Violent Elections and Citizens’ Support for Democratic Constraints on the Executive: Evidence From Nigeria

How do violent elections affect the willingness of citizens to defend democratic institutions? We argue that in the wake of violent elections, support for democratic constraints on the executive will diverge amongst ruling and opposition party supporters. To protect their position, ruling party supporters become more likely to endorse weakening constraints on executive power, even if it violates democratic principles. Opposition supporters, on the other hand, become more likely to reject democratic transgressions that de facto render them more vulnerable to political abuse. We examine these expectations using a vignette experiment embedded in a nationally representative 2019 post-election survey of 2400 Nigerians. Our findings suggest that incumbent supporters are overall more likely to endorse weaker constraints on the executive, but these attitudes are not reinforced by information about election violence. Opposition supporters, in contrast, become less likely to accept transgressions when informed about election violence.


Introduction
In elections across the world, many voters turn out to vote amidst threats, intimidation, and coercion.Election violence, affecting around a quarter of all electoral contests worldwide, undermines core democratic principles by violating the right of voters to express their political choice freely (Hafner-Burton et al., 2014;Norris, 2014).Elections not only represent a fundamental component of democracy by conferring the legitimate right to rule, but they also constitute particularly salient times for forming citizens' attitudes towards political institutions, their representatives, and democracy more broadly (McAllister & White, 2015;Moehler, 2009).What are the consequences of election violence on citizens' willingness to defend democratic institutions?
Existing research on this relationship paints a gloomy picture, suggesting that voters that are fearful of electoral violence also express lower support for and satisfaction with democracy in their country, rate the quality of their democracy lower, and are more supportive of a return to autocratic rule (Burchard, 2015;von Borzyskowski et al., 2022).The mechanisms invoked, which highlight voter disengagement, risk aversion, and subordination to threat, also resonate with studies demonstrating lower electoral participation (e.g., Bratton, 2008;Condra & Wright, 2019;Gutiérrez-Romero & LeBas, 2020) and stronger support for security-yielding candidates in elections held in the shadow of violence (e.g., Daly, 2019).We build on these important insights but nuance the assumption that all citizens react uniformly to contentious elections by endorsing more authoritarian forms of government.We argue, instead, that the effect of election violence on political attitudes will be conditional on citizens' affiliation with the ruling or opposition party.
Existing work suggests that citizens across the incumbent-opposition divide generally have different stakes in defending democratic institutions (Anderson et al., 2005;Mazepus & Toshkov, 2021).Ruling party supporters have fewer incentives to fear the expansion of power with an executive that (presumably) represent their interest.Opposition party supporters are, on the other hand, more vulnerable to the abuse of executive powers and thus more likely to support constraints.We argue that these positions are reinforced in electoral contexts where the partisan cleavage shapes not only political mobilization but also patterns of intimidation and violence.In these contexts, democratic regime features, such as democratic constraints on the executive and the protection of civil liberties and political rights, will carry real significance in structuring the political struggle between partisan groups. 1  Whereas ruling party supporters might seek stability and protection by conceding the executive more discretionary powers, we expect opposition supporters to reject transgressions on democratic principles that de-facto render them less secure.
Examining these theoretical arguments, we tie in with the broader research on popular underpinnings of democratic retrenchment by studying its determinants in a novel empirical setting.Research on citizens' willingness to condone violations of democratic principles by governments has highlighted polarization between partisan groups as an important driver (e.g., Graham & Svolik, 2020;Svolik, 2020).The empirical evidence, however, is drawn mainly from the US and other established democracies.We evaluate the salience of this political division in Nigeria, a case of a non-consolidated democracy, where partisan identity generally is weaker, but where violence and insecurity reinforce polarization along politically salient group cleavages.Understanding the possible micro-underpinnings of de-democratization in this type of setting is paramount.Not only do weak democracies, like Nigeria, constitute the most prevalent regime type today, but they also provide a context where the risk of de-democratization represents a genuine concern to voters.Nigeria's fragile political institutions and experience with electoral violence are characteristics shared by many countries on the continent and beyond. 2  We evaluate our pre-registered theoretical expectations using a vignette experiment embedded in a nationally representative survey of 2400 individuals fielded in Nigeria after the 2019 general elections.We randomly assign all 2400 individuals to one of four equally sized experimental conditions: A control group receiving a neutral introductory vignette and three treatment groups primed on the contentious electoral environment.Following the vignette, we probe citizens' support for democratic constraints on the executive by asking whether they (dis)agree with a series of presidential infringements on core democratic principles and checks and balances.These include bypassing other branches of government, placing restrictions on opposition parties, curtailing media freedom, and cracking down on citizens' political rights.Such democratic transgression by regimes in power is the most common form of current-day democratic erosion (e.g., Bermeo, 2016;Lührmann & Lindberg, 2019), and bears direct significance for our understanding of democratic backsliding in presidential systems, such as Nigeria.
We find that directing respondents' attention to the contentious features of the electoral environment influences their subsequent assessment of the importance of defending democratic institutions, but only for supporters of opposition parties.Whereas incumbent partisans are generally more accepting of presidential actions that violate democratic principles both in the absence and presence of a violence prime, they do not react to information about electoral contention in the direction implied by our hypothesis (as well as by much previous research).Opposition supporters primed on the violent electoral environment, on the other hand, strengthen their pro-democracy orientation and become more likely to condemn democratic transgressions.Our findings, therefore, indicate that many voters respond to violent electoral contention by reasserting their commitments to democracy.Rather than taking information about violent elections to testify to the limitations of democratic systems to ensure order, as existing research would suggest, many voters seem to believe that upholding democratic constraints on the executive becomes even more critical where elections are contentious.An empirical validation drawing on cross-national survey data from the Afrobarometer suggests that this result might not be unique to the Nigerian case.
