Innovative Uses of Response Rates for Survey Management

• Activity Number: 585
• Type: Invited
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Government Statistics Section
• Author(s): Amar Mann*+ and Tian Luo
• Companies: Bureau of Labor Statistics
• Keywords: Survey Management ; Responsive Design ; Econometric Modelling ; Predictive Analytics

Abstract

We will present approaches for predicting non-response and assessing non-response bias in the Bureau of Labor Statistics' (BLS) surveys. Exploratory data analysis gives us insight into what variables are related to response and are used to develop classifiers to predict response. The pros and cons of various classifiers such as decision trees, logistic regressions, linear discriminant analysis (LDA), and quadratic discriminant analysis (QDA) will be assessed using BLS survey and administrative data. Model- and non-model based criterion and penalization are explored and used to select a model that optimizes prediction. The outputs of the analysis are dynamic survey-management tools, which help improve the efficiency of collection as well as data quality.