Econ 488 – Applied Managerial Econometrics
Cameron Kaplan, Fall 2010
10/15/10
Lab 6 – Specification
This lab uses data from the General Social Survey (gss.gdt), which you can download from blackboard.
Background: Suppose you want to know whether happier people tend to make more income. In order to study this, you decide to look at the General Social Survey. You decide on the following specification:
(1) ln(rincdoli)= β0 + β1vhapi + β2phapi +β3femalei +β4marriedi +β5educi +β6agei +β7age2i +εi
Where:
rincdoli = the respondent’s income in dollars
vhapi = 1 if the respondent said they were “very happy” and 0 otherwise
phapi =1if the respondent said they were “pretty happy” and 0 otherwise
femalei = 1 if the respondent is female, and 0 if they are male
marriedi = 1 if the respondent is married, and 0 otherwise
educi = years of education completed
agei = respondent’s age in years
age2i = the square of age
Procedure: Many of these variables are not directly available in the data set, so you will have to create them.
i. Create the log of the income variable: In Gretl find the rincdol variable; click once on it to select it. Then go to: Add> Logs of Selected Variable. Notice that there is now a new variable called l_rincdol, which is the log of rincdol.
ii. Create dummy variables for happiness. In the dataset, the variable happy was coded as follows: 1= “very happy”, 2 = “pretty happy”, 3 = “not too happy.” Create a dummy variable for each category using the Add> Define New Variable function. For example create the variable vhap by using the numeric expression,
vhap = (happy=1)
Create dummy variables called phap, nothap, in a similar fashion for “pretty happy” and “not too happy.”
iii. In the GSS, the variable sex is coded as follows:
1 = “Male”
2 = “Female”
Create a dummy variable called female that equals 1 if the respondent is female and 0 if the respondent is male. Find the variable in the dataset that corresponds to gender, and use it to create the new variable.
iv. In the GSS, the variable marital is coded as follows:
1 = “Married”
2 = “Widowed”
3 = “Divorced”
4 = “Separated”
5 = “Never Married”
9 = “NA”
Create a dummy variable called married that equals 1 if the respondent is married, and 0 otherwise.
v. Create a variable that is the square of age. Click on the age variable and add the squares of the selected variable.
Write-Up:
- Run the regression in equation (1) and report your results.
(a) Do all of your coefficients have the expected sign? Which variables are statistically significant?
(b)Explain in words what your coefficient on vhap means.
(c)Explain in words what your coefficient on phap means.
(d)Why is nothap not included in the model?
(e)Holding other factors constant, what can you say about the incomes of “very happy” people compared to “pretty happy” people? What about compared to “not too happy” people?
(f)Write a few sentences describing the effect of age on income. Does income increase or decrease with age? Is this true for all ages? If not, in what ranges is income increasing with age, and what ranges is income decreasing with age? Do you think this makes sense?
- Suppose you believe that the effect of marriage on income differs for men and women.
(a)How would you modify equation (1) to test this theory? What additional variables do you need to create?
(b)Create the necessary variables and run the regression you specified in part (a). Report your results.
(c)Holding other effects constant, on average, how much more (or less) does a married man make than an unmarried woman in percentage terms?
(d)Based on your results, does the effect of marriage differ for men and women? Explain.