Methods for Estimating Historical Correlation Matrices

Summary

The document asks how researchers should estimate correlation matrices from historical financial data. It highlights choices that can affect the result, including observation frequency, sample size, and the method used to extract correlations from time series. It seeks literature that can guide those decisions and reduce estimation error.

The response points to three broad approaches: principal component analysis, random matrix theory, and shrinkage estimators. These methods offer ways to identify shared structure, address noise in estimated matrices, or regularize estimates. The document is a brief list of research directions rather than a practical comparison or step-by-step procedure. It does not specify how to select a method, sampling frequency, or window length, so those choices remain dependent on the data and intended use, such as portfolio or risk modeling.

Key ideas

• Historical correlation estimates depend on sampling frequency, sample size, and the estimation method.
• Principal component analysis can help identify common structure in asset returns.
• Random matrix theory is cited as an approach to cleaning noisy covariance estimates.
• Shrinkage is another referenced technique for improving covariance matrix estimates.
• The document lists research directions but does not compare methods or prescribe parameter choices.

Tags

# Sample size and historical correlation matrices


# Sample size and historical correlation matrices












> I was wondering whether any literatures existed on how to properly estimate correlation matrices from historical data.

Obviously the entire procedures allows a lot of leeway. The frequency of the data (daily, montly etc.), sample size, the atual method used to extract the correlation from the time-series and so on

Given the amount of decisions that have to be made and thus the amount of errors that can be made as well it would be nice to have some sources to turn to.

## Answer by lehalle (score 2)

https://quant.stackexchange.com/a/11531

There is a wide knowledge on correlation estimation, see other questions and answers:

- principal component analysis (PCA) - Equity Risk Model Using PCA

- random matrix theory (RMT) - Cleansing covariance matrices via Random matrix theory or Random matrix theory (RMT) in finance

- shrinkage - Portfolio Optimization : Shrinkage of Covariance Matrix when data is available

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.