Module 4: Panel Data Methods Flashcards
Why would we want to pool cross-sections?
1) To increase sample size -> improved efficiency
2) Study changes over time
For the following regression, interprect the intercept.
kids = 4.92-0.36y82-0.52y84 -0.16educ
Assume the base year is 1980 and y82 is 1982.
A woman without any education is predicted to have roughly 5 children in 1980.
For the following regression, interpret y82.
kids = 4.92-0.36y82-0.52y84 -0.16educ
Assume the base year is 1980 and y82 is 1982.
Holding education fixed, a woman in 1982 is predicted to have roughly 0.36 less children than a woman in 1980.
For the following regression, interpret y85.
log(wage) = 0.98+0.37y85+0.065educ -0.3female
Assume the base year is 1978 and y85 year is 1985
Holding everything else fixed, someone in 1985 earns roughly 37% higher hourly wages than they would have in 1978. (exact = 45%)
For the following regression, interprect educ.
log(wage) = 0.98+0.37y85+0.065educ -0.3female
Assume the base year is 1978 and y85 year is 1985
Holding everything else fixed, one additional year of education results is predicted to increase hourly wages by 6.5%
Given the regression, what was the gender pay gap in 1978?
log(wage) = B0 + B1y85 +B2educ +B3female+B4y85fem
When the base year is 1978.
-B3
log(wage)m78 - log(wage)w78 = (B0+B2)-(B0-B2+B) = B3
Given the regression, what is the gender pay gap in 1985?
log(wage) = B0 + B1y85 +B2educ +B3female+B4y85fem
Given that 1978 is the base year.
-(B3+B4)
log(wagem85)-log(wagew85) = (B0+B1+B2)-(B0+B1+B2+B3+B4) = -(B3+B4)
Given the following regression, what is the gender pay gap in 1985?
log(wage)=0.998+0.31y85+0.066educ-0.37female+0.14y85fem
Given that the base year is 1978.
Women in 1985 are predicted to earn 23% less per hour than men, all else equal. (26% with the formula)
Given the following regression, what are the returns to education in 197
log(wage) = 1.12+0.047y85+0.056educ-0.37female+0.14y85fem+0.021y85educ
Given that 1978 is the base year.
Every additional year of schol is prediced to increase wages by 5.6% in 1978.
Given the regression, what is the returns to education in 1895?
log(wage) = 1.12+0.047y85+0.056educ-0.37female+0.14y85fem+0.021y85educ
Given that the base year is 1978.
Every additional year of school in 1985 is predicted to increase wages by 7.6%
(educ+y85educ) = 0.056+0.021 = 0.076
How has the returns to education changed over time. Given that returns in 1978 were 0.056 and returns in 1985 were 0.077?
It has increased by more than a third and the change is statistically significant at the 5% level.
change in returns = (0.021/0.056) significance = (0.021/0.01 > 2)
What is the distinction between pooled independent cross-sections and panel data?
Pooled cross-sections are random samples drawn from different periods and pooled together
Panel data are observations of the smae units over time.
What are the trade-offs of using pooled cross-secions vs panel data?
1) Panel data follows the same units over time, so its better at accounting for unobservables, but…
2) It’s hard to collect and plagued by atrition
3) Over time, panel data may become less representative of the whole population
Given the regression, what happened to student performance over time?
math4 = 12.25+14.86y98 + 5.47log(rexpp)
Given the base year is 1997, math is math scores, and rexpp is…
…expenditures per student.
Holding expenditures fixed, there was a 14.86 percentage point increase between 1997 and 1998 in students receiving a “satisfactory” score on their math score.
Write a theoretical model including the fixed effect.
yit = B0+B1Tt+B2xit+ai+uit
where i is the observation, t is the time period, and a for FE