The first one is QJE of 199 1 year. Does Kruger's complicated school attendance rate affect school attendance and seriousness? . This paper studies an old problem: the influence of education on income. We know that direct OLS regression will have endogenous problems due to the lack of variables. This paper considers that because compulsory education in the United States is limited by age, people born in January can leave school before a certain age, but those born in June must stay in school until June, so people born in different months will end their education at different times, but the time when they start their education is the time when the school starts. Therefore, the month of birth will affect the length of education. Therefore, this paper uses whether you were born in the first quarter as an instrumental variable of education time, and finds that the results of instrumental variables are not significantly different from those of OLS. This instrumental variable was also questioned later, because it is a weak instrumental variable, and the birth month has little influence on the time of education. Even if there is only a weak correlation between the birth month and the missing variable, the estimated results will have considerable deviation.

The second is the lottery of living income and Vietnam era: evidence from the social security management records of Angrist on 1990 AER. This is also his doctoral thesis. This paper studies the influence of Vietnam War experience on subsequent income, which is obviously an endogenous problem. The author has noticed that when the Ministry of National Defense recruits, it draws a recruitment lottery number for men of school age, and then sets an upper limit. Those whose number is less than this upper limit are all within the scope of recruitment. So he defined whether the draft number is less than this upper limit as draft qualification. Obviously, people with conscription qualification of 1 are more likely to participate in the Vietnam War, and this number is randomly selected, so conscription qualification is a suitable tool variable. This is an example of clever use of natural experiments.

I know an interesting one recently. 1998 AER &； Labor supply of Evans' children and their parents: evidence from exogenous changes in family size. This paper studies the influence of the increase in the number of children on parents' participation in the labor market. The problem here is to find an exogenous shock that affects the number of children. Some studies in this field use twins, and some use abortion failure. Of course, the birth of twins is generally random and not designed as expected, so this is an exogenous influence that directly increases the number of children. This paper considers a factor that indirectly increases the number of children, namely the gender composition of the first two children. Why does this affect the number of children? Their theory is this: parents generally want their children's gender composition to be diversified, so if the first two children are both boys or girls, they want another child with a different gender. But if the first two children have a boy and a girl, and the task of gender diversity has been completed, there will be no more children, then families of the same sex will have more children. I think this idea is really amazing. In short, looking for tool variables is to use your brain and boldly associate.