Liquidity Value vs Contagion in Financial Networks

This article explores stochastic financial networks, focusing on liquidity, the value of key players, and contagion dynamics. It contrasts two approaches: enhancing markets through judicious liquidity injections versus limiting interactions due to contagion concerns.

Introduction to Stochastic Financial Networks

The discussion begins by acknowledging connections to previous lectures and outlining the core themes: - Stochastic Financial Networks: How markets vary over time and in response to shocks. - Economic Environment: Defining the network and its participants. - Market Structures: Differentiating between centralized markets (where all participants are connected) and fragmented markets (where participants are partitioned into clusters). - Ex-ante Liquidity Injections: Examining how pre-emptive liquidity injections can act as buffers against shocks, specifically identifying the most valuable players to receive such injections. - Positive Economics and Empirical Work: Relating theoretical concepts to real-world financial markets, using data from Thai villages to test the theory. - Financial Centrality and Contagion: Comparing the market-making aspect of liquidity injections with the current policy framework that aims to limit interactions due to financial contagion.

A crucial clarification is made regarding terminology: the term "financial centrality" used in the initial drafts of the paper and slides has caused confusion, as it is often associated with contagion. The authors intend to change this to "liquidity value of a player" to avoid misinterpretation, as their measure is distinct from other financial centrality metrics.

Market Disruptions and Shocks

Market disruptions are primarily caused by shocks that limit participation. Several existing literatures address this: - Over-the-Counter Markets (Duffie et al.): Focuses on search frictions, broker-dealer markups, and trade volume. - Monetary Models (Kyotaki and Wright): Involves random matching of traders and the role of assets in facilitating trade. - Random Market Participation: Models where participants leave early or arrive late, used to explain the Federal Reserve's need to inject liquidity. This concept is rooted in Milton Friedman's work on managing liquidity shortages.

Liquidity Value and Social Welfare

Liquidity value (or financial centrality, as initially termed) is defined as the marginal social value of providing additional purchasing power or goods to an agent, contingent on that agent's ability to trade with others. This is a social perspective, aiming to enhance overall social value by increasing not only the recipient's consumption but also that of other agents who trade with them.

The article also touches upon the possibility of endogenous market participation, where agents decide whether to enter a market. Subsidizing an agent who brings liquidity into the market could make it more attractive to others, creating positive externalities.

Interbank Market Example

An illustration of market variation is provided using the interbank federal funds market data from 2006. This data shows how the network of borrowing and lending between banks changes throughout the day, with denser graphs during midday and thinner markets at the beginning and end of the day. This variation can be attributed to both endogenous participation and shocks.

A side note mentions the current state of the federal funds market, which has largely collapsed due to excess liquidity. The repo market has become the primary venue for Federal Reserve monetary policy, involving money market mutual funds lending to hedge funds and pension funds, with collateral backing the loans.

Economic Environment and Risk Sharing

The underlying economic environment is one of risk sharing: - Agents: A finite number of risk-averse agents with concave utility functions. - Incomes: Agents have random incomes, representing a fundamental exogenous shock. An example uses constant absolute risk-averse utility functions and parameterized risk distributions, capturing income correlations. - Market Participation Shocks: An additional shock, represented by a binary vector (0 or 1), indicating whether an agent is in the market (1) or not (0). Agents not in the market consume their own income. - Resource Constraint: Agents in the market can share risk, but aggregate consumption cannot exceed aggregate income. - Community Objective: Maximize the expected utility of agents, weighted by lambda, considering both income and market participation shocks. Consumption choices aim to pool risk.

Network Aspects and Market Formation

The network aspect is introduced by considering how markets are formed: - Centralized Market: All participants are connected. - Bilateral Connections: More complex structures where connections are not universal.

An example of market formation involves a host node randomly chosen to send out invitations. Adjacent nodes are more likely to receive an invite (with probability q), and this probability diminishes with distance (q squared for two nodes away, etc.). This process generates stochastic market participation, which can be influenced by factors like latency or geographic proximity.

More generally, markets can be fragmented, with agents divided into non-overlapping clusters. A probability distribution governs the formation of these clusters. Within a cluster, agents are assumed to be fully interconnected, allowing for optimal risk sharing.

Financial Centrality (Liquidity Value) Definition

Financial centrality (or liquidity value) is formally defined as the marginal social value of an infinitesimal liquidity injection (epsilon) to agent i, made ex-ante (before market and income shocks are determined). This injection is like providing a liquid asset that converts to consumption goods on a 1:1 basis. The goal is to identify key players who bridge across others, especially when shocks are frequent.

The measure is derived from the community objective function: how much it is enhanced by an injection to agent i. This involves taking the derivative of the value function with respect to the injection.

Policy Experiment: Finite Liquidity Injection

If a finite amount A of liquidity is available for injection, the objective is to distribute it among agents to maximize social value. This involves ranking players based on their liquidity value. If A is small, it might be optimal to inject it all into a single highly valued player.

The value of liquidity stems from: - Risk Sharing Effect: Propagates through the risk-sharing network. - Participation Effect: Acts as a subsidy for market participation. - Income Distribution Effect: Changes endowments.

