High Performance Computing & Big Data for Finance Identifying Returns, Managing Risk and Reporting with Transparency

Transcription

1 HPC Finance Symposium London, March 14 th 2016 High Performance Computing & Big Data for Finance Identifying Returns, Managing Risk and Reporting with Transparency Erik Vynckier

2 Agenda Big Data Tools for Algorithmic Investing Scenarios for Portfolio Construction Balance Sheet Simulation and Capital Modelling Liquidity Planning Clearing and Collateral Regulatory Reporting Monitoring Systemic Risk Powerful Platforms Conclusions 1

3 What Does the Investor Get Compensated for? Risk Sharing Compensation for bearing risks or insuring against risks Structural Constraints Compensation for having fewer investment constraints Information Processing Compensation for processing information better Traditional Beta Specific Risk Events Liquidity Agency Motivations Investor Constraints Underreaction to Information Overreaction to Information Equity Risk Premium Commodity FX Carry Liquidity Low Beta Momentum Carry Quality Fine-tuning and back-testing investment strategies will benefit from Big Data Analytics 2

4 Unconstrained Thinking: Access a Broader Set of Risk Premiums Traditional Assets Carry Strategies Value Strategies Momentum Strategies Other Strategies Theme Representative Metric Intuition Ex Ret Vol SR Eq Corr Bd Corr Equities 4.9% 14.1% REITs 7.4% 15.4% HY Spreads 4.1% 11.6% (0.44) Government Bonds 2.4% 5.5% FX Carry Interest Rate Differential High carry currencies outperform 4.3% 7.5% (0.15) FI Carry Slope of Yield Curve High carry bonds outperform 3.2% 2.9% Comm Carry Roll of Futures Curve Backwardated commdity futures outperform 6.6% 10.9% EM FX Carry Interest Rate Differential High carry currencies outperform 7.0% 6.1% (0.19) EQ Capital Use Total Yield Efficient capital use results in higher returns 2.7% 5.6% 0.49 (0.02) (0.02) FX Value PPP/FX Cheap currencies outperform 1.6% 6.5% (0.12) FI Value Level of Yield Cheap bonds outperform 0.8% 2.5% Comm Value Past 5y Performance Cheap commodities outperform -0.4% 12.5% (0.04) EM FX Value PPP/FX Cheap currencies outperform 4.3% 3.9% EQ Current Value EBIT/EV Cheap stocks outperform 3.2% 6.0% 0.53 (0.03) 0.00 EQ Deep Value Price/Book Cheap stocks outperform 5.0% 7.3% FX Momentum Curreny Momentum Recent outperformance persists 3.3% 7.9% 0.42 (0.05) (0.04) FI Momentum Bond Momentum Recent outperformance persists 1.4% 3.0% 0.47 (0.13) 0.13 Comm Momentum Commodity Momentum Recent outperformance persists 6.5% 12.6% 0.51 (0.03) (0.08) EM FX Momentum EM Currency Momentum Recent outperformance persists 1.6% 5.2% 0.31 (0.08) (0.07) EQ Momentum Stock Price Momentum Recent outperformance persists 7.4% 10.3% 0.72 (0.09) 0.00 EQ Profitability ROE Highly profitable companies outperform 2.1% 6.2% 0.34 (0.06) 0.00 EQ Quality Accruals Higher quality companies outperform 5.1% 6.0% (0.01) EQ Size Market Cap Small companies outperform 5.2% 8.0% (0.01) EQ Beta Beta High beta stocks outperform 2.7% 9.6% EQ Risk Residuals High risk stocks outperform -1.2% 9.7% (0.12) 0.36 (0.03) As of July 31, 2014 Source: AllianceBernstein 1 Income for traditional assets(except HY) represented as total income for 100% long strategy 2 Income(Excess Return) for HY Spread represented as spread income(spread return) above the yield(return) of a government bond 3 Income for L/S strategies represented as ex-ante income generated by the L/S strategy only (does not include cash) 4 SR is the Sharpe ratio for traditional assets and Information ratio for Strategies 3

5 Risk Premiums Have Delivered 4

6 Risk Premiums Should Be Constructed as Pure Factors 5

7 Risk Factors Source of Common Exposures Across Various Asset Classes Building Blocks for Multi-Asset Solutions Carry Value Momentum Quality Bonds Commodities Currencies Equities Volatility + Allocating to Risk Factors Approach For Example Pros Cons Static Risk Parity Lower fees Irresponsive to market conditions Thematic Quality & Yield Diversification of other exposures Dynamic Factor Rotation Responsive to market conditions Irresponsive to market conditions Higher fees Focus on transaction costs and execution slippage to enable implementation 6

8 7 Macro- and Micro-Economic Data for Risk Factor Strategies Unstructured Content Structured Data Full Edition DJ Edition Granular Data Indicators Cloud Analytics Platform Entity Recognition Relevance Scoring Web Edition Event Detection Novelty Scoring API FTPS Sentiment Scoring FTPS HTTPS HTTPS

