S2 – Asset pricing

Asset pricing = Explain/Predict stock returns using systematic risk factors (MKT,size,value).
ALLOW :
Expected return estimation based on historical market data.
Risk exposure comparison across securities.
Support valuation, PTF optimization, Risk management.

Models assess how ≠ factors influences excess returns over rf.

3 CORE Asset pricing models

 Market Model = Basic model linking stock return to market return.
Estimate how much stock moves with mkt (mkt sensitivity)
Ri =  + Rm
Ri → Return of stock i

Rm → Market return (S&P 500)

𝛼 = Abnormal return (stock-specific risk)

β = Market risk

⚠️No error term bc theoretical model.

✅
• Simple model to measure systematic risk
(β & estimating E(r)
If β = 1.2 → stock is 20% more volatile than market.

 CAPM = Estimate stock E(r) based on market sensitivity (BETA) & risk free rate.
R = R +  (R - R )
i f m f

Ri → E(r) of asset i
Rf → Risk free rate
Rm → Market return
β → Sensitivity to market risk

Regression form : Ri - Rf =  +  (Rm - Rf)

→ Predicts return based on market premium.

Assumptions :
Investors = rational & risk averse
No taxes & transaction costs
Homogeneous expectations from all investors
Investors can borrow/lend at rf.

 Fama French 3-factor model = Improve CAPM by adding SIZE & VALUE FACTOR to better explain stock
returns.
Ri - Rf =  + mkt(Rm - Rf) + s•(SMB) + h•(HML)

s⋅SMB → Size effect (small-cap vs. large-cap).

β(Rm−Rf) → Market risk premium.

h⋅HML → Value effect (value vs. growth stocks).

 → Abnormal return not explained by these 3 factors

Assumptions:
Investors = risk-averse & demand compensation for systematic risk.
Firm characteristics (size, book-to-market ratio) influence E(r).
Nb: SMB & HML accounts for factor that CAPM fails to explain

SMB loading (s)→ Stock exposure to size factor.

s > 0  stock behaves like small-cap stocks (small firms)
s < 0  stock behaves like large-cap stocks (large firms)
HML loading (h)→ Stock exposure to value factor
h > 0  stock behaves like value stocks (high book-to-market).
h < 0  stock bejave like growth stocks (low book-to-market).
⚠️Limitations of Fama-French 3-Factor Model
FF3 more precise vs CAPM, but more complex (need more data & harder to apply).
SMB & HML not significant in all mkts/assets. (👍🏻 in 🇺🇸)
Misses important factor like Momentum or Liquidity risk.
Assumes linearity, so don’t capture non linear effects (during crisis)

Estimating Asset pricing models using Linear Regressions

Models estimated using LINEAR REGRESSION.
Types of regressions :
• Time-series regression → Estimates model for single stock or PTF over time.
• Panel data regression → Estimates model using multiple stocks at same time.
HOW  Method = Ordinary Least Squares (OLS) = econometric technique used to mini errors in model
predict°s.
🎯 Determines coef (α, β, s, h) to explain stock returns relative to risk factors.

Simple Linear Regression Model  𝑌𝑖=𝛼+𝛽𝑋𝑖+𝜖𝑖

Linear regressions models rltship btw dependent var (Y) & inde variables (X)

Yi= Dependent variable (e.g., stock return).

Xi = Independent variable (e.g., market return).
α = Intercept (abnormal return).
β = Impact of Xi on Yi.
ϵi = Error term (deviation from the model).
& SO  OLS mini sum of squared errors (SSE) min∑(Yi− α −βXi)2

 Estimating Market Model using regression

Market model equation  Ri =  + Rm+𝜖𝑖
We estimate α & β using OLS regression
Ri = Return of stock i.
Rm = Market return (e.g., S&P 500 index return).
α = Abnormal return not explained by market movements.
β = Sensitivity to market return (systematic risk).
ϵi = Error term (idiosyncratic risk) (factor affecting Ri not explained by Rm)
⚠️Expected value of error term = 0.

 Estimating CAPM using regression

CAPM equation  Ri – Rf =  + (Rm – Rf) +𝜖𝑖
We estimate α & β using OLS regression where
Ri = Return of stock i.
Rf = risk-free rate (Treasry bond yield)
Rm- Rf = Market risk prenium
α = Abnormal return not explained by CAPM.
β = Sensitivity to market excess return (systematic risk).
ϵi = Error term capturing firm-specific influences
⚠️Expected value of error term = 0.
INTERPRETATION OF CAPM
Alpha (α)  Indicates if stock is underpriced or overpriced.
Underpriced Stocks (α > 0) → stock generate excess returns beyond what CAPM predicts.
Overpriced Stocks (α < 0) → stock generate less excess returns than what CAPM predicts.

