Working papers:
Government Audits of Municipal Corruption and Belief Updating: Experimental from a Field Experiment(with Raymond Duch). Under Review
Abstract
We implemented a field experiment that assesses whether citizens update corruption beliefs when presented with audit information about malfeasance in their municipal government. The video treatment reports audit results for Chilean municipalities; the control is a placebo video. We measure incentivized pre-treatment and post-treatment corruption beliefs of 5,528 subjects. When informed about corruption in their municipality subjects update negatively. Updating is correlated with reported malfeasance and with trends in malfeasance. We find limited evidence of Bayesian learning. Treatment effects persist after one-month. We observe higher donations to local public goods by treated subjects in municipalities with more positive audits.
Choice Architecture for Messaging Corruption: Evidence from an Adaptive Experiment (With Raymond Duch).
Abstract
The presentation and framing of information are the foundations of many behavioural experiments. In the case of corruption, policymakers such as NGOs face the challenge of informing citizens about the levels of malfeasance found in their local constituency. This challenge implies identifying an optimal messaging strategy that is sufficiently compelling to attract the interest of citizens. We addressed this challenge by evaluating six strategies for information messages often used in corruption information experiments. Using historical data from local government audit reports in Chile and in partnership with the NGO Chile Transparente, we implemented an online adaptive experiment using a modified Thompson Sampling algorithm (Exploration sampling) in which the assignment probabilities of the information treatments were updated in 11 batches of 100 subjects each. The results showed no unique optimal information strategy for malfeasance messages. However, a loss-frame information strategy tends to be slightly more persuasive than other ways of conveying information about corruption. We also found evidence that more rudimentary information metrics of corruption can be equally persuasive compared to more sophisticated ones. Finally, we do not found significant differences between using spatial comparison (i.e. comparing corruption in a local government across local constituencies within the same region) versus a temporal comparison (i.e. comparing the same local government across time).
Do Government Audits Reduce Bureaucrats’ Rent Extraction? (with Luis Soto Tamayo).
Abstract
Governments have widely implemented audits to curb corruption, increase accountability, and discourage patronage practices within the public sector. This paper focuses on the impact of audits on civil servants by looking at two main aspects: first, whether audits effectively discourage 'rent-extraction' behaviour among civil servants, and second, whether the audits have a differentiated effect on civil servants with family affiliations. We employ a unique dataset from Chile's 2016 financial and performance audits of civil servants and leverage the random audit assignment across 17 governmental institutions over 8,640 public servants. Our preliminary analysis suggests that audits discourage public officials from engaging in rent extraction, but this impact is small and insignificant. We find some suggestive evidence that bureaucrats offset this drop in rent extraction by either extracting rent from other sources of government activities or by increasing the amount involved in current ones.
Measuring Corruption using Bayesian Randomise Item Response Model (with Ahra Wu).
Abstract
The political and economic consequences of corruption are substantial. Corruption can lead to a reduction in the provision of public goods and undermine trust in democratic institutions. Scholars have been trying to reliably measure the prevalence of corruption behaviors in the population using surveys using direct question format type-of-questions. However, these measures are likely biased due to social desirability and non-response biases. Indirect questioning survey techniques have been designed to minimize these biases and elicit truthful answers to sensitive topics and behavior. However, the canonical design of these techniques only allow the measurement of group-level estimates. This paper provides empirical evidence of an extension of the Randomized Response Technique called the Randomized Item Count Response Technique (RIRT) to estimate both group- and individual corrupt behaviors (n = 6058 and n = 3692). At the group level, we found prevalence rates from 60\% in the case of patronage to 1\% in administrative corruption. For individual-level estimates, we found that the distribution of respondents' underlying traits for engaging in corrupt behavior follows a power-law distribution. We implemented several approaches to identify inattentive study participants, showing that our results are robust once we exclude these respondents.
Exploring Global Perspectives from the Development Engagement Lab’s Database (With Jennifer Hudson, David Hudson, Soomin Oh, Paolo Morini) Under Review
Abstract
Public support is crucial for shaping effective development aid policies. The Development Engagement Lab (DEL) has conducted comprehensive surveys in France, Germany, Great Britain, and the U.S. between 2019 and 2024 to track and analyse the evolving profile of public attitudes, perceptions and behaviours around global poverty and inequalities, development aid, and the challenges facing the world. This paper presents multiple datasets curated by DEL in all four countries, including longitudinal panels (2019-2023), additional repeated cross-sectional tracking surveys (2019-2024), and a collection of subject-specific data surveys (2019-2023). To facilitate the access and dissemination of the DEL data, we present a new R package, \textit{DELdata}, which allows easy access to all datasets. This comprehensive release will enable researchers and students of public engagement and attitudes towards global development, poverty, and aid to generate new research advancing our knowledge of public opinion and political engagement in this field.
Working in progress:
A Meta-Analysis of the Effects of Patronage on Public Institutional Performance (Single-authored).
