Weighting Adjustments Can Help with Low Response Rates, but at What Cost to Data Quality?

• Activity Number: 311 - SPEED: Environment and Health, Governmental Policies and Population Surveys, Part 2
• Type: Contributed
• Date/Time: Tuesday, July 30, 2019 : 9:25 AM to 10:10 AM
• Sponsor: Government Statistics Section
• Author(s): Chrishelle Lawrence*
• Companies: U.S. Energy Information Administration

Abstract

The Residential Energy Consumption Survey (RECS) was traditionally an in-person survey, but in its most recent survey cycle, in-person, web, and mail surveys were used for data collection. Planning is underway to consider a web/mail only data collection. Moving from an in-person interview to a web/mail survey for data collection poses many challenges, and one major concern is the expected decline in the response rate. To compensate for low response rates, nonresponse and post-stratification weighting adjustments are often used produce more accurate estimates. In doing this, survey weights become larger and more error is introduced into the estimates. Using results from a national pilot study that was conducted via web/mail, this research will consider the impact varying levels of total and subpopulation nonresponse have on data quality with respect to weights and estimates.