/

Views on non-Normal markets


This presentation explores what we mean by 'extreme events' in the context of investment markets, what might cause these sorts of events, how they can be best analysed and catered for in portfolio construction and how to avoid being overconfident concerning them.

Slides
1Views on non-Normal markets
2Views on non-Normal markets
3Agenda
4Agenda
5Extreme events and fat-tails
6Examples of extreme events
7How really extreme has 2007-09 been
8Agenda
9Visualising fat-tailed behaviour in individual return series
10Example derivation of a quantile-quantile plot
11Fat-tailed behaviour depends partly on timescale
12More periods give more scope for extreme events
13Skew(ness) and kurtosis
14Interpretation via Cornish-Fisher asymptotic expansion
15Flaws in Cornish Fisher (and hence in skew/kurtosis)
16A better approach?
17Agenda
18What causes fat-tailed behavior?
19Time-varying volatility
20Explains some market index fat tails, particularly on upside
21Not just a developed market phenomenon
22A longer term phenomenon too
23Time-varying volatility
24Regime switching
25Regime switching (continued)
26Crowded trades
27Leverage
28Other examples of leverage
29Impact that leverage can have on portfolio behaviour
30Agenda
31Fat tails and portfolio construction
32Fat tails - in joint return series
33Visualisation of joint fat-tailed behaviour
34Quantile-quantile box plots
35Distributional mixtures - again!
36Principal components analysis (PCA)
37Applying PCA to sector relatives
38Adjusting for time-varying volatility in joint return series
39Longitudinal time-varying volatility adjustment
40Cross-sectional time-varying volatility adjustment
41Back-testing time-varying volatility adjustments
42Cross-sectional dispersion of pair-wise combinations
43Impact of time-varying volatility adjustment
44Agenda
45Possible selection effects arising from active management
46Implications for modelling
47PCA vs. ICA
48Including 'meaning' as well as 'noise'
49Selection effects are potentially very important
50Selection effects - Summary
51Agenda
52Portfolio construction
53Portfolio construction - sensitivities
54Portfolio construction - impact of fat tails (1)
55Portfolio construction - impact of fat tails (2)
56Other approaches - (1) distributional mixtures
57Other approaches - (2) lower partial moments
58Estimating lower partial moments
59Views on non-Normal markets
60Agenda
61Appendix A: PCA vs. ICA - Noise vs. Explanation?
62Independent components analysis
63E.g. Non-Normality and Projection pursuit
64Projection pursuit algorithm (1)
65Projection pursuit algorithm (2)
66Independent components analysis (ICA)
67Blending together PCA and ICA
68Identifying Principal Components one at a time
69Blending together PCA and ICA
70Appendix B: System-wide leverage effects
71Does marking-to-market exacerbate leverage effects?
72Systemic risk
73How best to regulate firms?
74Appendix C: Model risk and back-testing
75Model assessment
76Link with calibration
77Ideally model should also fit well 'period by period'
78Look-back bias and in-sample testing
79Look-back bias and out-of-sample testing
80Market consistent risk measurement
81Back-testing in other regulatory frameworks
82Appendix D: Stress Testing
83Stress Testing - Additional drivers
84Different meanings given to the term 'Stress Testing'
85Reverse stress testing
86Stress testing - other observations
87Stress testing - statistical techniques
88References
89Important Information



NAVIGATION LINKS
Contents | Next | Library


Desktop view | Switch to Mobile