Working paper
The Effects of a Politician-Targeted Social Accountability Initiative on Public Works Provision: Evidence from a Large-Scale Experiment in the Philippines
Nico Ravanilla, Eitan Paul, Allen Hicken, and Ronald Mendoza
In brief
Can civil society pressure aimed at politicians rather than voters improve public works? A nationwide experiment across 235 Philippine legislative districts finds advocacy shifted spending toward less corruption-prone projects but degraded contracting quality, and only the threat of media coverage prevented that backsliding.
Abstract
Politicians often use public works procurement to further their own interests. We test whether a social accountability initiative targeted at politicians can alter such behavior in a weak-oversight setting. We randomly assigned half of the legislative offices in the Philippines to receive advocacy from a civil society organization encouraging more transparent and efficient public works provision, half of which were also told their performance would be reported to media outlets. On average, the advocacy prompted legislators to prioritize less corruption-prone projects, but it had no effect on transparency or geographic allocation. Moreover, it decreased the competitiveness and quality of construction contracting. Politicians from dynasties, insecure seats, or seeking re-election were least responsive to the advocacy. The threat of media coverage offset these adverse effects. Overall, our findings suggest that bottom-up information campaigns can influence elite behavior even in settings with weak governmental oversight, though effects are modest and context-dependent.
Research annotationQuestion, variables, design, findings, and mechanism at a glance▾
- Research question
- Can civil society advocacy aimed directly at legislators, rather than at voters, change how those legislators allocate and procure public works in a patronage democracy where formal oversight is weak?
- Independent variable (X)
- Random assignment of legislative offices to an advocacy campaign in which civil society partners presented chiefs of staff with district-tailored data on public works distribution and procurement, offered specific reform actions, and told legislators their districts would be monitored through the May 2019 midterms. A cross-cutting second arm added an explicit threat that information on the office's response would be shared with local media around those elections.
- Dependent variable (Y)
- Four families of administrative outcomes: the share of funds going to corruption-prone project types such as roads and flood abatement versus less corruption-prone types such as health facilities and schools; procurement transparency, measured by posted contract documents and a composite index; contractor concentration and construction quality ratings; and per capita allocations across municipalities within a district, split by whether the mayor is clan-aligned with the legislator.
- Identification strategy
- Block randomized field experiment across the 238 legislative offices of the Philippine House, of which 235 single-member districts remain in the estimation sample, each serving roughly 450,000 constituents. Assignment was at the district engineering office cluster level to avoid spillovers, giving 171 units of randomization: 86 control clusters covering 128 districts, 42 advocacy clusters covering 49 districts, and 43 advocacy-plus-media clusters covering 58 districts. Blocking covariates were dynastic status, last-termer status, seat competitiveness, gender, and alignment with the incumbent party. Intent-to-treat effects are estimated by OLS on an advocacy indicator and a media indicator plus blocking covariates and baseline outcomes, with robust standard errors clustered at the district engineering office level. Outcomes come from roughly 65,000 awarded contracts involving more than 3,800 contractor firms, scraped from the public works department over the 2016 to 2019 congressional term.
- Main findings
- Advocacy shifted the composition of spending: the road share fell by 0.131 against a control mean of 0.334, the school share rose by 0.164 against a mean of 0.379, and the combined corruption-prone share fell by 0.153 against a mean of 0.478. Transparency did not move at all, with a composite index coefficient of 0.002. Contracting got worse: advocacy raised contractor concentration by 0.097, about a 30 percent increase on a control mean of 0.327, and lowered average construction quality ratings by 0.700 points. Legislators did not reduce favoritism toward clan-aligned municipalities but cut per capita allocations to all municipalities. The media arm reversed the contracting damage, cutting contractor concentration by 0.163 and restoring municipal allocations. Politicians from dynasties, in competitive seats, or seeking reelection were least responsive.
- Mechanism
- The initiative activates the belief that legislators are being watched without supplying a sanction. Where that sanction is absent, legislators comply on the cheap and visible margin, project type, while recouping the forgone rents on the margin no one is auditing, contracting. Adding the media threat supplies the missing sanction and reverses the perverse response.
- Why it matters
- Tests the politician-targeted variant of social accountability, which bypasses the voter collective action problem that has defeated voter-targeted information campaigns, at national scale in a patronage democracy. It documents that partial compliance is not benign: an advocacy campaign can improve one governance margin while degrading another, and a reputational sanction is what keeps the improvement from being offset.
- Speaks to
- social accountability and anti-corruption interventionsdistributive politics and public works procurementclientelism and patronage democracymedia and political accountabilityfield experiments on elite behavior
Cite
@misc{paul_accountability,
title = {The Effects of a Politician-Targeted Social Accountability Initiative on Public Works Provision: Evidence from a Large-Scale Experiment in the Philippines},
author = {Nico Ravanilla and Eitan Paul and Allen Hicken and Ronald Mendoza},
howpublished = {Working paper},
}