By honing in on how partisan affiliations condition voters' reactions to violent elections, we also link up with an influential strand of researchfocused mainly on more consolidated democracies-on the winner/loser gap in political opinion (e.g., Anderson et al., 2005).We build on recent studies suggesting that political alignments to parties inside and outside power might underpin gaps, not only in voters' sense of political efficacy or system legitimacy but also in the type of institutions they prefer (see Mazepus & Toshkov, 2021;Singer, 2018;S ¸as ¸maz et al., 2022).We extend these arguments to an electoral context where security concerns are rampant.This context, we argue, should be particularly likely to aggravate this gap.Our empirical evidence from Nigeria does not support the argument that election violence makes ruling party supporters more likely to endorse the expansion of discretionary executive powers.The ruling party's (APC) recent ascent to power possibly makes its supporters more vigilant in upholding democratic constraints than they would be in situations where power relations are more entrenched, electoral turnover less likely, and utility from winner status thus more secure.
In the following section, we situate our study in existing research on popular support for democratic constraints in general and the role of insecurity, violent political contention, and elections in particular.We move on to present our theoretical argument and discuss political alignments as a potential source of heterogeneity in how citizens respond to violent elections.After motivating our case selection, we introduce our data and the survey experiment.The final sections present our empirical findings and discuss their implications and limitations.

Existing Research
Popular support for democratic government is often held to be a core prerequisite of the resilience and stability of democratic systems (Almond & Verba, 1963;Claassen, 2019).Over the past decade, there have, however, been growing concerns about the "fading allure of democratic government" (Plattner, 2017; see also Foa and Mounk, 2016).Partly, researchers warn that citizens' negative evaluations of the performance of their democratic regimes spill over into wavering support for general democratic regime principles.Partly, researchers find that citizens' commitment to democracy runs shallow: they display an abstract allegiance to democracy but at the same time show acceptance of authoritarian alternatives; they generally embrace elections but are weakly committed to liberal values such as the protection of political rights (e.g., Graham & Svolik, 2020;Singer, 2018).Scholars have thus raised concerns about the public's ability and willingness to check the authoritarian ambitions of popularly elected executives.Cases such as US, Brazil, India, Tanzania, and Hungary illustrate how "executive aggrandizement"-the incremental erosion of democratic norms and institutions by elected leaders who weaken checks on executive powers, curtail political and civil liberties, and undermine the integrity of the electoral system-represents the main threat to democracy these days (e.g., Bermeo, 2016;Lührmann & Lindberg, 2019).
Against this background, much attention has been channeled towards understanding the popular endorsement of leaders that subvert democratic norms and institutions from within, mainly drawing on evidence from Western democratic contexts.Insecurity and threat perceptions play an essential role in these accounts.Some studies focus on diffuse threats, for example, how sociocultural and economic transformations of Western societies challenge traditional hierarchies, producing an "authoritarian reflex" among socially conservative groups of voters (e.g., Gidron et al., 2020;Norris, 2019).Others relate these attitudes to group-based identity threats originating in migration and changing demographics (e.g., Dinas et al., 2019;Dustmann et al., 2019;Nordø & Ivarsflaten, 2022).Scholars have also highlighted the pernicious role of polarization in undermining voters' willingness to punish the authoritarian temptations of elected leaders.Intense social and political divisions aggravate mistrust and aggregate, in the extreme, to an environment where "each camp comes to view the other as an existential threat to the nation or their way of life" (Somer & McCoy, 2019;p. 9; see also Graham &Svolik, 2020 andSvolik 2020).
Some research focuses specifically on how violent threats shape citizen support for liberal democratic principles.Studies on the consequences of terrorism for US public opinion suggests, for example, that fear of future attacks increases an individual's willingness to trade off civil liberties for political order and security (e.g., Davis & Silver, 2004;Huddy et al., 2007).Others find that popular support for institutions and incumbents increases in the wake of terrorist attacks (e.g., Dinesen & Jaeger, 2013;Hetherington & Nelson, 2003), as individuals cope with threats to their life by turning to institutions and leaders believed to be capable of delivering stability and security (e.g., Landau et al., 2004). 3 Little research, however, focuses on the erosion of support for liberal democratic principles in contexts where democratic institutions are already weak and contested.Even less work considers how the prevalence of violence and insecurity in the electoral environment feeds into these processes.This is a significant gap, given that a substantial proportion of elections worldwide see violence and coercion directly tied to the electoral contest or announced electoral result (Birch et al., 2020), while electoral periods are at the same time particularly critical for shaping public attitudes to political institutions and representatives (McAllister & White, 2015). 4An exception is von Borzyskowski et al. (2022), who use survey data from the Afrobarometer and find that voters who fear election violence express lower satisfaction with democracy, rate the quality of democracy in their country lower, and display more support for a return to autocratic rule (see also Burchard, 2015).The mechanisms they invoke highlight voter disengagement, risk aversion, and political subordination in the context of threat.We contribute to this literature by unpacking heterogeneity in individual responses to violence and insecurity within the electoral context and providing experimental evidence from a high-stakes field setting in a weakly consolidated democracy.We focus specifically on citizens' support for the constraining principles of a liberal democratic order as these are particularly salient in current debates about democratic retrenchment.

