and more enduring peace between signatories.

To test the explanatory power of external engagement theory and competing theories,14 I collected and analyzed the text of all 110 civil conflict settlements negotiated from 1975 to 2005 to identify which ones contain electoral participation provisions. I also developed measures to capture combatants' expectations of external engagement through the electoral process. Using these measures, I analyzed conflict recurrence between the signatories. In addition, employing both cross-national data and case evidence, I examined whether the electoral participation provisions were implemented, as well as whether external actors engaged in the elections.

I find that peace endured between signatories in 79 percent of civil conflict settlements with electoral participation provisions (33 of 42 cases), a considerably better outcome than for settlements with other provisions, where peace endured in only 44 percent of settlements (30 of 68 cases). Robustness checks confirmed these results. Further, the case of El Salvador, where the 1992 peace agreement provided for the Farabundo Martí National Liberation Front (FMLN) to compete in the 1994 elections, offers evidence consistent with the mechanism of external engagement.

This article proceeds as follows. The first section situates electoral participation provisions in the context of the literature on peace agreements. The second section introduces the theory of how electoral participation provisions can overcome crucial commitment problems to settle civil conflicts through external engagement. It also specifies testable implications that distinguish this theory from alternatives. The third section tests quantitatively the initial implications of my theory using new cross-national data. The fourth section tests the mechanism qualitatively using case evidence from El Salvador. The article concludes with a discussion of the implications of the theory for post-conflict elections and external enforcement.

## Commitment Problems and the Puzzle of Elections

Common bargaining approaches to civil war termination posit that signing a settlement is beneficial for combatants because fighting is typically costly.15 If, however, the combatants do not consider fighting sufficiently costly—for example, when they can pass the costs of fighting on to other actors or when they could lose social support if they stop fighting—they are unlikely to settle. An example is India since 2004, where the central state redistributed resources to regional governments to counter Maoist insurgencies, producing benefits and thus incentives at the local level to continue fighting.16 When outside sponsors or civilian supporters of the combatants tire of the conflict, it may become “ripe” for settlement, making a peace agreement more likely.17 But even with this scope condition met, if one side fears defection by the others, the settlement may remain precarious.18

Commitment problems complicate the ability of settlements to produce enduring peace.19 If one side is even temporarily weaker relative to its opponent during the peace process, that opponent has an incentive to grab more power than provided for in the settlement. Even the possibility of this happening may cause one side to defect because it believes that its opponent will do so. For example, a rebel group might interpret a slight delay or deviation in integrating the group into the state's power structure as a signal that the government is seeking to change the rules to allow it to maintain more power; the delay, however, might have been inadvertent, the result of low implementation capacity or a different understanding of the timeline. But in either case, the rebel group's concerns may drive the group to back out of the peace process. Commitment problems can thereby make peace precarious in several ways: (1) by incentivizing surprise attacks before disarmament, (2) by returning to the battlefield targets of reneging opponents (whose violations may be violent or solely political) to protect their share of power or to punish those seeking to reduce it, or (3) by driving those who fear such reneging to do the same preemptively.20 Commitment problems, therefore, have both an informational dimension and an incentive dimension.

Establishing a new government in the wake of a civil conflict that distributes power between the warring factions such that the payoffs from peace at least match those that the combatants could expect from continued fighting is difficult given the likelihood of commitment problems. Simultaneity and precision are hard to achieve, so even incremental processes of change (such as demobilizing units from each side in alternation or registering supporters of each side to vote) are inevitably accompanied by moments of weakness on each side.21

For a settlement to successfully end a civil conflict, therefore, commitment problems must be overcome. Sharing power is sometimes proposed as a mechanism to provide mutual benefits and reduce the marginal gains for each side from reneging on a peace deal.22

In some cases, elections, especially those with electoral participation provisions, may paradoxically complicate the ability of the parties to share power. Indeed, compared to fixed-formula power sharing, elections may not provide the same certainty that the sides will indeed share power.23

However, electoral participation provisions can be adjusted to provide some certainty.24 For example, in cases in which the core conflict includes an ethnic component, ethnic quotas can ensure that power distributed through voting matches each side's expected distribution of power from fighting. And importantly, many settlements with electoral participation provisions, such as those in Burundi and Bosnia and Herzegovina, as well as Colombia's peace agreement with indigenous fighters in the early 1990s, contained quotas or other mechanisms to guarantee the distribution of power.25 At the same time, however, by reducing the role of voters, such adjustments may make elections less normatively appealing. So why include electoral participation provisions in negotiated settlements?

## Electoral Participation Provisions and Enduring Peace

In this section, I discuss the importance of international actors in helping combatants overcome commitment problems to secure enduring settlements.26 Next, I introduce my theory of external engagement, which depends on electoral participation provisions. I then consider the theory's implications for enduring peace and those of some alternative theories.

### CREDIBLE COMMITMENTS AND OUTSIDE ACTORS

As discussed earlier, to ensure an enduring commitment to a peace agreement, the parties must expect noncompliance to be costlier than compliance. For example, if the government seeks to change laws before former rebels are integrated into governing bodies, then outside actors may withdraw funding for the government's parties and programs. In most cases, such sanctions are possible only when external actors who are not biased toward either side are available to enforce the settlement. I argue that these sufficiently impartial external actors must have credible leverage—that is, sanctions they can use against any party that fails to comply with the terms of the settlement. And to target sanctions appropriately, they must also have sufficient information about the parties' record of compliance. Finally, they must be able to enforce peace agreements at a sufficiently low cost to themselves.

The informational and incentive dimensions of the commitment problem require that external actors monitor compliance with the terms of the settlement, offer conditional incentives for compliance, and threaten sanctions for noncompliance. Because both sides have incentives to credibly commit to a mutually beneficial agreement, the incentives and sanctions need only be greater than what either side can seize through noncompliance.27 These incentives and sanctions, however, are effective only if they are applied over the entire implementation period to both sides. Asymmetries in monitoring or in the imposition of sanctions in the case of noncompliance will exacerbate the commitment problem and increase the probability of a return to conflict.

In addition, the enforcement mechanism is effective only if the external actor is credible, which may not be the case with armed intervention. Common wisdom has conceptualized enforcement as troops verifying compliance and threatening forceful punishment for violations.28 In theory, armed interveners could solve the commitment problem, but armed intervention is typically less credible than enforcement through elections because it can be costly for outsiders in terms of resources expended and lives lost over time. In practice, the evidence suggests that even armed interveners, often UN peacekeepers, employ conditional incentives rather than force to induce compliance—for example, shaming belligerents, withdrawing aid, or preventing party participation in political activities.29 Sending troops is still costly, though, and thus relatively rare, even if they are not threatening force.30

Combatants may also try to settle a civil conflict without the involvement of international actors. Settlements can establish territorial strongholds for the warring factions, allowing them to maintain control and keep their arms to defend themselves against violations by the other side. An example is the 1997 Chittagong Hill Tracts peace agreement in Bangladesh. The peace produced from such settlements, however, is often precarious.

### EXTERNAL ENGAGEMENT THEORY

According to external engagement theory, peace agreements with electoral participation provisions facilitate international enforcement of settlements. They do so by increasing coordination, lowering the cost for international actors to gather information, and applying leverage on combatants to incentivize compliance, which in many cases is augmented by cultural norms and institutions.31 (See table 1 for an overview of the theory.)

Table 1.

Essential Steps of the External Engagement Explanation

 CIVIL CONFLICTS ARE DIFFICULT TO END WITH SETTLEMENTSGovernment and rebel group(s) seek to settle by distributing power between themselves based on each side's strength during combat.Each side has incentives to grab more power when it is relatively strong and its opponent is complying with the terms of the settlement, producing credible commitment problems. OUTSIDERS CAN HELPOutsiders with sufficient information and credible leverage can enforce compliance, but how? ELECTORAL PARTICIPATION PROVISIONS ALTERNATIVE • Coordination around election benchmarks and milestones that culminate in the distribution of power increases the amount of information and leverage; outsiders can verify and incentivize compliance at these points. Armed actors monitor and threaten coercive force for noncompliance. • After the Cold War, interest in international involvement to help end civil conflict increased and democracy-promotion programs spread, facilitating external engagement through the electoral process. Armed intervention is typically less credible than external engagement because it can be costly for outsiders in terms of casualties and resource expenditures. Outside enforcement based on electoral participation provisions is low cost and therefore typically credible. Peace agreements with electoral participation provisions are thus more enduring, particularly when they have clear expectations of external engagement. Peace agreements that depend solely on armed intervention by outsiders are rare and may fail as the interveners' credibility falters.
 CIVIL CONFLICTS ARE DIFFICULT TO END WITH SETTLEMENTSGovernment and rebel group(s) seek to settle by distributing power between themselves based on each side's strength during combat.Each side has incentives to grab more power when it is relatively strong and its opponent is complying with the terms of the settlement, producing credible commitment problems. OUTSIDERS CAN HELPOutsiders with sufficient information and credible leverage can enforce compliance, but how? ELECTORAL PARTICIPATION PROVISIONS ALTERNATIVE • Coordination around election benchmarks and milestones that culminate in the distribution of power increases the amount of information and leverage; outsiders can verify and incentivize compliance at these points. Armed actors monitor and threaten coercive force for noncompliance. • After the Cold War, interest in international involvement to help end civil conflict increased and democracy-promotion programs spread, facilitating external engagement through the electoral process. Armed intervention is typically less credible than external engagement because it can be costly for outsiders in terms of casualties and resource expenditures. Outside enforcement based on electoral participation provisions is low cost and therefore typically credible. Peace agreements with electoral participation provisions are thus more enduring, particularly when they have clear expectations of external engagement. Peace agreements that depend solely on armed intervention by outsiders are rare and may fail as the interveners' credibility falters.

#### INCREASED COORDINATION

Electoral participation provisions increase coordination because electoral cycles provide multiple, easily observable cues as to whether participating combatant parties are complying with the terms of the settlement. As James Fearon writes, “[T]he institution of publicly understood rules for regular elections“ provides signals that actors can use to observe and coordinate their behavior.32 Following settlements, electoral participation provisions serve two important functions in coordinating enforcement.33 First, they enumerate regular public benchmarks agreed upon by the combatants and outside observers. They require publicly identifying who is eligible to run for office, the resources that will be made available to their campaigns, how the winner will be chosen, the resources that the winners will control and how they will exercise that control procedurally, and when each stage will occur. This stands in contrast to, for example, private promises of future government appointments and resource distribution. Second, electoral participation provisions constitute an ongoing set of regular milestones. The recurring electoral cycles consist of a timetable of observable events, including voter registration, the announcement of candidates, campaign periods, voting, and the installation of elected leaders, as well as non-electoral provisions pegged in the settlement to this timetable (such as demobilization, which may be established as a prerequisite for party registration).34 Thus, all of the actors know not only what is expected of the combatants in terms of compliance but also by when. These public benchmarks and regular milestones enable all of the actors to observe signals from various constituents about clearly defined standards of compliance.

The coordination built into electoral participation provisions can also increase leverage, which can be used to sanction noncompliance. Elections culminate in the redistribution of power, producing a risk for combatants of losing even their promised share of power within the state or within the party—a risk that other actors can influence. Politicians are especially vulnerable to the threat of withdrawal of international assistance; candidates and elected officials typically enjoy both domestic and international benefits provided to their parties in the form of salaries, stipends, access, and the ability to govern effectively in their communities or provide patronage to their supporters. Therefore, the threat of sanctions can be brought to bear on individual candidates in favor of other prospects or on parties in favor of other parties. A foreign state, for instance, might freeze the democracy and governance assistance funds that it provides to a government just before an election if the incumbent party attempts to change laws before its opponent is incorporated into the state's structures (which may make it harder for that party to campaign effectively to secure seats).

At the conclusion of a civil conflict, domestic actors—civil society, in particular—are often weak and typically polarized. They are thus unlikely to be able to monitor and incentivize compliance in a sufficiently impartial manner.35Although domestic actors can sometimes help provide information on combatant actions, which can be checked by outside actors, they have little credibility in detecting and sanctioning noncompliance.