Beta (β) 

 Ri - Rf =  + (Rm – Rf) + s•SMB + h•HML + 𝜖𝑖

 Estimating Fama-French Model using regression

We estimate α,β,s,h using OLS regression where

SMB = Return of small-cap stocks minus large-cap stocks.
HML = Return of high book-to-market stocks minus low book-to-market stocks.
s = Exposure to size factor (SMB).
h = Exposure to value factor (HML).
ϵi = Error term capturing firm-specific influences
⚠️Expected value of error term = 0.
INTERPRETATION OF FAMA-FRENCH
s > 0 → stock behaves like a small-cap stock
h > 0 → stock behaves like a value stock
(β) → Captures systematic market risk

SESSION 3 – Relative Valuation Using Regressions

RELATIVE VALUATION
= Price of asset is compared to prices assessed by market for similar assets.
= Values a Cy (Price of a CY) by comparing it to similar Cies, using valuation multiples.
Why RV ?
Easier & faster than DCF, reflects better the mkt mood.
VALUATION MULTIPLES
= financial ratios to estimate CY value by comparing them to similar CIES.
 Express value as function of financial metric.

Market Valuemetric
FORM  A Multiple =
Financial Metric
Express value as function of financial metric

⚠️multiples differ by industry, so compare within same industry.

Situation Action
Actual multiple < Predicted BUY
multiple (undervalued)
Actual multiple > Predicted SELL
multiple (overvalued)

Fundamentals of Multiples (Analyzing Multiple)

Fundamentals that determine & drive multiples
Stock Price D1 E 1∗Payout P Payout
🔹 P/E – Price to Earnings P/E = P0= = donne =
EPS r−g r −g E r −g
 ↑ Growth (gEPS) ↑ PE
 ↑ Risk (Beta + cost of equity  r)  ↓ PE
Dividends
 ↑ Payout ( )  ↑ PE (unless reduces growth)
Earnings
EV
🔹 EV/EBITDA EV/EBITDA=
EBITDA
 ↑ Growth rate  ↑ multiple
Total Debt
 ↑ Debt (Debt to Firm = )  ↓ multiple
Total Debt + MV of Equity
 ↑ Taxe rate  ↓ multiple

EV  PV of FCFF (FCF to firm)

∞
FCFF
EV = ∑ (1+r )tt So
t =1

AT steady state  simplifies to EV =

FCFF
WACC−g
FCFF = EBITDA – Taxes – Reinvestment
FCFF = EBITDA * ( 1 – Tax Rate) – Reinvestment

Market Cap ROE∗BV ROE

🔹 P/BV – Price to Book P/B = P0 =  P/BV =
Book value of equity r−gEPS r−gEPS
 ↑ ROE  ↑ P/B
 ↑ Growth (gEPS)  ↑ P/B
 ↑ Systematic Risk (Beta)  ↓ P/B
 Higher P/B = MKT expects strong future perfo.
 Lower P/B = may be weak profitability or undervalution
❗️Useful for comparing CIES in same industry.

Applying Multiple
2 ways to use valuation multiples
Direct Comparison (Comparing CIES Multiple) Non-Direct Comparison (
Compare CY multiple to peers. Estimate CY multiple based on fundamentels using regression.
Ex : Comparing CY P/E Ratio to industry avg. Ex: Use regression model to estimate appropriate P/B multiple
based on fundamentals.
When using comparable CIES When using regressions to predict multiples.
 Pick CIES in same industry w/ similar risk & growth.  Estimated a CY Predicted multiple using key fundamentals.
 Compute relevant valuation multiple (P/E, EV/EBITDA) That drive it.
 Compare CY mulitple To: Peer group median/average or a
specific firm’s multiple.  Higher actual = possible overvaluation.
 Lower actual = possible undervaluation.

