Research

Abstract. We test bright side versus dark side views of relationship lending in a dynamic setting, using daily quoted liquidity in the secondary syndicated loan market. We find bright side effects for prior term loan relationships and dark side effects for credit line relationships. We gauge these effects with bid and ask spreads on traded syndicated loans, where the originating lead banks act as market makers. Narrower spreads point to a more favorable secondary market that may support relationship borrowers' future borrowing, while wider spreads imply the opposite. The design delivers high statistical power and tight identification from fixed effects.

Presented at. 2025 IBEFA Annual Meeting, 2024 Southern Finance Association Annual Meeting.

SSRN

Abstract. Does digital finance help communities withstand drought? Using satellite vegetation data for 132 cities across 34 fragile states from 2000 to 2021, we show that a one standard deviation soil moisture deficit lowers the vegetation share by about five percentage points, and that this loss is far smaller where mobile money is available. Multiple research designs confirm the effect, and mobile money buffers drought most in cities far from formal banking. Bank branches and private credit do not reproduce the effect, so the channel is liquidity, especially mobile account ownership. The form of financial infrastructure, not only its level, shapes climate resilience.

Abstract. What earns consumers relief in financial disputes is verifiable facts, such as dollar amounts, dates, and regulatory citations, not fluent prose. Using nearly four million CFPB complaint narratives, we show that claim specificity, a count of the factual anchors a filer provides, predicts relief where firms exercise discretion, while linguistic polish does not. Specificity rises after state financial literacy mandates, validating it as a measure of consumer capability. Yet complaints increasingly reflect delegated authorship: credit repair firms file boilerplate and consumers paste in generative AI prose, both supplying fluency but omitting the specific facts that drive resolution. Removing delegated text strengthens the link between facts and relief.

Abstract. We find that bank deposits decline in markets where more households report digital asset activity on tax returns during periods of high cryptocurrency returns. Exploiting variation in banks' geographic exposure to these crypto investing households, we show that the effect aggregates to the bank level and is associated with lower loan growth at exposed banks. It also extends to real activity: counties that rely more on exposed banks experience slower employment and wage growth. Our results reveal a novel indirect exposure of banks to cryptocurrency markets and show how cryptocurrency participation can weaken financial intermediation.

Presented at. 2026 American Finance Association Annual Meeting, 2026 Columbia University 10th Annual Bank Regulation Research Conference, 2026 Fintech and Financial Institutions Conference (Federal Reserve Bank of Philadelphia), 2026 Loyola University Chicago Annual Policy Conference, 2025 Annual RCF-ECGI Corporate Finance and Governance Conference, 2025 Financial Management Association Annual Meeting, 2025 Federal Reserve Bank of Chicago's Conference, 2025 Southern Finance Association Annual Meeting, 2025 Friends of Women in Finance 4th Symposium in Greater New York.

SSRN

Abstract. Do capital requirements imposed on global banks in advanced economies carry real consequences for the developing countries those banks leave behind? We build a database of foreign bank exits following the wave of capital regulation after the crisis, using executives' own statements to identify departures they tie to Basel III. Where these banks exit, domestic credit contracts, and the banks that acquire the departing franchises do not replace the lost foreign credit. Manufacturing falls by nearly two percentage points of GDP, activity shifts toward agriculture, exports tilt back toward primary commodities, and output grows more volatile. Prudential regulation written for stability in advanced economies thus exports deindustrialization to the developing world.

Presented at. 2027 IBEFA Annual Meeting at the ASSA Conference, 2026 Financial Management Association Annual Meeting, 5th DC Finance Conference (Kogod School of Business, American University).

Abstract. We study how domestic sovereign risk is associated with household adoption of globally accessible crypto assets using mobile app data across G20 countries. In panel regressions, a 10% increase in sovereign CDS spreads is associated with a 2.8% to 3.6% increase in crypto app usage and downloads. These associations persist after controlling for a large set of macro-financial factors. For downloads, they are strongest in emerging markets and in environments with a high social cost of default. Event study estimates show an increase in crypto adoption following sovereign credit events associated with large CDS jumps and rating downgrades, particularly during episodes of high inflation or restricted capital mobility combined with political instability. These patterns are consistent with two complementary motives: asset preservation in high-inflation environments and asset mobility in politically unstable environments facing capital restrictions. The findings suggest that domestic financial vulnerability can create precautionary demand for access to global digital financial markets.

Presented at. 2025 Federal Reserve Bank of Atlanta's Conference, 2025 Federal Reserve Bank of Chicago's Conference, 2025 Interagency Risk Quantification Forum, 2025 MRS International Risk Conference, 2025 Southern Finance Association Annual Meeting, 2025 Friends of Women in Finance 4th Symposium in Greater New York.

SSRN

Abstract. We audit how well corporate disclosures measure firms' adoption of six general purpose technologies: AI and machine learning, cloud, blockchain, IoT, cybersecurity, and quantum computing. Using earnings calls and regulatory filings from 2014 to 2024, a 1,200 sentence human audit, and an ensemble of three Claude language models, we document three facts. Measurement precision varies more than tenfold across technologies, from 4 percent for blockchain to 94 percent for IoT, and agreement across disclosure channels is very low. The language model ensemble is 97 percent precise but misses many true cases. Estimated effects on firm value differ sharply across technologies.

Presented at. 2026 Financial Management Association Annual Meeting.

Abstract. Large U.S. banking organizations hold substantial amounts of high-quality liquid assets (HQLA) under the Liquidity Coverage Ratio (LCR). We estimate how LCR buffers respond to the opportunity cost of holding HQLA rather than loans, using within-bank variation in bank-specific loan–HQLA spreads. A one-percentage-point increase in this spread is associated with a 0.12 log-point decline in the LCR buffer. We estimate that a partial recalibration of the LCR floor from 100% to 85% of stressed 30-day net cash outflows for the largest banks would release approximately $0.2–$0.5 trillion of HQLA and $70 billion of system reserves. The implied effect on the EFFR–IORB spread is economically negligible in the current ample-reserves regime. We estimate that the shadow cost of LCR compliance for constrained banks is $44 billion per year.

Presented at. 2026 Federal Reserve Stress Testing Research Conference, 2026 Fischer-Shain Center Research Conference.

Abstract. Many firms depend on the same natural capital. One firm's damage to a shared ecosystem erodes the collateral of other borrowers and the books of their banks. For each syndicated loan, we ask which lender is chosen to lead it, and find that mandates for projects that damage nature flow away from exposed lenders. The selection is economically and statistically significant, and specific to the ecosystem the borrower's industry damages. Because the lead lender controls the loan, the damaging project ends up with the bank least motivated to restrain it, and its credit risk with the banks least attentive to the ecosystem.

Abstract. We study how advertising cryptocurrency acceptance affects liquidity in the standardized U.S. residential real estate market. Using MLS listings covering 42 million records from 2018 to 2025, we find that crypto acceptance is rare, appearing in only 1,912 listings. It carries no robust price effect but a large liquidity penalty. These listings stay on the market roughly 36 percent longer, are 15.5 percentage points less likely to sell within 180 days, and are 7.3 percentage points more likely to be withdrawn or canceled. The penalty is largest where local crypto familiarity is lowest, suggesting that unfamiliar payment terms raise perceived information asymmetry rather than signal asset quality.