The first part of the thesis consists of three chapters focusing on herd behavior in financial markets. Chapter one reviews the herding literature while chapter two studies a market where informed and noise traders show up sequentially and anonymously in front of a competitive and risk neutral market maker. Traders can in some cases observe whether some of their predecessors were informed, although they cannot observe their private in- formation. This creates an informational asymmetry between the traders and the market maker which generates herd behavior. I find that herd and contrarian behavior is gener- ated more easily in better-informed markets than in poorly informed ones. Informational cascades can never occur and the market learns in the limit. Moreover, I illustrate how a market dominated by herding features a price that is more informative of the asset value than the price of a market where traders always follow their signal. I also discuss how contrarianism has the exact opposite effect by decreasing price informativeness. In chap- ter two I consider the case of multiple trading rooms, where traders can in some cases observe whether some of the predecessors coming from the same room were informed. I first analyze herding conditions for the case of disconnected rooms where agents trading during the same time exhibit information correlation, and find that herding is more likely to occur in a market with positive correlation than in a market without correlation. I then link rooms by means of a network structure which dictates which rooms' predecessors one can observe. I check whether it is possible for a trader to herd with traders outside his own neighborhood instead of with his direct neighbors. I find that the answer to this question is negative and that herding cannot spread from one part of the market to another. Finally, I bring together information correlation and the network structure and I illustrate the example of a market where there are trading histories such that herd behavior can lead to the complete loss of information and, once herding has started, learning can be recovered only if noise traders enter the market. In the second part of the thesis I build a signalling model of delegated portfolio management where the manager can be of different qualities which affect the performance of the closed-end fund under his management. I find that in his effort to appear of high quality, the manager sends signals to the market which affect the share price of the fund in such a way that momentum and reversal are generated. While in the momentum phase, the price accumulates a discount with respect to its net asset value; during the reversal phase, the discount narrows and the price reverses back towards the net asset value of the fund.