Home Statistical Dictionary Response Rate
Response Rate
The proportion of units originally selected into a sample that ultimately provide a completed response, a widely reported (though imperfect) indicator of a survey's potential exposure to nonresponse bias.
In Plain English
Response rate is simply the fraction of the people you tried to survey who actually completed it, out of 1,000 people contacted, if 400 respond, that's a 40% response rate. It's one of the most commonly reported numbers in survey methodology because it's easy to calculate and gives a rough sense of how much room there is for nonresponse bias to creep in, but a high response rate doesn't guarantee an unbiased survey, and a low one doesn't guarantee a biased one, what actually matters is whether the people who didn't respond differ from those who did, on the specific thing being measured, which the response rate alone can't tell you.
Definition
Response rate is the ratio of completed (or, under some definitions, completed plus partial) responses to the total number of eligible units originally sampled, , expressed as a percentage. The American Association for Public Opinion Research (AAPOR) publishes several standardized response-rate definitions (RR1 through RR6) that differ in how they treat partial interviews and units of unknown eligibility, reflecting the fact that 'response rate' is not a single unambiguous quantity and comparisons across surveys require checking that the same definition was used; while response rate is a widely reported and easily interpretable survey quality indicator, it is only a proxy for the risk of nonresponse bias, not a direct measure of it, since the actual bias depends on how different respondents are from nonrespondents on the outcome of interest, a quantity the response rate alone does not reveal.
Formula
Notation
Properties
- Response rate should be interpreted as an upper bound on concern, not a direct measure of bias, the actual nonresponse bias depends on the product of the nonresponse rate and the respondent-nonrespondent gap on the outcome, so a low response rate with similar respondents and nonrespondents can still produce accurate estimates, and a high response rate with very different holdouts can still be badly biased.
- Response rates have declined substantially across the survey research industry over recent decades, particularly for telephone surveys, a well-documented trend that has driven increased reliance on multi-mode designs, incentives, and more intensive follow-up protocols to maintain data quality despite lower response.
- Because 'response rate' can be computed several different, standardized ways (AAPOR's RR1 through RR6 differ in how partial interviews and cases of unknown eligibility are counted), comparing response rates across different surveys or organizations requires confirming both used the same definition, otherwise the comparison can be misleading.
At a Glance