A comprehensive assessment of measurement equivalence in operations management

Abstract

This paper provides a comprehensive framework for treating equivalence both prior to data collection and during subsequent analyses, and assesses the extent to which equivalence is considered in survey research in six leading empirical Operations Management (OM) journals (Decision Sciences, International Journal of Operations & Production Management, International Journal of Production Research, Journal of Operations Management, Management Science and Production and Operations Management). Measurement equivalence of latent variables in survey data is an important condition that should be met in order to meaningfully pool and/or compare data stemming from apparently heterogeneous sub-groups. We assess 465 survey articles from a six-year period from 2006 to 2011 and document these articles in relation to the four main stages of our comprehensive framework: identifying sources of heterogeneity; maximising equivalence prior to data collection; testing measurement equivalence after data collection; and dealing with partial and non-equivalence. We conclude that pooling of data from heterogeneous sub-groups is common practice in OM, but that awareness and testing of equivalence remains limited. Given these findings, we further elaborate the best practices detected in those few OM studies that do address equivalence in some way. We conclude that to improve the quality of OM survey research, authors, editors and reviewers should pay greater attention to equivalence, and we provide a pragmatic checklist of measurement equivalence issues across the four stages.

Keywords:

• survey research
• measurement equivalence
• Operations Management (OM)

Acknowledgements

We gratefully acknowledge the methodological support of Willem Saris. In addition, we thank Sinéad Roden, Michael Lewis, Benn Lawson, Brian Squire, and Carey Queenan for their helpful comments on earlier versions of this paper.

Notes

1. Please note that 22 of these papers did not explicitly state that they addressed translation equivalence, but these papers use data collected through multi-country research collaborations (IMSS, HPM and GMRG) for which translation equivalence actions have been described in other references.

2. In extremis, each individual respondent can be considered to have a unique response slope (Schwartz 2007). Consequently, data on his human values scale, which is widely employed in cross-cultural research, are commonly ‘centred’ by respondent. We thank one of the reviewers for raising this point.