Understand correlation risk across positions — Stock-Market notes

WHY does correlation risk exist?

WHAT we want: to spread money so that when one thing goes down, another cushions it.

WHY it fails: stocks are driven by common factors — interest rates, oil prices, the overall market ("beta"), sector news. If two positions share the same driver, they crash at the same time. Diversification only reduces risk to the extent that positions are not perfectly correlated.

HOW we measure it: with the correlation coefficient $\rho$ , which ranges from $-1$ to $+1$ .

Deriving portfolio risk from scratch

We never dump the formula — we build it.

Step 1 — Portfolio return. Two positions with weights $w_A, w_B$ ( $w_A+w_B=1$ ): $R_p = w_A R_A + w_B R_B$ Why this step? Portfolio return is just the weighted average of the pieces — money-weighted.

Step 2 — Variance is the risk measure. Risk = variance of $R_p$ : $\sigma_p^2 = \operatorname{Var}(w_A R_A + w_B R_B)$

Step 3 — Expand using the algebra of variance. For any $X,Y$ : $\operatorname{Var}(aX+bY)=a^2\operatorname{Var}(X)+b^2\operatorname{Var}(Y)+2ab\,\operatorname{Cov}(X,Y)$ Why? Square the deviation $(a(X-\mu_X)+b(Y-\mu_Y))$ and take expectation; the cross term survives.

So: $\sigma_p^2 = w_A^2\sigma_A^2 + w_B^2\sigma_B^2 + 2w_Aw_B\operatorname{Cov}(R_A,R_B)$

Step 4 — Replace covariance with correlation. Since $\operatorname{Cov}=\rho\,\sigma_A\sigma_B$ :

What the $\rho$ term does (the punchline)

Take equal weights $w_A=w_B=\tfrac12$ and equal risk $\sigma_A=\sigma_B=\sigma$ : $\sigma_p^2 = \tfrac14\sigma^2+\tfrac14\sigma^2+2\cdot\tfrac14\rho\sigma^2 = \tfrac{\sigma^2}{2}(1+\rho)$ $\sigma_p=\sigma\sqrt{\tfrac{1+\rho}{2}}$

$\rho$	$\sigma_p$	Meaning
$+1$	$\sigma$	No diversification — same as one stock
$0$	$\sigma/\sqrt2\approx0.71\sigma$	Real risk reduction
$-1$	$0$	Risk fully cancels (perfect hedge)

Figure — Understand correlation risk across positions

Worked examples

Common mistakes (steel-manned)

The 80/20 takeaway

Recall Feynman: explain to a 12-year-old

Imagine you bet on a soccer game. To be safe you also bet on... the same team in the same game. That's not two bets — it's one big bet! Real safety is betting on different games that don't affect each other. In the stock market, stocks that go up and down together are like the same team. Owning ten of them is still one big bet. To be truly safe, own things that move for different reasons — then when one has a bad day, another might have a good one.

Flashcards

What does correlation risk mean?

The risk that positions you think are diversified actually move together, so combined losses are far larger than expected.

Formula for two-asset portfolio variance?

\sigma_p^2 = w_A^2\sigma_A^2 + w_B^2\sigma_B^2 + 2w_Aw_B\rho\sigma_A\sigma_B

Where does ALL the diversification benefit live in that formula?

In the correlation cross term

2w_Aw_B\rho\sigma_A\sigma_B

; low/negative ρ shrinks portfolio risk.

For equal weight & equal σ, what is σ_p?

\sigma_p=\sigma\sqrt{(1+\rho)/2}

At ρ = +1, what is portfolio risk vs a single stock?

Identical — no diversification benefit at all.

At ρ = −1 with equal weights and σ, what is σ_p?

Zero — the positions perfectly hedge.

Diversification floor for N equal-ρ positions as N→∞?

\sigma_p\to\sigma\sqrt{\rho}

; correlation caps how much risk you can remove.

Why is "I own 10 stocks so I'm safe" wrong?

Diversification depends on correlation, not the number of tickers; 10 highly correlated names ≈ one bet.

What happens to correlations during market crashes?

They spike toward +1, so diversification weakens exactly when you need it most.

Does negative correlation give free profit?

No — it reduces risk but also cancels return; it's a hedge, not an edge.

Connections

Diversification — the practical application of low correlation.
Position Sizing — size the combined correlated block, not each name.
Beta and Market Risk — common factor driving correlation.
Value at Risk (VaR) — uses the covariance matrix directly.
Hedging Strategies — deliberately using negative ρ.
Standard Deviation and Variance — the building blocks of these formulas.
Portfolio Theory (Markowitz) — the efficient frontier is drawn from exactly this math.

Concept Map

Hinglish (regional understanding)

Intuition Hinglish mein samjho

Dekho, correlation risk ka matlab simple hai: agar tumne 5 alag-alag stocks kharide hain lekin woh saare ek saath upar-neeche jaate hain, to tumne 5 bets nahi lagayi — tumne ek hi badi bet lagayi hai jiska size 5 guna hai. Log sochte hain "maine 10 stocks le liye to main safe hoon", par asli safety tickers ki ginti se nahi, unke correlation se aati hai. Do bank stocks ek crisis mein saath mein girenge — woh ek hi bet ke barabar hai.

Maths bilkul seedhi hai. Portfolio ka risk (variance) hota hai: $\sigma_p^2 = w_A^2\sigma_A^2 + w_B^2\sigma_B^2 + 2w_Aw_B\rho\sigma_A\sigma_B$ . Poora diversification ka fayda us aakhri term mein chhupa hai — us $\rho$ (correlation) waale term mein. Agar $\rho$ high (jaise 0.9) hai to risk mushkil se kam hota hai; agar $\rho$ kam ya negative hai to risk kaafi ghat jaata hai. Equal weight ke liye formula banta hai $\sigma_p=\sigma\sqrt{(1+\rho)/2}$ .

Ek important baat yaad rakho: market crash ke time correlations $+1$ ki taraf bhaagte hain — matlab sab kuch ek saath girta hai kyunki sab log cash ke liye bechte hain. To jab tumhein diversification ki sabse zyada zaroorat hoti hai, tab woh partly gayab ho jaati hai. Isliye position size hamesha stressed (zyada) correlation maankar rakho, calm-time correlation par bharosa mat karo. Aur negative correlation ko "free paisa" mat samjho — woh risk kam karta hai par return bhi cancel kar deta hai; woh hedge hai, edge nahi.

Test yourself — Risk & Money Management

Connections