Published online by Cambridge University Press: 14 September 2018
We investigate the strict-exogeneity assumption, a necessary condition for estimator consistency in many finance panel-data applications. We outline tests for strict exogeneity in both traditional (non–instrumental variable (IV)) and IV settings. When we apply these tests in common traditional finance panel regressions, we find that the strict-exogeneity assumption is often strongly rejected, suggesting large inference errors. We test for strict exogeneity in specific finance panel-data IV settings and illustrate the potential for these tests to help confirm, or rule out, the validity of common panel-data IV estimators. We offer recommendations to address the strict-exogeneity issue in finance research.
We thank an anonymous referee, Bernard Black, Jennifer Conrad (the editor), Todd Gormley, Jeff Wooldridge, and seminar participants at the University of Iowa, the University of Nebraska, the 2015 London Business School (LBS) Symposium on Causal Inference, and the 2016 Financial Research Association Meetings for helpful comments and suggestions. All errors remain our own.
Full text views reflects PDF downloads, PDFs sent to Google Drive, Dropbox and Kindle and HTML full text views.