With dual trading, brokers trade both for their customers and for their own account. We study dual trading and find that customers who are less likely to be informed have higher expected profits with dual trading while customers who are more likely to be informed have higher expected profits without dual trading. We also examine the effects of frontrunning. We test the major empirical implications of our model. Consistent with the model, dual traders earn higher profits than non-dual traders, and customers of dual-trading brokers do better than customers of non-dual-trading brokers.