This thesis includes three interconnected essays which, building on the work by Hart and Zingales (2011), lay down the foundations for a new theory of financial intermediation. The first essay explains the Hart and Zingales (HZ) framework and shows that their results are not general. In the HZ model, there is a lack of simultaneous double coincidence of wants, and future income is not pledgeable. This implies that agents need money to trade. However, holding money entails an opportunity cost that leads to a waste of resources. Because of this inefficiency, pecuniary externalities have welfare consequences that private price-taking agents fail to internalize. I find that HZ's result, whereby the market produces inefficiently high levels of liquidity, cannot be generalized, because the conflict between private and social incentives to create money depends on agents' preferences. In the second essay I construct a framework that explains the transactions, precautionary and speculative demand for money. Again, the welfare analysis indicates that, depending on individuals' preferences, the market may produce inefficiently high or low levels of liquidity. The results also evidence that the speculative demand for money exists only when households are risk averse in their wealth. In that case, private and social incentives to hold money are stronger, but the market produces insufficient means of payment relative to the social optimum. The third essay introduces active financial institutions, and examines the role played by moral hazard in the provision of and demand for liquidity. Limited liability and the non-contractibility of bank investment policy induce highly levered financial institutions to invest in an inefficient gambling asset. I find that, when the probability that banks gamble is non-zero, the primary goal of public intervention is to address the moral hazard problem by restricting the creation of liquidity. Several policies to address this inefficiency are discussed and analyzed.