Association for Public Policy Analysis & Management

In many metropolitan areas, critical infrastructure such as electricity, gas, water, and wastewater systems are owned and operated by a complicated array of public and private entities. Fragmented, polycentric systems are often theorized to improve metropolitan outcomes through competition, flexibility, and responsiveness. However, highly fragmented institutional environments create many challenges, including that local providers can lack the institutional and fiscal capacity to maintain service quality and that fragmented authority makes it hard to account for system interdependencies. While much attention has been paid to the geographic distribution of particular service outcomes (e.g., drinking water quality) we still do not know very much about how fragmentation affects the quality and equality of service outcomes in urban areas--and how outcomes vary when interdependent water and energy systems are considered in concert.

To understand how fragmentation and interconnectivity affect service outcomes, this project focuses on the case of electricity, gas, water, and wastewater systems in the 382 metropolitan statistical areas (MSAs) in the United States. We conceive of each MSA as a two mode, or bipartite, network wherein census tracts are first level nodes and water, electricity, gas, and wastewater providers are second level nodes. To build these networks, we combine data from the US Census, the US EPA’s Safe Drinking Water Information System, annual pipeline operator safety reports from the US Department of Transportation, and several other national databases. We then use text mining and web scraping tools to supplement these data with public financial records, geographic location data, and other contextual attributes.

We address two primary questions: (1) How does the structure of an MSA’s service delivery network--for instance, whether an MSA has high functional fragmentation or geographic fragmentation--relate to service outcomes?; and (2) How do outcomes vary by economic and social strata given system structure? To analyze these questions, we use a two-stage analysis. The first stage will use graph modeling to characterize the overall structure of a given MSA network. This will provide systematic measures of the extent to which management and operations are centralized or fragmented, and quantify the position of individual localities within a given network. The second stage will then use a joint likelihood, multilevel model that regresses three outcome variables--water quality violations, gas leaks, and electricity outages--on infrastructure network characteristics generated in the first stage, as well as key environmental, economic, and social covariates. Using a multilevel specification will allow for active modeling of variables of interest at the local, regional, and state levels. This will allow us to assess how drivers at multiple scales, such as state rules governing institutional formation and grant programs, affect the distribution of service outcomes.