Last Draft: September 1, 2026

    This paper examines how banks' exposure to repo markets and reliance on deposit funding shaped their financial performance and balance sheet strength across monetary policy regimes. Using linear models, quasi-natural experiments, and a theoretical framework, I show that post-crisis monetary policy transformations reshaped banks' equilibrium behavior in heterogeneous ways across risk types. Before the crisis, lending and financing behavior diverged: risk-seeking banks prioritized lending income, while risk-averse banks prioritized balance sheet strength to maintain low repo financing costs. After the crisis, these behaviors converge, as the extensive use of repo during QE helped risk-seeking banks shift toward greater repo reliance, with net interest margins rising by 2.27% as a result of a one-standard-deviation increase in repo dependence, while risk-averse banks extend more credit, strengthening Tier 1 capital ratios by 4.61% as a result of a one-standard-deviation increase in delinquency rates, reflecting structural changes and competitive pressures in the post-crisis environment.

Figures on the right: 

Top panel: Average effect of increasing repo dependency on the net interest margins of the risk-seeking banks during ZIRP

Bottom panel: Average effect of increasing loan portfolio risk on the tier 1 capital ratios of the risk-averse banks between the start of QE3 and the onset of the COVID-19 crisis

Presentations: 

2025: University of Georgia

2026: Southwestern Finance Association, Case Western Reserve University, Southern Finance Association

Last Draft: April 1, 2026

This paper exploits the tonal and topical structure of press conferences following Federal Open Market Committee (FOMC) meetings to train large language models (LLMs) that generate realistic responses to journalists’ questions by Federal Reserve Chairs. Using both actual and LLM-generated responses, I study the extent to which language and information-based features predict short-horizon movements in Treasury ETF trading volume, returns, volatility, and quoted and effective spreads during the press conference window. Forecasts based on LLM-generated responses exhibit strong average predictive performance across a range of econometric and machine-learning models, often delivering higher explanatory power and lower forecast error for liquidity and trading-activity measures. The results highlight a clear distinction between mechanical market responses to linguistic structure and informational reactions to policy communication, demonstrating both the usefulness and the limitations of LLMs in analyzing central bank communication and short-horizon market dynamics.

Figure on the right: 

Feature importance (based on the percentile of contribution to the forecast) of the questions and answers' tone-topical content in forecasting microstructure variables 

Presentations: 

2026: Southern Finance Association

Awards and Nominations: 

Outstanding Doctoral Student Paper (Runner-up) - 2026 Southern Finance Association Annual Meeting 

Fed Council Model:

https://huggingface.co/spaces/vasyakimenko/FedChairChat 

Last Draft: September 4, 2026 

Under Initial Review, The Journal of Finance

Dealer reciprocity is the fraction of a dealer's directed interdealer links matched by trade in the opposite direction. Using municipal-bond transaction chains, we show that dealers with more reciprocal networks return more often to recent counterparties and route fewer bonds directly to customers. Reciprocity's association with chain markups depends on credible alternatives: it is positive in narrow networks and negative in broad networks, especially when alternatives are relevant to the bond being traded. After an abrupt collapse in a counterparty's activity, dealers with a reciprocal pre-event relationship subsequently have narrower networks and higher markups than dealers whose relationship with the same counterparty was one-way. The evidence is consistent with reciprocity as a persistent form of relational organization whose pricing consequences depend on credible alternatives.

Table on the right: 

Average transaction markups, transaction chain length, and network size based on the chain-initiating dealer's degree of centrality and reciprocity

Presentations (* for presentations by a coauthor): 

2024: University of Georgia

2025: Virtual Municipal Finance Workshop*, 14th Annual Municipal Finance Conference at Brookings*, Financial Management Association*, German Finance Association*, Southern Finance Association

2026: Midwest Finance Association*, Eastern Finance Association*, Financial Management Association European Conference*, Asian Finance Association (accepted, withdrawn), University of North Carolina at Chapel Hill Alumni Conference*

Media Footprint: 

https://www.bondbuyer.com/news/dealer-quid-pro-quo-shapes-bond-price-markups-paper-says 

Awards and Nominations: 

Best Microstructure Paper (Semifinalist) - 2025 Financial Management Association Annual Meeting 

Last Draft: February 22, 2026 (Available upon request)

We study smithereening, the rapid pass-through of newly issued municipal bonds from institutional allocations into retail hands through tightly linked interdealer and dealer--customer trades. Using EMMA transactions, we identify smithereening trades by exact matches on CUSIP, timestamp, and par between an interdealer purchase and a customer sale, and classify a bond as smithereened if it exhibits such activity in the issuance window. Smithereening is common and retail-oriented, and smithereened bonds exhibit substantially higher offer-relative markups in the early aftermarket. In regressions with rich controls and fixed effects, bond-level exposure predicts 20-23 bps higher markups concentrated in small tickets. After Rule G-15 mark-up disclosure in 2018, the excess markups compress, consistent with transparency limiting intermediary rents.

Figure on the right: 

Average Effect of the MSRB Rule G-15 Amendment on the Markups of Smithereened Bonds

Presentations (* for presentations by a coauthor): 

2026: University of Georgia, 15th Annual Municipal Finance Conference at Brookings*, Southern Finance Association