Addressing endogeneity in operations management research: Recent developments, common problems, and directions for future research

Abstract

Addressing endogeneity can be a challenging task given the different sources of endogeneity and their impacts on empirical results. While premier business journals typically expect authors to rigorously address endogeneity, this expectation is relatively new to many Operations Management (OM) scholars, as exemplified by a recent editorial in Journal of Operations Management that calls for more rigorous treatment for endogeneity. This study serves two purposes. First, we summarize recent OM literature with respect to the treatment for endogeneity by reviewing studies published in leading OM journals between 2012 and 2017. The review provides evidence that endogeneity problems have received increasing attention from OM scholars. However, we also find some common problems that may render the chosen techniques for addressing endogeneity less effective and potentially lead to biased analysis results. Second, since instrumental variable regression is the most prevalent technique for dealing with endogeneity in the OM literature according to our review, we provide an empirical illustration tailored to OM researchers for using instrumental variable regression in the post‐design (data analysis) phase. Using variables from a publicly available healthcare dataset, our analysis sheds light on the importance of examining instruments' quality and triangulating results based on more than one test/estimator.

Emerging Market Footholds and Knowledge: An Examination of New Product Launch Performance

“…For instance, when omitted variables potentially influence the dependent variable and are correlated with one or more explanatory variables, bias can be introduced into the parameter estimates (Lu et al. ). To address this issue, we controlled for the influence of a customer segmentation strategy by using the following item on a 7‐point Likert‐type scale (Ehrgott et al.…”

“…Next, we addressed potential simultaneity issues in our dependent and explanatory variables (Lu et al. ) by asking the respondents to rate knowledge levels prior to the product launch and to rate performance 12 months after the product is launched.…”

A New Perspective on Returns to Scale for Truckload Motor Carriers

“…As using invalid instruments is more problematic than estimation without instruments (Rossi 2014), our inability to find a valid instrument led us to not use an instrumental variable estimator. While we have sought to include theoretically justified controls per best practice (Lu et al 2018), our inability to use instrumental variable estimation is a limitation nonetheless.…”

“…A fourth limitation is that we were unable to identify a valid instrument to evaluate whether carrier size was endogenous in our model. As noted by Lu et al (2018), finding valid instruments can be challenging, especially when working with archival data where authors' selection of measures is limited. As using invalid instruments is more problematic than estimation without instruments (Rossi 2014), our inability to find a valid instrument led us to not use an instrumental variable estimator.…”

Strategic alliance between information intensive services and supply chain integration: impact on firm performance

“…Future studies may be directed to examine the proposed relationship by incorporating other contextual factors, such as firm size, and supply chain structure and complexity, which is a company's position in the supply chain (Lu et al, 2018). It will be intriguing to examine how IIS differs across firm size within an integrated supply chain.…”