You have a problem with heteroskedasticity. Explain in which situations would you prefer WLS, FGLS, or OLS with White heteroskedastic consistent standard errors. Discuss the pros and cons of each method.

You are estimating the effect of savings rates when firms offer "matching" 401k savings rates, which is when firms offer to match the percent of savings of income of an employee for a 401k retirement account. Typically there is a maximum match rate offered. The policy evaluator wants to see if matching incentive programs increase retirement savings of an individual. The researcher only has information at the firm level and estimate the following equation:

savingsf = 0 + 1matchprogramf + 2sizef + 3agef + 4revf + 5profitf + kind + uf (1)

matchprogram is a binary variable, size is the number of employees, age is the age of the firm, rev is total firm revenue, profit is the profit rate of the firm, ind is the industry that the firm is in (a series of dummy variables).

(a) Given this information, what potential issue do you think the researcher may face? (hint: remem- ber this is an exercise on heteroskedasticity)

(b) What information could they use to resolve this issue? Explain. How would they utilize it in the model?