III Payment System Oversight Workshop
CEMLA – Central Bank of Guatemala
Guatemala City, 16-18 October 2013

Quantitative O...
Agenda
• Introduction and Background
• Elements of Quantitative Oversight
–
–
–
–
–
–

Network Maps
Relevant Metrics
Real-...
Quantitative Oversight
• The recent financial crisis prompted the need and created the
expectation for regulators to colle...
Systemic Risk

News articles mentioning “systemic risk”, Source: trends.google.com

Not

“The risk that a system composed ...
Dragon King

Black Swan

(Sornette 2009)

(Taleb 2001, 2007)

vs.
Network Models
• The financial crisis brought to light the interconnected nature of
modern financial systems. Academia and...
Network Theory

Financial
Network Analysis
Social Network
Analysis

Network Science
NETWORK
THEORY

Graph & Matrix
Theory
...
Main Premise of Network Theory
Structure of links between nodes matters
• The properties and behavior of a node cannot be ...
Networks Brings us Beyond the Data Cube
For example:

Entities:
100 banks
Variables:
Liquidity, Opening Balance, …
Time:
D...
Network Research
• A growing body of empirical research on financial networks
• Interbank payment flows
– Soramäki et al (...
The Journal of Network Theory in Finance
• Financial institutions and markets are highly
interconnected, but only recently...
Network Visualization
• Network layout
– The relative position of nodes
– Many algorithms are available

• Two (or 3) dime...
Intelligence Amplification
•

Intelligence Amplification vs Artificial
Intelligence
William Ross Ashby (1956) in ‘Introduc...
“In the absence of clear guidance from existing analytical
frameworks, policy-makers had to place particular reliance on
o...
Elements of QO

1.
2.
3.
4.
5.
6.

Network Maps
Relevant Metrics
Real-time Monitoring
Predictive Modeling
Stress Testing
A...
1. Network Maps
Fedwire Interbank
Payment Network, Fall
2001
Around 8000 banks, 66
banks comprise 75% of
value,25 banks co...
More Maps
Federal funds
Bech, M.L. and Atalay, E. (2008), “The Topology of
the Federal Funds Market”. ECB Working Paper No...
Interactive Maps
Exploration of
data
and
Presentation of
information

18
2. Relevant Metrics
Starting point: BSBC Monitoring tools for intraday liquidity management
“It is envisaged that the intr...
Network Metrics
Recently developed financial
system specific metrics:
•

Core-Periphery
–

•

DebtRank
–

•

Craig and von...
Process in Payment Systems
Central bank
Payment system

4 Payment account

5 Payment account

is debited

Bi

is credited
...
Payment
System
Instructions

Summed over
the
network, instructi
ons arrive at a
steady rate

Liquidity
When liquidity is h...
Payment
System
Instructions

Liquidity

Time

Payments

Frequency

Time

Reducing liquidity leads to
episodes of congestio...
How to Calculate a Metric for Payment Flows
Depends on process that takes place in the network!
Trajectory
–
–
–
–

Geodes...
SinkRank
•

Soramäki and Cook (2012), “Algorithm for
identifying systemically important banks
in payment systems”

•

Meas...
26
3. Real-time Monitoring
Outlier Detection
Visual alerts
Visualizations can
highlight banks or their
links that are out of
...
4. Predictive Modeling
• Predictive modeling is the process by which a model is
created to try to best predict the probabi...
Distance from Sink vs Disruption
Relationship between
Failure Distance and
Disruption when the most
central bank fails

Hi...
Simulations
• Methodology to understand complex systems – systems that are large
with many interacting elements and or non...
Application areas

Enhance
understanding of
system mechanics

Evaluate alternative
design features

Why
Simulate?
Stress t...
Data Needs for Simulations
• Historical transaction data
– From interbank payment systems
– At minimum: date, time, sender...
Simulations are Difficult
Major challenge to make them
– tie in to recent/real-time data
– easier to carry out
– easier to...
5. Stress Testing
• Scenario types
–
–
–
–

Historical
Probabilistic
Extreme but plausible market conditions
Worst-case

•...
Stress Simulations
Proper simulations need information on
payment flows between all banks –
feedback effects!
It is a Comp...
Bank Behavior
Payment Systems are “Complex Adaptive Systems”: Interaction of simple
events (one debit and one credit) crea...
The FNA Platform
FNA has developed a proprietary software
platform that runs a wide range of applications
(either cloud-ba...
Automation
• End-to-end Automation of Analytics
– Speed, no set-up costs
– Organizational continuity

• Continuous Calcula...
Interactive Workshop
Agenda

1.

Network Theory and Payment Networks
– Network concepts
– Network metrics

2.

