413 – Surveys of Children, Students, and Schools
Research and Development for Methods of Estimating Poverty for School-Age Children
Sam Hawala
U.S. Census Bureau
Ciara Nugent
U.S. Census Bureau
The American Community Survey (ACS) does not publish single year estimates for areas with a population of less than 65,000. The majority of school districts in the United States contain fewer than 2,000 relevant school-age children. As a result, direct survey estimates for school districts are often highly variable and unreliable due to small sample sizes. This paper looks at two different approaches to modeling that address the problem of small samples in small areas. These small sample sizes can also cause a high frequency of zeros estimates. First, we look at a reweighting method proposed by Schirm and Zaslavsky (1997). Secondly, we look at a censored Fay-Herriot model, similar to that in Slud and Maiti (2011). This model takes up the specific case of no observations of poverty in the ACS sample, resulting in a zero estimate: a problem for over 25% of school districts.
