Reducing nonresponse and nonresponse error in a telephone survey: An informative case study

This paper presents the results of an effort to raise the response rate for a national survey of adults aged 40 years or older about a number of issues related to Medicare costs. The survey was initially planned as a list-assisted random digit dialing telephone survey, but when the early response rate experience was low, nonrespondents were sent a paper version of the survey by mail with a $10 incentive and asked to respond. The result was that about half of the telephone nonrespondents for whom we had a mail address returned a mail survey, and the 24 percent telephone response rate was raised to 58 percent overall. Moreover, of the sixteen key estimates on which the survey was focused, seven were significantly different as a result of the mail returns. Our experience demonstrates the potential for mail protocols to improve response rates on national surveys and provides another example of low response rates having important effects on the estimates made from survey data.