Heteroscedasticity Problem Set 7 Refer to the attached “wage” data

Heteroscedasticity Problem Set 7

Refer to the attached “wage” data.  Data on 523 workers include Wage measured in dollars per hour, Education measured in years of schooling, and Experience measured in the number of years of work experience (as well as other variables that won’t be analyzed in this problem).

Note: Don't let the fancy names confuse you.  The problems of multicollinearity, heteroscedasticity, and autocorrelation are real, but solvable.  

For example, in the case of heteroscedasticity, all we have to do to solve the problem is (once detected) is to use the option vce (robust) when estimating a regression model, that's it.  For example, if we regress DV on IV₁ and IV₂, we can obtain the correct standard errors by using the simple modification of the regress command:

. regress DV IV₁ IV_2, vce (robut)

1. Obtain the residuals from the regression of Wage on Education and Experience.  Do you see evidence of potential heteroscedasticity?
2. Obtain the absolute values of the residuals as well as their squared values and plot each against education.  Is there evidence of heteroscedasticity?
3. Use the residuals to conduct the Park test.  What do you conclude?
4. Finally, use White’s heteroscedasticity corrected standard errors to solve the problem of heteroscedasticity.