Re-election in High-Crime Settings

The ability to run for re-election fundamentally changes the incentives of politicians, voters, and interested groups. Politicians rely on voter support for electoral success, prompting credit claiming, personalistic campaigns, and local infrastructure investments (for example, Austen-Smith and Banks, Reference Austen-Smith, Banks and Ordeshook1989; Ch, Reference Ch2022; Ferejohn, Reference Ferejohn1986; Ferraz and Finan, Reference Ferraz and Finan2011; Motolinia, Reference Motolinia2021). Re-election may reduce corruption, improve governance, and increase government accountability, especially when oversight and transparency are high. While incumbents may also use the resources of office for electoral success (for example, Baron, Reference Baron and Ordeshook1989; Fouirnaies and Hall, Reference Fouirnaies and Hall2014; Ofosu, Reference Ofosu2019; Rueda and Ruiz, Reference Rueda and Ruiz2020), re-election is often associated with improved governance and better accountability.

However, in high-crime contexts, re-election may generate pernicious consequences (Fouirnaies and Hall, Reference Fouirnaies and Hall2018; Trejo and Ley, Reference Trejo and Ley2020; Snyder and Ting, Reference Snyder and Ting2008; Snyder and Durán-Martínez, Reference Snyder and Durán-Martínez2009). Criminal groups can leverage violence, bribery, and intimidation to shape political competition (Magaloni et al., Reference Magaloni, Franco-Vivanco and Melo2020; Ley, Reference Ley2017; Matanock and Staniland, Reference Matanock and Staniland2018; Moncada, Reference Moncada2021). Re-election may drive politicians to rely on armed groups that use illicit means to secure power, particularly in resource-poor or institutionally weak settings (Barnes, Reference Barnes2017; Castillo-Quintana, Reference Castillo-Quintana2024; Dal Bó et al., Reference Dal Bó, Dal Bó and Di Tella2006; Lessing, Reference Lessing2017).Footnote ² By influencing who runs for and wins elections, criminal groups undermine voters’ capacity to hold officials accountable and reduce the benefits of re-election (Durán-Martínez, Reference Durán-Martínez2015; Magaloni et al., Reference Magaloni, Gosztonyi and Thompsonn.d.; Ponce et al., Reference Ponce, López Velarde and Santamaría2022).

Criminal groups shape elections using tactics like violence, threats, bribery, voter intimidation, and fraud (for example, ICG, 2023; Trejo and Ley, Reference Trejo and Ley2020; Durán-Martínez, Reference Durán-Martínez2015). In regions where criminal organizations are powerful, these groups can play a significant role in the initial selection of candidates, often steering the process toward those more likely to co-operate.Footnote ³ When unable to select an official, such as in places with strong party dynamics or federal oversight, criminal groups often find other strategies to build or force a relationship with whoever gets elected. Like voters, criminal groups watch how mayors behave during their first term to decide whether they can work with them. A mayor who avoids confrontation or tolerates criminal activity may be seen as a useful partner, while one who increases patrols or targets local interests may face threats or be pushed out in the next election. This period allows criminal groups to test strategies and find the best ways to pressure local officials.

Re-election amplifies this dynamic by giving both parties longer and more stable time horizons. Mayors are better able to deliver lasting protection, policy favors, and access to state resources, and criminal groups can exert more pressure on an incumbent. For example, officials who initially resist organized crime may later collaborate under pressure or in exchange for campaign funding or protection. Over time, these relationships may evolve from ad hoc bargains into stable arrangements of mutual benefit, where criminal groups secure access to state resources and incumbents receive financial and electoral support. Without re-election, criminal groups may seek to capture or collude with more stable roles, such as party brokers and police chiefs (Morris, Reference Morris2009; Hernández, Reference Hernández2013; Trejo and Ley, Reference Trejo and Ley2020). Thus, re-election shifts the locus of power from parties to individual politicians, making elections a pivotal opportunity for criminal groups to ensure their preferred candidates gain and retain office.