Liquidity in Payment Networks
– ...
Basics of Network Theory
• Constituents
– Nodes (vertices)
– Links (ties, edges or arcs)

• Links can be
– Directed vs und...
Graph Terminology

Trivial Graph

Complete graph, K4

Empty Graph

Complete graph, K7

Undirected Graph

Directed Graph
Directed Weighted Graph
Graphs

Star

Lattice

Ring

44
Graphs

Random (Erdos-Renyi)

Scale-free (Barabasi-Albert)
Note: Scale free property assumes infinite size
45
Bipartite Graph
Example Application Areas

Banks

Countries

Banks

Infrastructures

47
Tree
Example Application Areas
FNA HeavyTails

Node color indicates latest
daily return
- Green = positive
- Red = negative
Nod...
Hands-on: Types of Layouts

Circle

Data, Remittances 2011

Force-Directed

Orbital

50
Common Centrality Metrics
Centrality metrics aim to summarize some notion of importance

Degree: Number of links
Closeness...
Degree

Degree

In-Degree

Out-Degree
Degree Distribution

The topology of interbank
payment flows. Soramäki et
al. Physica A: Statistical
Mechanics and its
App...
Paths, Trails, Walks

A Walk is any free movement along the links
A Trail is a walk where a given link is visited only onc...
Components

55
Strongly Connected Graph
Starting from any red
node allows one to
reach any other red
node with a walk along
the links.
De...
Closeness
• The Farness of a node is defined
as the sum of its distances to all
other nodes
• The Closeness of a node is
d...
Betweenness Centrality
• Measures the number of shortest paths going through a vertex or an arc
• Algorithm
– Calculate sh...
Eigenvector Centrality
Problem: Eigenvector Centrality can be (meaningfully) calculated only for
“Giant Strongly Connected...
PageRank
• Algorithm used by Google to rank web pages. Random surfer model.

• Solves the problem of dead-ends with a “Dam...
Hands-on: Centrality

Exploring centrality
of countries in
Remittances 2011
network

61
Agenda

1.

Network Theory and Payment Networks
– Network concepts
– Network metrics

2.

Liquidity in Payment Networks
– ...
Standard Reporting
• System turnover (value, volume)
–
–
–
–
–

Daily
Monthly peak/low/average
Yearly total
Unsettled paym...
BCBS Monitoring tools
Starting point: BCBS Monitoring tools for intraday liquidity management
“It is envisaged that the in...
(i) Daily maximum liquidity requirement usage

65
Systemic Risk in Payment Systems
• Credit risk has been virtually eliminated by
system design (real-time gross settlement)...
How to Calculate a Metric for Payment Flows
Depends on process that takes place in the network!
Trajectory
–
–
–
–

Geodes...
Distance to Sink
Absorbing Markov Chains give distances:

(66.6%)

(100%)

To B

1

From C

To A

From B

2

From A

From ...
SinkRank
•

Soramäki and Cook (2012), “Algorithm for
identifying systemically important banks
in payment systems”

•

Meas...
SinkRank
SinkRank is the average distance of a unit of liquidity to the sink.
The actual liquidity distribution can be use...
Core-Periphery Structure
•

Craig and von Peter (2010)

•

Interbank markets are tiered in a
Core-Periphery structure
The ...
How good is it? Experiments:
• Design issues
– Real vs artificial networks?
– Real vs simulated failures?
– How to measure...
SinkRank vs Disruption
Relationship between
SinkRank and Disruption

Highest disruption by
banks who absorb
liquidity quic...
Hands-on: Systemically Important Banks

74
3. Real-time Monitoring
Outlier Detection
Visual alerts
Visualizations can
highlight banks or their
links that are out of
...
4. Predictive Modeling
• Predictive modeling is the process by which a model is
created to try to best predict the probabi...
Distance from Sink vs Disruption
Relationship between
Failure Distance and
Disruption when the most
central bank fails

Hi...
Agenda
1.

Introduction

2.

Payment Networks
– Network concepts
– Network metrics

3.

International
Remittances
network
...
What are simulations?
• Methodology to understand complex systems – systems that are large
with many interacting elements ...
Simulations vs analytical models
• Simulations (e.g. Koponen-Soramaki 1998, Leinonen, ed.
2005, work at FRB, ECB, BoC, BoJ...
Short history of LVPS simulations
•

1997 : Bank of Finland
– Evaluate liquidity needs of banks when Finland’s RTGS system...
Tools
• Bof-PSS2
–
–
–
–

Bank of Finland, 1997- (BoF-PSS1)
RTGS, RRGS, Net, many optimization methods
www.bof.fi/sc/bof-p...
Hands-on: Simulation on LSM

Register at www.fna.fi
83
Framework – Liquidity Optimization

Settlement speed

END OF DAY

IMMEDIATE
MIN

Liquidity Used

HIGH

Source: Koponen-Sor...
Framework – RTGS

Instructions

RTGS

Settlement

Queue

RTGS

Settlement

Instructions
Queue
Optimization
85
Framework – LSM

Time critical
payments

RTGS Stream

Settlement

Queue
Instructions
Optimization
LSM Stream
Non-time
crit...
Optimization
What? For example:

How?

