Track 4: New Dimensions in Financial Risk Management Risk Visualization: Presenting Data to Facilitate Better Risk Management 1:30pm 2:20pm Presenter: Jeffrey Bohn Senior Managing Director, Head of Portfolio Analytics and Valuation, State Street Global Exchange
2 R&D and Product Strategy Improving Portfolio Risk Management: Risk Data Visualization March, 2014
The wisdom of Archilochus (Greek lyric poet from Paros, lived 680 BCE to 645 BCE) The fox knows many things, but the hedgehog knows one big thing 3
Setting the stage Challenges, trends and tools 4
Questions often asked of risk management Risk management groups are increasingly asked to join strategy discussions How much concentration/correlation risk do we have in our portfolio? What factors are driving this risk? Can it be hedged? What exposures/positions are driving this risk? Based on a particular macro-economic stress scenario, which sub-portfolios are most impacted? How does our portfolio compare against comparable benchmarks? How does our portfolio today compare to portfolios in the past? How does our portfolio compare to stressed portfolios? What is the optimal possible portfolio? How does my capital supply stack up to my capital demand in a severestress scenario? What portfolio/capital allocations make sense in light of risk appetite, constraints, macro-economic conditions and opportunity set? 5
Portfolio Analysis Workflow Analytics Data warehouse Data Sources Executive Risk Appetite Setting Target Returns vs. Tolerated Risks Maximum drawdown, Allocation Strategy Benchmarking Asset Allocation Risk Parameter Setting Limits, Stress Loss, VaR, DV01, Absorption Ratio, etc. Performance Evaluation Sharpe Ratio, Economic Profit, Capital efficiency, Backtesting, etc. Decision Support Risk analysis & reporting Tactical reallocation / hedging Optimization routines Model Validation / Benchmarking Analyst Compliance reporting Macro scenario analysis ALM analysis Compliance and Regulatory reporting 6
Market Needs Each can benefit from better data visualization By Segment Pension/SWF Insurance Commercial Banks Order Management - - Compliance Reporting Data infrastructure & process Portfolio Risk / Return evaluation Expected Tail Loss / Stress Loss analysis Surplus / Capital management - Macro scenario / Regime shift analysis Risk Appetite Setting Benchmarking Risk factor analysis, attribution - Asset-Liability analysis Re-allocation & "Optimization - Profitability analysis / pricing - Franchise P&L impact analysis - Dashboards Tool implementation Knowledge transfer & training Investment Solutions - 7
Basic constituents of Bottom-up Portfolio Analytics Optimizer Searches for better portfolio allocations / trades subject to constraints and criteria Simulation engine A forward-looking simulation framework iteratively generates a set of random outcomes for the factor realizations, representing different states of the world. Macro scenarios Macro scenarios are translated into specific changes in inputs to a portfolio model via linking functions. Simulation engine Portfolio valuation engine Generates a distribution of horizon portfolio values to depicting the potential evolution of the portfolio. This distribution characterizes the portfolio return and risks, stress loss, etc. Position valuation engine Generates a distribution of horizon position values for each individual position based on its exposure to the risk factors. Proxy Factors Elements attributed to explaining performance (like value or momentum ) that proxy for the Factors. Easier to interpret than the abstract Factors, but may be unnecessary if latent Factors suffice. Latent Factor model A model for the set of explanatory variables that drive the position (and ultimately the portfolio) values. Different states of the world impact the portfolio value via the factors. 8
A word on Factors A consistent terminology is needed Positions equity fixed income Investible positions being valued Proxy Factors Interpretive proxies like value or momentum, often confusingly called Factors Factors Implicit underlying drivers of value..most current frameworks address Proxy level, not Factor level 9
Tools Data visualization and decision-support tools are much improved Multi-dimensional OLAP Linked to real-time data sources Object oriented Data preparation tools becoming more sophisticated 10
Principles of risk data visualization Risk data tend to be defined by outliers Match output to use cases Concentration risk assessment Risk appetite assessment (stress testing) Position-level limits/allocation Prepare for multiple dimensions (e.g., region, sector, asset class, customer type, size) Incorporate drill-down capability Contextualize output (e.g., benchmarks, time series, scenario-based) Use robust statistics (e.g., median, inter-quartile, mean absolute deviation) Use techniques to address data difficulties (e.g., Winsorization, shrinkage) Target near-instantaneous rendering of decision-support output 11
The challenge of moving between portfolio and positions Dashboard examples 12
Major Return and Risk Metrics Breakdown Table by Segment Category Exposure Size Expected Return Volatility Stress Loss 13
Major Return and Risk Metrics Breakdown Table by Segment Low return, high vol & stress loss but small exposure 14
Major Return and Risk Metrics Breakdown Table by Segment 15
Volatility Dashboard (1) By Sector 16
Volatility Dashboard (2) Company Drill-Down for Scatter 17
Volatility Dashboard (3) Geographic Drill-down 18
Volatility Dashboard (4) Drill-down levels: Product > Geography > Sector Vol by Product Vol by Sector Vol by Geography 19
The challenge of multiple dimensions Dashboard examples 20
Expected Returns vs. Volatility by Exposure Size Looks sophisticated but is it useful? Exposure size Exposure size 21
The next slides in this section all come from one chart that provides the ability to select different dimensions. 22
Expected Return vs. Volatility as a scatter Almost the same information as the 3D visualization 23
The same 3-D plot introduce Size as a dimension Expected Returns vs. Volatility by Exposure Size Highest Expected Returns Highest Volatility 24
The same plot introduce Color as another dimension Expected Returns vs. Volatility by Exposure Size Sharpe Ratio as Color High Sharpe Ratios, but small positions OK Sharpe Ratios, and larger position 25
Drill down, if you need the numbers 26
Zoom out aggregate by sector Software & Computer Services 27
Zoom in Display every position 28
Zoom in Display every position Inspect performance clustering High Sharpe Ratio names Return proportional to Volatility High Volatility, Low Return 29
Expected Return vs. Stress Loss Contribution By Sector Color indicates Vasicek Ratio 30
Expected Return vs. Stress Loss Contribution by Company Color indicates Vasicek Ratio 31
Expected Return vs. Stress Loss Contribution by Position Color indicates Vasicek Ratio 32
So which is richer, from a data insight perspective? 33
The challenge of wide-ranging data Dashboard examples 34
The Loss Distribution not very insightful 35
Zooming: The ability to filter data ranges Loss distribution On-the-fly filters 36
Zooming: The ability to filter data ranges Log of the Loss, to eyeball the simulation error 37
Distributions Exposure Size by Sector 38
Stress Loss by Sector Showing the weighted average stress loss as a point 39
Stress Loss by Sector: drilling into the distribution Why not display every single data point? 40
Using heat maps for stress-test output Dashboard examples 41
Stress Loss Contributors by Sector (1) Heat Map 42
Stress Loss Contributors by Sector (2) Heat Map Company Drill-down 43
Drill down to the Exposure Higher Stress Loss is redder, grouped by Sector and Company 44