The Application of Visual Analytics to Financial Stability Monitoring

Transcription

1 The Application of Visual Analytics to Financial Stability Monitoring Mark D. Flood 1, Victoria Lemieux 2, Margaret Varga 3 and B.L. William Wong 4 1 Office of Financial Research, U.S.A. 2 University of British Columbia, Canada 3 Seetru Ltd, U.K. 4 Middlesex University, U.K.

2 Content Challenges in Financial Stability Monitoring Visualization Visual Analytics Conclusions Future Research

3 Financial Stability Monitoring Macro-prudential Supervisors Identify new sources of financial instability Maintain situational awareness of developing stresses Make decisions and rules Promote transparency of information to market participants

4 Macro-prudential supervisors challenges 1. Financial system complex, enormous, highly diverse, and constantly changing 2. The system generates voluminous, dynamic and heterogeneous data/information, at ever-increasing rate (but, of varying reliability, completeness and uncertainty) 3. Large, diverse and growing financial and economic models to comprehend threats to financial stability 4. Transparency and accountability in certain regulatory activities can complicate / restrict the choices of tools and approaches

5 Challenges in monitoring financial (in)stability

6 Challenge 1: The Multifaceted Nature of Systemic Risks There is no universally accepted definition of systemic risks No analytical solution will address all the requirements Human judgment will remain an integral part of the analytical process Computational tools such as visualizations can support the human effort

7 Challenge 2: Acquiring The Right Type and Adequate Quality Of Data Need to know the quality, accuracy, reliability and (lack) of coverage of data (e.g. G-20 initiated a process to identify data gaps)

8 Challenge 3: Organizational Issues Macroprudential supervisors: Convert data on institutional and financial market conditions into analyses and formal decisions 1. General monitoring track potentially stressful conditions in the financial sector Situational awareness requires diverse and flexible tools to facilitate the analysts understanding of evolving conditions 2. Formal supervision regulatory enforcement actions Formal decisions with tangible consequences require techniques and tools which are pre-defined, but streamlined for clarity and ex-post accountability

9 Challenge 4: Identifying and Visually Representing Underlying Processes Well designed visualizations are an effective communicator Vital to understand the relationships among multiple entities and concepts

10 Challenge 4: Identifying and Visually Representing Underlying Processes Identify over-arching function relating core concepts and their relationships across different processes in a system Choose appropriate renderings to enable rapid assessment of overall system state, with capability to rapidly interrogate state of sub-systems and components One technique from Cognitive Systems Engineering is the Abstraction-Decomposition Hierarchy 1&2 Unified Model of Systemic Risk 3 provides an abstract description of financial system components and sub-systems, projecting contracts, market behaviors, income, revenues and liquidity into abstract cash-flow patterns and relationships for a composable representation of financial risk. Visually representing financial stability: To unambiguously show relationships integrating crucial financial stability factors, such as credit exposure, leverage, or liquidity 1 Rasmussen, J., Pejtersen, A. M., & Goodstein, L. P. (1994). Cognitive systems engineering. New York: Wiley. 2 Vicente, K. J. (1999). Cognitive Work Analysis: Toward safe, productive, and healthy computer-based work. Mahwah, NJ: Lawrence Erlbaum Associates, Inc., 3 Brammertz, W., Akkizidis, I., Breymann, W., Entin, R., & Rustmann, M. (2009). Unified Financial Analysis: The Missing Links of Finance. Chichester, UK: John Wiley & Sons Ltd.