International actors are therefore needed, at least initially, to “mediate and supervise joint disarmament and state-building.”36 Because enforcement works only if the parties believe that they will be sanctioned for noncompliance with the peace agreement but not for compliance, outside actors must be seen by combatants as both strong and sufficiently impartial. During an electoral cycle, opposing combatant parties, as well as the population and civil society, can provide their views on how well each side is complying with the electoral laws and other rules established by the settlement; but because signals may be biased, international actors must check them and respond appropriately.

#### INTERNATIONAL INVOLVEMENT

After the Cold War, international actors had both an interest in and the tools needed to foster peace. Major powers, no longer fighting proxy wars, were in a particularly good position to help end civil conflict.37 UN peacekeeping became more prevalent, but neither pervasive nor persistent, as mentioned earlier.38 Meanwhile, international interest in low-cost intervention, such as unarmed involvement, increased rapidly.39

Democracy-promotion programs spread gradually, providing new tools for external actors to use along with electoral participation provisions to overcome two sets of commitment problems. The first set of tools consists of international election observation processes. Before 1989, international election observation occurred in less than 10 percent of all elections worldwide; by the 2000s, more than 80 percent of all elections were being observed by international monitors.40 International election observation provides information about whether power is being distributed according to the agreed-upon rules throughout electoral cycles, extending the coordination capabilities of electoral participation provisions described above. As democracy-promotion programs spread around the world, starting in regions neighboring Western powers,41 they produced expectations of external engagement in states not yet affected by them.42 In states with civil conflicts, such expectations changed the form of peace settlements and combatants' behavior. When some states are subjected to rigorous international observation and receive incentives for playing by established rules, it is more likely that neighboring states' noncompliance will be detected and sanctioned.43 These sorts of mechanisms are found in many states, including many post-conflict states.44 Extant work suggests that international observers base their benchmarks on the terms of settlements, privileging them over other democratic norms, in post-conflict contexts.45 In Bosnia and Herzegovina, for example, the European Union (EU) measured the elections beginning in 1996 using the “yardstick” of the 1995 Dayton agreement,46 and showed that the combatant parties had complied with that set of standards, even though the agreement's design of elections around ethnic quotas conflicted with EU democracy standards.47 The EU's report on Burundi in 2010 noted similar findings. Focusing on whether the parties meet the terms of the settlements, established as electoral participation provisions, is useful for overcoming commitment problems and stabilizing peace.48

