A comparison of aggregation approaches for second-order data

Confusion and misunderstanding over the issues and problems associated with aggregation and second-order data have resulted in unwarranted aversion to, and suspicion of a useful approach for incorporating the perspectives of two parties in an attempt to understand their relationship. In this article, two new approaches are proposed for aggregating individual-level responses to capture the concept of mutuality (e.g,, mutual trust or mutual satisfaction) in dyadic relationships. The conceptual approach involves mapping paired responses, either symmetrically or asymmetrically, into rank-ordered sets. The mathematical approach takes the square root of the product of paired responses. These approaches are compared empirically with the traditional approach of an arithmetic average using both simulated and actual dyadic data. Results suggest that the choice of aggregation approach does have empirical implications, particularly in situations of highly divergent responses. Adopting the approach most consistent with the conceptual meaning of a construct is recommended. Mixing aggregation approaches should be avoided. (C) 1999 Elsevier Science Inc. All rights reserved.