Estimating Disadvantaged Populations Based on Public Data

About the report:

This report develops robust estimation models to address data challenges for identifying disadvantaged populations, defined by low-income, racial minority, and limited English proficiency (LEP) status, which aligns with the Virginia Department of Transportation’s (VDOT) SMART SCALE definition. The current reliance on American Community Survey data presents granularity limitations, double-counting risks, and large margins of error, particularly at finer geographic levels.

A custom tabulation from the U.S. Census Bureau, adjusted for noise errors and suppressed data, served as the foundation. The researchers developed three regression models at the census tract level: Basic, Option 1 (including LEP households), and Option 2 (including LEP population). All models achieved high reliability, with adjusted R² values exceeding 0.95 and mean absolute errors of 11% (169 persons) for the Basic model. Option 2 demonstrated superior accuracy, reducing errors to 8% (118 persons) and ensuring estimates for 95% of tracts were within the census margin of error. For Option 2, only 97 out of 2,188 tracts exhibited errors exceeding the census margin of error, averaging 599 persons.

At the block group level, a single regression model incorporated racial minority, poverty, and LEP populations. This model achieved an adjusted R² of 0.97, with a mean absolute error of 15.6% (88 persons). Approximately 90% of block groups met census margin-of-error thresholds, with average discrepancies for outliers at 227 persons.

The findings demonstrate that reliable estimates can be produced using public data, with models reducing error rates significantly. The Option 2 model at the census tract level and the block group model are recommended for statewide application to enhance transportation equity planning. VDOT’s Transportation and Mobility Planning Division can leverage these tools to prioritize SMART SCALE projects effectively, ensuring equitable access for disadvantaged populations.

Authors