Home » Advisor » Michelle Connolly

Category Archives: Michelle Connolly

Predicting the Work Task Replacement Effects of the Adoption of Machine Learning Technology

by Shreya Hurli

This paper develops a methodology to attempt to predict which tasks in the workforce will be resistant to the replacement of labor by machine learning technology in the near future given current technology and technology adoption trends. Tasks are individual activities completed as parts of a job. Prior research in the field suggests that characteristics of tasks (non-roteness, creativity, analysis/cognitive work) that make them harder for machine learning technology to complete are good predictors of whether those tasks will be resistant to replacement in the workforce. This study utilizes O*NET (Occupational Information Network) task description and education data from October 2015 to August 2020 and Bureau of Labor Statistics salary data to use task characteristics to predict tasks’ resistance to replacement. Normalized scores, average salaries, and average worker education levels are calculated to quantify the relative presence or absence of non-roteness, creativity, and cognitive work in a task. This paper then uses the calculated scores, salary, and education data, as well as a number of interaction terms as inputs to a support vector machine (SVM) model to predict which tasks will be resistant to decline in their shares of workplace tasks weighted by the jobs under which the tasks fall. Using task characteristics, the SVM predicts that just approximately 39% of tasks are likely to be resistant to replacement. These tasks tend to be highly non-deterministic (very non-rote, analytical/cognitive, and/or creative) in nature.

View Thesis

Advisors: Professor David Berger, Professor Michelle Connolly | JEL Codes: J23, J24, O33

Closing the Digital Divide: Evaluation of FCC’s Connect America Fund

by Yurong Jiang

The Connect America Fund is the largest federal support for broadband buildout to unserved areas. This paper provides the first econometric assessment of its effectiveness. The absence of evidence in favor of the program suggests more productive policies should also address the sluggish demand.

View Thesis

Advisors: Professor Michelle Connolly | JEL Codes: H43, H32, H45

The Effects of Pharmaceutical Price Regulation on Probability of Patenting in OECD Countries

by Rachel Korn

The introduction of parallel trade mechanisms allowing for the free trade of pharmaceutical goods in the European Economic Area represents a significant departure from the standard monopolistic competition pricing structure in the pharmaceutical market, in which firms have a great deal of control over pricing. Another mechanism, external reference pricing, also contributes to undermining traditional price structures by imposing a price ceiling on drugs. As these methods of regulating pricing in the healthcare market are receiving growing interest in countries such as the United States, it is prudent to consider their effects. It is apparent that parallel trade and external reference pricing decrease average drug costs, but little has been said about their effects on drug availability. Using global patent data from the European Patent Office PATSTAT database as a proxy for drug availability, I investigate how parallel trade and external reference pricing affect the decision of firms to file a pharmaceutical patent in a given country. I accomplish this through a logistic regression model with a difference-in-differences approach to estimate the probability of patenting a pharmaceutical in an OECD country, given that a patent has previously been approved in the United States. I find that the presence of parallel trade in a country significantly decreases the probability of patenting and increases patent lag time while external reference pricing unexpectedly increases the probability of patenting and decreases patent lag time. These findings demonstrate the complexity in attempting to create policy to regulate rising pharmaceutical prices, as doing so may increase affordability of existing drugs in a country while decreasing availability of new ones.

View Thesis

Advisors: Professor Michelle Connolly | JEL Codes: I1, I11, I19

Analysis of Brain Diagnoses and the Incidence of Chronic Traumatic Encephalopathy (CTE)

by Arjun Lakhanpal

Chronic traumatic encephalopathy (CTE) has become a significant area of scientific inquiry in relation to various sports with contact exposure, specifically boxing and professional football, resulting from many individuals who participated in these sports being diagnosed with CTE neuropathology after death. This paper contributes to the CTE literature by analyzing the various predictors of the progression of neurodegenerative disorders, including CTE, that are associated with a history of head impact exposure. In addition, it analyzes how manner of death shifts depending on an individual’s clinical brain diagnosis, which is a decision based upon the clinical record and case review of a patient.
Through data from the NIH NeuroBioBank, the VA-BU-CLF Brain Bank, and data self-collected from living individuals with symptoms associated with CTE, this paper explores an analysis of various brain diagnoses through a large control population and small subset of athletes and veterans. Logistic regression models are established to analyze explanatory variables of clinical brain diagnosis, manner of death, and CTE presence and severity.
These logistic regression models confirm previous research surrounding the potential racial influence present in Black populations with schizophrenia related diagnoses and illustrate the degree to which neurodegenerative disorders, specifically Parkinson’s Disease, are influenced by increased age. Specific to CTE, the analysis conducted through the sample population illustrates the influence of an extra year of football played at the professional level, while counteracting existing literature regarding the association between position and CTE.

