Part 2 — Financial Risk Assessment Basics

🎯 Learning Objective

By the end of this section, you will understand portfolio risk fundamentals, Value at Risk (VaR), Expected Shortfall (CVaR), and how Monte Carlo methods are used in financial risk assessment.

2.1 — Portfolio Risk: The Foundation

What is Risk?

In finance, risk is the uncertainty of future returns. A risky investment is one whose outcome is unpredictable. Risk management is the science of quantifying and controlling this uncertainty.

Three Key Concepts

Variance (σ²)

Variance measures how spread out the returns of a single asset are:

$$\sigma^2 = \frac{1}{N-1} \sum_{i=1}^{N} (r_i - \bar{r})^2$$

Where: $r_i$ = individual return, $\bar{r}$ = mean return, $N$ = number of observations.

Interpretation: Higher variance = higher uncertainty = higher risk.

Covariance (σ_ij)

Covariance measures how two assets move together:

$$\sigma_{ij} = \frac{1}{N-1} \sum_{k=1}^{N} (r_{i,k} - \bar{r}_i)(r_{j,k} - \bar{r}_j)$$

Positive covariance: Assets tend to move in the same direction
Negative covariance: Assets tend to move in opposite directions
Zero covariance: No linear relationship

Diversification: The Free Lunch of Finance

💬 Famous Quote

"Diversification is the only free lunch in investing." — Harry Markowitz

The portfolio variance of two assets:

$$\sigma_p^2 = w_1^2 \sigma_1^2 + w_2^2 \sigma_2^2 + 2 w_1 w_2 \sigma_{12}$$

When $\sigma_{12} < 0$ (assets move oppositely), the portfolio risk is less than the sum of individual risks.

Given:

Asset	Weight	Annual Return	Volatility
Stock A (Tech)	60%	15%	25%
Bond B (Govt)	40%	5%	8%
Correlation		-0.3

Question: What is the portfolio volatility?

Solution:

First, compute $\sigma_{12} = \rho \times \sigma_1 \times \sigma_2 = -0.3 \times 0.25 \times 0.08 = -0.006$

Portfolio variance:

$$\sigma_p^2 = (0.6)^2(0.25)^2 + (0.4)^2(0.08)^2 + 2(0.6)(0.4)(-0.006)$$ $$= 0.0225 + 0.001024 - 0.00288 = 0.020644$$

Portfolio Volatility = $\sqrt{0.020644}$ ≈ 14.37%

Stock A alone has 25% volatility, but the portfolio has only 14.37% — the negative correlation with bonds reduced overall risk by nearly half! This is the power of diversification.

2.2 — Value at Risk (VaR): The Industry Standard

Definition

📐 Definition

Value at Risk answers one simple question:
"What is the maximum expected loss over a given time horizon, with a specified confidence level?"

Formal definition:

$$\text{VaR}_\alpha = -\inf\{x : P(L \leq x) \geq \alpha\}$$

In plain language: VaR is a quantile of the loss distribution.

Practical Examples

Confidence Level	Time Horizon	Interpretation
95% VaR	1 day	"We are 95% confident that we will not lose more than ₹X in one day."
99% VaR	10 days	"We are 99% confident that our 10-day loss will not exceed ₹Y."

Visual Intuition

Loss Distribution ████████████████████▓▓▓░░ ←── 95% of outcomes ──→│ │ VaR₉₅ │ 5% of outcomes are worse than this ──→

Three Methods to Compute VaR

Method	Description	Pros	Cons
Historical	Use actual past returns	Simple, no model assumptions	Past ≠ future
Parametric	Assume normal distribution	Fast calculation	Returns aren't truly normal
Monte Carlo	Simulate thousands of scenarios	Handles any distribution, non-linear payoffs	Computationally expensive

⚠️ Key Point for This Workshop

Monte Carlo VaR is the most flexible and powerful method, but also the most computationally demanding. This is exactly where quantum computing can help.

Scenario: Your portfolio is worth ₹10,00,000 (₹10 Lakh). The 1-day 95% VaR is ₹50,000.

Questions:

What does this mean in plain English?
How often might you lose more than ₹50,000?
Does VaR tell you the worst-case loss?

Answers:

On 95 out of 100 trading days, your loss will be less than ₹50,000.
On approximately 1 out of 20 trading days (5%), you could lose ₹50,000 or more. That's roughly 12-13 days per year.
No! VaR only tells you the threshold. On those 5% of bad days, the actual loss could be ₹55,000, ₹1,00,000, or even ₹5,00,000. VaR says nothing about the severity beyond the threshold — this is VaR's biggest weakness, and why CVaR was invented.

