Handling Short Histories in Covariance-Based VaR Estimates

Summary

The document considers how to estimate a stock-return covariance matrix for portfolio value-at-risk when some holdings have only one or two years of data, while the desired estimation window is longer. It asks whether to fill unavailable observations with zero returns or exclude those stocks from the risk calculation. The setting is a long-only equity portfolio where risk measures also need to be explainable to decision-makers.

The response suggests using a factor model, such as a single-index model, to reduce the difficulty caused by incomplete histories. For a stock with little or no usable data, it proposes estimating exposure and residual risk from comparable companies in the same industry. This offers a practical proxy-based direction rather than a full estimation recipe. The document does not specify factor estimation details, proxy selection rules, or how to validate the resulting VaR, so those choices remain important implementation limits.

Key ideas

• Short return histories complicate covariance estimates used for portfolio VaR.
• The response recommends a factor model to reduce dependence on complete pairwise return histories.
• Industry peers can provide proxy estimates of beta and residual risk for data-poor stocks.
• The answer does not specify proxy selection or validation procedures.

Tags

# Covariance matrix for VaR: what to do with missing data?


# Covariance matrix for VaR: what to do with missing data?












I need to compute a covariance matrix using three years of stock returns from a portfolio which has a couple of stocks with only one or two years of history (being relatively new stocks).

Should I replace the missing data with zeroes, or should I omit the offending stocks from my VaR calc?

edit: I'm interested in knowing how other quants deal with this common occurrence. To put it in context, I'm the sole quant at an old-school, long-only stock-picking firm. I like to keep risk metrics like VaR simple enough to explain to my bosses, but they still need to be correct.

## Answer by Alexandre Oliveira (score 1)

https://quant.stackexchange.com/a/33872

There are several ways, but to keep things simple: use a factor model like single index model and you will reduce a lot this problem. If you have no data (very iliquid stock), consider using a proxy from average beta and residual risk for stocks in similar industry.

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.