Violent Elections and Partisan Support for Constraining the Executive
An important building block that is under-theorized in existing work is the role of citizens' political alignments in shaping their political attitudes in the wake of violent elections. 5To explore such heterogeneous impacts, we draw on an influential strand of research pointing to a gap in political attitudes between voters of the ruling and the opposition party, manifest, for example, in their perception of system performance, legitimacy, trust in institutions, and the degree to which they consent to government authority (e.g., Anderson et al., 2005;Esaiasson, 2011).These patterns suggest that citizens derive different utility from the democratic system depending on whether they align with parties inside and outside power.As a consequence, scholars worrying about democratic resilience have long highlighted the risk that opposition supporters turn their back on democracy and emphasized the importance of securing loser's consent to avoid interruptions of democratic order (Anderson et al., 2005;Moehler & Lindberg, 2009).Yet, a recent strand within this literature suggests that equally consequential for the fate of democracy are the attitudes and behavior of electoral winners.Whereas the risks that electoral processes derail may critically depend on losers accepting unfavorable outcomes, the protection of liberal democratic principles-such as checks and balances on the executive, respect for civil liberties, and independence of legislature and media-might depend as much on the willingness of regime supporters to monitor and constrain their leaders and protest undemocratic behavior.Complacent and loyal ruling supporters, confident in being represented by those holding power, might be more willing to give leaders discretion to act on their behalf and less willing to monitor their behavior (Mazepus & Toshkov, 2021;Moehler & Lindberg, 2009;Singer, 2018).They might also offer more benign interpretations of democratic transgression by the regime (Clark et al., 2019).Unconditional allegiance by electoral winners could thus underpin democratic erosion, making them less willing to sanction democratic transgression by leaders with authoritarian dispositions.Opposition supporters, conversely, could be more likely to sanction abuses of executive power due to their vulnerable political position (Mazepus & Toshkov, 2021).
Building on these insights, we argue that the influence of violent electoral contention on citizen support for the constraining principles of a liberal democratic order will be conditional on whether they support the ruling regime or the opposition.A central assumption in the above literature is that election winners and losers have different stakes in keeping the executive constrained.We propose that these preferences will be reinforced in electoral contexts where political actors, such as party agents or political leaders, deploy violence and intimidation to influence voters.To start with, election violence enhances the salience of the political cleavage between ruling and opposition party supporters.In violent elections, partisanship will shape not only patterns of political mobilization but also direct the patterns of threats, intimidation, and coercion.Political violence, in turn, is known to make the particular cleavage along which violence is committed more important to selfidentity (Hadzic et al., 2020;LeBas, 2006;Wood, 2008); to make in-and out-group divisions along this cleavage more salient (Bauer et al., 2014;Beber et al., 2014;Canetti et al., 2016;Scacco & Warren, 2021;Tajfel & Turner, 1979); and to produce divergent opinions and political preferences across these groups (e.g., Lupu & Peisakhin, 2017).Situations of insecurity can also induce prejudice that increases political intolerance and an unwillingness to grant rights and liberties to out-groups perceived to endanger the societal order (e.g., Conrad et al., 2018;Gibson, 1998).In short, we argue that contexts of violent polarization make inter-group relations more combative and citizens thus more likely to privilege institutions and policies that offer protection to their own group and restrict the influence of other groups.
Election violence also heightens the stakes attached to partisan group belongings.Losers of elections are generally more at risk of abuse of political power (Mazepus & Toshkov, 2021).Where violence becomes integral to the political competition, those on the losing side become particularly vulnerable to victimization.Groups outside of power overall face a heightened risks of illtreatment and violent marginalization, as repressive actions and restrictions of political rights and liberties are directed towards the opposition and those most likely to dissent (Bhasin & Gandhi, 2013;Davenport, 2004;Fjelde et al., 2021).For those constituencies represented by the regime, on the other hand, incumbency status offers security and control over resources that protect their privilege.Winning and losing status becomes, therefore, particularly consequential in violent contexts.
The combination of more divisive political cleavages and the heightened benefits and risks conferred by winner/loser status leads us to expect divergent preferences for upholding constraints on the executive across ruling and opposition party supporters.Ruling party supporters-who perceive benefits under the current distribution of powers-should become more likely to support the expansion of executive powers.These might be seen as tools to protect the interests and well-being of one's own group in a tense and insecure political environment while simultaneously reducing the ability of out-groups to pursue their interest.We expect the opposite for opposition supporters.Although violence might lead this group to withdraw their support for the particular regime and downgrade their satisfaction with how it is performing (as previous research has argued), we do not expect it to weaken their demand for democratic constraints.Violence should rather have the opposite effect.Excluded from power, opposition supporters are in particular need of the protection that constraints on the executive entail.Placing more discretionary power with the executive only renders opposition supporters more vulnerable to the threat of violence, intimidation, and coercive marginalization. 6 In sum, we expect election violence to affect incumbent and opposition supporters differently.Whereas incumbent supporters should become more likely to support the expansion of the executive's discretionary powers (Hypothesis 1), opposition supporters should become less likely to do so and instead become more vigorous supporters of democratic constraints (Hypothesis 2).
We make no specific prediction for non-partisans.Our theoretical mechanisms invoke group belonging, meaning they have a weak bearing for predicting how awareness of contentious elections impacts this group.Considering the benefits of democratic constraints for the population at large, there are reasons to believe that non-partisans will reject loosening these when alerted to violent threats.Without an incumbent partisan lens to legitimize non-liberal moves by the regime, non-partisan supporters-as opposition supporters-might be likely to defend democratic principles (Svolik, 2020).On the other hand, lacking the political efficacy and collective agency bestowed by partisan belonging, non-partisans might be less easily mobilized in support of democratic principles compared to opposition partisans (c.f.Bratton, 2008;Young, 2020).We return to this question in the analysis.

