1. When insurers leave: what happens to credit markets?

The insurance-protection gap has widened sharply, with the proportion of uninsured U.S. homes more than doubling, to 12%, between 2019 and 2025 as private insurers retreated from high-risk states. Globally, only roughly one-third of the $318 billion in natural- disaster losses were insured in 2024. Insurers’ decisions to leave locations impacts credit markets: Properties without insurance cannot be mortgaged, municipal tax bases erode as property values fall, and company face higher unhedged operational risk.

• Does insurance market withdrawal predict widening in municipal bond spreads for affected geographies?
• Does insurer retreat signal corporate credit risk that physical hazard scores alone do not capture?
• Can insurance availability data serve as a leading indicator for credit stress?
• What is the relationship between the insurability threshold and geographic bond issuance activity over time?

• Fixed income markets, municipal bond familiarity
• Regression analysis or difference-in-differences methods, Python or R
• Knowledge of insurance markets helpful but not required

2. Climate migration, fiscal stress, and sovereign bond risk

The sovereign-bond market is among the largest asset markets globally, yet it has received less attention for its sensitivity to physical climate risk compared with equities or corporate bonds. What the literature has not yet addressed is the role of climate-induced human migration as a fiscal transmission channel — the pathway from physical- hazard stress to population displacement to deteriorating public finances to sovereign credit risk. Repeated extreme weather events trigger internal or cross-border displacement; displaced populations reduce tax bases, increase demand for social services, and create fiscal and political pressure in receiving countries. For heavily indebted, climate-exposed developing economies, this could create a self-reinforcing cycle of physical stress and fiscal deterioration.

• Does climate-induced internal displacement predict fiscal deterioration in exposed countries?
• Is the fiscal impact transmitted into sovereign bond spreads or credit rating changes, and over what time horizon?
• Is the effect asymmetric by income group?
• Can a composite sovereign climate vulnerability score improve on existing proxies in predicting sovereign yield levels?

• Macroeconomics and international finance background
• Panel analysis / Econometrics
• Familiarity with sovereign debt markets
• Local projections methodology a plus

3. Does adaptation pay? Testing an adaptation premium in credit markets

Recent evidence from development finance suggests that every dollar invested in adaptation and resilience generates more than $10 in economic benefits over the next decade. Yet it is unclear whether capital markets price corporate adaptation at all. Do issuers that demonstrate meaningful adaptation and resilience (A&R) activity — investing in flood-resilient facilities, business-continuity planning, supply- chain redundancy, or climate-adjusted operations — earn a financing advantage in bond markets? MSCI has developed Adaptation & Readiness metrics that score companies on their adaptation activities using AI-assisted extraction from disclosure documents, creating the first opportunity to test this empirically.

• Do issuers with higher MSCI Adaptation & Readiness scores trade at different credit spreads?
• Do high-A&R issuers recover spread levels faster following a hazard event?
• Is the adaptation premium sector-specific?
• Do adaptation bonds carry a greenium analogous to green bonds?

• Fixed income or credit analysis background
• Panel regression methods, Python or R
• Some familiarity with ESG data and disclosure frameworks helpful

4. Exploring supply-chain insights through knowledge graphs

This project investigates what new analytical questions become tractable when global supply chain data is structured as a knowledge graph rather than a flat dataset. Students will build a graph representation of a sector-level supply-chain dataset and explore use cases such as multi-hop exposure tracing, and cluster identification.

• What new insights become visible when supply chain data is structured as a knowledge graph?
• How can knowledge graphs aid multi-hop exposure tracing, cluster identification, and anomaly detection that a flat dataset cannot?

• Familiarity with graph databases or graph data structures (Neo4j, NetworkX, or similar)
• Python or R for data manipulation and analysis
• Experience with network analysis methods a strong plus
• Some familiarity with global supply chains and trade flows helpful

5. Clustering companies by supply-chain similarity

This project explores whether companies grouped by the shape of their supply chains form more useful peer sets than companies grouped by conventional industry classifications. Students will develop a similarity measure based on supply chain structure and test whether the resulting clusters provide better peer comparisons for risk analysis, valuation, and thematic investment.

