Research

This paper examines the incentives behind, and the consequences of, firms’ engagement with employee reviews. Using 367,749 reviews and firm responses from 237 firms between 2007 and 2024, together with a survey of 148 HR professionals, I address two questions: whether, why, and how firms engage with and respond to employee reviews, and how current employees (insiders) and prospective employees (outsiders) react to those responses. I apply an information-processing framework to employee reviews, under which firms’ decisions to acquire and integrate the information are unobservable but inferable from the responses they post. I classify the sentences in firms’ responses by whether they could have been written without reading the review (template) or presuppose that the firm read it (tailored), and, within tailored content, by whether they address the situation as it stands or seek further information and commit the firm to act. I find that firms produce more tailored content where employees are less satisfied and where reviews raise previously unsurfaced topics, consistent with processing being undertaken where its return is higher. Conditional on responding, tailored content is associated with subsequent improvements in employee satisfaction while template content is associated with deterioration, and topic-level engagement with the concerns raised in reviews is followed by fewer complaints on the same topic. Survey evidence further indicates that firms escalate review content internally and act on it, as committed in their responses. By contrast, response composition is largely unrelated to labor market attractiveness, with the exploratory exception of responses that publicly contest the reviewer’s account. The findings indicate that employee review platforms can inform managerial decision-making, conditional on firms processing what appears on them rather than replying generically.

We examine the nature and consequences of disclosed corporate reputation-building actions following major violations of environmental and social (labor and consumer safety) laws. We construct a novel dataset of potential reputation-building actions that firms publicly disclose, based on hand-classification of 21,932 press releases surrounding 411 serious violations (e.g., an oil spill or a major labor lawsuit verdict). We find that firms disclose more reputation-building actions after such violations, but the disclosed actions typically do not benefit the stakeholders affected by the underlying violation. Instead, firms’ public disclosures disproportionately emphasize customer- and investor-relevant actions, including customer-oriented actions irrespective of violation type and employee- and shareholder-oriented actions after environmental (but not social) violations. Regarding consequences, increased environmental or social disclosures correlate with lower future rates of recidivism when topically aligned with the subsequent compliance outcome. Finally, the stock market responds positively to disclosed actions targeted at shareholders and customers, but not to those aimed at other stakeholder groups, suggesting that investor responses are, on average, primarily financial rather than broadly prosocial. These findings highlight a divergence between remedial needs and the actions firms choose to highlight publicly, emphasizing reputation management toward financially relevant stakeholders.

Researchers have long sought to understand how firms internally drive value. Management accounting research, in particular, has developed for decades the seminal concept of value drivers, which represent the internal financial and non-financial processes through which firms create value. However, the lack of consistent and publicly available data capturing these internal activities has constrained empirical research from broadly examining firm-level value creation. In this study, we develop new text-based measures to systematically capture firms’ financial and non-financial value drivers by applying a large language model (LLM) to over 318,000 archived snapshots of public U.S. firms’ website homepages (1995–2020) from the Wayback Machine. Leveraging established conceptual frameworks from accounting literature, we construct standardized annual measures of nine value drivers. Our measures exhibit meaningful variation over time and demonstrate distinct patterns both within and across industries. We validate each of the nine value drivers through associations with established financial-statement and external-rating proxies. Additionally, we demonstrate the usefulness of our measures by testing two hypotheses about long-term firm value: that emphasis on non-financial value drivers predicts higher long-horizon firm value, and that misalignment between firms’ stated organizational objectives and their realized homepage emphasis predicts lower long-horizon firm value. By offering a replicable methodology for quantifying firms’ value drivers using publicly accessible historical website data, our study enables researchers to more comprehensively understand how firms internally drive value.

Communication is a pervasive organizational activity that is difficult to observe from the outside. We construct novel firm-year measures of internal communication quality using a two-stage NLP pipeline applied to Glassdoor employee reviews from 2008 to 2024. A fine-tuned BERT classifier identifies communication-related sentences, which a GPT model then classifies by direction (top-down versus bottom-up) and evaluates on a sentiment scale. Applying these measures to S&P 1500 firms, we find that firms in which employees perceive communication more favorably exhibit higher Tobin’s Q, greater R&D intensity, more patent activity, and faster earnings announcements. The associations are directionally asymmetric: top-down communication perception is most strongly correlated with productivity, while bottom-up communication perception is the primary correlate of innovation and information environment quality. The productivity associations are strongest when both communication directions score highly together. Together, these results suggest that internal communication is a meaningful organizational attribute that varies systematically across firms and correlates with economically important outcomes.

Using a sample of Chinese listed firms from 2007 to 2017, this study investigates the impact of trust on corporate innovation. We find that firms located in provinces with higher trust levels have both higher innovation inputs and outputs, indicating trust plays an important role in promoting corporate innovation. We further document that this association is more pronounced for firms with a higher degree of information asymmetry between boards and managers, and when CEOs have stronger incentives to engage in opportunistic behavior in innovation activities. In addition, we find that forced CEO turnover-performance sensitivity is lower and managers are less likely to underinvest in R&D in high trust areas, suggesting trust enhances corporate innovation through relationship alignment between shareholders and managers. We also find that trust promotes not only the quantity but also the quality of corporate innovation. Our study complements existing research that investigates factors encouraging corporate innovation and extends a growing body of literature that examines the impact of trust on corporate decisions.

This project examines whether and how AI reshapes the classic trade-off over where within the firm decision rights should reside. AI creates a new margin in the economics of authority: it can embed information, expertise, and coordination support at the point of decision while leaving final judgment with employees who possess local knowledge, changing both which allocations of authority are feasible and how motivating delegated authority is to exercise. We test this argument in scheduling decisions at two care organizations: a randomized field experiment at a U.S. home care organization that shifts scheduling authority to caregivers and randomly varies their access to AI-enabled decision support, and a study of a European mental health care organization’s transition to AI-supported self-scheduling.

This project examines the role of AI in employee selection systems. AI can streamline the acquisition of the private information that job seekers reveal in interviews and integrate it more consistently into hiring decisions, but it may also embed bias, both in how it acquires and processes soft information and through the data on which it is trained. We take both archival and field approaches. In the archival analysis, we identify when and which firms adopt AI interview systems and relate adoption to recruiting-related outcomes. In the field study, we partner with an organization that staggered the rollout of an AI interview system for its blue-collar jobs to study the effect on employee–firm match and the underlying mechanisms.