The Nigerian Context
We evaluate our hypotheses using evidence from Nigeria, which represents a good case for examining how priming citizens on violent features of their elections influence their support for constraints on the executive.First, Nigeria is an unconsolidated democracy, and the risk of autocratization through government transgression on democratic principles is a realistic concern for voters. 7The country transitioned from military to multi-party electoral rule in 1999 but quickly turned into a dominant party state under the People's Democratic Party's (PDP) tight control of the Presidency (Lewis & Kew, 2015).There is historically a strong culture of executive dominance, with abuses and encroachments into the jurisdiction of both the legislature and the judiciary (Baba, 2018, p. 271).Although the electoral cycle has remained uninterrupted since the transition, elections have been compromised by voter suppression, large-scale violence, bribery, ballot fraud, and administrative irregularities (e.g., Bratton, 2008;Kerr, 2018;Omotola, 2010).In recent years, however, the electoral landscape has become more competitive and, with that, more politically polarized.Until the 2015 elections, no opposition party could establish itself beyond a limited regional or sectarian constituency.Although the opposition used institutional venues to compete for power, PDP maintained its hegemony.The victory of the All Progressives Congress (APC) in the fiercely contested 2015 elections represented the first time in the history of independent Nigeria that the opposition won.Both parties carried their mobilization potential to the 2019 elections (Obe, 2019).Jointly, APC (with 56%) and PDP (with 41%) received 97% of the votes cast in the Presidential election.While Nigeria's political party system is weakly institutionalized, with fierce intra-party competition and frequent party-switching amongst the elite (Angerbrandt, 2020), the prevalence of one visible political cleavage between two dominating parties over the 2015 and 2019 electoral races provides a suitable context for exploring our theoretical expectations.Whereas partisan identities are weaker than in many Western democracies, recent research shows that partisan allegiances are not only prevalent in African politics (including Nigeria) but also structure voting behavior and political participation (Mattes & Krönke, 2020).Importantly, it is not the depth of ideological commitment to a party that matters for our argument but rather whether citizens identify as ruling or opposition party supporters more generally-and thus are election winners or losers.We acknowledge, however, that the recent entry of the ruling party into Nigerian politics and its relatively short history of incumbency will probably make Nigeria a hard case to test our hypotheses about the importance of ruling and opposition party affiliation.In terms of policy, the APC and PDP are by and large "indistinguishable from each other" (Campbell, 2020, p. 110), meaning that divergent political attitudes between regime and opposition supporters exposed to our vignette treatments are less likely to stem from underlying ideological differences than from the political status conferred by these allegiances.While the incumbent is a former military general and proclaimed tough on electoral security violations (Onuoha et al., 2020), the opposition also made security a central campaign issue in 2019. 8 The electoral environment in 2019 was tense.According to the EU observer mission, the 2019 elections were marked by "severe operational and transparency shortcomings [and] electoral security problems" (EU Election Observation Mission, 2019, p. 3).Security problems manifested in fatalities, contentious events involving security agencies, and the abuse of incumbency at the federal and state levels (p.32-33).The observer report also notes the failure of leading parties to reign in acts of violence and places the death toll from violence during the campaign and the three election days at 145 (p.33).Whereas the contentious electoral environment brings credibility to the experimental design because vignettes about violence present realistic scenarios to respondents, there is also a risk that respondents are already "treated" with knowledge about electoral insecurity. 9Yet, citizens' perceptions of the 2019 elections are sufficiently ambiguous to provide our priming experiment with significant leverage.Election quality in both the 2015 and the 2019 elections was significantly improved compared to the extremely violent and fraudulent elections preceding them.Reminding citizens of the elections, particularly against the background of Nigeria's recent authoritarian past, is also likely to prompt voters' perceptions of greater political efficacy and participation.This ambiguity in connotations-contentious elections on the one and improved electoral security and political participation on the other hand-provides leverage for the priming experiment while making Nigeria a hard case for testing our theory.
Nigeria is also a relevant but under-researched case.It is one of the world's largest electoral democracies, Africa's most populous state, and perhaps the most powerful country in the region.Its democratization trajectory and the influence of violent and troublesome elections could shape political developments beyond its borders.Understanding how violent elections influence political attitudes in Nigeria is thus important in its own right.

Data and Research Design
To examine the relationship between citizens' awareness of violent elections and support for constraints on the executive, we draw on evidence from a nationally representative post-election survey of 2400 Nigerian citizens.The survey was fielded between March 23 and April 16, 2019, after the conclusions of the 2019 elections. 10The target population was adult Nigerians aged eighteen years and above.Respondents were selected using a clustered, stratified, multi-stage random selection procedure across all of Nigeria's 37 states (including the Federal Capital Territory).Within each state, local government areas were first stratified based on urban and rural status and then selected based on probability proportional to size random sampling.To this end, we used the official population projections for 2016 based on the 2006 census.The respondents were selected randomly across 132 urban and 168 rural primary sampling units (PSUs), with a 50/50 quota for men and women.The survey was conducted face-to-face, using tablets.For a discussion of ethical considerations related to the study and steps taken to protect the integrity of our respondents, see Section A in the Supplementary appendix.