• I have 20m and a payment of
30m to settle

• Splitting (eg. CLS)

• I have a 30m ...
Optimization
• What is being optimized?
– Minimize idle liquidity
– Maximize use of queued incoming payments

• Five main ...
Bilateral offsetting with new payments

Bilaterally offset
liquidity needs

New payments aimed at
minimizing multilateral
...
Issues
• What LSM mechanisms to use? LSM simulations
(have a long tradition and) are standard now
• Some options may have ...
Stress Testing
• Scenario models
–
–
–
–

Historical
Probabilistic
Extreme but plausible market conditions
Worst-case

• S...
Stress Simulations

Proper simulations need information on payment flows
between all banks – feedback effects!
It is a Com...
FNA Payment Simulator
• Allow testing of Liquidity saving mechanisms
–
–
–
–
–

Queuing (FIFO + priorities + bypass)
Bilat...
Hands-on: Stress Simulations

Register at www.fna.fi
94
Interactive Workshop Agenda

1.

Network Theory and Payment Networks
– Network concepts
– Network metrics

2.

Liquidity i...
Hands-on: FNA Platform

http://www.fna.fi

96
Thank you
Blog, library and demos are available at www.fna.fi

Dr. Kimmo Soramäki
kimmo@soramaki.net
Twitter: soramaki

97
Upcoming SlideShare
Loading in …5
×

Quantitative Oversight of Financial Market Infrastructures

1,183 views

Published on

Slides from a one day seminar and interactive workshop hosted by Bank of Guatemala and CEMLA in Guatemala City.

Published in: Economy & Finance, Business
  • Kimmo, taas kerran loistava esitys. Kyllä meidän on jossain välissä nähtävä oikein livenä! Leena Wienistä
       Reply 
    Are you sure you want to  Yes  No
    Your message goes here

