I’m exploring the relationship between macroeconomic variables and the time-varying correlation between asset classes. Specifically, I’ve constructed a monthly rolling correlation (e.g., between stock and bond returns) and would like to use it as the dependent variable in an ARDL framework.

Would this approach be methodologically valid? Should I be mindful of any statistical concerns or precautions (e.g., induced autocorrelation, stationarity issues)?

Any insights or relevant references would be greatly appreciated.