Research

In this paper, I study the effect of the political assignment of temporary public jobs—referred to here as short-term patronage—on the strategic behavior of temporary employees in developing public administrations. I examine the interplay between job uncertainty and political patronage as determinants of temporary employees' work effort, operationalized as time allocation to a project. The central argument is that temporary employees strategically adjust their time allocation to mitigate future job uncertainty, and patronage can incentivize increased or reduced effort contingent upon such uncertainty. Using a novel vignette experiment with Colombian public employees, I find that lower job uncertainty can lead to an increase in job effort by temporary employees. Furthermore, when future job uncertainty is lower (higher), patronage can increase (decrease) temporary employees' time dedication to their tasks. These findings underscore the complex and often unintended consequences of the interaction between employment flexibilization reforms and the political dynamics of public administration, particularly in the context of weakly professionalized bureaucracies.

How does patronage—the political appointment of bureaucrats—affect the coordination and joint delivery of policies? Research has examined patronage's effects on bureaucratic performance, but mostly within hierarchical, top-down policymaking. However, in many domains, governance depends on horizontal networks of intergovernmental collaboration. I argue that patronage's effect on collaboration depends on the fit between the resources appointees carry and the demands of the positions they fill. Patronage facilitates coordination when appointees' political capital can be deployed for brokerage and negotiation but undermines it when loyalty-driven hiring displaces technical expertise and organizational stability. To test this argument, I analyze the network of environmental collaboration agreements among Colombian public agencies using hierarchical Exponential Random Graph Models. The evidence shows that managerial patronage fosters collaboration, especially along co-partisan lines, while professional-level patronage inhibits it. These findings underscore patronage's contingent role in governance networks and the importance of bureaucratic politics for collaborative policy delivery.

Cross-boundary collaboration to produce public services or govern policy problems can emerge as contingent dynamics initiated by local governments or driven by incentives and mandates from higher-level governments. This research focuses on understanding collaboration regimes in the Global South, a region where decentralization processes have left a centralist legacy of underfunded local governments with minimal autonomy. We synthesize elements from the Collaborative Governance Regime and Institutional Grammar tool as research lenses to propose a cross-boundary collaboration framework for the Global South to understand how national-level system contexts and decentralization processes shape the institutional arrangements that regulate cross-boundary interactions. As a proof-of-concept demonstration, we apply the framework across three Latin American countries: Chile, Colombia, and Mexico. This initial application suggests that different paths of decentralization not only set the devolution of capacity and autonomy to local governments, but also determine the level and type of rules under which cross-boundary collaboration emerges.

Bureaucratic autonomy is one of the most important institutional determinants of government performance, yet its effects have been studied almost entirely within hierarchical settings. This paper extends the comparative bureaucratic politics literature to polycentric systems, a form of governance that depends on the coordination of multiple autonomous decision-making centres, and that characterises much of contemporary governance. Public bureaucracies function as primary nodes in most polycentric systems, and the political arrangements that govern their recruitment, tenure, and organisational independence reshape the conditions under which polycentric coordination can succeed or fail. I propose three analytical dimensions through which bureaucratic autonomy shapes polycentric systems: capacity, intertemporal dynamics, and political capital. Because no single configuration of autonomy optimises all three dimensions simultaneously, the framework argues that the level of autonomy best suited to polycentric coordination shifts as the governance system develops. The paper proposes mechanisms to respecify enabling conditions and dysfunction syndromes identified in polycentric theory in terms of their upstream political-administrative determinants, and proposes a research agenda that bridges polycentric governance with comparative bureaucratic politics and public administration reform.

Intergovernmental collaboration has become a central mechanism for governing complex environmental problems. However, empirical evidence on whether collaboration improves governance outcomes remains limited. Governments select into partnerships based on shared problems, political connections, and institutional constraints, making it hard to separate selection from the causal effects of collaboration. This paper proposes a strategy for estimating the effects of collaboration networks on policy outcomes while addressing endogenous tie formation. I exploit the fact that authorities participate in several policy networks at once, using an exogenous network to build instruments. Using a friction-only gravity model, I estimate the structural propensity to form ties, and combine it with partner spending shocks to instrument for collaboration intensity. I apply this approach to municipal drinking water governance in Colombia, using over 4,700 inter-administrative collaboration agreements. Entering collaboration depresses water quality, while additional partnerships and greater collaborative investment produce sizeable improvements. The findings highlight the importance of distinguishing network formation from network effects when evaluating collaborative governance.

Environmental governance worldwide increasingly operates through multi-level, decentralised institutional architectures. Yet the environmental outcomes of these arrangements vary enormously, and we still know relatively little about the macro-institutional conditions under which the coordination mechanisms that polycentric and multi-level systems rely upon can actually flourish. I argue that a foundational but overlooked conditioning factor is the politicisation of bureaucratic appointments—the extent to which politicians control the recruitment of public servants. Using a panel of 91 countries over the period 2005–2022, I examine how the interaction between decentralisation and bureaucratic politicisation shapes four environmental outcomes: overall environmental performance, ecosystem vitality, environmental health, and climate change mitigation. Results from within-between random-effects models, complemented by hierarchical GMM estimators, show that whilst decentralisation is associated with short-run declines in environmental quality, countries with structurally more decentralised systems perform better in the long run. Bureaucratic politicisation amplifies the short-term environmental costs of decentralisation, but its moderating role varies by environmental domain: climate mitigation—where coordination demands are highest and political credit-claiming is most difficult—is the most severely affected, whilst environmental health—comprising locally tangible goods such as drinking water and sanitation—is the least vulnerable.

