GSRisk.org Project

Don Berndt • dberndt@usf.edu • dberndt.link/home

Our overall goal for the GSRisk.org Project is to accelerate research into systemic financial risk monitoring approaches that leverage granular data. More traditional economic models capture aggregate behavior while making certain simplifying assumptions. This project aims to create more "bottom-up" approaches that model many fine-grained actions, which lead to system-wide emergent behaviors. Our ultimate goal is design simulation-based tools for financial regulators and practitioners that complement existing economic modeling approaches.

We were fortunate to receive a grant from the Office of Financial Research (OFR) to explore computational approaches for better understanding systemic risk using granular data (administered by the NSF). This grant got us started on what has become a truly interesting stream of research for all involved (with a chance of producing some very useful tools). For a description of the original funded project, have a look at the office NSF Award 1445403 summary. While our focus at the very start was fine-grained computational models, we had not chosen any specific technology or application area. We experimented with various approaches during the first year or so and became convinced that agent-based modeling and simulation was a great fit. In addition, we found tthat the Python programming environment and extensive collection of data science libraries provided an excellent tool set for our approach, reducing the need to assemble disparate data science and financial resources. This early work narrowed our focus and allowed us to make progress on designing our first models.

Contributors

• Don Berndt (PI)
• David Boogers
• Carles Cabre
• Saurav Chakraborty
• James McCart

Agent-Based US Corporate Bond Market Model

The initial agent-based model being implemented as part of the project is focused on the US corporate bond market. Our first experiments were aimed at the potential for bank-run like behavior in mutual funds since these investment vehicles provide liquidity transformation (with potentially short-term investors and longer-term bond investments). Mutual (open end) funds have been growing as a corporate bond market participant and causing some concern since these funds can face redemption-driven feedback loops. We continue to refine this model and develop additional experiments.

Learn More: An Agent-Based Model of the US Corporate Bond Market (BND)

Agent-Based Federal Funds Market Model

The success of our first agent-based model inspired us to choose a second application area with two goals. Again, we looked for another financial market in which there is considerable debate and concerns regarding systemic risk. Clearly, the Federal Reserve took unprecedented action, along with other major central banks, after the financial crisis. In particular, the Fed responded by lowering the target Federal funds rate, providing credit to financial institutions, and eventually pursing actions comprising quantitative easing (QE). As the Federal Reserve seeks to unwind its positions, the impacts of monetary policy normalization are a matter of wide-ranging debate.

In addition, designing a second model allows us to find some of the commonalities in using agent-based modeling approachs for financial markets. The Federal Reserve model is also interesting because is hierarchical, combining a coarse-grained Federal Reserve model linking participating agents through common accounts with a fine-grained model of the inter-bank lending market. We have implemented a couple versions of this model and begun running simulation experiments.

Learn More: An Agent-Based Model of the Federal Funds Market (FED)

Future Work

Currently, we are actively working with both the US corporate bond market model and Federal Reserve funds market model. Some of the next steps include the following.

US Corporate Bond Market Model

• Next up is fine-tuning and further empirical testing of the existing fund redemption price impact model (the basic model used in our first experiments). This includes cross-referencing against findings from the Bank of England (BoE). BoE analysis, Simulating stress across the financial system: the resilience of corporate bond markets and the role of investment funds, using stress simulation finds that weekly bond fund redemptions equivalent to 1% of NAV could increase yields on Investment Grade bonds by 40 basis points.
• In the first model, we assume static trading/investing behavior of the value investor community (long-term real money buy side players). In other words, the model assumes that value investors don’t change investment/trading parameters when faced with redemption-driven price declines (which they view as short-term technical price fluctuations). Effectively, the model assumes value investors act in a counter-cyclical manner. Recent research has questioned the capacity of large buy-side players (such as life insurance companies) to "lean against the wind," especially in light of new constraints on regulatory capital. For example, see the BoE analysis: Insurance companies: amplifiers or the white knights of financial markets?
• Include modeling of the funding flow dynamics for market participants (e.g. hedge funds, dealers) whose activities rely on short-term wholesale funding from institutional cash pool (money market mutual funds or banks). The interconnections between market liquidity and funding liquidity are widely documented. Specific to funding markets, the normalization of monetary policy (unwinding of the Fed’s balance sheet/increase in the short term policy rate) combined with new liquidity constraints on the banking sector (LCR and NSFR) is expected to cause significant changes in institutional money markets (which are critical to the functioning of liquidity providers in bond markets). We look to model the funding decisions and related constraints of all players who rely on wholesale money market financing (dealers, hedge funds, banks) and examine the bond market dynamics under stressed funding markets.

Publications

1. Donald J. Berndt, David Boogers, Saurav Chakraborty and Carles Cabre, “Using Agent-Based Modeling to Assess the Implications of Changing Liquidity Conditions,” Proceedings of the Summer Simulation Conferencei> (SummerSim-SCSC), Berlin, Germany, July 22-24, 2019.
2. Saurav Chakraborty, Donald J. Berndt and David Boogers, “A Hierarchical Model of The Federal Funds Market,” Winter Conference on Business Analytics (WCBA), Snowbird, Utah, March 7-9, 2019.
3. Donald J. Berndt, David Boogers, Saurav Chakraborty, “An Ensemble Agent-Based Model of the Federal Funds Market,” Proceedings of the Summer Simulation Multi-Conference (SummerSim-SCSC), Work-in-Progress Track, Bordeaux, France, July 9-12, 2018.
4. Donald J. Berndt, David Boogers, Saurav Chakraborty and James McCart, “Using Agent-Based Modeling to Assess Liquidity Mismatch in Open-End Bond Funds,” Systems, 5(4), 54, 2017. See https://doi.org/10.3390/systems5040054.
5. Don Berndt, David Boogers, Saurav Chakraborty, and Ratish Dalvi, “Data Visualization and Sonification for Financial Agent-Based Models,” Americas Conference on Information Systems (AMCIS), Boston, MA, August 10-12, 2017. See https://aisel.aisnet.org/amcis2017/HumanCI/Presentations/12/.
6. Donald J. Berndt, David Boogers, Saurav Chakraborty and James A. McCart, “Using Agent-Based Modeling to Assess Liquidity Mismatch in Open-End Bond Funds,” Proceedings of the Summer Simulation Multi-Conference (SummerSim-SCSC), Bellevue, WA, July 9-12, 2017.
7. Saurav Chakraborty, Jordan Gaeta, Kaushik Dutta and Donald Berndt, “An Analysis of Stability of Inter-Bank Loan Network: A Simulated Network Approach,” Proceedings of the 50th Hawaii International Conference on System Sciences (HICSS 50), January 4-7, 2017.
8. Donald J. Berndt, David Boogers and James McCart, “Agent-Based Models of the Corporate Bond Market,” Data Science for Macro-Modeling with Financial and Economic Datasets Workshop (DSMM), ACM SIGMOD/PODS Conferencei>, San Francisco, CA, June 26 – July 1, 2016.
9. D. Berndt, D. Boogers, J. McCart, “Agent-Based Simulation of Corporate Bond Market Liquidity,” Academy of Business Research Fall Conference, Boca Raton, FL, November 10-12, 2015.

Acknowledgements

This material is based upon work supported by the National Science Foundation under NSF Award 1445403 and the Office of Financial Research (OFR), which initiated the funding program. Disclaimer: Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.