9 8 Bond Futures Strategy from Economic News Data Thalesians ( Bond Over Big Data Jan 2015) - Example trading rules: Go long futures when economic news indicator is bad and falling Go flat futures when economic news indicator is at an extreme Otherwise remain in the previous position

10 Agenda Big Data Tools for Algorithmic Investing Scenarios for Portfolio Construction Balance Sheet Simulation and Capital Modelling Liquidity Planning Clearing and Collateral Regulatory Reporting Monitoring Systemic Risk Powerful Platforms Conclusions 9

11 Efficient Frontier Optimization with Sequential Quadratic Programming One-for-one porting of mathematical techniques from the physical sciences into investment practice has often produced catastrophic outcomes 10

12 Percent Historically Low Yields Pose Challenges for Bond Investors Range of Compound Growth Rates over 10 Years (in Pounds) 20 Normal 10th Percentile Normal Median Dec 31, 2014, Median Normal 90th Percentile (10) UK Inflation UK Cash Global 7-Year Sovereign Bonds Global 7-Year Invest. Grade Global 7-Year High Yield Global Public Real Estate Global Large- Cap Global Small- Cap Emerging Markets View of the future by means of probabilities, not point forecasts As of December 31, 2014 Returns hedged into GBP and reported in GBP Data do not represent past performance and are not a promise of actual returns or range of future results. Source: AB 11

13 Stochastic Economic Scenario Generator Why Focus on Modelling the Economic Building Blocks of Return? Global real-world scenarios Multiple horizons, multi-currency Economically underpinned Consistent across asset classes Equity Return on Equity Credit Alternatives Spreads by Rating Dividend to Book Macroeconomy Sovereign Bonds Currency Exchange Rates Ratings Change Price to Book Return on Equity Profitability Profitability Payout Valuation Book Growth = ROE D/B Price Appreciation = Book Growth + Valuation Change Dividend Yield = D/B B/P = D/B / P/B Typically 10,000+ scenarios are needed for converging, reliable outcomes 12

14 Monte Carlo Simulation in Finance often Embarrassingly Parallel Entrenched method for modelling financial risks and deriving metrics, such as option prices, risk sensitivities (Greeks) and Value at Risk 13

15 Allocation Dynamic Stochastic Programming: Optimizing across Scenarios Defined Contribution Glide Path Adventurous Managed Balanced Managed Cautious Managed Defensive Managed 100% 80% Target Years Short Duration Index-Linked Gilts Long Duration Index Linked Gilts Short Duration Gilts Medium Duration Gilts 60% 40% 20% Long Duration Gilts UK Long Duration Corporate Bonds Global Corporate Bonds Global Property Emerging Market Equities 0% 40 yrs 30 yrs 20 yrs 10 yrs yrs Global Developed Market Equities UK Equities 2050 to to to to to to 2004 How should an investor allocate money to investments before and in retirement to increase wealth and consumption while reducing the risk of running out of savings? 14

16 Agenda Big Data Tools for Algorithmic Investing Scenarios for Portfolio Construction Balance Sheet Simulation and Capital Modelling Liquidity Planning Clearing and Collateral Regulatory Reporting Monitoring Systemic Risk Powerful Platforms Conclusions 15

17 Regulations regarding Capital, Derivatives and Trading Regulatory Overview A Global Picture CRR / CRD IV OTC Derivatives capital costs: Completed Basel III Solvency II Leverage constraint NSFR Capital requirement LCR Pillar 1: Valuation and risk-based capital requirements Pillar 2: Governance and risk management requirements, Pillar 3: Supervisory reporting and public disclosure. Dodd-Frank Act Title VII Trade Reporting: Completed Mandatory clearing: Completed Mandatory Execution: Completed EMIR Trade Reporting: Completed Mandatory clearing: anticipated March - December 2016 Margining requirements: phase in from Dec 2016 to Dec 2019 MiFID II / MiFIR Market structure, execution venues, trading transparency requirements: anticipated January

18 Applying Scenarios to Risk & Capital Management Internal Models for Solvency II/Basel III Capital Exposures Market risks Insurance risks Stresses Correlations Internal model Scenario generator High Performance Computation Businesses and assets which are poorly covered by the standard model Accurate tail risk metrics require a massive number of scenarios! 17

19 Asset and Liability Reporting Templates for Solvency II Pillar 3 D1 - line-item reporting of assets held for insurer D2 - derivatives (trades and open positions) D3 - comprehensive insurer report (including bank deposits & self-occupied property) D4 - look-through table (fund-of-funds) awaiting clarification on materiality D5 - securities financing (securities lending and borrowing, repurchase agreements) D6 - collateral (posted and received) XBRL format for Big Data reporting, searching and analysing Supervision focusing on systemic effects now a real prospect 18

20 Agenda Big Data Tools for Algorithmic Investing Scenarios for Portfolio Construction Balance Sheet Simulation and Capital Modelling Liquidity Planning Clearing and Collateral Regulatory Reporting Monitoring Systemic Risk Powerful Platforms Conclusions 19

21 Persistent Illiquidity Premium in Infrastructure, Real Estate and SME Loans Range of Indicative Illiquidity Premiums (basis points) Senior Unsecured Private Placement Senior Infrastructure Debt Junior Infrastructure Debt Senior Secured Unitranche Loans Transitional Commercial Real Estate Loans As of December 2014 Source: AB 20