The second set of tools available to external actors consists of conditional economic and political incentives—that is, promises of rewards and threats of punishment. The former includes foreign aid, especially democracy and governance assistance, party aid, and election aid, as well as membership in popular intergovernmental organizations and treaties. The latter includes the threat or actual withdrawal of any of these benefits, as well as other sanctions and bans.49 After the Cold War, the availability of foreign aid and loans increased significantly and became, in many cases, conditional on compliance with set standards, as reported by international election observers.50 Democracy and governance aid, in particular, which could most easily be tied to compliance with electoral rules, increased over the 1990s and into the 2000s.51 Many post-1989 settlements also have included the establishment of trust funds or other assistance for former rebel groups' political parties, for whom even small reductions in resources can be costly.52 For instance, the rebel group RENAMO decided against returning to fighting during Mozambique's first post-conflict election when it was threatened with the loss of an $18 million UN-administered trust fund.53 International actors apply much of this conditionality, however, to the incumbent government, which is more able than the rebel group to renege on settlements from its position of strength in state institutions. In Guatemala, for example, the UN mission's greatest successes following the 1996 settlement came when the mission was able to convince international aid donors to impose “‘constructive conditionality’ on the government to follow through with the peace process.”54 The United Nations leads most observation missions during initial post-conflict elections and also coordinates foreign donors on conditional incentives. As stated earlier, for international observation and conditional incentives to be effective, international actors must be willing to criticize and sanction combatant parties' noncompliance with the terms of settlements. Existing evidence suggests that international actors do use these mechanisms against combatant parties, evaluating noncompliance rigorously and enacting sanctions when needed,55 including withdrawing aid and other benefits.56 This type of external engagement is also effective because it is low cost, partly because participatory electoral cycles increase coordination over time (as discussed earlier) and partly because cultural norms and institutions support this involvement in elections. Cultural norms and institutions have changed since the end of the Cold War. First, with the fall of the Soviet Union, Western power secured elections as the method to legitimately distribute power and produced conventions on when to assess compliance with state rules on power distribution.57 With these liberal norms in place, elections began to attract international media attention, monitoring, and related aid.58 Second, democracy-promotion programs became a widely acceptable method of international involvement among domestic actors,59 removing the need for interveners to overcome domestic actors' sovereignty concerns, and thus reducing resistance to this form of intervention.60 Third, the institutionalization of observation and assistance provisions that accompanies elections eases the logistical burden of enforcement. External engagement through electoral participation provisions, particularly with the spread of democracy-promotion programs, is therefore less costly, and thus more credible, than alternatives such as armed international intervention. ### IMPLICATIONS OF EXTERNAL ENGAGEMENT AND ALTERNATIVE THEORIES Two hypotheses follow from the above discussion of external engagement theory. First, if electoral participation provisions serve as low-cost, long-term mechanisms for credibly monitoring and incentivizing compliance, compared to alternatives that are likely more precarious, signatories to a settlement should expect marginally better payoffs for compliance, making them less likely to return to conflict. Hypothesis 1: Peace agreements with electoral participation provisions are associated with a lower likelihood of conflict recurrence than peace agreements without these provisions. Evidence of H1 would be consistent with external engagement theory and would challenge the existing literature's expectations that post-conflict elections have no effect, or perhaps even a negative effect, on peace. It would also be consistent, however, with three potential alternative explanations for this relationship. First, given that the parties to the conflict negotiate the provisions of the settlement, it is possible that a selection effect drives the relationship between provisions and peace. For example, the decision to include electoral participation provisions might reflect combatants' private (and potentially unobservable) information about the stability of a settlement.61 Thus, the presence of electoral provisions may simply be correlated with, rather than causally contribute to, peace.62 A second alternative explanation is that combatants view elections as the most legitimate mechanism for distributing power, and are therefore more likely to abide by their results. Third, elections might be associated with better governance, which could increase accountability, and thereby the propensity for peace.63 Second, if external engagement theory is correct, the relationship between electoral participation provisions and peace should hold when combatants expect international actors to monitor elections and provide incentives conditional on compliance. If a positive relationship between electoral participation provisions and peace is found (H1), it should hold in cases in which combatants expect enforcement of these provisions by external actors. In this sense, these external engagement expectations are moderating variables. Hypothesis 2: The pacifying effect of electoral participation provisions increases with the expectation of external engagement to enforce the peace agreement. ## Testing External Engagement Theory This section offers tests of my two hypotheses using cross-national data. It also provides summary statistics on whether external actors engage as expected when electoral participation provisions are implemented. ### DEPENDENT VARIABLE: CONFLICT RECURRENCE ACROSS SETTLEMENTS The universe of cases for this cross-national analysis comprises settlements from 1975 to 2005 signed to solve, regulate, or outline a process to resolve the incompatibilities between two or more sides fighting any type of internal conflict that resulted in twenty-five or more battle deaths per year.64 The period thus provides time after 2005 to assess the recurrence of conflict and other election-related outcomes; the period also coincides with the third wave of democratization, which began in 1974, limiting the comparison to cases that experienced normative pressure to hold elections.65 (To produce comparisons only after 1989 to further standardize the counterfactual, I also include an indicator of the Cold War.) I cluster continuous peace processes, treating sets of negotiations that end with a single solution as a settlement, because peace is not expected to follow each negotiation, but rather the settlement these negotiations produce.66 I also convert these data to dyadic agreements between each signing rebel group and the government because conflict recurs on a dyadic level.67 The variation I explore is conflict recurrence within each dyad that signs a peace agreement to terminate civil conflict. Using the standard Uppsala Conflict Data Program/Peace Research Institute Oslo (UCDP/PRIO) Armed Conflict Dataset,68 I identify whether the rebel group that signs the peace agreement reenters the dataset under the same name or that of a UCDP/PRIO–coded alliance or merger, indicating a return to conflict at the threshold of at least twenty-five battle deaths per year.69 I code whether the conflict ever recurred for the dyad through 2010; however, because most dyads that fail do so within five years, I use an indicator of whether the agreement fails within five years as the dependent variable in the main analyses.70 The resulting dataset contains 110 dyadic peace agreements (with 12 emerging from settlements with multiple signatories) between 81 different dyads in 49 civil conflicts in 43 states from 1975 to 2005.71 In 47 instances (43 percent of these agreements), the dyads returned to civil conflict within five years.72 ### INDEPENDENT VARIABLE: ELECTORAL PARTICIPATION PROVISIONS As noted earlier, electoral participation provisions are clauses concerned with holding elections and allowing rebel groups to participate as political parties. I analyzed the text of all of the accords that produced the 110 settlements to identify whether they include (1) clear expectations that an election would be held on a set date, which I also code as extant if elections have been held regularly for two cycles and the settlement does not interrupt them, and (2) clear expectations that both sides would field candidates. In practice, governments typically conducted these elections. To include rebel groups, therefore, the settlement must have (1) legalized rebel groups as political parties; (2) created a transitional government set to include rebel groups as parties; or occasionally (3) included rebel groups that already participated in elections during the conflict, either on their own or through alliances with political parties. In Djibouti, for example, elections were held regularly for at least two cycles before the signing of a settlement in 1994; the settlement stipulated that “once the present Peace agreement has been signed, FRUD [the rebel group Front for the Restoration of Unity and Democracy] will become a legal political party.”73 Of the 110 dyadic peace agreements, 42 (38 percent) include electoral participation provisions.74 Those without such provisions usually outline other institutional arrangements for distributing political power, such as territorial division to allow some control by each side. ### MODERATING VARIABLES: EXPECTATIONS OF EXTERNAL ENGAGEMENT External engagement theory implies not only that electoral participation provisions correlate with more enduring peace (H1), but also that this relationship should hold when combatants expect international actors to engage through elections to monitor and incentivize compliance (H2). Therefore, it is necessary to measure combatant expectations of external engagement in elections that theoretically supports peace and stability (rather than backing one side as happens in proxy wars, for instance). This variable is difficult to measure directly, but there are three variables that can serve as proxies. First, I use the end of the Cold War as a proxy variable because it coincided with increased international coordination on civil conflict termination and the development of tools that allowed electoral processes to be more easily observed and incentivized. The end of the Cold War should thus increase expectations of appropriate external involvement.75 However, because the end of the Cold War likely produced change through multiple mechanisms (e.g., lessening support to armed actors), I focus primarily on the two other proxy variables, both of which are closely tied to the spread of democracy-promoting programs by region,76 and each of which should therefore increase expectations of appropriate external engagement. The first is the percentage of legislative elections in the region being observed by international monitors in the year prior to the agreement, excluding the state being examined.77 The second is the percentage of all development assistance in the region devoted to democracy and governance promotion, averaged over the two prior years.78 Theories about democracy promotion argue, and show empirically, that external engagement spreads within regions, driven by systemic characteristics such as international interest in a region, making it more likely that other states within a particular region can expect the same type of external engagement in future elections. Regions neighboring the United States and Europe were the first subjects of external engagement. International election observation, crucial for detecting noncompliance in post-conflict situations, followed these regional trends,79 as did democracy and governance aid.80 These regional trends, and the diffusion theories associated with them, suggest that states in regions with higher rates of international election observation and associated aid should be more likely to expect external engagement to monitor and incentivize compliance with the terms established in peace agreements.81 Importantly, these lagged regional variables are also likely exogenous: inviting international election observers and receiving associated aid could be highly endogenous to expectations about peace and an aspect of implementation.82 In contrast, the regional variables, which are from past years and even exclude the state in question, are unlikely to be determined by the settlement's implementation or by how promising peace is. Interviews that I conducted with UN and Carter Center officials indicate that nongovernmental organizations tend to engage in unstable states and regions, places that are less promising for peace, further supporting the exogeneity of these variables. Also, existing work shows that regional trends of increasing external engagement do not reflect the spread of democracy or other aspects of governance, which could foster stability.83 Regional variables thus represent plausible proxies for combatants' expectations of international engagement and are less likely than provisions to face selection effects. To test H2, then, I interacted each measure with electoral participation provisions in the analyses.84 ### ELECTORAL PARTICIPATION PROVISIONS AND ENDURING PEACE Consistent with external engagement theory, the data show that settlements that contain electoral participation provisions are more enduring (see figure 1). Indeed, five years (sixty months) after settlements have been signed, about 75 percent of signatories to settlements with electoral participation provisions are still at peace, compared to about 50 percent of those without such provisions. The result is statistically significant.85 This simple comparison is consistent with H1. Figure 1. Peace Endurance across Peace Agreements (Kaplan-Meier Estimates) Figure 1. Peace Endurance across Peace Agreements (Kaplan-Meier Estimates) In addition, when peace agreements fail, they usually do so within the first five years (again, see figure 1). This result reflects the impact of commitment problems, given that at the start of the implementation process, combatants are most likely to return to fighting at any sign their opponent will defect. The variation of interest is therefore whether they fail more or less frequently with electoral participation provisions (and, as a moderating factor, expectations of external engagement). This article thus uses a logistic regression model to identify covariates of peace among combatants, or, more precisely, negative correlations with conflict recurrence between the signatories within five years of the settlement.86 I cluster the standard errors by state in all model specifications because observations may be related.87 The results show that electoral participation provisions are associated with more enduring peace, thus supporting H1. These results are stable across other methods of estimating these effects. Conflict recurs within five years in 21 percent of agreements with electoral participation provisions, in contrast to 56 percent of those without these provisions.88 The relationship is shown in table 2.89 Substantively, the effect translates into an 80 percent increase in the probability of enduring peace (44 percent compared to 79 percent). Table 2. Conflict Recurrence after Peace Agreements (Logistic Regression) Independent VariablesModel 1Model 2 Participation provisions −1.54*** (0.56) −1.45** (0.56) Cold War 0.44 (0.74) Participation provisions *Cold War Provisions no cases Number of observations 110 110 Pseudo R-squared 0.09 0.09 Log pseudo likelihood −68.48*** −68.23 Independent VariablesModel 1Model 2 Participation provisions −1.54*** (0.56) −1.45** (0.56) Cold War 0.44 (0.74) Participation provisions *Cold War Provisions no cases Number of observations 110 110 Pseudo R-squared 0.09 0.09 Log pseudo likelihood −68.48*** −68.23 note: Method is logistic regression analysis. Dependent variable is conflict recurrence by government-rebel group dyad within five years (binary). Numbers in parentheses are robust standard errors, clustered by state (maximum = 43). ***p < 0.01 **p < 0.05, *p < 0.1 ### PERSISTENCE OF PEACE WHEN PARTIES EXPECT EXTERNAL ENGAGEMENT To further test external engagement theory, I examine H2 using the same models, but including the proxy variables interacted with electoral participation provisions to assess the conditional effects.90 The results are also consistent with external engagement theory: the association between electoral participation provisions and peace holds when combatants expect external engagement. First, during the Cold War, combatants should not have expected external actors to detect and sanction compliance through elections, and, indeed, no peace agreements with electoral participation provisions were signed prior to 1989.91 This means that the coefficient on electoral participation provisions can be read as the effect of electoral participation provisions on peace in the post-1989 period (model 2 in table 2), and it shows a large and statistically significant reduction in the risk of conflict recurrence. Second, the interactions of the regional democracy-promotion variables and electoral participation provisions are negative and statistically significant; to interpret these relationships, I calculate the adjusted predicted probabilities of each type of peace agreement at each value of the two regional proxies.92 The predicted probability of conflict recurrence with electoral participation provisions compared to those without these provisions indicates that when at least 40 percent of elections held in the region during the previous year were internationally observed, electoral participation provisions are associated with a decreased risk of conflict recurrence (and the effect is statistically significant when at least 60 percent of elections were observed) (figure 2). The same is true when democracy and governance assistance constitute at least 3 percent of aid distributed to the region over the previous two years (and at 7 percent, the effect is statistically significant) (figure 3). The proxy variables for combatant expectations of external engagement do not reduce the risk of conflict recurrence on their own in these models, or later as controls; rather, they moderate the relationship between electoral participation provisions and peace, as external engagement theory implies. Figure 2. Predicted Probabilities of Conflict Recurrence (Monitoring) Figure 2. Predicted Probabilities of Conflict Recurrence (Monitoring) Figure 3. Predicted Probabilities of Conflict Recurrence (Conditional Incentives) Figure 3. Predicted Probabilities of Conflict Recurrence (Conditional Incentives) ### ASSESSING ALTERNATIVE EXPLANATIONS AND ROBUSTNESS CHECKS The above results stand in contrast to studies that expect a negative association between electoral participation provisions and enduring peace. These results are consistent with H1 and H2 implied by external engagement theory, but there are several alternative explanations for some of these results. A selection effect could drive the relationship between electoral participation provisions and the endurance of peace: the strategic decision made by combatants to include electoral participation provisions might reflect their private (and potentially unobservable) information about the stability of the conflict's end. For example, the combatants may know that contraband funding that has financed weapons purchases is no longer available, making peace a more attractive option. If combatants then include electoral participation provisions in those cases, the provisions should correlate with, but not cause, more enduring peace.93 Other alternatives posit other causal logics or omitted variables that could produce a relationship between electoral participation provisions and enduring peace. In this section, I describe some testing of the most obvious selection effects, which shows that these effects do not seem to produce the correlation between electoral participation provisions and more enduring peace. I then I discuss other empirical strategies to increase confidence in a causal relationship between electoral participation provisions and enduring peace. Next, I describe the controls that I use to test some alternative explanations as part of this approach. Finally, I discuss the results of these models and additional robustness checks, highlighting evidence that is most consistent with external engagement theory. #### PROVISIONS NOT LIMITED TO EASIER CASES FOR PEACE Existing work has found that conflicts with more deaths and displacement, those that feature contraband funding, and identity conflicts are more likely to recur and thus are harder to settle than other types of conflicts; in contrast, secessionist conflicts are less likely to recur and therefore are easier to settle.94 Moreover, weak states, measured by variables such as low gross domestic product per capita and more mountainous terrain, also predict conflict occurrence, likely making them harder cases for enduing peace.95 To test the selection effect, I assess whether electoral participation provisions are included in easier or harder settlements. If anything, the results demonstrate that electoral participation provisions are included in peace agreements in conflicts that are harder to settle,96 including those involving more conflict deaths, lower state capacity, more mountainous terrain, and more contraband financing.97 In addition, as external engagement theory would imply, electoral participation provisions are more likely to be included when domestic actors expect external engagement through democracy-promotion programs.98 #### STRATEGIES TO FURTHER ASSESS ALTERNATIVE EXPLANATIONS Initial evidence, then, runs contrary to an explanation based on selection effects, implying that the correlation between electoral participation provisions and enduring peace may be causal. Still, there could be unobservable selection that I cannot capture, or another explanation could be at work. To address this issue, I use multiple tests and sources of data to increase confidence in a causal relationship between electoral participation provisions and enduring peace. First, H2, suggesting that the relationship should hold only when combatants expect external actors to engage, is a test