View Thesis

Advisors: Professor Jason Luck, Professor Michelle Connolly | JEL Codes: I10, Z20, J15

Impacts of Housing Interventions on Neighborhoods in Durham County

by Cassandra Turk

Housing intervention models intended to revitalize neighborhoods and empower homeowners are frequently observed in cities across the United States. To determine the efficacy of these programs, this study analyzes the effects of a housing intervention on the price of the home and the changes in neighborhood characteristics that may lead to neighborhood stability or instability in the long run, including the home prices, the racial makeup, the median income, and crime rates of the neighborhood. To study these characteristics and how they interact with interventions, I implement a propensity score matching model to reduce variation in unobservable characteristics and to isolate the effect of interventions on the block group characteristics of interest. In addition, I implement a non-parametric kernel regression to allow for the possibility of a non-linear relationship between home prices and home interventions. The results show significant evidence that interventions increase neighborhood home values at the bottom 10th percentile and at the median of each block group, suggesting that housing interventions do serve to increase the quality of the neighborhood. However, there is evidence that these effects taper off after a certain percent of the households in the neighborhood have been intervened upon, reducing the marginal benefit of completing a new housing intervention.

View Thesis

Advisors: Professor Christopher Timmins, Professor Michelle Connolly | JEL Codes: R2, R23, J10

Maternal Grandparent Living Arrangements and the Motherhood Wage Penalty for Mothers in China

by Mary Wang

Living arrangements of mothers in China significantly impact their annual wages and motherhood wage penalties. I study how the presence of mothers’ parents, or the maternal grandparents, affect mothers’ wages for each child living in the mothers’ households. Existing literature finds that mothers in China not only experience a motherhood wage penalty, but also observe wage impacts from the living arrangements of their family members, such as the paternal and maternal grandparents. Although existing research on motherhood wage penalties references the China Health and Nutrition Survey, I use data from the China Family Panel Studies, the most recent and comprehensive panel survey that reflects the social and economic transformations of contemporary China. To extend and update the analysis of living arrangements on the motherhood wage penalty, I present evidence of the impact of living arrangements on the motherhood wage penalty, distinguishing between the presence of the maternal grandmother, maternal grandfather, and both maternal grandparents. While I find clear evidence that the presence of the maternal grandmother in the household counters the motherhood wage penalty, due to the lack of data on single mothers, I am not able to find conclusive evidence of a difference in the impact of grandparents on the motherhood wage penalty for single mothers compared with married mothers.

View Thesis

Advisor: Professor Peter Arcidiacono, Professor Michelle Connolly | JEL Codes: J12, J16, J21

The Effects of Leveraged Buyouts on Health Outcomes

by Robert Williams

Private equity firms first began acquiring hospitals in the United States during the early 1990s, yet the effects of private equity ownership on patient outcomes and treatment costs are still not clear. Some argue that although private equity firms are adept at improving operating efficiencies and introducing managerial expertise, these cost-cutting measures may come at the expense of patient outcomes.

Because acute myocardial infarctions (AMIs) serve as proxies for patient outcomes and treatment costs, I collect information on 30-day mortality rates and Medicare reimbursements for treatments of AMIs at US Medicare-certified short-term acute care general hospitals from 2014 to 2019. This paper uses fixed effects models to analyze the impact of leveraged buyouts, relative to strategic acquisitions, on patient outcomes. After integrating both hospital and time fixed effects, I find that private equity ownership does not lead to significant changes in Medicare reimbursements or mortality rates for AMI treatments.

