Home / Portfolio Risk / Correlation Risk
Portfolio riskIntermediate

Correlation Risk

Correlation risk is the danger that positions assumed to be independent actually move together, so their combined loss is far larger than expected, a danger that intensifies because correlations converge toward one in a crisis.

Quick answer: Correlation risk is the danger that positions assumed to be independent actually move together, so their combined loss is far larger than expected, a danger that intensifies because correlations converge toward one in a crisis.

In simple words

You may think you hold several different trades, but if they secretly rise and fall together, you really hold one big trade in several costumes. Correlation risk is the chance that this hidden togetherness shows up at the worst time, turning what looked like spread-out risk into a single concentrated blow. It is dangerous because correlation is not fixed: in a crash, positions that normally drift apart snap together and fall as one. Managing it means measuring how your positions move together, not just how large each one is.

Purpose

This page defines correlation risk, shows how correlation drives the true risk of a multi-position book, and explains why crisis convergence makes historical correlations unreliable exactly when they matter.

Visual explanation

Correlation Risk

A grid of pairwise correlations across positions; clusters of high values reveal hidden concentration the position count hides.

Correlation MatrixNiftyNiftyBankNiftyBankNiftyITITGoldGold1.000.820.55-0.200.821.000.48-0.150.550.481.00-0.05-0.20-0.15-0.051.00Low or negative correlation is what diversification needs

Professional explanation

Correlation, not count, drives portfolio risk

The risk of a multi-position book is governed by the correlations between its positions, not by how many there are. Two positions with a correlation of 0.9 barely diversify each other and behave almost like a double-sized single position, while two with a correlation of −0.3 substantially offset. This means a book can look diversified by name count yet carry the concentrated risk of a single bet if its positions are highly correlated. Correlation risk is the gap between the diversification a trader believes they have and the far smaller diversification the correlation structure actually delivers.

The maths: how correlation moves portfolio volatility

In the two-position volatility formula, the cross term 2 w1 w2 ρ σ1 σ2 carries the entire influence of correlation. When ρ is high and positive, this term is large and portfolio volatility approaches the weighted average of the parts, meaning almost no diversification. As ρ falls toward zero the cross term shrinks and combined volatility drops, and at negative ρ the positions actively hedge each other. A trader who plugs in a comfortable historical ρ but faces a stressed ρ near one has understated their true portfolio volatility, sometimes dramatically, because the cross term they assumed small has quietly become large.

Crisis convergence: the correlation you measured is not the correlation you get

The defining feature of correlation risk is that correlations are unstable and rise sharply in stress. Assets that appear independent in calm periods, different sectors, stocks and options positions, tend to fall together when liquidity dries up and leverage is unwound, so realised crisis correlations converge toward one. A portfolio sized on peacetime correlations can therefore suffer a loss far larger than its risk model predicted, because the diversification it counted on vanishes exactly during the drawdown. This convergence is why correlation risk is not a minor modelling nuance but the mechanism through which apparently prudent portfolios blow up together.

Hidden and indirect correlation

Correlation risk is often invisible because it is indirect. Several positions may share a common driver, interest rates, the US dollar, crude oil, or broad risk sentiment, without any obvious pairwise link, so they move together through a factor the trader never charted. Long positions in banking, real estate and autos, for instance, are all sensitive to interest rates and can slump together on a rate shock despite being different sectors. Detecting correlation risk therefore means thinking in common factors and scenarios, not only in pairwise history, because the most damaging correlations are the ones a naive correlation matrix does not obviously show until the driver moves.

Correlation is not causation and is regime-dependent

A measured correlation is a backward-looking statistic over a chosen window; change the window or the regime and the number changes. Two assets can show a strong historical correlation that reflects a shared regime rather than a stable structural link, and that relationship can break or reverse without warning. This has two consequences: relying on a single correlation estimate to justify a hedge or a diversification claim is fragile, and correlations near zero over one period offer no promise of independence in the next. Treat every correlation as an uncertain, regime-dependent estimate, and size for the possibility that it moves against you.

Managing correlation risk

Because the dangerous scenario is convergence, correlation risk is managed by limiting aggregate exposure to any common factor, not just by capping individual positions. Desks group positions by shared driver, sector, factor, direction, and cap the net exposure of each group, then stress-test the book under the assumption that all correlations go to one. The practical rule is to size the portfolio so it survives the crisis-correlation case, treating calm-market diversification as a bonus rather than a support. Hedges and cash reserves, whose value rises when correlations spike, are the direct countermeasure to correlation risk.

