Abstract:
|
Nonresponse bias can be tested by comparing characteristics of respondents who returned completed surveys and non-respondents who failed to return a completed survey. For years, statisticians have proposed theories and findings in regard to nonresponse as well as the statistical models that link non response to bias in estimates. Drawing inference about the population from a sample with non-response may generate biased results. Nonresponse can lead to bias in the estimates even if nonrespondents are similar to respondents in observable characteristics. Thus, the focus of this study was to empirically measure nonresponse and for sole-proprietor non-employers in the current Survey of Business Owners (SBO). Response rates were first examined at subgroup levels to determine areas for potential nonresponse. In this paper, we use data on nonrespondents obtained by using administrative data from the IRS and the American Community Survey (ACS). Using this data, we test for nonresponse bias and illustrate the effect of bias on aggregate benefit estimation.
Keywords: Nonresponse, empirical, bias
|