| Summary: | The objective of this thesis is to avoid misspecifications and to seek efficiency improvements in cross sectional and time series econometric applications using semiparametric methods. We restrict our attention to single equation models and the use of conditional moment restrictions as well as maximum likelihood methods. The first part of the thesis deals with cross sectional studies on the United Kingdom car market and the second part deals with time series studies of the United States consumption function. There are five main contributions of the thesis. First of all, we have suggested minor extensions of existing semiparametric models; secondly, we have suggested the use of a dimensional reduction method prior to nonparametric estimation; thirdly, we have investigated the use of various rules of subjective and automatic bandwidth selection methods using real and simulated data; fourthly, we have suggested a new approach to overcome problems in the hedonic approach for cross sectional studies,; and finally, we have established a relationship between expected real interest rate and consumption using US time series data.
|
|---|
