Three Applications of Quantile Regression Approach on Government Size, Wage Inequality and Educational Achievement

• Main Authors: Sheng Tung Chen, 陳盛通
• Other Authors: 陳吉仲
• Format: Others
• Language: en_US
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/58068425546337494317

博士 === 中興大學 === 應用經濟學系所 === 95 === As Koenker and Hallock (2001) indicate that quantile resression may be viewed as a natural extension of classical least squares estimation of conditional mean models to the estimation of an ensemble of models for conditional quantile functions. Recently, large researches have pay attention to this econometric method. For instance, there are a lot of empirical quantile regressions researches in labor economics (Chamberlain, 1994; Buchinsky, 1994) and educational achievement analysis (Arias et al., 2001). The relationship between variables can go beyond models for conditional mean while the quantile regression can be considered. If the estimation of quantile regression with panel data can be defined clearly, the empirical research with short data span can be solved. So, the empirical models which apply quantile regression can estimate the result with panel data. The problems of using quantile regression to estimate and compare the robust of the estimation of panel data quantile regression analysis will be identified. Therefore, I firstly applied the issue to an empirical investigation on the relationship between economic growth and government size in OECD countries as the first part of the dissertation. We find that the effect magnitude of government size to economic growth varies through quantiles. When the economic growth is low, increasing its government size may have positive effect and stimulate economic growth. Secondly, I will use the Manpower Survey data in Taiwan to examine the returns on schooling in Taiwan from 1991 to 2006 as the second part of my dissertation. It is found that wage inequality is associated with the returns to education in Taiwan. Besides, the marginal returns to education change with time much slighter in lower quantiles than higher quantiles and returns to education in higher quantile wage distribution group have an increase trend from 1991 to 2003 but the trend is inversed in lower quantiles after 2000. Finally, the original quantile regression methodology is applied into analyze the factors which affect the score of Basic Competence Test for Junior High School Students of Taiwan in 2005. It is found that the difference between urban and country affects the relative higher score group much more than other one. And the family factor has the vital effect on the educational achievement. Besides, increasing educational expenditure can improve the performance more in the relative lower achievement ones than the other ones.