Abstract:
|
In 2011, the Director the Census requested researchers look into the application of small area estimation for a small area/small domain problem relating to provisions under the Voting Rights Act using ACS data. A small area model was developed on short order leaving room for improvement and refinement of methods. For 2016's application there were additional challenges. Primarily that the Census, previously used as a control on fine race and ethnicity information, is no longer a directly useful proxy. Secondarily, the methods utilized previously exhibit issues related to assumptions in regards to the use of design-effects in approximating likelihood functions and the lack of useful regression forms. This paper seeks to demonstrate two approaches to the overall problem and to establish the respective utility of the presented methods. Specifically, their performance with regards to five-percent and ten thousand person language minority citizenship limited-English proficiency criteria.
|