Estimating bias due to nonresponse in mail surveys

Abstract

A procedure for minimizing nonresponse error in a self-administered mail waterfowl harvest survey was tested on a stratified sample of 3,360 Canada Migratory Game Bird Hunting Permit purchasers in Ontario. On the assumption that follow-ups probe deeper into the core of nonrespondents, a linear regression model for estimating parameter values of the population while correcting for nonresponse bias was devised using cumulated responses over three successive mailings. It was estimated that nonrespondents who tended to have a significantly lower level of participation and involvement in the topic investigated were younger and resided in rural areas of the province. Nonresponse bias was as high as 14.4 percent for waterfowl kill per day of hunting and 11.1 percent for age of hunters. Results confirm the usefulness of follow-ups of nonrespondents as a means of exploring and correcting for nonresponse error. © 1974 Columbia University Press.