This dissertation consists of three parts. In the first part, we study an oligopoly model in which firms compete across several geographic regions. This networked competition is prevalent in many markets, such energy, metals, and agricultural commodity markets. Firms operating in these industries are constrained by physical limits on production capacities. Our paper provides the first analytical study on firms' competition in industries where players are capacity constrained. We find that a reduction in import-export taxes can have qualitatively different effects on consumer welfare depending on whether or not the impacted firm is capacity constrained. Our results imply that policies that promote free trade (e.g. NAFTA, European Union) may have unintended consequence and reduce the consumer surplus in capacity constrained industries.
The second part of this dissertation analyzes the pros and cons of Bitcoin payment systems. The creator of Bitcoin envisioned a decentralized payment system in which mining can be performed by anyone using their home computers. Since it was introduced in 2008, Bitcoin attracted significant attention, both by public media and by investors. This led to a surge in the bitcoin price, and its market capitalization exceeded $170 billion (as of February 2, 2020). With the rise of bitcoin price, firms started to invest in developing efficient hardware to increase their probability of successfully mining blocks. As a result, mining operations became vertically integrated with single firms designing and manufacturing mining chips, and operating them in data centers. These major developments in mining technology bring up the following question: Does Bitcoin's proof-of-work protocol serve its intended purpose of enabling and supporting a decentralized payment system? We propose a two-stage game to answer this question. Firms first invest in research and development to subsequently compete in a Bitcoin mining game. We show that firms fail to capture the surplus created from their research, because higher research expenditures induce a more aggressive mining game. We calibrate our model to rewards and operational costs observed in the Bitcoin system, and quantitatively demonstrate that the mining industry has a tendency towards centralization, against the core principles of cryptocurrencies.
The third part of this dissertation studies the emerging robo-advising industry. Roboadvisors are threatening traditional wealth management firms due to their ability to offer lower fees and minimum balance requirements, as well as transparent and systematic advise. Robo-advisors had $300 billion in assets under management during 2016, and are projected to reach $2.2 trillion by 2020. Currently, robo-advising firms employ questionnaires to assess the risk preference of investors. While appealing, the use of questionnaires presents various shortcomings: (i) investors' answers do not account for emotional responses observed when the loss is incurred, (ii) survey responses are subject to noise, and (iii) risk tolerance assessments are sensitive to the specific wording and formats used in questionnaires. To overcome these limitations, we propose a reinforcement learning framework for retail roboadvising. The robo-advisor does not know the investor's risk preference, but learns it over time by observing her portfolio choices in different market environments. We develop an exploration-exploitation algorithm which trades off costly solicitations of portfolio choices by the investor with autonomous trading decisions based on stale estimates of investor's risk aversion. We illustrate how, by correcting for the investor's mistakes, the robo-advisor may outperform a stand-alone investor regardless of the investor's opportunity cost for making portfolio decisions.
| Identifer | oai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/d8-v2gw-g545 |
| Date | January 2020 |
| Creators | Alsabah, Humoud |
| Source Sets | Columbia University |
| Language | English |
| Detected Language | English |
| Type | Theses |
Page generated in 0.01 seconds