Abstract

Globally, bureaucrats are responsible for the day-to-day matters of policy implementation and government service provision. However, we know very little about how bureaucrats’ own interests and biases influence the ultimate distribution of public goods and services. To measure how citizens’ identities and the characteristics of their petitions affect municipal bureaucrats’ bias and efficiency, researchers conducted a national-scale audit of two of Colombia’s largest social service programs. They found evidence of moderate levels of discrimination on the basis of socioeconomic class and migrant status when callers asked how to access certain social benefits.

Policy Issue

While there have been many studies on how the actions of elected officials affect the provision of public services, less has been written on how access to those services is shaped by the behavior of the street-level bureaucrats who implement these programs. Bureaucrats influence the distribution of public goods and services in two respects: efficiency (how well they implement directives) and bias (who they choose to serve). Additionally, bureaucrats’ behavior is affected by their relationship with the elected officials they are accountable to. This research aimed to advance our understanding of how the behavior of local-level functionaries shapes access to social services, as well as the conditions under we observe variation in this behavior.

Context of the Evaluation

This study focused on two programs related to Colombia’s national social support system: Sistema de Identificación de Potenciales Beneficiarios de Programas Sociales (SISBÉN) and Más Familias en Acción (MFA). SISBÉN is a national survey of income and assets that determines if one qualifies for many of Colombia’s social support programs. MFA is Colombia’s main national conditional cash transfer program. These programs were selected for the audit in collaboration with officials from the Departamento Administrativo de la Function Pública, Departamento Nacional de Planeación, and Prosperidad Social. Both programs are national government programs with large numbers of beneficiaries and some implementation at the local level.

Details of the Intervention

Researchers partnered with the Departamento de la Prosperidad Social (DPS) and Departamento Nacional de Planeación (DNP) to conduct a randomized audit of two of Colombia’s largest social services, SISBÉN and MFA, in order to measure how applicants’ identities and the characteristics of their petitions affected municipal bureaucrats’ bias and efficiency.

Researchers conducted a total of 1,836 audits across 618 randomly selected alcaldías (municipalities). All municipalities (and localities in Bogotá) with an estimated population over 35,000 were included in the sample. Smaller localities were sampled proportional to population.

In these audits, confederates supervised by IPA called local government service offices posing as citizens requesting information on how to access certain SISBÉN or MFA benefits. Researchers randomly varied the script of the callers to measure how traits like socioeconomic class and migrant status affected bureaucrats’ responses.

The following aspects of surveyors’ scripts were randomized:

• Efficiency factors:
  • Time of day – surveyors called either before or after lunch, as worker absenteeism is expected to increase after lunch.
  • Difficulty of request -  surveyors either asked about the requirements for enrollment in the program (easy) or a more technical question about how to change enrollment in the program (difficult)
• Bias factors:
  • Regional accent – surveyors spoke with one of three regional accents (Bogota, Paisa and Costeño), to study whether responses varied when the petitioner’s accent was not native to the region.
  • Socioeconomic class – surveyors either spoke with vocabulary and phrasing that signaled low socioeconomic class, or in a neutral way.
  • Migrant status – surveyors either stated that they were a migrant from another locality or did not mention their migrant status. Migrants are less likely to be voters and more difficulty to handle administratively.

By comparing how the outcomes of the phone audits varied across these randomly assigned groups, researchers were able to measure the extent to which bias and efficiency factors affect access to SISBÉN and MFA. Researchers collected data on the following outcome measures:

• Absenteeism of bureaucrats
• Delegation of calls and access to program officers
• Adequacy of response to petitions
• Disposition of bureaucrats

Results and Policy Lessons

Researchers found moderate evidence of bias against Costeñas, low-income callers and migrants. Audits also revealed that the response rate to petitions was low across both programs, and the proportion of petitions that received complete, accurate answers was even lower.

Trajectory of calls: Across both programs, only about 65% of calls were answered within six attempts. Only about 50% of callers received a direct response to their question, and fewer still – about 35% - received any correct information. 

Response patterns: Calls were slightly more likely to be answered in the morning than in the afternoon. Across the 618 alcaldías surveyed, there was a large amount of variation in the number of calls answered:

• 148 never answered any calls,
• 155 answered some calls but not others,
• 315 answered all calls.

Accuracy of answers: In general, the quality of information provided to petitioners was low: fewer than 30% received a complete, correct answer to their question. As predicted, this rate was even lower for technical questions, fewer than 5% of which received a full, accurate answer.

Bias: Researchers examined evidence of bias in several areas:

• Socioeconomic class: Callers with “low-class” speech patterns received substantially less information and were more likely to be asked to visit in person.
• Migrant status: Callers that mentioned they were migrants received somewhat less information and were much more likely to be asked to visit in person.
• Regional accent: There was no evidence on the basis of regional accent in the national sample. Further there was no evidence of in-region favoritism in the regions to which these accents are native.

These results suggest that Colombian citizens face discrimination in their access to services on the basis of their socioeconomic class and migrant status. Importantly, these programs are ostensibly targeted to the populations that were least able to access information. Additionally, the overall low response rate to petitions suggests that much could be done to improve the accessibility of alcaldías and reduce bureaucrat absenteeism. The limited accuracy of answers points to potential opportunities to increase bureaucrats’ knowledge of these programs to provide more uniform service.