Experimental Design
In the survey, we embed a vignette experiment to isolate the effect of priming citizens on the violent aspects of their electoral environment.We randomly assign all 2400 individuals to one of four equally sized experimental conditions (N = 600): one control group receiving a neutral introductory vignette that reminds them of the election and its two main contenders and three treatment groups. 11The vignette in the three treatment conditions primes respondents on the contentious nature of the electoral environment, including the presence of violence and intimidation.In addition, we include a reference to an explicit threat of violence where we randomly vary the actor that issues the threat: a politician from the incumbent party (APC), a politician from the main opposition party (PDP), or a commander in the Nigerian insurgent group Boko Haram: "The elections saw repeated incidents of violence and intimidation.There have been reports of armed thugs attackingeven killingpolitical candidates, party supporters, and voters.Both voting materials, INEC offices, and civilian houses have been set ablaze.Many have accused political leaders of inciting this violence by paying and arming thugs and through hate speech.This could, for example, be a high-level [APC politician/ PDP politician/Boko Haram commander] stating in public that anyone that [does not vote for APC/does not vote for PDP/does vote] is a 'betrayer', and encouraging his followers to violently attack anyone [that votes for any party other than APC/that votes for any party other than PDP/that votes]."To enhance the credibility of the prime and for ethical reasons, the vignettes present the respondents with fictional yet realistic accounts resembling statements made by politicians, as well as by Boko Haram, during elections. 12 While our interest lies in violence by political parties, we add the insurgent treatment to test some of the assumptions underlying our argument and to address alternative explanations.If electoral violence enhances the salience of the partisan cleavage and makes it consequential for the type of regime features one prefers, as our theory suggests, citizens should not react in the same way to anti-systemic threats that do not invoke partisan identities.We return to this in the analysis.
After the treatment vignettes, we measure our outcome variable by presenting respondents with statements describing five policies that reduce constraints on the executive and thereby bestow more discretionary powers with the President.These constraints refer to both formal institutions that check the power of the executive, such as a capable legislature, as well as liberal democratic principles that de facto limit the reach of executive power.We frame our outcome items as presidential bypassing and infringements on such democratic principles, as exemplified by sidelining other branches of government (item 5), placing restrictions on opposition parties (item 2), curtailing media freedom (item 4), as well as cracking down on citizens' political rights (items 1 and 3).We then ask respondents to what extent they agree or disagree with these: 1.To ensure electoral integrity, it is sometimes necessary for the President to allow the security forces to use violence against those participating in a protest, strike, or political meeting.2. To prevent violent elections, it is sometimes necessary for the President to restrict the freedom of opposition parties to form and to compete in the elections.3. To prevent violent elections, it is sometimes necessary for the President to restrict the number of polling stations in areas where political competition is very high.4. To ensure electoral integrity, it is sometimes necessary for the President to ban journalists that accuse the President of misconduct, without revealing their sources.
5. To ensure that Nigeria can be governed efficiently, the President should be able to increase his power if the Senate and House of Representatives are very divided along partisan lines.
Rather than asking about support for democracy in the abstract, we use indirect questions to measure respondents' support for specific constraints on the executive that are important to uphold democratic governance.Individuals might be supporters of democracy as a political system at large but still willing to compromise with specific democratic principles, that is, their democratic commitments may be shallow in practice (see Svolik, 2020).Indirect questions also alleviate concerns about social desirability bias as respondents will be more likely to answer truthfully compared to direct questions.For the same reason, we also formulate our items as democratic trade-offs, referring to the need to promote electoral integrity or political order.

Measurement and Estimation Strategy
We construct our outcome variable-support for constraints on the executive-as an additive index that sums respondent responses across all five policy items.Our outcome items are measured using a 5-point scale (4 = "Strongly agree," 3 = "agree somewhat," 2 = "neither agree, nor disagree," 1 = "disagree somewhat," 0 = "strongly disagree").The resulting index can take on a minimum value of 0 and a maximum value of 20, where higher values imply more support for the expansion of discretionary powers, and lower values indicate more support for constraining these. 13Table A1 in the supplementary appendix shows descriptive statistics for our outcome variable and all other variables used in the main analysis by experimental condition.Figure A1 shows the distribution of all outcome items.
Our treatment variable-electoral violence prime-is a binary indicator coded 0 for individuals in the control group and 1 for individuals in treatment groups receiving either the incumbent or the opposition violence prime.In our main analysis, we pool across the two treatment groups to measure whether individuals hear threats of violence from any of the two political party actors. 14We do so because our argument holds that the overall violent polarization stemming from threats issued by political parties, rather than risk assessment concerning a specific actor, shapes citizens' attitudes towards expanding executive power.To explore whether the alleged actor is important, we report additional tests distinguishing between incumbent and opposition violence.Last, we examine the impact of the Boko Haram issued threat in a separate regression to test the underlying assumptions of our theoretical argument and to address alternative explanations for our findings.
The conditioning variable-partisan alignment-is measured before the experimental part by asking whether the respondent feels close to a political party. 15Individuals responding "no" are coded as non-partisans.Those that answer "yes" receive a follow-up question asking them what party they feel close to.Respondents choosing APC are coded as ruling party supporters, whereas respondents choosing PDP or any other opposition party are coded as opposition supporters.
In addition to the above mentioned variables, we include a set of demographic variables likely to influence individual political attitudes.These include age (integer), sex (binary, 0 = men), urban/rural residency (binary, 0 = urban), socio-economic status (continuous), ethnicity (categorical, 0 = other), and education (binary, 0 = low).We use these variables to assess the integrity of randomization in our treatment assignment by testing whether the experimental groups are balanced across these variables, as well as to adjust our treatment effects to get more precise estimates.In both robustness and balance tests, we also include several variables indicating whether an individual had previously experienced (electoral) violence, either personally or in its proximity.All covariates are balanced across the experimental conditions.For more information, see Section C and E in the Supplementary appendix.
