Only in simple cases can econometric theory deduce the properties of a statistics sampling distribution. In most cases, theory is forced to use asymptotic algebra, producing results that apply only when the sample size is very large. Although in many cases these asymptotic results provide remarkably good approximations to sampling distributions associated with typical sample sizes, one can never be sure. Because of this, econometricians have turned to the computer to discover the sampling distribution properties of statistics in small samples, using a method called Monte Carlo.
This thesis implements the approach of Monte Carlo simulation as an alternative to asymptotic approximation on the rational addiction model of coffee developed by Olekalns and Bardsley (1996). First, the study concentrates on the importance of the literature on the addiction theory. Second, it provides a more rigorous account of the rational addiction theory, through developing the model for rational addiction for coffee. Thirdly, the estimation of the model is analysed through a Monte Carlo technique and also another simulation method known as the bootstrap re-sampling procedure.
Performing the Monte Carlo experiments and bootstrapping has deepened the understanding of the small sample properties of the estimators of the model for coffee. Some insights are provided on the comparative performance of different estimation techniques like Ordinary Least Squares and Two-Stage Least Squares.