Search

This thesis is organized in three chapters. In the first two chapters, an econometric model selection procedure and a method to improve some existing estimators are proposed. In the third chapter, a theoretical microeconomic analysis of herd behavior is performed under a fairly new set of assumptions. In chapter one, a model selection procedure based on the Penalized Empirical Likelihood(PEL) technique is developed, and guidelines are provided for the extensionof the procedure to the setting of Generalized Empirical Likelihood (GEL). The procedurewas initially applied to linear models and was called "Least Absolute Shrinkage and Selection Operator" (LASSO). It was subsequently extended to Generalized Method of Moments models in, and we now extend it to Empirical Likelihood (EL)models. Its main advantage over classical methods is in the combination of model selection and model estimation into a single step, while improving the post-selection properties of the resulting estimators. This procedure is easy to implement, and it remains computationally feasible even in models with a large number of parameters. A simulation study is performed to compare the newly proposed procedure to someclassical methods such as AIC, BIC, and DT. The simulation results show a better performance of the new procedure. In chapter two, we define the modulation technique for the EL estimator modulation technique pertains to the class of methods generally known as "shrinkage methods".Shrinkage methods are frequently used to improve the properties, in particular smallsampleproperties, of existing estimators. In this paper, a general theoretical analysisof modulation estimators is developed for EL models, along with a discussion of howthey can be implemented in special cases.In chapter three, a theoretical model of imitation and herd behavior is considered.It is assumed in that some participating agents have specific abilities to affect otherpeoples behavior. Results are provided on how "stars"