• CY Multiple much higher = possible overvaluation;

• Much lower = possible undervaluation.
Example : Using P/E for Valuation Example : Predicting P/E Ratio w/ Regression.
We estimate stock price using P/E ratio of comparable CY.
Regression Estimated Model :
Formula : P0 = P/Ecomp * EPSfirm PE=α+b1⋅Beta+b2⋅gEPS+b3⋅Payout
Data : Similar CY has P/E = 18 P/E = 11.89 + 1.47 + 32.44gEPS + 13.18Payout Ratio
Our CY EPS = 4.20
Applying Regression Model
SO estimated price : P0 = 18 * 4.20 = 75.60 Given CY fundamentals
The compare to actual stock price ($82.00) - Beta = 1.2
Since $82 > $75.60  CY appears overvalued - gEPS = 4 ;0
- Payout Ratio = 0.5
To better assess valuation  Compare w/ industru avg.
Predicted P/E Calculation :
P/E = 11.89 + (1.47 * 1.2) + (32.44 * 4.0) + (13.18 * 0.5)
P/E = 11.89 + 1 ;76 + 129.76 + 6.59
P/E = 149.99

 Our CY has higher P/E ratio than its peers. Comparison w/ Actual Multiple
 Estimated price using comparables suggesrs overvaluation. - If CY actual P/E = 140  Undervalued
- If CY actual P/E = 160  Overvalued
 S4 - Mutual Funds & ETFS: Ratings, ESG
Scores, Controversy and Industry Involvement

• Investment fund = Pool money from multiple investors to invest in diversified PTF of assets.
Pro Management + Diversification + Access to MKT hard to individually + Risk sharing
 Investors receive returns based on fund perfo & propor to their invest.

• 2 way to manage Invest Funds :

Active management = Managers try beat market by picking adequate stocks. Higher fee
Passive management = Just replicate perfo of benchmark index. Lower fee.
How choose?  Based on risk tolerance, cost preferences, return expectations.

• Types of Invest Funds

 ≠ structures, ≠ investors needs , ≠ risk/liquidity/return characteristics
MF Actively/Passively managed, priced once day at NAV
ETF Often passivively managed, replicate an index and traded
like stocks on stock exchange.
HF Actively managed, use complex strat, offer high return
but high risk, demand large mini requirment & exclusive
to accredited investors.
PE Invest in pv CIES, focus on LT growth
Real Estate Investment Trusts Invest in income-generating real estate properties.
(REITs)

🔎MF
= Actively/Passively managed, priced once day at NAV
HOW IT WORKS ? Investors buy share of MF & Net Asset Value (NAV) determines share price.
❗️Value of MF measured by it NAV
Total assets−Total liabilities
NAV =
Total outstanding shares
Total assets = Stocks, Bonds, Cash & Other secu in fund.
Total liabilities = Management fees, Operational costs, Other exp.
 Updated at end of each trading day (after MKT closing)

🔎ETFs
= Often passivively managed, replicate an index and traded like stocks on stock exchange.
 Diversification at ↘️costs than MF.
⚠️UNLIKE MF  Can be bought or sold throughout day at MKT prices

Advantages ✅ :
- Liquidity – Trade like stocks, so allow real-time price execution
- Cost Efficiency, - Lower expenses ratios than actively managed MF
- Transparency – Holdings disclosed daily (unlike MF = monthly or quarterly)

ESG & Investment Funds

• Why does ESG Matter ?

For Investors
- Help spot LT risks linked to ESG issues
- Aligns invest w/ ethical + sustainability values.
- Lead to better LT risk-adjusted return.
For Investment Bankers
- ESG now part of valu models
- ESG funds & products are booming.
- Clients demand ESG invest options.
For Society
- Encourages responsible Corpo behavior
- Supports transition to greener eco.
- Decrease global envi & social risk.

• Quanti measure of ESG

Via ESG scores based on 3 factors :
- Environmental (E) – CO2, Energy efficiency, Resource use, Waste
- Social (S) – Labor practices, Diversity, Human rights policies
 - Governance (G) – Board composition, Transparency, Shareholder rights
 Overall ESG score = Aggregate scores across 3 cat to provide single perfo measures.
⚠️ESG ratings can vary across providers due to ≠ methodo & weightings. (NOT STANDARDIZED), so
carreful when interpretating.

• ESG investing & Funds

 ESG investing integrates E,S,G criteria into decisions.
ESG Investment Strategies :
- Negative screening  Exclusing bad industries (tobacco, weapons, fossils fuels).
- Positive screening  Select CIES w/ strong ESG perfo.
- Best-in-class  Pick best ESG CY in each industry.
- Thematic investing  Focus on ESG sectors (Renewable Energy).

