Abstract:
|
Regulations state that any U.S. government survey with an overall unit response rate of less than 80 percent shall conduct a nonresponse bias analysis. The Quarterly Financial Report (QFR) falls below 80 percent unit response; therefore, QFR staff performed the first nonresponse bias analysis for this survey since 2006. First, regression analyses are performed upon response rate measures to show which industries have abnormally low response and should be targeted for follow-up. Then, sampling frame data is used to perform equivalence of means tests to determine whether nonrespondent and respondent data differ across various crosstabs, and to calculate relative bias. This analysis is generalizable to many surveys across the U.S. Census Bureau and other agencies, and is comprehensive in that it addresses both sources of nonresponse bias and bias magnitude.
|