Honduran money
Honduran money
How should governments communicate reforms to fight corruption without reinforcing citizens’ negative perceptions? A carefully designed information intervention in Honduras provides a promising approach to update citizen beliefs about corruption in the tax agency, which translated into increased tax compliance.
Editor’s note: For a broader synthesis of themes covered in this article, check outIssue 1 of our VoxDevLit on Taxation.
Reforms aimed at fighting corruption and improving public perception of institutions face a significant challenge: citizens with entrenched beliefs about corruption often resist change, instead reinforcing their existing views. Multiple theoretical frameworks, each emphasising different psychological processes, underscore the crucial role of prior beliefs in shaping how individuals interpret and absorb new information. These priors do not simply filter information, but actively shape how individuals interpret and respond to evidence, making institutional reform particularly challenging (Kunda 1990, Zaller 1992).
Indeed, a growing body of research suggests that anticorruption messaging can sometimes have unintended consequences. Rather than reducing perceptions of corruption, such messages may reinforce citizens’ scepticism and reduce compliance with government policies (Peiffer 2020, Cheeseman and Peiffer 2022, Corbacho et al. 2016), a particularly relevant effect in contexts of political polarisation (Enríquez et al. 2024). On the other hand, while previous research has suggested that there are often ‘backfiring’ effects, more recent studies indicate that these negative outcomes are not necessarily inevitable and may be less common than previously thought (Erlich and Gans-Morse 2025).
We investigate this paradox by leveraging a major institutional reform of the Honduran tax administration (Ajzenman et al. 2025). The reform, implemented between 2014 and 2019, overhauled the country’s tax agency, including dismissing 85% of existing personnel to address corruption concerns and hiring new staff under competitive and transparent procedures. We conducted a survey experiment that was complemented by a government-led field experiment with over 30,000 taxpayers testing whether different anticorruption messages could influence perceptions about the tax authority and tax compliance. Our findings reveal that the way anticorruption efforts are communicated is crucial: while simply announcing government-led reforms to combat corruption did little to improve perceptions and the willingness to pay taxes, a carefully designed sequenced approach—first addressing citizens’ negative priors, and then delivering information about reform efforts to address corruption within the tax agency—significantly improved trust in tax officials and boosted tax compliance.
How citizens interpret anticorruption messages
People process information with context in mind. Psychological research shows that individuals interpret new data through the lens of their preexisting beliefs (Kunda 1990, Zaller 1992). When it comes to corruption, citizens who already believe that government institutions are corrupt are likely to dismiss—or even react negatively to—messages that claim progress in fighting corruption. This phenomenon, known as the ‘backfire effect’, has been documented in various contexts, including anticorruption campaigns in Latin America and Africa (Peiffer 2020, Cheeseman and Peiffer 2022). Recent research suggest that credibility is key: when messages are perceived as trustworthy, they are more likely to influence beliefs (Erlich and Gans-Morse 2025), while the failure to update beliefs can significantly impact critical outcomes, such as assessments of government performance (Enríquez et al. 2024).
Honduras provides an ideal setting to evaluate these dynamics. Before the tax administration reform was completed, public trust in tax authorities was extremely low: in 2016, over 40% of Hondurans believed that ‘all or almost all’ tax officials engaged in corruption (Latinobarómetro 2016). Given this entrenched scepticism, how should governments communicate their efforts to improve governance without reinforcing citizens’ negative perceptions?
Evaluating the effectiveness of anticorruption messages: A survey experiment
To answer this question, we conducted an online survey experiment with 1,411 Hondurans. Respondents were randomly assigned to one of four groups:
Control group: Received neutral information about the national tax authority (SAR), describing SAR's role in tax collection, along with a map with the locations of SAR offices across Honduras.
Reform treatment: Informed about government efforts to combat corruption within the tax administration by highlighting the extent of personnel renewal for the purpose of reducing corruption within the tax agency.
Debiasing treatment: Presented independent public opinion data emphasising that the average Honduran perceives corruption in tax administration to be relatively low compared to citizens in the rest of Latin America.
Sequential treatment: A combination of the debiasing message followed by the reform message.
Our main outcome captures respondents' beliefs about corruption in the tax agency. Respondents rated their beliefs on a scale from 0 (not likely at all) to 10 (very likely) measuring the likelihood of SAR officials engaging in corruption and seeking gifts or money for resolving or preventing tax-related issues. We also measured tax compliance intentions using the crosswise technique, which allows respondents to answer sensitive questions without explicitly admitting to noncompliant behaviour, ensuring a more accurate estimate of the prevalence of tax evasion.
