Risk Attribution from Returns Without Portfolio Weights

Summary

The document asks how to attribute the risk of a broad equity portfolio when its constituent sector and regional weights are unavailable. It proposes risk-based performance attribution: regress the portfolio’s returns on sector index returns and use the estimated coefficients as exposures to calculate each sector’s contribution to performance over time.

The method replaces known holdings weights with statistically estimated return sensitivities. Its evidence is a brief explanation rather than an empirical example, and it does not provide a risk decomposition formula, diagnostics, or validation results. Regression coefficients need not equal actual portfolio weights: correlated sector returns can make exposures unstable or difficult to interpret. The approach therefore offers a possible attribution framework, not a direct recovery of the unknown holdings or a complete account of portfolio risk.

Key ideas

• Risk-based attribution can use return exposures when portfolio holdings weights are unavailable.
• Regressing portfolio returns on sector index returns estimates sensitivities to those sectors.
• Estimated coefficients can be used to calculate period-by-period factor performance.
• Correlated indices and model specification can affect the interpretation of estimated exposures.

Tags

# Risk attribution model without weights data


# Risk attribution model without weights data












I was just wondering if there are risk attribution model that does not require the asset weights data. It appears that most risk attribution models do require asset weight data. I am looking for model that does not require the asset weight data;

I am trying to build the risk attribution model for MSCI ACWI, where the underlying assets are sectors and regions data. The model should assume that I do not know the weights for the sectors and regions.

Thank you!

## Answer by Tim Wilding (score 1, accepted)

https://quant.stackexchange.com/a/40909

I think you are looking for a technique called Risk-Based Performance Attribution (see http://www.northstarrisk.com/004riskbased-performance-attribution for a simple overview). That uses various risk factors or risk exposures to decompose the performance of the portfolio.

In your case, multiple regression of the MSCI ACWI returns will generate regression coefficients against the sector indices. The regression coefficients are the risk exposures and can be used to generate the factor performance for each period.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

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