Monday, Feb 1 at noon, 6050 ISR-Thompson
Peytcheva, E., and Robert M. Groves. 2009. "Using Variation in Response Rates of Demographic Subgroups as Evidence of Nonresponse Bias in Survey Estimates." Journal of Official Statistics, 25(2): 193-201.
Survey researchers frequently compare the demographic distributions of respondents to the corresponding distribution from the target population (often using census data). Many researchers hope that if the respondents in their study have demographic characteristics similar to those of the target population, their research findings do not suffer from nonresponse bias. Hence, it is useful to ask the question whether the nonresponse biases of demographic statistics are predictive of biases in key substantive statistics of the same surveys. A meta-analysis of 23 methodological studies designed to estimate levels of nonresponse bias in statistics of interest was conducted. The data from the 23 studies suggest that the difference between respondent and nonrespondent means for demographic variables is not predictive of the difference between respondent and nonrespondent means for substantive variables of the same survey.