LABOR ECONOMICS PROJECT
INSTRUCTIONS
Economics 5850 w Dr. Doetsch
Due December 3, 2018 at 2:20 pm
Late submissions are NOT accepted.
This assignment is intended to give you experience in
researching, analyzing, organizing, interpreting, evaluating,
and presenting realworld data and applying economic theory and
analytical techniques to realworld labor markets. It is highly
suggested that you work on it continually throughout the
semester as we cover relevant topics.
The work you produce must be your own. Substantial collaboration with other students constitutes academic misconduct. Sharing statistical results with other students constitutes academic misconduct.
You are to answer the following questions using an unweighted sample from the 2016 American Community Survey ACS1. These microdata are available for download in the Files section of this courses Carmen page.
The sample is a 11,000 sample n 28,285 of persons in the noninstitutionalized workingage population age 1864 residing in Ohio, Michigan, Pennsylvania, Indiana, Kentucky, or West Virginia. It is contained in files by the name of ProjectSample2019 on Carmen. Each observation is a real, live human being living in Ohio or a neighboring state in 2016 and each column is a different variable. Files are available in .csv comma separated value, .dta Stata, and .gdt Gretl formats on Carmen. You can import .csv files into R.
The codebook for this data is in the appendix to this
assignment. Another appendix describes using binary
categorical variables in your analyses. You should review and
understand each appendix before you begin.
1 For simplicity, the sample you are receiving is unweighted. This means that it is not representative. For the sake of this assignment, assume that it is. Your estimates will be biased, so take your results with a small grain of salt. Because your sample is unweighted, you will produce slightly different numbers than any official BLS reports.
Page 1 of 13
You should be familiar with running regressions and using
statistical analysis software from the prerequisites for this
course. If you are rusty with regressions or displaying data,
you may avail yourself of multiple online tutorials.
Tables are graded on presentation and clarity. Every table
should have an appropriate title. There should be labels on
rows and columns as appropriate. A good table attempts to
convey the most information with the least number of cells,
columns, and rows. Tables should be intuitive and easily read.
A reader should be able to understand what a table presents from
reading the table itself. Carefully think about and plan out
what your table should look like and how it should be
effectively organized before you create it.
Do not submit raw regression outputs; instead, make a nice table. For regression report tables, please include the relevant tstatistic in parentheses under the coefficient. Report every coefficient, every tstatistic, R2, adjustedR2, sample size, and any sample restrictions. If you create new variables, please define them. Regressions are graded on presentation, clarity, and application of theory.
Discussions should run the indicated word count. They should
reflect your grasp of labor economic theory and basic empirical
results. You should be able to judge which information in each
table or results from each analysis are interesting and worth
commenting on. Try to include as much important detail as you
can, given the small amount of space. Discussions will be
graded on relevance, insight, and application of theory.
The text of your report must be doublespaced in 12point
professional font with oneinch margins. Use a single staple in
the upper lefthand corner to attach multiple sheets of paper
together. Present your labeled answers to each section in
order, please.
Do not turn in a copy of these instructions. Do not repeat the
questions in the text of your answers.
If you create any new variables e.g. natural log
transformations, interactions, binaries from the ones you are
given, please specify the details on how you created them and
include them in any summary statistics.
Page 2 of 13
Please keep in mind that part of the learning process is the
frustration of figuring out how to do things yourself. Also
note that there are often no clear answers on how to best
approach the various little questions that come up during an
analysis. Do your best and explain what you do when such
situations arise.
Page 3 of 13
1 EXECUTIVE SUMMARY
1A Table: Summary Statistics
Produce a table of means, standard deviations, minimums, and
maximums for every variable that you use in the foregoing
analysis. You may wish to create this table last.
1B Table: Labor Market Statistics
Produce a table of estimates of salient aggregate labor market
statistics mean earnings, labor force participation rates,
unemployment rates, employmentpopulation ratios by state.
1C Discussion: Executive Overview Summary
Consider your results from 1B. Compare and contrast the
aggregate labor market of Ohio with that of its neighboring
states. How do conditions in the Ohio labor market compare?
How do these states compare to the country as a whole recall
this is 2016 data? Do you believe these numbers are
satisfactory to make crossstate comparisons of labor market
conditions? Explain. 300 words
Page 4 of 13
2 LABOR FORCE PARTICIPATION
2A Table: Labor Force Participation
Using the data provided, produce one table of statistics
presenting the following:
a Labor force participation rate by sex
b Labor force participation rate by sex and marital status
c Labor force participation rate by sex and educational
attainment dropout, high school, some college, college,
postgrad
2B Regression: Labor Force Participation
Use an OLS linear probability model to model the effects of
correlates on labor force participation and how they differ by
sex. Run three regressions: one including both sexes, one for
men, and one for women.
