Multi-Asset Allocation Through Tail-Risk-Focused Portfolio Optimization

Summary

This report challenges mean-variance optimization assumptions that returns are normally distributed, volatility captures risk symmetrically, and portfolios should maximize return per unit of risk. It instead frames investor concerns as preserving principal and meeting investment targets. Kernel density estimation and a multivariate normal transformation are used to model skew, kurtosis, and cross-asset dependence; Monte Carlo simulations then estimate the proposed risk measure. Portfolio weights are chosen by minimizing that measure.

The reported backtests combine A-shares, Hong Kong and US equities, government bonds, and gold over roughly a decade, with different results at two rebalancing frequencies. The authors attribute the approach’s behavior mainly to tail-risk control and report relatively low momentum exposure. These are historical backtest claims, not evidence of future performance; the summary provides limited detail on implementation, costs, constraints, or robustness. The authors also identify broader monitoring dimensions and conditional probabilities as areas for future development.

Key ideas

• The report argues that normal returns and symmetric volatility risk can misrepresent real portfolios.
• It defines risk around principal loss and failure to reach an investment target.
• Kernel density estimation and dependence modeling generate inputs for Monte Carlo risk estimates.
• The allocation process minimizes the resulting risk measure across a multi-asset portfolio.
• Historical backtests are reported, but the summary gives limited information about costs and robustness.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.