Is a picture worth a thousand words? Image usage in ESG reports
Disen Huang Albrecht, Divya Anantharaman, Keyi Zhao
Abstract
We collect a novel dataset on images in Environmental, Social, and Governance (ESG) reports using deep learning and artificial intelligence (AI), to test hypotheses arising from experimental literature that visual impression management enhances audience perception. We further combine AI and manual methods to contrast images that lack information content (“generic images” evoking broad ESG themes) against images with information content (“specific images” depicting firm-specific ESG activities). We hypothesize and find that firms (1) operating in socially problematic industries (e.g., “sinful”, polluting, or controversial industries), (2) issuing less extensive textual ESG disclosures, and (3) experiencing poor ESG performance tend to use more images, especially generic images, consistent with strategic motivations. We then examine the impact of image usage on stakeholders, observing unduly enhanced perception in retail traders but seldom in institutional investors. We observe that ESG award committees reward specific images and penalize generic images. Our results show that ESG report visual impression management affects stakeholders in the real world, though its impact can be overcome by attention and expertise.