QMSA Thesis Projects – 2022 Cohort

The economic effects of COVID-19 have been widespread and multifaceted. Researchers are just beginning to understand the extent of the economic recession and the effectiveness of implemented measures. To further expand knowledge around this issue, the current research examines post-pandemic economic shocks and their disproportionate impact on employment and education. The study relies on survey data from CPS (Current Population Survey) and BLS (Bureau of Labor and Statistics) that is aggregated on IPUMS (Integrated Public Use Microdata Series). Multiple regression analysis is performed to isolate specific category effects of the pandemic. The results show the expected distributional effects of the economic recession. These include widespread employment losses and disproportionate effects on low-paying occupations compared to high-paying occupations. Disproportionate impacts are also witnessed in different demographics. Notably, it is observed that African Americans and Hispanics experience higher employment impacts compared to their Caucasian counterparts. In response to the pandemic, liquidity-constrained households seem to have invested less in education.

There are many explanations for why armed groups cooperate. In this paper, I zoom in on South Asia from the 1990s through 2016 to explore some of these explanations using Additive and Multiplicative Effects (AME) models alongside descriptive tools. At the dyadic level, I focus on the roles of shared ideology and group identity in structuring networks of cooperation among armed groups in South Asia. I also bring states into the mix by modeling armed groups’ relations with each other and with states as coevolving longitudinal networks. In doing so, I look at the role of shared state supporters or rivals in drawing armed groups together or pushing them apart. I judge my models by their ability to predict cases of armed group cooperation out-of-sample.

Linguistic discrimination occurs when an individual receives unfair or unjustifiable treatment due to his or her native language or other characteristics of his or her language skills such as accent, grammar, or vocabulary. In particular, individuals speaking a second language may experience prejudice and stigma solely because of their accents. This research designs two cognitive interventions, namely raising individuals’ awareness of the impact of linguistic discrimination and exposure to a foreign accent, to investigate how linguistic discrimination can be mitigated by the two interventions, as well as if there exists an interaction between the two interventions. Participants are randomly assigned to three groups: one experimental group and two control groups. Participants in the experimental group are primed with an audio introducing the phenomenon of linguistic discrimination, while participants in the first control group are presented with an audio about dialects and accents. Participants in all three groups then listen to a pedagogical story read by a Chinese-accented speaker, and complete a lexical decision task where the Chinese speaker iterates auditory stimuli for the participants to determine as word or nonword. I hypothesize that participants exposed to the educational materials on linguistic discrimination as well as participants exposed to the Chinese-accented audio will show shorter lexical response times. I also hypothesize that there exists an interaction between the two independent variables, so that the aggregate effect of the two interventions is greater than the sum of parts. The results of my research can be applied to school settings, where the implementation of the proposed cognitive interventions might be able to help reduce students’ linguistic discrimination against students from minority ethnic groups to uphold equity and avert interpersonal conflicts in the academic community.

Do people thoughtlessly support positions taken by their party leaders, or carefully alter their beliefs when given reason to do so? Many studies examine the effects of cues from party leaders on policy preferences and cast voters as party loyalists, but rarely compare information from party leaders to information from other political and nonpartisan sources and thus cannot disentangle whether people rationally update their preferences or blindly follow party leaders. To investigate, I vary cues to identify the comparative strength of party leader cues and test issue importance and previous knowledge as potential moderators. I find that when asked to support or oppose a discrete policy, partisans respond to cues from party leaders but not other cues. When respondents respond with a continuous range of policy preferences, however, party leader cues are not inherently stronger — and are sometimes weaker — than cues from other sources. I find limited evidence to suggest either issue importance or political knowledge significantly moderates partisan sensitivity to elite cues, no matter the source. These results suggest that while party leaders draw partisans to express support for individual policy planks, leaders’ influence on underlying beliefs is far more complicated and voters engage in more cognition than previously suggested.

The trafficking in wildlife in Southeast Asia continues to increase and exceeds an annual value of US$20 billion. Southeast Asian countries account for under 3% of the world’s land mass and 8% of the global population, but the region is estimated to account for 25% of the global illegal wildlife trade (OECD, 2019). The demand for rare wildlife products has grown in many Asian markets due largely to increased wealth in the region. Illicit trade is motivated by profit and wildlife crime is low risk with high reward (OECD, 2019). Even though a lot of wildlife related problems have been addressed previously, not many of the literature focus specifically on Southeast Asian countries. To further understand the motivation behind these transactions and the common modes of transportation used by the traffickers in the region, I analyzed the top five most trafficked species in Vietnam, Cambodia, Thailand, Malaysia, and Philippines in the past ten years based on the data from TRAFFIC International Wildlife Trade Portal and CITES (Convention on International Trade in Endangered Species of Wild Fauna and Flora) Trade Database. I used different generalized linear models to explain why some species are trafficked the most and how the transportation modes are decided. And the models are tested with all possible combinations among the recorded variables and their interactions.

The economic development and competitiveness of nations are increasingly deemed to be underpinned by the growth and dynamism of regional economic hubs. Within the popular Regional Innovation Systems analysis approach, higher education institutions such as universities are featured prominently as generators of research, innovative technologies and human capital. However, one group of institutions whose role in development remains under-researched are vocational education institutions (VEIs), which in the US largely comprise public community colleges and private vocational/technical schools at the post-secondary level. In this study, we aimed to uncover the possible contributions of VEIs to regional growth and innovation by conducting a quantitative panel regression analysis. Collecting publicly-available data from over 700 counties in the US between 2010 to 2015, we assessed suitability of various linear models for analyzing the relationship between the supply of vocationally-trained human capital (VHK), GDP growth and utility patent generation within counties. Results of our fixed effects regression analyses do not support the hypothesis of positive and significant effects of VHK on county-level growth and innovation, but they showcase the value of future in-depth quantitative research into the possible channels through which VEIs contribute to regional development.

This paper explores the quantitative impacts that corporate cultures have on US publicly-traded firms’ rate of return on stocks. By studying companies contained in the 2019-launched MIT SMR/Glassdoor Culture 500 index, I employ both the Carhart four-factor model and the Carhart four-factor model with the Culture Factor, a self-generated long-short strategy, to study the excess returns on asset prices associated with differences in corporate cultures. The analysis shows that there is no statistically-significant difference in companies’ excess returns on asset prices for those with the best (top 20%) and the worst (bottom 20%) corporate cultures. In addition, the inclusion of the additional risk factor (i.e., the Culture Factor) fails to add explanatory powers to the Carhart four-factor model. Therefore, according to this analysis, company cultures do not offer additional information on asset prices for investors, and they shall not be major factors in investment decisions.  

As of 2020, China’s car ownership has exceeded 280 million units, tied with the United States for first place in the world. The huge car ownership provides a broad base of car sources and room for development for China’s used car market. However, China’s used car market has always remained at the stage of barbaric growth, and there has never been a set of valuation system with market authority. However, the advantages of machine learning in data analysis and prediction can effectively help dealing with information asymmetry. In this paper, I will first introduce the mainstream qualitative methods of used car valuation and point out their shortcomings and problems; then introduce quantitative methods to value used cars. The data I will utilize come from a large online trading platform for used cars. The data will be pre-processed and exploratory analyzed, including missing values, outliers, data feature transformation and other processing work. Then descriptive analysis and visualization will be performed based on the preprocessed data to initially explore the distribution of the data and the correlation between variables. After feature selection and extraction, predictions will be made using two different models whose predictive effectiveness is based on MSE values. Finally, I will draw conclusions and further discuss them in relation to more realistic factors and suggest possible future research directions.