Experimenting with Pre-Contact Strategies for Reducing Nonresponse in an Economic Survey
- 1. Experimenting with Pre-Contact Strategies for Reducing Nonresponse in an Economic Survey Richard S. Hough, Kenneth Pick, and Alfred D. Tuttle, U.S. Census Bureau 1 Jeri M. Mulrow, National Science Foundation1 Introduction Non-response in economic surveys poses at least as great a challenge as it does in demographic surveys, but the reasons for it are less well understood. Finding the right person within an organization with the necessary authority and access to requested information to facilitate completion of a survey can be very challenging, especially when the survey requires responses from multiple people throughout the company. However, the assignment of responsibility for the survey is one aspect of the survey process that is outside of the control of the survey organization. The Business R&D and Innovation Survey (BRDIS) is a new survey sponsored jointly by the National Science Foundation and the U.S. Census Bureau that measures research and development (R&D) activities of U.S.-based companies. The sample is comprised of approximately 40,000 companies. Of these, approximately 4,000 companies with total R&D costs of at least three million dollars are sent the “long” form (BRDI-1) and the remainder, with total R&D not known or known to be under three million dollars, receive a “short” form (BRDI-1A). The new survey is organized into several topical sections and explicitly instructs the “survey coordinator” to seek the assistance of various types of specialized personnel within the companies. Therefore the selection of a coordinator capable of obtaining assistance from likely respondents throughout the company is critical to accurate and complete response. To maximize the success of the new survey we decided to investigate an alternative strategy for getting the survey to the appropriate respondents, and thus, reducing nonresponse. We conducted a split-sample experiment with historically problematic respondents and non-responders to the SIRD who would be receiving the BRDI-1. In the control treatment, we sent the new survey to the last company contact for the SIRD or to an updated contact from the Census Bureau’s Business Register. In the experimental treatment, we sent a pre-survey letter to a company executive (CEO, president, VP of research, etc.) with a request that s/he provide contact information for an employee capable of coordinating the company’s response to the survey. We will present the overall results of this experiment and compare responses to the pre-survey contact update request as well as to the survey itself. 1 This report is released to inform interested parties of research and to encourage discussion. The views expressed on methodological, technical, or operational issues are those of the authors and not necessarily those of the U.S. Census Bureau or the National Science Foundation.