Quantitative Oversight of Financial Market Infrastructures

  1. 1. III Payment System Oversight Workshop CEMLA – Central Bank of Guatemala Guatemala City, 16-18 October 2013 Quantitative Oversight of Financial Market Infrastructures One Day Interactive Workshop Dr. Kimmo Soramäki Founder and CEO Financial Network Analytics, www.fna.fi
  2. 2. Agenda • Introduction and Background • Elements of Quantitative Oversight – – – – – – Network Maps Relevant Metrics Real-time Monitoring Predictive Modeling Stress Testing Automation of Analysis • Applying Quantitative Oversight • Interactive Workshop using FNA 2
  3. 3. Quantitative Oversight • The recent financial crisis prompted the need and created the expectation for regulators to collect more data about the financial system and to analyze it more efficiently • At he same time, know-how and tools for analyzing large data sets so-called Big Data - have become more prevalent • The continuing reduction in storage costs and increase in computing power has meant that data stored economy wide is growing exponentially. Regulatory data – especially in a data-intensive field such as finance – is no exception • The expectations set by the public and the opportunities created by advances in data analytics will both necessitate and enable a more quantitative approach to the Oversight of Financial Infrastructures Demand – Supply - Expectations 3
  4. 4. Systemic Risk News articles mentioning “systemic risk”, Source: trends.google.com Not “The risk that a system composed of many interacting parts fails (due to a shock to some of its parts)” In Finance, the risk that a disturbance in the financial system propagates and makes the system unable to perform its function – i.e. allocate capital efficiently. Or Domino effects, cascading failures, financial interlinkages, … -> i.e. a process in the financial network 4
  5. 5. Dragon King Black Swan (Sornette 2009) (Taleb 2001, 2007) vs.
  6. 6. Network Models • The financial crisis brought to light the interconnected nature of modern financial systems. Academia and policy-makers have recently developed a stronger awareness of the need for new analytical methods • These new approaches often involve network models, which naturally capture the interconnectedness of the financial system. • In payment system oversight the links may be related to bilateral payment flows, overnight lending relationships, or common participation in different financial infrastructures. • Payment system have led the research on financial networks as the first area where data of the needed granularity has been available from interbank payment systems operated by central banks. 6
  7. 7. Network Theory Financial Network Analysis Social Network Analysis Network Science NETWORK THEORY Graph & Matrix Theory Computer Science Biological Network Analysis
  8. 8. Main Premise of Network Theory Structure of links between nodes matters • The properties and behavior of a node cannot be analyzed on the basis its own properties and behavior alone. • To understand the behavior of one node, one must analyze the behavior of nodes that may be several links apart in the network. • Financial contexts – Trading networks, payment networks, exposure networks – Networks of interconnected balance sheets – Networks of asset dependencies • Topics: Centrality, Communities, Layouts, Spreading and generation processes, Path finding, etc. 8
  9. 9. Networks Brings us Beyond the Data Cube For example: Entities: 100 banks Variables: Liquidity, Opening Balance, … Time: Daily data Links: Bilateral payment flows Links are the 4th dimension to data Information on the links allows us to develop better models for banks' liquidity situation in times of stress 9
  10. 10. Network Research • A growing body of empirical research on financial networks • Interbank payment flows – Soramäki et al (2006), Becher et al. (2008), Boss et al. (2008), Pröpper et al. (2009), Embree and Roberts (2009), Akram and Christophersen (2010) … • Overnight loans networks – Atalay and Bech (2008), Bech and Bonde (2009), Wetherilt et al. (2009), Iori et al. (2008) and Heijmans et al. (2010), Craig & von Peter (2010) … • Flow of funds, Credit registry, Stock trading, Markets, … – Castren and Kavonius (2009), Bastos e Santos and Cont (2010), Garrett et al. 2011, Minoiu and Reyes (2011), (Adamic et al. 2009, Jiang and Zhou 2011), Langfield, Liu and Ota (2012) • More at www.fna.fi/blog
  11. 11. The Journal of Network Theory in Finance • Financial institutions and markets are highly interconnected, but only recently has literature begun to emerge that maps these interconnections and assesses their impact on financial risk and returns • The Journal of Network Theory in Finance (JNTF) is an interdisciplinary journal publishing academically rigorous and practitioner-focused research on the application of network theory in finance and related fields • If you have a paper you would like to submit to the journal, or are interested in contributing in any other way please contact Editor-in-Chief: Kimmo Soramaki – kimmo@fna.fi • For information on subscribing please contact Journals Manager of Risk Journals: Jade Mitchell jade.mitchell@incisivemedia.com 11
  12. 12. Network Visualization • Network layout – The relative position of nodes – Many algorithms are available • Two (or 3) dimensions are available for visualization – One variable in One dimension – Many variables in Many dimensions • All details must convey unique meaning – – – – High data-ink ratio No ‘chartjunk’ Correct dimensions No duplication (from Tufte 2001, Tufte’s Rules)
  13. 13. Intelligence Amplification • Intelligence Amplification vs Artificial Intelligence William Ross Ashby (1956) in ‘Introduction to Cybernetics’ • Visual Cortex is very developed • Technology, products and practices change constantly, market knowledge is essential • Algorithms don’t fare well in periods of abrupt change, algorithms do not think outside the box • Game of Go (from China). Computer programs only get to human amateur level due to good pattern recognition capabilities needed in the game. Build intuition and mental maps, provide tools for scenario thinking 13
  14. 14. “In the absence of clear guidance from existing analytical frameworks, policy-makers had to place particular reliance on our experience. Judgment and experience inevitably played a key role.” in a Speech by Jean-Claude Trichet, President of the European Central Bank, Frankfurt, 18 November 2010 14
