Please read the attached article and answer the following questions:

1. Suppose the instrument variable in this article is valid. Take a look at Table 5, which shows estimates from the "first stage" regression. (Table 7 shows estimates from the joint IV estimate). That IV yields different estimates from ANOVA (regression) indicates endogeneity, more for some predictors than others. Which predictors are most sensitive to the inclusion of the instrument? Tell me in English (not economics) what that means about those predictors.

2. The findings that more intensive cardiac care improves mortality for Medicare patients is very stable (e.g. http://economics.mit.edu/files/10570). Researchers like to study the Medicare population because their claims are easier to get than commercial or Medicaid claims. How might the latter two populations differ from the Medicare population in terms of the relationship between spending and outcomes?

3. Bonus (hard!): who are the "compliers" in this paper? (Recall that IV estimates the local average treatment effect, i.e. the effect of treatment on the average complier).