• How ESG Funds & ETFs are constructed ?

 use various stat :
- Passives fund track ESG indices (MSCI ESG leaders Index, S&P500 ESG Index)
- Active funds select individual stocks based on ante ESG research & scoring models
PTF construction may involve :
- ESG tilts  Give more weight for CIES w/ high ESG scores.
- ESG factor integration  Adjust weights based on specific ESG risks.
- Engagement strategy  Push CIES to improve ESG via active ownership.
💡ESG ETFs  offer transparency, liquidity, ↘️fees compared to actively managed ESG funds.
 S5 – Funds Performance

• MFs & ETFS

🔑≠:
- FEES – MFs fees > ETFs fees bc active management & higher distribution costs.
- TRADING – ETFs traded like individual stocks on stock exchanges & MFs trade only at NAV once a day.
- TRANSPARENCY- ETFs disclose holdings daily & MFs quarterly/monthly
- TAX EFFICIENCY – ETFs more tax-efficient.

Why more MFs than ETFs ?

MF more numerous bc earlier origin (MFS  20s & ETFs  90s) & widely used by retail investors, offer more
active strategies (diversification), easier to set up.
BUT, ETFs growing faster today (cheaper & more flexible).

• Compare funds by (GENERAL COMPARISON)

1) Inception YEAR
= Date fund officially launched & available to investors.
2) Management FEES
= Annual costs charged by fund manager to operate & administer an invest funds
Why fee comparison = important ?
 Even small ≠ in fees can significantly affect LT invest returns
3) Sustainability Scores
= Composite indicator reflecting funds exposure to ESG risks.
Higher ESG score = Lower ESG risks.
WHy ESG scores comparison = important ?
 ETFs often track ESG indices
 MFs may have more flexibility in ESG screening or active engagement
 Help eval LT responsbility & risk.

• Compare fund with MKT PERFO

1) Returns
High-frequency returns (daily, monthly, quarterly)
• Show ST vol, shocks, mkt reactions
• Reveal patterns (seasonality, crisis periods)
• Ideal  Risk analysis (STDdev, Sharpe ratio)

Low-frequency returns (yearly, multi-year trailing)

• Show better LT strat & perfo consistency
• Reduce temporary noises & mkt fluctu
• Ideal  Benchmarking & PTF eval.
🔸 ETFs → High frequency return more relevant (traded daily)
🔸 MFs → Low frequency return more relevant (less traded)
Quarterly Returns Year-to-date Return (YTD) X-Year trailing Returns
= Fund perfo over individual 3 = Fund perfo from start of current = Fund perfo over fixed past
month periods year to latest available date periods(3yo, 5yo, 10yo)

2) Sector compo
 Shows fund focus/ theme (techno, energy, healthcare)
 Help explain fund risk & return sources.
 Help w/ diversification analysis.

Types of funds & sector focus :

ETFs  Track sector indices (S&P500 tech) or Broad indices w/ fixed sector weights.
MFs  Often active managed, w/ adjust of sector exposure over time.
What we expect to see
- Thematic funds  Sector concentration (all tech)
- Diversified ETFs  Balanced sector spread.
- Active MFs  Sector mix evolves tactically

How Sector Compo reflects risk ?

Sector weight  Exposure to eco & market risks
Sector shocks (tech crash)  strong impact if fund is concentrated
More diversified sector allocation = lower portfolio sensitivity

🔎FOCUS SECTORS
Financial Services Healthcare Technology
 RPZ core of Kmkts &  Defensive sector (less  High growth
banking systems. cyclical).  Highly volatile
 Highly sensi to MP, IR, Regu Innovation driven  React to rapid mkt shifts,
changes.  Stable demand geopo tensions, regu.
 Leading indicator of
macroeco shifts.
 HOW TO ANALYZE SECTOR COMPO IMPACT ON FUND PERFO  via simple linear
REGRESSIONS
Use regression to see which sectors drive returns:
Fund trailing return ytd = β0 + β1·Financial + β2·Healthcare + β3·Technology + ε
Purpose ?
 Qtify impact of each sector on return.
 Test if funds w/ higher expo to specific sectors perform ≠.

3) RATINGS
Funds ratings = Inde scores provided by 3rd-party agencies based on risk-adjusted perfo vs similar funds.
Why matters ?
 Quick benchmarck for fund quality
 Help make informed decisions

Morningstar Ratings
 Assigns funds rating from 1 to 5 .
Based on risk-adjusted return relative to peers in same category.

