Can Weighting Compensate for Sampling Issues in Internet Surveys

While Internet surveys have increased in popularity, results may not be representative of target populations. Weighting is commonly used to compensate for sampling issues. This article compared two surveys conducted in the Netherlands—a random mail survey (n = 353) and a convenience Internet survey (n = 181). Demographic characteristics of the samples were weighted by three variables (sex, current residence, age) using Census data. Prior to weighting, the mail sample approximated the population in half of the joint distributions formed by the weighting variables. The Internet sample differed from the population on all 12 cell-by-cell comparisons and no respondents were over age 65. After weighting, the two samples yielded different estimates for non-weighting variables. The Internet sample overrepresented those in the highest education category and appears to have overrepresented those who are ambivalent toward wildlife. Caution is advised when generalizing results from open access Internet surveys.