A Rasch analysis of the psychometric properties of the High Potential Trait Indicator (HPTi) using a South African sample

Abstract

Psychometric assessments commonly form part of the decision-making process in various industries. The High Potential Trait Indicator (HPTi) is one such psychometric assessment; a six-trait personality-based questionnaire designed for targeting senior leadership positions. The psychometric tool began its development in the UK by Ian MaCrae and Adrian Furnham (2014; 2016), and has since been utilized worldwide. Notwithstanding the history of South Africa and the use of psychometric instruments in the country, such tools are generally required to possess scientific evidence of being reliable and valid. However, given the history, South African legislation requires that all psychological assessments are found to be valid and reliable in order to be used for, and not limited to, their use in human resource related contexts. A measurement theory, known as Rasch Measurement Theory – and its related analytical techniques, – is regarded as achieving the standards required to be considered the definition of measurement. Therefore, should a psychometric instrument meet certain requirements of the Rasch model, the instrument may be regarded as a tool that meets the definition of measurement. This study, therefore, evaluates the reliability and validity of the HPTi using Rasch analysis techniques. While there are many criteria to evaluate an instrument on within the family of Rasch analysis techniques, this study focused on the unidimensionality and local independence of each subscale, the fit to the Rasch model, person reliability and separation, and differential item functioning (DIF) as key components to determining elements of the reliability and validity of the HPTi. Secondary data, obtained from copywrite holder, Thomas International, was obtained and used for analysis. The initial sample consisted of 1257 respondents, with further reductions taking place for each of the six traits based on the person fit statistics. The analysis of reliability found that one of the six traits, Ambiguity acceptance, would not be considered reliable, with the other five ranging from a reliability index of .70 to .76, where a score of .70 is the minimum required to be considered reliable. An item fit analysis revealed 14 of the 78 HPTi items did not fit the Rasch model. Of the 14, 10 underfit the model (more varied than the expected model) and 4 underfit the model (less varied than the expected model). In analysing the differential item functioning of the instrument, no items were found to be functioning differently between genders, female and male. However, 12 items were evaluated to possess differential item functioning across ethnicities and home language groups. The most severe instance of DIF occurred in trait Competitiveness, yet this scale had only one item experiencing DIF. On the other hand, trait Conscientiousness contained four items experiencing various severities of DIF.