1
a) It is necessary to use IV to analyze MDVE because it eliminates selection bias. It allows us to capture the causal effect of treatment on the treated in spite of the nonrandom compliance decisions made by participants in experiments.
b) Compliers – Assigned to arrest treatment, and delivered the arrest treatment
Always Takers – Assigned to a coddle treatment, and delivered the coddle treatment
Never Takers – Assigned to coddle treatment, and delivered the arrest treatment
Defiers – Assigned to arrest treatment, and delivered coddle treatment
c) Always Takers = 177
Never Takers = 45
2
a) Si = a + B Ai +ei
b) The original authors hypothesize that college educated children encourage their parents to engage in healthier behaviors which is why they live longer. Dynaski does not believe this causal relationship is proven because the empirical test is inconclusive.
c) Zi would not be a valid instrument because it only shows a relationship between the differences of a child having one more year of education. I do not believe that comparing those who parents who have children with only a one year difference in education would be relevant.
d) First Stage: A = af + ØZ + ef
Reduced Form: A = af +ÞS + er
3
a) PC ownership could be correlated with u because a person’s income, need for a computer, or availability of a computer all come into effect when analyzing whether a person owns one or not
b) PC is likely related to parents’ annual income because a student with high income parents will more likely own a PC due to having the funds available to purchase one. I do believe this means parental income is a good instrument for PC because most students are not working and probably can’t afford a PC on their own, so must rely on their parents.
c)
d) First Stage:
Reduced Form:
e)
4
a)

For every additional year of education, one makes ~.17% more in wages. This does not seem right so it is not likely to capture the causal effect of education on earnings.

b)
i.

ii.

These results do not match the ones from part 4b(i).

iii.

These results match part 4b(ii) but not 4b(i). For every additional sibling, there is an additional 12.2% of education.
c)
i.

ii.

These results do not match the ones from part 4c(i). iii. These results are similar to 4c(ii), but not exact. They do not match 4c(i) at all. For each step in the order of birth, there is an additional 13.1% of education.
d) I believe that both the sibs and brthord instruments are valid because both the amount of siblings and what order they were born in would feasibly have an impact on education. Because the siblings instrument more closely matched the regression in step ii, I believe it is a better instrument. However, both instruments displayed close matches.