RISK MODELLING FLOW CHART VAR SUMMARY

Equity Portfolio Equity spot, volatility

RISK FACTOR REDUCTION USING PCA optional
MAPPING

Interest Rate Portfolio IR Curve, CS Curve, IR Vol

RISK

Link Portfolio Loss to Risk Factor returns P2

FX Portfolio Fx Spot, Fx Vol
Commodity Portfolio Commodity Spot, Vol

VALUATION LAYER

Local Valuation Full Valuation

∆𝑽 = 𝒇𝒏(𝑹𝒇 𝒔𝒆𝒏𝒔𝒊𝒕𝒊𝒗𝒊𝒕𝒚, 𝑹𝒇 𝒔𝒉𝒐𝒄𝒌) ∆𝑽 = 𝑽@𝑹𝒇 𝑺𝒉𝒐𝒄𝒌𝒆𝒅 − 𝑽@𝑹𝒇 𝒐𝒓𝒊𝒈𝒊𝒏𝒂𝒍

Equity Monte Carlo Simulation suited for exotic options

having path dependency that do not have analytical
➢ 𝑳𝒊𝒏𝒆𝒂𝒓 ∆𝑽 = 𝑽 × 𝒓 Worst case
form. Ex – MBS, Callable Bond, Puttable Bond
➢ Delta Normal - ∆𝑽 = 𝒅𝒆𝒍𝒕𝒂 × ∆𝑺 = 𝒅𝒆𝒍𝒕𝒂 × 𝑺 × 𝒓 Worst
➢ Delta Gamma approximation – case
𝟏
∆𝑽 = 𝒅𝒆𝒍𝒕𝒂 × ∆𝑺 + × 𝒈𝒂𝒎𝒎𝒂 × ∆𝑺𝟐
𝟐

Bonds
𝟏 𝟐
∆𝑽 = 𝑴. 𝑫 × 𝑷 × ∆𝒀 + × 𝑪𝒐𝒏𝒗𝒆𝒙𝒊𝒕𝒚 × 𝑷 × ∆𝒀
𝟐
Worst case

𝑷𝟎
𝑷𝟏 𝑷𝟎
𝑷𝟏

WORST CASE DETERMINATION LAYER

𝒓𝟎 𝒓𝟏
W W W . P E A K S 2 T A I L S .C O M
𝒓𝟎 𝒓𝟏
Return metric on risk factors to be moduled -
1. Absolute Return [∆𝒚]
2. Relative Return [r (simple)]
3. Log Return [r]

Generate Risk Factor Return distribution

Parametric Methods Historical Simulation Monte Carlo Simulation

Parametric Normal future = past

6%
z 𝝁
M1 𝝁 − 𝒁 × 𝝈 (returns) Percentile
𝝁 + 𝒁 × 𝝈 (loss) IMPROVEMENTS
P1 = Age Weighted
(hybrid)
yes no 𝝀𝒕−𝟏
Historical 𝟏−𝝀 M1 Brownian Motion
IV
Returns 𝟏 − 𝝀𝒌
𝒅𝒔
w = 𝒖𝒅𝒕 + 𝝈 𝒅𝒕 × 𝒛
𝒔
Drift Shock
∑𝒓𝟐
Weigh historical 𝝈= or w
𝒏
returns equally 𝟐 ∑𝒓𝟐
𝝈 = Volatility Weighted
𝒏−𝟏
𝝈𝒄
𝑬𝑾𝑴𝑨 = 𝜶𝒓𝟐𝒏−𝟏 + 𝜷𝝈𝟐𝒏−𝟏 𝑨𝒅𝒚 𝒓𝒚 = 𝒓𝒕 ×
More weight 𝝈𝒐
𝑨𝑹𝑪𝑯 = 𝒘 + 𝒅𝟏 𝒓𝟐𝒏−𝟏 − 𝒅𝟐 𝒓𝟐𝒏−𝟐 𝑺𝑻 𝟏 𝟐
to recent obs. 𝒍𝒏 = 𝒖− 𝝈 𝒕+𝝈 𝒕×𝒛
𝑮𝑨𝑹𝑪𝑯 = 𝒘 + 𝑺𝟎 𝟐
Kernel Density est. 𝟏
𝒖− 𝝈𝟐 𝒕+𝝈 𝒕×𝒛
Similarity to 𝒙−𝑿
𝟏 𝑺𝑻 = 𝑺 𝟎 𝒆 𝟐
−𝟐
𝑲 𝒙, 𝑿 = 𝒆 P2
economic state 𝒏 𝒅𝒓 = 𝝀 𝑲 − 𝒓𝟎 𝒕 + 𝝈 𝒕 × 𝒛
Normalise Kernel
(Term Structure Models)
W W W . P E A K S 2 T A I L S .C O M
M2 Parametric Log Normal Filtered Historical Simulation
Bootstrapping
(𝟏 − 𝒆𝒖−𝒛×𝝈 )

M2 EVT (Extreme Value Theory) - Fat tails M1 Normal Models

(Regime Based Models)
Block Maxima Peak over Thresholds

u1 u2
𝝈𝟏 𝝈𝟐

Generate Portfolio Loss distribution Fed into capital models

Basel III – 10 day VaR @ 99%
CALCULATE RISK MEASURE FRTB – LH ES @97.5%

Value at Risk Expected Shortfall

Max possible loss under Loss under abnormal circumstances

Properties Normal circumstances (Average of Loss beyond VaR)
ES
z u
VAR Monotonicity

𝑬 𝒀 > 𝑬 𝑿 , 𝑷 𝑿 > 𝑷(𝒀)

Translation Invariance

𝑷 𝑿+𝑪 =𝑷 𝑿 −𝑪
ES
Positive Homogeneity
VAR
𝑷 𝝀𝑿 = 𝝀𝑷(𝑿)

Sub - Additivity
𝑷 𝑿 + 𝒀 ≤ 𝑷 𝑿 + 𝑷(𝒀)
ES 𝒛
𝒖−𝒛×𝝈 𝒖 − × 𝒇(𝒁𝒅 )
𝜶
W W W . P E A K S 2 T A I L S .C O M
 CALCULATE RISK MEASURE

Profit and Loss

Backtesting
Attribution Test

1. Types of PnL Actual PnL, Hypo PnL, 1. Normal Hypothesis testing using
RTPL Binomial distribution
2. KS Test 2. Traffic Light approach (Regulated)

PORTFOLIO LEVEL
AGGREGATION

Parametric Methods Historical Simulation Monte Carlo Simulation

W 𝒓 ∆𝒚 𝒍𝒐𝒔𝒔
𝛺 𝑾𝑻
CORRELATION
𝑾𝑺 𝟏 𝑾 𝑺𝟏 𝝈𝟐𝟏 𝝈𝟏𝟐 𝑾𝑺 𝟏 + =
𝝈𝟏𝟐 𝝈𝟐𝟐 𝑾𝑺 𝟐

𝑽𝑨𝑹𝟏 𝑽𝑨𝑹𝟐 𝟏 𝒇𝟏𝟐 𝑽𝑨𝑹𝟏 Improvement –

𝒇𝟏𝟐 𝟏 𝑽𝑨𝑹𝟐 Correlation weighted
Historical Simulation
=
GAUSSIAN COPULA
Updata correlation between
1. Z1 , Z2
set of Historical Obs.
2. Z1 , Z2*
Z2*= 𝒓 × 𝒛𝟏 + (𝟏 − 𝒓𝟐 × 𝒛𝟐
3. Back to Marginals

W W W . P E A K S 2 T A I L S .C O M