Presentation at Delft University's Mathematics and Computer Science department on Financial Networks, on analyzing and modeling them and on the www.fna.fi service.
1. Presentation at TU Delft
2 November 2011
Financial Network Analytics
Kimmo Soramäki
kimmo@soramaki.net
2. “When the crisis came, the serious limitations of existing economic
and financial models immediately became apparent.
[...]
As a policy-maker during the crisis, I found the available models of
limited help. In fact, I would go further: in the face of the crisis, we
felt abandoned by conventional tools.”
in a Speech by Jean-Claude Trichet, President of the
European Central Bank, Frankfurt, 18 November 2010
3. We are talking about systemic risk ≠ systematic risk
News articles mentioning “systemic risk”, Source: trends.google.com
• The risk of disruption to a financial entity with spillovers to the real
economy
• Risk of a crisis that stresses key intermediation markets and leads to their
breakdown, which impacts the broader economy and requires
government intervention
• Risk that critical nodes of a financial network cease to function as
designed, disrupting linkages
-> some chain of events that starts or gets magnified in the finance sector and
makes us all worse off
4. Agenda
• Three components of models
– Topology of financial networks
– System mechanics
– Behavioral dynamics
• How to bring research to policy?
• Financial Network Analytics -software
5. Payment systems
2.50E+15
~1939 tr
2.00E+15
1.50E+15
1.00E+15
5.00E+14
~194 tr
~120 tr
0.00E+00
~5 tr
Annual value (euros) Liquidity need Age of the universe (hours)
Age of the universe (days)
Bech, Preisig and Soramäki
(2008), FRBNY Economic Review, Vol.
6. Topology of interactions
Degree distribution
Total of ~8000 banks
66 banks comprise 75% of value Soramäki, Bech, Beyeler, Glass and Arnold
25 banks completely connected (2006), Physica A, Vol. 379, pp 317-333.
7. System mechanics
Central bank
4 Payment account Payment system
5 Payment account
is debited is credited
Bi Bj
6 Depositor account
3 Payment is settled is credited
or queued
Qi Bi > 0 Di Liquidity Dj Qj > 0 Qj
Market
2 Depositor account Bank i Bank j 7 Queued
is debited payment, if any, is
released
1 Agent instructs
bank to send a
payment
Productive Agent Productive Agent
Beyeler, Glass, Bech and Soramäki
(2007), Physica A, 384-2, pp 693-718.
8. Payment
System
Instructions Payments
Time
Liquidity
Time
Summed over
the
network, instructi When liquidity is high
ons arrive at a payments are submitted
Payments
steady rate promptly and banks
process payments
independently of each
other
Instructions
9. Payment
System
Instructions Payments
Liquidity
Time Time
Reducing liquidity leads to
episodes of congestion
when queues build, and
cascades of settlement
Frequency
Payments
activity when incoming
payments allow banks to
work off queues. Payment
processing becomes
Cascade Length
coupled across the
Instructions
network
10. System mechanics
Central bank
4 Payment account Payment system
5 Payment account
is debited is credited
Bi Bj
6 Depositor account
3 Payment is settled is credited
or queued
Qi Bi > 0 Di Liquidity Dj Qj > 0 Qj
Market
2 Depositor account Bank i Bank j 7 Queued
is debited payment, if any, is
released
1 Agent instructs
bank to send a
payment
Productive Agent Productive Agent
Beyeler, Glass, Bech and Soramäki
(2007), Physica A, 384-2, pp 693-718.
11. Economic behavior
• Example: How much liquidity to post?
• Cost for a bank in a payment system depends on
– Choice of liquidity and
– Delays of settlement
• Banks liquidity choice depends on other banks‟ liquidity choice
• We develop ABM
– payoffs determined by a
realistic settlement process
– reinforcement learning
– look at equilibrium
Galbiati and Soramäki
(2011), JEDC, Vol. 35, Iss. 6, pp 859-
14. Data tsunami
• Digital information is doubling every 1.2 years. Open data, data
science, …
• Regulatory response to recent financial crisis was to strengthen
macro-prudential supervision with mandates for more regulatory
data
• The challenge will be to
understand and analyze the data
• “Analytics based policy”, i.e. the
application of computer
technology, operational research,
and statistics to solve regulatory
problems
Katsushika Hokusai. The great wave off
Kanagawa ~1830
15. Network maps
• Recent financial crisis brought to light the need to look at
links between financial institutions
• Natural way to visualize the financial system
• „Network thinking‟ widespread by regulators
• Mapping of the financial system
has only begun
Eratosthenes' map of the known
world, c.194 BC.
16. Intelligence
• Financial crisis are different and
rare
• Technology, products and
practices change
• Data is not clean, actions are not
„rational‟
• Hard to develop algorithms
• A solution is to augment human
intelligence
(in contrast to AI and algorithms)
18. Screen elements
Access via browser Result panel shows
Explain screen in intranet, internet
or desktop
command output And creates files
(charts, data, etc)
Each command has
‘Visualize’ screen Operation based
different Submit command
shows created on commands
parameters
charts and layouts
Switch between
‘point-and-click’ and Files and database
command line view connections are in
file panels
18
20. Dashboard (concept) The dashboard can
combine multiple
views to the data
on a single screen
It can be available
e.g. on the
intranet and
20
updated daily
21. Command line
All commands can
be submitted using
command syntax History provides an
easy way to make
new scripts for
research or for the
dashboard
All commands
submitted (also from
point-and-click) are
shown in history 21
22. Command line
Scripts can be run
from the scripts panel
or as regular jobs by
the server 22
23. Objectives
• Provide a tool for exploration, analysis and
visualization of regulatory financial data
• Make online financial available for easy
analysis
• Provide a extendible platform for custom
functionality, agent based models and other
simulation models
• Make advances in research available to policy
25. Technical details
• Performance
– Client server architecture allows use of high performance servers, computer
clusters and cloud computing
– High-performance graph engine (neo4j.org)
– Fast client application (Google web toolkit e.g. as in gmail.com)
• Security
– Sensitivity and confidentiality of data creates addition constraints for analysis
– Data is stored on server where it can be protected better (vs analysts desktops)
– Each user accesses FNA with her own account
– SSL encryption of traffic
– Logging and analytics, audit trail
• Integration to corporate IT
– Integration to databases possible
– FNA accounts can be managed centrally by IT (integration to LDAP systems)
– Can run on most application servers
– Modular structure allows easier updates