Beyond Bias: AI as a Proxy Advisor

We evaluate the use of AI as an unbiased proxy advisor for shareholder proposals. When provided with ISS guidelines, our model matches ISS recommendations in 75%  of  proposals and better  predicts  shareholder  support than ISS recommendations alone. Extensive tests confirm this alignment is a result of AI’s rule application rather than lookahead bias. AI-ISS disagreements are more likely when firms disclose a governance consultant, consistent with firm efforts and capacity to influence recommendations. AI recommendations on ES proposals also predict subsequent ES incidents. These findings illuminate the proxy advisor industry’s conflicts of interest and demonstrate AI’s potential to improve transparency.