Formula

σp = √(w1²σ1² + w2²σ2² + 2 w1 w2 ρ σ1 σ2); crisis case: set ρ → 1

σp = portfolio volatility; w1, w2 = capital weights; σ1, σ2 = individual volatilities; ρ = correlation between the two positions. The correlation enters only through the cross term 2 w1 w2 ρ σ1 σ2. Stress-testing correlation risk means recomputing σp with ρ forced to 1, which gives σp = w1σ1 + w2σ2, the fully undiversified worst case; the gap between the calm-ρ and ρ=1 figures is the size of the correlation risk you are carrying.

Assumed independence vs realised crisis correlation

AspectCalm-market assumptionCrisis reality
Correlation ρLow, near 0.2 to 0.4Converges toward 1
Diversification benefitLarge, volatility well below averageVanishes, volatility near the sum
Portfolio lossCushioned by offsetAll positions fall together
Effective betsSeveral independentEffectively one
What protectsSpread across namesHedges, cash and lower exposure

Practical example

Illustrative example (Indian market)

A trader with Rs 5,00,000 holds three long positions, a Nifty future, an IT largecap and an auto largecap, and feels diversified across sectors. In calm markets these correlate around 0.4, so the combined volatility is comfortably below the sum. On a sharp risk-off day, say a global shock, all three fall together as correlation jumps toward 0.9; if each position could lose Rs 15,000 in such a move, the trader who assumed offset budgeted perhaps Rs 30,000 of portfolio loss but actually faces close to Rs 45,000, a 9 percent hit rather than the 6 percent modelled. Recomputing the two-position formula with ρ near one shows portfolio volatility rising back toward the undiversified weighted average, which is the correlation risk that peacetime numbers concealed.

On NSE, banking, NBFC, real estate and auto stocks all carry heavy interest-rate sensitivity, so a hawkish RBI surprise can push them down together even though they sit in different sector buckets. A book that looks spread across four sectors can behave as one leveraged bet on rates, the classic hidden correlation.

Limitations

  • Correlation estimates are backward-looking and unstable, so any single number is unreliable
  • Pairwise correlation misses common-factor risk that only shows when the factor moves
  • Crisis correlations converge toward one, defeating the diversification the model assumed
  • Correlation measures only linear co-movement and can miss non-linear tail dependence
  • A short estimation window can show near-zero correlation that is spurious for the next regime

Common mistakes

  • Judging diversification by the number of positions rather than their correlations
  • Using calm-market correlations to size a book that must survive a crisis
  • Missing that different sectors share a common driver like rates or the dollar
  • Treating a historical correlation as a stable, causal relationship
  • Assuming a hedge with 0.8 historical correlation will hold in the next stress event
  • Ignoring that a near-zero correlation over one window can flip in the next

Professional usage

Risk desks model the full correlation matrix of the book, decompose exposure into common factors, and cap the net exposure to each factor rather than only to each position. They routinely re-run the portfolio risk with all correlations forced to one, sizing so the book survives that convergence, and they favour hedges whose value rises exactly when correlations spike. Correlation is treated as an unstable, regime-dependent estimate to be stressed, never a fixed input to be trusted.

Key takeaways

  • Portfolio risk is driven by correlations between positions, not their number
  • High correlation means apparent diversification hides a concentrated bet
  • Correlations converge toward one in a crisis, so peacetime numbers understate risk
  • Manage it by capping common-factor exposure and stress-testing with all correlations at one