Electoral violence remains one of the most visible and consequential tools used by criminal groups to shape political outcomes, often deployed to ‘cleanse the candidate pool’ by targeting rival candidates with intimidation or assassination (Blume, Reference Blume2017; Schedler, Reference Schedler2014). While criminal groups may use violence against incumbents to coerce policy change, this is less common in areas where these groups are already entrenched. In such settings, groups often influence candidate selection or establish cooperative relationships with incumbents early in their first term. As violence is systematically used to deter entry, punish opposition, and suppress competition, the theory remains agnostic about effects on incumbents and focuses empirically on the observable targeting of challengers. Following this logic, I hypothesize:

Violence against candidates is a costly but effective signal of a criminal group’s strength, commitment, and threat to other politicians. Attacking one challenger signals imminent danger in running against the incumbent, therefore dampening future rivals and repressing current ones. After re-election, incumbents in areas where criminal groups control local illicit markets are more likely to enact policies that benefit criminal groups and help cement their own political power, such as selectively deploying police, allocating government contracts to criminally controlled businesses, embezzling, or sharing intelligence. With stable state power, criminal groups and the incumbents can make more credible future commitments, and use the power of the state to achieve their own goals. Thus, re-election, despite the positive incentives of voter accountability, can also incentivize capture and collusion with powerful actors in high-crime settings. Therefore, I hypothesize:

In areas where criminal groups are not locally powerful, elected officials operate in a context where voters remain the dominant force shaping electoral outcomes. These settings serve as a baseline in which standard models of democratic accountability are more likely to apply, and where the risks of criminal capture are lower. In contrast, in municipalities with entrenched criminal groups, re-election may have unintended consequences. The prospect of continued power can incentivize collusive relationships between mayors and illicit actors, further embedding criminal influence in local governance.

Staggered Introduction of Mayoral Re-election

For more than eighty years, Mexican politicians were ineligible for consecutive re-election, allowing the party-driven system of control to flourish under the Partido Revolucionario Institucional (PRI). However, Pres. Enrique Peña Nieto (PRI) announced the Mexican Pact Accord in 2014, facing legislative gridlock, pressure from rival parties, and major reform negotiations. This reform permitted state legislators to run for re-election and, critically, gave them power to determine if mayors could also run.Footnote ⁴ To maximize their influence, governors used their agenda-setting power to push for votes on re-election reform near the end of their terms; as gubernatorial cycles were staggered, most states (twenty-five of thirty-one) approved the reform in 2015 and 2016 (Motolinia, Reference Motolinia2023).Footnote ⁵ By the 2018 election, six states had not yet approved it, although most did in the following months and years.Footnote ⁶ These mayors, as autonomous local authorities in charge of municipal budgets, intelligence, and policing, are the primary targets for criminal capture and collusion (Ch, Reference Ch2022; Blume, Reference Blume2017; Trejo and Ley, Reference Trejo and Ley2019).

I utilize the staggered approval of mayoral re-election as the treatment for a staggered difference-in-differences design. While voters could not re-elect a mayor in a treated state until the 2018 election, this reform was closely followed by mayors and other local politicians, whose behaviors and incentives changed when their political horizons did (Motolinia, Reference Motolinia2023). Mayors were unable to select into a treatment condition, nor is there evidence that reform adoption was a result of concerns about or influence from criminal collusion or influence.Footnote ⁷ Because states held elections in different years on staggered three-year cycles, I group municipalities into three-year electoral bins, ensuring each held one election per period and avoiding biased comparisons between electoral and non-electoral years.Footnote ⁸

Finally, the theory anticipates that re-election will have different effects depending on criminal groups’ entrenchment in local illicit markets. While precise measurement is difficult, I proxy criminal entrenchment by observable poppy fields (Farfán-Mendez, Reference Farfán-Mendez2021; Felbab-Brown, Reference Felbab-Brown2020), a measure to reduce false positives, drawing from federal press releases, NGO reports, and investigative journalism, as an intentionally conservative measure.Footnote ⁹ This is only a portion of the national stage (202 municipalities), and represents where criminal groups are more likely to be well-established and entrenched compared to places where groups are competing over new markets. This indicator is validated with pre-treatment NGO and government reports and is uncorrelated with reform approval.Footnote ¹⁰

Empirical Approach and Results

I examine the observable effects of re-election on criminal–politician dynamics during the campaign period and in the post re-election period. I show that in places where criminal groups are already entrenched, re-election increases fatal attacks against candidates who challenge the incumbent. Then, I show that in municipalities where criminal groups are entrenched and mayors could run for re-election, the federal government was significantly more likely to audit municipalities for corruption and more likely to find evidence of significant misappropriation. I estimate this model using a stacked difference-in-differences estimator, which creates separate ’stacks’ for each treatment cohort and aligns comparisons to untreated units in a pre-treatment window (Baker et al., Reference Baker, Larcker and Wang2022; Cengiz et al., Reference Cengiz, Dube, Lindner and Zipperer2019; Butts and Gardner, Reference Butts and Gardner2022).Footnote ¹¹ The estimating equation is:

$$Y_{i,e,c}=\beta {\it Re {\text{-}} election}_{i,e,c} + \theta _{c} X_{i,e} + \alpha _{s,c} +\lambda _{e,c} + \epsilon _{i,e,c}$$

