Improved estimates of poverty of place for the United States: A county-level Supplemental Poverty Measure

• In Event: Why'd We Have to Make Things So Complicated? The Case for a Simpler, More Transparent SPM

Abstract

Properly capturing poverty across space in the United States remains challenging. Although national estimates of poverty using modern and robust measures are available, when working at the sub-state level researchers are forced to rely upon the Official Poverty Measure (OPM) of the United States—a measure widely viewed as problematic and limited. The Supplemental Poverty Measure (SPM), developed in response to the shortcomings of the OPM, has increasingly supplanted the OPM within research and policy analysis. However, the SPM is functionally tied to the Current Population Survey and its publicly available geographies, meaning that sub-state estimates of the SPM (e.g. county) are unavailable. In this paper, we present new county-level estimates of the SPM using restricted access version of the American Community Survey and Current Population Survey. To generate our estimates, we follow the procedure used to create the ACS-SPM by the Census Bureau as closely as possible within the FSRDC. The key challenge with producing a county-level SPM is the fact that far more data on household income and expenditures is needed to calculate the SPM than is collected in the ACS. The CPS-ASEC, at best, allows researchers to generate estimates at the state level by metropolitan and nonmetropolitan status. It is not possible to generate smaller estimates with the limited sample of the survey. Thus, the ACS—which replaced the long-form Census in the 2000s and operates on a five-year nationwide sampling protocol—is the only option. To overcome the challenges imposed by this limitation, the Census Bureau-produced ACS-SPM uses the CPS to impute values within the ACS. 

In our final paper, we will use 2008-2023 restricted-access CPS-ASEC and ACS data within the FSRDC to: first, impute variables needed to construct the SPM that are in the CPS but not in the ACS using 5-year pooled CPS data; and second, generate SPM poverty estimates at the county-level. Finally, we will validate our SPM constructed from ACS data by comparing the demographic composition of the population of people determined to be in poverty by the ACS SPM we construct to the population of people determined to be in poverty based upon the Census’s SPM; to do this, we use demographic breakdowns of poverty across gender, race, geographic regions, etc. presented in yearly SPM reports.  Beyond providing poverty estimates at a scale previously unavailable, we also address another limitation of the SPM related to its coarse geographic adjustment for cost of living by introducing a new geographic adjustment at the scale of the commuting zone. In the remainder of the paper, we assess the validity of our estimates by comparing our estimates to previous estimates and other measures. Overall, the ACS-SPM we create within the Federal Statistical Research Data Center and present in this paper offers an essential data product for accurately assessing the impacts of policy at the local level, particularly in rural areas.

Author

• Darcy Lee Sullivan, University of Kansas
  Presenting Author