11 Base Architecture for Mapping Risk Factors to Risk Measures Brammertz, W. (2013), The Office of Financial Research and Operational Risk, in Victoria Lemieux, ed., 'Financial Analysis and Risk Management: Data Governance, Analytics and Life Cycle Management', Springer, pp

12 Potential Solutions Visualization Visual Analytics

13 Visualization A Picture is worth a thousand words The process of analyzing and transforming non-spatial data into an effective visual form to amplify cognition Exploiting the humans visual perception and cognition A highly efficient way for the human to directly perceive data and discover knowledge and insight from it Users know or have decided what they want to look at

14 Classification of Visualization Techniques STATIC DYNAMIC NON-INTERACTIVE No user input after initial rendering, and image does not change, Fixed Example: Newspaper infographic No user input after initial rendering, but image may change Example: Animated GIF INTERACTIVE Ongoing user input, but rendering does not change between input events Example: Dental x-rays Ongoing user input, and rendering may change between input events Example: Video game

15 Static Visualization Pre-composed views of data Multiple static views are used to present different perspectives on the same information

16 Example 1: The Depiction Of Economic Value For Individual Firms And Sectors Concentrated risk exposure is an unsustainably large contingent obligation (or aggregation of them) If triggered would lead to the failure of a financial firm or system Federal Reserve recently introduced Y-14 to collect instituted contract level data Individual firm Basel Committee on Banking Supervision (BCBS) has codified standards for reporting bank capital, riskweighted assets and acceptable leverage Key abstractions for systemic risk analysis - bank failures and bond defaults

17 Example 1a: The Depiction Of Economic Value For Individual Firms This, for example, is a visualization of the People s Bank of China s balance sheet for managing currency, it shows the accumulation of risk over time Risk Accumulation at the Bank Level Koning, J. P. (2012), 'Managed Currency: The People's Bank of China balance sheet since 2002', Technical report, Financial Graph and Art

18 Example 1b: The Depiction Of Economic Value For Sectors Risk Accumulation at the Bank Level Ranking of large banking system in risk concentration (1) Loss-absorption capacity, (2) asset quality, (3) profitability and (4) reliance of wholesale funding Standardisation of accounting data enables cross-firm aggregations Normalised ratios are aggregated nationally and compared for a broad crosssection of countries The closer a banking system is to the centre, the more adjustment it still needs to undertake Global Financial Stability Report, Old Risks, New Challenges, IMF, April 2013,

19 Example 1c: US Bank Suspensions Agriculture-related bank failures Urban and Eastern

20 Example 2: Network Analysis Systemic Interconnectedness FedwireInterbank Payment Network 6,600 nodes and 70,000 links Core Network payment - 75% transferred value 66 nodes and 181 links Potential loss of information and/or context, e.g. anomaly Soramäki, K.; Bech, M. L.; Arnold, J.; Glass, R. J. and Beyeler, W. E. (2007), The topology of interbank payment flows, Physica A, 379,

21 Network Abstraction Aggregate Nodes/Links with Similar Properties 233 nodes Level 1 Level 2 Level 3 Level 4 Level 5 Level 6 Level 7 Level 8 Provide visual clarity Maintain underlying network characteristics 2 nodes Bjørke, J. T., Nilsen, S. and Varga, M. J., Visualization of network structure by the application of hypernodes, International Journal of Approximate Reasoning 51, pp , 2010.

22 Example 3: Revealing Changes Over Time US Treasury yield curve Short term rate driven by fluctuations in investment demand -decline in inflation rates Durden, T. (2010), Visualizing the Past of the Treasury Yield Curve, and Deconstructing the Great Confusion Surrounding Its Future, Internet resource, ZeroHedge.

23 Advantage Static Visualization When alternative views are not necessary nor desired, for example when publishing in a static medium. Disadvantages Limited number of data dimensions when all visual elements are presented on the same surface at the same time Difficult to represent multi-dimensional data fairly, thus could result in some aspects of the data not being explored / exploited or even missed. Users have to decide or know in advance what they want to look at

24 Dynamic Financial Systemic Risk Analysis Visual Analytics Users do not necessary know in advance what to look at or the characteristics of the data There will always be new emergent risks that require new approaches to their detection and identification Must incorporate diverse perspectives Continuous re-evaluation of the evolving structure of the financial system Adapt systemic risk metrics to dynamic and unpredictable changes Perform: Sense-making Decision-making Rulemaking Transparency

