Dissertation
Applications of Computational Methods to Contemporary Economic and Political Issues
Washington State University
Doctor of Philosophy (PhD), Washington State University
2023
DOI:
https://doi.org/10.7273/000005100
Abstract
The three analyses contained in this manuscript use computational methods to address pressing economic and political questions. In the first chapter, I examine the welfare implications of state level responses to COVID-19 using a structural model of household response and millions of Monte-Carlo simulations. I find that moderately restrictive response regimes promote the lowest degree of welfare loss. In the second manuscript, I use millions of lines of text data to assess the causal role that polarized media plays in driving political partisanship in the United States. I find significant evidence of a causal effect of partisan media on political polarization. Finally, in the third chapter I use a recurrent neural network with long-short term memory (LSTM) to predict monthly inflation rates. Using Shapely values, I decompose the the LSTM network's variables to determine which economic factors are most predictive of inflation over the years 1982-2022. We find that lagged inflation, the unemployment rate, and measures related to international trade are most predictive of inflation. Finally, we compare the performance of the LSTM network to more traditional time-series models in a host of forecasting applications. We find that the LSTM network does not offer statistically significant performance improvements over more conventional models.
Metrics
14 File views/ downloads
32 Record Views
Details
- Title
- Applications of Computational Methods to Contemporary Economic and Political Issues
- Creators
- Michael Mahoney
- Contributors
- Salvador Ortigueira (Advisor)Jinhui Bai (Committee Member)Jia Yan (Committee Member)
- Awarding Institution
- Washington State University
- Academic Unit
- Economic Sciences, School of
- Theses and Dissertations
- Doctor of Philosophy (PhD), Washington State University
- Publisher
- Washington State University
- Number of pages
- 152
- Identifiers
- 99901019234001842
- Language
- English
- Resource Type
- Dissertation