5 Top 10% of funds in the cat.

4 Next 22,5%
3 Middle 35%
2 Next 22,5%
1 Bottom 10%

Ratings mean for each fund type :

 ETFs – Matter less if ETF simply tracks index
 MFs – Highly relevant

⚠️LIMITATIONS OF RATINGS
• Based on past perfo  No future result guarantee.
• Ignore fund strategy, liquidity, cost structure.
 COMPLEMENT but don’t replace deep financial analysis & perfo eval.

⚠️RETURN  Fundamental perfo measure BUT not sufficient for comprehensive eval of MFs or ETFs.
(even w/ ratings)

SO IMPORTANCE OF Relative performance

Absolute return = not enough as we care about how fund perfom (return) relative to risk taken.

 Modern PTF Theory (MPT) emphasizes fundamental trade-off btw RISK & RETURN
• ↗️E(r) come w/ ↗️risk
• Efficient PTFs = max return for given level of risk or mini risk for given leven of return.

Why relative perfo matters ?

2 funds can have same return but very ≠ levels of volatility. → Prefer one with better risk-adjusted
performance.

 SO Use of risk-adjustd metrics

SHARPE RATIO
= Measures return earned per unit of risk taken.
R p−R f
SR = Higher SR  More return per unit of risk  more efficient
σp
Rp = PTF or Fund return
Rf = Risk free rate of return
p = STDDev of PTF returns (risk)
 S6 – Prices Returns & VAR

1) Working w/ Price Data

Price type (OHLC format):
- Open --> Price at which tarding begins
- High --> Highest transa price during trading period
- Low --> Lowest transa price
- Close or Adjusted Close --> Final price at end trading period.
Nb : Close price = most stable & widely used ?

• Trade Volume & Dollar Volume

 Volume = # of shares traded during given time period.
⚠️Measured as # shares, not in $.
 Dollar volume = Volume * Close price
Volume∗Close
❗️For visu or comparison, scaled in BILLIONS : Dollar Volume (bn) = 9
10
• Returns Types
Returns = % change in asset price over time interval.
- Daily return – Change in price from 1 trading day to next
- Monthly return – Change from end of 1 month to next
- Quarterly return – Computed over 3month period
- Annual return – Year-over-year price change.

P t−P t−1 Pt
rt = or r t =ln ⁡( )
P t−1 P t−1
 Compounded Returns
To compute total returns over multiple periods
R = (1+ r1)(1+r2)…(1 +rn) -1

 Visualising Returns
Common visual tools :
 Lines plots – Trace return path over time
 Histograms – Study return distrib
 Faceted plots – For daily VS monthly return comparisons.

2) Basic features of assets

 Expected (mean) return
= Avg perfo return of asset over time/

Where ri,t – return of asset i at time t (ex : Daily return at 01/04/25)

T - # of time periods (ex : how many days we have in sample)
. NB : High frequency returns tend to be centered around 0.

 Variance & Volatility

STDDev of asset = Measure volatility of asset

(Variance = STDdev au Carré)

 Measure of risk – Higher values =More uncertainty
Volatility = economic interpretation of standard deviation.

 Covariance & Correlation (for more than 1 assets)

Assets don’t exist in ISOLATION  We care about how they move TOGETHER.
• Covariance
 • Correlation
NB : Low or Negative correlation btw assets enhance diversification benefits.

Variance-Covariance matrix summarize asset interactions

variances (e.g., σ²₁) on the diagonal

covariances (e.g., Cov(r1,r2)) off-diagonal.
 Key Feature: Symmetric (Cov(r1,r2) = Cov(r2,r1)).

Sharpe ratio for Asset i

= Measures excess returns of asset per unit of risk (volatility).
 perfo of asset i compared to risk-free invest.

Where i – Mean return of asset i.

Rf – Risk-free rate (Return of T-Bill)
 i - STDdev (volatility) of asset i returns.
 Tell how much excess return (above rf rate), we receive per unit of volatility
Higher = Better risk-adjusted perfo.

WHY INVESTORS CARE ABOUT RISK ?

Bc care about risk of losing money.
 High-return assets often have high loss risk.
Investors fear losses more than they value gains (loss aversion).

