From Shares to Shields: The Role of Employee Ownership in Mitigating Data Breach Risks
Job market paper
Employees have a comparative advantage in monitoring their peers and other employees—an advantage that managers and executives do not have. I call this mutual monitoring and investigate whether it can protect firms from risks where a small mistake by one individual could cause significant harm, such as data breaches. To explore this, using AI(BERT models), I conduct a comprehensive textual analysis to identify distinct data breach incidents reported by the same firm. I then examine how the ratio of active participants in employee stock ownership plans (ESOPs) to total employees (the active ratio) affects the probability of a data breach incident. My findings indicate that a higher active ratio is associated with a lower probability of such incidents. Moreover, by analyzing the extensive and intensive margins of ESOP ownership, I find that two firms with the same ESOP value per employee can experience different levels of protection against data breaches due to variations in their active ratio; the distribution of ESOP assets within the firm matters in protecting against data breach incidents. Next, using a staggered differencein- differences model, I analyze how the first noticeable data breach in an industry impacts the active ratio of peer firms within the same industry. I find that ESOP firms increase their active ratio by 3 to 4 percentage points following the industry shock, driven by a 7 to 10 percent increase in the number of ESOP owners within those firms. Notably, this change remains persistent after excluding financial firms and industry shocks coinciding with the years 2007 and 2008.
My paper is the first work to:
Investigate the role of broad stock-based compensation as a risk management tool.
Study the effects of data breaches on rank-and-file employees.
Utilize large language models to identify distinct data breach incidents within one of the most widely used datasets in the literature.
Highlights of my methodology include:
Causal inference
Staggared difference in difference
Matching
BERT models
Natural language processing