View Thesis

Advisors: Professor Ryan McDevitt, Professor Grace Kim, Professor Michelle Connolly | JEL Codes: I0, I110, G340

Municipal and Cooperative Internet on Broadband Entry and Competition

by Tianjiu Zuo

The broadband market is unique for municipal (government-owned) and cooperative (member-owned) competitors. Their participation, however, raises conflict of interest concerns. Both municipalities and cooperatives are often owners of utility poles that are an essential input for broadband deployment. Internet service providers (ISPs) must lease pole attachment space. While most pole attachment rates are regulated, municipal and cooperative pole owners are exempt by Section 224 of the Telecommunications Act. This paper, therefore, studies the competitive effects of municipal and cooperative ISPs, and the effect of potential entry by municipal and cooperative electric utilities (non-ISPs), on broadband entry and quality. I add to the existing literature by building a dataset of municipal and cooperative non-ISP service areas, designing a method to clean the Federal Communications Commission’s (FCC) broadband data, developing a novel geographic entry threat model, and analyzing municipalities and cooperatives in conjunction. I categorize markets into three types: rural, urban clusters (2,500 to 50,000 people), and urbanized areas (≥ 50,000 people). Looking at Illinois from June 2015 to June 2018, I find that the presence of a municipal ISP lowers the probability of market entry and service quality in urbanized areas. The presence of a cooperative ISP lowers the probability of market entry and service quality in rural areas and urban clusters. The presence of a municipal non-ISP has little to no effect on the probability of market entry or service quality. The presence of a cooperative non-ISP appears to increase the probability of market entry in rural and urbanized areas, but depress service quality in urbanized areas, though these effects could be attributed to bad data.

View Thesis

Advisor: Professor Michelle Connolly | JEL Codes: L32, L41, L96

The Effects of Health IT Innovation on Throughput Efficiency in the Emergency Department

By Michael Levin  

Overcrowding in United States hospitals’ emergency departments (EDs) has been identified as a significant barrier to receiving high-quality emergency care, resulting from many EDs struggling to properly triage, diagnose, and treat emergency patients in a timely and effective manner. Priority is now being placed on research that explores the effectiveness of possible solutions, such as heightened adoption of IT to advance operational workflow and care services related to diagnostics and information accessibility, with the goal of improving what is called throughput efficiency. However, high costs of technological process innovation as well as usability challenges still impede wide-spanning and rapid implementation of these disruptive solutions. This paper will contribute to the pursuit of better understanding the value of adopting health IT (HIT) to improve ED throughput efficiency.

Using hospital visit data, I investigate two ways in which ED throughput activity changes due to increased HIT sophistication. First, I use a probit model to estimate any statistically and economically significant decreases in the probability of ED mortality resulting from greater HIT sophistication. Second, my analysis turns to workflow efficiency, using a negative binomial regression model to estimate the impact of HIT sophistication on reducing ED waiting room times. The results show a negative and statistically significant (p < 0.01) association between the presence of HIT and the probability of mortality in the ED. However, the marginal impact of an increase in sophistication from basic HIT functionality to advanced HIT functionality was not meaningful. Finally, I do not find a statistically significant impact of HIT sophistication on expected waiting room time. Together, these findings suggest that although technological progress is trending in the right direction to ultimately have a wide-sweeping impact on ED throughput, more progress must be made in order for HIT to directly move the needle on confronting healthcare’s greatest challenges.

View Thesis

View Data (Email for Access)

Advisors: Professor Ryan McDevitt, Professor Michelle Connolly | JEL Codes: I1, I18, O33

Redefining Resource Allocation in Computing Systems

By Jacob Chasan

A new kernel1 is in town. The current industry-standard for resource allocation on computers does not take the user’s preferences into account, rather programs are given access to resources based on the time that each requested to be run. Although this system can lead to solutions that minimize the time it takes for a program to receive an allocation, it often leads to an incentive misalignment between the programs and the user. This misalignment is exacerbated as the current queue based systems have no inherent mechanism to prevent a tragedy of the commons issue, whereby programs take more resources from the system than the value they provide to the user. By shifting to a market-based approach, where computing resources are allocated to programs based on how much utility the user receives from each program, the incentives of the programs and the users align. With inherent market mechanisms to keep the incentives aligned, this new paradigm leads to at least superior levels of utility for a user.

1As described in subsequent parts of this paper, the kernel is the core program within an operating system which is given the authority to allocate the hardware resources amongst the programs on the computer.

View Thesis

Advisors: Professor Benjamin C. Lee, Professor Atila Abdulkadiroglu, Professor Michelle Connolly | JEL Codes: C8, C80


Undergraduate Program Assistant
Dwayne Russell

Director of the Honors Program
Michelle P. Connolly