Randomization in Surveys with the Halton Sequence

Abstract

Randomization has been widely used in surveys for various purposes such as within household respondent selection, rotation of questions and answer choices, and split sample (or ballot) technique for survey experiments. The randomization is usually based on a random number generating process whereby the computer generates random numbers which are then used to classify respondents in different groups. In this study, we use an alternative randomization based on the Halton sequence. The method is used in a survey with a political experiment which requires randomization of the political candidate’s characteristics. Our survey results demonstrate that the Halton sequence can be quite effective in randomly assigning respondents into groups, especially in surveys with small sample sizes.