  15. 15. Elements of QO 1. 2. 3. 4. 5. 6. Network Maps Relevant Metrics Real-time Monitoring Predictive Modeling Stress Testing Automation 15
  16. 16. 1. Network Maps Fedwire Interbank Payment Network, Fall 2001 Around 8000 banks, 66 banks comprise 75% of value,25 banks completely connected Similar to other sociotechnological networks Soramäki, Bech, Beyeler, Glass and Arnold (2007), Physica A, Vol. 379, pp 317-333. See: www.fna.fi/papers/physa2007sbagb.pdf 16
  17. 17. More Maps Federal funds Bech, M.L. and Atalay, E. (2008), “The Topology of the Federal Funds Market”. ECB Working Paper No. 986. Italian money market Iori G, G de Masi, O Precup, G Gabbi and G Caldarelli (2008): “A network analysis of the Italian overnight money market”, Journal of Economic Dynamics and Control, vol. 32(1), pages 259-278 Unsecured Sterling money market Wetherilt, A. P. Zimmerman, and K. Soramäki (2008), “The sterling unsecured loan market during 2006–2008: insights from network topology“, in Leinonen (ed), BoF Scientific monographs, E 42 Cross-border bank lending Minoiu, Camelia and Reyes, Javier A. (2010). A network analysis of global banking:1978-2009. IMF Working Paper WP/11/74. 17
  18. 18. Interactive Maps Exploration of data and Presentation of information 18
  19. 19. 2. Relevant Metrics Starting point: BSBC Monitoring tools for intraday liquidity management “It is envisaged that the introduction of monitoring tools for intraday liquidity will lead to closer co-operation between banking supervisors and the overseers in the monitoring of banks’ payment behaviour.“ A(i) A(ii) A(iii) A(iv) Daily maximum intraday liquidity usage Available intraday liquidity at the start of the business day Total payments Time-specific obligations B(i) B(ii) Value of customer payments made on behalf of correspondent banking customers Intraday credit lines extended to customers C(i) Intraday throughput 19
  20. 20. Network Metrics Recently developed financial system specific metrics: • Core-Periphery – • DebtRank – • Craig and von Peter 2010, Optimal classification that matches theoretical core-periphery model Battiston et al, Science Reports 2012, Cascading failures -model SinkRank – Soramäki and Cook, Kiel Economics DP, 2012, Absorbing Markov chains World's Ocean Currents NASA Scientific Visualization Studio 20
  21. 21. Process in Payment Systems Central bank Payment system 4 Payment account 5 Payment account is debited Bi is credited Bj 6 Depositor account 3 Payment is settled is credited or queued Qi 2 Depositor account Bi > 0 Di Bank i Liquidity Market Dj Qj > 0 Qj Bank j is debited payment, if any, is released 1 Agent instructs bank to send a payment 7 Queued Productive Agent Productive Agent Beyeler, Glass, Bech and Soramäki (2007), Physica A, 384-2, pp 693-718.
  22. 22. Payment System Instructions Summed over the network, instructi ons arrive at a steady rate Liquidity When liquidity is high payments are submitted promptly and banks process payments independently of each other Time Payments Time Payments Instructions Beyeler, Glass, Bech and Soramäki (2007), Physica A, 384-2, pp 693-718.
  23. 23. Payment System Instructions Liquidity Time Payments Frequency Time Reducing liquidity leads to episodes of congestion when queues build, and cascades of settlement activity when incoming payments allow banks to work off queues. Payment processing becomes coupled across the network Payments Cascade Length Instructions Beyeler, Glass, Bech and Soramäki (2007), Physica A, 384-2, pp 693-718.
  24. 24. How to Calculate a Metric for Payment Flows Depends on process that takes place in the network! Trajectory – – – – Geodesic paths (shortest paths) Any path (visit no node twice) Trails (visit no link twice) Walks (free movement) Transmission – Parallel duplication – Serial duplication – Transfer Source: Borgatti (2004)
  25. 25. SinkRank • Soramäki and Cook (2012), “Algorithm for identifying systemically important banks in payment systems” • Measures how big of a “sink” a bank is in a payment system • Based on theory of absorbing markov chains: average transfer distance to a node via (weighted) walks from other nodes • Provides a baseline scenario of no behavioral changes by banks • Allows also the identification of most vulnerable banks SinkRank on unweighted networks
  26. 26. 26
  27. 27. 3. Real-time Monitoring Outlier Detection Visual alerts Visualizations can highlight banks or their links that are out of normal bounds by marking them with different color or other visual cues E-mail/SMS alerts In a real-time monitoring environment the system can sen e-mails if aspects of the system are out of normal bounds 27
  28. 28. 4. Predictive Modeling • Predictive modeling is the process by which a model is created to try to best predict the probability of an outcome • Given a distribution of liquidity among the banks in the morning, how is it going to be in the afternoon? – What is the distribution if bank A has an operational disruption at noon? – Who is affected first? – Who is affected most? – How is Bank C affected in an hour? • Valuable information for decision making – Crisis management – Participant behavior 28
  29. 29. Distance from Sink vs Disruption Relationship between Failure Distance and Disruption when the most central bank fails Highest disruption to banks whose liquidity is absorbed first (low Distance to Sink) Distance to Sink
  30. 30. Simulations • Methodology to understand complex systems – systems that are large with many interacting elements and or non-linearities (such as payment systems) • In contrast to traditional statistical models, which attempt to find analytical solutions • Usually a special purpose computer program is used that takes granular inputs, applies the simulation rules and generates outputs • Take into account second rounds effects, third round, … • Inputs can be stochastic or deterministic. Behavior can be static, preprogrammed, evolving or co-learning
  31. 31. Application areas Enhance understanding of system mechanics Evaluate alternative design features Why Simulate? Stress testing and liquidity needs analysis Platform for communication among stakeholders
