Avoiding bias and fallacy in survey research: A behavioral multilevel approach

Much like researchers in general management and organization theory interested in organization-level phenomena, operations management researchers theorize and analyze entire operational systems, such as factories, supply chains, airports, or hospital wards. In discussing threats to measurement validity, we tend to focus on motivational factors such as social desirability, acquiescence, and other self-serving biases. In this paper, I seek to complement these discussions by examining the broader question of how individual-level responses can provide information on system-level traits. In an empirical illustration, I apply multilevel factor analysis to examine the factor structure of organizational commitment in a sample of 2,355 informants embedded within 265 manufacturing plants. The conclusion is that understanding the multilevel essence of theoretical concepts is crucial and that confounding levels may lead to inferences that are not only biased but-much worse-fallacious.