Data Breach Disclosures and Stock Price Crash Risk: Evidence from Data Breach Notification Laws (International Review of Financial Analysis, Volume 93, May 2024, 103164)
Hung Cao, Hieu V.Phan, and Sabatino Silveri
Abstract
Exploiting the staggered adoption of data breach notification (DBN) laws, which obligate firms to disclose data breaches when they occur, as an exogenous shock to data breach disclosures, we find that the adoption of these laws leads to higher future stock price crash risk. The positive relation between DBN laws and crash risk is more pronounced for firms with weaker corporate governance, higher financial constraints, and higher information asymmetry. Our findings suggest that investors’ concerns about the consequences of data breaches and the vulnerability of breached firms’ data security heighten stock price crash risk.
Legal Liability and Stock Price Crash Risk: Evidence From a Quasi‐Natural Experiment (The Financial Review, forthcoming)
Hung Cao, Hieu V.Phan, and Matteo Arena
Abstract
We investigate the impact of a 2001 Nevada corporate law reform, which reduced the legal liability of corporate directors and officers, on stock price stability. Our analysis addresses previous conflicting evidence and reveals a significant decrease in stock price crash risk following this legal change, particularly among small, young, and weakly governed firms. We attribute this reduction primarily to diminished earnings management practices and improved quality of corporate information disclosures. Our findings underscore the influence of reduced managerial legal liability on shaping the corporate information landscape and mitigating stock price crash risk.
Cash vs. Stock: The Role of Cybersecurity Risk in M&A Payment Decisions
Hung Cao, Hieu Phan
[Under Review]
Abstract
This study investigates the impact of cybersecurity risk on the choice of payment methods in mergers and acquisitions (M&A). We find a negative relationship between the acquirer’s cybersecurity risk and the likelihood of stock payments. Moreover, the results are stronger for acquiring firms that have larger cash reserves and higher degrees of information asymmetry. Our finding is consistent with the view that acquirers prefer to pay cash to mitigate possible negative effects of cybersecurity risk on their stock values while target firms prefer cash payments to minimize their exposure to acquirer cybersecurity risk post-acquisition.
Institutional trading around the 52-week and historical highs
Hung Cao, Jeff Black, and Pankaj Jain
Abstract
This study explores the association between institutional trading and psychological barriers around 52-week and historical highs. Our results indicate a positive correlation between institutional trading and the proximity of the stock price to its 52-week high, while a negative correlation is observed when the stock price is close to its historical high. Conditioning to these high price levels, we find that institutional trading can strongly predict future stock returns. Overall, these findings suggest that institutional investors exploit other investors’ behavioral biases, which could have implications for investment decisions and market outcomes.
Handling imbalanced input dataset for Machine Learning predictive models: A case study for banking fraud detection
Hoai Phan, Hung Cao, Oanh Nguyen, Thanh To, and Tu Nguyen
Abstract
The effectiveness and efficiency of machine learning (ML) predictive models are greatly influenced by the quality of data. This study explores various techniques such as random under-sampling (RUS), random over-sampling (ROS), Synthetic Minority Oversampling Technique (SMOTE), and Adaptive Synthetic (ADASYN) to tackle imbalanced problems in credit card transaction datasets. To evaluate the performance of these techniques, a generated balanced sub-dataset is used in four popular ML predictive models, namely Logistic Regression (LR), Random Forest (RF), Gaussian Naive Bayes (GNB), and Extra Trees (ET). We find that the ET model using the SMOTE techniques wins the majority of horseraces, while the RUS technique performs worst in terms of precision, recall, and F1 score. These findings suggest that selecting an appropriate selection of data pre-processing techniques and ML algorithms is crucial for effective credit card transaction fraud detection.