Asian researchers should be more critical: The example of testing mediators using time-lagged data

Kenneth S. Law, Chi Sum Wong, Ming Yan*, Guohua Emily HUANG

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

10 Citations (Scopus)

Abstract

In the past decade, there has been call for Asian researchers to be more confident and not limit themselves to follow only the footsteps of Western studies. In this paper, we follow up the discussion in Western literature about the importance of testing mediators with longitudinal data. The prevailing way of testing mediation is the use of time-lagged models. That is, the predictor or mediator is collected at prior time points than the outcome variable. We believe this is not sufficient. Instead, cross-lagged models, which measure all three types of variables at different time points, are necessary for testing mediation. Unfortunately, Asian researchers have again followed the footsteps of the suboptimal practice of time-lagged models. Using computer simulation data and a real-life dataset collected in China, we show that erroneous conclusions may be drawn even when the predictor, the mediator, and outcome variables are measured at different time waves under the time-lagged model. We propose a more appropriate procedure to use the cross-lagged model to test the exact causal ordering among the predictor, the mediator, and the outcome variable.

Original languageEnglish
Pages (from-to)319-341
Number of pages23
JournalAsia Pacific Journal of Management
Volume33
Issue number2
DOIs
Publication statusPublished - 1 Jun 2016

Scopus Subject Areas

  • Business and International Management
  • Economics, Econometrics and Finance (miscellaneous)
  • Strategy and Management

User-Defined Keywords

  • Asian management research
  • Cross-lagged model
  • Mediation
  • Organizational socialization
  • Time-lagged model

Fingerprint

Dive into the research topics of 'Asian researchers should be more critical: The example of testing mediators using time-lagged data'. Together they form a unique fingerprint.

Cite this