Narcissism in CEO research: a review and replication of the archival approach

Abstract

Using an unobtrusive archival narcissism measure (Chatterjee and Hambrick in Adm Sci Q 52:351–386, 2007; Adm Sci Q 56:202–237, 2011), CEO narcissism has been linked to important strategic issues: overpayment for acquisitions, risk taking, fraud, etc., but measurement unreliability raises important questions about this research. Six studies (N = 791, comprising 5.3% of CEOs in this review) reported Cronbach’s alphas that met minimum standards (.71 ≤ α ≤ .75) along with other reliability results that were often mixed (e.g. poor test–retest reliability and factor analysis results). However, 37 studies (N = 14,165, 94.7% of CEOs) had unacceptable reliability or failed to report any evidence of reliability at all. Out of 43 studies (total N = 14,956 CEOs), 10 studies (N = 3582) obtained unreported (calculated) Cronbach’s α ≤ .58, 10 studies reported no evidence of reliability or indicator correlations, despite using multiple indicators, and 10 studies justified a measure of narcissism by citing Chatterjee and Hambrick’s (2007) work, but actually relied on only a single archival indicator. Recent studies using the unobtrusive archival approach for CEO narcissism have failed to find acceptable reliability, failed to report low reliability, and have engaged in questionable data transformations. In an operational replication study, data collection procedures were carefully followed, but acceptable results were not obtained in three CEO samples. This article concludes by considering implications for CEO narcissism research and recommendations for future research involving unobtrusive measures.