Comparing Models for Multi-Asset Portfolio Allocation

Summary

This Chinese-language survey discusses fundamental allocation, mean-variance optimization, Kelly-CVaR, Black-Litterman, and risk-parity approaches. Fundamental methods include mapping economic-cycle phases to asset weights and using macroeconomic factors to forecast both return direction and magnitude. The text reports historical backtest metrics for several approaches, including the economic-cycle model, a two-dimensional macro model, a mean-variance variant using Kaufman’s adaptive moving average for expected returns, and Kelly-CVaR. These figures are reported by the source and are not independently validated here.

The discussion highlights model trade-offs. Mean-variance weights can be sensitive to small changes in estimated returns and risk. Kelly-CVaR combines growth-oriented allocation with a constraint on average portfolio loss. Black-Litterman blends equilibrium return assumptions with subjective investor views through Bayesian inference, but results depend heavily on how those views and their uncertainty are specified. Risk parity balances asset risk contributions; the survey also describes risk budgets, leveraged variants, factor-based approaches, and a downside-risk version. The supplied text is only an abstract, so details of data, test design, and implementation are unavailable.

Key ideas

• The survey compares fundamental, mean-variance, Kelly-CVaR, Black-Litterman, and risk-parity allocation methods.
• Economic-cycle and macro-factor approaches translate economic information into portfolio weights.
• Mean-variance allocations can change sharply when estimated returns or risks shift slightly.
• Kelly-CVaR targets portfolio growth while constraining average loss to a chosen level.
• Black-Litterman results depend on subjective views and their uncertainty inputs.
• Risk parity balances portfolio risk contributions, with variants based on budgets, factors, leverage, or downside risk.

Tags

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