The Usefulness of the Two-Step Normality Transformation in Retesting Existing Theories

The Two-Step normality transformation has been shown to reliably transform continuous variables toward normality. The procedure offers researchers a capable alternative to more prominent methods, such as winsorization, ranking, and power transformations. We demonstrate its utility in the context of the Productivity Paradox literature stream, which is renowned for inconsistent results. This paper demonstrates that the Two-Step normality transformation, which has not been used in Productivity Paradox research, may produce greater goodness-of-fit and affect theoretical understandings on the topic. We use a classic Productivity Paradox dataset to show that compared to the prominent normality transformations, the Two-Step produces unique findings, including 1) regression coefficients more closely resembling the original data, 2) different effect sizes and significance levels, and 3) strengthening evidence for fundamental theories in Productivity Paradox literature. We demonstrate results that challenge uncertainties about the relationship between IT investment and firm performance. Our results imply that the Two-Step procedure should be considered a viable transformation option in future information systems research.