Estimating Postcensal Household Size for NYC’s Neighborhoods

While the decennial census provides a full count of people in households once every ten years, small area population estimates from the American Community Survey (ACS) have proven unreliable for postcensal time points. The housing unit method, however, can be used to estimate the population: Administrative data on net new construction in the post-census period can be used to estimate the housing stock, and then combined with estimates of household size and occupancy levels to create population estimates for postcensal time points. However, a major stumbling block is the unavailability of reliable household size measures between censuses. Typically, housing unit method users default to data on persons per household (PpH) from the previous census, despite changes in PpH that may have occurred at the local level. Building on prior regression-based PpH research at the county level, we develop postcensal estimates of household size at a sub-county level. We first employ a regression-based model that predicts PpH in 2010 controlling for lagged PpH in 2000 and 2000–2010 changes in symptomatic indicators, which include housing stock and birth rates from local administrative data, and socio-demographic variables from the ACS. The regression coefficients are then applied to symptomatic changes from 2008–2012 to 2013–2017 to produce more recent and reasonable household size estimates for New York City’s neighborhoods. To the best of our knowledge, this is the first paper that employs a lagged model to predict PpH at the sub-county level, using a combination of administrative and 5-year ACS data.