Survey response rate calculator
Count your final dispositions and get every AAPOR response rate, so you can name the exact formula you report.
Leave e blank to skip RR3 and RR4. AAPOR requires e to rest on evidence you can state; never pick it to raise the rate. Example counts are for illustration.
What is a survey response rate, and why are there six of them?
A response rate is the share of the cases you tried to reach that ended up as usable interviews. The complication is that a sample contains cases whose status you never learned: a phone that rang unanswered, a letter that may never have arrived. Whether such a case was eligible (a household, an adult, a customer) is unknown, and how you treat it changes the answer. The American Association for Public Opinion Research (AAPOR) therefore defines six response rates in its Standard Definitions. RR1 is the strictest: complete interviews divided by everything, including all unknowns. RR2 also counts partial interviews as responses. RR3 and RR4 count only an estimated share e of the unknowns as eligible. RR5 and RR6 assume no unknown case was eligible, so they are the most generous, with RR6 the maximum. AAPOR asks researchers to say exactly which rate they report, as in “Response Rate 2 was 45%”.
What do the disposition codes mean?
The calculator uses the final disposition codes of the AAPOR 10th edition (2023). I is a complete interview and P a partial interview. R is a refusal or break-off, NC a non-contact and O any other eligible non-interview (for example, the person was away or ill). The unknown group has three codes: UH, unknown whether the household or housing unit is occupied (an address that was never reached); UR, unknown whether an eligible respondent lives in the housing unit; and UO, unknown for other reasons. e is your estimate of the proportion of the unknown cases that were eligible. If you work from an older edition that has no UR code, leave UR at 0.
Worked example: 1,000 sampled cases
Suppose the final dispositions were I = 400, P = 50, R = 150, NC = 100, O = 20, UH = 200, UR = 30 and UO = 50, which is 1,000 cases (these numbers are invented for illustration). RR1 = 400 ÷ 1,000 = 40.0%, and counting the partial interviews as well, RR2 = 450 ÷ 1,000 = 45.0%. If external evidence suggests that half of the 280 unknown cases were eligible, the base becomes 450 + 270 + 140 = 860, so RR3 = 400 ÷ 860 = 46.5% and RR4 = 52.3%. Assuming none of the unknowns were eligible gives RR5 = 55.6% and RR6 = 62.5%. The same data therefore yield response rates from 40.0% to 62.5%, which is why naming the formula matters. The cooperation rate (I ÷ [(I + P) + R + O]) is 64.5%, the refusal rate is 15.0% and the contact rate is 62.0%.
How should I choose e?
AAPOR says plainly that e must be estimated from the best available scientific information about what share eligible cases make up among the unknown ones, that the basis must be stated, and that one must not choose a proportion to boost the response rate. Typical evidence includes the eligibility rate among the cases whose status you did learn, results of extra follow-up on a sample of the unknowns, or known features of the frame such as the share of non-working numbers. When you cannot justify an estimate, report RR1 and RR2, which need none. That is why the calculator leaves e blank by default and shows RR3 and RR4 only when you fill it in.
Is a low response rate always bad?
Not automatically. A low response rate raises the risk of nonresponse bias, but bias depends on whether responders and non-responders differ on what you are measuring, not on the rate itself. Still, a high rate gives less room for that to matter, and many funders and journals ask for the rate to be reported. Compare like with like: a web panel survey and a mailed survey of customers have different typical rates, and online surveys of volunteers have no defined sampling frame, so the AAPOR rates do not apply cleanly to them. For internal surveys, such as an employee questionnaire where the denominator is all employees invited, RR1 is simply completes divided by invited, and this calculator gives the same answer with the other codes set to zero.
What are the cooperation, refusal and contact rates?
They describe where the non-response came from. The cooperation rate is the share of eligible cases that were reached and agreed to an interview, so it ignores the cases you never contacted. The refusal rate divides refusals and break-offs by the full base used for RR1. The contact rate is the share of all cases in which someone was reached, whether or not they completed the interview. A survey can have a poor response rate because of a bad contact rate (wrong numbers, inaccurate addresses) even when cooperation is good, and the fix is different from a refusal problem. AAPOR defines several versions of each; the page shows the first.
Common mistakes
Counting only the people who answered the phone or opened the email shrinks the denominator and flatters the rate; AAPOR’s RR1 and RR2 keep every unknown-eligibility case in the base. Mixing codes is a second trap: a refusal belongs in R, not O, and a break-off counts as a refusal. Reporting “the response rate” without the number of the formula is a third, since the same data can give rates that differ by more than 20 percentage points, as the example above shows. Finally, if the sample was weighted or selected with unequal probabilities, the standard definitions call for weighted counts, which this calculator does not handle.
For general information only. Sources: The American Association for Public Opinion Research, 2023 Standard Definitions: Final Dispositions of Case Codes and Outcome Rates for Surveys, 10th edition, section on outcome rates (pp. 85–88). The 9th edition (2016) was not opened. Related: margin of error calculator, sample size calculator and Likert scale calculator. How we check our calculators.