An Agent-Based Model For Crisis Liquidity Dynamics

An Agent-Based Model For Crisis Liquidity Dynamics by OFR

Richard Bookstaber

Regents of the University of California

Mark Paddrik

Office of Financial Research


Financial crises are often characterized by sharp reductions in liquidity followed by cascades of falling prices. Researchers are making progress in work to understand the levels of liquidity on a daily basis, but understanding the vulnerability of liquidity to market shocks remains a challenge. We develop an agent-based model with the objective of evaluating the market dynamics that lead the market supply of liquidity to recede during periods of crisis. The model uses a limit-order-book framework to examine the interaction of three types of traditional market agents: liquidity demanders, liquidity suppliers, and market makers. The paper highlights the implications of changes in market makers’ ability to provide intermediation services and the heterogeneous decision cycles of liquidity demanders versus liquidity suppliers for crisis-induced illiquidity.

An Agent-Based Model For Crisis Liquidity Dynamics – Introduction

Financial crises are often characterized by sharp reductions in liquidity followed by cascades of falling prices. Researchers are making progress in measurement of liquidity day-to-day, but little is understood on the vulnerability of liquidity to crisis-like shocks. Unfortunately, its difficult to evaluate the vulnerability of markets to the liquidity shocks that accompany such events. Most research on asset liquidity focuses on day-to-day market functioning during non-crisis periods, employing measures based on market statistics such as bid-ask spreads and daily volumes drawn from these typical market periods. But these data provide only limited insight into large liquidations during periods of sharp price declines and related fire-sale dynamics.

Modeling liquidity during market crises is difficult because of the complex, nonlinear dynamics of market participants interacting. As a result, measuring the normal relatively small transactions does not give insight into the impact of larger ones during crisis. In addition, the infrequent and sporadic natures of liquidity declines make for limited knowledge of what defines a large order and where the limits of liquidity supplier lie.

Rather than extrapolating from non-crisis to crisis periods, we propose a model in this paper meant to incorporate market participant dynamics into the liquidity assessment process. Using a similar technique as Kyles (1985) seminal paper, we build a continuous market model using an agent-based modeling frame work to test the concerns of liquidity shocks. We specifically examine two important characteristics of liquidity specific to crises periods that cannot typically be assessed through statistical measures of spread, depth, or resiliency.

The first is the effect of the crisis on the balance sheet of market makers, reducing their ability to take on inventory. Notably, market making issues can be seen on both liquid exchanges and less liquid over-the-counter markets. In electronic exchanges, like those of equity, options, and futures, automated algorithmic trading systems have come to dominate the liquidity supply, acting as market makers in many ways but having much tighter balance sheets and no requirements on making markets in illiquid periods. In many of the less liquid over-the-counter markets, regulatory changes have required bank/dealers to maintain a larger balance sheet, which reduces the incentive  for market making by limiting their ability to trade on their own account against the market making flows. The second is the heterogeneous decision cycles among those in the market, specifically the difference in time-frames between the liquidity demanders that require immediacy and the liquidity suppliers that continue to have a longer-term decision cycle. The 1987 crash and the failure of Long Term Capital Management (Bookstaber, 2007) showed the effect of the difference in the operating speed of liquidity demanders and suppliers.



See full PDF below.