As part of our agent-based modeling and simulation (ABMS) stream of research (and GSRisk.org Project), we are designing models of financial markets. We are focusing on markets that exhibit signs of systemic risk with vigorous debate by regulators and practitioners. Our first model is being used to investigate issues around the US corportate bond market. Our second model implements aspects of the Federal Reserve system (and funds market) aimed at investigating some open questions regarding the impact of extraordinary post-crisis monetary policies.
The Global Financial Crisis of 2008 and reverberating effects during the GreatRecession had profound impacts on financial markets and the monetary system. In response to the crisis, central banks around the world took unprecedented action in order to ease financial conditions. In the US, the Federal Reserve (Fed) responded by providing credit to financial institutions, lowering interest rates (through the target federal funds rate), providing additional forward guidance, and eventually engaging in a series of large-scale asset purchase programs known as quantitative easing (QE). As a result, the Fed's balance underwent a more than five-fold increase after the crisis, with total assets growing from roughly $850 billion to more than $4.5 trillion and banking reserves expanding from $20 billion to more than $2 trillion. At this point, the Federal Reserve has embarked on a program of monetary policy normalization which implies an unwinding of these extraordinary positions.
The impact of this new phase of monetary policy normalization is a matter of wide-ranging debate. While the Fed has acted cautiously and been quite transparent about its gradual approach toward reducing the balance sheet, significant uncertainty around the effects of the unwind remains. The exact impact on financial markets and institutions depends on the actions (and reactions) of all financial system participants. Policy normalization will unfold from a unique starting point, at a time when the US Treasury is embarking on a significant fiscal stimulus program and within a new untested regulatory environment. There is little or no historical precedent to inform policy makers about the range (and likelihood) of possible scenarios. So, debate is understandable, but what tools are available to help us evaluate policy alternatives?
Clearly, the US Federal Reserve and counterparts around the world will continue to use existing models to better understand unfolding challenges and evaluate proposed policy alternatives. This project looks at adding agent-based modeling and simulation to the toolkit. We are desgining a collection of agent-based models to assess the implications of the reserve drain resulting from the unwind of the Fed's balance sheet. The Federal Reserve model is also has interesting hierarchical structure, combining a coarse-grained Federal Reserve model linking participating agents through common accounts with a fine-grained model of the inter-bank lending market (as decribed below).
Note: This paper is based on a second-generation model that uses shared financial account structures to capture the flow of funds between representative agents, such as the Federal Reserve, US Treasury, banking and non-banking sectors.
Abstract: The Great Recession (2007-2009) and reverberating effects during the global financial crisis had profound impacts on the US Federal Reserve system, especially the Federal funds market. In response to the financial crisis, the Federal Reserve took unprecedented action, along with other major central banks. In particular, the Fed responded by lowering the target Federal funds rate, providing credit to financial institutions, and eventually engaging in a series of large-scale asset purchase programs 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. Clearly, the US Federal Reserve will continue to use traditional economic models to better understand unfolding challenges and evaluate proposed policy alternatives. This paper looks at adding agent-based modeling and simulation to the toolkit. In particular, a combination of coarse-grained and fine-grained models are used to simulate monetary policy normalization.
This figure from the paper shows a more detailed view of the Fed agent balance sheet that highlights some important balance sheet changes, including the reduction in the Fed’s US Treasuries holdings as per the planned runoff. One of the most important simulation results is the decrease in banking sector (BKS) reserves. It is this critical account that is likely to cause stress to appear in the banking sector as reserves become scarce. The drop here is not too severe with banking reserves rising at the outset, but then dips below $2 trillion. However, it is important to note that this is basically a one year simulation (252 ticks) and the runoff schedule starts somewhat gently becoming much larger toward the end of the year. The brief rise in the banking reserves is due to the initial runoff in the Reverse “Repo” Program (over the first two months), see the initial drop in NBS RRP.
See the paper: DJ Berndt, D Boogers, S Chakraborty and C Cabre, “Using Agent-Based Modeling to Assess the Implications of Changing Liquidity Conditions,” Proceedings of the Summer Simulation Conference (SummerSim), 2019. For now, please see the draft: Berndt_SummerSim2019FED_Draft.pdf
We gratefully acknowledge the support of the National Science Foundation ( NSF Award 1445403) and the Office of Financial Research (OFR) in providing funding for this research.