Notes on "Risk Topography" (Markus K. Brunnermeier, Gary Gorton, and Arvind Krishnamurthy)

Risk Topography
Markus K. Brunnermeier, Gary Gorton, and Arvind Krishnamurthy
May 31, 2011

a risk topography outlines a data acquisition and dissemination process that informs about systemic risk.

systemic risk:

1. cannot be detected based on measuring cash instruments
2. typically builds up in the background before materialising in a crisis
3. is determined by market participants' endogenous response to various shocks

Outline a system of measuring risks and liquidity on the financial sector and producing a risk topography for the economy:

·      Improve on the standard accounting paradigines on capturing the risk

·      Reveal risk and liquidity pockets on the economy

·      Improve on current macroeconomic models that for the most part do not incorporate a financial sector

Macro models with financial frictions focus on leverage and the dynamics of net worth/capital, limiting the leverage ratio, while models in finance highlight in addition the important role of liquidity

·      Bernanke,Gertler and Gilchrist (BBG) (1999)

·      Kryotaki and Moore (KM) (1997)

·      Brunnermeier and Sannikov (2010)

BBG: the “net worth” of the financial sector is an important state variable in driving macroeconomic phenomena. Net worth is commonly thought of as the equity capital of the financial sector. Thus, in this model, when banks take losses that deplete their equity, they increase the rates charged on loans and/or act back on lending, thuscausing a credit crunch.

KM : add an ingredient to BGG

Agents in the model have collateral that they pledge to raise funds from lenders. Since the market value of agents’ collateral is partly dependent on their financial health, it affects the value of capital. With high leverage, losses deplete capital more dramatically and feedback to further reducing the market value of collateral and so on.

Diamond and Dybvig (1983):

·      It is not just borrowing or leverage of the financial sector that is salient, but rather the proportion of debt that is comprised of short-term demandable deposits.

·      When the financial sector holds illiquid assets financial by short-term debt, the possibility of “counterparty run” behaviour emerges that can precipitate a crisis.
feedback mechanism between capital problems and liquidity problems. Triggered by run by lenders, the sector sells, the sector sells assets whose prices then reflect an illiquidity. The lower asset prices lead to losses that deplete capital, further compromising liquidity.

Brunnermeier and Pedersen(2009):

• Interaction between funding liquidity and market liquidity for modern collateralised (wholesale) funding market.
• Liquidity spirals and “collateral runs”
• An adrerse shock heightens volatility leading to higher margins. This lower fund liquidity and forces institutions to fire-seil their assets, thus depressing market liquidity of assets and increasing volatility further.

Leverage is well-defined in simple stylized models, but is an ill-defined measure in practice.

Liquidity measurement

•   Ignore the future risk
•   Problem of duration
•   Liquidity mismatch index is designed to capture the sensitivities to these kinds of issues, which were particularly captured by any current reporting system.

The choice of scenarios is critical:

• the propagation and patterns of a crisis are similar across events. By collecting data on a core set of factors that are held constant ones time, the data can shed light on the common propagation patterns that underline all financial crisis.
• history suggests that the tigger for crises varios from event to event. Thus at any time the regulator needs to choose factors that are informed by prevailing economic conditions.
• in most case, particular cross-scenarios are of special interest.
• scenarios can include events that have never happened before, that is events that are not in recorded experience.

In the spirit of Kaldor (1961), potential stylized facts concerning the interaction between the financial sector and macro could be :

1.   The risk-deltas tend to display a high coherence with more traditional measures of economic activity, such as output or hours worked. That is, the risk-deltas of different sectors tend to be positively correlated with output.

2.   The risk-deltas in financial firms rises from trough to peak, and falls from peak to trough.

3.   Risk becomes more concentrated over the cycle. This does not necessarily mean that on average risk in individual firms becomes concentrated in certain risk sectors.

4.   Real estate-related risk is the main type of risk for 1 and 2.

5.   The liquidity aggregate is countercyclical, declining as output rises.

6.   The liquidity aggregate is positively related to commercial and industrial loans.

7.   Liquidity risk is procyclical.

8.   Risk-deltas and liquidity are negatively related to the commercial Paper-Treasury Bill spread.

To model systemic risk, ideally we would like data that includes periods of extreme financial crises with large real economic fallouts. These extreme events are rare. However, these are numerous medium-size crises. These crises all reflected significant shocks to financial intermediacies and involved negligible real effects.

It is worth highlighting the commonality with and differences to extreme event analysis on general. Extreme value theory and other methods covering rare events rely critically on certain statistical assumptions. The probability distribution of outcomes deep in the tails is typically assumed. In comparison, macroeconomic modeling involves assumptions about structural parameters that govern behavior both in medium events and in tail events. such modeling is less subject to the Lucas critique. In addition, models of financial market frictions often describe behavior in terms of constraints, rather than beliefs or preferences. It constraints are tighter in extreme events, then it seems plausible that the models may better approximate behavior during such events so that a modeling exercise may perform better than a statistical exercise for extreme events or behaviors is extreme events are avoidable due to limited data.