Socioeconomic Status Index Methodology

The Socioeconomic Status (SES) Index was developed by Common Ground Health and calculated using income, education, and occupation indicators from the U.S. Census Bureau’s American Community Survey (ACS). Each ZIP code and census tract is assigned a SES ranking from 1 (low) to 5 (high). The lower SES areas tend to have lower income levels, lower educational attainment, and lower employment rates in high status jobs. The ACS data used for the current version of the SES Index is the 5-Year Estimates for 2016 - 2020.

We used a classical definition of socioeconomic status with three conceptual components: education, income, and occupation. Together, these dimensions provide a comprehensive view of social and economic standing.

The methodology described below was implemented for all areas of New York State except the 10 counties in the New York City metropolitan area. 

The NYC area was excluded because some of the socioeconomic indicators are very different than the rest of the state and we didn’t want the NYC data to skew the index in a way that made it harder to see distinctions across the rest of the state. These are the excluded counties: Bronx, Kings, Nassau, New York, Putnam, Queens, Richmond, Rockland, Suffolk, Westchester

Across the state, specific ZIP codes and tracts were excluded from the analysis if there were substantial gaps in the availability of ACS data.

For each of the three components—education, income, and occupation—we identified data available at ZIP code and tract levels. We chose American Community Survey (ACS) data because it offers the necessary geographic granularity and is updated regularly. A wide range of ACS metrics was considered, and we examined their distributions, completeness, and intercorrelations before final selection of variables for the model.

We used Principal Components Analysis (PCA) to synthesize the eight metrics into a single ‘score’. PCA is a dimensionality reduction technique that identifies the most important patterns from the correlated input variables and condenses them into one index, capturing maximum information while reducing redundancy. This approach provides a clearer, more accurate summary than simply averaging metrics. This methodology was applied separately to both the ZIP-level and tract-level data sets.