Abstract
Diversity faultline strength – the extent to which diversity attributes within a group converge in such a way that they split a group into homogeneous subgroups – can decrease group performance. Based on the categorization–elaboration model (CEM) of workgroup diversity, we assumed that task motivation can overcome the detrimental effects of faultlines. We further assumed that this effect is contingent on the groups' diversity beliefs and that it is mediated by the elaboration of task-relevant information. Forty-three student groups worked on a computer-simulated complex problem solving scenario. The group faultline strength was calculated over the attributes gender, age and educational background. Information elaboration was elicited by means of behavioural coding. Results revealed a three-way interaction of faultline strength, diversity beliefs and task motivation: groups overcame the detrimental effects of faultlines only if they exhibited pro-diversity beliefs and high task motivation. The three-way interaction of faultline strength, diversity beliefs and task motivation was partly mediated by information elaboration. Practical and theoretic implications are discussed.
Notes
1 Both the Fau measure by Thatcher, Jehn, and Zanutto (Citation2003) and Shaw's FLS (2004) measure are based on the concept for faultline strength calculation proposed by Lau and Murnighan (Citation1998). As both Fau and FLS exhibit high correlations with Lau and Murnighan's (Citation1998) example values, they measure the same construct. As the Fau measure has been employed successfully in previous studies (Bezrukova et al., Citation2009; Zanutto, Bezrukova, & Jehn, in press), we employed the Fau measure in this study.
2 We computed all of the presented regressions a second time with affiliation as a further control variable. This did not change the pattern of the reported results; all main effects and interactions remained at the reported levels of significance. For reasons of parsimony, we thus refrained from including affiliation as a further variable in the reported analyses.
3 Note that when employing non-standardized (i.e. non-z-transformed) variables and control variables as in the current case, the scale of the y-axis may be incorrect, but the overall pattern of the interaction remains correct (Dawson & Richter, Citation2006).