On Presenting and Interpreting a Response Rate: Lessons from Double Samples of Non-respondents

Good survey practice requires the computation and presentation of the response rate for the realized sample. The response rate is an indicator of the potential bias associated with sample-specific estimates of population parameters, but it is only an indicator, since the extent of bias due to non-response also hinges on the differences in the population between respondents and non-respondents. This latter quantity is oft-surmised but rarely known, and in practice reactivity to response rates is generally a gestalt, with summary assessments on the order of "too low" or "good enough". Following a survey with any degree of non-response, a second (double) sample can be taken from among the initial non-respondents, and used to measure and/or adjust for bias.