# Psychometric evaluation of the self-undermining scale in South Africa using the Rasch model

## TL;DR

This study evaluates a scale measuring employee behaviors that hinder job performance in South Africa using a statistical model.

## Contribution

The first Rasch model-based validation of the self-undermining scale in South Africa, highlighting its limitations and suggesting improvements.

## Key findings

- The scale showed acceptable reliability with Cronbach’s alpha of 0.77 and person separation of 1.57.
- Most items fit the Rasch model well, but one item needs refinement.
- The scale lacks strong discriminatory power for self-undermining behaviors in this context.

## Abstract

The self-undermining scale is used to assess employee behaviours that undermine job performance, including making mistakes, creating conflict, creating confusion and creating a backlog in work tasks. To date, its psychometric properties have not been thoroughly investigated using item-response theory applications, especially in South Africa. Applying the Rasch Rating Scale Model, this study aimed to investigate the reliability and internal validity of the self-undermining scale, including item fit and rating scale functionality. Data were collected from 318 South African employees using a non-experimental, cross-sectional survey design. The instrument demonstrated unidimensional scaling with Cronbach’s alpha and person separation values of 0.77 and 1.57, respectively. Item and category fit statistics showed satisfactory fit to the Rasch model, with only one item warranting further attention. Some refinements regarding item wording and rating scale optimisation are provided.

This study is the first to investigate the reliability and validity of the self-undermining scale through the Rasch Measurement Model. It also offers cautionary insights into the applicability of the scale to measure self-undermining among South African employees because of the lack of discriminatory power. Recommendations for further validation studies are provided.