Question

6. The minimum legal drinking age in the United States is 21. Suppose we conduct a survey of students at URI. We collect data on their age (in months), the number of drinks they consumed in the last week, number of minutes of exercise in the last week, their body mass index (BMI), and other characteristics such as education, race, gender, class year, etc.

   1. Using this observational data, what would be an appropriate identification strategy if we are interested in estimating a causal effect of being legally allowed to drink (age >= 21) on: (1) number of drinks consumed in the last 7 days, (2) minutes of exercise in the last 7 days, and (3) BMI. Explain, in general, how the identification strategy works, and its assumptions. (10 points)

   2. Suppose we are interested in the effect of being legally allowed to drink (i.e. age >=21)on an individual’s BMI. Would you expect to find an effect of being legally allowed to drink on weight gain (measured by BMI) for individuals “very close” to age 21? Explain.

Answer 1

1. Here we wish to predict the dependent variable that a person is legally allowed to drink or not. Now number of drinks consumed in the last 7 days is a good indicator of this. If a person has a heavy amount of alcohol intake in the last 7 days, then most probably he is >=21 years of age (as people with age <21 would not have that easy access to drinks as others).

Minutes of exercise is a moderate proxy to determining legal age.People usually start their exercises post the age of 20 and hence may be used to determine age.

Heavy drinkers usually tend to have larger BMI as compared to others which they try to make up through exercise. Hence this can be used to determine approximate legal age.

Thus a person with high alcohol intake, heavy exercise and high BMI usually has his age >-21.

Very close to age 21, a person usually drinks a lot. Hence his BMI is expected to be on the higher end(beer belly and all :P)