Small-area analyses using public American Community Survey data: A tree-based spatial microsimulation technique

Abstract

The American Community Survey (ACS) is the largest household survey in the United States and indispensable for detailed analysis of specific places and populations. This paper introduces a technique to produce “small area” (e.g. census tract) estimates of any person- or household-level phenomenon that can be derived from ACS microdata variables. This is demonstrated by producing novel, tract-level estimates of 1) excess housing capacity, 2) prevalence of traditional living arrangements, and 3) household energy burden. We combine conventional spatial microsimulation techniques with binary-split decision trees to efficiently select local population margins from a large set of candidates. The result is place-specific microdata samples that are calibrated to match an information-rich set of known constraints (e.g. number of households by income group). A validation exercise indicates agreement between model output and known values (mean R2 = 0.78). We conclude by discussing potential extensions of the technique to derive small area estimates of variables observed in surveys other than the ACS.