Likert scales and questions: uses and abuses

Abstract

when conducting business and management research, the most common strategy employed for collecting primary data is either interviews or questionnaires (or both). When questionnaires are used to collect primary data a very common approach used is to construct questions which can be called Likert-type questions or Likert scales. Numbers are allocated to the question responses of each question.. Often, this is followed by performing arithmetic or statistical operations on the allocated numbers. In many cases encountered by the author, the analysis has even included techniques such as parametric statistics and factor analysis. This paper explains why such simplistic approaches are completely invalid and should never be used. It goes on to explain how analysis can be done whilst avoiding typical hazards. Often, the writers of such papers do not understand or explain levels of measurement. Of course, Likert scales are at an ordinal level of measurement which would normally preclude the use of arithmetic, statistical, factor analysis techniques there is an additional problem of reliability because different people will interpret terms such as strongly disagree and disagree, differently. Likert was aware of these problems when writing the original paper in 1932 although at that time the term “levels of measurement” was not in use. This paper provides approaches and suggestions for avoiding the problems of data analysis when using Likert-type questions. This paper should be of assistance to those who intend to use Likert-type questions in the questionnaire.