In the first two essays, I study value discovery in discrete-time dynamic markets with imperfect information. One model examines a market with no payoff externalities while the other studies a market with a negative or crowding payoff externality. Short-lived buyers encounter infinite-lived sellers who provide heterogeneous quality goods. The number of trades at each seller in the prior period is public information. At first, buyers do not have information about the type of the good offered at any seller, but they can acquire private information about goods at a single seller. When there is no crowding, I find that coordination frictions and informational cascades prevent efficient outcomes in markets with no search costs as well as those with costly search. These markets continue to be inefficient even as the number of buyers goes to infinity. With crowding, the market endogenously segments into areas of known quality and unknown quality. In regions of unknown quality, informed traders drive out uninformed traders, reducing trade and forcing uninformed traders into regions of known quality--a pattern that superficially resembles risk-aversion, though buyers are risk-neutral. As time goes on, public information grows, in contrast to many herding models where information is lost. This market becomes more efficient over time as fewer buyers incur search costs each period. The third essay is a study of a dynamic principal-agent scenario where there is no commitment. These situations suffer from hold-up problems--that is, projects that benefit both parties are not initiated. Research has been done on dynamic principal-agent games where gradual investment ameliorates the hold-up problem. The model here is a dynamic game where the total amount of investment to complete the project is a random variable whose realization is unknown to both principal and his agent. Because completion of the project is uncertain, the dynamics of the game evolve into a ongoing relationship. For some parameter values, the project is initiated and has positive surplus in expectation. In certain cases, however, the project is never started even though the social planner's first-best solution has positive expected surplus.