Suppose we estimate the following model:
Passengersi = 1 + 2Populationi + ui
a) Generate a scatter plot with passengers on the vertical axis and population on the horizontal axis. Do you see a reason to be concerned about heteroscedasticity? Present your graph with your answer.
b) Can you come up with a story that could explain why variation of the error term might be increasing with population? (It should make at least some sense.)
c) Perform Goldfeld-Quandt test to see if there is heteroscedasticity of this speci c form. To do this, estimate the model using the 36 lowest and 36 highest populated airports separately, and use the sum of squared residuals from these regressions. Don't forget to sort the data by population rst! Make sure to state clearly your null and alternative hypotheses, test statistic, critical value, and decision.
Now suppose we estimate the following model instead:
Passengersi = 1 + 2Populationi + 3Incomei + ui
a) Test this model for heteroscedasticity using White's test. Make sure that your answer includes the following:
- Null and alternative hypotheses
- Test statistic (copy from the Stata output)
- Critical value (you may also use the p-value from Stata output)
- Decision (heteroscedasticity or no heteroscedasticity)
b) Estimate the model with heteroscedasticity-robust standard errors using White's method. Include your Stata output for this regression.