where the outcome Y _{i, e, c} is the outcome for municipality i in bin e, within treatment cohort c (stack group). The indicator Re-election _{i, e, c} is a binary indicator that turns on in the post-treatment bin for municipalities in cohort c. $X_{i,e}$ are pre-treatment controls such as population size, partisan misalignment, access to major highways, international ports, oil pipelines, and railways. The model includes state α _{s, c} and electoral bin fixed effects γ _{e, c} , both defined within each stack group. All coefficients are estimated by cohort and averaged to produce an average treatment effect on the treated by cohort, following Cengiz et al. (Reference Cengiz, Dube, Lindner and Zipperer2019); Callaway and Sant’Anna (Reference Callaway and Sant’Anna2021); Goodman-Bacon (Reference Goodman-Bacon2021). Standard errors are clustered at the state level using 1,000 wild cluster bootstrap iterations. The appendices present complete regression tables and alternative estimations, including analyses of criminal group behavior and governance quality, to assess the robustness of the findings and alignment between theoretical expectations and measurement strategy.

Increased Violence Against Rival Candidates

While many illicit campaign strategies are intentionally hidden, violence against politicians is an intentionally visible signal of strength, meant to shape who chooses to run for office (Blume, Reference Blume2017; Schedler, Reference Schedler2014; Trejo and Ley, Reference Trejo and Ley2019). I collected an original dataset on fatal attacks against local politicians, using newspaper reports from 1 January 2012 through 30 June 2022. This dataset captures fatal attacks on incumbents, mayoral candidates, and former mayors, including systematic details about the attack, including the location, timing, and context.Footnote ¹² This dataset is unique in in its detail and temporal scope. Other useful, yet spatial or temporally limited datasets, such as Votar Entre Balas (CIDE), Alcaldes de México, and ACLED were used to validate these data. In total, there are 378 recorded fatal attacks against local politicians.Footnote ¹³

Table 1 shows that re-election eligibility significantly increases political violence in areas with entrenched criminal groups. Using a conservative definition of entrenchment, the analysis compares post-treatment outcomes in municipalities where mayors could run for re-election to similar municipalities where they could not (n = 202). For contrast, the table also reports results for 2,031 municipalities without entrenched criminal presence. In both contexts, Column 1 shows a positive effect on violence against any candidate, but the effect size in entrenched areas is over twenty-seven times larger, indicating a substantial rise in fatal attacks against politicians. A formal test confirms that this difference in effect size between entrenched and non-entrenched areas is statistically significant, particularly for challengers (see Appendix G).

Table 1. Effect of staggered treatment on attacks, by criminal entrenchment and candidate type

Note: clustered standard errors at the state level in parentheses. ⁺ p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001.

The reform disproportionately increased violence against challengers in high-crime areas, not incumbents.Footnote ¹⁴ Column 2 of Table 1 shows no significant change in violence against incumbents in high-crime areas, likely because re-election may both provoke attacks on uncooperative mayors and reduce them for cooperative ones, offsetting effects. In contrast, Column 3 reveals a significant rise in violence against challengers in areas with entrenched criminal groups. In these municipalities, the average number of attacks on challengers rose more than tenfold from a baseline of just 0.03 per bin following the reform.Footnote ¹⁵

Appendix F presents the results of an event study, indicating no pre-treatment trends that violate the parallel trends assumption. Appendix G outlines alternative specifications and estimators, including the estimator from Callaway and Sant’Anna (Reference Callaway and Sant’Anna2021) and a triple difference-in-differences approach, suggesting the same results as shown here. In line with these results, when mayors can run for re-election and criminal groups are entrenched, mayoral challengers are more likely to face violence compared to low-crime areas that experience the same reform.

Local Collusion Following Mayoral Re-election in High-Crime Areas

If re-election facilitates and strengthens existing relationships between elected officials and criminal groups in high-crime areas, we should expect increases in indications of collusion where both re-election and criminal presence coexist. While direct measurement of collusion is challenging, we can observe how federal institutions allocate anti-corruption investigations in response to concerns about collusion and corruption, as well as the findings of their investigations. Appendices I and J present the effects of the reform on additional indicators, such as the quality of governance, reported perceptions of municipal and police corruption, and civilian reports of the presence and organization of criminal groups at the municipal level, which show that perceptions of corruption by municipal police increase and criminal groups are less likely to fragment following mayoral re-election in high-crime areas.

