Problem Set 3 Solved

Impact of restaurant smoking restrictions on smoking rate (55 points)

In this problem set, you are going to use SMOKE.DTA data. In this data set, you have information on individuals’ smoking behaviour and some other individual and locational characteristics for a random sample of single adults from the United States.


    1) (10 points) What is the share of people in the sample who smokes (you need to generate a new binary variable indicating whether the person is "smoking")? What is the share of people who resides in a state with restaurant smoking restrictions? What is the di erence in average smoking probability between states with restaurant smoking restrictions and states without restrictions?

    2) (10 points) Using the information available in the data set, estimate a linear probability model that examines the determinants of smoking probability (note that the dependent variable will be a binary variable). Include all potentially relevant variables that might a ect smoking behavior (you need to decide which variables might a ect smoking behavior). Interpret the signs and magnitudes of the coe cients that are signi cant at 10% signi cance level.

    3) (10 points) Estimating a new speci cation, check whether the e ect of age on smoking probability is quadratic or not (assuming that you included "age" in question 2 in linear form). At what age does the impact of age on smoking probability becomes negative?

    4) (5 points) Now, estimate your model including the income and cigarette prices in logarithmic form instead of level form (assuming that you included them in question 2 in level form). Interpret the signs and magnitudes of the estimated coe cients for these two variables.

    5) (10 points) You might think that the impact of restaurant smoking restrictions on smoking probability might be di erent for white and non-white individuals. Using an interaction variable approach, test whether this hypothesis is true at 10% signi cance level.

    6) (10 points) Can we consider the coe cient for restaurn as the causal e ect of restaurant smoking restrictions on individuals’ smoking probability? Discuss whether there might be an endogeneity problem here. Provide an example of omitted state speci c factor that could lead to a bias in the estimated e ect of smoking restrictions and discuss the direction of the potential bias that might arise because of this omitted factor.






2
Impact of Job Training Grant (25 points)

Use the data from JTRAIN.DTA for this exercise.


7) (10 points) Consider the simple regression model


log(scrap) =  0 +  1grant + u;

where scrap is the rm scrap rate and grant is a dummy variable indicating whether a rm received a job training grant. Can you think of some reasons why the unobserved factors in u might be correlated with grant? Provide example.

    8) (5 points) Estimate the simple regression model using the data for only 1988 (You should have 54 observations.) Does receiving a job training grant signi cantly lower a rm’s scrap rate?

    9) (10 points) Now, add an additional explanatory variable (to the model in question 8) indicating the log scrap rate of the company in year 1987 (lscrap1). Interpret the coe cient on grant. Is it statistically signi cant at the 10% signi cance level? How do you explain the change in the coe cient of grant between two models (from question 8 to 9)?



Marijuana usage and Wage (20 points)

Suppose you collect data from a survey on wages, education, experience, and gender. In addition, you ask for information about marijuana usage. The original question in the survey is: "On how many separate occasions last month did you smoke marijuana?"

    10) (5 points) Write an equation that would allow you to estimate the e ects of marijuana usage on wage, while controlling for other factors. You should be able to make statements such as, "Smoking marijuana ve more times per month is estimated to change wage by x%."

    11) (5 points) Write a model that would allow you to test whether marijuana usage has di erent e ects on wages for men and women. How would you test that there are no di erences in the e ects of marijuana usage on wage for men and women?

    12) (5 points) Suppose you think it is better to measure marijuana usage by putting people into one of four categories: nonuser, light user (1 to 5 times per month), moderate user (6 to 10 times per month), and



3

heavy user (more than 10 times per month). Now, write a model that allows you to estimate the e ects of marijuana usage on wage by using this categorical variable.

    13) (5 points) Discuss whether it is possible to estimate the causal e ect of marijuana usage on wage based on this survey data? What might be the problem in identifying the causal e ect here? Provide an example.























































4