Abstract

Noncoverage of nontelephone households is a limitation of random-digit-dialing surveys, because households without telephone service may differ from telephone households on key survey measures. Poststratification is the most common method of compensating for the exclusion of nontelephone households. One alternative approach uses data on interruptions in telephone service from telephone households. Another method uses logistic regression to model households’ propensity to be a nontelephone household. We evaluate the three methods using data from the National Health Interview Survey, a large in-person interview survey conducted by the National Center for Health Statistics. Our results show that the interruption-in-telephone-service method generally has the lowest mean squared error. A variant of the interruption method is suggested for the situation where no independent estimate of telephone coverage is available for the target population.