25 What is Visual Analytics? Visual Analyticsis the science of analytical reasoning supported by interactive visual interfaces (Thomas and Cook, 2005) 25

26 NLP, text analysis, knowledge management, entity extraction, semantic analysis methods eglatent Semantic Analysis Data analysis Tukey, J. W. (1962). The future of data analysis. Annals of Mathematical Statistics, 33(1), 1-67 What is Visual Analytics? Human sensemaking Klein, G., Philips, J. K., Rall, E. L., & Peluso, D. A. (2007). A dataframe theory of sense-making. In R. R. Hoffman (Ed.), Expertise Out of Context: Proceedings of the Sixth International Conference on Naturalistic Decision Making(pp ). New York: Lawrence Erlbaum Associates. Visualization + Representation Ware, C. (2004). Information visualization: Perception for design(2 nd ed). San Francisco, CA: Morgan Kaufmann Publishers. Vicente, K. J. (1999). Cognitive Work Analysis: Toward safe, productive, and healthy computerbased work. Mahwah, NJ: Lawrence Erlbaum Associates, Inc., Publishers. Assembly of evidence Interactive dynamics Heer, J., & Shneiderman, B. (2012). Interactive dynamics for Visual Analytics. Communications of the ACM, 55(4), Analysis, results, and representationthat are in a tightly coupled perception-action loop Wigmore, J. H. (1913). "The problem of proof". Illinois Law Review 8(2): Toulmin, S. (1958). The Uses of Argument. Cambridge, England: Cambridge University Press.

27 Visual Analytics Users explore the data interactively to gain an understanding of the data Detect the expected Discover the unexpected For situational awareness, decision- and rule-making It is not fixated on any model or dependent on any specific data Both the questions and the nature of the answers are unknown

28 Dashboard -Wire-transfer Monitoring (WireVis) Heatmap keywords for each account Search by example Strings and beads wire transactions Keyword graph Displays the relationships among accounts, time, and keywords within the transactions - overview of the data. User can interactively compare, aggregate, and organize groups of transactions and drill down into and comparing individual records Chang, R.; Ghoniem, M.; Kosara, R.; Ribarsky, W.; Yang, J.; Suma, E.; Ziemkiewicz, C.; Kern, D. & Sudjianto, A. (2007), WireVis: Visualization of categorical, time-varying data from financial transactions, in 'IEEE Symposium on Visual Analytics Science and Technology, 2007', pp

29 RiskVA Consumer credit risk analysis Entity heatmap users entity rules Product comparison Trend analysis Interactive analysis of the 30-day delinquency rate in market across a range of consumers Wang, X.; Jeong, D.; Chang, R. & Ribarsky, W. (2012), 'RiskVA: A Visual Analytics System for Consumer Credit Risk Analysis', Tsinghua Science and Technology 17(4),

30 Dashboard Failed Banks State - total assets Temporal pattern Treemap Overview bank headquarter locations, acquirors, assets, timing of bank failures

31 Interactive Exploration Texas Bank Failures 12 failed banks, their total assets and when were they acquired by the 10 institutions

32 Visualization and Visual Analytics of Failed Banks Static Visualization Exploratory Visual Analytics

33 Conclusions Visual analytics and visualization are promising approaches for monitoring financial (in)stability: detecting, identifying, monitoring and managing threats to global financial stability

34 Future Research 1 Definition of core abstractions Bridging the gap between financial domain abstractions and visualization Definition of canonical algorithms Need for precise and semantically relevant (fast) algorithms that can be embedded in visual analytics Users Interaction with and support from the users during the conception, development and evaluation of the visual analytics algorithms / tools / systems

35 Future Research 2 Provision of standardised data For development, implementation and evaluation of visual analytics tools Evaluation techniques Need to be able to assess the performance and effectiveness of the visual analytics algorithms / tools / systems