of external engagement theory. However, it is also useful that the proxy variables included in the interactions—the end of the Cold War and the regional trends in democracy-promotion programs—are exogenous. In other words, these variables are unlikely to be affected by whether a particular conflict is easy to settle. The evidence in the prior subsection strongly supports H2, and thus provides evidence for external engagement theory, but it counters alternative theories, especially a selection explanation for the correlation based on settlement ease. Second, to isolate the relationship between electoral participation provisions and enduring peace, I control for possible confounding variables, including other predictors of conflict recurrence, settlement design, and post-conflict elections.99 I include controls for conflict severity (including conflict size and duration; state strength, measured using gross domestic product per capita; and the balance between the rebel group and the government),100 settlement difficulty (including number of past failed agreements; number of factions signing and not signing; and whether further negotiations were stipulated in the agreement), group goals (including territorial aims; ethnic and religious agendas; Marxist platforms; and attempts to totally overthrow the state), international relationships (including peacekeeping missions and indicators of settlements signed during the Cold War or after the terrorist attacks of September 11, 2001),101 and measures of democracy and development (including democracy levels at the state and regional levels, as well as population size).102 These are measures used in other studies, and they may affect settlement design and conflict recurrence.103 In addition, I control for region and time-period indicators, as well as the interaction of these variables, which should soak up some of the variation specific to particular geographic areas or temporal periods.104 Third, I assess whether electoral participation provisions are selected in conflicts that are easier to settle by examining whether they are more likely to be included with other power-sharing provisions. The explanation for the correlation could be that settlements include many provisions when combatants expect post-conflict stability. Again, their expectation of stability, rather than the provisions, could then be driving the correlation with more enduring peace. I therefore control for power-sharing provisions besides electoral participation provisions, including those to reform the security sector; to disarm, demobilize, and reintegrate combatants; and to integrate government or civil service105—each of which may reduce conflict recurrence. If provisions are simply included as a result of such selection effects, rather than driving a causal effect, they should be negatively correlated with conflict recurrence and positively correlated with electoral participation provisions (results showing the opposite are in the section below).106 Finally, the case study also helps assess the mechanism, further testing the theories. #### CONTROLS AND ADDITIONAL ROBUSTNESS CHECKS With controls included in the baseline models with electoral participation provisions (table 2), the correlation between electoral participation provisions and enduring peace remains strong across specifications, which is consistent with external engagement theory but not with alternative explanations (table 3).107 These specifications include the controls described above for other power-sharing provisions, conflict severity, settlement difficulty, group goals, international relationships (including peacekeeping missions), and measures of democracy and development, as well as a final set with all variables that are statistically significant in any specification. Across all of these specifications, the coefficient on electoral participation provisions remains negative and statistically significant.108 Table 3. Conflict Recurrence across Peace Agreements (Sets of Controls) Independent VariablesDirectionStatistical Significance Electoral participation provisions − **(* OTHER PROVISIONS Disarmament, demobilization, and reintegration provisions − **(* Security-sector reform provisions − Government power-sharing provisions Civil service power-sharing provisions CONFLICT SEVERITY Major war (* Duration of the dyad's conflict Real GDP per capita (1,000s, lagged) (*** Balance between group and government − OTHER OBSTACLES TO PEACE Past agreement(s) Number of fighting factions not signing Number of factions signing (** More negotiations stipulated in the settlement GROUP GOALS Territorial conflict (+/−) Identity conflict − Marxist conflict − Rebel croups with total goals (* INTERNATIONAL DEMOCRACY AND DEVELOPMENT Cold War − Post–September 11, 2001, terrorist attacks UN peacekeeping mission (present) Regional election observation (percent, lagged) − Regional democracy level (lagged) − (* Level of democracy (lagged) − Population (1,000s, lagged) − Independent VariablesDirectionStatistical Significance Electoral participation provisions − **(* OTHER PROVISIONS Disarmament, demobilization, and reintegration provisions − **(* Security-sector reform provisions − Government power-sharing provisions Civil service power-sharing provisions CONFLICT SEVERITY Major war (* Duration of the dyad's conflict Real GDP per capita (1,000s, lagged) (*** Balance between group and government − OTHER OBSTACLES TO PEACE Past agreement(s) Number of fighting factions not signing Number of factions signing (** More negotiations stipulated in the settlement GROUP GOALS Territorial conflict (+/−) Identity conflict − Marxist conflict − Rebel croups with total goals (* INTERNATIONAL DEMOCRACY AND DEVELOPMENT Cold War − Post–September 11, 2001, terrorist attacks UN peacekeeping mission (present) Regional election observation (percent, lagged) − Regional democracy level (lagged) − (* Level of democracy (lagged) − Population (1,000s, lagged) − note: Method is logistic regression analysis. Dependent variable is conflict recurrence by government-rebel group dyad within five years (binary). Full models, as well as additional controls, are shown in the supporting information at https://dataverse.harvard.edu/dataverse/matanock (table A7.0–2). Direction and statistical significance indicators in parentheses indicate that they only occur in some specifications. ***p < 0.01 **p < 0.05 *p < 0.1 Meanwhile, few controls are statistically significant, and the evidence is not consistent with alternative theories. The results suggest that larger and longer conflicts, for instance, are more likely to recur following a negotiated settlement (statistically significant effects across specification); rebel groups that do not aim for total control of the state or that seek control in only part of the state's territory are less likely to return to conflict after signing a peace agreement; and conflicts involving more factions are more likely to dissolve back into fighting after a settlement (statistically significant effects in some model specifications but not across all specifications). UN peacekeeping does not have a statistically significant correlation with peace in these models,109 which suggests that the main effect is not driven by peacekeepers who push elections, for example, but then increase peace through other actions alone. The results also do not support a good-governance argument: governance controls are not significantly associated with peace, and they do not change the association between peace and electoral participation provisions, suggesting that they are not driving the relationship.110 Finally, including region and time-period controls, and their interaction, and still seeing the same correlation between electoral participation provisions and more enduring peace, suggests that these relationships are not based on simple temporal or geographic trends. Moreover, not all provisions reduce the likelihood of conflict recurrence. Provisions to integrate the civil service and government are associated with a greater likelihood of conflict recurrence (table 3 shows the relationships are positive, and the former sometimes statistically significant). The coefficient on disarmament, demobilization, and reintegration is negative and statistically significant. Taken together, these findings suggest that some other aspects of peace agreements are also important, but that not all of them are associated with peace. Electoral participation provisions still have a statistically significant effect when these other provisions are included, and the provisions are not highly correlated with each other,111 suggesting that an omitted propensity for peace, perhaps private information held by combatants that could lead them to include many different provisions, is not driving the results. In sum, when all of these other provisions and possible confounds are included, electoral participation provisions are still positively correlated with more enduring peace, which is consistent with external engagement theory. ### CORRELATES OF PEACE How does the relationship between electoral participation provisions and enduring peace explored in this article compare to existing findings on conflict termination? First, previous studies have found that any negotiated settlement, as compared to a military victory by one side, is associated with less enduring peace.112 To assess how well settlements with electoral participation provisions perform relative to these other options, I modeled all peace periods of at least one year following a civil conflict.113 Compared to conflicts that drop below the twenty-five-battle-death threshold—when international actors may be especially able to push one side to victory or to help negotiate a settlement—a victory reduces the risk of renewed conflict by 90 percent, while an agreement with electoral participation provisions reduces this risk by 74 percent. The former is therefore still the best option, but the latter is better than any other type of peace agreement or cease-fire. Second, as discussed earlier, existing studies are pessimistic about the ability of post-conflict elections to produce enduring peace, but they do not pay considerable attention to the type of election that follows a conflict. I therefore examine whether any electoral provisions, including those that do not require all sides to compete as parties,114 have the same correlation with more enduring peace as do those calling for participation by rebel groups.115 The correlation of any election and conflict recurrence is small and not statistically significant in these data, which is consistent with some previous studies and challenges the current practice that does not distinguish between these categories.116 Only by narrowing the focus to electoral participation provisions is the negative correlation statistically significant. This evidence, too, supports the mechanism and is consistent with external engagement theory. ## External Engagement in the Case of El Salvador The mechanism of external engagement suggests that outside actors should be likely to engage in electoral processes once peace agreements are signed. Comparing levels of engagement is difficult both because the counterfactual is not entirely clear and because such comparisons require selecting successful implementation of post-conflict elections, which potentially introduces bias. Nevertheless, I can examine rates of international observation, democracy and governance aid, and conditionality around elections, comparing those elections produced by participation provisions as opposed to any other post-conflict elections, to evaluate if the results are plausibly consistent with external engagement theory's expectation that post-conflict elections that follow electoral participation provisions attract more external engagement than those that do not. The results show that this is indeed the case.117 In addition, conditionality—punishments or rewards conditioned on abiding by the rules of the game118—is also statistically significantly higher prior to these same elections (present in 77 percent of these cases, compared to 44 percent otherwise). These results suggest that participation provisions are associated with external engagement through the electoral process as the theory implies. To further test the mechanism, I analyzed El Salvador's 1992 settlement using process tracing to assess whether the mechanism of external engagement plausibly explains the correlation between electoral participation provisions and peace.119 In choosing this case, I sought a clear change in expectations of external enforcement around which to examine the causal chain. A moment in which these expectations should increase, at least in the regions closest to the West, is the end of the Cold War.120 As discussed earlier, its end decreased coordination problems around terminating civil conflict and coincided with the spread of democracy-promotion programs. As the Cold War wound down, then, expectations about external enforcement increased, driving the causal chain implied by external engagement theory.121 Once a peace deal is signed, the theory implies, former government and rebel forces would try to take advantage of moments in which they were relatively stronger. The theory also implies, however, that the electoral process established by participation provisions would reverse these incentives and foster peace. Specifically, the theory implies that international actors would monitor and supply aid conditioned on compliance, and that this mechanism, rather than troops threatening coercive punishment, would ensure a sufficient level of compliance by each side to maintain peace. I therefore examined all civil conflicts that began before the collapse of the Soviet Union, featured continuous fighting between rebel groups and the government, and then terminated through settlements after the end of the Cold War. Ten such cases included electoral participation provisions, and each shows extensive external engagement around the participatory electoral processes, including international election observation and related aid available to be conditioned on compliance.122 In selecting one case to process trace in this piece, I sought a hard case from among these: it should be especially difficult to persuade rebel groups that fought against U.S.-backed governments during the Cold War that international actors, particularly the United States and intergovernmental organizations to which it belongs, will enforce compliance on both sides. Leftist guerrilla groups from Central America fit these criteria. If any rebel group would not need international support, it would be a stronger group. For this reason, I selected El Salvador over Guatemala, which had a weaker rebel group, making this the hardest possible case among the set in which to expect this shift. The robust insurgency in El Salvador, which began in 1980, had its roots in land rights and labor policies.123 The right-wing government of El Salvador, led by the Nationalist Republican Alliance, which had U.S. backing during the Cold War, fought the leftist FMLN guerillas, which had foreign support from the Soviet Union. At times, the FMLN was a robust rebel group that looked as if it could defeat the government. However, the government and the FMLN reached a stalemate sometime in 1987 at the latest, following resource losses and war weariness among supporters on both sides.124 No settlement had been possible during the Cold War because the FMLN, in particular, was concerned about commitment problems vis-à-vis the government, and it did not have a credible foreign guarantor of a peace agreement because the only international actors involved (the United States and the Soviet Union) had instead taken sides in the conflict. Leaders on both sides were well aware of this bias.125 U.S. and Soviet sponsorship also kept the cost of conflict low for combatants through the supply of weapons and other resources. By 1989, both the FMLN and the El Salvadoran government had lost their sponsors, and international actors had begun to engage in a more impartial push for peace.126 In January of that year, the FMLN took the “unprecedented” step of signaling its willingness to commit to democratic procedures by asking to participate in the March elections, if the rules were changed (which the government did not yet do).127 The rebels had become more confident as guarantees were “awarded to the electoral processes” to incentivize all sides to comply with the rules; these guarantees came in part from “international pressure” wherein actors no longer backed a particular side but rather sought to stabilize peace through their involvement.128 For example, after the election of George H.W. Bush, the FMLN approached the administration about taking steps toward ending El Salvador's civil conflict, emboldened by the administration's less partisan stance in the region.129 Peace negotiations followed, featuring the FMLN's proposal to participate.130 U.S. officials met with guerrilla commanders in 1991 to assure them that the United States would “encourage lasting stability in the region” with elections at the center of those efforts, and it seemed to work: Ana Guadalupe Martínez, an FMLN leader, remarked, “Our attitude has changed [and now] we think the U.S. military group [U.S. military advisers and the U.S. embassy, according to the report] can help in the transition to peace.”131 In 1992, the FMLN and the El Salvadoran government signed a peace agreement, after both sides had come to believe that sufficiently impartial international actors would be available to monitor and incentivize compliance with the peace agreement through a participatory electoral process. Looking to the elections resulting from the peace process, the FMLN would depend “more on international pressures than on legislative clout or popular mobilization to achieve its goals.”132 The deal, even with complete compliance, left in place a relatively closed electoral system as part of an elite bargain between the sides.133 At the same time, however, the electoral process provided extensive external opportunities to both monitor and incentivize compliance. The combatants thus were able to overcome their commitment problems, particularly the FMLN's concern that the government would fail to comply, by “accept[ing] COPAZ [the National Commission for the Consolidation of Peace] and the United Nations as guarantors of [the FMLN's] security.”134 The United Nations established a peacekeeping mission during the negotiations to enhance the credibility of its guarantee, but the mission's size was small and its scope limited; it had no punishment mechanism that entailed force.135 Much of the international involvement in post-conflict El Salvador surrounded the elections. The 1994 elections attracted an influx of international attention, including about 4,000 international election observers. Crucially, these observers, many supervised by the United Nations, were key in unlocking aid needed by both sides.136 Official assistance, including national reconstruction funds that assured each side of “substantial funding,” was conditioned on compliance with the peace agreement and thus provided incentives to comply.137 All sides “needed financial resources to implement institutional reforms, land transfers, and reconstruction projects”; the United Nations judged whether both sides demonstrated “good faith and democratic vocation,” and, in doing so, it “[influenced] the prospects for major international funding.”138 Importantly, these international actors were no longer seen as backing a particular side, as they had during the Cold War. As such, they were able to provide international assurance that the rebels would be protected and that each side would receive aid if it complied with the terms of the peace agreement.139 Mechanisms for enforcement were brought to bear by international actors against each side when implementation of the settlement's terms stalled (suggesting that compliance would not otherwise have been each side's consistently preferred strategy). Entering into the implementation process, both sides initially failed to reach the benchmarks set in the settlement, either deliberately or because of logistical challenges. In response, international actors threatened to block participation and withhold aid that could have lost the noncomplying party essential support among voters. Ultimately, this involvement, particularly by the United Nations and the United States, created incentives to obtain sufficient compliance and continued peace.140 The government, for example, was slow to demobilize its forces and establish less partisan policing institutions. As the elections approached, and the incumbent party's presidential candidate began to face stiff competition, the United Nations issued public statements about the government's slow progress to pressure the latter into complying. These efforts were accompanied by specific threats from the United States and other international actors to withdraw aid.141 Analysts emphasized that the timing of these threats—before the vote—was crucial, because costing the state these critical funds would have been especially unpopular in the elections.142 The FMLN also initially failed to fully implement the settlement. An arms cache linked to the group exploded in 1993, causing the United Nations to forcefully criticize the FMLN's failure to disarm, and the government to suggest banning the rebel group from elections. The threat proved effective as the FMLN revealed and destroyed many other weapons caches.143 Thus, as both sides increased their accountability prior to the elections, the process also fixed firm deadlines for the most crucial components of the peace agreement.144 Noncompliance with the accords—usually impeding efforts to redistribute power—persisted beyond the disarmament phase, and external engagement in the electoral process continued to enforce the agreement. For example, when the government failed to issue voter registration cards in a timely manner—many to supporters of the Left who had not previously voted—the United States froze$70 million in U.S. Economic Support Funds; the freeze produced an increase in the pace of voter registration, leading the United Nations to declare that 90 percent of potential voters would likely be registered by the November 1994 deadline.145 Similarly, when the government attempted to move polling stations from locations where the FMLN had substantial support, the UN electoral observation mission disputed the government's claim that the move was motivated by security concerns and threatened further fund freezes, forcing the government to capitulate.146 These accountability mechanisms continued to be employed by external actors through subsequent elections, including in 1997; they also received substantial international observation,147 and donors continued to provide substantial funding that could be conditioned on compliance.148

