The data file mexican contains data collected in 2001 from the transactions of 754 Mexican sex workers.

a. Using OLS, estimate the hedonic log-linear model with LNPRICE as the dependent variable and independent variables BAR, STREET, SCHOOL, AGE, RICH, ALCOHOL, ATTRACTIVE. Interpret the estimated coefficients.

b. Test for heteroskedasticity related to ATTRACTIVE using the \(N R^{2}\) test at the \(1 \%\) level of significance.

c. Estimate the model separately by OLS for observations with ATTRACTIVE \(=1\) and ATTRACTIVE \(=0\). Using the results, carry out the Goldfeld-Quandt test for heteroskedasticity across the two regressions. Use a two-tailed test at the 5\% level. Which regression has a larger estimated error variance?

d. Compare the estimates from the two estimations in (c). Do they appear similar or dissimilar? Which coefficients are noticeably different? Use OLS to estimate the model that includes the original variables and interactions between ATTRACTIVE and the other explanatory variables. Test the joint significance of ATTRACTIVE and the interaction variables at the \(1 \%\) level of significance. Is this a "valid" Chow test? Is homoskedasticity a necessary condition for this test? Recall that the test is described in Section 7.2.3.

e. Using the estimation results in (d), test for heteroskedasticity related to ATTRACTIVE using the \(N R^{2}\) test at the \(1 \%\) level of significance.

f. Use OLS with heteroskedasticity robust standard errors to estimate the model that includes the original variables and interactions between ATTRACTIVE and the other explanatory variables. Test the joint significance of ATTRACTIVE and the interaction variables at the \(1 \%\) level of significance. Is this a "valid" Chow test?