The following is an excerpt from "The Labor Market Effects of Rising Health Insurance Premiums," by Katherine Baicker, Amitabh Chandra. If workers in a certain sector of the economy or those who are married are systematically more likely to have different levels of unobservable characteristics that affect health insurance premiums, then such a correlation is possible. This problem is identical to the standard endogeneity problem in program evaluation, where receipt of the treatment is correlated with unobservable characteristics of the person receiving treatment. A solution to this problem is to instrument for imputed premiums using variables that are uncorrelated with &; and m; but are correlated with imputed health insurance premiums. In our analysis we use state-level, per-capita medical malpractice payments as an instrument for imputed premiums. In other words, in order for malpractice payments to be a valid instrument for health insurance premiums, it must be the case that medical malpractice payments _with health premiums. It must also be the case that malpractice payments_ _ with the unobservable characteristics of workers that might impact the outcome variable. X are not correlated; are not correlated are strongly correlated; are strongly correlated O are not correlated; are strongly correlated O are strongly correlated; are not correlated