Panel (a) of Figure 1 shows that the reform message alone had no significant impact on corruption beliefs and, if anything, slightly reinforced negative beliefs. In contrast, the debiasing treatment led to a significant improvement in perceptions, reducing perceived corruption levels by 7%, while the sequential treatment—first addressing scepticism and then presenting the reforms—produced the largest improvement, reducing perceived corruption levels by 13%.
Tax compliance attitudes mirrored these findings. Panel (b) of Figure 1 shows that the reform message alone had no effect, and while the debiasing treatment did not show an improvement in compliance intentions, the sequential treatment significantly reduced respondents’ stated willingness to evade taxes by 11 percentage points—a 21% reduction relative to the control group. Additionally, the effect of the debiasing treatment was particularly pronounced among individuals who initially held more pessimistic views of corruption. Among respondents who entered the survey rating government officials as highly corrupt (a score of 8 or higher on a 10-point scale), the sequential treatment had an even greater impact, reducing stated willingness to evade taxes by 14 percentage points.
Figure 1: Effects of treatments on outcomes
Panel (a): Corruption beliefs
Panel (a): Corruption beliefs
Panel (b): Willingness to evade taxes
Panel (b): Willingness to evade taxes
Panel (b): Willingness to evade taxes
Notes: Panel (a) shows the effects of the reform, debiasing, and sequential arms on corruption beliefs, while Panel (b) displays their effects on stated willingness to evade taxes, where we multiplied the outcome by 100 in order to get a direct interpretation of the coefficient. Each coefficient is presented with 90% and 95% confidence intervals, using robust standard errors.
From beliefs to behaviour: A field experiment on anticorruption messaging
While survey experiments provide valuable insights, they do not always translate into real-world behaviour. We thus collaborated with the Honduran tax authority on a large-scale field experiment to test whether these messages could influence actual tax compliance.
In the weeks before the 2023 income tax filing deadline, the tax authority sent emails to over 30,000 taxpayers, randomly assigning them to receive either the debiasing or sequential message. The control group received only a basic tax reminder. Based on our survey findings, the government deliberately omitted the reform-only treatment as it may have backfired.
The results aligned with our survey experiment. Taxpayers who received the sequential message declared 6.8% more taxable income than the control group. Among high-risk taxpayers—those more likely to evade—reported income increased by 11.5%. By contrast, the debiasing message alone had a positive but more minor and statistically insignificant effect. These results confirm our hypothesis that our sequential treatment was the most effective for reducing tax evasion.
Policy implications for anticorruption messaging interventions
Our study demonstrates that beliefs about corruption are both critical and malleable. Effective communication in anticorruption campaigns requires a two-step approach: first, altering perceptions of corruption, and second, emphasising government actions. By carefully sequencing these messages, governments can foster more positive views of institutional competence and reduce tendencies toward tax evasion or corrupt behaviour.
These findings carry significant implications for policymakers. In low-trust contexts, exaggerated beliefs about corruption can lead anticorruption messages to inadvertently reinforce negative perceptions, while scepticism about government intentions may result in outright rejection of such efforts. These challenges are critical in many developing countries caught in a vicious cycle of weak institutions and widespread tax evasion. Addressing distorted beliefs through theoretically grounded interventions can help rebuild trust in key public institutions such as tax authorities and foster voluntary compliance, thus enhancing revenue collection.
References
Ajzenman, N., Ardanaz, M., Cruces, G., Feierherd, G. and Lunghi, I. (2025). "Unraveling the paradox of anticorruption messaging: Experimental evidence from a tax administration reform." Unpublished manuscript.
Cheeseman, N. and Peiffer, C. (2022). "The curse of good intentions: Why anticorruption messaging can encourage bribery." American Political Science Review, 116(3): 1081–1095.
Corbacho, A., Gingerich, D. W., Oliveros, V. and Ruiz-Vega, M. (2016). "Corruption as a self-fulfilling prophecy: Evidence from a survey experiment in Costa Rica." American Journal of Political Science, 60(4): 1077–1092.
Corporación Latinobarómetro. (2016). "Latinobarómetro 2016 Data Set."
Enríquez, J. R., Larreguy, H., Marshall, J. and Simpser, A. (2024). "Accountability under polarization." Forthcoming.
Erlich, A. and Gans-Morse, J. (2025). "Can norm-based information campaigns reduce corruption?" American Journal of Political Science, (forthcoming): 1–20.
Kunda, Z. (1990). "The case for motivated reasoning." Psychological Bulletin, 108(3): 480.
Peiffer, C. (2020). "Message received? Experimental findings on how messages about corruption shape perceptions." British Journal of Political Science, 50(3): 1207–1215.
Zaller, J. (1992). "The nature and origins of mass opinion." Cambridge University Press.
Honduras tax reform anti-corruption taxation tax tax compliance Information