With dwindling participation rates, selective unit nonresponse has become a major issue in the analyses of mail, telephone and web surveys. This is especially true when the main research question of the survey is closely related to situations that can have a direct in uence on the well being of the population.

The GES study is an ongoing monitoring program that studies the impact of Amsterdam Schiphol Airport on the health and well being of the population in the surrounding areas. Over the last decades the situation around the airport has changed considerably. While the number of aircraft has increased dramatically, the noise emitted by each aircraft has decreased due to the phasing out of noisier aircraft. The growth of the airport has led to the installation of a new runway in 2002. Over the years, the growth of the airport has sparked strong opposition from residents and the airport is almost constantly on the policy agenda. In this volatile situation, three postal questionnaire surveys have been carried out since 1996 to assess, among other topics, annoyance and sleep disturbance due to aircraft noise exposure. With participation rates of approximately 50% there was ample possibility for selective nonresponse bias. To assess this in uence, a subsample of nonresponders was contacted by telephone or visit to the house to answer a subset of critical questions from the main questionnaire. In 2002 we were able to reach 83% of the nonresponders in the subsample. The paper will discuss the in uence of selective unit nonresponse on the dierent study outcomes and the methods that were used to correct for this in uence.