Peace persisted in El Salvador following the 1992 peace agreement. Despite some setbacks and slow democratization, international actors kept the deal on track during its implementation, monitoring and incentivizing former combatant party compliance. This evidence of the mechanism is consistent with external engagement theory.

## Conclusion

Some settlements that end civil conflicts endure, but others do not. This article demonstrates that electoral participation provisions, clauses included in some peace agreements to transition combatants into political parties, help stabilize peace. In particular, they can facilitate the engagement of external actors, intergovernmental organizations, and foreign donors, who can detect and sanction former combatant parties' noncompliance with such settlements. Electoral participation provisions, which facilitate a particularly low-cost and long-term mechanism for enforcement, became especially effective as democracy-promotion programs spread following the end of the Cold War. By making noncompliance less beneficial than compliance throughout the implementation of a mutually beneficial deal, this enforcement mechanism can help overcome commitment problems and contribute to enduring peace.

Cross-national evidence indicates that agreements with electoral participation provisions are an excellent option for stabilizing peace, second only to military victory in their effectiveness. The results are robust, even when accounting for potential selection effects such as whether these provisions are simply included in easier cases for peace (they are not) and other alternative explanations. Case-study evidence also supports the argument. If an international actor wants to help foster peace, but will not support the stronger side in winning a conflict, supporting a settlement with these provisions is its best option.

Counteracting growing concern that post-conflict elections are ineffective or contribute to renewed conflict, these findings also demonstrate which kinds of elections can help peace endure. Post-conflict elections that are participatory and internationalized, both of which are demanding conditions, have a discernibly positive impact on peace. The results also offer an important implication about international intervention. The use of peacekeeping troops is costly and therefore relatively rare; there are, however, complements and substitutes to threatening punishment by force. Provisions for electoral participation by former combatant political parties, in combination with international involvement in elections, can detect and increase the cost of noncompliance.

Beyond the implications for scholars' understanding of stable civil conflict settlements, post-conflict elections, and international intervention, this article suggests two major extensions to current research. First, in addition to disaggregating post-conflict elections to examine their relationship with peace, future research should consider whether different post-conflict elections affect other outcomes, particularly democracy. External engagement theory suggests that electoral participation provisions work because they are elite arrangements likely accompanied by low levels of liberalism.149 A stability-democracy trade-off that may occur through these elections, which are often highly engineered and evaluated for their ability to resolve conflict, merits further investigation. Not many elections carry great uncertainty, however,150 and as a state develops a post-conflict electoral history, it may move away from some of the rigid guarantees of representation that secure the peace but potentially impede democracy.

Second, future work should consider the external engagement mechanism in light of larger issues of compliance and enforcement in post-conflict settlements. If my theory is correct, aid and other nonmilitary mechanisms can be used to help address domestic political issues beyond the immediate commitment problems associated with settling a civil conflict. These issues include many of interest to the global community, such as improving domestic human rights compliance, reducing fraud during leader transitions, and reducing corruption by incumbent governments. In practice, external engagement will not always be an option; for example, absent electoral cycles with the relevant actors participating as parties, the costs may be too high. Nevertheless, external engagement theory helps identify a much-expanded role for international actors in helping to solve domestic political problems.151

Two policy implications follow from this study. First, the United Nations and other international peacemakers should propose the inclusion of electoral participation provisions in every future peace agreement. Once an agreement is signed, international actors can make themselves available to identify and sanction instances of noncompliance. Their credibility can be enhanced by being explicit about the goals of post-conflict elections. If the explicit goal is external enforcement of compliance with a settlement, policymakers may be more able to effectively monitor such compliance. For instance, they might send missions out to former combatant strongholds in the lead-up to elections, rather than only to competitive regions. They might also provide more trust funds for combatant parties that have compliance as an explicit condition. A recent example of a trust-fund recipient is Nepal. Some policymakers already implement aspects of this external engagement effectively, but more customization to encourage compliance could be done across cases.

Second, there are still cases in which external engagement through elections is likely to be impractical, perhaps because combatants perceive international organizations as strongly committed to one side or because combatants perceive them to be apathetic. International actors must be interested but not excessively invested in either combatant side if they are to be trusted to detect and sanction noncompliance. In such scenarios, international peacekeepers should identify a bipartisan foreign state, or a collective of impartial states, and seek to signal their commitment to peace. A potentially useful option would be to strengthen intergovernmental organizations so that no single state is entrusted with determining noncompliance or imposing sanctions unilaterally. Intergovernmental bodies that design policy through coordination among many of their members—particularly those that may be on different sides of geostrategic debates—could work around concerns that one side will receive special treatment. The involvement of strong international organizations would take careful crafting, but the possibility that they could decrease the likelihood of renewed fighting in post-conflict states may make this approach worth serious consideration.

Much work still needs to be done on ways to ensure lasting peace in states that experience civil conflict. This research represents a first step in identifying electoral participation provisions in post-conflict settlements as a means toward achieving that objective.

## Acknowledgments

The author is grateful to Stanford University, the National Science Foundation, and the Scowcroft Institute of International Affairs, which funded research for this article. Eli Berman and the Institute on Global Conflict and Cooperation at the University of California, San Diego, provided a postdoctoral fellowship with funding from the Department of Defense's Minerva Research Initiative, which allowed time to write this article. Comments from the anonymous reviewers have considerably strengthened the article. For their feedback, the author also thanks Leo Arriola, Dawn Brancati, Sarah Bush, Dara Cohen, Luke Condra, Gary Cox, Martha Crenshaw, Christian Davenport, Daniela Donno, Thad Dunning, Lynn Eden, Tanisha Fazal, James Fearon, Tom Flores, Page Fortna, Scott Gates, Desha Girod, Caroline Hartzell, Ron Hassner, Reyko Huang, Susan Hyde, Bethany Lacina, David Lake, Katerina Linos, Michaela Mattes, Irfan Nooruddin, Maggie Peters, Jon Pevehouse, Alison Post, Robert Powell, Lauren Prather, Bryan Price, Amanda Robinson, Kenneth Schultz, Beth Simmons, Jack Snyder, Rachel Stein, Barbara Walter, Chris Warshaw, Barry Weingast, and Jeremy Weinstein. Ben Allen, Katie Beall, Natalia Garbiras-Díaz, Chelsea Johnson, Adam Lichtenheld, and Evan Ramzipoor provided outstanding assistance. The author received further useful comments from participants in panels at national conferences and in seminars at the Conflict Consortium Virtual Workshop, Columbia University, Duke University, George Washington University, Emory University, the University of Michigan, the Naval Postgraduate School, the Peace Research Institute Oslo, Stanford University, Universidad de los Andes, the University of California, Berkeley, the University of California, Los Angeles, the University of California, Merced, the University of California, San Diego, the University of Chicago, the University of Texas at Austin, Uppsala University, and Yale University.

## Notes

1.

Approximately 42 percent of peace agreements from 1975 to 2005 featured provisions for rebels to compete in post-conflict elections. See Aila M. Matanock, “International Insurance: Why Militant Groups and Governments Compete with Ballots instead of Bullets,” Ph.D. dissertation, Stanford University, 2012, chap. 2; Aila M. Matanock, “External Engagement: Explaining the Spread of Electoral Participation Provisions in Civil Conflict Settlements,” University of California, Berkeley, 2016; and Aila M. Matanock, Electing Peace: From Civil Conflict to Political Participation (Cambridge: Cambridge University Press, forthcoming), chap. 4.

2.

Paul Collier et al., Breaking the Conflict Trap: Civil War and Development Policy (Washington, D.C.: World Bank and Oxford University Press, 2013), p. 3.

3.

James D. Fearon and David D. Laitin, “Ethnicity, Insurgency, and Civil War,” American Political Science Review, Vol. 97, No. 1 (February 2003), p. 75.

4.

Peacekeeping missions coincide with post-conflict elections so often that some scholars suggest that elections are part of the United Nations' standard operating procedure. See, for example, Paul Collier, Wars, Guns, and Votes: Democracy in Dangerous Places (New York: Harper Perennial, 2009), p. 80. Others, however, have noted that international actors' procedures evolve across cases. See Marina Ottaway, “Promoting Democracy after Conflict: The Difficult Choices,” International Studies Perspectives, Vol. 4, No. 3 (August 2003), pp. 314–322. On other international actors, see Matanock, “International Insurance”; and Matanock, Electing Peace.

5.

I. William Zartman, “Negotiating the South African Conflict,” in I. William Zartman, ed., Elusive Peace: Negotiating an End to Civil Wars (Washington D.C.: Brookings Institution Press, 1995), p. 152.

6.

Lise Morjé Howard, UN Peacekeeping in Civil Wars (Cambridge: Cambridge University Press, 2008), pp. 52–87.

7.

Sheri Berman, “The Vain Hope for ‘Correct’ Timing,” Journal of Democracy, Vol. 18, No. 3 (2007), pp. 14–17; and Thomas Carothers, “The ‘Sequencing’ Fallacy,” Journal of Democracy, Vol. 18, No. 1 (2007), pp. 12–27.

8.

Jack Snyder, From Voting to Violence: Democratization and Nationalist Violence (New York: W.W. Norton, 2000); Roland Paris, At War's End: Building Peace after Civil Conflict (Cambridge: Cambridge University Press, 2004); Lars-Erik Cederman, Simon Hug, and Lutz F. Krebs, “Democratization and Civil War: Empirical Evidence,” Journal of Peace Research, Vol. 47, No. 4 (July 2010), pp. 377–394. Some empirical evidence suggests, however, that elections do not lead to conflict recurrence. See Roy Licklider, “Democracy and the Renewal of Civil Wars,” in Harvey Starr, ed., Approaches, Levels, and Methods of Analysis in International Politics: Crossing Boundaries (Hampshire, U.K.: Palgrave Macmillan, 2006), pp. 95–116. Prominent empirical studies on post-conflict elections and peace, specifically, show across cases that these contests fail to improve prospects for peace and can be detrimental to peace if held too soon after termination. See Paul Collier, Anke Hoeffler, and Mans Söderbom, “Post-Conflict Risks,” Journal of Peace Research, Vol. 45, No. 4 (2008), pp. 461–478; Dawn Brancati and Jack Snyder, “Time to Kill: The Impact of Election Timing and Sequencing on Post-Conflict Stability,” Journal of Conflict Resolution, Vol. 57, No. 5 (2012), pp. 822–850; and Thomas E. Flores and Irfan Nooruddin, “The Effect of Elections on Post-Conflict Peace and Reconstruction,” Journal of Politics, Vol. 74, No. 2 (April 2012), pp. 558–570.

9.

See, for example, Barbara F. Walter, “Designing Transitions from Civil War: Demobilization, Democratization, and Commitments to Peace,” International Security, Vol. 24, No. 1 (Summer 1999), pp. 127–155.

10.

Some post-conflict elections, such as those in Afghanistan in 2009, have been held by governments without the participation of primary rebel groups (or any signed settlement). Such elections may encourage conflict.

11.

The few studies that examine participatory post-conflict elections do not discuss at length their effect on conflict recurrence. See Jeroen de Zeeuw, ed., From Soldiers to Politicians: Transforming Rebel Movements after Civil War (Boulder, Colo.: Lynne Rienner, 2007); Mimmi Söderberg Kovacs, “From Rebellion to Politics: The Transformation of Rebel Groups to Political Parties in Civil War Peace Processes,” Ph.D. dissertation, Uppsala University, 2007, p. 216; and Carrie Manning, The Making of Democrats: Elections and Party Development in Postwar Bosnia, El Salvador, and Mozambique (New York: Palgrave Macmillan, 2008).

12.

See especially Barbara F. Walter, Committing to Peace: The Successful Settlement of Civil Wars (Princeton, N.J.: Princeton University Press, 2002).

13.

See data in Susan D. Hyde, The Pseudo-Democrat's Dilemma: Why Election Observation Became an International Norm (Ithaca, N.Y.: Cornell University Press, 2011).

14.

An alternative explanation is that enduring peace is simply the result of easier settlements. This explanation is refuted in, for example, Virginia Page Fortna, Does Peacekeeping Work? Shaping Belligerents' Choices after Civil War (Princeton, N.J.: Princeton University Press, 2008). Similarly, such selection does not seem to explain the results on enduring peace shown in this article, as the section on data analysis discusses.

