TAX COLLECTION AND BUREAUCRAT PERFORMANCE: EXPERIMENTAL EVIDENCE FROM THE DRC

We exploit random variation in whether local bureaucrats known as avenue chiefs were responsible for property tax collection (treatment), or whether agents of the tax ministry collected taxes within avenue chiefs’ jurisdictions instead (control). We study the impacts of working as a tax collector on the distribution of scarce benefits from an antipoverty program. Additionally, we cross-randomize two additional interventions. First, in one treatment arm, we provide information to randomly selected households about the antipoverty program as an encouragement for individuals to seek out the chief and ask for a share of the benefits. Second, in randomly chosen neighborhoods, citizens will have a chance to vote for audits of the chief or the provincial government; audits will occur in the neighborhoods with a highest percentage of voters.

2019-07-21

2019-10-31

The economic need of lottery ticket recipients, errors of inclusion and exclusion, receipt of a lottery ticket for the poverty program, diversion and nepotism by the chief, bias of the chief toward coethnics or taxpayers.

The current project builds on experimental variation in tax collection responsibilities to evaluate how collecting taxes affects the performance of local bureaucrats. Because the mode of collection for the 2018 property tax campaign was assigned on the neighborhood level, there is random variation in whether chiefs (whose jurisdictions are roughly coterminous with neighborhoods) have been charged with tax collection in 2018.

In addition to the underlying random variation in whether avenue chiefs collected property taxes in 2018, we cross-randomize three experimental arms specific to this study in bureaucrat performance. The first arm has no additional intervention. This arm will enable us to measure the first possible mechanism through which tax collection could impact chief performance: altering chiefs’ public spiritedness. This is one channel consistent with the classic tax-accountability hypothesis.

In the second arm, we will randomly select 20% of households within a selected neighborhood to receive information about the lottery.3 This treatment will create variation across neighborhoods in whether citizens have received information, but also within neighborhoods among citizens. Through door-to-door visits to households in selected neighborhoods, citizens will receive written and oral information about (a) the name and rank of the chief who will be responsible for distributing lottery tickets on the avenue, (b) the time period during which the chief will be distributing lottery tickets, (c) the total number of winners per neighborhood. The flier also notes that citizens can “see the chief for more information.” The information treatments function as an encouragement for citizens to talk to the chief and, if eligible, ask for lottery tickets.

Finally, in a third arm, we shock the capacity of citizens in the neighborhood to engage in collective action and exert pressure on their chief. We randomly give citizens an opportunity to vote for community audits of the chief who worked on the program as well as DIVAS itself. The audits will be conducted by well-known and respected local civil society organizations: RIAC (the Network for Transparency and Anti-corruption), which specializes in promoting transparency and fighting corruption, and SOCICO (the Civil Society of Congo), which focuses on government accountability in the areas of violence, conflict, and elections.

All randomizations will be done in Stata unless stated otherwise in our pre-analysis plan.

Tax collection responsibility was randomized on the neighborhood level. Cross-randomized interventions are assigned at the chief level.

Yes

361 neighborhoods

This varies by the outcome variable, as described in detail in the PAP. The estimated sample for the "need" outcome should have a sample size of more than 9000 observations.

See pre-analysis plan.

Pre Analysis Plan (updated 9/16/19)

Pre Analysis Plan