Thesis: This thesis is made of three independent chapters. The first chapter contributes to the literature on expectations. I argue that they may learn a misspecified model instead of learning the rational expectation model. I consider a simple economy with two types of agents. Rational learners learn the true model of the economy whereas consistent learners learn an autoregressive model. I show that a long run equilibrium exists in which consistent learners dominate. Simulations show that the economy may converge towards it. The second chapter deals with the intertemporal choice. I consider a model with wealth in the utility. I study the case of nonseparability. This disentangles between the income effect on labor supply and the intertemporal substitution effect. I derive several implications for economic policy. Then, I estimate the two new parameters introduced in the paper. I find large and positive values for both. The third chapter builds a model of corporate investment under adverse selection. My contribution is to provide a tractable model easy to embed into a macroeconomic model. Borrowers differs by the riskiness of their investment project like in Stiglitz and Weiss (1981). They have infinite horizon and signal their type by borrowing a fraction of their retained earnings. I get an analytic solution for the incentive constraint. I integrate the relation into a dynamic model and derive some implications.