Suppose we know nothing about assignment mechanism: what can we say?
Suppose all we know is Yx∈[0,1] (w.l.o.g.: with bounds [Y_, ˉY] replace Y by Y−Y_ˉY−Y_)
Theorem: E[Y1−Y0]∈[{E[Y|X=1]P(X=1)−E[Y|X=0](1−P(X=1))}−P(X=1),{E[Y|X=1]P(X=1)−E[Y|X=0](1−P(X=1))}+(1−P(X=1))]
Corollary: Width of possible interval learnable from data is 1 (as opposed to 2 without data) and is [0,1] at largest, [−1,0] at smallest, so worst case interval always contains 0.
Proof: E[Y1−Y0]=E[Y1X+Y1(1−X)]−E[Y0X+Y0(1−X)] =E[Y1X]−E[Y0(1−X)]+E[Y1(1−X)]−E[Y0X]
Have E[Y1X]=E[Y|X=1]P(X=1), E[Y0(1−X)]=E[Y|X=0](1−P(X=1))
Largest possible effect when Y1=1 when X=0 and Y0=0 when X=1, so E[Y1(1−X)]−E[Y0X]=1−P(X=1)
Smallest possible effect when Y1=0 when X=0 and Y0=1 when X=1, so E[Y1(1−X)]−E[Y0X]=P(X=0)
Upper and lower intervals in corollary follow when E[Y|X=1]=1,E[Y|X=0]=0 vs E[Y|X=1]=0,E[Y|X=0]=1, respectively
Paper | Topic | Theoretical Motivation | Intervention | Outcome |
---|---|---|---|---|
Kremer and Glennerster (2011) | Demand | Law of demand | price of health goods | quantity ↓ |
Jensen and Miller (2008) | Demand | Giffen goods | coupons for grain by income | quantity ↑ |
Caunedo and Kala (2021) | Industrialization | Lewis dual sector models | Subsidize tractor rental | non-farm labor supply ↑ |
Crépon et al. (2013) | Labor Search | Matching function | job search assistance by city | congestion ↑ |
Breza, Kaur, and Shamdasani (2018) | Wage Rigidity | Bewley (1999) | unequal wages within teams | productivity ↓ |
Balboni et al. (2021) | Poverty trap | inflection point in returns | give people asset (cows) | inflection point found |