Dependent variable: lfpart
Independent variables: Use your judgment, guided by theory Sample restrictions: None
2C Discussion: Labor Force Participation
Consider your results from 2A and 2B. Discuss the differences
in the labor force participation rate of workingage people of
different demographic groups and different individual
characteristics for this sample. Does economic theory
adequately explaining these differences in labor force
participation? Did you expect to find what you did here? Is
anything noteworthy? Interpret any interesting coefficients.
Are there any important omitted variables that may introduce
bias into your estimates? Discuss any weaknesses in your
analysis. Explain. 300 words
Page 5 of 13
3 HUMAN CAPITAL EARNINGS
3A Table: Wage and Salary Income by Educational Attainment
Using the provided data, produce one table presenting the
following statistics for the sample for each of the five
educational attainment categories dropout, high school, some
college, college, postgrad:
a Mean yearly wage and salary income of employed workers
b Mean weekly hours of employed workers
c Employmentpopulation ratio
d Unemployment rate
e Mean age
f Mean marriage rates
g Mean children
h Nativity status composition
i Sex composition
j Racial composition
3B Regression: The Determinants of Earnings
Use a Mincerian Wage equation OLS model to model the
correlates of yearly labor market earnings.
Dependent variable: earnings or lnearnings
Independent variables: Use your judgment, guided by theory Sample restrictions: Employed persons only
3C Calculation: Earnings Predictions
Use your coefficient estimates and values from your personal characteristics2 to predict your personal earnings at age 40 in 2106 with the 2016 coefficients. Provide a brief table explaining the values used and showing your work. Briefly discuss if you think your prediction is accurate. 50 words
3D Discussion: Earnings Determination
Consider your analyses from 3A, 3B, and 3C regarding the value
of human capital. Did you find what you expected to find? Is
anything noteworthy? Discuss your findings and any weaknesses
of your analysis. 300 words
2 You will have to use your best guess for what your personal characteristics by age 40 will be. Be realistic, though the data can guide you here. This, of course, assumes youll live in Ohio or a neighboring state.
Page 6 of 13
4 WORK HOURS
4A Table: Usual work hours per week
Produce one table of summary statistics means, standard
deviations of variables for employed persons, comparing usual
work hours per week by sex, race, and education level.
4B Regression: Determinants of Work Hours
Use a simple OLS model to model the correlates of usual weekly
hours.
Dependent variable: hours or lnhours
Independent variables: Use your judgment, guided by theory Sample restrictions: Employed persons only
4C Discussion: Work Hours
Consider your results from 4A and 4B. Do substitution or income
effects dominate? Do your results confirm what one expects from
economic theory? Did you find anything noteworthy? Explain your
reasoning and explain any weaknesses in your analysis. 300
words
Page 7 of 13
5 SEX DISCRIMINATION
5A Table: Sex Comparisons
Produce one table of summary statistics means, standard
deviations of variables for the sample comparing employed males
and employed females on the following variables:
a Wage and salary income by sex
b Educational attainment by sex
c Wage and salary income by sex and educational attainment
5B Oaxaca Decomposition by Hand: Men and Women
Estimate two Mincerian OLS wage equations restricted to males
and females, respectively.
Dependent variable: lnearnings
Independent variables: use your judgment, guided by theory Sample restrictions: Employed persons only
Using the Oaxaca Decomposition technique and your coefficient estimates, calculate by hand aided by Excela calculatorwhatever how much of the log malefemale earnings gap is explained and unexplained3. Report your findings. Show your work.
5C Discussion: Sex Discrimination
Consider your analyses from 5A and 5B regarding malefemale
differences in labor market outcomes and earnings. Interpret
your results. Discuss your findings and any weaknesses of these
analyses. What variables have you left out that are important?
How do you think their inclusion would affect your conclusions?
300 words
3 Use as many decimal places as you can in all your calculations for better numbers. It makes a very large difference.
Page 8 of 13
6 WAGE INEQUALITY
6A Table: Wage Structure Summary Statistics
Break up the sample of employed persons into four earnings quartiles. The first quartile is the lowest 25 of earners. The second quartile is the next 25 25th to 50th percentiles. The third quartile is the next 25 50th to 75th percentiles. The fourth quartile is the top 25 of earners above 75th percentile.
