Consider the following model for married women's hourly wages:

where i indexes the individual, lwage is the log of hourly wages, educ and experare years of education and years of experience, respectively (and expersq is the square). We are particularly interested in estimating the causal rate of return to education. The model is estimated using the Heckit method, and the output is reported below. Note that inlf is an indicator for being employed in the labour force, nwifeinc is the family income excluding the wife's labour market income, kidlt6 is an indicator for the presence of children less than 6 years of age.

a. Why do we need to use the Heckit method to estimate the model above?

b. Describe the Heckit method in this context. Write the conditional expectation equation for observed wages (you can assume joint normality of the error terms here). Describe the intuition for the extra term, and what are the necessary conditions to identify it?

c. Explain how you would construct the extra term in the conditional expectation equation and how you would use that to estimate the second stage. Point out these steps with reference to the reported results.

d. What is the estimated return to education based on the selection-corrected results above? Do you suspect this estimate to be very similar to the OLS estimate using the selected sample? Why or why not? Provide an explanation based on a statistical test (use =0.05 for the size of the test). Can you guess the direction of the bias when estimating model (1) on the selected sample?