LT 7 Regression Discontinuity Sam Marden s h
- Slides: 7
LT 7: Regression Discontinuity Sam Marden s. h. marden@lse. ac. uk
Basic Idea Explain in no more than three sentences why we think about regression discontinuities as natural experiments. Remember that throughout the first part of the term, we are looking for ways to construct a valid counterfactual. In the case of the example regarding college scholarships and the PSAT test that we discussed in lecture, what is the counterfactual we are trying to construct? Explain using either math or words, how the arbitrary eligibility rule of the National Merit Scholarship helps us construct this counterfactual.
Q 1 scatter emp age if year==1986 & region=="Quebec“, xline(29. 5)
Q 2 & 3 twoway (scatter emp age if year==1986 & region=="Quebec" & age>=25) /// (lfitci emp age if year==1986 & region=="Quebec" & age>=25 & age<30, fcolor(none)) /// (lfitci emp age if year==1986 & region=="Quebec" & age>=25 & age>=30, fcolor(none))
Q 4 reg emp treat age if year==1986 & region=="Quebec" & age>=25, r outreg 2 using "C: UsersmardenDesktopEC 455lt 6. d oc", word replace reg emp treat age 0 treat if year==1986 & region=="Quebec" & age>=25, r outreg 2 using "C: UsersmardenDesktopEC 455lt 6. d oc", word append (1) (2) VARIABLES emp treat -0. 0383*** -0. 0494*** (0. 0103) (0. 00727) age -0. 00804*** -0. 00217 (0. 00103) (0. 00226) age 0 treat -0. 00658** (0. 00249) Constant 0. 882*** 0. 723*** (0. 0284) (0. 0643) Observations 15 15 R-squared 0. 954 in parentheses 0. 964 Robust standard errors *** p<0. 01, ** p<0. 05, * p<0. 1
Q 5 VARIABLES treat age 0 treat Constant Observations R-squared (1) emp 0. 0403 (0. 0242) -0. 00777** (0. 00278) 0. 826*** (0. 0772) 15 0. 503 (2) emp 0. 00761 (0. 0149) 0. 00951*** (0. 00224) -0. 0194*** (0. 00361) 0. 360*** (0. 0603) 15 0. 755 Robust standard errors in parentheses *** p<0. 01, ** p<0. 05, * p<0. 1
Q 6 (1) (2) VARIABLES hours treat -1. 419*** -1. 782*** (0. 442) (0. 342) age -0. 310*** -0. 118 (0. 0562) (0. 0852) age 0 treat -0. 215* (0. 106) Constant 34. 50*** 29. 32*** (1. 540) (2. 414) Observations 15 15 R-squared 0. 921 in parentheses 0. 928 Robust standard errors *** p<0. 01, ** p<0. 05, * p<0. 1
