none

• Main Authors: Bo-jyun Wang, 王柏鈞
• Other Authors: Yu-hsia Chen
• Format: Others
• Language: zh-TW
• Published: 2009
• Online Access: http://ndltd.ncl.edu.tw/handle/41351087072464740419

碩士 === 東吳大學 === 經濟學系 === 97 === Using the data sets” Taiwan Social Change Survey ” of 1997 and 2002 , this thesis first uses the conventional MNL model(Multinomial Logit Model) and the general MNP model(Multinomial Probit Model) without IIA(Independence of Irrelevant Alternatives) restrictions to analyze the impact of different individual characteristics and family background on the incidence of over/undereducation and compares the empirical results of these two models. The empirical results show that the IIA restrictions are not rejected, the marginal effects of the explanatory variables of the two models on the incidence of over / undereducation and the average predicted probability of over / undereducation are very similar. As far as the marginal effect is concerned, both models show that the variable ” rapid expansion of universities ” has the highest marginal probability among all factors which have significant impact on overeducation by both models. Unlike the existing literature which uses only the sample of workers in studying, the incidence of overeducation, this thesis further takes into account the role of non-workers and compares the results estimated by both the Probit Model and Probit Model with Sample Selection. The empirical results indicate that the incidence of overeducation is significantly influenced by the individual’s choice to work or not. Therefore, the estimated cofficients will be inconsistent if the role of non-workers is ignored.