  32. 32. Data Needs for Simulations • Historical transaction data – From interbank payment systems – At minimum: date, time, sender, receiver, value – More data on type of payment, economic purpose, second tier (if any), type of institution, etc. useful • Artificial transaction data – Based on aggregates (possible with Entropy maximization methods) – Based on a network model (defining bilateral flows) – Assumptions • Timing of payments • Value distribution • Correlations – System stability (net flows over longer times)
  33. 33. Simulations are Difficult Major challenge to make them – tie in to recent/real-time data – easier to carry out – easier to understand 33
  34. 34. 5. Stress Testing • Scenario types – – – – Historical Probabilistic Extreme but plausible market conditions Worst-case • Scenarios in BCBS document – Own financial stress : a bank suffers, or is perceived to be suffering from, a stress event – Counterparty stress : a major counterparty suffers an intraday stress event which prevents it from making payments – Customer bank’s stress: a customer bank of a correspondent bank suffers a stress event – Market-wide credit or liquidity stress 34
  35. 35. Stress Simulations Proper simulations need information on payment flows between all banks – feedback effects! It is a Complex Adaptive System A well set-up simulation environment allows exploration of many different stress scenarios Large body of research and policy work on various stress testing has been carried out with data from interbank payment systems 35
  36. 36. Bank Behavior Payment Systems are “Complex Adaptive Systems”: Interaction of simple events (one debit and one credit) creates complex overall behavior A bank’s ability to settle payments (its liquidity risk) depends on its available liquidity and other banks ability to settle payments, which depend … Galbiati and Soramäki (2011), An Agent based Model of Payment Systems. Journal of Economic Dynamics and Control, Vol. 35, Iss. 6, pp 859-875
  37. 37. The FNA Platform FNA has developed a proprietary software platform that runs a wide range of applications (either cloud-based, via intranet, or on individual desktops) for financial data analysis and visualization. The focus is on providing unique analysis capabilities not available from any other solution vendors. The FNA Platform is operational and offers more than 200 functions for modeling, analysing and visualising complex financial data - ranging from graph theory to VaR models. The platform represents more than six man-years of work from highly experienced data scientists and financial market professionals. • FNA’s "secret sauce" is network analysis—algorithms and visualization • Network approaches are the best way for modeling complex systems • FNA leads the way in this new market segment
  38. 38. Automation • End-to-end Automation of Analytics – Speed, no set-up costs – Organizational continuity • Continuous Calculation – Analytics available when needed – Predictions ready when needed • Security – Only available to authorized people • Integration – Analysis available in different application and different formats for online and print 38
  39. 39. Interactive Workshop
  40. 40. Agenda 1. Network Theory and Payment Networks – Network concepts – Network metrics 2. Liquidity in Payment Networks – Liquidity metrics – Modeling liquidity 3. Simulating Payment Networks – Liquidity Saving Mechanisms – Stress testing Payment Networks 4. International Remittances network Central Bank of Guatemala payment data FNA Platform 40
  41. 41. Basics of Network Theory • Constituents – Nodes (vertices) – Links (ties, edges or arcs) • Links can be – Directed vs undirected – Weighed vs unweighted • Graph + properties = Network 41
  42. 42. Graph Terminology Trivial Graph Complete graph, K4 Empty Graph Complete graph, K7 Undirected Graph Directed Graph
  43. 43. Directed Weighted Graph
  44. 44. Graphs Star Lattice Ring 44
  45. 45. Graphs Random (Erdos-Renyi) Scale-free (Barabasi-Albert) Note: Scale free property assumes infinite size 45
  46. 46. Bipartite Graph
  47. 47. Example Application Areas Banks Countries Banks Infrastructures 47
  48. 48. Tree
  49. 49. Example Application Areas FNA HeavyTails Node color indicates latest daily return - Green = positive - Red = negative Node size indicates magnitude of return Bright green and red indicate an outlier return 49
  50. 50. Hands-on: Types of Layouts Circle Data, Remittances 2011 Force-Directed Orbital 50
  51. 51. Common Centrality Metrics Centrality metrics aim to summarize some notion of importance Degree: Number of links Closeness: Distance from/to other nodes via shortest paths Betweenness: Number of shortest paths going through the node Eigenvector: Nodes that are linked by other important nodes are more central, visiting probability of a random process
  52. 52. Degree Degree In-Degree Out-Degree
  53. 53. Degree Distribution The topology of interbank payment flows. Soramäki et al. Physica A: Statistical Mechanics and its Applications 379 (1), 317-333 53
  54. 54. Paths, Trails, Walks A Walk is any free movement along the links A Trail is a walk where a given link is visited only once A Path is a walk where a given node is visited only once A Geodesic Path is the Shortest Path A Cycle is a path starting and ending to the same vertex
  55. 55. Components 55
  56. 56. Strongly Connected Graph Starting from any red node allows one to reach any other red node with a walk along the links. Depending which black node one begins from, one can either reach other black nodes or red nodes
  57. 57. Closeness • The Farness of a node is defined as the sum of its distances to all other nodes • The Closeness of a node is defined as the inverse of the farness • Needs a connected graph (or component) • Directed/undirected • Weighed/un-weighted 57
  58. 58. Betweenness Centrality • Measures the number of shortest paths going through a vertex or an arc • Algorithm – Calculate shortest paths between each pair of nodes – Determine the fraction of shortest paths that pass through the node in question – Sum this fraction over all pairs of nodes • Directed/undirected; Weighed/unweighted Freeman, Linton (1977). "A set of measures of centrality based upon betweenness". Sociometry 40: 35–4 58
  59. 59. Eigenvector Centrality Problem: Eigenvector Centrality can be (meaningfully) calculated only for “Giant Strongly Connected Component” (GSCC) Solution: PageRank
