Master of Arts, Miami University, 2020, Economics

This paper studies the relationship between oil prices and United States stock market from January 1987 to May 2020. It has been documented in previous studies that oil prices cannot be taken as strictly exogenous. Stock market returns and oil prices are endogenously determined. To address this issue, the use of a Structural Vector Autoregression is employed where the target shock is identified using an external instrument. Impulse responses are obtained and disaggregated between the total U.S. market and 11 chosen sectors. The results of the SVAR-IV model are compared with results from a standard SVAR where shocks are identified with Cholesky decomposition. Cumulative impulse responses are taken to illustrate the change in stock market responses over time. The results show that oil prices generally don't have a strong impact on the U.S. stock market. This paper also illustrates the importance of having current data observing the dramatic changes occurring in the U.S. oil market.

Committee: Nam Vu Dr. (Advisor); Jonathan Wolff Dr. (Committee Member); Jing Li Dr. (Committee Member) Subjects: Economics

Doctor of Philosophy, The Ohio State University, 2021, Physics

Reservoir computing (RC) is a machine learning method especially well suited to solving physical problems, by using an internal dynamic system known as a 'reservoir'. Many systems are suitable for use as an internal reservoir. A common choice is an echo state network (ESN), a network with recurrent connections that gives the RC a memory which it uses to efficiently solve many time-domain problems such as forecasting chaotic systems and hidden state inference. However, constructing an ESN involves a large number of poorly- understood meta-parameters, and the properties that an ESN must have to solve these tasks well are largely unknown. In this dissertation, I explore what parts of an RC are absolutely necessary. I build ESNs that perform well at system forecasting despite an extremely simple internal network structure, without any recurrent connections at all, breaking one of the most common rules of ESN design. These simple reservoirs indicate that the role of the reservoir in the RC is only to remember a finite number of time-delays of the RCs input, and while a complicated network can achieve this, in many cases a simple one achieves this as well. I then build upon a recent proof of the equivalence between a specific ESN construction and the nonlinear vector auto-regression (NVAR) method with my collaborators. The NVAR is an RC boiled down to its most essential components, taking the necessary time- delay taps directly rather than relying on an internal dynamic reservoir. I demonstrate these RCs-without-reservoirs on a variety of classical RC problems, showing that in many cases an NVAR will perform as well or better than an RC despite the simpler method. I then conclude with an example problem that highlights a remaining unsolved issue in the application of NVARs, and then look to a possible future where NVARs may supplant RCs.

Committee: Daniel Gauthier (Advisor); Amy Connolly (Committee Member); Ciriyam Jayaprakash (Committee Member); Gregory Lafyatis (Committee Member) Subjects: Physics

MS, University of Cincinnati, 2020, Engineering and Applied Science: Computer Science

Reinforcement learning is a machine learning model to train software agents for making sequential decisions to maximize long-term reward. It is able to interact with an uncertain and potentially complex environment. Reinforcement learning for multiple time series is one of the critical areas worth exploring. Ever since the reinforcement learning software program, AlphaGo, defeated one of the best Go board game players Lee Sedol in 2016, reinforcement learning has attracted the interest of many different areas, especially the financial community. People are enthusiastic about using reinforcement learning as the key to solving financial trading problems. In this thesis, we explore how to implement reinforcement learning algorithm on multiple time-series data to achieve predefined goals. We pick foreign exchange markets as our testing ground and develop a reinforcement learning automated trading agent to gain rewards through transactions. We propose a reinforcement learning algorithm with the Vector Autoregression (VAR) models describing the market pattern and Exponential Moving Average (EMA) serving as the trading indicator when no information gets extracted from the market. To evaluate the feasibility and robustness, we apply the proposed method on three years daily rates of four currency pairs over the period from 2016 to 2019. With proper parameter tuning, our online reinforcement learning agent is able to achieve more than 20% long-term profit with only one-month pre-training data consumed beforehand. The test is then extended on more datasets with different currency pairs and periods. Though the performance in the following tests is not as stable as we expected due to lack of parameter tuning, we believe the framework we outline provides many advantages as a proof of concept and could easily get extended in the future works.

Committee: Raj Bhatnagar Ph.D. (Committee Chair); Yizong Cheng Ph.D. (Committee Member); Ali Minai Ph.D. (Committee Member) Subjects: Computer Science

MS, University of Cincinnati, 2019, Engineering and Applied Science: Computer Science

In this thesis we have investigated reinforcement Learning for multiple time series data as a solution for finding optimal actions at each time point. The main aspect of the problem addressed by us is the manner in which the state of the system is defined, and the way it is inferred from the observed values of the multiple time series. Our contribution is to define the states in terms of the Vector-Auto-Regression models that fit the multiple time series for short time windows. We then present a reinforcement learning (Q-Learning) technique for predicting optimal actions based on the observed multiple time series data. Most of the existing systems use binning of the time series data to define states, primarily in terms of the bin identities. We have demonstrated the superior performance with our proposed VAR-model based description of the state of the system. This method for state description is effective in characterizing the state of the system and includes the modeling of inter-dependencies among various time series in the state description. The state descriptions are constituted using four currency exchange rate time series data and the Q-learning based system is used for predicting buy, sell or hold actions for specific currencies. We have compared our method of defining states with traditional technique using binning of the observed data. The results of our experiments show that our proposed method for state descriptions performs much better and is very robust in the context of choosing some parameters for our overall learning framework.

Committee: Raj Bhatnagar Ph.D. (Committee Chair); Yizong Cheng Ph.D. (Committee Member); Ali Minai Ph.D. (Committee Member) Subjects: Computer Science

Doctor of Philosophy, The Ohio State University, 2004, Agricultural, Environmental and Development Economics

For nearly a half century, food aid has aroused considerable debate among economists. However, a definitive answer to the fundamental question - “What is the impact of food aid?”- has proven elusive. Theodore W. Schultz's 1960 article warned that program food aid likely had a disincentive impact on farmers in recipient countries. More recently, Christopher Barrett has maintained that food aid has little effect on local production, but rather displaces imports (2002). Both ideas are based on an examination of program food aid; food that is sold on the recipient country's market. Since the 1960s, assistance has evolved beyond program food aid to include targeted food aid, which is at least intended for free distribution to the hungry poor. In this study a welfare analysis is performed to develop hypotheses regarding the relationships among targeted food aid, program food aid, imports and production. The central hypotheses resulting from the theoretical framework are that program food aid discourages production and it may displace imports. Targeted food aid displaces imports and may discourage domestic production. These hypotheses are tested using a vector autoregression similar to that used by Barrett et al. (1998). Departures from Barrett's study include the use of fixed effects to control for differences among countries and differentiation distinguish between targeted and program food aid. INTERFAIS data on food aid (provided by the World Food Programme) are used along with FAOSTAT data on per capita cereal production and imports by country; the data span the years 1988 to 2000 and 64 countries. The main findings of the empirical work are that neither targeted nor program food aid affect food production in the countries receiving them and that both result in import displacement. However, the degree of import displacement is greater for program food aid than for targeted food aid. The implications of this research for policy makers are that improvements to social wel (open full item for complete abstract)

Committee: Douglas Southgate (Advisor) Subjects: