Working papers

• Command and Can't Control: Assessing Centralized Accountability in the Public Sector (submitted)- with Saad Gulzar (Notre Dame), Muhammad Zia Mehmood (Princeton), Daniel Rogger (World Bank)

A long-established approach to management in government has been the transmission of information up a hierarchy, centralized decision-making by senior management, and corresponding centralized accountability; colloquially known as "command and control". This paper examines the effectiveness of a centralized accountability system implemented at scale in Punjab, Pakistan for six years. The scheme automatically identified poorly performing schools and jurisdictions for the attention of central management. We find that flagging of schools and corresponding de facto punishments had no impact on school or student outcomes. We use detailed data on key elements of the education production function to show that command and control approaches to managing the general public sector do not induce bureaucratic action towards improvements in government performance.

• Pay-to-play: Campaign Contributions and Long-term Distortions in Public Procurement - with Saad Gulzar (Notre Dame) and Juan Felipe Riaño (Georgetown). 

Why do so few firms and individuals engage in political contributions despite mounting evidence of their political and economic returns? This paper addresses this generalized version of Tullock's puzzle by examining (i) how campaign donations translate into persistent advantages in public procurement, (ii) why current policies fail to prevent such advantages, and (iii) why, despite these long-term returns, relatively few individuals and firms donate to political campaigns. Using  data linking the universe of political contributions to every contract-level procurement record in Colombia from 2011 to 2025, we document that (1) donor are more likely than non-donors to win public procurement contracts, and the contracts they receive are often more discretionary and of higher value; (2) donors continue receiving contracts even \textit{after} losing political connections in later election cycles, suggesting donations provide prolonged access not explained by immediate reciprocation; and (3) donor-secured contracts frequently exceed budgetary limits and tend to benefit riskier firms and inexperienced individuals. To assess whether policy can mitigate these dynamics, we conduct a randomized controlled trial (RCT) with investigative journalists, NGOs, and Colombia's Inspector General Office to study how newly elected mayors respond to deterrence messages warning of scrutiny over contracts awarded to donors. We find that deterrence is more effective when messages include actionable information and explicitly identify donor firms and individuals. To understand why donations remain limited, we combine the intervention with administrative data and estimate a dynamic discrete choice model of selection into donations and public procurement. Our analysis shows that, despite the significant returns, one potential explanation to the puzzle is the presence of high entry costs and operational frictions deterring competition and widespread contributions.

Best paper in Public Procurement at the World Bank Conference on Public Institutions for Development

Articles

• One Step Ahead of the Law: The Net Effect of Anticipation and Implementation of Colombia's Illegal Crops Substitution Program - with Santiago Saavedra (URosario) and Daniel Wiesner (C-Analisis). Journal of Public Economics 202, 104498 [Data&Code]

Pre-announced policies often generate unintended consequences due to individuals’ acting to take advantage of the policy conditions. Little is known about  the extent to which unintended consequences from the early announcement of a policy can be larger than the implementation effect, especially in contexts of weak state capacity. We use detailed 1 km grid square data on coca cultivation to estimate the net effect of the announcement and implementation of coca crop substitution payments in Colombia. Our fine-grain data also enable us to estimate geographical spillovers of the program to non-targeted neighboring areas. Using a difference-in-differences empirical strategy, we find that program recipients reduced coca acreage. Surprisingly, the reduction in neighboring grid areas is of a similar magnitude. However, the effectiveness was reduced by half, because farmers increased coca cultivation in order to be eligible for the program after the announcement and before implementation. But the policy’s net effect is negative: due to weak state capacity, the program could not be implemented in every region of the country. 

Work in progress

• Election Day is Firing Day: Political Turnover, Patronage and Monitoring - with Anton Arbman-Hansing (IIES) and Arstchil Okropiridse (IIES). 

• The Political Economy of Environmental Protection: Evidence from India - with Muhammad Zia Mehmood (UC Berkeley) and Suraj R. Nair (UC Berkeley)