Frequently asked questions

What is correlation risk?
Correlation risk is the danger that positions assumed to be independent actually move together, so the portfolio's combined loss is far larger than a diversified book would suggest. It is most acute in a crisis, when correlations converge toward one and supposedly separate positions fall as a single concentrated bet.
Why does correlation matter more than the number of positions?
Because the risk of a book depends on how its positions move together, not how many there are. Two highly correlated positions behave like one double-sized position, so a book of many correlated names carries concentrated risk despite looking diversified by count.
How does correlation affect portfolio volatility?
Through the cross term 2 w1 w2 ρ σ1 σ2 in the portfolio volatility formula. High positive correlation makes this term large and pushes volatility toward the weighted average of the parts, meaning little diversification; low or negative correlation shrinks it and lowers combined volatility.
Why do correlations rise in a crisis?
In stress, forced selling, margin calls, deleveraging and a flight to cash push almost all risk assets down together, so realised correlations converge toward one. The independence seen in calm markets reflects the absence of a common shock, and it disappears once a systemic shock forces everything to move as a block.
What is hidden correlation?
Hidden correlation is when positions share a common driver, such as interest rates, crude oil or the dollar, without an obvious pairwise link. They can slump together when that factor moves, so a book that looks spread across sectors is really one bet on the shared factor.
How do I measure correlation risk?
Compute the pairwise correlations across the book, group positions by common factor, and recompute portfolio volatility with all correlations forced to one. The gap between the calm-market and stressed figures is the size of the correlation risk you carry, and the factor groupings show where it concentrates.
Is a correlation of zero the same as safety?
No. A near-zero correlation over one window can be spurious or regime-specific, and it offers no promise of independence in the next period, especially in a crisis. Zero correlation is an estimate, not a guarantee, and it should be stressed rather than trusted.
Does correlation imply causation?
No. A measured correlation is a backward-looking statistic that can reflect a shared regime rather than a stable structural link. It can break or reverse when the regime changes, which is why relying on a single correlation estimate to justify a hedge is fragile.
How is correlation risk different from concentration risk?
Concentration risk is too much capital in one visible position; correlation risk is too much exposure to one hidden common driver spread across several positions. Correlation risk is concealed concentration, because the book looks spread out until the shared factor moves and the positions fall together.
How do desks manage correlation risk?
They model the full correlation matrix, decompose the book into common factors, cap net exposure per factor, and stress-test with all correlations forced to one. They also hold hedges and cash whose value rises when correlations spike, sizing so the portfolio survives the convergence case.
Can a hedge fail because of correlation risk?
Yes. A hedge relies on a stable negative or offsetting correlation with the position it protects, and that relationship can weaken or break in a stress event. A hedge sized on an 0.8 historical correlation can under-protect if the realised correlation shifts, so hedges are stressed, not assumed perfect.
Does correlation capture tail risk?
Only partly. Standard correlation measures linear co-movement and can miss non-linear tail dependence, where assets are loosely related in normal times but crash together in extremes. This tail dependence is exactly what correlation risk is about, and it is understated by ordinary correlation figures.
Why does correlation risk understate my true portfolio risk?
Because sizing a book on comfortable historical correlations assumes a small cross term in the risk formula. If the realised correlation rises toward one, that cross term becomes large and the portfolio's volatility and loss exceed what the model predicted, so the true risk was understated all along.
What common factors drive correlation on NSE?
Broad risk sentiment, interest rates, the rupee and the dollar, and crude oil are major shared drivers. Banking, NBFC, real estate and auto stocks are all rate-sensitive, so they can move together on a policy surprise, creating correlation risk across what look like distinct sectors.

Voice search & related questions

Natural-language questions people ask about Correlation Risk.

What is correlation risk?
It is the risk that your different trades secretly move together, so what looks like several bets is really one big bet that can lose all at once.
Why do my diversified trades all fall together?
Because in a scare almost everything drops at the same time. Positions that normally move apart suddenly move together, which is correlation risk showing up.
Is having many positions enough to be safe?
Not if they move together. What matters is how correlated they are, not how many you hold. Many similar positions are just one hidden bet.
Do correlations stay the same over time?
No, they change with the market. In calm times things look independent, but in a crisis they snap together, so you cannot trust a single correlation number.
What is hidden correlation?
It is when several trades quietly depend on the same thing, like interest rates or oil, so they all move together when that one thing moves, even across different sectors.
How do I protect against correlation risk?
Cap how much you bet on any single driver, and test your book as if everything falls together. Keep some cash and hedges for when that happens.

Sources & references

    Last reviewed 12 July 2026. Educational content only — not investment advice. Markets and rules change; verify current conventions with SEBI, NSE/BSE and your broker.

    ← Previous
    Diversification
    Educational content only — not investment advice. Examples use illustrative numbers and simplified models. Risk-management techniques reduce but never remove risk, and trading derivatives involves substantial risk of loss. See our Risk Disclosure and SEBI Disclaimer.