This paper examines how AI chatbots influence public opinion on climate-related fiscal policy. Using a randomised controlled survey experiment in Germany (N=1,074) and the UK (N=1,200), we test whether AI-generated recommendations affect support for fuel taxation and climate investment. Participants viewed identical policy advice framed either as ChatGPT output or general opinion, enabling causal inference regarding framing effects. The design includes pre-treatment measures of AI trust, attitudes, usage patterns, and demographics to examine heterogeneous treatment effects. Results show that pro-climate arguments consistently increased policy support; however, ChatGPT attribution did not uniformly enhance persuasion. In Germany, ChatGPT attribution produced a marginally significant convergence toward the recommended fuel tax level, whereas in other conditions, respondents shifted slightly more when advice was unattributed, though the differences were not statistically significant. The evidence suggests that LLM-generated pro-climate content is persuasive, while attribution effects remain context-dependent.

One of the main causes of this forest conversion is international demand for agricultural commodities, which creates a fundamental trade-off for tropical countries between short-term market rents and long-term environmental conservation. This trade-off is politically asymmetric: concentrated agribusiness interest groups typically overpower diffuse conservation constituencies. Under what conditions can domestic institutions resist these pressures? This paper argues that bureaucratic autonomy mitigates the effect of international commodity price shocks on agriculture-driven deforestation. Autonomous bureaucracies are better insulated from political and business interest-group interference, enabling more stringent and effective environmental regulation, monitoring, and enforcement, even when commodity booms heighten incentives for frontier expansion. Using satellite-measured forest loss attributed to permanent agricultural conversion, I test this argument with a global panel of approximately 430,000 cell-year observations across 101 countries from 2001 to 2022. I find that a 10% price increase raises deforestation by approximately 7.1% in low-autonomy countries but only 3.9% in high-autonomy countries. Findings are robust to competing moderators, alternative specifications, and moderation diagnostics. These results extend the literature on bureaucratic and regulatory independence to environmental protection and identify personnel autonomy as an upstream institutional lever for mitigating agricultural deforestation.

Recent work on patronage argues that political appointments differ in whether officials are chosen for loyalty or for competence, and that this follows from how parties are organised and how they win votes. I build on these arguments to classify governments in 121 democracies between 1970 and 2019. Using a fuzzy-membership procedure, I give each government a graded profile across four types of patronage appointment. I validate the measure against existing case-study evidence on patronage, map how countries compare, how they change over time, and examine how each patronage type relates to other government characteristics such as impartiality and corruption. The comparison shows that the kind of patronage politics that operates in a country and its intensity are separable dimensions, and that they carry different implications for the quality of government.

Climate adaptation is often treated as a public good, yet its benefits are spatially uneven and excludable, making it closer to an impure, club, or even private good. Climate politics research has shown that citizens may support climate action in principle, particularly after they experience the risks of climate change, but might withdraw support when they become net contributors to protecting others. Focusing on flood risk management in England, in this paper we examine how people perceive climate adaptation as a good and how such perception varies depending on their flood risk exposure. Empirically, we embed a discrete choice experiment in an information provision experiment with 1,500 respondents. Participants choose between flood protection strategies varying in excludability, distributional fairness, funding mechanism, effectiveness, and household cost, allowing estimation of willingness to pay for each attribute. A randomly assigned treatment group learns their address-level flood risk before the choice tasks. We test whether personal risk awareness shifts support toward localised, excludable measures, with implications for equitable and politically sustainable adaptation policy.

Exponential random graph models (ERGMs) are widely used to study political networks, but standard implementations assume homogeneity across actors once covariates are included. This assumption rarely holds in political systems, where networks are routinely embedded in administrative, partisan, or territorial hierarchies. In this paper, I propose a mixed ERGM framework that incorporates group-level random effects as a computationally tractable way of modelling hierarchical structures. Building on the iterative estimation procedure of Kevork and Kauermann (2022), I generalise their framework to group-level intercepts, combining simulation-based maximum likelihood for structural parameters with penalised pseudolikelihood for group effects. This yields partial pooling across groups at computational costs feasible for the large sparse networks common in political research, and without the bias in structural parameter estimation that afflicts pure pseudolikelihood approaches. I introduce a pseudo-intraclass correlation coefficient to quantify the degree of group-level clustering and assess whether a hierarchical specification is needed. Monte Carlo experiments and applications to demonstrate that fixed-effects and pooled alternatives produce biased estimates under heterogeneous data-generating processes, and that the proposed estimator outperforms both, particularly when groups vary in size. I also illustrate the method using canonical policy network data.

In Territorial Association, Multi-level Governance and Decentralisation (Sanabria-Pulido, Rodríguez-Caporalli & Leyva, eds.). World Bank and Icesi University, Cali.

In Citizen Security from a Metropolitan Governance Perspective (Gallego, Leyva & Mesa, eds.). EAFIT University, Medellín.

In Citizen Security from a Metropolitan Governance Perspective (Gallego, Leyva & Mesa, eds.). EAFIT University, Medellín.