1) In this exercise you are going to investigate the determinants of wages in a panel data setting. The data for this exercise contains information on 545 young men in the United States covering the period 1980-1987. The dataset is an extract from the National Longitudinal Survey of the U.S.Department of Labor. The dataset consists of three different parts which contains the following variables: nr person identier year year lwage log of the wage rate educ education (years of schooling) exper work experience (years) union dummy = 1 if persion is in a union black dummy = 1 if person is black hisp dummy = 1 if person is hispanic married dummy = 1 if person is married Use the dataset to answer the following questions: a) As a rst step ignore the fact that the data are a panel and regress the logarithm of the wage on the dummy variable married, which is equal to one if a person is married in a particular year. What could explain your results? b) Now include the variables for experience {exper), membership in a union (union)1 the years of education (educ), and the dummy variables (black) and {hisp}. How does the inclusion of these variables in the regression change the marriage premium? What do you conclude from this? c) A key omitted variable from the regression in question 3 is the unobservable variable "talent". To capture the effect of this variable include "person xed effects" in the regression and estimate the regression with the help of the within transformation {you need the STATA commands mtreg for this. If you need help type help into the command line). What light do the results of this regression shed on the possible explanations for the observed correlation between marriage and wages? Why does STATA now drop the variables educ, black and hisp from the regression? d) In the lecture you have been told that estimating a regression with xed effects with the within transformation is completely equivalent to using the dummy variable regression. To check that this is true {and that it take; a lot longer to estimate the dummy variable regression) estimate the model using the dummy variable regression. Why are the results not completely identical with those that you obtained using the within transformation