Working Papers
Permanent URI for this collection
Browse
Recent Submissions
Publication Soverign Debt Restructuring An Assessment of Investor Pricing Behavior(2023-05-01) Stewart, OliverThis research paper focuses on the analysis of investor behavior in sovereign bonds, with the aim of assessing their ability to detect impending default risk. Using a dataset of external sovereign bonds spanning from 1820 to 1980, the paper performs a dynamic difference in difference regression. The paper finds that sovereign bond investors begin to react to early signs of default, with defaulting bonds beginning to be priced down 40 months before default. The prices reduce steadily relative to non-defaulting bonds until around 10 months before the default date. At around 10 months the drop in prices accelerates until the default date. Post-default, the prices continue to decrease, but at a decreasing rate, reflecting the uncertainty in the haircut.Publication The Effect of Upzoning on Single-Family Building Permits: An Intermediate Term Analysis of the Minneapolis 2040 Plan(2023-04-01) Selvamani, SonaIn 2020, over 30% of households had unaffordable rent. Policymakers currently believe the solution to this affordable housing crisis is to allow for denser development in areas by implementing less restrictive zoning policies. While evidence shows that increased restrictiveness in zoning can negatively affect prices or the supply of housing, there is little evidence on whether upzoning can reverse these consequences. This paper uses Minneapolis city’s 2019 ban on single-family (R1) zoning to close this gap and find out what effects a more liberal change in zoning policy might have on building permits in the area. I use the U.S Census Bureau Building Permits Survey to test the effect of this policy on construction permitting from 2016-2021. Statistical results prove that there was a reduction in single-family permitting in Minneapolis after the policy implementation date compared to other communities in Hennepin County, MN. However, there is no conclusive difference in total units and multi-family units permitted after the policy was implemented. This paper sheds light on the intermediate term effects of upzoning and encourages policymakers to assess these policies more carefully.Publication Analyzing Key Events & Casual Factors Impacting Cryptocurrency Lending Rates(2023-05-01) Bobovich, NicoleDigital asset markets reached peak valuations of more than 3 trillion in 2021. However, the strong growth in crypto asset value in 2021 was met with turmoil and value destruction in subsequent years with a series of bankruptcies of cryptocurrency-related businesses in a period dubbed the "crypto winter." This study considers the impacts of the events of the crypto winter on decentralized lending protocols, specifically using the Aave protocol as a case study. Looking at data centered on interest rates, available liquidity, and the total value of invested funds on the Aave protocol, this study seeks to better understand how investors' perceptions of the safety of borrowing cryptocurrencies have changed after the events of the crypto winter. The study will aim to look at predicting spreads between variable and stable interest rates with an OLS regression and analyze the impacts of the crypto winter using historical Aave lending data.Publication Congressional AI: A Framework for Task Generalization and Alignment with Expert Language Modes(2024-04-08) Ramji, KeshavAs foundation models have facilitated rapid adaptation to downstream tasks, a challenge remains in efficiently and flexibly improving their instruction-following capabilities and alignment to human preference distributions. We propose a novel modular architecture, Congressional AI, consisting of parallel trained "experts", such that the top-k relevant experts can be activated during inference. These experts are obtained by fine-tuning LoRA adapters on interpretable data mixtures; for instruction-tuning, each dataset corresponds to a task cluster, while for preference alignment to improve steerability, each dataset represents a group or persona. Our experiments show that instruction-tuning with Congressional AI through low-rank adapter merging is effective via evaluation of cluster-specific adapters across various domains on the MMLU benchmark. These findings demonstrate that Congressional AI is a hardware-efficient and interpretable mixture-of-experts (MoE)-style framework for adapting language models to new tasks and domains, and can be used to further improve both pre-trained and fine-tuned LLMs.Publication Stories Told in Multistakeholder Governance Inititatives: Understanding the Grievance Process of the Roundtable on Sustainable Palm Oil (RSPO)(2024-12-01) Shen, AngelaThe Roundtable on Sustainable Palm Oil (RSPO) brings private sector and civil society groups under a system of governance to ensure the sustainable, equitable, and humane production of palm oil. The RSPO includes a grievance mechanism that seeks to remedy the reported violations of its members. This paper builds upon a large body of scholarship examining the effectiveness of the RSPO’s conflict resolution through an in-depth investigation of the published correspondences of a complaint case involving PT. PP London Sumatra Indonesia Tbk (a subsidiary of Indofood Agri Resources Ltd) and three non-governmental organizations (NGOs). Using methods of textual analysis through NVivo and temporal analysis, this research identifies and distinguishes narratives of collaboration, opposition, and compliance used by the RSPO, NGOs, and palm oil suppliers to influence outcomes in a grievance process. The interaction of these narratives highlights strengths and gaps in the RSPO’s ability to produce trust and mediate conflict.