15.

James D. Fearon, “Rationalist Explanations for War,” International Organization, Vol. 49, No. 3 (Summer 1995), pp. 379–414; Robert Powell, “Bargaining Theory and International Conflict,” Annual Review of Political Science, Vol. 5, No. 1 (2002), pp. 1–30; and Dan Reiter, “Exploring the Bargaining Model of War,” Perspectives on Politics, Vol. 1, No. 1 (March 2003), pp. 27–43. I use the terms “combatants,” “warring factions,” and “sides” to refer collectively to the rebel group(s) and the government in conflict with each other, including when discussing the period after the settlement. Fighting has to take place between a government and at least one rebel group to qualify as a “civil conflict,” and negotiations for settlements take place between these actors, too. Some cases, however, have more than one rebel group, or even involve multiple governments. Additional studies could further explore dynamics of more combatants in these contexts. “Rebel groups” are defined here as nongovernmental organizations using violence to achieve a political agenda. This definition fits closely with the UCDP/PRIO definition of “armed actors.” See Nils Petter Gleditsch et al., “Armed Conflict 1946–2001: A New Dataset,” Journal of Peace Research, Vol. 39, No. 5 (2002), pp. 615–637, which I use as the basis for the cross-national data. “Governments” are defined here as the actors that rebel groups oppose. In most cases, “government” refers to the established state before the civil conflict began. In other cases, especially conflicts after a colonial power withdraws (as in the Republic of the Congo), I consider the incumbent or, if none exists, the side that has more control and international recognition, as the “government.” I generally treat rebel groups and governments as unitary actors for simplicity, although they may have different factions with different preferences. Further study could also consider these dynamics.

16.

See, for example, Jason Miklian, “Revolutionary Conflict in Federations: The Indian Case,” Conflict, Security & Development, Vol. 11, No. 1 (March 2011), p. 43.

17.

See, most notably, I. William Zartman, Ripe for Resolution: Conflict and Intervention in Africa (New York: Oxford University Press, 1985).

18.

Identifying a settlement acceptable to warring factions can be difficult if the factions disagree on their relative strength. Most theorists, however, believe that information asymmetries that inhibit identification are resolved soon after conflict starts because fighting reduces combatants' abilities to bluff about capabilities and resolve. See, for example, Robert Powell, “War as a Commitment Problem,” International Organization, Vol. 60, No. 1 (Winter 2006), pp. 169–203.

19.

Barbara F. Walter, “The Critical Barrier to Civil War Settlement,” International Organization, Vol. 51, No. 3 (Summer 1997), pp. 335–364; Walter, Committing to Peace; James D. Fearon, “Why Do Some Civil Wars Last So Much Longer Than Others?” Journal of Peace Research, Vol. 41, No. 3 (May 2004), pp. 275–301; Caroline A. Hartzell and Matthew Hoddie, Crafting Peace: Power-Sharing Institutions and the Negotiated Settlement of Civil Wars (University Park: Pennsylvania State University, 2007); David A. Lake, “International Relations Theory and Internal Conflict: Insights from the Interstices,” International Studies Review, Vol. 5, No. 4 (2003), pp. 81–90; and Michaela Mattes and Burcu Savun, “Fostering Peace after Civil War: Commitment Problems and Agreement Design,” International Studies Quarterly, Vol. 53, No. 3 (September 2009), pp. 737–759.

20.

Similarly categorized by Fortna, Does Peacekeeping Work? pp. 82–85.

21.

Combatants are also wary because it is hard to confidently predict the consequences of each institutional change. In particular, rebels, who may not know the existing state structures as well as governments, may seek external guarantees to ensure that governments do not use the state apparatus to resist changes to the distribution of power. See James D. Fearon and David Laitin, “Civil War Termination,” paper presented at the annual meeting of the American Political Science Association, Chicago, Illinois, 2007.

22.

Walter, Committing to Peace; Hartzell and Hoddie, Crafting Peace; and Mattes and Savun, “Fostering Peace after Civil War.”

23.

Walter, “Designing Transitions from Civil War.”

24.

Walter, Committing to Peace; Leonard Wantchekon, “The Paradox of ‘Warlord’ Democracy: A Theoretical Investigation,” American Political Science Review, Vol. 98, No. 1 (February 2004), pp. 17–33; T. Clark Durant and Michael Weintraub, “How to Make Democracy Self-Enforcing after Civil War: Enabling Credible yet Adaptable Elite Pacts,” Conflict Management and Peace Science, Vol. 31, No. 5 (November 2014), pp. 521–540; and Caroline A. Hartzell and Matthew Hoddie, “The Art of the Possible: Power Sharing and Post–Civil War Democracy,” World Politics, Vol. 67, No. 1 (January 2015), pp. 37–71.

25.

See, for examples, Benjamin Reilly and Per Nordlund, eds., Political Parties in Conflict-Prone Societies: Regulation, Engineering, and Democratic Development (Tokyo: United Nations University Press, 2008). More generally, see Arend Lijphart, The Politics of Accommodation: Pluralism and Democracy in the Netherlands (Berkeley: University of California Press, 1975).

26.

Walter, Committing to Peace.

27.

The agreement's design can also further decrease the amount of leverage needed. Through incrementalism, it can minimize the amount of power each side could grab by noncompliance. For instance, the agreement can specify that units of each side demobilize in alteration, so that each side only ever has slightly less relative strength. The agreement might call for land redistribution or distribution of other political benefits to occur in a similar fashion.

28.

Based on Walter, Committing to Peace, although Walter does not specify the mechanism as forceful coercion. Others suggest numerous potential mechanisms. See, for example, Fortna, Does Peacekeeping Work?

29.

Fortna, Does Peacekeeping Work? pp. 89–93, 98–102; and Aila M. Matanock and Adam G. Lichtenheld, “How Does International Intervention Work? Mechanisms for Securing Settlements to Civil Conflicts,” paper presented at the annual meeting of the American Political Science Association, San Francisco, California, 2015.

30.

Even when deployed, armed peacekeepers often draw down or withdraw their forces rather than punish noncompliers, and they are often deployed for shorter periods than needed to implement peace settlements. See Fortna, Does Peacekeeping Work? pp. 87–89.

31.

Initial evidence suggests that electoral participation provisions may engage external actors. See, for example, Krishna Kumar, ed., Postconflict Elections, Democratization, and International Assistance (Boulder, Colo.: Lynne Rienner, 1998); Terrence Lyons, Demilitarizing Politics: Elections on the Uncertain Road to Peace (Boulder, Colo.: Lynne Rienner, 2005); and Rafael Lopez-Pintor, “Postconflict Elections and Democratization: An Experience Review” (Washington, D.C.: USAID, 2005).

32.

See James D. Fearon, “Self-Enforcing Democracy,” Quarterly Journal of Economics, Vol. 126, No. 4 (2011), p. 1662 (emphasis in the original). Coordination, a crucial element in this argument, can be accomplished through a variety of institutions, but electoral processes may work best to provide information from multiple actors and respond to noncompliance with an agreement.

33.

The elections that they produce, however, need not be open. In fact, the elections need to be designed so that payoffs from the electoral process are greater than those gained from war, as discussed above. Any settlement requires such payoffs. Depending on the relative power of the combatants, the payoffs may include party funding (e.g., Colombia 1990–91), substantial institutional bias toward the status quo (e.g., South Africa 1991), or even ethnic quotas (e.g., Bosnia and Herzegovina 1995).

34.

Indeed, Matanock, “International Insurance”; and Matanock, Electing Peace demonstrate that in more than 90 percent of cases in which first elections occurred, second elections were held without major delays.

35.

In Central America, for example, those working with the United Nations found domestic society to be susceptible to manipulation by politicians as well as alienated from the political process. See Jean Arnault, “Good Agreement? Bad Agreement? An Implementation Perspective,” Princeton University, 2011, p. 11; and William Stanley, Enabling Peace in Guatemala: The Story of MINUGUA (Boulder, Colo., Lynne Rienner, 2013), p. 289.

36.

Wantchekon, “The Paradox of ‘Warlord’ Democracy,” pp. 17, 27. Elections also offer the possibility of incremental change by gradually allowing more input from citizens as they become more effective at carrying out these tasks themselves. See ibid.; and Emily Beaulieu, Electoral Protest and Democracy in the Developing World (Cambridge: Cambridge University Press, 2014), p. 132.

37.

See Michael E. Brown, “The Causes and Regional Dimensions of Internal Conflict,” in Brown, ed., The International Dimensions of Internal Conflict (Cambridge, Mass.: MIT Press, 1996).

38.

See, for example, Michael Gilligan and Stephen John Stedman, “Where Do the Peacekeepers Go?” International Studies Review, Vol. 5, No. 4 (December 2003), pp. 37–54; and Fortna, Does Peacekeeping Work?

39.

Jennifer L. McCoy, “Mediating Democracy: A New Role for International Actors?” in David C. Bruce, ed., New World Order: Social and Economic Implications (Atlanta: Georgia State University Business Press, 1993), pp. 129–140.

40.

Hyde, The Pseudo-Democrat's Dilemma, p. 8.

41.

Alberto Simpser and Daniela Donno, “Can International Election Monitoring Harm Governance?” Journal of Politics, Vol. 74, No. 2 (2012), pp. 501–513.

42.

Hyde, The Pseudo-Democrat's Dilemma; and Judith G. Kelley, Monitoring Democracy: When International Election Observation Works, and Why It Often Fails (Princeton, N.J.: Princeton University Press, 2012). Similar diffusion arguments can be found in, for example, Beth A. Simmons and Zachary Elkins, “The Globalization of Liberalization: Policy Diffusion in the International Political Economy,” American Political Science Review, Vol. 98, No. 1 (February 2004), pp. 171–189.

43.

Hyde uses regional percentage of elections observed to predict election observation in the next year. See Hyde, The Pseudo-Democrat's Dilemma, pp. 81–83. Simpser and Donno use regional percentage of elections observed by high-quality monitors as an instrument. They argue that these regional trends in observation are likely driven by the preferences and capabilities of international actors sending these missions. They then perform tests showing that these trends predict engagement in the particular state. See Simpser and Donno, “Can International Election Monitoring Harm Governance?”

44.

Post-conflict elections, in general, receive more observation than other kinds of elections, though the degree varies by region. Cases of post-conflict elections that did not receive electoral observation include Mali after 1991, Djibouti after 1994, and the Philippines after 1995. (Peace agreements in these cases did not include participation provisions, likely because of low expectations of engagement.)

45.

Hyde, The Pseudo-Democrat's Dilemma; and Kelley, Monitoring Democracy.

46.

Kelley, Monitoring Democracy, p. 63.

47.

Ibid.; and Kovacs, “From Rebellion to Politics,” pp. 134–156.

48.

Kelley refers to the differences in the standards employed in post-conflict contexts as “bias.” See Kelley, Monitoring Democracy, p. 62. These differences, however, help make observation effective in binding combatants to a settlement. See McCoy, “Mediating Democracy,” especially pp. 131–133; and Thomas Carothers, Aiding Democracy Abroad: The Learning Curve (Washington, D.C.: Carnegie Endowment for International Peace, 1999). Observers prioritizing the settlements' terms when defining compliance, rather than democratic norms, is consistent with external engagement theory.

49.

Other work has identified similar mechanisms used to improve elections and human rights protections. See Simpser and Donno, “Can International Election Monitoring Harm Governance?”; and Richard A. Nielsen and Beth A. Simmons, “Rewards for Ratification: Payoffs for Participating in the International Human Rights Regime?” International Studies Quarterly, Vol. 59, No. 2 (June 2015), pp. 197–208.

50.

McCoy, “Mediating Democracy,” p. 133.

51.

Much of this assistance was provided by the United States. See Dinorah Azpuru et al., “What Has the United States Been Doing?” Journal of Democracy, Vol. 19, No. 2 (2008), pp. 150–159, (data in supplementary files). See also Carothers, Aiding Democracy Abroad, pp. 6, 85; Steven Levitsky and Lucan A. Way, Competitive Authoritarianism: Hybrid Regimes after the Cold War (Cambridge: Cambridge University Press, 2010), pp. 14–32; and Hyde, The Pseudo-Democrat's Dilemma, pp. 143–144.

