Transparency Reforms and Big Data

My second research program examines urban public sector reform initiatives involving the collection and dissemination of new types of data. This work examines the efficacy of such initiatives, as well as how such data enables scholars to revisit standard approaches to the study of local public services and comparative politics.

The project consists of a set of five papers anchored by a field experiment in Bangalore, India. Three of the papers represent collaborations with Isha Ray (Associate Professor, Energy and Resources Group, U.C. Berkeley). 
This research has been funded by Berkeley's Development Impact Laboratory under USAID Cooperative Agreement AID-OAA-A-13-00002.  Please do not hesitate to contact me if you would like to read one of these papers.


Flows, Leaks, and Blockages in Informational Interventions: Field Experimental Evidence from Bangalore's Water Sector 2018. World Development. 106: 149-160. (with Tanu Kumar and Isha Ray)

Can providing citizens with information compensate for unreliable public services? Existing scholarship examines whether increasing transparency motivates individuals to hold governments accountable. No one, however, has studied whether information that increases service predictability reduces coping costs or changes how citizens relate to the state. We conducted a field experimental evaluation of a program providing households in Bangalore with advance warning of the timing of intermittently provided water services. Contrary to expectations, we detected no impacts on household welfare, citizen attitudes about the state utility, or propensities to contact the state utility directly rather than through intermediaries. Noncompliance by street-level bureaucrats charged with supplying water timing information both reduced the power of the experiment and undermined the program’s viability. These findings suggest that scholars should investigate and report how frontline worker actions drive positive or null experimental findings. We introduce a framework for analyzing “information pipelines” to assist with such efforts.

“Frontline Worker Compliance with Transparency Reforms: Barriers Posed by Family and Financial Responsibilities” (with Christopher Hyun and Isha Ray) 2018. Governance. 31: 65-83. (Formerly "Why Street Level Bureaucrats Don't Comply: Insights from the Valvemen of Bangalore")

Significant development funding flows to informational interventions intended to improve public services. Such “transparency fixes” often depend on the cooperation of frontline workers who produce or disseminate information for citizens. This article examines frontline worker compliance with a transparency intervention in Bangalore’s water sector.  Why did compliance vary across neighborhoods, and why did workers exhibit modest rates of compliance overall? Drawing on ethnographic observation and an original dataset, this article finds that variation in workers’ family responsibilities and financial circumstances largely explains variation in compliance with the intervention. Furthermore, workers often prioritize longstanding responsibilities over new tasks seen as add-ons, leading to modest rates of compliance overall.  Perceptions of “core” jobs can be sticky—especially when reaffirmed through interactions with citizens. This study represents one of the first multi-method companions to a field experiment, and illustrates how the analysis of qualitative and observational data can contribute to impact evaluation.

“Crowd-Sourced Data for Comparative Politics: Insights from Urban India.” Forthcoming.  Studies in Comparative International Development. (with Anustubh Agnihotri and Christopher Hyun)

In much of the developing world, states fail to collect accurate information on social and political phenomena such as protest activity, the extent and reach of public services, and corruption. Crowd-sourced data—data sourced electronically from disparate individuals or groups—now offers researchers new information that can potentially correct for biases in existing data about such processes. In this paper, we highlight ways in which crowd- sourced data can be used to cast new light on major problems in comparative politics and show how important analytic pitfalls can be avoided. Descriptive inference using crowd- sourced data can be improved by explicitly examining how data is contributed and conducting ground truthing exercises. Assessments of the causal impact of crowd- sourced data on political processes can be strengthened by randomizing exposure at the cluster rather than individual level, or by measuring information spillovers. We illustrate strategies for addressing inferential difficulties in urban India, through a study of the collection of crowd-sourced information on water delivery. In this study, crowd-sourced data on water allocation is used to revisit long-standing debates on distributive politics and newer debates regarding the potential effects of informational interventions on citizen political participation.

“Distributive Politics in Networked Systems: The Political Geography of Water Intermittency” (with Tanu Kumar, Megan Otsuka, Francesc Pardo-Bosch, and Isha Ray) (in preparation)


Alison Post,
Dec 13, 2016, 10:47 AM