Thesis Title: "Essays on Digital Economics"
Best Dissertation Award
Committee:Daniel Arce (Chair), Catherine Tucker, Anne M. Burton, Anton Sobolev
Methodology: Causal Inference, Econometrics, Observational Data, Game Theory
Topics: Health IT, Cybersecurity, Privacy, AI
Working Papers
Abstract:
Hospital data breaches have been shown to disrupt care, incur costs for remediation, and under
mine investor confidence. However, less attention has been paid to the causes of these data breaches. This
study explores the effects of mergers on hospital data breaches. Drawing upon U.S. hospital merger data over
ten years, I empirically show that data breach rates double during merger execution. The results suggest that
increases in data breach risk during merger execution are mainly caused by intensified external threats. I
present some evidence revealing online visibility and information system integration increase external threats.
These findings have managerial implications for predicting and managing the risk of data breaches.
Mergers and Data: Evidence From Healthcare, with Catherine Tucker and Amalia Miller
Abstract:
While antitrust concerns often focus on market power arising from the merger of two firms
if they both have access to large datasets, our study highlights that data itself can be an impediment to
consolidation and realizing benefits from consolidation across firms. We build a panel data set from 2007
to 2019 that deploys hospital information from the American Hospital Association’s annual surveys; cost
data from the Healthcare Cost Report Information System; and system compatibility measures from the
Healthcare Information and Management Systems Society. Using difference-in-differences estimation, we find
that compatible mergers achieve cost synergies consistent with prior literature while incompatible mergers
show no significant cost improvements. Compatibility especially benefits geographically local mergers, where
data compatibility facilitates efficiency gains from market share expansion and administrative consolidation.
Further analysis reveals that this pattern of differential synergy realization is particularly pronounced in
administrative costs. Mergers characterized by initially incompatible systems achieve a 5% cost reduction in
the first post-merger year when they utilize cloud-based infrastructure, suggesting that cloud technology may
serve as a mediating mechanism for overcoming data system incompatibilities.
Marketing, AI and Data: Evidence from Medical Devices, with Catherine Tucker
Abstract:
It is unclear how privacy policies shape the development of AI technologies. On the one hand, when
a state passes a privacy policy, compliance costs and legal risks increase. On the other hand, an established
set of privacy regulations may give consumers and firms confidence about how training data sets will be
used in the development of AI. We explore these questions empirically in the setting of radiology software, a
frontier setting for the use of AI in medicine. In states with stronger privacy protections, companies are less
aggressive in marketing AI-powered devices. Using detailing records from the Open Payment project, FDA
documents, national demographic data on doctors and clinicians, and the FDA’s list of radiology AI/ML
Software as Medical Devices, we explore how current AI and privacy law shape how firms approach marketing
payments to physicians. Our analysis reveals that payments from AI medical device manufacturers decrease
by 22.6% when the physician’s primary residency state enacts new privacy policies.
Detailing and Ownership with Yaa Akosa Antwi and Catherine Tucker
Abstract:
The strategic decisions that guide business-to-business marketing are often hard to observe. For
example, how does the organizational structure of the potential client affect how the firm markets toward that
client? How do firms improve their efficient marketing? We explore this question in the context of detailing
decisions towards physicians. We use the sharp discontinuity of when a physician becomes the owner of an
ambulatory surgical center, rather than an employee how this changes detailing behavior towards them. We
find that the probability of detailing activities in a quarter increases, and most increases are in food and
beverage. We find that female physicians, in particular, often receive lower marketing pay, to the point that
when they become owners, they receive significantly more from food and beverages.
Monitoring in Healthcare IT with Xiru Pan, Niam Yaraghi, and Catherine Tucker
Abstract:
Healthcare IT has transformed healthcare. However, a challenge has been that often doctors feel
burdened by having to interact with their computers consistently. Prescription Drug Monitoring Program
(PDMP) were designed as powerful tools to combat prescription drug misuse. These state-wide systems allow
physicians to access patients’ controlled substance prescription histories across providers. Using a natural
experiment from policy shocks in Washington, we show that Oregon physicians near the Washington border demonstrated significant increases in PDMP usage, both in new user adoption and query frequency among
existing users. Despite the increase in patient volume, doctors are using PDMP to deter doctor-shopping.
Publications
Article in Advance, Information Systems Research
Abstract:
Cloud services exist under a shared security environment with a dynamic nature; users trade fixed costs for variable costs over time and both cloud service providers (CSP) and users contribute to overall security. We investigate the nature of shared security in a dynamic game where users' security contribution takes into account both their users as well as competition with other CSPs. The Markov Perfect Equilibrium reveals the long-term time patterns of security of the cloud. In particular, we identify a novel form of time-path strategic complementary between usage and a CSP's Markov state of security. This implies cloud security is an unusual form of impure public good whereby individual contributions bolstering a CSP's security endow a selective incentive (private benefit) on others, rather than on the contributor alone. Since this increases usage, CSP vulnerability increases over time. At the same time, CSP competition on security may lead to both welfare improvements for users and lock-in.
Selected Works in Progress
Opioid and Information Systems, with Catherine Tucker, Funded by MIT Sloan Health Systems Initiative
AI and Operation Management: Evidence from Healthcare
Fellowships, Grants, Honors, and Awards
WEIS’23 Best Paper Award,
UC Berkeley SLMath Algorithms, Approximation, and Learning in Market and Mechanism Design invited attendee,
Charles C. McKinney Scholarship,
DFW Research Data Center Grant ($10,000),
NBER 2023 Workshop of Digital Economics invited attendee,
NBER 2023 Digital Economics Tutorial invited attendee,
Irving J. Hoch Scholarship,
Office of Graduate Education Research funding,
NSF graduate student travel grant,
Alfred P Sloan Foundation student travel grant,
Betty Gifford Johnson Travel Award,
NBER 2022 Fall Economics of Privacy Tutorial invited attendee
Services
Information System Research, Journal of Industrial Economics, Journal of Cybersecurity Reviewer,
ASHEcon 2024 discussant, WEIS'24 Session Chair,
ASHEcon 2023 Newsletter writer