UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

On dynamic choice models, agent's beliefs and unobserved heterogeneity Hao, Yu


Chapter 2 provides a tractable model that separates firms’ incentive problems and coordination problems during the initiation of collusion. In the Chilean pharmaceutical industry, firms collude through price leadership. Collusion gradually diffuses among markets: firms collusively raise prices in a couple of markets per week. We propose a model of price leadership under the dynamic pricing game framework to incorporate the coordination problems by allowing firms’ beliefs about competitors’ conduct to be biased towards a competitive equilibrium. As firms observe supracompetitive prices, they adaptively learn that competitors are willing to collude. We show that gradualism is explained by the heterogeneous market characteristics as well as firms’ learning to coordinate. Chapter 3 develops the likelihood-ratio based test of the null hypothesis of a m₀-component model against an alternative of (m₀+1)-component model in the normal mixture panel regression. I show that the normal mixture panel regression does not suffer from the Fisher Information matrix degeneracy under the reparameterization proposed in Kasahara and Shimotsu [89]. As a result, the likelihood ratio test statistic can be approximated by a local quadratic expansion of squares and products of the reparameterized parameters. Moreover, I obtain the data-driven penalty function via computational experiments to attend to the unbounded likelihood ratio. In addition, I apply the test to random coefficient Cobb-Douglas production function estimation following the framework of Gandhi et al. [63] and Kasahara and Shimotsu [90]. The empirical findings suggest evidence of heterogeneous production technology beyond the Hicks-neutral technology factor. Chapter 4 develops a modified expectation-maximization(EM) algorithm to incorporate unobserved heterogeneity for the dynamic discrete choice model that does not require the finite dependence property. Following the Euler Equation(EE) representation of dynamic discrete decision problems, we provide an alternative conditional choice probability (CCP) value function representation that relies only on the CCP of one action. We illustrate the computational gains with Monte Carlo simulations.

Item Media

Item Citations and Data


Attribution-NonCommercial-NoDerivatives 4.0 International