We account for expectations of heterogeneous treatment effects by including interaction terms between the treatment and individual-level partisanship variables alongside the constitutive terms, using opposition partisans as the reference category.Because we are primarily interested in the conditional average treatment effects within two subgroups (ruling and opposition party supporters), we do not interpret the interaction effects directly but use them to calculate the quantity of relevance for our hypotheses.

Analysis
Table 1 provides an overview of the main results from linear regression models where we focus on the pooled treatment variable and the control group as the reference category. 16While our hypotheses focus on conditional average treatment effects based on political alignment with the ruling party, we start reporting the average treatment effect of our pooled election violence prime as a benchmark in Model 1.We find no systematic difference in support for constraints on the executive between those receiving the treatment condition and those not receiving it when averaging across partisanship.
To examine hypotheses 1 and 2 proposing heterogeneous treatment effects of the electoral violence prime across ruling and opposition party supporters, Model 2 reports conditional average treatment effects, which are also displayed in Figure 1.Contrary to our first hypothesis, the violence prime does not make ruling party supporters more likely to endorse more discretionary powers with the President.The estimated effect of the violence prime on support for executive constraints amongst incumbent supporters is not statistically significant (see Figure 1).In support of our second hypothesis, we  1), with 95% confidence intervals.
find that opposition supporters become less likely to accept violations of democratic principles when receiving a violence prime.The coefficient for the treatment condition is negative and statistically significant.Opposition supporters subject to the treatment condition score .773points lower on our outcome variable than opposition supporters in the control group (7.879 compared to 8.652).This is consistent with our argument that information about electoral insecurity for opposition partisans strengthens their preferences for constraining the executive.In line with the argument that incumbent and opposition supporters have different stakes in defending democratic institutions, incumbent supporters report higher levels of supporting the expansion of discretionary powers than their opposition counterparts across all experimental conditions.In the absence of a prime, they score .942points higher than opposition supporters (9.594 vs. 8.652, Table 1 Model 2), and in the presence of a prime, 1.107 points higher (7.879 vs. 8.986, Table A3).These differences are substantively and statistically significant.
A full set of covariates is included in Model 3. Our violence treatment is randomized at the individual level, and balance tests suggest that treatment assignment is not significantly correlated with important alternative explanatory variables (see Section E in the Supplementary appendix).The political alignment variable is not randomized, however, and although not correlated with our treatment, likely to be systematically related to demographic characteristics that predict political attitudes.Results with the full set of covariates are largely similar to the ones above.The conditional average treatment effect for opposition supporters remains negative and statistically significant, and the conditional average treatment effect for ruling party supporters remains non-significant.When including the full set of covariates, the coefficient showing the baseline support for democratic constraints for ruling supporters in the absence of threat is not precisely estimated, while ethnicity is a significant predictor of accepting democratic transgressions.Ethnicity is closely entwined with support for the ruling party and, not surprisingly, these two are hard to disentangle.In the presence of our election violence prime, however, ruling party supporters are systematically and substantially more supportive of expanding discretionary executive powers, even when controlling for ethnic differences (Table A3, appendix).This is in line with our argument, suggesting that contentious elections activate partisan alignment as a salient identity that shapes preferences for or against constraining the executive.We return to this in the concluding section.
We find that violent threats do not seem to have a significant effect on individuals who report being non-partisans-neither statistically nor substantively (see Figure 1 and more information in Table A4, supplementary appendix).Not surprisingly, non-partisans generally display less support for expanding discretionary powers than incumbent partisans.Yet, while nonpartisans do not differ from opposition partisans in the absence of a threat, they become more likely to endorse the expansion of discretionary power in the presence of a threat compared to opposition supporters.The difference in treatment effects between the groups is statistically significant at the 95% confidence level.Without the efficacy granted by an opposition affiliation, information on election violence does not, in the same way, seem to mobilize non-partisans to defend democratic institutions.We return to this in the conclusion.

Robustness Tests and Alternative Explanations
Our findings are robust to a range of model specifications.Treatment effects remain similar when analyzing sub-samples of ruling party supporters, opposition supporters, and non-partisans separately (Table A5, supplementary appendix), as well as when including PSU-level clustered standard errors, while models using PSU-level fixed effects lead to slightly less precise estimates (Table A6).To show that our results do not hinge on the construction of our dependent variable, we also conduct several robustness tests with alternative measures (see Section H.3 in the Supplementary appendix).
We also address several theoretical and empirical concerns to increase confidence in our findings.First, we control for citizens' self-reported experience with election violence using an item on whether individuals have witnessed violence at polling stations, as well as a more objective measure of their proximity to election-related violence using data from the Armed Conflict Location & Event Data Project (ACLED) (Raleigh et al., 2010; see Table A9 and Section H.4 in the appendix for more details).Nigeria represents a high-stake field setting that brings credibility to our experimental prime.At the same time, the contentious electoral environment also means that respondents are, to varying degrees, already treated with information about or experience with violent elections.Although we randomize the treatment conditions in the experimental design, the experience of threat might not be orthogonal to individual characteristics, including political affiliation.The results remain robust when controlling for respondents' experience of election violence, both using the self-reported measure and the observational data on violent events.
