Skip to content
Learning path Gold Professional operator

Diversification

A process that combines weights, exposures and dependencies to reduce portfolio concentration. Its effect depends on data and scenario and does not guarantee protection.

Who this is for — Anyone checking whether many positions genuinely distribute risk or reproduce the same exposure under different names.

Diversification is the process of combining weights, sensitivities and dependencies with the objective of reducing portfolio concentration. The result depends on exposures, data and the scenario considered: it does not guarantee that no source will dominate under every condition. Counting instruments, strategies or accounts does not measure the achieved effect.

In the linear volatility case:

σₚ² = ∑ᵢ (wᵢ² × σᵢ²) + 2 × ∑ᵢ<ⱼ (wᵢ × wⱼ × σᵢ × σⱼ × ρᵢⱼ)

The cross terms explain why two portfolios holding the same instruments at different weights can have different risk. The formula alone does not capture nonlinearities, liquidity, default, jumps or tail dependence.

Portfolio risk comes from exposures and dependencies Addition, offsetting and diversification change with metric, model and scenario Portfolio risk comes from exposures and dependencies Addition, offsetting and diversification change with metric, model and scenario POSITIONS DEPENDENCIES AGGREGATE VIEWS Position A weight · notional · factors Position B currency · payoff · liquidity Position C leverage · counterparty · horizon 1.00ρABρACρAB1.00ρBCρACρBC1.00 Correlation measures a linear relation in asample; it is not causation and does not byitself describe tails. Gross / net Accounting offsets do not removeevery form of risk. Contributions Which positions and factors drivethe chosen measure? Stress What if dependencies, volatilityand liquidity change? Cyclepedia · source-checked visual explainer
A diversified structure aims to distribute contributions and vulnerabilities; the result must be measured, not inferred from ticker count.

Look through the labels

Dimension Control question
Name / issuer do several instruments depend on the same entity?
Sector and geography do revenues, funding or regulation react to the same shock?
Market factor do equity beta, rates, credit, volatility or currency dominate?
Strategy do different signals create similar positions at the same time?
Liquidity do exits require the same market or time window?
Counterparty and custody are capital and collateral concentrated at one entity?

Five long equities are not “one trade” by definition: they may reduce idiosyncratic risk while retaining strong common market or sector exposure. Likewise, trend, mean reversion and market-neutral labels do not create diversification automatically. Measure them after weights, leverage, costs and behaviour in adverse scenarios.


A verifiable process

  1. Choose the unit: market value, beta, delta, DV01, volatility or scenario loss.
  2. Look through to actual instruments, issuers, currencies, factors and counterparties.
  3. Calculate weights and contributions at the same date and over the same scope.
  4. Estimate dependence on aligned returns and test sensitivity to window and regime.
  5. Apply common scenarios and liquidity constraints; low average correlation is not enough.
  6. Set concentration limits for the mandate rather than adopting universal thresholds.

Ex-ante risk reduction is an estimate. New data, weight changes, leverage and market moves can quickly transform portfolio structure. Diversification therefore requires monitoring and reconciliation, not permanent asset labels.

Common mistake — Calling a second position a “hedge” merely because it belongs to another asset class. A hedge needs a defined economic relation and size, and may add basis, liquidity and counterparty risk.


Sources