Bureaucratic Organization and Service Provision: Field Experimental Evidence from Colombia

How to cite

Slough, Tara. 2022. "Bureaucratic Organization and Service Provision: Field Experimental Evidence from Colombia." AEA RCT Registry. AEARCTR-0009963. https://doi.org/10.1257/rct.9963-1.0

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Tara Slough*

* Primary investigator

Registry ID
AEARCTR-0009963
DOI
https://doi.org/10.1257/rct.9963-1.0
First registered
Last updated
Status
Completed
Trial dates
2018-02-10 – 2018-04-01
Countries
Colombia ()
JEL codes
H4, H75, I38, P17
Registry record
http://www.socialscienceregistry.org/trials/9963

Abstract

A substantial literature in political economy seeks to explain inequality in access to public goods and services. However, bureaucrats are typically absent or otherwise assumed to be non-strategic actors. Nevertheless, the typical rationale for bureaucrats as “producers of public goods” challenges these assumptions. As the producers and ultimate distributors of public goods and services, I argue that bureaucrats’ role in the distribution of public goods and services should be conceptualized along two dimensions: efficiency and bias. In consultation with three national government agencies in Colombia, I conduct a nationwide factorial audit experiment that measures both efficiency and bias in the administration of two large national social welfare programs.

This trial was preregistered at Evidence in Governance and Politics prior to the receipt of any data. It was retrospectively added to the AEA registry. The EGAP pre-registration and pre-analysis are available at https://osf.io/wg7pc,

Intervention

This is an audit experiment that studies bureaucratic service provision in Colombian local governments. Specifically, I measure how bureaucrats respond to informational requests from petitioners from different groups (varying class, regional accent, and status as an internal migrant). I vary the difficulty of the petition to see how identity-based biases in effort respond to more challenging requests. Finally, I randomly assign whether calls were conducted in the morning (when absenteeism is lower) or afternoon (when absenteeism is higher) to measure absenteeism and ensure adequate response rates.

Experimental design

This audit experiment consists of a 2x2x3x2x2 experimental design to measure biases in service provision for two social service programs. I vary five factors experimentally:
- The social class of the petitioner (lower or lower-middle)
- The regional accent of the petitioner (Bogotano, Paisa, or Costeño)
- The migrant status of a petitioner (municipal resident or internal migrant)
- The difficulty of the question (hard or easy): Note that these questions are unique to each program
- The timing of the call (morning or afternoon).

I sample 618 municipalities, including all municipalities over 35,000 residents (n = 220) and a sample of 398 municipalities sampled with sample weights increasing in population for the municipalities under 35,000 residents. I then conduct 6 calls over a period of a month in municipalities with a population over 100,000; 4 calls over a period of a month in municipalities with between 35,000 and 100,000 residents; and 2 calls over a period of a month in municipalities with a population lower than 35,000 residents. This permits within-municipality comparisons.

I use block random assignment to ensure that the randomized attributes exhibit variation along every dimension within every municipality.

Design details

Randomization method
I conducted blocked random assignment using a computer.
Randomization unit
The unit of randomization is the call.
Clusters
There are no clusters.
Observations
1836
Arms
This study uses a 2x2x3x2x2 factorial design, so there are 48 cells. All of the treatments were assigned with equal probability except for one (the difficulty of request), which varied across population strata (see pre-analysis plan). As such, the expected number of observations per cell for the easy petition was 41.58 and the expected number of observations per cell for the difficult petition was 34.91.
Primary outcomes
I measure multiple attributes of calls. First, I measure how the calls progress through the alcaldía (seeing if they reach the relevant program officer from a predefined list). Second, I measure the accuracy of information provided.

Cited by

No citation data available for this registration yet.

Keywords