  60. 60. PageRank • Algorithm used by Google to rank web pages. Random surfer model. • Solves the problem of dead-ends with a “Damping factor” to modify the adjacency matrix (S) – Gi,j= Si,j • Effectively allowing the random process out of dead-ends (dangling nodes), but at the cost of introducing error • Effect of – Centrality of each node is 1/N – Eigenvector Centrality – Commonly is used which is used
  61. 61. Hands-on: Centrality Exploring centrality of countries in Remittances 2011 network 61
  62. 62. Agenda 1. Network Theory and Payment Networks – Network concepts – Network metrics 2. Liquidity in Payment Networks – Liquidity metrics – Modeling liquidity 3. Simulating Payment Networks – Liquidity Saving Mechanisms – Stress testing Payment Networks 4. International Remittances network Central Bank of Guatemala payment data FNA Platform 62
  63. 63. Standard Reporting • System turnover (value, volume) – – – – – Daily Monthly peak/low/average Yearly total Unsettled payments Distribution • – Payments value over day – Intraday pattern/throughput (value/volume) – Delays due to lack of liquidity • Technical – Processing times – Settlement mode (by algorithm) • Individual payments – Average/min/max value – Value distribution – Payment type breakdown (interbank, ancillary, cb operations, etc) – Priority (urgent, normal) – Breakdown by bank Intraday statistics • Static information – – – – • Number/types of participants Opening/closing balances Intraday credit limits Bilateral limits Incident reports 63
  64. 64. BCBS Monitoring tools Starting point: BCBS Monitoring tools for intraday liquidity management “It is envisaged that the introduction of monitoring tools for intraday liquidity will lead to closer co-operation between banking supervisors and the overseers in the monitoring of banks’ payment behaviour.“ A(i) A(ii) A(iii) A(iv) Daily maximum intraday liquidity usage Available intraday liquidity at the start of the business day Total payments Time-specific obligations B(i) B(ii) Value of customer payments made on behalf of correspondent banking customers Intraday credit lines extended to customers C(i) Intraday throughput 64
  65. 65. (i) Daily maximum liquidity requirement usage 65
  66. 66. Systemic Risk in Payment Systems • Credit risk has been virtually eliminated by system design (real-time gross settlement) • Liquidity risk remains – “Congestion” – “Liquidity Dislocation” • Trigger may be – Operational/IT event – Liquidity event – Solvency event • Time scale is intraday, spillovers possible
  67. 67. How to Calculate a Metric for Payment Flows Depends on process that takes place in the network! Trajectory – – – – Geodesic paths (shortest paths) Any path (visit no node twice) Trails (visit no link twice) Walks (free movement) Transmission – Parallel duplication – Serial duplication – Transfer Source: Borgatti (2004)
  68. 68. Distance to Sink Absorbing Markov Chains give distances: (66.6%) (100%) To B 1 From C To A From B 2 From A From C (33.3%) To C From A From B (100%) 1
  69. 69. SinkRank • Soramäki and Cook (2012), “Algorithm for identifying systemically important banks in payment systems” • Measures how big of a “sink” a bank is in a payment system • Based on theory of absorbing markov chains: average transfer distance to a node via (weighted) walks from other nodes • Provides a baseline scenario of no behavioral changes by banks • Allows also the identification of most vulnerable banks SinkRank on unweighted networks
  70. 70. SinkRank SinkRank is the average distance of a unit of liquidity to the sink. The actual liquidity distribution can be used in calculating SinkRank Uniform (A,B,C: 33.3% ) C A “Real” (A: 5% B: 90% C:5%) C PageRank (A: 37.5% B: 37.5% C:25%) C B A Note: Node sizes scale with 1/SinkRank B A B
  71. 71. Core-Periphery Structure • Craig and von Peter (2010) • Interbank markets are tiered in a Core-Periphery structure The algorithm determines the optimal set of core banks that achieves the best structural match between observed structure and perfectly tiered structure • • • • Core banks are connected to each other. Periphery banks are not connected to other periphery banks. Core banks are connected to (some) periphery banks. Ben Craig and Goetz von Peter (2010). Interbank tiering and money center banks, BIS Working Papers No 322. 71
  72. 72. How good is it? Experiments: • Design issues – Real vs artificial networks? – Real vs simulated failures? – How to measure disruption? • Approach taken 1. 2. 3. 4. Create artificial data with close resemblance to the US Fedwire system (BA-type, Soramäki et al 2007) Simulate failure of a bank: the bank can only receive but not send any payments for the whole day Measure “Liquidity Dislocation” and “Congestion” by non-failing banks Correlate 3. (the “Disruption”) with SinkRank of the failing bank
  73. 73. SinkRank vs Disruption Relationship between SinkRank and Disruption Highest disruption by banks who absorb liquidity quickly from the system (low SinkRank)
  74. 74. Hands-on: Systemically Important Banks 74
  75. 75. 3. Real-time Monitoring Outlier Detection Visual alerts Visualizations can highlight banks or their links that are out of normal bounds by marking them with different color or other visual cues E-mail/SMS alerts In a real-time monitoring environment the system can sen e-mails if aspects of the system are out of normal bounds 75
  76. 76. 4. Predictive Modeling • Predictive modeling is the process by which a model is created to try to best predict the probability of an outcome • Given a distribution of liquidity among the banks in the morning, how is it going to be in the afternoon? – What is the distribution if bank A has an operational disruption at noon? – Who is affected first? – Who is affected most? – How is Bank C affected in an hour? • Valuable information for decision making – Crisis management – Participant behavior 76
  77. 77. Distance from Sink vs Disruption Relationship between Failure Distance and Disruption when the most central bank fails Highest disruption to banks whose liquidity is absorbed first (low Distance to Sink) Distance to Sink