Build a table of summary statistics by quartile, showing inter
quartile differences in characteristics like earnings, age, sex,
race, education, hours, and any other variables you think
pertinent or interesting.
6B Discussion
Consider your table in 6A. Discuss what you find interesting.
Discuss your findings and any weaknesses in your analysis.
Explain. 300 words.
Page 9 of 13
7 CHOOSE YOUR OWN ADVENTURE
7A Table andor Regression
Produce an interesting question that may be answered from this
data. Creative, challenging, interesting, or rigorous questions
will earn comparatively more points. Use the data to answer
your question with a table andor regression. Explain what you
do and why.
7B Discussion
Discuss your answer to your question from 7A. Discuss your
findings and any weaknesses in your analysis. Explain. 300
600 words.
Page 10 of 13
APPENDIX
CODEBOOK
Variable Name
Type
Description
age
Continuous
Persons age in years
age2
Continuous
Age squared
asian
Binary
1 if person identifies as Asian
racial category
black
Binary
1 if person identifies as African
Americanblack racial category
children
Continuous
Number of own children in household
college
Binary
1 if educational attainment is a
Bachelors degree
dropout
Binary
1 if educational attainment is less
than high school or equivalent
earnings
Continuous
Yearly labor market wage and salary
earnings 2016
employed
Binary
1 if person is currently employed
female
Binary
1 if person is female
hispanic
Binary
1 if person identifies as Hispanic
ethnic category. Note: this is an
ethnic category; one can be Hispanic
of any race
hours
Continuous
The number of hours usually worked
per week
hsonly
Binary
1 if educational attainment is high
school graduation, GED, or equivalent
immigrant
Binary
1 if the individual was foreignborn
lfpart
Binary
1 if person participates in labor
force employed or unemployed
lnearnings
Continuous
The natural log of earnings
lnhours
Continuous
The natural log of hours
lnotherincome
Continuous
The natural log of other income
male
Binary
1 if person is male
married
Binary
1 if person is currently married
native
Binary
1 if person was born in USA or
territories
notinlf
Binary
1 if person is not in the labor
force
otherincome
Continuous
Includes all nonlabor income: income
earned and unearned from other
family members, interest, business
income, dividends, capital income,
rents, government transfers, etc.
2016
Page 11 of 13
otherrace
Binary
1 if the individual identifies as
American Indian, biracial,
multiracial, or other racial category
postgrad
Binary
1 if educational attainment is a
Masters degree, PhD, or professional
degree e.g. MD, JD
somecollege
Binary
1 if educational attainment is
Associates dergree, trade school, or
some college no completed degree
statefip
Categorical
A categorical variable indicating
state of residence. Codes are
available at this link.
unemployed
Binary
1 if the individual is unemployed
seeking work
vet
Binary
1 if the individual is a veteran of
US military service during wartime
or peacetime
white
Binary
1 if person identifies as white
racial category
yearsinus
Continuous
If a person is foreignborn, how many
years have they been in the US.
Caution: value is 0 if native born!
Page 12 of 13
APPENDIX
BINARY CATEGORICAL VARIABLES HELP
Mutually exclusive binary variables
Some of the binary variables above are mutually exclusive
because they fall into categories. You cannot include all
variables within each group in your regressions because of
perfect collinearity. Hence, you must choose a base group,
leave it out of your regressions, and interpret binary in
reference to that base group. Below is a list of all such
cases. Be careful: you generally want a large base group.
Stars indicate conventional base groups in economics.
Sex: female, male
Educational Attainment: dropout, hsonly, somecollege, college,
postgrad
Race: asian, black, otherrace, white
Employment Status: employed, unemployed, notinlf Labor Force Status: lfpart, notinlf
Birthplace: immigrant, native
Categorical variables
Statefip is categorical in nature. You cannot include it in a regression without transforming it first. There are numerous ways to do this. You could make a single binary variable for each category, e.g. an Ohio binary variable, an Indiana binary variable, etc. Even better, you can use your program of choice to create a dummy variable for each category aka factor variables. Be sure to choose your base category carefully as it affects your interpretation on each binary variable.
Here is how to do it in Stata: link
Here is how to do it in R: link
Here is how to do it in Gretl: link Note: First you must
declare a variable as discrete in Gretl link.
Page 13 of 13