Use the following information to answer the next two questions: In their empiric
ID: 3220582 • Letter: U
Question
Use the following information to answer the next two questions:
In their empirical analysis of faculty salaries, Haignere, Lin, Eisenberg, and McCarthy (1996, p. 65) suggest that it is preferable to use "salary" rather than "ln(salary)" as the dependent variable in an earnings equation: "Our preference for Salary rather than Logged Salary as the dependent variable is based on simplicity and the logic of studying what people actually take home in their paychecks."
Is this an appropriate rationale for selecting between these functional forms? Why or why not?
Could a specification test help to resolve this issue? If so, which test(s)?
Explanation / Answer
The rationale is not approriate the reason being following:
when we run regression, we look for variables that are more normally distributed
When we use logarithm of income instead of income, the data becomes more normally distributed. Ths becomes a more effective dependent variable.
.
yes we can surely make use of specification test to help us decide which form of dependent variable is more effective in regression model
The test that can help us in this case is the "normality and distribution tests"
examples: Normality tests in scipy stats & normal_ad