This preview shows pages 1–8. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full DocumentThis preview has intentionally blurred sections. Sign up to view the full version.
View Full Document
Unformatted text preview: Olivier Deschenes, UCSB, Econ 140B, Winter 2011 Lecture 4: Regression Specification Dummy / indicator variables Interactions with dummy variables Interactions with “continuous” variables Functional form: quadratics and log models S&W Ch.8 Olivier Deschenes, UCSB, Econ 140B, Winter 2011 Regression with dummy variables Suppose we are interested in calculating the percent difference in earnings between males and females Y i = log weekly earnings of person i D 1i = 1(if person i is female) Y i = 0 + 1 D 1i + u i 0 = Average log weekly earnings of males 1 = Difference in average log weekly earnings between fem ales and males ( percent difference in weekly wages) Why: E[Y i D 1i =0] = 0 , E[Y i D i1 =1] = 0 + 1 , so 1 = E[Y i D i1 =1]  E[Y i D i1 =0] Olivier Deschenes, UCSB, Econ 140B, Winter 2011 STATA application (CPS) . regress lwkearn female, robust; Linear regression Number of obs = 8454 F( 1, 8452) = 487.02 Prob > F = 0.0000 Rsquared = 0.0542 Root MSE = .54535  Robust lwkearn  Coef. Std. Err. t P>t [95% Conf. Interval]+ female  .2613968 .0118448 22.07 0.000 .2846155 .2381782 _cons  6.799713 .008433 806.32 0.000 6.783182 6.816244 Olivier Deschenes, UCSB, Econ 140B, Winter 2011 Current Population Survey (CPS) Monthly survey of about 60,000 households Administered by U.S. Census Bureau for the Bureau of Labor Statistics (BLS) Sample represents the civilian noninstitutional U.S. population The survey asks about the employment status of each member of the household 15 years of age or older in the reference week ⇒ Used to construct official unemployment rate series Also ask about demographics, education, wages and income (some months), industry, etc Olivier Deschenes, UCSB, Econ 140B, Winter 2011 Adding more dummy variables to the model D 2i = 1(if person i has a college degree) Y i = 0 + 1 D 1i + 2 D 2i + u i 0 = Average log weekly earnings of males without a college degree 1 = Percent difference in average weekly earnings between females and males, for those without a college degree 2 = Percent difference in average weekly earnings between college graduates and noncollege graduates, irrespective of gender Olivier Deschenes, UCSB, Econ 140B, Winter 2011 In regression algebra: 00 = E[YD 1 =0, D 2 =0] = 0 [omit i subscript] 10 = E[YD 1 =1, D 2 =0] = 0 + 1 01 = E[YD 1 =0, D 2 =1] = 0 + 2 11 = E[YD 1 =1, D 2 =1] = 0 + 1 + 2 So: 0 = 00 1 = 10  00 2 = 01  00 = 11  10 Olivier Deschenes, UCSB, Econ 140B, Winter 2011 STATA application...
View
Full
Document
This note was uploaded on 09/04/2011 for the course ECON 140b taught by Professor Staff during the Winter '08 term at UCSB.
 Winter '08
 Staff

Click to edit the document details