![IconGems-Print](images/IconGems-Print.png)
179 – Combined Data (Surveys + Administrative Data, etc.)
Bayesian Decision Theory to Optimize the Use of Administrative Records in Census NRFU
Yves Thibaudeau
U.S. Census Bureau
Darcy Steeg Morris
U.S. Census Bureau
Morris, Keller and Clark (2016) show how to identify reliable administrative records for enumerating the occupants of housing units in the context of the U.S. Decennial Census Non-Response Follow-up. We propose using Bayesian decision theory to extend the approach of these authors and account for costs and response propensities of field follow-ups. We elicit a loss function that emphasizes the importance of a correct enumeration for each unit. We exploit the properties of the loss function to make decisions between conducting new field follow-ups and utilizing administrative records to complete an enumeration. This leads to a general Bayesian decision theory problem. We attempt approximating the Bayes (optimal) solution of this problem through a version of "backward induction" (DeGroot 1970; Brockwell Kadane 2003). We give explicit formulas applicable to specific situations and derive possible strategies.