David L W. answered • 06/17/19
PhD Tutor in Math, Statistics and Economics and Writing
The number of instrumental variables needs to be equal to or greater than the number of endogenous variables. Instrumental variables are variables that are correlated with the endogenous variable but not any other unobserved variable that affects the dependent variable. It's not easy to find good instrumental variables so if the independent variables are judged to be endogenous are more than the variables judged to be adequate instruments then one cannot completely address the bias caused by endogenous variables. However, it may be a little naive to expect to find uncontroversial, unbiased results from non-experimental data, as is common to economic data. Someone will always have an incentive to argue that an estimate is biased if the implication tends to be policy changes that benefit others more than them.
For example, If one wanted to find the impact of more education on log-wages then you'd need to find a variable that increases the amount of education someone gets but doesn't have anything to do with other qualities that affect a person's ability to command a higher average wage. But typically, people who can afford to get more education, also have other things that help them to earn more later on. This makes it harder to argue that simply helping a lot more people go to college, perhaps by making the first year or two free, would be a good investment for the economy. Those who teach at colleges or might get that free year of education would be in favor of it, those who already are educated or might be taxed more would be against it.
