Arun Gupta

Machine Learning and AI for Finance

• Agentic AI for Credit Risk (work in progress) Summary
  Building a multi-agent system in Python that automates risk analysis for large credit and bond portfolios. The agents autonomously generate cash-flow projections, estimate tail risk, and produce standardized, easy-to-use risk reports.

• Machine Learning for Financial Tail Risk Forecasting

Under revision at the Journal of Alternative Investments. Paper 

In comparison to traditional nonlinear GARCH models, I find that machine learning models (LightGBM) produce 67% to 85% more accurate tail-risk and volatility forecasts for highly volatile asset returns.

• Estimating Macroeconomic Risk Premia in Private Credit (work in progress)
  Benchmarks ten machine-learning methods for forecasting default risk and uses the best-performing model (random forests) to quantify macroeconomic risk and estimate risk premia for illiquid credit assets.

Analysis of Alternative Investments

• The Performance of Private Equity: Evidence from Confidential Filings (working paper) Paper
  Using novel data, I show that (1) larger PE funds systematically outperform, (2) industry-diversified funds deliver lower returns but superior reward-to-risk ratios, and (3) leveraged buyouts do not yield excess return premia.

• Risks in Hybrid Securities (Federal Reserve project)
  Reviewed risks in callable and convertible bonds, credit-linked notes, and insurance-linked products, focusing on their performance under market stress scenarios.

• Risks in Private Credit (Federal Reserve project)
  Analyzed the performance of different types of private credit loans, e.g., mezzanine, unitranche, and NAV-based facilities, under an environment of weakened borrower fundamentals or market conditions. 

• Risks in Project Finance Credit (Federal Reserve project)
  Reviewed project finance credit agreements within large U.S. banks’ wholesale credit portfolios, assessing covenant breach risk, sponsor strength, and collateral coverage for loans associated with infrastructure and energy projects.

Credit and Macroeconomic Growth

• Extracting Predictive Macro Signals from Bank Decisions During Crises (with H. Sapriza)

Under revision at the Journal of Money Credit and Banking. Paper
Banks often have private information about borrower conditions that markets do not fully observe. We show that costly bank actions during crises reveal this information, which can be extracted into a predictive signal that significantly improves forecasts of macroeconomic outcomes.

• The Collateral Channel and Bank Credit (with H. Sapriza, V. Yankov)

Under revision at the Journal of Financial Economics. Paper

Provides causal evidence that asset price booms contribute to economic booms, as increases in collateral values significantly expand credit capacity and investment by corporations. Specifically, a 1 dollar rise in collateral values causes a 6 cent increase in future credit and a 3 cent increase in future capital expenditures.

• The Usability of Bank Capital Buffers and Credit Supply at SMEs during the Pandemic (with J. Berrospide and M. Seay)

International Journal of Central Banking (IJCB) 2024 Vol 20 No 3 Paper

We show causal evidence that capital regulation backfired during COVID. Poorly designed regulation caused banks to curtail business credit 10% more than otherwise warranted by the Covid recession (further leading to 2% slower employment growth in affected areas), adding salt to the economic wound.

• A Natural, yet Imperfect Hedge for Interest Rate Risk: Estimating the Value of Deposit Franchise. Working Paper. SSRN

I show direct empirical evidence that markets price in deposit franchise (worth on the order of $1 trillion USD) when valuing bank equities, and quantify how imperfect this deposit franchise is as an interest rate hedging tool.

• Measuring Global Bank Complexity (with Linda Goldberg and Nicola Cetorelli), FRBNY Liberty Street Economics, 2014. Link

Addressing the macro-prudential concerns of the growing “complexity” of global financial institutions, we look present comprehensive data analyses on various metrics for measuring the “business line”, “geographical”, and "organizational” complexity of the largest financial intermediaries.

• Ring-Fencing and “Financial Protectionism” in International Banking (with Linda Goldberg), FRBNY Liberty Street Economics, 2013. Link

We discuss recent policy and regulatory efforts to induce banks to shift away from global activity, including ring-fencing of domestic banking operations, other forms of financial protectionism, and enhanced oversight and prudential measures.