【摘要】This dissertation investigates the role of subjective expectations in the process of strategic decision-making.;Chapter 1 investigates the role of subjective expectations in the strategic problem of buyers and sellers participating in a double auction experiment. The experiment elicits agents' beliefs about the bidding decisions of other market participants using a quadratic scoring rule. I show that subjective beliefs cannot be modeled by equilibrium beliefs or by empirical/historical beliefs, and that elicited subjective beliefs help explain observed choices. Observed choices often deviate from the predictions of the Bayesian Nash Equilibrium model. Evidence suggests that the failure of the game to converge to equilibrium is due to subjective beliefs not converging to equilibrium beliefs.;Chapter 2 studies first- and second-order subjective expectations in strategic decision-making. I propose a method to elicit probabilistically not only first- but also second-order expectations and apply such method to a Hide-and-Seek experiment. The coherency of the elicited second-order expectations is supported by both nonparametric and parametric analysis. By measuring second-order beliefs probabilistically, not only do I find results in contrast with previous experimental work that uses nonprobabilistic measures, but I can also examine characteristics of the decision process that were before unobservable.;Chapter 3 develops the Quantal Response Equilibrium (QRE) model for a double auction environment. I characterize the solution to the QRE model through a system of differential equations and I show that the solution exists and is unique. The concept of QRE has the intuitive property that a deviation from the best-response decision is less likely the higher the cost associated with the deviation itself. Thanks to such property, the concept of QRE enables us to accommodate stochastic elements in the analysis of the strategic decision-making that arises in the double auction, where players' best responses are likely to be noisy. By providing a theoretical alternative to the BNE model, the QRE model offers an appealing tool for analyzing data of double auction experiments.