Systemic risk in derivative
markets: a graph-theory analysis
Objectives
• Empirical study on systemic risk in derivative markets • Integration as a necessary condition for systemic risk
to appear
• Previous literature on integration
• Influence of physical as well as derivative markets • Three-dimensional approach
- Observation time - Spatial integration - Maturity
Selected markets
• Choice directed by:
- Concerns about speculation in commodities and particularly in the crude oil market
- Development of bio fuels - Portfolio management :
- Commodities as a new class of assets
- Weight of the energy products in commodity indexes • Three sectors :
- Energy products
- Agricultural products
1. Energy products
- Two crude oils (US and Europe) - Heating oil (US)
- Gas oil (Europe)
- Two natural gases (US and Europe) 2. Agricultural products
- Soy beans and soy oil (US) - Corn (US)
- Wheat (US)
3. Financial products - Gold
- Interest rates : Eurodollar - Exchange rates : EUR/USD
14 markets
> 655 000 daily futures prices
(settlement prices)
Statistical physics and graph-theory
• Complex evolving system
• Use of recent methods originated from statistical physics • Graph-theory
• Application to financial markets: Mantegna, 1999 • Minimum Spanning Tree (MST) :
Filter the information given by the correlation matrix of price returns
Graph built on correlations transformed into distances The shortest path for a shock transmission
Methodology
• Construction of MST
• Topology of the filtered networks • Evolution of the MST over time
Main results
1. Topology
- Star-like trees in the spatial and 3-D dimensions - Chain-like trees in the maturity dimension
Star-like organization dominates 2. Emerging taxonomy
- Trees organized around the three sectors of activity - Center of the graph : two crude oils
- Links between the agricultural and financial sectors
- Crude as best candidate for the transmission of shocks 3. Integration
- Increases in all dimensions
Section1. Minimum spanning trees :
a correlation based method
Section 2. Topology of the MST
Section 1. Minimum spanning trees
• Synchronous correlation coefficients ρ of prices returns r :
with: F(t), futures prices at t - Correlation matrix C, (N x N)
From correlation to distances
• Non linear transformation
• Distances d defined as follows:
• Distance matrix D, (N x N) • Full connected graph
• Represents all the possible connections between N nodes (vertices), with N-1 links (edges)
Minimum spanning tree
• All the nodes of the graph are spanned • No loops
• Result: links of the MST are a subset of the initial graph • The information space is reduced from (N(N-1)/2) to (N-1) • In this study : shortest path linking all nodes
Section 2. Topology of the MST
• Static analysis on the whole time period • Spatial, maturity and 3-D
Spatial dimension
Finance Agriculture
Maturity dimension
Heating oil – Month 1 to 18
M1
M18
Three dimensions
Finance Agriculture
Allometric properties of the MST
• Quantifying the degree of randomness in the tree • The root is the node with the highest connectivity • Starting from this root, two coefficients Ai and Bi
are assigned to each node i:
Section 3. Dynamical studies
3.1. Correlation coefficients 3.2. Node’s strength
3.3. Normalized tree’s length 3.4. Survival ratios
3.5. Pruning the trees
3.1. Mean correlations (a) and their variances (b)
(3-Dimensions)
3.2. Node’s strength
• Full connected graph
• The node’s strength Si indicates the closeness of one node i to the others:
Energy
US EU
3.3. Normalized tree’s length
• Sum of the lengths of the edge belonging to the MST:
• The more the length shortens, the more integrated the system is
3.4. Survival ratios
• Robustness of the topology over time
• The survival ratio SR refers to the fractions of edges in the MST, that survives between two consecutive
trading days:
3.5. Pruning the trees
• Analysis of inter-market and inter-sectors reorganizations
• Consider only the links between markets, whatever the maturity is considered
Energy
Finance
Conclusions
• Topology of the graphs and its consequences for shock transmission
Star-like / Chain like • Taxonomy of the graph
Crude oil at the center Connections intra / inter sector Maturities and connections (?) • Robustness over time of the trees
Further studies
• Empirical studies
- Enrichment of the graph :
Introduction of transaction volumes and open interest - Analysis in the maturity dimension
- Analysis of shocks