• Do companies grouped by the shape of their supply chains form more useful peer sets than those grouped by conventional industry classifications?
• Do clusters built on supply chain similarity outperform conventional industry classifications for risk analysis, valuation, and thematic investment?

• Clustering and unsupervised machine learning (k-means, hierarchical clustering, graph-based methods)
• Python or R for data analysis and modelling
• Some familiarity with industry classification systems (GICS, SIC, NAICS) helpful

6. From broad sector classifications to specific minerals

This project develops a methodology for breaking down broad supply chain activity classifications into specific minerals such as lithium, rare earth elements, cobalt and gallium. Students will combine sector-level data with external sources covering mineral production, processing and trade to map geographic concentrations and acute dependencies in global supply.

• Where does a broad sector classification end and a critical mineral dependency begin — and can a repeatable methodology bridge that gap?
• Which minerals are hiding in plain sight within global supply chains, and where are the geographic choke points that investors and risk managers aren’t yet seeing?

• Python or R for analysis and methodology development
• Geospatial analysis or mapping tools a strong plus
• Familiarity with critical minerals trade and supply chains helpful but not required

7. The analyst of 2030: measuring AI’s impact on financial-research productivity

AI productivity claims in financial research are widespread, but measured evidence is scarce. There are very limited published study compares human-only, AI-only, and human + AI performance on the specific tasks that financial research teams do every day. This workstream will produce that study.

The student will design and run a controlled experiment on research tasks drawn from MSCI’s work: extracting data from corporate filings, writing research summaries, spotting anomalies in structured datasets, and drafting methodology notes. The experiment will measure speed, quality and error rates for each mode and each task type. The output will be a research paper with task-level productivity multipliers and a benchmark task suite with a scoring rubric that MSCI can rerun on its own teams. The resulting dataset and methodology become a permanent internal asset for workforce planning and AI adoption decisions.

• When you pit human-only, AI-only, and human+AI against the same financial research tasks, which mode wins on speed, which wins on quality, and does the answer change depending on the task?

• Experimental design, or organisational behaviour with quantitative methods
• Interest in AI and knowledge work
• Ability to design and run structured experiments

8. Project: AI-powered corporate data exchange — from manual reporting to intelligent infrastructure

Disclosure has become one of the defining operational priorities of the modern firm. Companies today engage with an expanding set of reporting obligations – to investors, regulators, data providers and industry bodies – reflecting growing recognition that structured, comparable data is essential to well-functioning markets. The ambition is right; the infrastructure to achieve it efficiently is still catching up.

AI and natural-language capabilities now make it possible to redesign the disclosure experience entirely. This project will examine whether that experience can be redesigned to create genuine value for both the reporting company and the data consumer. The most compelling version of this is an AI-powered disclosure interface that incorporates ease of submission, transparency of use, and meaningful intelligence by design, turning a resource-intensive obligation into a more-efficient, genuine exchange.

• Can an AI-powered natural language interface meaningfully streamline the reporting process for companies responding to investor, regulatory, and industry data requests while maintaining or improving data quality and completeness?
• What would corporate reporting teams find most valuable to receive in return for their data — and how might that shape the design of a more reciprocal exchange?
• Could a well-designed corporate data exchange serve as shared infrastructure across the disclosure ecosystem, making structured issuer data more accessible to analysts, regulators, and smaller investors?
• What governance and trust conditions would make companies confident sharing sensitive operational data through an AI-mediated interface?

• UX/product design, human-computer interaction, or business strategy with an interest in AI applications
• Familiarity with ESG reporting workflows, open data standards, or corporate disclosure frameworks helpful
• Regulatory compliance, platform design, or financial markets exposure a bonus