|
|
| Let us continue the schooling sample selection example. We want to measure the effect of an extra year of schooling on earnings, but we cannot because schooling is correlated with unobserved ability. |
| What we would like to do, is design an experiment that neutralizes the effect. This would require us to take a bunch of young people and at random give some of them, say, 10 years of schooling, and others 11 years. Given this randomized treatment of individuals, ability would now not be correlated with schooling, and would no longer matter on average. This is the idea behind clinical trials in medicine. |
| Of course, no parent would allow a bunch of pointy-headed academics to do this to their kids, so we cannot conduct this experiment. As a subsitute, we look for a natural experiment. A natural experiment is some accident about the way the world works that -- in this example -- has the effect of inducing some randomness in the years of schooling. |
| Researchers have indeed found a natural experiment for this problem. Children must stay in school in the US until they are 16. In many states, children must also start school in the year they turn 6, if they are born before 1 September, and in the year after they turn 6 if they are born after 1 September. This has the effect of requiring a different number of years of schooling according to whether a child was born before or after 1 September. Children born before 1 September will tend to have one more year of schooling than those born after 1 September. So splitting our data up into two groups -- those born before 1 September and those born after -- would create two groups with different years of schooling on average, but no difference in innate ability. |
| Natural experiments cannot always be found. But some of the best empirical economists succeed in the careers because they are better than others at finding natural experiments for problems in which they are interested. Of course, what we don't know is whether the best empirical economists are those with an extra year of schooling or those with more innate ability . . . . |