The Child Opportunity Index (COI) 3.0 is the most recent version of the COI. It succeeds COI 1.0 (published in 2014) and COI 2.0 (published in 2020). Each successive version has added new relevant indicators, improved the underlying methodology and increased the amount of available data. The feedback we have received from our growing data user community has been invaluable in this process. The following table summarizes key differences across the three versions and charts the evolution of the COI.
More recent—and annual—data
COI 3.0 is available for ten consecutive years (2012 through 2021), while COI 2.0 was available for two years (2012 and 2017) and COI 1.0 was available for one year (2012). (Note that most COI component indicators combine data from a five-year period and that we have changed how we label these periods. For COI 2.0, we had labeled five-year periods by their mid-point year, i.e., “2015” for data covering 2012-2017. We now label five-year periods by their terminal year, i.e. “2017” for data covering 2012-2017.)
We intend to update COI 3.0 annually, using the same protocols for the foreseeable future.
New and revised indicators
The indicators that comprise the Child Opportunity Index are selected for their relevance to child wellbeing, their quality and reliability and their availability across all census tracts. The number of indicators has grown from 19 in COI 1.0 to 29 in COI 2.0 to 44 in COI 3.0.
New indicators in COI 3.0 include measures of social capital, wealth, broadband access and density of local nonprofits. See a full table of COI 3.0 indicators and their definitions.
Updated methodology
Changes to COI 3.0 go beyond the addition of new indicators. In constructing COI 3.0, we changed the sources of some component indicators to improve measurement quality and facilitate more timely updates. We also revised methods used to process component indicators and to construct the composite index. We relied on new tools, such as embeddings published by OpenAI, to classify food retailers and on machine learning algorithms to improve data quality and predictive validity. We added improved measures for school quality and access to green space, and we introduced new subdomains, including those that capture safety- and health-related resources, housing quality, wealth and socioeconomic inequity. Because of these changes in its construction, COI 3.0 data cannot be compared to COI 2.0 data. Nevertheless, COI 2.0 and 3.0 are very highly correlated (Pearson’s rho = 0.91).
Another major change in COI 3.0 is in the way the composite index is constructed. For COI 2.0, we grouped component indicators into three domains (education, health and environment, social and economic). We then constructed the overall index by first averaging indicators within these domains into three domain scores, and then averaging the three domain scores into the COI composite index. For COI 3.0, we group similar indicators into 14 subdomains, and then compute the overall COI composite index and the three domain scores from these subdomain scores. By introducing subdomains, we enable users to perform more fine-grained analyses with subdomain metrics (e.g., for early childhood education, elementary education, secondary and post-secondary education, educational resources) in addition to the broader domain metrics (e.g., education).
Built at the census block level
COI 1.0 and 2.0 were constructed for census tracts (2010 Decennial Census definition). COI 3.0 is constructed at the census block level, for both 2010 and 2020 Decennial Census blocks. Constructing COI 3.0 at the block level allows us to compute precise COI estimates at higher geographic summary levels—not just census tracts, but also ZIP code tabulation areas, school districts, congressional districts or cities, for example. COI 3.0 is currently publicly available for 2010 census tracts only.
More predictive of adult outcomes
Our validation analyses, available in our Technical Document, show that COI 3.0 is moderately to highly correlated with other leading composite neighborhood indices—but that it outperforms these indices (including COI 2.0) both in terms of predictive validity and equity validity.
These results suggest that COI 3.0 is well-suited for equity-focused research and applications. COI 3.0 will often reveal neighborhood effects and inequities that are as large as or larger than those found using similar composite indices. Policymakers and program administrators will find COI 3.0 to powerfully measure local structural inequities across neighborhoods that are highly predictive of long-term health and socioeconomic outcomes.
Learn more about the Child Opportunity Index 3.0
Read how we built the COI 3.0 in our Technical Document
