Scale reduction techniques for email and web-based surveys

27 Mar 2015

Surveys have been used by research communities for many decades to collect data, particularly in the social sciences. The advent of the world-wide web has facilitated a proliferation of surveys both for scientific and market research. The main tool for conducting surveys has been questionnaires, mainly because they are relatively easy to administer and can efficiently gather sufficient data at low cost. In the current era, both email and web-based questionnaires have been firmly entrenched as a tool of choice for researchers. The ease of being able to administer questionnaires using Internet protocols has motivated postgraduate students to incorporate questionnaires in their research design. However, the question that arises concerns the appositeness of questionnaire design techniques as well as issues pertaining to the length of electronic questionnaires. The authors, having conducted an analysis of a sample of master?s studies at a post-graduate centre of a South African University, have found that scant regard is given to proper survey instrument development technique. This is applicable to questionnaires that extensively utilise Likert scales. While most students have piloted their questionnaires, they have not really explored the statistical options available to refine the questionnaire at this stage of design. This article aims, first, to highlight concerns pertaining to electronic survey design, and to stimulate debate on the use of statistical analyses during instrument development. The primary objective of the article is to report on three web-based and emailed surveys that were conducted by postgraduate students. The authors use these three studies as case studies to demonstrate how, by application of statistical techniques for scale reduction, questionnaires can be refined and prepared through a pilot process. In particula, the article describes the use of item-analysis as a tool to prepare questionnaire instruments after an initial pilot phase. The article further demonstrates that if item-analysis were used in these three cases, the results of the surveys could have been enhanced.