Cross−impact and price bubbles in hybrid financial markets

We explore cross-impact in a hybrid experimental market with human and artificial agents, varying liquidity across treatments. In treatment Separated participants hold distinct portfolios for two stocks, while in Integrated they hold a unique portfolio, i.e., can freely move capital between assets. Larger bubbles and asymmetric cross-market impact occur with unique portfolios and decreasing value asset. When comparing experimental and synthetic data, cross-impact is attributed to human players, especially when stock values are close to each other. Artificial players also react to human presence, contributing to cross-impact.