Dissertation Project

This project analyzes the effect of weak party systems on the nature of clientelism. I argue that when political parties lack the capacity to select and monitor party brokers, national legislators will use mayors as brokers in order to capture local voter networks. I use a formal model in order to predict when mayors will send the signal that they use clientelism for secure voter networks and when they will be reliable brokers for national legislators. I test this model in the context of Colombia. I use a Bayesian mixed-membership model in order to estimate the prevalence of clientelism at the municipal level. I then conduct a large-N analysis of municipalities on how many club goods projects are contracted and the discretionary transfers received. I also run two survey experiments- one on citizens and one on local elites- in order to test when mayors will deliver votes by attributing credit to national legislators for club goods received. Finally, I use survey data from interviews conducted with mayors, legislators, and local bureaucrats from July-December 2018.


Measuring Local-Level Clientelism: A Bayesian Mixed-Membership Approach

Scholars of distributive politics often discuss the importance of the use of clientelism. However, since clientelist exchanges occur “under the table”, we lack strong observational data to estimate the actual prevalence of clientelism. The current solution is to use survey experiments designed to elicit sensitive information in order to create population-level estimates.This measurement strategy, however, does not permit us to estimate within-country variation in the use of clientelism. In order to address this challenge, I propose estimating local-level clientelism using a Bayesian Mixed-Membership model. Using data from Colombia on the employees who are employed as temporary teachers, I model whether the teachers are hired based on their meritocratic qualifications or to fulfill political aims,  Specifically, I consider the type of vacancy the teacher fills, the level of education the teacher has, whether their education aligns with what they are teaching, where the school is located within the municipality, how the teacher is paid, and any bonuses received. I use these municipal hiring patterns to estimate the presence of clientelism at the municipal level. These estimates align with survey questions from the same time period. 

For a Working Paper, click here

Club Goods Provision: The Effect of Weak Parties on Clientelism Strategies

How do politicians strategically allocate funds for political gain in the context of weak political parties? In much of the literature on distributive politics, political parties play an instrumental role in determining the allocation of funds. Particularly in clientelist systems, political parties are seen as instrumental for finding reliable brokers. However, in many democracies, political parties lack both the discipline to insure a cohesive group of elected party members and the internal capacity to build and maintain clientelist networks. I argue that when parties are unable to oversee clientelist machines, national politicians will use relationships with local politicians to determine where to allocate discretionary funds in the form of club goods. I argue that regardless of the political party in office, national politicians are more likely to target municipalities where mayors have personal clientelist networks in place.  I explore these relationships in the context of Colombia. Using a Bayesian Mixed-Membership model, I estimate municipal-level clientelism. I find that municipalities with higher levels of clientelism are likely to receive more club goods projects and more discretionary transfers.

For a Working Paper, click here

(Draft Prepared in advance of the American Political Science Association Annual Meeting- August 2019)

When do Club Goods Buy Votes? Mayoral Cooperation in Clientelist Exchanges

When are mayors reliable clientelist brokers? In order for clientelism to work, brokers need to deliver votes. However, when those brokers are also politicians, they have incentives to deviate from the clientelist agreement to maximize their own reputation with voters. I focus on clientelist benefits in the form of targeted infrastructure investment as a benefit that legislators can use to induce mayors to deliver voters. I use a signaling model to predict the conditions under which mayors will votes. In order to test this model, I use survey experiments on both municipal politicians and citizens in Colombia to observe (1) when politicians are most likely to attribute credit for club goods to legislators and (2) whether these signals change citizen’s voting behavior.

For a Working Paper, click here

(This version explicitly presents the formal model, last updated 9/13/2019)

Additional Research


Testing the Effects of Nationally Mandated Participatory Reforms

Coauthored with Stephanie McNulty, Franklin and Marshall College 

Do participatory governance reform lead to increased citizen participation, improved accountability, and reduced clientelism? As participatory democratic institutions emerge around the world, we still know too little about their potential to impact democracy and governance in the developing world. This article explores these issues though the use of an original database of one hundred and twenty five countries in the developing world, which captures eighteen cases of nationally mandated participatory reforms passed and implemented between 1985 and 2015. The analysis suggests that nationally mandated participatory reforms do engage citizens more effectively in public-policy decision-making processes and can, especially in weakly decentralized environments, improve accountability. However, they do not curb clientelism, rather, they open up new venues for particularistic spending in subnational governments. Importantly, as the degree of decentralization increases, positive effects on participation and accountability get smaller and smaller. This suggests that these venues for participation may be most important in weakly decentralized contexts, where few additional means for effective participation exist in local and regional governments. In highly decentralized contexts, alternative means for reducing clientelism, increasing accountability, and even engaging citizens in decision-making will probably be more effective.

Chapel Hill, North Carolina

©2019 by Katherine McKiernan. Proudly created with