Final Application Flashcards
Instrumental variables in schooling:
Assume we have omitted ability from the regression
ln(wagei) = β0 + β1 Educi + εi
What must a good instrument for Educi satisfy? Use the 2 assumptions for instruments
Instrument relevance - must strongly correlate with Educ
Instrument exogeneity - must be uncorrelated with ε. ( so must be uncorrelated with ability!!!!!)
What IV did they choose then, and intuition
(Exact same in development module!)
Quarter of birth!
Children born first quarter can drop out earlier compared to those born in last quarter (they found they get 6 months less on average!)
Thus quarter of birth is correlated with education (years of schooling) , and not correlated with ability (2 assumptions satisfied!)
Slave trade cited cause of underdevelopment. So regress effect of slave exports.
ln(Yi)=β₀+β₁ln(Exports/Areai)+controls+εi
Where controls represents controlling for other factors such as latitude, endowments geographics etc
Issues with this concept (2)
Slave exports meassuremnt erros
Reverse causality - underdevelopment causes slave trade, or slave trade causes underdevelopment
Good instrument for ln(exports/area)
Outline relationship with assumptions
Distance to important destinations of exports.
Relevance: it is correlated Cov<0 , since when closer, exports increase.
Exogeneity: distance is uncorrelated with geographic characteristics that can impact development since are controlled for!
3rd case: faculty quality on student outcomes
(Since may not effect, since smart students go to the good unis, so may not be down to the teachers!)
What would be a good instrument for faculty quality
And how it works with assumptions
Expulsion of Jewish professors from Nazi Germany (Einstein etc)
Relevance: Cov<0 , as we expell the professors, faculty quality falls! Since they are smart people leaving!
Exogenous: Cov=0 since no anticipation effects
