An Evaluation of Nonresponse Bias in the Business R&D and Innovation Survey

• Activity Number: 339
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Government Statistics Section
• Author(s): Richard Hough*+ and Brandon Shackelford
• Companies: U.S. Census Bureau and Twin Ravens Consulting
• Keywords: Non Response Bias ; R&D ; Innovation ; Business Surveys

Abstract

The Business R&D and Innovation Survey (BRDIS) is a sample survey that produces estimates on R&D and innovation conducted in for-profit companies operating in the U.S. The BRDIS is conducted under a joint partnership agreement between the National Science Foundation and the U.S. Census Bureau and uses the Census Bureau's business register as its frame, from which it samples over 40,000 companies annually. Unit response rates to BRDIS have declined since its inception in 2008 from 74% to 68% in 2011. In 2012, a new version of the survey that was significantly shorter than those used in prior years was mailed to the majority of sampled companies. This paper will analyze the impact of this shorter form on response and use its introduction as an opportunity to evaluate nonresponse bias in BRDIS using both survey data as well as para data from the Census Bureau business register. This analysis will examine the characteristics of the non-response population within two subsets of the survey sample: companies for which information about R&D performance is known prior to sample selection and companies for which no information about R&D performance is known prior to sample selection.