52.

See, for example, Carothers, Aiding Democracy Abroad.

53.

Adriano Nuvunga, “Post-war Reconstruction in Mozambique: The United Nations' Trust Fund to Assist the Former Rebel Movement RENAMO” (Maputo: Centro de Integridade Publica, 2007), pp. 10–14.

54.

Stanley, Enabling Peace in Guatemala, p. 240.

55.

Whereas pressure is common, punishment is relatively rare, because the threat of conditionality should be enough to prevent violations. Barry R. Weingast, “Off-the-Path Behavior: A Game-Theoretic Approach to Counterfactuals and Its Implications for Political and Historical Analysis,” in Phillip E. Tetlock and Aaron Belkin, eds., Counterfactual Thought Experiments in World Politics: Logical, Methodological, and Psychological Perspectives (Princeton, N.J.: Princeton University Press, 1996), pp. 230–243.

56.

See Carew Boulding and Susan D. Hyde, “Political Terror, Election Fraud, and Foreign Aid: When Do Donors Withdraw Aid to Promote Democracy?” paper presented at the annual conference of the Midwest Political Science Association, Chicago, Illinois, 2005; and Inken von Borzyskowski and Felicity Vabulas, “The Punishment Phase: IGO Suspensions after Political Backsliding,” paper presented at the annual conference on the Political Economy of International Organizations, 2014. These measures, of course, rely on counterfactuals, and it is hard to assess “the amount of aid that might have been received under different circumstances.” See Boulding and Hyde, “Political Terror, Election Fraud, and Foreign Aid,” p. 22. Case evidence can also be found in Carrie Manning and Monica Malbrough, “Bilateral Donors and Aid Conditionality in Post-conflict Peacebuilding: The Case of Mozambique,” Journal of Modern African Studies, Vol. 48, No. 1 (March 2010), pp. 143–169.

57.

Author interview with Condoleezza Rice, May 28, 2013, Stanford University, Palo Alto, California.

58.

On the media, see Guy Golan and Wayne Wanta, “International Elections on US Network News: An Examination of Factors Affecting Newsworthiness,” International Communication Gazette, Vol. 65, No. 1 (2003), pp. 25–39; evidence on observation and assistance is discussed below. Kelley also shows that two-thirds of the post-conflict elections in her sample were observed, compared to just one third of other elections. See Kelley, Monitoring Democracy, p. 33.

59.

See Levitsky and Way, Competitive Authoritarianism, p. 14; and Hyde, The Pseudo-Democrat's Dilemma.

60.

A special issue of Governance makes this argument regarding legitimacy in the context of intervention. See Stephen D. Krasner and Thomas Risse, “External Actors, State-Building, and Service Provision in Areas of Limited Statehood,” Governance, Vol. 27, No. 4 (October 2014), pp. 545–567; and Matanock, “Governance Delegation Agreements: Shared Sovereignty as a Substitute for Limited Statehood,” Governance, Vol. 27, No. 4 (October 2014), pp. 589–612.

61.

Fortna examines and overturns the same argument about armed peacekeeping. See Fortna, Does Peacekeeping Work?

62.

A correlation between provisions and peace still suggests that provisions have effects, even if their strategic selection causes those effects. See Kenneth A. Schultz, “The Enforcement Problem in Coercive Bargaining: Interstate Conflict over Rebel Support in Civil Wars,” International Organization, Vol. 64, No. 2 (Spring 2010), pp. 281–312.

63.

See, for example, James D. Fearon, “Governance and Civil War Onset” (Washington, D.C.: World Bank, 2011); and Barbara F. Walter, “Why Bad Governance Leads to Repeat Civil War,” Journal of Conflict Resolution, Vol. 59, No. 7 (October 2015), pp. 1242–1272.

64.

Lotta Harbom, Stina Högbladh, and Peter Wallensteen, “Armed Conflict and Peace Agreements,” Journal of Peace Research, Vol. 43, No. 5 (September 2006), pp. 617–631; and Stina Högbladh, “Peace Agreements, 1975–2011,” in Therese Pettersson and Lotta Themnér, eds., States in Armed Conflict, 2011 (Uppsala, Sweden: Department of Peace and Conflict, Uppsala University, 2012), pp. 39–56. The UCDP includes more agreements than do other datasets, such as the Peace Accord Matrix (PAM), encompassing almost all the agreements in other datasets. The PAM data are quite similar to the data I use when clustering peace processes (see below).

65.

Samuel P. Huntington, “Democracy's Third Wave,” Journal of Democracy, Vol. 2, No. 2 (Spring 1991), pp. 12–34; and Levitsky and Way, Competitive Authoritarianism, p. 14.

66.

Clustered peace processes include, for example, the UN-run negotiations in El Salvador. In other cases, earlier accords in processes are not recorded, so clustering also reduces reporting bias. I control for provisions calling for further processes that do not take place, and the results hold. Other datasets, such as the PAM noted above, also cluster.

67.

For example, in Cambodia, three rebel groups signed the Paris agreement in 1991, but the Khmer Rouge soon returned to combat. In my analysis, most of the comparison is across agreements (only 11 percent of the settlements have multiple rebel groups signing). Therefore, though I analyze dyadic results to accurately test the theory, the results are similar, though sometimes weaker, on the agreement level. This weakness is not surprising, given that the sample size is smaller. In addition to this robustness check, I also always cluster the standard errors by state throughout the analyses.

68.

Nils Petter Gleditsch et al., “Armed Conflict 1946–2001”; and Lotta Themnér and Peter Wallensteen, “Armed Conflicts, 1946–2011,” Journal of Peace Research, Vol. 49, No. 4 (2012), pp. 565–575.

69.

For robustness checks, I use the UCDP coding of whether implementation failed, including failure because of lower-level conflict or renegotiation, and the results hold. Additionally, a few initial agreements were renegotiated without intervening conflict; I exclude renegotiations because keeping them would inflate the success ex post (as noted in table A1 in the supporting information). The results hold either way.

70.

Other specifications of the dependent variable, including different specifications of time until failure as well as the level of failure, described above, are also included.

71.

Among all agreements, 18 cases are missing full text. In 8 cases, detailed summaries provide confidence for interpreting provisions, but in 10 cases, all Chadian, even detailed summaries are missing. The results hold when dropping the uncertain cases.

72.

The data are shown in the supporting information (table A1). The supporting information, as well as the data, code, and results are available at https://dataverse.harvard.edu/dataverse/matanock.

73.

A few less clear cases are recoded as robustness checks: peace agreements that open the way for multiparty competition broadly and those where combatants participated in elections before the conflict are included; and those that integrate groups in regional but not national transitional governments before elections are excluded (table A1). The results hold.

74.

This definition of electoral participation provisions includes provisions that are not implemented. Comparing electoral participation with other provisions whose implementation may be harder to observe could bias my results. In rare cases when the government or the rebel group failed to implement electoral provisions, the uncooperative side was usually punished, as external engagement theory anticipates. Evidence on this appears in Matanock, “International Insurance”; and Matanock, Electing Peace. Full implementation by 2015 occurred in 34 agreements with electoral participation provisions, including 6 that were later renegotiated, producing peaceful electoral participation by the ex-rebel parties.

75.

Existing intervention literature also portrays the end of the Cold War as a shock that should increase expectations and thus involvement of international actors. I follow one prominent study in using 1989 as the date of the Cold War's end. See Fortna, Does Peacekeeping Work?

76.

I follow Hyde, The Pseudo-Democrat's Dilemma, in designating five regions: sub-Saharan Africa, the Americas, Asia, Europe, and North Africa and the Middle East. The region measure is the same for both proxy variables discussed below. I also code alternative specifications, however, such as continents, clusters of countries sharing regional organizations, and UCDP/PRIO's classification of regions. None of these alternative specifications alter the results substantially.

77.

Susan D. Hyde and Nikolay Marinov, “Which Elections Can Be Lost?” Political Analysis, Vol. 20, No. 2 (Spring 2012), pp. 191–201.

78.

Azpuru et al., “What Has the United States Been Doing?”

79.

Hyde, The Pseudo-Democrat's Dilemma; and Kelley, Monitoring Democracy.

80.

Azpuru et al., “What Has the United States Been Doing?”; and Richard Youngs, “What Has Europe Been Doing?” Journal of Democracy, Vol. 19, No. 2 (April 2008), pp. 160–169.

81.

Hyde and Simpser and Donno use the regional election observation trend in similar ways. See Hyde, The Pseudo-Democrat's Dilemma; and Simpser and Donno, “Can International Election Monitoring Harm Governance?” Beth A. Simmons similarly uses regional trends in other forms of international involvement. See Simmons, Mobilizing for Human Rights: International Law in Domestic Politics (Cambridge: Cambridge University Press, 2009).

82.

The effect of receiving observers or aid during the elections, like that of holding elections, is quite strong. Aside from being highly endogenous, however, this effect is based on an ex post measure of implementation of provisions. Successful implementation should, of course, be correlated with peace.

83.

Simpser and Donno, “Can International Election Monitoring Harm Governance?” p. 507.

84.

I argue that electoral participation provisions facilitate international engagement (both international observation and associated aid), but observation and aid alone might also produce enduring peace. The interaction terms test each proposition.

85.

A log-rank test for equality of survivor functions suggests a statistically significant difference between the curves (at a 0.01 level).

86.

This variable captures all but four failures, including the failure of the peace agreement between the Colombian government and Ejército Popular de Liberación, for example, which failed after thirteen years. Scholars commonly use such methods to model correlations between the type of termination and conflict recurrence. See, for example, Monica Duffy Toft, “Ending Civil Wars: A Case for Rebel Victory?” International Security, Vol. 34, No. 4 (Spring 2010), pp. 7–36. Given the small size of the dataset and small number of controls, however, I re-ran the analysis using generalized linear models; the results are similar. Finally, conflict recurrence could also be conceptualized via analysis that explicitly models the time to recurrence as a function of the conflict and settlement characteristics—and also potentially of post-settlement characteristics. I therefore also re-ran the analysis using duration models. Note that failure within five years is a hazard rate, as well. See Janet M. Box-Steffensmeier and Christopher J.W. Zorn, “Duration Models and Proportional Hazards in Political Science,” American Journal of Political Science, Vol. 45, No. 4 (October 2001), pp. 972–988. The results hold. The supporting information shows these results.

87.

The results also hold when clustering by conflict.

88.

See the cross-tabulations in the supporting information (table A5). If the independent variable is the implementation of participatory post-conflict elections, rather than the inclusion of electoral participation provisions, conflict recurs in 7 percent of cases, compared to 55 percent of those in which participatory post-conflict elections were not held. This result is even stronger than the result for inclusion of electoral participation provisions. Including implementation, however, is determined post-treatment and potentially biased; thus, provisions are the preferred measure.

89.

See table 7.0-2 in the supporting information, which shows that a similar relationship holds with potentially confounding variables. All analysis was run in Stata 13, and the interpretations were produced with the new margins and marginsplot commands, which can be applied to interactions. See Richard Williams, “Using the Margins Command to Estimate and Interpret Adjusted Predictions and Marginal Effects,” Stata Journal, Vol. 12, No. 2 (2012), pp. 308–331.

90.

The two democracy-promotion variables are most plausible as proxies for expectations of sufficiently impartial external engagement. These variables introduce exogeneity, and they could even be thought of as instruments for electoral participation provisions. As a robustness check, I used both variables as instruments, separately and together, and the results hold.

91.

Coding electoral participation provisions is more difficult for a few early cases, so those cases are included in the re-coding robustness check noted above, and the results hold.

92.

The models can be found in the supporting information (table A6). The interpretation follows the recommendation of Thomas Brambor, William Roberts Clark, and Matt Golder, “Understanding Interaction Models: Improving Empirical Analyses,” Political Analysis, Vol. 14, No. 1 (2006), pp. 63–82.

93.

