2012年-IMF国际货币组织全球_Monitoring_Systemic_Risk_Basedon_Dynamic_Thresholds_36页_2mb

Summary of "Monitoring Systemic Risk Based on Dynamic Thresholds"

Core Content

This paper, authored by Kasper Lund-Jensen and published by the IMF in June 2012, presents a method for monitoring systemic risk in real time using a dynamic threshold approach. The study focuses on the conditional probability of a systemic banking crisis as the measure of systemic risk, and proposes a fixed effect binary response model to forecast this probability based on a set of time-varying risk factors.

Main Contributions

• Real-Time Monitoring: The paper argues that systemic risk can be monitored in real time, as it contains a predictable component that varies over time.
• Dynamic Model Structure: A dynamic binary response model is used to capture the time-varying nature of systemic risk, which allows for forecasts based solely on real-time data.
• Risk Factor Identification: Several risk factors are identified, including:
  • Banking sector leverage
  • Equity price growth
  • Credit-to-GDP gap
  • Real effective exchange rate appreciation
  • Changes in banks' lending premium
  • Degree of bank interconnectedness (measured by the ratio of non-core to core bank liabilities)
• Dynamic Thresholds: It demonstrates that risk factor thresholds are not static but depend on the values of other risk factors, enhancing the realism and effectiveness of policy signals.
• Empirical Validation: The model is validated using an out-of-sample analysis (2001–2010) and shows that it provides reliable early warning signals for systemic banking crises in several economies.

Key Findings

• Predictability of Systemic Risk: The conditional probability of a systemic banking crisis is found to be predictable in real time. The most significant risk factors include credit-to-GDP growth, equity price growth, and banking sector leverage.
• Contagion Effect: There is a significant contagion effect between economies, where a systemic banking crisis in one country can increase the systemic risk level in others.
• Credit Growth and Systemic Risk: Rapid credit growth is associated with higher systemic risk. However, credit growth can also reflect healthy market responses to productivity gains, and the paper emphasizes the importance of distinguishing between these scenarios.
• Equity Price Growth as a Signal: The paper finds that credit growth increases systemic risk more when accompanied by high equity price growth, suggesting that equity price trends can help identify whether credit expansion is healthy.
• Threshold Analysis: The optimal credit-to-GDP growth threshold is not static but depends on the values of other risk factors. For example, the threshold is around 10% if equity prices have decreased by 10% and banking sector leverage is around 130%, but only around 0% if equity prices have grown by 20% and leverage is 160%.
• Model Performance: The binary response model outperforms the signal extraction approach in terms of minimizing type I and type II errors in crisis signaling.

Methodology

• The model assumes that the binary crisis variable $ y_{i,t} $ is drawn from a Bernoulli distribution that depends on $ k $ systemic risk factors $ \mathbf{x}_{i,t - h} $.
• The probability of a systemic banking crisis is modeled as:
  $$
  \Pr(y_{i,t} = 1 \mid \mathbf{x}{i,t - h}; \alpha_i, \boldsymbol{\beta}) = G(\alpha_i + \mathbf{x}{i,t - h}' \boldsymbol{\beta})
  $$
  where $ G $ is a link function (logit, probit, or linear).
• The paper uses an unbalanced annual panel dataset of 68 advanced and emerging economies over the period 1970–2010.
• The model is estimated using maximum likelihood, and the fixed effect estimator is used to account for unobserved country-specific effects without requiring independence between risk factors and country fixed effects.

Risk Factor Definitions

• Credit-to-GDP Growth: Measured as the difference between actual credit growth and its long-term trend, estimated using a backward-looking Hodrick-Prescott filter with a smoothing parameter $ \lambda = 1600 $, reflecting longer financial cycles.
• Banking Sector Leverage: Defined as private credit by deposit money banks as a percentage of demand, time, and saving deposits.
• Equity Price Growth: Reflects changes in stock prices, which can indicate financial instability.
• Real Effective Exchange Rate Appreciation: Captures the impact of international trade competitiveness on systemic risk.
• Lending Premium: The difference between the interest rate charged on loans to the private sector and the risk-free interest rate (treasury bill).
• Bank Interconnectedness: Measured by the ratio of non-core to core bank liabilities. Non-core liabilities include funding from other financial institutions, which increases systemic risk.

Policy Implications

• The model provides a framework for policymakers to derive dynamic thresholds for risk factors, allowing for more responsive and accurate macroprudential policy decisions.
• The early warning signals generated by the model were reliable in predicting the recent financial crisis in several countries, including the U.S. subprime crisis in 2007.
• The paper emphasizes the importance of distinguishing between healthy credit growth and risky credit booms, which can be aided by monitoring equity price trends.

Conclusion

The study contributes to the literature on systemic risk by introducing a dynamic threshold approach that improves the accuracy of early warning signals. It demonstrates that systemic risk is not entirely unpredictable and that a fixed effect binary response model can effectively monitor and forecast it in real time, providing valuable insights for macroprudential policy design.