  78. 78. Agenda 1. Introduction 2. Payment Networks – Network concepts – Network metrics 3. International Remittances network Liquidity in Payment Networks – Liquidity metrics – Modeling liquidity 4. Simulating Payment Systems – – 5. Liquidity Saving Mechanisms Stress Testing Payment Networks Central Bank of Guatemala payment data FNA Platform 78
  79. 79. What are simulations? • Methodology to understand complex systems – systems that are large with many interacting elements and or non-linearities (such as payment systems) • In contrast to traditional statistical models, which attempt to find analytical solutions • Usually a special purpose computer program is used that takes granular inputs, applies the simulation rules and generates outputs • Take into account second rounds effects, third round, … • Inputs can be stochastic or deterministic. Behavior can be static, preprogrammed, evolving or co-learning
  80. 80. Simulations vs analytical models • Simulations (e.g. Koponen-Soramaki 1998, Leinonen, ed. 2005, work at FRB, ECB, BoC, BoJ, BoE) have so far not endogenized bank behaviour – behaviour has been assumed to remain unchanged in spite of other changes in the system – or to change in a predetermined manner – due to the use of actual data, difficult to generalize • Game theoretic models (e.g. Angelini 1998, Kobayakawa 1997, Bech-Garratt 2003) need to make many simplifying assumptions – on settlement process / payoffs – topology of interactions – do not give quantitative answers 80
  81. 81. Short history of LVPS simulations • 1997 : Bank of Finland – Evaluate liquidity needs of banks when Finland’s RTGS system was joined with TARGET – See Koponen-Soramaki (1998) “Liquidity needs in a modern interbank payment systems: • 2000 : Bank of Japan and FRBNY – Test features for BoJ-Net/Fedwire • 2001 - : CLS approval process and ongoing oversight – Test CLS risk management – Evaluate settlement’ members capacity for pay-ins – Understand how the system works • Since: Bank of Canada, Banque de France, Nederlandsche Bank, Norges Bank, TARGET2, and many others • 2010 - : Bank of England new CHAPS – Evaluate alternative liquidity saving mechanisms – Use as platform for discussions with banks – Denby-McLafferty (2012) “Liquidity Saving in CHAPS: A Simulation Study”
  82. 82. Tools • Bof-PSS2 – – – – Bank of Finland, 1997- (BoF-PSS1) RTGS, RRGS, Net, many optimization methods www.bof.fi/sc/bof-pss Free, Support & Training available, Annual workshop • FNA – Financial Network Analytics Ltd. (UK), 2009– RTGS, RRGS, many optimization methods, visual exploration of results, network analysis – www.fna.fi – Free online, License, Support & Training available • Proprietary tools or general purpose programs – Matlab, SAS, Excel, …
  83. 83. Hands-on: Simulation on LSM Register at www.fna.fi 83
  84. 84. Framework – Liquidity Optimization Settlement speed END OF DAY IMMEDIATE MIN Liquidity Used HIGH Source: Koponen-Soramäki (1997). Intraday liquidity needs in a modern interbank payment system - a Simulation Approach , Bank of Finland Studies in Economics and Finance 14. 84
  85. 85. Framework – RTGS Instructions RTGS Settlement Queue RTGS Settlement Instructions Queue Optimization 85
  86. 86. Framework – LSM Time critical payments RTGS Stream Settlement Queue Instructions Optimization LSM Stream Non-time critical payments 86
  87. 87. Optimization What? For example: How? • I have 20m and a payment of 30m to settle • Splitting (eg. CLS) • I have a 30m payment to B, but only 20m on account. B has a 10m payment to me • Bilateral offsetting (eg. CHAPS) • I have no money, a payment of 10m to B, B has a payment of 10m to C, C has a payment of 10m to me • Multilateral netting (eg. CLS) • Partial netting (e.g. Kronos) 10 10 10 30 10 10 20 10 10 10 10 87
  88. 88. Optimization • What is being optimized? – Minimize idle liquidity – Maximize use of queued incoming payments • Five main methods – – – – Splitting Ordering/Bypassing Bilateral netting (offsetting) Multilateral netting (offsetting, circles processing, gridlock resolution) – New payments 88
  89. 89. Bilateral offsetting with new payments Bilaterally offset liquidity needs New payments aimed at minimizing multilateral positions Bilaterally offset, reduced liquidity needs Source: www.lmrkts.com89
  90. 90. Issues • What LSM mechanisms to use? LSM simulations (have a long tradition and) are standard now • Some options may have legal ambiguity (status of split payments) or issues with netting agent in multilateral netting • Banks need to learn to co-use the LSM features. “If you build it they will come” is not likely to work 90
  91. 91. Stress Testing • Scenario models – – – – Historical Probabilistic Extreme but plausible market conditions Worst-case • Scenarios in BCBS document – – – – Own financial stress Counterparty stress Customer stress Market wide credit or liquidity stress • Predictive modeling 91
  92. 92. Stress Simulations Proper simulations need information on payment flows between all banks – feedback effects! It is a Complex adaptive system A well set-up simulation environment allows exploration of many different stress scenarios Large body of research and policy work on various stress testing has been carried out with data from interbank payment systems 92
  93. 93. FNA Payment Simulator • Allow testing of Liquidity saving mechanisms – – – – – Queuing (FIFO + priorities + bypass) Bilateral limits, overdraft limits, opening balances Two-stream operation Bilateral offsetting (first, fifo, best) Queue optimization (Bech-Soramaki) • Allow simulation of Stress scenarios, such as the scenarios in BCBS document “Monitoring Indicators for Intraday Liquidity Management” – – – – (i) Own financial stress (ii) Counterparty stress (iii) Customer stress (iv) Market wide credit or liquidity stress • Any functionality can be implemented in a custom Payment Simulator application 93
  94. 94. Hands-on: Stress Simulations Register at www.fna.fi 94
  95. 95. Interactive Workshop Agenda 1. Network Theory and Payment Networks – Network concepts – Network metrics 2. Liquidity in Payment Networks – Liquidity metrics – Modeling liquidity 3. Simulating Payment Networks – Liquidity Saving Mechanisms – Stress testing Payment Networks 4. International Remittances network Central Bank of Guatemala payment data FNA Platform 95
  96. 96. Hands-on: FNA Platform http://www.fna.fi 96
  97. 97. Thank you Blog, library and demos are available at www.fna.fi Dr. Kimmo Soramäki kimmo@soramaki.net Twitter: soramaki 97

×