Abstract
Patronage remains a widespread phenomenon that shapes the functioning of political systems and influences governance outcomes. Scholars across disciplines have explored the impact of patronage on various organisational, institutional, and individual outcomes, highlighting how patronage affects performance, efficiency, and public trust. Numerous studies provide empirical evidence on the strategic allocation of patronage benefits by officials and brokers based on the characteristics of beneficiaries, such as loyalty and social connections, and the nature of benefits provided. While there is some agreement on the general direction of patronage’s impact on these outcomes, there remains significant variation in the size and scope of these effects across contexts. To the best of my knowledge, no study has yet undertaken a systematic assessment of the pooled impact of patronage. This project aims to systematically identify and analyse published quantitative studies on the effects of patronage over the past decade. Through meta-analysis, I estimate the inverse variance-weighted average effect of patronage on organisational performance, institutional efficacy, and public perceptions of institutional integrity. Furthermore, using meta-regression, I investigate the heterogeneity of these effects by examining how outcomes differ based on beneficiary characteristics (e.g., loyalty, socioeconomic status, historical connections) and types of benefits (e.g., material resources, career positions, or preferential treatment). This analysis helps clarify the extent and variability of patronage’s impact on governance and identify the factors that drive these effects.
Deterring Public Procurement Favouritism: Evidence from a Natural Experiment (Single-authored).
Abstract
To be added
The Spillover Effects of Audits (Single-authored).
Abstract
Governments across the globe have set audits as one of the primary mechanisms to measure and deter corruption. This paper focuses on the spillover effects of audits within local governments in Chile. Using a fuzzy regression discontinuity design, I exploit the discontinuity in allocating audits within local governments/departments to estimate the effects of audits within departments within each municipality. The fuzzy RD results will be completed using panel data from audits conducted between 2014-2024. I expect to present the preliminary results of both estimation strategies pre-registered in the analysis plan at the conference.
Explaining Public Support for Foreign Aid Projects: Evidence from France, Germany, Great Britain and the United States (With Jennifer Hudson, David Hudson, Soomin Oh, Paolo Morini).
Abstract
Public support is understood to be a condition for foreign aid spending in donor countries. But do donor publics have preferences on which aid projects foreign aid is spent? Drawing on a new dataset examining four donor countries – France, Germany, Great Britain and the United States (n= 26,169) – we leverage a unique conjoint experiment to better understand which attributes of aid projects generate more or less support with donor publics. Our results show that respondents are more likely to support projects targeting water, sanitation, and hygiene (WASH), food security, and heath programmes. Projects focused on gender equality, environment, infrastructure and social protection are negatively associated with support. In line with our expectations, we find effects for need and aid effectiveness: higher levels of extreme poverty and effectiveness of aid are associated with support for the development project, whereas low levels of need and low effectiveness are negatively associated with support. Second, we find few effects for region, with less support for projects targeted to the Middle East and Northern Africa. Finally, our main findings are robust when we remove need and effectiveness as attributes, demonstrating while these are important drivers of support, they do not crowd out other attributes or dimensions. Our findings show that in addition to preferences on overall levels of foreign aid spending, donor publics have a consistent set of preferences on how aid is spent.
The Power of Activation: Assessing the Effect of Imagery in Global Poverty Campaigns: Evidence from a Visual Conjoint. (With Jennifer Hudson, David Hudson, Soomin Oh, Paolo Morini). 
Abstract
This project seeks to understand how the images development organisations use in campaigns and communications affect public engagement, and specifically, donations and information seeking behaviours. In our design we study the way individuals respond to these images by studying their emotional response, and senses of efficacy and urgency. We then study the effect of these responses on their choices to donate or seek more information about the people in the image.
Deeds and Words: Linking Attitudes and Actions on Global Poverty and Development (With Jennifer Hudson, David Hudson, Soomin Oh, Paolo Morini).
Abstract
Political scientists have often looked at factors affecting attitudinal and behavioural indicators to study public engagement with political issues, but studies which consider the effects of each on the other are less common, due, amongst other factors, to data availability constraints. In this paper, we leverage on a novel four-year longitudinal panel study of public engagement with global poverty and sustainable development in France, Germany, Great Britain, and the United States (2019-2022) to draw connections between attitudinal and behavioural engagement in this domain. Specifically, we look at ten behavioural engagement indicators (from news consumption to volunteering, to participating in demonstrations), and attitudinal indicators capturing the public’s moral views on global poverty and development, their concern for the issue, and their world view on inequalities and global challenges. We exploit variation in time and across individuals in a cross-lagged panel model to identify causal effects of each on the other. We show that changes in a respondent’s moral views at any one time predicts changes in behavioural engagement in the following time in a series of behavioural indicators, but also that people taking actions also change their views on global poverty. We draw implications for charities, NGOs and governments engaged in efforts to involve publics in donor countries with their work to tackle poverty worldwide.
Adaptive Experimentation to Explore Treatment Effect Heterogeneity Efficiently (Single-authored).
Abstract
To be added soon :)
</details>