Second, we assess whether sub-national dynamics related to partisanship matter (Table A10, Section H.5 in the supplementary appendix).Nigeria is a federal system, and respondents could thus be in the opposition at the center while being incumbent party supporters within their respective states.The treatment effect that we identify for opposition supporters could potentially be driven by respondents that lost both the general and gubernatorial election, that is, that are in the opposition at both the central and state level.We do not find evidence for this specific pattern.Third, differences in support for executive violations of democratic principles between ruling and opposition partisans could be caused by factors other than the winner/loser gap we propose with our theoretical argument.For instance, systematic differences in how concerned respondents are about security, in general, could determine how they react to our electoral violence prime.If such security concerns differ systematically between ruling and opposition party supporters, this could produce differential responses to treatment.We empirically assess this alternative explanation in several ways but find no support.First, we show that there are no systematic differences between ruling and opposition party supporters regarding security concerns in the absence of our treatment that could produce our results (see Table A12, supplementary appendix).Second, we show that controlling for non-electionrelated violence does not alter our results (see Table A14 in the supplementary appendix).Third, we show that security threats issued by actors not affiliated with any political contender do not affect respondents' support for the expansion of discretionary executive powers, which is opposite to what we would expect if general security concerns were to drive our effects.Our Boko Haram issued threat has no effect on our outcome-neither averaged across partisanship nor conditional on it (Table A11).This also supports our argument that violence and threats issued by political parties bear specific relevance for shaping citizens' attitudes towards executive powers, as thesein contrast to more general security threats-directly relate to inter-groups relations that are salient in the electoral context.Fourth, we show that our results remain robust when including fixed effects at the level of Nigeria's six geopolitical zones, where partisan identities and insecurities may cluster in ways that could produce our results (see Table A13).
Finally, we also do not find the effect of the violence prime to be conditional on the party issuing the threat.This is in line with our theoretical argument that proposes that it is the overall violent polarization stemming from threats issued by political parties, rather than risk assessment concerning specific perpetrators, that influences political attitudes.In an exploratory analysis, we disaggregate our treatment into threats attributed to an APC politician (the ruling party) or a PDP politician (opposition party; see Section H.6, Table A11 in the supplementary appendix).Differentiating between perpetrators does not matter much for the overall interpretation of the results, neither regarding the direction of effects nor statistical uncertainty.The average treatment effects show no systematic differences between individuals in the control group and those receiving the ruling party or opposition-issued threat, respectively.Considering heterogeneous treatment effects, we find that both ruling and opposition party supporters react to the violence prime across both alleged party perpetrators by decreasing their support for executive discretionary powers-just as in our main analysis relying on the pooled threats.The uncertainty around these estimates is comparable across perpetrators, 17 and only reaches conventional levels of statistical significance for opposition supporters faced with an opposition threat when a full set of covariates is included.Although slightly less precisely estimated, this effect is similar for the incumbent-issued threat (p = .063).In sum, there is little to suggest that results differ based on the perpetrator.

Discussion and External Validity
Why do incumbent partisans not adjust their democratic attitudes when informed about election violence, while opposition supporters do?Contrary to hypothesis 2, we do not find any effect of our treatment on ruling supporters' preferences for weakening executive constraints.One explanation could be that incumbent supporters do not interpret the information about electoral contention as much as a signal of risk as their opposition counterparts.Affiliation with the regime already in power may lead incumbent party supporters to be more content and generally have more stable preferences towards the political regime, independent of the level of political contention.Opposition supporters, on the other hand, might perceive themselves as more vulnerable in the face of threat and shift their preferences more swiftly.
Another explanation is related to the case.Compared to PDP, APC made a relatively recent entry into Nigerian politics, being formed as a merger of smaller opposition parties in 2013 and controlling the Presidency since 2015 only.Due to the recent alternation in power, ruling party supporters in Nigeria thus have recent experience being excluded from it.This could act as a buffer against the temptation to counter threats by expanding the power of the President.Against this background, however, their higher baseline willingness to support the weakening of democratic constraints is particularly noteworthy.PDP, conversely, has a short history as an opposition party.On the one hand, PDP's dominance in contemporary Nigerian politics could render their constituents with a weaker sense of marginalization than opposition supporters elsewhere.At the same time, the recent political turnover might make PDP constituents particularly cognizant of the importance of constitutional limits on the executive.Research by Moehler (2009) suggests that electoral turnovers in Africa have narrowed the winner/loser gap in perceived institutional legitimacy, and S ¸as ¸maz et al. (2022) show how uncertainty in electoral turnover can shape the extent to which individuals are willing to expand executive power in Turkey.In line with such arguments and consistent with our findings, the effect of election violence on partisan gaps in preferences for democratic constraints might be stronger in countries where the authoritarian successor party has ruled uninterruptedly.
