Organizational Dynamics and Administrative Data Production

The experimental interventions are embedded in a new data collection effort by a Colombian government agency. To understand how communication within local bureaucracies influences administrative data submission, we use an encouragement design to increase the level of communication or communication between bureaucrats and politicians or other bureaucrats in their local government when collecting data for submission to the national government agency. Municipal governments assigned to treatment receive communication promoting this type of collaboration or communication. Those assigned to control do not receive this encouragement. To measure bureaucratic uncertainty, we use a within-subjects randomized elicitation of uncertainty.

2022-10-24

2022-12-07

I measure the distribution of responses across all questions on the instrument, validate accuracy of responses (where objective data are available), and measure bureaucrats' level of certainty in select responses.

I use enumerator assessments of bureaucratic performance in the data submission process.

The primary experimental manipulation utilizes an encouragement RCT design. In this design, municipal governments are randomly assigned to a treatment or control condition. The treatment encourages bureaucrats to consult with other individuals in their local government—whether other bureaucrats or the elected mayor—or others in their municipality. I contrast this to a control condition in which bureaucrats are simply asked to prepare responses. In the data collection, enumerators measured levels of collaboration or communication within the institution (the endogenous treatment). In the first stage, this design identifies the effect of the encouragement on levels of collaboration within local governments. I then can use this experimental manipulation to identify the ITT of encouragement or the complier average causal effect (CACE or LATE) of collaboration on attributes of the data submitted to the national government. Additionally, the design permits identification of quantile treatment effects, which measure the effects of encouragement on the distribution of responses.

The secondary experimental manipulation provides novel measurement of bureaucratic uncertainty. This within-subjects design elicits respondents' level of certainty three times, once per each block of two questions. In each block, of two questions, a municipal government is randomly assigned to provide their level of certainty on the first or second question. This design allows for measurement of uncertainty as well as studying the effects of certainty elicitation on subsequent responses. For the latter, this design facilitates the identification the ATE (and quantile treatment effects) of certainty elicitation on subsequent responses.

Computer

For the encouragement design: the municipal government. For the secondary manipulation: the municipality-question (a within subjects design)

No

1102

1102

551 per arm for the encouragement design.

For the ITT analysis, a standardized effect of 0.2 control group standard deviations.