Using Conditional Correlations to Study Asset Relationships

Summary

The document raises the question of whether splitting asset return correlations by direction can reveal relationships hidden by a single overall correlation. It considers comparing correlations during upward and downward moves, including lagged relationships, and asks how to distinguish causation from association when modeling asset prices. These are useful questions for feature design and dependence analysis, but the document does not provide a proposed estimator, empirical examples, or answers about available software.

Its central motivation is that one correlation number may obscure differences across market regimes or return signs. Conditional analysis can be informative, but results depend on how the subsets are defined and may be unstable when observations are sparse. Lagged correlation alone does not establish causation; common drivers, selection effects, and time variation can produce misleading patterns. The text cites a paper as its inspiration but presents no evidence from it, so it serves as a research prompt rather than a tested method or trading strategy.

Key ideas

• Overall correlation can hide different relationships during positive and negative returns.
• Separate conditional correlations may help examine asymmetric asset dependence.
• Lagged correlation indicates temporal association but does not by itself establish causation.
• The document poses these questions without supplying an estimator or empirical results.

Tags

# How to extract informative value from correlations of assets? Subadditivity of correlation calculation an issue


# How to extract informative value from correlations of assets? Subadditivity of correlation calculation an issue












I was reading Nassim Taleb's Paper: Fooled by Correlation and found it very informative. I had always struggled with finding value in correlation in Finance, especially seeing a lot of bad applications in the workplace, however this paper got me thinking.

Would splitting up correlation into subsections as shown here have informative value for model building?

Say I wanted to model the % price change (or the log % price change) for a pair of assets, would it be better to look at the correlations for the downwards movements and upwards movements separately?

Are there any risks to doing so for lagged correlations between the variables?

A more general question as well, how do we prove causation in asset price relationships with different features in our model?

Also, do there exist any Python libraries that calculate these subsets of the overall correlation?

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.