Fortna examines and overturns a similar selection argument about armed peacekeeping. See Fortna, Does Peacekeeping Work? Even if such a selection argument held, combatants' decision to include these provisions suggests that the provisions have an effect on enduring peace; otherwise they should not be systematically selected. This point about selection is made by Schultz, “The Enforcement Problem in Coercive Bargaining.”

94.

To define which cases are hard to settle, Fortna chooses a sample (conflicts before 1989) in which international involvement is uncommon and therefore does not drive which cases successfully settle. See Fortna, Does Peacekeeping Work? pp. 187–189.

95.

See, for example, Fearon and Laitin, “Ethnicity, Insurgency, and Civil War.”

96.

This result builds on Fortna, Does Peacekeeping Work?

97.

See table A4 in the supporting information. These results generally suggest that strong rebel groups operating in weak states may be better able to secure electoral participation provisions, compared to weak rebel groups operating in strong states. The former may have more leverage in negotiations, and, crucially, may also be more likely to win seats in elections after settlements. Cases with strong rebels (and a weak government) may nevertheless be harder to settle than those with weak rebels (and a strong government), because strong rebels can fight harder than weak rebels, making a settlement more precarious. The correlations between strength and electoral participation provisions are not robust, however. Many of the measurements of relative strength are flawed—rebel troop count, for example, is coded for the entire conflict period prior to the peace agreement even though it changes over time—and the correlations between strength and enduring peace are not consistent once other controls are added. In addition, rebel groups vary substantially in their relative strength at the ballot box, obtaining from 0 to 80 percent of the vote share in post-conflict elections (see the supporting information, table A4.0-1, figure A2). Nonetheless, I control for measures of strength when evaluating the effect of electoral participation provisions on peace, and the results hold, suggesting that the provisions may have an effect beyond signaling rebel group strength.

98.

Matanock, “International Insurance”; Matanock “External Engagement”; and Matanock, Electing Peace.

99.

See, for example, Walter, Committing to Peace; Barbara F. Walter, “Does Conflict Beget Conflict? Explaining Recurring Civil War,” Journal of Peace Research, Vol. 41, No. 3 (May 2004), pp. 371–388; Michael W. Doyle and Nicholas Sambanis, Making War and Building Peace: United Nations Peace Operations (Princeton, N.J.: Princeton University Press, 2006); Bumba Mukherjee, “Why Political Power-Sharing Agreements Lead to Enduring Peaceful Resolution of Some Civil Wars, but Not Others?” International Studies Quarterly, Vol. 50, No. 2 (June 2006), pp. 479–504; Hartzell and Hoddie, Crafting Peace; J. Michael Quinn, T. David Mason, and Mehmet Gurses, “Sustaining the Peace: Determinants of Civil War Recurrence,” International Interactions, Vol. 33, No. 2 (2007), pp. 167–193; Collier, Hoeffler, and Söderbom, “Post-Conflict Risks”; Fortna, Does Peacekeeping Work?; Michael J. Gilligan and Ernest J. Sergenti, “Do UN Interventions Cause Peace? Using Matching to Improve Causal Inference,” Quarterly Journal of Political Science, Vol. 3, No. 2 (2008), pp. 89–122; Toft, “Ending Civil Wars”; Fearon, “Self-Enforcing Democracy”; Brancati and Snyder, “Time to Kill”; and Flores and Nooruddin, “The Effect of Elections on Post-Conflict Peace and Reconstruction.”

100.

101.

102.

I also include good governance controls (including degree of corruption, bureaucratic quality, and law and order) in the supporting information (table A7.2). Elections may foster the rule of law and increase the effectiveness of bureaucracies, while reducing corruption, which could explain why peace endures. See, for example, Fearon, “Governance and Civil War Onset”; and Walter, “Why Bad Governance Leads to Repeat Civil War.” These variables, however, are not highly correlated with electoral participation provisions.

103.

Summary statistics and data sources for all of these controls—and for settlement provisions, discussed below—are available in the supporting information. I also include alternative variables and the results hold. See table A2 and codebook.

104.

I show ten-year indicators, but the results also hold with five-year indicators. One-year indicators interacted with region indicators approach collinearity with the regional democracy promotion variables, producing models that do not converge.

105.

I use the UCDP data to capture these provisions, which broadly define each of these categories. See Harbom, Högbladh, and Wallensteen, “Armed Conflict and Peace Agreements.”

106.

Strategic inclusion may explain some cases in which peace agreements include multiple provisions. See, for example, Hartzell and Hoddie, “The Art of the Possible.”

107.

Full models are shown in table A7.0 in the supporting information.

108.

Robustness checks include dropping each conflict and using alternative control variables. The results hold.

109.

This result differs from other findings, perhaps because the sample is limited to settlements.

110.

See table A7.2 in the supporting information.

111.

Aside from demobilization, disarmament, and reintegration provisions (36 percent), no other provisions are highly correlated with electoral participation provisions. Economic, military, and other political power-sharing provisions represent three of the four main categories of provisions identified in the literature. See Hartzell and Hoddie, Crafting Peace. Variables representing the fourth category of provisions, territorial power-sharing, are negatively correlated with participation provisions. This evidence is also not consistent with the selection effect.

112.

I appreciate an anonymous reviewer's suggestion to examine the first comparison. Much of the work on conflict recurrence indicates that victories have an important effect. See, for instance, Toft, “Ending Civil Wars.”

113.

Data are from Joakim Kreutz, “How and When Armed Conflicts End: Introducing the UCDP Conflict Termination Dataset,” Journal of Peace Research, Vol. 47, No. 2 (2010), pp. 243–250. This additional analysis could suffer from bias, however, because it requires that a conflict successfully ends for a year to be included in the sample, unlike the other analysis in this article.

114.

Former combatants instead may be given non-elected positions in the cabinet, military, or other government bodies, for instance.

115.

Another relevant comparison is how much this reduces conflict recurrence in all post-conflict elections that occur once all militant groups have signed on. Considering data on such post-conflict elections (Flores and Nooruddin, “The Effect of Elections on Post-Conflict Peace and Reconstruction”), I examine a variable for those produced by electoral participation provisions, and show it has a larger effect on enduring peace. See Aila M. Matanock, “Using Violence, Seeking Votes: Introducing the Militant Group Electoral Participation (MGEP) Dataset,” Journal of Peace Research, Vol. 53, No. 6 (2016), pp. 845–853. This additional analysis could, however, suffer bias because it requires successfully holding participatory elections.

116.

See, for example, Collier, Hoeffler, and Söderbom, “Post-Conflict Risks.”

117.

The first comparison is statistically significant at the standard levels. See the supporting information (table A13).

118.

Daniela Donno, Defending Democratic Norms: International Actors and the Politics of Electoral Misconduct (Oxford: Oxford University Press, 2013), p. 204.

119.

John Gerring, “Is There a (Viable) Crucial-Case Method?” Comparative Political Studies, Vol. 40, No. 3 (2007), p. 232.

120.

This could also be described as a shock across which I can compare the state to its best control, itself.

121.

Given that these changes are driven by geostrategic factors exogenous to the civil conflict in any particular state, they are a particularly clean shock to examine.

122.

The cases are Angola, Cambodia, Colombia, El Salvador, Guatemala, Mozambique, Philippines/Mindanao, South Africa, Sudan/South Sudan, and the United Kingdom/Northern Ireland. They are discussed in more depth in Matanock, “International Insurance”; and Matanock, Electing Peace.

123.

Enrique A. Baloyra, “El Salvador: From Despotism to Partidocracia,” in Kumar, Postconflict Elections, Democratization, and International Assistance, pp. 15, 17, 31.

124.

Terry L. Karl, “El Salvador's Negotiated Revolution,” Foreign Affairs, Vol. 71, No. 2 (Spring 1992), pp. 148–149.

125.

As a member of the FMLN General Command, Shafick Jorge Handel, noted in a policy document. See Handel, “Análisis de la Situación Político-Militar Hecho por el Miembro de la Comandancia General del FMLN,” box 3, 1984, El Salvador, Salvadoran Subject Collection (1976–94), Hoover Institution Archives, Stanford, California, p. 15.

126.

Alberto Martín Álvarez, From Revolutionary War to Democratic Revolution (Berlin: Berghof Research Center, 2010), pp. 24–25, 28; Karl, “El Salvador's Negotiated Revolution,” pp. 151–159; and James LeMoyne, “Out of the Jungle—in El Salvador, Rebels with a New Cause,” New York Times, February 9, 1992.

127.

Elizabeth Jean Wood, Forging Democracy from Below: Insurgent Transitions in South Africa and El Salvador (Cambridge: Cambridge University Press, 2000), p. 80.

128.

Álvarez, From Revolutionary War to Democratic Revolution, p. 24.

129.

Handel, “Salvadoran Guerrilla Letter to POTUS,” December 8, 1988, box CF00196B1, Latin American Directorate Files, National Security Council, George H.W. Bush Presidential Library, College Station, Texas, p. 1. Talks initially seemed to be put on hold, however. See “Re: GOES/FMLN Talks,” September 18, 1991, box CF00196B2, Latin American Directorate Files, National Security Council, George H.W. Bush Presidential Library, p. 2.

130.

Wood, Forging Democracy from Below, p. 80.

131.

Interview with FMLN leadership quoted in LeMoyne, “Out of the Jungle,” for example.

132.

Herman Rosa and Michael Foley, “El Salvador,” in Shepard Forman and Stewart Patrick, eds., Good Intentions: Pledges of Aid for Postconflict Recovery (Boulder, Colo.: Lynne Rienner, 2000), p. 151.

133.

See, for example, Wood, Forging Democracy from Below.

134.

Karl, “El Salvador's Negotiated Revolution,” p. 160.

135.

Fortna, Does Peacekeeping Work?

136.

Joseph G. Sullivan, “How Peace Came to El Salvador,” Orbis, Vol. 38, No. 1 (Winter 1994), pp. 86–87.

137.

Rosa and Foley, “El Salvador”; Sullivan, “How Peace Came to El Salvador,” pp. 86–87; and Wood, Forging Democracy from Below, p. 94.

138.

William Stanley and David Holiday, “Peace Mission Strategy and Domestic Actors: UN Mediation, Verification, and Institution-Building in El Salvador,” International Peacekeeping, Vol. 4, No. 2 (Summer 1997), p. 24.

139.

See interviews with the FMLN rebel leaders and well as international actors in LeMoyne, “Out of the Jungle.”

140.

Fortna, Does Peacekeeping Work? pp. 90–94, 162; Howard, UN Peacekeeping in Civil Wars, pp. 94, 101; Rosa and Foley, “El Salvador,” p. 137; and Wood, Forging Democracy from Below, pp. 90–91.

141.

Fortna, Does Peacekeeping Work?; and Howard, UN Peacekeeping in Civil Wars, p. 94.

142.

Rosa and Foley, “El Salvador,” p. 142.

143.

Howard, UN Peacekeeping in Civil Wars, p. 94; and Stanley and Holiday, “Peace Mission Strategy and Domestic Actors.”

144.

Charles T. Call, “Assessing El Salvador's Transition from Civil War to Peace,” in Stephen John Stedman, Donald Rothchild, and Elizabeth M. Cousens, eds., Ending Civil Wars: The Implementation of Peace Agreements (Boulder, Colo.: Lynne Rienner, 2002), p. 581.

145.

Human Rights Watch, “Americas Report, El Salvador” (New York, N.Y.: Human Rights Watch, 1994); and William LeoGrande, “Foreign Aid to Central America: Is the Era of Aid Over?” in Richard Grant and Jan Nijman, eds., Global Crisis in Foreign Aid (Syracuse, N.Y.: Syracuse University Press, 1998), p. 108.

146.

United Nations Observer Mission in El Salvador document and other evidence, cited in Tommie Sue Montgomery, “International Elections, Observing Elections, and the Democratic Transition in El Salvador,” in Kevin J. Middlebrook, ed., Electoral Observation and Democratic Transitions in Latin America (La Jolla: Center for U.S.-Mexican Studies, University of California, San Diego, 1998), p. 131.

147.

Ibid., p. 137; and Kelley, Monitoring Democracy, p. 225.

148.