How Decay and Newey–West Adjustments Shape Barra Covariance Estimates

Summary

The document asks how Barra risk models construct covariance matrices from volatility and correlation estimates. It identifies exponential decay through half-life parameters and asks how Newey–West lag adjustments and a separate volatility-regime half-life enter the calculation.

It provides an illustrative decay example for a 90-day volatility half-life, but offers no explanation or answer to the questions. It therefore serves mainly as a description of parameters a researcher would need to investigate when replicating a Barra model. The treatment is incomplete: no estimation formulas, data requirements, or evidence about Barra’s proprietary implementation are supplied, so the questions cannot be resolved from this document alone.

Key ideas

• Barra covariance estimates may use half-life decay for both volatility and correlation inputs.
• The document asks how Newey–West lag adjustments affect volatility and correlation estimates.
• It also raises the role of a separate half-life parameter for volatility regime adjustment.
• No implementation method or answer is provided, so the post identifies research questions rather than presenting a reproducible procedure.

Tags

# Barra covariance matrix construction


# Barra covariance matrix construction












I am trying to replicate the covariance matrix used by Barra risk models.

All Barra models have half life parameters for volatilities and correlations (e.g. if the half life for volatlity is 90 days, the decay rate is $(1/2)^{90} \sim 0.991782$).

And there is Newey-West volatility and correlation lags with $n$ number of days. Does anybody know how do these Newey-West adjustments work within Barra framework? Another parameter that I am not sure of is half life for volatility regime adjustment. How is this incorporated in building covariance matrix?

Thank you!

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.