Augmenting Traditional Estimation with Nondesigned Data: Application to the U.S. Unemployment Rate

• Activity Number: 369
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Robert Montgomery* and Martin Barron and Nicki Dunnavant and Yongheng Lin and Ilana Ventura
• Companies: NORC at the University of Chicago and NORC at the University of Chicago and NORC at the University of Chicago and NORC at the University of Chicago and NORC at the University of Chicago
• Keywords: big data ; social media ; non-designed data

Abstract

One of the key emerging questions in the use of Big Data, Social Media, and other forms of non-designed data is how to best leverage their strengths in producing generalizable population estimates. Many researchers have attempted to employ these data to replace traditional estimation methods, which in some cases has led to embarrassing failures. We advocate an approach that combines both traditional design-based data with Social Media and other data to enable better estimation at lower cost. As an example, we share our work exploring the use of non-designed data in the estimation of the US unemployment rate. We will discuss the challenges of using non-designed data, our particular data sources, some models we have estimated, and how these techniques may be used to enhance the utility of the source data.