Dissertations, Theses, and Capstone Projects
Date of Degree
9-2024
Document Type
Doctoral Dissertation
Degree Name
Doctor of Philosophy
Program
Economics
Advisor
Sebastiano Manzan
Committee Members
Wim P. Vijverberg
Lilia Maliar
Subject Categories
Econometrics | Finance
Keywords
Machine Learning, Econometrics, Computational Economics
Abstract
Chapter 1 Data-Driven Graph Neural Networks And Realized Volatility Prediction through Graph Discovery: Financial markets’ interconnectedness plays a crucial role in financial predictions. However, existing network structures often fall short in aligning with specific prediction tasks. This paper presents a novel model that integrates graph discovery techniques with Graph Neural Networks (GNNs) to address this challenge, focusing on short-term intraday realized volatility prediction. Our investigation from a graph theory perspective reveals that precision-based graph discovery methods substantially enhance GNNs accuracy by generating meaningful graphs and node embeddings. In contrast, conventional covariance-based approaches exhibit limited effectiveness due to their consideration of indirect influences, resulting in suboptimal graph structures. These findings underscore the crucial role of graph topology in optimizing GNNs, offering valuable insights for enhancing predictive models in financial markets.
Chapter 2 Agent-based Model Parameter Estimation with Wasserstein Distance (Co-authored with Sebastiano Manzan): The goal of this chapter is to propose an estimation method that is accurate when the interest is to estimate the parameters of high-dimensional models, such as DSGE and ABM. In particular, we propose a Simulated Minimum Distance (SMD) estimator based on the Wasserstein distance between the model- simulated distribution and the empirical distribution. There are several advantages of this methodology. First, the SMD does not require costly and difficult approximations of the likelihood function. Second, the Wasserstein distance is more stable, and it is defined also for distributions with no overlapping support. We first conduct a comparison exercise of the Wasserstein distance estimator and other estimation techniques on the Brock and Hommes (1998) asset pricing model and a set of classical time series models. The results show the verified statistical properties of the estimator and its significantly higher accuracy relative to established estimation methods such as Bayesian estimation, Simulated Method of Moments, and other novel information theory-based estimation methods. Finally, we estimate the Brock and Hommes model to financial data and discuss the results relative to the existing literature.
Chapter 3 Endogenous Demand Driven Supply Chain Cycles with Exogenous Shocks: This chapter presents a model of inventories and productions within a demand- driven supply chain, encompassing upstream, midstream, downstream, and retail markets. The dynamic system is conceptualized as a network of two-dimensional oscillators, accounting for positive feedback from lower stream demand and negative feedback stemming from the need to mitigate inventory costs. Other features of the supply chain are also integrated into the model: bullwhip effect, sluggish response to changes in demand, and upstream industries’ sluggish expansion process. This chapter shows that supply chains exhibit endogenous fluctuations without the need for external shocks. Notably, under certain parameters within our model, the supply chain system presents a minor degree of chaos.
Recommended Citation
Wang, Lulu, "Essays in Machine Learning, Econometrics, and Computational Economics" (2024). CUNY Academic Works.
https://academicworks.cuny.edu/gc_etds/6418
