^aUniversity of Amsterdam, ASCOR, p.a. Klein Muiden 38, 1393 RL Nigtevecht, The Netherlands. E-mail: robertvoogt@gmail.com

Abstract

Traditional correction methods in survey research are mainly focused on correcting for nonresponse bias. The possible presence of response bias in the data is usually ignored, while past research has given clear indications that response bias as a consequence of stimulus effects and over-reporting caused by answer effects can be a serious problem, especially in election research. In this paper, an integrated method for correcting for bias is developed, in which stimulus effects, over-reporting as a result of answer effects and nonresponse bias are dealt with separately. The results of this integrated correction method are compared with the results obtained by traditional nonresponse weighting, using the 1998 Dutch National Election Study data. Finally, it is suggested how the quality of election study data can be improved and how the correction methods can be optimized.