Abstract

In recent years，the dynamic analysis of imperfect competition market grows rapidly. Unlike traditional research in IO study which based on comparative static models or reduced form models，these studies exploit fully dynamic models of strategic competition between enterprises. These studies based on the Markov Perfect Equilibrium（MPE）concept which was proposed by Tirole et al.，and the Ericson-Pakes Framework. They combined game theories，econometric methods and numerical methods in dynamic IO research. Compared with traditional research methods，there are three characteristics in the new analysis paradigm. Firstly，the process of competition between enterprises are fully dynamic. Secondly，the new paradigm emphasizes the importance of strategic competition. Firms’ decision are based on their rational expectation of competitors’ action. Thirdly，as a fully dynamic and more realistic model，the new paradigm does not use analytical methods in the computation of equilibrium，but turn to compute-intensive numerical method. Thus it relaxes the constrain on dynamic model setting.

This thesis based on the new dynamic IO Research paradigm，explores three aspects of dynamic market analysis：dynamic modeling，calculation of Markov Perfect Equilibrium（MPE），and the structural econometric estimation. Taking advantage of the new paradigm and numerical algorithms，this thesis explores two problems in network industries and Chinese mobile communication market. One problem is the influence of asymmetric regulation policy，another problem is number portability policy.

This thesis contains four sections.

Firstly，we study dynamic modeling and equilibrium calculations in chapter 2. The Ericson-Pakes Framework is at the core of dynamic IO study，modeling ideas of different literatures are similar，but implementation details are different. Therefore，we summarize the development of computational algorithms of MPE，discuss different algorithms and its variations，as well as the links between different algorithms. This summary is the methodological basis for applied research in chapter 3 & 4. We sort out the literatures on dynamic IO model construction，compare the differences between different literatures. Then we discuss the definition and computation of MPE. There are several computation algorithm of MPE，e.g.，the Pakes-McGuire Algorithm，the Backward Recursion Method，the Stochastic Iteration Algorithm，the Homotopy Continuation Methods，etc. Each algorithm has its advantages and drawbacks. The computation of the MPE is still in its infancy，facing the convergence failure，heavy computational burden，and multiple equilibria problem. Thus this summary is useful to select the appropriate method in the following applied research. Because computing MPE needs complicated mathematical tools，requiring high programming skills，it is rarely used in domestic IO research. Sorting out these methods will be helpful to promote dynamic IO research in China.

Secondly，we analysis the influence of asymmetric regulation in network industry in chapter 3. We analysis the tipping trends in network industry，and the influence of asymmetric regulation. Dynamic simulation shows that asymmetric regulation could prevent market tipping，but it also softens competition. We find that asymmetric regulation will raise the long-term average price. Due to network effects，consumers’ demand are inelastic，thus the costs of regulation shift to consumers. Therefore，policy maker should carefully evaluate asymmetric regulation policy in order to make informed decisions.

Compared with the traditional research paradigm，this thesis has several contributions. First，we could avoid the end-of-the-world effects in the two-period static model. Second，compared with single state variable markov model，our multi states model can handle more than two companies，and avoid the “zero-sum game” defect. Third，we use both Pakes-McGuire Algorithm and Backward Recursion Method to calculate the equilibrium，which make our study more stable.

Thirdly，we study the coexist of switching cost and network effect in chapter 4. By numerical simulation，we find that when network effect exists，the switching cost has non-monotonic influence on market dominance. Strong switching cost will lead user lock-in，dampen market tipping，while moderate switching cost makes large firm more aggressive，leads to market dominance. In addition，combination of lower switching cost policy and pro-compatibility policy will reduce market tipping，and increase competition，thus they are preferred. From this perspective，we propose a explanation of delay of mobile number portability（MNP）policy in China.

Fourthly，in chapter 5，we turn to structure estimation of dynamic interaction model，which used to be a difficult problem. Recently，the development of two-step estimation subtly bypasses the difficulty of computing burden in computing equilibrium. We propose an estimator for incomplete game with continuous actions，which introduce nonparametric estimation in the first stage. Numerical simulations are used to demonstrate the finite sample properties of the two-step estimator.

Keywords：Applied Dynamic IO Analysis，Ericson-Pakes Framework，Markov Perfect Equilibrium，Network Effect，Two-step Estimation