Whereas our experimental approach enhances the internal validity of our findings, the generalizability of these findings to other cases remains a more open question.Nigeria is known for having high-stakes elections-held against the backdrop of an entrenched patronage economy, a "winner takes all" electoral system and unconsolidated institutions.Yet, the country is certainly not unique in these respects, neither in the region nor beyond.Many other countries witness an even more salient incumbent-opposition divide, which could underpin divergent responses to violence amongst the electorate.A comprehensive examination of the external validity is beyond the scope of this study.Still, below, we present some evidence to suggest that the relationships we uncover are not simply a peculiarity of the Nigerian context.To describe the relationship between election violence and citizen support for the expansion of executive powers across the ruling and opposition party divide in Africa, we draw on data from the Afrobarometer survey data (rounds 4 to 7, conducted over the years 2008-2019). 18Our argument holds that elections are salient periods for shaping citizens' attitudes to their political institutions (e.g., Mattes, 2014) and to make this connection explicit, we include countries where respondents were surveyed no more than 2 years after a general election. 19We draw on the Varieties of Democracy dataset (v. 10, Coppedge et al., 2020) to identify countries/electoral rounds where voters faced a risk of violence and intimidation when exercising their democratic right to vote. 20 We code a survey respondent as supportive of the expansion of executive powers at the expense of democratic constraints if they "approve" or "strongly approve" of any of the following three non-democratic modes of governance: military rule, one-party rule or personal rule, or when asked about their democratic preference answers that "In some circumstances, a non-democratic government can be preferable" (see section I, supplementary appendix).We thus tap into forms of authoritarian governance where all, or many, democratic constraints have been dismantled.Similar to our experiment, we code the variable incumbent partisan in the affirmative if the respondent states that they feel close to the ruling party (our own coding, based on the most proximate electoral results), an opposition partisan if they feel close to any party other than the ruling party, and a non-partisan if the respondent does not report to feel close to any political party.Since our dependent variable is dichotomous, we run logit models where we regress non-democratic attitudes on election violence, political affiliation, and their interaction term.We include the same covariates as in our previous models (except ethnicity, which is hard to operationalize across countries): age, gender, poverty, urban, and level of education.
The results for the validation are reported in Table 2.As shown from Model 1, the coefficient for election violence is negative and statistically significant, suggesting that for opposition partisans (the excluded reference category), there is a lower probability of condoning non-democratic government in a context of violent elections.The component term for ruling party supporters is positive and significant, suggesting that in the context of peaceful elections, ruling party supporters are generally more supportive of non-democratic governance than their opposition counterparts.The interaction term with election violence is positive, as we postulate in our theoretical argument, but again not statistically significant.We also divide our cases into sub-samples based on whether they have seen an alternation of power within the context of multi-party elections or not. 21The direction of the effects is largely the same across these two sub-samples.But as seen in Table 2, Model 2 (no electoral turnover), and Model 3 (electoral turnover), the higher baseline support for non-democratic government in peaceful elections is more robust in the context where the regime has not lost power in multi-party elections, but these attitudes are not reinforced by election violence.The mobilizing effect of election violence on opposition supporters' democratic commitments is, on the other hand, more robustly estimated in countries that have seen regime turnover.This validation indicates that our experimental findings might be relevant beyond the Nigerian context, but we leave it to future research to probe these relationships more comprehensively.

Conclusion
Does information about violent elections make citizens more willing to accept weaker democratic constraints on the executive, or does it make them stronger proponents of the same?Whereas many voters globally cast their votes against a backdrop of violence and coercion, we know little about how electoral insecurity influences their attitudes towards specific democratic institutions.We address this gap by examining how priming respondents on their country's violent elections affect their willingness to expand the discretionary powers of the executive, using survey data from Nigeria.We predict that an election violence prime will spur different reactions amongst ruling and opposition party supporters: the former are likely to endorse more discretionary government powers to ensure order, and the latter more likely to strengthen their preferences for constraining the executive.
Although there is some evidence that respondents supporting the ruling party report higher support for dismantling democratic constraints on the executive compared to other voters, we do not find evidence that they shift in a more authoritarian direction when subjected to information about election violence.Opposition supporters, on the other hand, become more vocal condemners of non-democratic executive power grabs when being primed on a contentious electoral environment.These findings nuance existing work that suggests that citizens' fear of electoral violence is associated with increased support for non-democratic forms of governance.
Our findings thus tie in with the literature showing that violence might have mobilizing effects, particularly for those that are embedded in networks and structures that provide agency in the face of threat (Aytac & Stokes, 2019;Dorff, 2017;Young, 2019).Specifically, our findings indicate that the efficacy afforded by affiliation to the opposition might be associated with an increased commitment to democratic politics rather than a withdrawal from it.We show that many voters respond to violent electoral contention by reasserting their willingness to stand up for democratic principles.Election losers in nonconsolidated democracies might, therefore, not necessarily present barriers to democratization, but a force for it.This is encouraging in times when concerns about popular underpinnings for democratic backsliding are deep.Our finding that the division between non-partisans and opposition partisans increases when provided with information about contentious elections might still be concerning, as it indicates that election violence could aggravate gaps in political attitudes within the electorate-between those that are embedded in networks and structures that provide agency in the face of threat and those that are not.This, in itself, might be troubling for democratic consolidation.
Future research could probe these relationships further by explicitly examining the role of factors such as social embeddedness, political engagement, and efficacy in shaping citizens' reactions to election violence.Our experiment and pre-registered hypotheses focused specifically on the role of ruling and opposition party affiliation.Based on the mixed results for partisanship in terms of identifying heterogeneous responses, combined with suggestive evidence that other group affiliations (such as ethnicity) matter for shaping citizens political attitudes, a valuable agenda for future research could be to unpack heterogeneity along other dimensions that relate to the vulnerability and risks that voters face in contentious electoral environments.
Finally, although we find no evidence that violent elections lead to an increase in popular support for executive violations of democratic principles, election-related violence might precipitate de-democratization along multiple other pathways.Violence could directly stifle mechanisms of democratic accountability, targeting civil society organizations, opposition candidates, or media.While our findings may alleviate some concerns about the democratic erosion of non-consolidate democracies via citizens' support for undemocratic policies, other pathways may exist and warrant empirical investigation.

Table 1 .
Average and Conditional Average Treatment Effects (OLS).
Figure 1.Conditional average effects of electoral violence prime on support for dismantling democratic constraints, by partisanship.Based on Model 2 (Table

Table 2 .
Election Violence and Authoritarian Attitudes-Cross-National Evidence.