Front-runners: predators or liquidity providers?

I develop a dynamic model of strategic trading where front-running occurs in equilibrium. I show that the presence of traders who front-run may uniformly improve liquidity in all states, while episodic presence shifts liquidity towards the states where he is present. Improved liquidity increases  welfare, but state-contingent restrictions of front-running may be beneficial because it moves liquidity to periods where it is most needed. The effect of restricting front-running without considering the effect on liquidity is different and may improve welfare for some traders. This reconciles the conventional "predatory" view of front-running with the "welfare- and liquidity-improving" view of this paper. The model provides several novel empirical predictions. I provide empirical support for key predictions of the model using an event where the departure of a non-fundamental trader caused a long-lasting reduction in liquidity.