DFQ-2020: A Factor Risk Model for Chinese A-Shares

Summary

DFQ-2020 is a risk model for mainland Chinese equities that combines style factors, industry factors, and a market factor. It retains 40 factors and revises the construction of the Trend, Growth, and state-owned-enterprise factors relative to its predecessor. The model estimates factor exposures, factor returns, residual returns, factor covariances, and residual risk for portfolio analysis and optimization.

To improve risk forecasts during abrupt market moves, the approach models factor and residual returns with ARMA-GARCH processes, accounting for serial dependence and changing volatility. The document reports stronger explanatory power for CSI 300 constituents than for CSI 500 constituents or the full market. It also describes comparisons with the earlier model: closer individual-stock volatility and correlation forecasts during the 2015 turbulence, lower turnover in a minimum-variance portfolio, and improved tracking error and drawdown results for an enhanced CSI 300 portfolio. These are reported historical findings; the summary does not provide enough detail to assess test design, transaction costs, or performance outside the periods discussed.

Key ideas

• The model combines style, industry, and market factors to estimate A-share risk.
• ARMA-GARCH models are used for factor and residual returns to capture serial dependence and changing volatility.
• The 2020 version revised its Trend, Growth, and state-owned-enterprise factor definitions.
• The document reports more accurate stock risk forecasts during the 2015 market disruption than its predecessor.
• The risk estimates can support portfolio optimization, risk analysis, and performance attribution.

Tags

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