2.3 — Expected Shortfall (CVaR): Beyond VaR

The Problem with VaR

VaR tells you the threshold, but nothing about what happens beyond it. If the 99% VaR is ₹10 million, VaR says nothing about whether the actual loss in that 1% worst case might be ₹11 million or ₹100 million.

Definition

📐 Definition

Expected Shortfall (ES) or Conditional Value-at-Risk (CVaR) answers:
"Given that we have exceeded the VaR threshold, what is the expected (average) loss?"

$$\text{CVaR}_\alpha = E[L \mid L > \text{VaR}_\alpha]$$

Why CVaR Matters

Property	VaR	CVaR
Tells you the threshold?	✅	✅
Tells you the tail risk?	❌	✅
Coherent risk measure?	❌	✅
Preferred by regulators (Basel III)?	✅ (still used)	✅ (increasingly preferred)

Visual Comparison

Loss Distribution ████████████████████▓▓▓░░░░ ←── 95% of outcomes ──→│░░│ │ │ VaR CVaR (threshold) (average of the shaded tail)

Scenario: Two portfolios have the same 95% VaR of ₹5,00,000. But their tail risks are very different:

Portfolio	95% VaR	95% CVaR	Worst Observed Loss
Portfolio A (Diversified)	₹5,00,000	₹6,50,000	₹8,00,000
Portfolio B (Concentrated)	₹5,00,000	₹15,00,000	₹40,00,000

Key insight: VaR says both portfolios are equally risky. CVaR reveals that Portfolio B is far more dangerous — when things go wrong, they go very wrong.

This is why Basel III regulations are increasingly moving toward CVaR (Expected Shortfall) as the primary risk measure.

2.4 — Monte Carlo Methods in Finance

The Core Idea

Monte Carlo simulation works by:

Generate random scenarios according to a probabilistic model
Evaluate outcomes under each scenario
Aggregate results to estimate quantities of interest

Applications in Finance

Application	What is Simulated	Output
Stock Path Simulation	Future stock prices under GBM	Projected price trajectories
Derivatives Pricing	Payoff of complex options across paths	Fair value of the derivative
Stress Testing	Extreme market conditions	Loss under worst-case scenarios
Risk Forecasting	Portfolio P&L distribution	VaR, CVaR, risk capital

The Monte Carlo Convergence Problem

The fundamental limitation of classical Monte Carlo:

$$\text{Error} \sim O\left(\frac{1}{\sqrt{N}}\right)$$

Question: Given the $O(1/\sqrt{N})$ convergence, fill in the blanks:

Simulations (N)	Relative Error	Time (if 10,000 sims = 1 sec)
10,000	1%	1 second
1,000,000	?	?
100,000,000	?	?
10,000,000,000	?	?

Solution:

Error scales as $1/\sqrt{N}$. If $N = 10{,}000$ gives 1% error:

Simulations (N)	Relative Error	Time	Explanation
10,000	1%	1 second	Baseline
1,000,000	0.1%	100 seconds	100× more sims → 10× less error
100,000,000	0.01%	~2.8 hours	10,000× more sims → 100× less error
10,000,000,000	0.001%	~11.6 days	1,000,000× more sims → 1,000× less error

⚠️ The Convergence Wall

To get 10× more accuracy, you need 100× more simulations. This is the fundamental bottleneck that quantum computing aims to break — quantum amplitude estimation converges at $O(1/N)$ instead, needing only the square root of the simulations.

2.5 — Setting the Stage

Now that we understand:

✅ How portfolio risk is measured (variance, covariance)
✅ What VaR and CVaR mean and why they matter
✅ Why Monte Carlo simulation is the industry workhorse
✅ Why Monte Carlo has a fundamental convergence limitation

🚀 What's Next

In Part 3, we will build a complete Monte Carlo risk simulator in Python — hands-on, from scratch. Then we will return to that convergence problem and show how quantum computing offers a fundamentally better scaling.

📚 References

Markowitz, H. (1952). Portfolio selection. The Journal of Finance, 7(1), 77–91. doi:10.1111/j.1540-6261.1952.tb01525.x
Jorion, P. (2007). Value at Risk: The New Benchmark for Managing Financial Risk (3rd ed.). McGraw-Hill.
Glasserman, P. (2003). Monte Carlo Methods in Financial Engineering. Springer. doi:10.1007/978-0-387-21617-1
McNeil, A. J., Frey, R., & Embrechts, P. (2015). Quantitative Risk Management: Concepts, Techniques and Tools (Revised ed.). Princeton University Press.
Wilkens, S., & Moorhouse, J. (2023). Quantum computing for financial risk measurement. Quantum Information Processing, 22(1). doi:10.1007/s11128-022-03777-2