Correlated factors under ridge and lasso

Two standardised factors are nearly interchangeable. How might ridge and lasso distribute their exposure, and what stability checks would you run?

Ridge tends to distribute exposure across correlated predictors, whereas lasso may select one and suppress the other. Lasso selection can change sharply under small sample changes; exact duplicates can make lasso coefficients non-unique. Refit across chronological windows and perturbations, comparing coefficients, selected groups and predictions. Elastic net can encourage grouped retention, but validation must still establish predictive value.