Work
Individual ResearchAsset Allocation · Risk · Robustness

Portfolio Construction & Risk Allocation

Two portfolios can hold similar assets while concentrating risk very differently. The study compares both weights and risk contribution, then tests whether sophisticated optimisation earns its complexity.

Central findingIn the development sample, simple 60/40 led while most sophisticated methods were statistically indistinguishable from equal weight on Sharpe; minimum variance mainly reduced risk rather than raising return.

Evidence

The result in context

ETF proxies
9
Development 60/40 Sharpe
0.82
Minimum-variance development volatility
3.66%
Maximum-Sharpe average monthly turnover
39.55%

Question

Which portfolio-construction methods remain useful once estimation error, concentration, costs and changing market regimes are taken seriously?

A comparison of simple and optimised allocation methods under estimation error, concentration, turnover, stress periods and frozen confirmation.

Risk contributions. Two portfolios can hold similar assets while concentrating total portfolio risk very differently.

Comparison

Weights are not the same thing as risk

The study compares equal weight, traditional 60/40, inverse volatility, minimum variance, risk parity, hierarchical risk parity and maximum Sharpe portfolios across nine liquid asset proxies.

Risk contribution estimates how much each holding contributes to total portfolio volatility. A small weight can still dominate risk if the asset is volatile or highly correlated with the rest of the portfolio.

Average capital weights provide the companion view: allocation and risk are related, but they are not interchangeable.

Development

Complexity did not guarantee a better result

From the common 2011 start, 60/40 produced 8.93% CAGR and 0.82 Sharpe. Most sophisticated methods were statistically indistinguishable from equal weight on Sharpe.

Minimum variance delivered 3.66% annual volatility and a −13.10% drawdown, showing a clearer risk-reduction role than a return-enhancement role.

Portfolio method explorer

Weight is not risk contribution

Compare source-backed portfolio-level results by method and period.

Method
Period
Capital weightHow much money is allocated
Risk contributionHow much total volatility it creates
CAGR
4.17%
Sharpe
0.39
Volatility
8.85%
Maximum drawdown
−18.54%
Turnover
3.10%
Effective assets by weight
8.99

Equal capital does not create equal risk: equity and real-asset exposures still dominate.

The figures on this page show asset-level weights and risk contributions. The explorer reports source-backed portfolio-level metrics only.

Stress-period comparisons show that a method’s full-period average can conceal materially different regime behaviour.

Frozen confirmation

A strong short period is not proof of superiority

From 2 January 2025 to 24 August 2026, a 10% volatility-targeted maximum-Sharpe portfolio produced 26.13% CAGR and 1.82 Sharpe, versus 19.05% and 1.61 for equal weight.

Minimum Variance Shrinkage recorded the shallowest confirmation maximum drawdown at −2.32%. The period is short and was not treated as a new optimisation sample. Its 1.82 maximum-Sharpe result is context, not a superiority claim.

Limitations

What this evidence does not establish

  • The confirmation period is too short to establish stable method rankings.
  • Optimised weights remain sensitive to estimated returns, covariance and implementation assumptions.
  • The nine-ETF universe is selected ex post and remains exposed to product-survivorship bias.
  • SHY is a simplified cash proxy and hurdle rather than a complete financing model.
  • Transaction costs and financing assumptions are stylised rather than broker- or mandate-specific.

Source and reproducibility

Trace the evidence

Source code, evaluation outputs and supporting material are available in the repository.

View repository
  1. Risk contributionsoutputs/figures/risk_contributions.pngCommit / evidence ID: 9a40b2bd97ca3e54ef105a964ff9f375f98be7f4
  2. Average weightsoutputs/figures/average_weights.pngCommit / evidence ID: 9a40b2bd97ca3e54ef105a964ff9f375f98be7f4
  3. Stress-period comparisonoutputs/figures/stress_period_comparison.pngCommit / evidence ID: 9a40b2bd97ca3e54ef105a964ff9f375f98be7f4
  4. Frozen protocolconfig/frozen_protocol_v1.jsonCommit / evidence ID: 9a40b2bd97ca3e54ef105a964ff9f375f98be7f4