Reducing Nontelephone Bias in RDD Surveys

Telephone surveys face the challenge of assessing, and compensating for, bias from noncoverage of nontelephone households. Work by Keeter (1995) suggested that some persons and households who do not have a telephone at a particular time are actually between spells of telephone service. This basic observation, coupled with an estimation strategy proposed in a different context by Politz and Simmons (1949), led to the potential for a statistical adjustment for nontelephone bias in random-digit-dialing (RDD) surveys. This adjustment uses data from sample persons in households that experienced spells of telephone interruption in the recent past. Brick, Waksberg and Keeter (1996) examined several bias-reducing estimates using data on interruptions in telephone service from the National Household Education Survey (1993) and the National Survey of Veterans (1993). Frankel, Ezzati-Rice, Wright and Srinath (1998) applied these and other adjustment methods to two RDD health surveys for the states of Iowa and Washington. Both of these investigations were limited by the inability to compare the adjusted estimates with corresponding estimates that included nontelephone respondents.