Homework -
1) Given the following variable and assumed causal relations:
- Adult wages [label this Wage]
- Going to college (affects adult wages) [label College]
- Attending a private or public high school (affects wages and whether you go to college) [label Private]
- Family income (affects wages, college-going, and attending a private or public high school) label [label SES]
- Living near a private high school (affects wages and whether you attend a public or private high school, is correlated with family income) [label Location]
- Religious affiliation (affects whether you attend a public or private high school) [label Religion]
a) Draw a causal diagram incorporating all the above variables and effects.
b) Assume that the causal diagram as you drew it is true. What variables would you condition on (control for) if you wanted to identify the effect of Private on Wage using the "back-door method"? Don't include any unnecessary controls.
c) Assume that the causal diagram as you drew it is true. How would you go about identifying the effect of Private on Wage using an exclusion restriction / instrumental variable? What control variables would you include in the first stage, if any? What control variables would you include in the second stage, if any? Explain why this method works and what it's doing.
d) Assume that the causal diagram as you drew it is true. How would you go about identifying the effect of Private on Wage using the "front-door method"? Is this possible in this diagram? Don't include any unnecessary controls.
e) So far we've assumed that the causal diagram as you drew it is true. But thinking about the education system in the United States, what assumptions in the diagram are likely to be untrue? List at least one, although there are at least three big problems.
f) Go back to assuming the causal diagram as you drew it is true. What kind of effect will the exclusion restriction / instrumental variable method give you? In other words, you're estimating the average effect among what group?
g) You see a study that makes a good case that this exact causal diagram should hold in the particular region sampled for the author's data set. However, looking at the sample statistics, you note that this particular region is much wealthier than other regions. The author runs either a back-door or IV method to estimate the effect of Private on Wage, finding in both cases that going to a private school raises wages by 10%. The author then argues that this 10% estimate means that the government should enact a policy encouraging many more people to go to private school. Even assuming that the causal diagram does apply in the author's sample, what further assumption is the author's advocacy making about the effect of Private on Wages? Is that assumption likely to be true? What will happen if the policy is enacted but the assumption isn't true?