The example given for the sharp RD case is legal drinking age and mortality rate. From what I heard, people under the age of 21 also can their hands on alcohol. In other words, being under the age of 21 reduces one's chance to consume alcohol but not diminish it 100%. Following this line of argument, it makes me wonder if this example can also be an example of fuzzy RDD.
For the fuzzy RDD example, if the numerator is already statistically insignificant, do we still need to proceed to get the denominator and the bootstrap to get the ATE? I thought the denominator (or results from the first stage) would always carry the same sign, so if the numerator is not different from 0, scaling it with a denominator of the same sign would move the distribution of the ATE to be entirely to the right or left of 0? Maybe I missed something here.
Reader asqued:
The example given for the sharp RD case is legal drinking age and mortality rate. From what I heard, people under the age of 21 also can their hands on alcohol. In other words, being under the age of 21 reduces one's chance to consume alcohol but not diminish it 100%. Following this line of argument, it makes me wonder if this example can also be an example of fuzzy RDD.
For the fuzzy RDD example, if the numerator is already statistically insignificant, do we still need to proceed to get the denominator and the bootstrap to get the ATE? I thought the denominator (or results from the first stage) would always carry the same sign, so if the numerator is not different from 0, scaling it with a denominator of the same sign would move the distribution of the ATE to be entirely to the right or left of 0? Maybe I missed something here.