22 Liability Driven Investment Investments in Assets and Derivative Overlays Requiring Collateral Many assets require derivatives overlays to match liabilities but few assets provide eligible collateral or cash for clearing 21

23 Best-in-Class Analytics CSA Discounting Example accounting for Collateral Optionality Market data: - SONIA - Fed Funds - EONIA - Cross currency basis swaps GBPUSD and EURUSD - Spot GBPUSD and EURUSD - GBP, USD, EUR Libor 3m - GBP Libor 6m - Term structures out to 60Y 22

24 Best-in-Class Analytics Revaluation of GBP 4 billion under different CSAs GBP 4 billion notional in 400 fixed receiver swaps maturing over deeply in the money since traded in 2008 Approximately GBP 800 million of mark-to-market interest in re-couponing? Mix of bilateral CSAs accurate valuations and risk sensitivities matter in trading! Value SONIA US Fed Funds EONIA Libor GBP6m DV01 Libor USD3m Libor EUR3m Libor GBP3m GBPUSD spot EURUSD spot FullyCollateralizedInGBP 829,450, ,077-4,618,226 FullyCollateralizedInUSD 820,412, , ,554,093-8, , ,481 FullyCollateralizedInEUR 806,398, ,860 2, ,453,480-7, , , ,379 MulticurrencyOISCollateral 803,071, , ,617,132-4,941 14, , , ,722 FullyCollateralizedInLiborGBP6m 812,917, ,186-5,045,374 FullyCollateralizedInLiborUSD3m 820,910, , ,558,806-8, , ,631 FullyCollateralizedInLiborEUR3m 807,003, ,779 2, ,458,799-7, , , ,275 MulticurrencyLiborCollateral 802,955, , ,579,365-4,900 30, , , ,857 Variations of GBP 27 million depending on collateral policies Risk sensitivities of approximately GBP 5.3 million smeared out over different market parameters You require the same analytic capabilities as your brokers! 23

25 Rates Yield Curve Evolution Sample Scenario Yield Curve Evolution Jan22 Jan21 Jan20 Jan19 Jan18 Jan17 Jan16 Jan15 Jan14 Observation Date Tenor (Months)

26 Mark-to-Market of Each Swap at Every Step for One Scenario 25

27 Mark-to-Market of Entire Swap Portfolio for All Scenarios Multiple time-step, path-dependent simulations across the complete book of derivatives, collateral assets and counterparties over the entire life of the trades 26

28 Portfolio Exposure Across Scenarios indicates Potential Liquidity Sink Potential Future Exposure (PFE) the maximum expected counterparty exposure or collateral requirement within a specified high level of confidence Expected Positive Exposure (EPE) the Expected or average Exposure up to a given future date t. Credit Valuation Adjustment (CVA) the adjustment to the price of a derivative to take into account counterparty credit risk = Expected Positive Exposure x Probability of Default x Loss Given Default 27

29 EPE & CVA Estimations for a GBP 4 billion Receiver Swap Book Market Value of receiver swap book approximately GBP 800 million Counterparty Value Adjustment GBP 91 million Would be charged as approximately 20 bp adjustment on the quoted swap rate Some companies post collateral even if not obliged by law 28

30 Agenda Big Data Tools for Algorithmic Investing Scenarios for Portfolio Construction Balance Sheet Simulation and Capital Modelling Liquidity Planning Clearing and Collateral Regulatory Reporting Monitoring Systemic Risk Powerful Platforms Conclusions 29

31 Cray Supercomputing? 30

32 GPU Multi-Core Processors: Cores per Chip 31

33 Computational Pipeline for Dataflow: Use All Si All of the Time 32

34 Computing with the Cloud: Capacity Burst on Demand 33

35 Agenda Big Data Tools for Algorithmic Investing Scenarios for Portfolio Construction Balance Sheet Simulation and Capital Modelling Liquidity Planning Clearing and Collateral Regulatory Reporting Monitoring Systemic Risk Powerful Platforms Conclusions 34

36 Scope for Innovative Technology in Banking, Investment and Insurance Affordable computational platforms and accessible big data analytics are here & now A variety of sensible applications are easily identified! Investment: information processing, portfolio construction, implementation, dynamic hedging, reporting Balance sheets: risk and capital management, liquidity planning, strategic planning Regulators ask for more risk reporting, are on the cusp of systemic monitoring Do we have the right staff, the right organization, the optimal value chain? Are our organizations fit for harnessing technology? 35

37 Merging Quantitative Finance and Technology Distributed Ledgers for Immediate Collateral and Minimal XVA Cost Algorithmic Differentiation, Initial Margin Computation, and Numerical Optimization Big Data for Trading and Risk Management Fast Analytics and High Performance Computing WBS London Canary Wharf Thu 21st - Fri 22nd April

38 Clearing and Collateral: Read Up and Connect with Colleagues Clear Path Analysis study: Collateral Management for Institutional Investors International Securities Services: Collateral Management, Insurance Asset Management