Skip to content
All library documents

Constant and Dynamic Models for Forecasting Asset Correlation

Article Quant Q&A · Author: Thomas Johnson
Read the original

Summary

The document asks how to forecast future correlations between securities, contrasting this problem with more familiar volatility and return forecasting. Its brief answer identifies constant conditional correlation as a baseline modeling assumption, as used in a CCC multivariate GARCH framework. This provides a simple reference point: correlations are held fixed while conditional variances may still evolve over time.

For models that allow a richer treatment of dependence, the answer points readers toward a survey of multivariate GARCH methods. It does not compare model specifications, give estimation procedures, or present empirical forecasting results. The material is best read as an entry point to the subject; deciding whether constant or time-varying correlation is appropriate requires further modeling and evidence for the assets and data at hand.

Key ideas

  • A constant correlation assumption can serve as a baseline for forecasting dependence.
  • CCC multivariate GARCH models pair constant correlations with conditional volatility modeling.
  • Multivariate GARCH surveys provide a route to more flexible correlation specifications.
  • The document offers no empirical comparison to establish which model forecasts best.

Tags

StatisticsVolatilityPortfolio construction
Full text
# How can I forecast future correlation?


# How can I forecast future correlation?












There are some standard models for forecasting volatility (e.g., GARCH) and for forecasting returns (e.g., factor models). What kind of standard models exist for forecasting future correlation between securities?

## Answer by Student tea (score 2, accepted)

https://quant.stackexchange.com/a/16510

One common "model" is to assume the correlation to be constant, such as in a CCC-MVGARCH model. If you want a review of different multivariate GARCH models, you could look at:

Silvennoinen and Täräsvirta 2009, Multivariate Garch models, in Handbook of financial time series.

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.