关键词 > ECMT6007/6702
ECMT6007/6702: Econometric Applications Problem Set 11
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 11
Semester 2 2022
Question 1. Computer Exercise: The Effects of Unions on Wages
The data for this exercise are taken from a study of the effects of trade unions on labour market
outcomes. This exercise is based on a panel survey which follows a random sample of male workers. The focus is on the effect of unions on a worker’s ln (wage), which is the natural log of the real hourly wage, and the key explanatory variable is union status (an indicator of whether the worker is a union member). The cross-sectional unit is an individual worker (each with a unique value for the identifier id), and the sample period covers the years 1990 to 1997 (with time identifier year).
(i) Download the dataset wagepan10 .dta and report the pooled sample mean, min and max
values for each variable. How many individual workers are included in the dataset? (ii) Estimate the model:
ln (wageit) = β0 + β1 unionit + β2 educit + β3 experit + β4 experit(2)
+ β5 manufit + β6 marriedit + β7 rurit + uit (1)
using OLS on the pooled sample. Present the results in a table.
(iii) What is the interpretation of β1 ? Is this likely to represent a causal effect? Explain. (iv) Estimate the model:
ln (wageit) = β0 + β1 unionit + β2 educit + β3 experit + β4 experit(2)
+ β5 manufit + β6 marriedit + β7 rurit + ai + uit (2)
using the random effects estimator. Add the results to the table. How does the estimate for β1 differ to that for the pooled OLS in (ii)? What, if any, are the advantages of the random effects (RE) estimator over the pooled OLS estimator?
(v) Now re-estimate (2) using the fixed effects (FE) estimator, and add the results to the table.
(vi) What is the FE estimate for β1 ? Is it similar to the RE estimate? Explain with reference to the reported 
for the RE estimation.
(vii) Re-estimate model (2) using the first difference estimator. Add the results to the table. How
do the coefficient estimates compare to those for the FE estimator? Briefly discuss.
(viii) To allow for a time trend in wages, re-estimate model (2) with the inclusion of year as an
additional explanatory variable using the FE estimator. What is the estimated coefficient on year? Explain this result for the estimator.
(ix) Name two possible characteristics of a worker that are captured by ai . Do you expect these would be correlated with unionit?
(x) What effect does union membership have on the expected wage? What is the best estimate of the effect, and why?
[Hint: In STATA look in the help guide for information on using the commands: xtsum and xtreg]
Note: The wagepan10 .dta dataset can be downloaded from the course Canvas website. The dataset contains 888 observations and 10 variables. The variables correspond to:
• id: person identifier (the cross-sectional unit)
• year: values of 1990 to 1997 (the time period identifier)
• lnwageit : ln(real hourly wage) of worker i in year t
• unionit : = 1 if worker i is a union member in year t (= 0 otherwise)
• educit : years of completed education of worker i in year t
• experit : years of labour market experience of worker i in year t
• exper2it : = exper2
• manufit : = 1 if worker i is employed in manufacturing in year t (= 0 otherwise)
• marriedit : = 1 if worker i is married in year t (= 0 if single)
• rurit : = 1 if worker i is located in a rural area in year t
