Asking questions, analysing answers: relative importance revisited

TLDR

Earlier observations by Holt highlighted the importance of appropriate questionnaire response scale design and the use of the relative importance index (RII) method in construction management research, and the editorial notes that uncertain application of RII has been observed, underscoring its relevance to the field. The paper aims to discuss issues surrounding questionnaire response scale design and the application of RII in construction management research, including contrasting model practicality, comparing outputs, and highlighting limitations. The authors examined several RII models by applying them to a data set, contrasting their practicality, comparing outputs, and highlighting limitations. The study finds that careful design of response scale components is essential, alternative RII models yield similar results, simple percentage models may not produce true percentages, the minimum rating scale integer must match its descriptor, only scales with a zero response or adjusted percentage models achieve unity, and RII outputs should be interpreted cautiously, with no additional literature linking RII usage to construction management research.

Abstract

Purpose – Construction management research (CMR) surveys frequently use questionnaires that employ “Likert-type” items, Likert-items, and Likert scales to capture respondents ' self-reported attitudes. This Editorial builds substantially on earlier observations by Holt regarding the significance of appropriate questionnaire response scale design and, use of the relative importance index (RII) method to analyse those data such scales yield. The paper aims to discuss these issues. Design/methodology/approach – Several RII models are examined by their application to a data set. Objectives include to contrast their practicality in use, compare outputs, highlight limitations of some models and discuss methods of RII data analysis, interrogation and interpretation. Findings – Principal messages are: all components of response scales should be carefully designed; alternative RII models can produce similar results of analysis; “simple percentage” RII models may not yield “real” percentages; the minimum rating scale integer should accurately reflect its descriptor; only rating scales offering a “0” response, or an “adjusted percentage” RII model can achieve unity; and RII outputs should be interpreted cautiously. Research limitations/implications – Uncertain application of the RII has been observed within CMR literature, so implications of the Editorial will be of direct relevance to this research community (and beyond). Originality/value – Aside from that cited above, no additional literature has been identified that specifically links RII usage to CMR.