关键词 > ECMT6007/6702
ECMT6007/6702: Econometric Applications Problem Set 12
发布时间：2022-11-19
Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit
ECMT6007/6702: Econometric Applications
Problem Set 12
Semester 1 2022
Note: Solutions to this Problem Set must be submitted via the Course Canvas Dropbox. Please attach a copy of your computer output (i.e. your log file) to your typed answers.
Question 1. Analysing Drinking Behaviour
You will use the data in alcohol_consumption .dta to examine individual’s consumption of alcohol and the effects of alcohol consumption on health, productivity and earnings.
The categorical variable drinker indicates an individual’s drinking status, where 0 = abstainer (i.e. non-drinker), 1 = moderate drinker, and 2 = heavy drinker (i.e. more than 8 standard drinks on any one day during the week).
(i) Download the dataset from the course Canvas website and report the descriptive statistics for the sample. What fraction ofthe sample are abstainers, moderate drinkers and heavy drinkers?
(ii) Estimate the following choice model using the Ordered Logit estimator:
(
+ β4 edba + β5 married + β6 dchron) (1) and report the results (coefficient estimates, standard errors, 
2 , and llf) in a table.
(iii) Based on the coefficient estimates, what is the pattern of drinking status by age, education, and marital status?
(iv) Report the marginal effects of each variable on the probability of each of the 3 drinker cate- gories. In terms ofmagnitudes, what are the important factors associated with drinking status?
(v) Re-estimate the model with the inclusion of lninc = log(income) and lninc2 = lninc2 :
(
+ β4 edba + β5 married + β6 dchron + β7 lninc + β8 lninc2) (2)
and test whether income is an important determinant of drinking status by testing the joint significance of lninc and lninc2 using a 5% significance level. (That is, use the Likelihood Ratio Test to test the null hypothesis H0 : β7 = 0, β8 = 0.)
(vi) Based on model (2) test whether the quadratic term, lninc2, is individually significant, using
a two-sided alternative and a 5% significance level.
(vii) Estimate the model (1) using the Multinomial Logit estimator, and report the results (coefficient estimates, standard errors, 
2 , and llf) in the table.
(viii) Report the marginal effects of each variable on the probability of each of the 3 drinker cate- gories. Are there any differences to the marginal effects from the ordered logit model in (ii)? Discuss your answer.
(ix) Re-estimate model (2) using the Multinomial Logit estimator and test the joint significance of lninc and lninc2 using a 5% significance level. What do conclude about the relationship between drinking status and income?
(x) Estimate the model with a linear term in lninc:
P (drinker = j | x) = Λ(β0 + β1 age40s + β2 age50s + β3 ednchs
+ β4 edba + β5 married + β6 dchron + β7 lninc) (3)
using the Multinomial Logit estimator, and report the marginal effect (and standard error) of lninc on each of the outcome categories. Do the results suggest that alcohol consumption is a normal, or interior, good?
Note: The alcohol_consumption.dta dataset can be downloaded from the course Canvas website. The dataset has 858 observations and 8 variables. The sample is drawn from a survey of males aged 20-59 years who worked full-time. The definition of each variable is:
• drinker : = 0 ifthe individual is an abstainer, = 1 if a moderate drinker, = 2 if heavy drinker
• age40s: = 1 if aged 40–49 years (= 0 otherwise)
• age50s: = 1 if aged 50–59 years (= 0 otherwise)
• ednchs: = 1 if did not complete high school, (= 0 otherwise)
• edba: = 1 if complete bachelors degrees (or higher), (= 0 otherwise)
• married: = 1 if married (= 0 if single)
• dchron: = 1 if have a chronic (long-term) health condition (= 0 otherwise)
• income: annual income
