Imputing Proxy Advisor Recommendations

Abstract

Proxy advisor recommendations play a central role in research on shareholder voting and corporate governance, yet they have become largely unavailable to academics. I develop a method to impute benchmark recommendations from Institutional Shareholder Services (ISS) and Glass Lewis using publicly available institutional voting data and estimated investor “follow rates”—the investor's likelihood of voting in line with each advisor. The method applies Bayes' theorem to infer recommendations from observed votes and iteratively updates both follow rates and imputations over successive rounds. To improve performance on contested proposals and avoid systematic bias, I estimate follow rates that vary by context—such as whether management and the other advisor agree with the recommendation. Validation against actual recommendations shows high accuracy: 96.4% of ISS recommendations imputed with 99.6% accuracy and 90.8% of Glass Lewis recommendations with 99.0% accuracy. Coverage improves substantially over prior approaches, especially for hard-to-classify proposals. I provide the full dataset of imputed recommendations as an Online Appendix for academic use.