ECON6300/7320/8300 Advanced Microeconometrics Linear regression basics
Christiern Rose 1University of Queensland
Practical 2
1/7
Introduction
This class will review:
Data management (refer to T2 Introduction to STATA.pdf)
Descriptive analysis
Linear regression
Prediction and analysis of residuals
Individual and joint hypothesis tests (Wald, t, F) Regression diagnostics
We begin with a demonstration from Microeconometrics using STATA Chapter 3 looking at whether private health insurance reduces medical expenditures
We move on to a practical looking at the gender gap in earnings of Australian clinicians
2/7
Demonstration – Introduction (1)
This follows Chapter 3 of the course textbook
We analyse data on medical expenditures of individuals aged 65+ who qualify for health care under the U.S. Medicare program
The data is from the Medical Expenditure Panel Survey
Medicare does not cover all medical expenditures.
Around 50% of individuals take out additional private cover
to ensure against out-of-pocket expenses.
3/7
Demonstration – Introduction (2)
Question: Is it worth it? By how much does private cover reduce medical expenditure?
Need to control for any factors which might determine medical expenditures and the propensity for individuals to take out private insurance
Apply multiple regression to estimate the treatment effect controlling for observable factors
4/7
Practical – Earnings decomposition (1)
We have earnings data for a sample of Australian GPs from the MABEL survey
It is well known that female doctors earn significantly less on average than male doctors. We use will use the Oaxaca-Blinder decomposition to decompose the earnings gap
5/7
Practical – Earnings decomposition (2)
The Oaxaca-Blinder decomposition is as follows. For two groups A, B let define the linear model:
Yg=Xg′βg+ug, E[ug]=0 Then it can be shown that
R = E[YA] − E[YB]
= (E[XA] − E[XB])′βB
+ E[XB]′(βA − βB)
+ (E[XA] − E[XB])′(βA − βB)
(Endowment effect) (Coefficient effect) (Interaction effect)
6/7
Practical – Earnings decomposition (3)
1. Load the data into STATA
2. Describe and summarise the data for the pooled sample,
males only and females only. What is the mean difference
in earnings between males and females? The median?
3. Determine a suitable dependent variable for a regression
which measures annual earnings
4. Regress your dependent variable on yhrs female expr
exprsq fellow pgradoth pracsize childu5 visa. Interpret the results. Is there evidence of heteroskedasticity? Is the model correctly specified? Are there any outliers?
5. Perform the regression separately for males and females. Interpret the results.
6. Perform a single regression in which males and females have heterogeneous coefficients. Test equality of the coefficients for males and females.
7. Use the oaxaca command to perform the Oaxaca-Blinder decomposition. You may need to install it first with the line ssc install oaxaca. See also the STATA journal article.
7/7
Interpret the results.