When Do Operational Events Become a Systemic Concern: an - PowerPoint PPT Presentation

When Do Operational Events Become a Systemic Concern: an Agent-Based Model of the Large Value Transfer System. Nicholas Labelle 10 February 2009 1

Introduction • Basic assessment method of an outage. Value of payments - March 2008 25 20 Average Can$ billion 15 Max. Min. 10 Outage 5 - 10:30 11:30 12:30 13:30 14:30 15:30 16:30 17:30 0:30 1:30 2:30 3:30 4:30 5:30 6:30 7:30 8:30 9:30 Time 2

Outline 1. discuss the analysis of payment systems 2. demonstrate ABM application 3. propose future improvements and applications 3

1. Agent-Based Model Contribution 1.1 Problems with Standard Simulation Approach a) Fixed order of payment b) Cumbersome input data manipulation 4

1.2 Solutions Offered by Agent-Based Modeling a) Replicate the characteristics of the payment system b) Replicate assumed behavioural responses c) Simulate hypothetical outages The outcome is then measured _ 5

2. Demonstrate ABM Application When do operational events become a systemic concern? 2.1 Payment System Features 2.2 Assumed Behaviour of Banks 2.3 Outage Simulation 2.4 Data 2.5 Parameterization 2.6 Preliminary Results 6

2.1 Payment System Features: LVTS 1. LVTS Tranche 2 only 2. Bilateral Credit Limits (BCLs) 3. Bilateral and Multilateral Risk-Control Test 4. Central queue a) Release algorithm b) Jumbo queue algorithm c) Queue-expiry algorithm 7

2.2 Assumed Behaviour of Banks Payments Internal Central Processed queue queue •We take the BCLs as granted by the participants. •Unallocated collateral is significant in the LVTS. 8

2.3 Outage Simulation •To simulate outages, the simulator intercepts and releases payments at certain times. 9

2.4 Data • Data on March and June 2008 is provided by the Canadian Payments Association (CPA). • 2 files: payments and BCLs. • Distinction between payment submission times versus processing times. 10

2.5 Parameterization Variable Description Value outstart Beginning of the outage. 8:30 duration Duration of the outage. 1 to 9.5 hours Period after which participants stop sending payments to the 35 – 60 – 120 reaction impacted participant until the outage ends. minutes The maximum amount of payments a participant can send maxpaysec 60 payments per minute from its internal queue through the LVTS. 11

2.6 Preliminary Results A) Assumed Behavioural Response Validation Context: • One day in March 2008, a large bank experienced a partial outage from 7:24 to 13:37. • The CPA sent a notification at 8:11. • Participants held back payments at around 8:25. • We can compare the outage payment distribution with simulation results. 12

2.6 Preliminary Results A) Assumed Behavioural Response Validation Value of payments - March 2008 45 40 35 30 Outage Can$ billion 25 Simulations Average 20 Simulations max. Simulations min. 15 10 5 - 0:30 1:30 2:30 3:30 4:30 5:30 6:30 7:30 8:30 9:30 10:30 11:30 12:30 13:30 14:30 15:30 16:30 17:30 Time 13

2.6 Preliminary Results B) Concerns for Operational Outages Context: • All the parameters were set as in the parameterization table. • The month of June 2008 was chosen: 21 business days, 672 simulation runs. • The outage starts at 8:30. • The impacted participant is Bank 1. Warning: • Results depend on the assumed behaviour. 14

2.6 Preliminary Results Can the payment system settle? Proportion of Unsettled Transaction Volume 45% 40% 35% 30% Mean Reaction 35 Mean Reaction 60 25% Mean Reaction 120 20% Max. Reaction 35 Min. Reaction 35 15% 10% 5% 0% 16:30 17:00 17:30 18:00 15 Outage end time

2.6 Preliminary Results How many delays and costs does the outage entail? Average Intraday Queue Value 45 40 35 Mean Reaction 35 30 Can$ billions Mean Reaction 60 25 Mean Reaction 120 20 Max. Reaction 35 15 Min. Reaction 35 10 5 0 9:30 10:30 11:30 12:30 13:30 14:30 15:30 16:30 17:30 Outage end time 16

2.6 Preliminary Results How many delays and costs does the outage entail? Network Average Intraday Queue Value 2.5 2.0 Mean Reaction 35 Can$ billion 1.5 Mean Reaction 60 Mean Reaction 120 Max. Reaction 120 1.0 Min. Reaction 120 0.5 0.0 9:30 10:30 11:30 12:30 13:30 14:30 15:30 16:30 17:30 Outage end time 17

2.6 Preliminary Results How many delays and costs does the outage entail? Average Network Delay Indicators (Reaction 35 min.) 18

3. Improvements and Applications 1. Improvements: - better behavioural rules that are empirically and/or theoretically founded; - methods to validate these assumptions; - search and develop better metrics. 2. Other applications: - change in parameter (SWP, BCL%, central queue); - change in assumed behaviour; - multi-operational outages; - risk assessment of periods and participants; - interaction with other payment systems. 19

Conclusion • ABM: a black box that gives the end-results of our assumptions about participant behaviour. • ABM might make our oversight approach more quantitative and empirical. • Possible preliminary implications: 1. confidence in the system robustness; 2. proactive approach on certain payment system rules related to participants’ reaction to outages; 3. efficient ways to manage payments during outages. 20

References Arciero, L., C. Biancotti, L. D’Aurizio and C. Impenna. 2008. “Exploring agent-based methods for the analysis of payment systems: A crisis model for StarLogo TNG.” Bank of Italy Working Paper No. 686. Arjani, N. 2006. “Examining the Trade-Off between Settlement Delay and Intraday Liquidity in Canada’s LVTS: A Simulation Approach.” Bank of Canada Working Paper No. 2006-20. Arjani, N. and D. McVanel. 2006. “A Primer on Canada’s Large Value Transfer System.” Bank of Canada. <http://www.bankofcanada.ca/en/financial/lvts_neville.pdf> (5 January 2009). Bank for International Settlements. 2003. “A glossary of terms used in payments and settlement systems.” March. <http://www.bis.org/publ/cpss00b.pdf?noframes=1> (5 January 2009). Belisle, C. 2005. “Event Study of LVTS Participants in Situations of Partial Outages.” Bank of Canada. Canadian Payments Association. <http://www.cdnpay.ca/> (5 January 2009). Galbiati, M. and K. Soramäki. “An agent-based model of payment systems.” Bank of England Working Paper No. 352. Lefebvre, S. and K. McPhail. 2003. “Pannes du STPGV: note explicative.” Bank of Canada. FN-03-036. Leinonen, H. and K. Soramaki. 1999. “Optimizing Liquidity Usage and Settlement Speed in Payment Systems.” Bank of Finland Discussion Paper No. 16/99. 21 McPhail, Kim and A. Vakos. 2003. “Excess Collateral in the LVTS: How Much is Too Much?” Bank of Canada Working Paper No. 2003-36.