Friday, June 19, 3:30 pm
The Child Opportunity Index 2.0: A New Index of Neighborhood Opportunity for All US Neighborhoods
Presenter: Clemens Noelke, Research Director, Institute for Child, Youth and Family Policy, Heller School for Social Policy and Management, Brandeis University
Dolores Acevedo-Garcia, Brandeis University
Mikyung Baek, Ohio State University
Erin Hardy, Brandeis University
Rebecca Huber, Brandeis University
Nick Huntington, Brandeis University
Nancy McArdle, Brandeis University
Nomi Sofer, Brandeis University
Michelle Weiner, Brandeis University
The Child Opportunity Index (COI) 2.0, to be launched publicly in January 2020 on diversitydatakids.org, is a new composite index of neighborhood opportunity. Unlike opportunity metrics capturing long-term outcomes of past neighborhood conditions, COI 2.0 is based on 29 contemporaneous predictors of healthy child development. COI 2.0 is based on COI 1.0, which has been used in academic studies and applied research with a focus on the Boston metro area and has featured prominently in local media. COI 1.0 used 2010 data for the 100 largest metro areas and only permitted neighborhood comparisons within metro areas. COI 2.0 has comparable data for all US census tracts for 2010 and 2015. In this presentation, we will survey the construction and explore the predictive validity of the COI 2.0 in relation to other metrics of opportunity.
Source data on 29 component indicators were collected from open-source and restricted datasets. Indicators were standardized using 2010 means and standard deviations. The resulting z-scores were combined into three domain scores (education, health/environment, social/economic) and domain scores were combined into an overall index score. Weights were used that reflect how strongly each component indicator or domain score predicts four health and socioeconomic outcomes. Census tract-level outcome data was taken from the Opportunity Atlas and the 500 Cities health indicator database. To explore the predictive validity of the COI, we used additional outcome data from the same sources and CDC data on life expectancy. We regressed tract-level outcomes on COI domain and overall scores and compared the percent variance explained by COI 2.0 and two other metrics of neighborhood opportunity, an indicator of intergenerational economic mobility taken from the Opportunity Atlas and a socio-economic index of Concentrated Disadvantage.
Results show that the COI 2.0 overall index score explains more variation in outcomes than the three domain scores or any of the component indicators, illustrating the gains in terms of predictive validity from pooling information across many indicators. Furthermore, the COI 2.0 overall index score explain as much or more variation in outcomes compared to the two other metrics of neighborhood opportunity/conditions examined.
The COI 2.0 is a new tool for researchers who want to better understand the role neighborhood conditions play in shaping children’s outcomes, and it’s a new tool for professionals across many sectors who want to better understand the neighborhood conditions children face in the communities they serve. Given past uses of COI 1.0, we expect that COI 2.0 will be broadly used across the health, education, and housing sectors for academic research on the role of neighborhoods in children’s health and long-term outcomes, to conduct community needs assessments, to analyze and highlight local inequity in access to opportunity, for strategic planning, resource allocation, and in-forming place-based and mobility interventions.
What You Need to Know about Differential Privacy and the 2020 Census: Tradeoffs and Real-World Implications
Presenter: Cliff Cook, Senior Planning Information Manager, City of Cambridge Community Development Department
Bailey Werner, City of Cambridge
Privacy and security of personal data is a significant concern for the public sector. When citizens share data with the government there is an implicit expectation that the general public will not have access to personally identifiable information.
The Census Bureau must balance data privacy with the release of a large number of highly detailed tables at many levels of geography. Many organizations depend on these tables to develop data products and analyses that touch every part of the economy and society. Recent technical developments, namely the increase of computing power and improved methods for data reconstruction, have undermined older data protection methodologies and informed the Bureau’s decision to apply Differential Privacy (DP) protection to data collected by the 2020 Census.
DP methods inject noise to “blur” the data in a way that retains the generalized properties of the population while making it difficult to extract data about individuals. However, this method does not come without costs— by definition, DP provides a quantifiable tradeoff between accuracy and privacy. Our presentation provides a brief, non-technical introduction to DP then focuses on some of the real-world issues presented, using examples from planning at the municipal level
US census data, differential privacy, and spatial uncertainty: implications for using census tract data to quantify health inequities
Presenter: Nancy Krieger, PhD, Professor of Social Epidemiology, American Cancer Society Clinical Research Professor, Department of Social and Behavior Sciences, Harvard T.H. Chan School of Public Health
Brent Coull, PhD, Professor of Biostatistics, Associate Chair of the Department of Biostatistics, Department of Biostatistics, Department of Environmental Health, Harvard T.H. Chan School of Public Health
Jarvis T. Chen, ScD, Research Scientist, Department of Social and Behavioral Sciences, Harvard T.H. Chan School of Public Health
Rachel Nethery, PhD, Assistant Professor, Department of Biostatistics, Harvard T.H. Chan School of Public Health
Pamela D. Waterman, MPH, Project Director, Department of Social and Behavioral Sciences, Harvard T.H. Chan School of Public Health
Emily Wright, BA, Doctoral student, Department of Social and Behavioral Sciences, Harvard T.H. Chan School of Public Health
Tamara Rushovich, MPH, Doctoral student, Department of Social and Behavioral Sciences, Harvard T.H. Chan School of Public Health
Despite the importance of accurate census data for public health, little is known about how census tract (CT) estimates of health rates and inequities are affected by: (1) differential privacy, whereby statistical “noise” is injected into the publicly released decennial census (DC) data to protect individual privacy; and (2) spatial uncertainty, which for the American Community Survey (ACS) is due to the small number of different persons surveyed each year in a given CT, producing high margins of error for point estimates.
We accordingly compared estimates of the 2008-2012 average annual premature mortality rate (death before age 65) and inequities in Massachusetts using CT data from the 2010 DC, 2010 DC with differential privacy, and 2008-2012 ACS 5-year estimate data.
For these 3 denominator sources, the age-standardized premature mortality rates (per 100,000) for the total population respectively equaled 166.4 (95% CI 162.2, 170.6), 166.4 (162.2, 170.6), and 166.4 (162.1, 170.6), and ranged across the best to worst quintile for CT racialized economic segregation (aggregating across CTs in each quintile) from 103.4 to 260.1, 102.9 to 258.7, and 102.8 to 262.4. These results suggest estimates are insensitive to the denominator source, a finding which held across racial/ethnic groups and by gender.
Moderator: Beth Huang, Director, Mass Voter Table