Lagrangian and Shadow Prices

The community objective is maximized subject to resource constraints. The Lagrange multiplier (qs) for each state s represents the shadow price of resources in that state, indicating how much the objective function would increase if the constraint were relaxed.

The liquidity value for agent i is the expected product of agent i's participation shock (xi_i) with these shadow prices (qs). If agent i is not in the market (xi_i = 0), their contribution is zero. If they are in the market (xi_i = 1), their contribution is the value of liquidity in that market, reflected by qs.

Examples and Insights

Identical Agents and IID Incomes

• If all agents are identical, have common utility, and IID incomes, the optimal risk-sharing solution is to give everyone the average income of those participating in the market.
• The marginal utility (shadow price) is the marginal utility of the common utility function evaluated at this average income.

General Case and Prudence

• The liquidity value is related to the mean income, marginal utility, and the prudence (third derivative of utility) of agents.
• Higher variance in incomes increases the ex-ante expected utility of a liquidity injection, as there's more gain from pooling risk.
• It diminishes with the number of agents in the market due to diminishing returns.
• In segmented markets, the division is by the number of agents in the specific cluster.
• The intuition is that liquidity is more valuable when injected into agents who are frequently in the market.

Host-Based Market Formation

• If agent i is chosen as a host with probability 1/n, and markets form through invitations to adjacent nodes, the liquidity value depends on the agent's direct connections and the connections of their neighbors.
• Counterintuitively, an agent might be more valued if their neighbors have fewer connections, as this implies a thinner market when that agent is active, making their liquidity more impactful.
• Valued players are those who are in the market when the market is thin, when risk is high, and when they are highly valued by the community.

General Environment

• The theory extends to heterogeneous preferences, non-IID incomes, and varying Pareto weights.
• Agents are more central if:
  • The market is small.
  • Average volatility is high.
  • Income shocks are positively correlated (indicating high aggregate risk).
  • Their Pareto weights are high.
  • The average agent is poor or has high risk aversion.

Positive Counterparts to Financial Centrality

The theory connects to real-world financial concepts: 1. Market Implementation of Risk Sharing: The optimal risk-sharing contract can be decentralized through financial assets. A "personalized bond" that pays off only when agent i is in the market would have a price exactly equal to our measure of agent i's liquidity value. This links the concept to standard asset pricing formulas. 2. Nash Bargaining: If agents bargain ex-ante over resource allocation, their bargaining power (and thus their share of the "pie") would be higher if their "threat point" (utility in autarky) is higher. This implies that agents with higher liquidity value (social benefit) should receive a larger share. The liquidity value itself depends on these bargaining weights (lambdas), creating a fixed-point problem.

Empirical Work: Thai Villages

The theory is tested using data from Thai villages: - Risk Sharing Regression: A regression of individual consumption on an intercept, individual income (counterfactually expected to be zero in perfect risk sharing), and a time-varying fixed effect (capturing aggregate risk). The intercept (Ai) is of particular interest. - Hypothesis: The higher the intercept (Ai), the higher the agent's liquidity value. - Findings: The empirical results show that agents with higher participation in thin markets and higher variance in their income (or the income of those they interact with) tend to have higher intercepts. This suggests that communities implicitly value and reward individuals who consistently participate in markets, especially when risk is high, by allowing them higher average consumption. This is interpreted as a "premium" for their market-making role.

Contagion vs. Liquidity Value

The article contrasts its approach with the prevailing view of financial contagion: - Contagion Perspective: Views financial crises as analogous to disease transmission in networks. Policy implications focus on managing systemic risk by limiting interactions, especially for large, interconnected players, to prevent the spread of problems. This is an ex-post measure guiding ex-ante policy. - Liquidity Value Perspective: Aims to use ex-ante policy (liquidity injections) to mitigate adverse shocks and maintain market thickness and functionality.

The author suggests that while these two perspectives are distinct, a hybrid policy reflecting both might be beneficial, though such literature is currently lacking.

US Repo Markets (A Missing Third Lecture)

The discussion briefly outlines a third, unpresented model focusing on the US repo markets: - Repo Market Structure: Money market funds (with excess liquidity) lend to hedge funds and pension funds (seeking liquidity) through broker-dealers. Treasuries serve as collateral. - Coordination Problem: The market's thickness depends on coordination among participants, leading to potential multiple equilibria. - Regulation and Contagion: Post-financial crisis regulations (e.g., Basel rules) limit broker-dealers' balance sheet expansion, even for simultaneous borrowing and lending (rehypothecation). This can restrict their participation in the repo market, leading to liquidity shortfalls. - Fed Intervention: When repo rates spike (e.g., 10% in October 2023), the Federal Reserve intervenes as a lender to inject liquidity. - Policy Implications: The author's research, in collaboration with the US Treasury, analyzes repo market data using flow decomposition algorithms to understand interconnectedness, cycles, and chains among dealers. This work aims to assess the impact of regulations and the true risk absorption by dealers, even for off-balance-sheet activities.