& STDDev  NOT ENOUGHT TO VALUE RISK

- Flaw 1: Treats upside & downside volatility equally - we only fear downside (losses).
- Flaw 2: Assumes returns are symmetric & normal - not always true.
- Flaw 3: Can’t answer key questions like “What’s my worst case-loss at 5% proba?”
💡Solution VaR (Value at Risk) = method to directly estimate potential losses under specific
scenarios.

3) Value at Risk (VaR)

VaR = Estimate max potential loss in value of asset/PTF/firm, over specific horizon , at given
confidence level.
“How much could I lose with X% confidence over Y days?” (“95% chance of losing max
1million in 1 day”).

BUT ⚠️--> Limits

VAR quantifies : Magnitude of potential loss, Probability loss will occur, Time frame over
wich loss may happen.
❌LIMITS :
- Not a guarantee, just an estimate.
- Don’t show losses beyond threshold.
- Assumes normal conditions—extreme events can cause bigger losses.
 SO, useful tool, but must be interpreted w/ care & other risk metrics

VaR MODELS
I. Historical Simulation
Estimates VaR using historical data returns – without making assumptions about underlying
distributions of returns.
STEP 1 --> Collect historical returns
STEP 2 --> Sort returns from worst loss to greater gain
 STEP 3 --> Select return corresponding to desired confidence level.
⚠️Assume past behavior retuns = good indicator of future risk
Assumptions :
 Historical return data reflects range of future outcomes
 Mkt conditions in past are rpz of current/future risk
 No specific distributions (normality) is assumed.

r – historical returns (sorted)

α: Significance level (5% for 95% confidence)
 VaR reported as positive number (the magnitude of the loss)
✅ ❌
• No distributional assumptions • Sensitive to historical window (Choice
• Easy to implement of time period can over/under estimate
• Data-driven risk)
• Outliers dominate (Extreme past return
can hevaily influence estimate)
• Assume past = predective.

II. Parametric (Variance-Covariance) VaR

Estimates VaR by assuming returns follow normal distribution. Use mean & STDdev.
STEP 1 --> Use asset/PTF mean return + STDDeV
STEP 2 --> Quantifies potential losses in term of STDdev events from mean.
Assumptions :
 Returns follow normal distribution.
 Correlations btw assets remain constant over time.
 Assets independently & identically distributed (i.i.d.) across time.
 Historical data used only to estimate return distribution parameters.

µ: Expected return
σ: STDdev of returns (volatility)
zα: z-score corresponding to desired confidence level (1.645 for 95%)
α: Significance level (5% for 95% confidence)
NB : For ST horizon (1 DAY), µ≈0 so
❌
• Underestimates extreme losses.
• Assumes stability in correlations – not
true during crisis.
• Poor perfo w/ non-linear instru.

III. Monte Carlo VaR

Estimates VaR by simulating thousand of random possible return paths.
STEP 1 --> Define return distribution (N-D or T-student)
STEP 2 --> Generate large nbr (10000) of simulated return scenarios
STEP 3 --> Compute PTF value or loss for each simulated return
STEP 4 --> Sort losses & extract α-quantile to obtain VaR.

(conceptual formula)
NB : Accuracy of MC VaR depends on # of simulation & qlty of distributional assumptions.

BONUS  Expected Shortfall (Conditional VaR)

Complement Risk measure  Avg loss beyond VaR treshold (worst case scenario)
 « If losses exceed VaR, how bad could they be on avg ? »

VaRα: Value at Risk at confidence level α

L: Loss
E[·]: Expected value operator
Interpretation: ES is average loss in the worst (1−α)% of cases.
Example: If VaR at 95% is 1M, ES tells us what avg loss is among the 5% worst outcomes.

4) Relaxing Normality Assumption

 Most VaR models (parametric, Monte Carlo) assume returns = normally distributed.
BUT in reality  Returns show fat tails, asymmetry, and joint crashes – leading to
underestime risk during market stress.
Solutions ? 💡COPULAS
 Tools to model how assets depend on each other, separate from their individual
distributions.
Used to generate realistic scenarios in Para & Monte Carlo VaR .
Copula-Based Simulation = Function linking individual asset distribution (univariate/marginal)
into joint (multivariate) distribution.
→ Models how variables move together, even in extreme cases.
 S7 – PTF & VaR for Multiasset Instruments

I. From assets to PTFs

PTF = Collection of financial assets (stocks, bond, ETFs) that an investor holds.
🎯 Combine them to achieve desired balance btw return & risk.
💼 Each asset gets a weight (% of K invested).
🎯 These weights guide risk & return.
📊 Judge whole portfolio — not just each piece!
Mathematical rpz of PTF
 Using Weight vector

ωi = proportion of K invested in asset i

n = # of assets
⚠️sum of weights =1 (100%) for fully invested PTF

• PTF Expected Return

= Weighted avg of individual expected returns.

au CARRE les petits trucs

ωi : weight of asset i
µi : expected return of asset i

• PTF Risk (= STDDev)

PTF Stddev = Depends on each asset volatiliy & correlation btw assets.

