Business studies that probe how one factor influences another over time are often vulnerable to a subtle kind of error. Rigorous tests are therefore needed to avoid the drawing of false conclusions, as explained in a paper by HKUST’s Jiatao Li and mainland colleagues. They show how a long-standing problem in econometrics may have compromised the results of numerous studies—and recommend methods and guidelines to address this issue in future.
Academics routinely study cause-and-effect by measuring how differences in one quantity, known as the independent variable, are related to measurable changes in another quantity, the dependent variable. For example, finance researchers might investigate how variations in a character trait of CEOs correlate with the investment behavior of their firms. Some measurement of this trait will be input into a mathematical model together with investment data.
Li and colleagues studied how one effect—dynamic endogeneity—can bias the results of such studies in international business research (IBR). Ideally, the independent variable is relatively fixed, like the CEO character-trait example or a historical fact. However, in some cases it might not be truly independent. “For instance,” write the authors, “a study examining how a shared language between headquarters and a subsidiary affects the flow of knowledge between these two entities is at risk of dynamic endogeneity.”
Such situations violate the exogeneity assumption, namely that “current observations of the independent variables be completely independent of the previous values of the dependent variable.” In the IBR context, this assumption is particularly risky in some research domains that are based on comparing multiple time periods. If past values of the dependent variable affect current values of the independent variable, any inference about causality between them may be spurious. Practical recommendations on this basis could then be critically flawed.
The problem can be addressed by choosing an appropriate model. Unfortunately, after reviewing 80 IBR studies published over two years, the authors found that most of them had not clearly acknowledged the problem in the first place. Even when dynamic endogeneity was recognized, it was not adequately mitigated: “Among the 12 papers in which there is any mention at all of a potential problem, just three applied a dynamic panel model and only one of those used the estimator in a rigorous way.”
The study recommends a sophisticated model—the system generalized method of moments (GMM) estimator—for research in which the independent variable is likely to be affected by the historical dependent variable. Li’s simulations using econometric data showed that the system GMM estimator outperformed traditional models in reducing dynamic endogeneity: “Regardless of the size of the … effects or of the structure of the panel, the fixed effects estimator is always severely biased, whereas the system GMM estimator is not.”
The authors then showcased the system GMM estimator in an example study of the link between managers’ international experience and their firms’ internationalization. Finally, they provided an easy-to-follow checklist on spotting the risk of dynamic endogeneity and judging whether to use the system GMM estimator or another tool. This comprehensive study will be invaluable for IBR researchers who aim to meet the highest standards of validity for their conclusions.