Capturing federal concerns about corruption, I use data from La Auditoría Superior de la Federación (ASF), a federal office which conducts annual audits of municipal governments in Mexico. Each year, the ASF selects a subset of municipalities to audit based on a predetermined risk assessment formula. This selection process considers factors such as the volume of federal transfers, the share of municipal budgets reliant on federal funds, prior audit history, and previously identified irregularities, often in high-risk or politically sensitive contexts (Larreguy et al., Reference Larreguy, Marshall and Snyder2014; García-Ponce and Rios, (Reference García-Ponce and V RiosND)). Places with rail lines, higher populations, and poppy fields are all more likely to be selected for audits, while there is little evidence to suggest partisan targeting of opposition parties.Footnote ¹⁶ Approximately one-third of all municipalities have been audited at least once since 2000. Studies have regularly examined the ASF’s consistent history of political neutrality in both the selection and findings of their audits, and it is well-regarded as an independent auditing body (Chong et al., Reference Chong, De La AL, Karlan and Wantchekon2015; Guajardo and Schwindt-Bayer, Reference Guajardo and Schwindt-Bayer2023; Larreguy et al., Reference Larreguy, Marshall and Snyder2020).

Audit selection serves as an indirect signal of federal concern about corruption risk in high-crime areas. Rather than treating audit selection as a direct measure of corruption or collusion, I interpret it as an institutional response to suspected vulnerabilities, such as local capture, fund misuse, or weak state capacity. If re-election entrenches corruption where criminal groups are active, we should expect greater oversight and more frequent findings of misconduct.Footnote ¹⁷

Using the same estimating equation from the previous section, I estimate the effect of re-election eligibility on the likelihood that a municipality is selected for audit, contingent on the presence of criminal organizations, presented in Table 2. In places where criminal groups may be entrenched in local economies, re-election eligibility dramatically increases the likelihood of audit, representing an eleven-fold increase (79 per cent) relative to the baseline. In contrast, where groups are not entrenched, re-election eligibility has no significant effect on audit selection. This pattern suggests that the federal government perceives elevated risk in municipalities where elected officials can consolidate power and collaborate with organized crime.

Table 2. Average treatment effect on audit selection

Note: clustered standard errors at the state level in parentheses. Significance levels: + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001.

Not all municipalities are audited each year, and those selected repeatedly tend to differ from the average in terms of budget size, governance quality, and corruption risk. To gain insight into audit outcomes, I compare the outcomes of audited municipalities with and without entrenched criminal presence following the reform. Table 3 estimates the effect of re-election eligibility on the share of Fondo de Aportaciones para la Infraestructura Social Municipal (Funds for Contributions to Municial Social Infrastructure, or FISM) that are correctly allocated. Lower values indicate greater misuse, such as missing funds or improper spending. Due to the limited sample size and the fact that audit outcomes occur post-treatment, these findings should be interpreted cautiously and not as definitive causal effects.

Table 3. Treatment effect on audit outcomes: portion of FISM budget correctly allocated

Note: clustered standard errors at the state level in parentheses. Significance levels: + p < 0.1, * p < 0.05, ** p < 0.01, *** p < 0.001.

The results show significantly lower rates of correct fund allocation in areas where criminal groups are entrenched and mayors could run for re-election. In these cases, correctly apportioned FISM funds dropped by more than 75 per cent, a striking result given that these funds make up most of the municipal budget. This suggests that such municipalities were not only more likely to be audited but also more likely to exhibit detectable corruption in high-crime areas. Appendix I shows service quality is stable, but perceptions of police corruption – who are often at the forefront of direct collusion with criminal organizations (for example, Flores-Macías and Zarkin, Reference Flores-Macías and Zarkin2021; Ravelo, Reference Ravelo2009; Hernández, Reference Hernández2013) – rise. Further, criminal organizations in non-entrenched areas are more likely to splinter and fragment following the reform, suggesting some stability and protection in criminally entrenched areas where mayors are eligible to run for re-election (see Appendix J).

These findings suggest that in areas where criminal groups are powerful, re-election fails to promote accountability or better governance and instead enables collusion. Criminal groups can capitalize on the reform by backing cooperative incumbents and keeping them in office across terms, gaining state protection and pushing out challengers to strengthen control over illicit markets, using that control to build out trafficking corridors and regional drug economies (Felbab-Brown, Reference Felbab-Brown2020). With access to substantial funds and military-grade weapons, these groups are well-positioned to influence who enters office, unlike newer or competing organizations. Appendix K shows suggestive evidence that homicides increased in criminally entrenched areas after the reform, while other crimes like kidnapping, extortion, and business robbery remained unchanged.Footnote ¹⁸