Even if assets are individually risky  combining them can ↘️total risk.

Sharpe Ratio = Measures risk-adjustd perfo.

µp : expected return of PTF

rf : risk-free rate
σp : standard deviation of PTF
Example : 3-Asset PTF

II. MODERN PTF THEORY

MOTIVATION FOR PTF THEORY

 Investors diversify by spreading K across multiple assets !
BUT, equally weighted PTFs don’t consider :
1) ≠ in E(r)
2) ≠ in Risk (STDDev)
3) Interaction btw assets (Correlation)
SO, MODERN PTF THEORY --> Identify most efficient allocation of K for given objective
= Risk mini or Return max.
💡MPT tourne autour du Mean-variance framework
 🎯 Helps construct PTFs balacing E(r) & Risk.
Assumes that :
- Investors = Risk-Averse (prefer higher E(r) for lower risk)
- PTF choice based on 2 parameters only  Expected return of PTF (µp) +
Risk (Variance or STDdev - (σp))
CORE IDEA  Among all feasible PTF, choose 1 that :
 Mini var for given E(r)
Or
 Max E(r) for given var
🔑 Concepts :
- Mini-variance PTF (MVP) = PTF that deliver lowest possible variance as ref
point.
- Efficient PTF = Any PTF that lies on efficient Frontier
- Efficient Frontier = Set of PTFs that are M-V efficient.

 MVP
= PTF that has lowest possible risk (variance) among all feasible PTFs.
Why Important ?
Lowest risk achievable via diversification
😌 Optimal PTF for risk-averse investors
️Foundation for building other efficient portfolios (

📊 Graph Insight:
📍 MVP at bottom-left point on Efficient Frontier
🧩 All other efficient portfolios = MVP + higher return assets

 Efficient PTFs
• Eco Motivation Behind  Not ALL PTFS = Desirable (some clearly inferior)
Fundamental trade-off  E(r) (reward) VS Volatility (uncertainty)
 Efficient PTF rpz rational choice for risk-averse investor (BALANCE BOTH) :
For given level of risk, provide max possible E(r)
For given level return, provide lowest possible Risk.

 Efficient PTF from UTILITY MAXIMIZATION (one that max

satisfaction (utility) & best balance e® & risk).
 Efficient PTF from Sharpe Ratio (one that maximize SR)

 Efficient Frontier
= Rpz sets of PTFs that are M-V efficient
(Highest E(r) for given level risk OR Lowest Risk for given level E(r))
🩷PTFs on frontier (top-edged only – au dessus du MVP sur line) = Efficient
PTFs
❌PTFS above frontier = Impossible
💩PTFS below frontier = Suboptimal

 Efficient frontier is a curve (parabola in risk-return space)

 Shows minimum risk for any expected return.

• More return = more risk

• 📉 Left-most point = MVP (lowest possible risk)
• 📈 PTFs combining Rf asset + Risky Asset  Capital Market Line (CML) =
Straight line from Rf asset touching frontier
💼 Market Portfolio = Where the CML kisses curve 💋

III. VaR For PTFs

 Useful measure for PTF perfo.
 For Historical & Parametric VaR
 Calculated same way as indivudial stocks BUT, employ PTFs returns for
calculations !

 For Copula Based VaR

Why use copula in VaR estimation ?
📉 Historical VaR → Only sees past return combos
🧮 Parametric VaR → Assumes normal distribution + linear correlation
❌ Both miss non-linear, extreme (tail) risks
SCopulas, allows :
- Separate marginal behavior (individual asset behavior) from dependence (how
asset move together)
- Captures:
 💥 Tail dependence (extreme events happening together)
 🔀 Non-linear relationships (beyond simple correlation)
- Allow to Simulate realistic & new return combinations even if not observed
historically.
 Copulas = smarter, more flexible risk tools for modeling how bad
things can go wrong — together 🔥📉