When we return from Spring Break, we will begin discussing time-series.
To prepare for those discussions, please read Stock-Watson chaps. 12, 13 and 14
and please read Kennedy chaps. 10 and 19. And for a good set of notes on
serial correlation, take a look at
Sharyn O’Halloran's notes.
I particularly like the way she shows how
serial correlation may occur over space as well as time.
To prepare for our discussion, I have updated the
that I use to analyze the
To account for serial correlation over time,
the new script tests for a unit root in the residuals. And to account for serial correlation
over space, the new script also includes a "Spatial Autoregressive Model with Autoregressive
Enjoy the break! See you soon.
(posted: 13 April 2017)
The purpose of the course project is to help you "learn by doing."
In other words, you will learn more about econometrics by conducting
an econometric analysis than by reading the textbook.
If you all learned how to conduct an econometric analysis, that would be "good."
If you all learned how to assemble a dataset and conduct an econometric analysis,
that would be "perfect." But assembling a dataset is consuming too much of time,
so let's all focus on doing a "good" job (not a "perfect" job).
So a good place to start is with one of the datasets that I have already given you.
For example, you might visit
download some additional variables and append them to the
(Hint: At my econometrics
videos page, I have posted instructions on how to append data in
Or if you're interested in finance, you may wish to develop your knowledge of time-series.
For example, you might want to know how stock prices respond to interest rate changes.
If so, then you might download and analyze a dataset with such variables. Or you
might explore the time-series questions in the
Can you obtain an estimate of the Phillips Curve that does not suffer from serial correlation of
Or as an alternative to working with a conventional dataset, you may also conduct a
statistical analysis of human language. I started developing some notes on the subject,
but I never finished. So another good project would be to help me develop my
text mining notes.
Regardless of which dataset you choose, the most important thing to do is think about the
null hypotheses that you wish to test. For example, when working with the employment datasets,
we want to know what factors affect the employment rate, so one null hypothesis that you might
test is the null hypothesis that employment protection does not affect the employment rate.
Another that you might test is the null hypothesis that the minimum wage does not affect
the employment rate.
Assuming that you have rejected the null hypothesis that the minimum wage does not affect
the employment rate, your next step might be to predict the effect that an increase in the minimum
wage would have on the employment rate. So, for example, if the minimum wage rose 10 percent,
how much would the employment rate rise/fall? What is the standard error of your estimate?
Work through these types of questions, then -- when you begin writing your paper -- imagine that
you are an economist working for a government agency and your boss wants to know what actions they
could take to increase the employment rate.
Should they increase the minimum wage? Should they protect workers from dismissal? Should they
require all employment opportunities to be full-time, permanent jobs (as opposed to temporary and
contract work)? Should they encourage workers to join a union?
What actions should they take? How much will those actions increase the employment rate?
What is the standard error of your estimate?
(posted: 01 April 2017)
The purpose of the midterm exam
is to force each of you to read the textbook and develop a good set of notes
I am not looking for any particular page count. I just want to see that you
learned the material. What I want to see is that you developed your own notes.
What I do not want to see is whole paragraphs copied from the internet.
As you know, I have been too flexible about the submission date.
So let's set the deadline at the last class before spring break.
That's Thurs. 06 April or Sat. 08 April.
Interpreting Our Regression Models
(posted: 01 April 2017)
In our last class sessions, we discussed the
minimum wage and monopsonist problem.
The problem shows that when employers have monopsony power over their workforce,
increasing the minimum wage could increase employment.
To test this hypothesis, we need to add the minimum wage to the OECD data.
We also need to add the minimum wage to the dataset, so that our estimates
of the effect of employment protection on employment rates control for the
effect that the minimum wage has on employment rates.
But some countries -- Austria, Denmark, Finland, Germany, Iceland, Italy,
Norway, Sweden and Switzerland -- do not have a minimum wage.
But very few people are willing to work for zero, so these countries have
an effective minimum wage that is greater than zero.
Because real GDP per capita is highly correlated with the minimum wage
(among the countries that have a minimum wage), we can use real GDP per capita
to predict the effective minimum wage in the countries that do not have one.
"Due to the severe weather,
Queens College will be closed on Tuesday, March 14.
Please note that the college’s shuttle bus service will not be operating.
All events scheduled on campus have been canceled."
Enjoy the snow. Stay safe. See you on Thursday.
Gauss-Markov and Panel Data
(posted: 08 March 2017)
During our next class session, I would like to wrap up our
discussion of the Gauss-Markov assumptions. Then I would like
to discuss panel data and explore a pair of panel datasets.
For background on the Gauss-Markov assumptions, please read
Kennedy chaps. 5, 6 and 7. And for background on panel data,
please read Stock and Watson chap. 8 and Kennedy chap. 18.
Realfonzo and Tortorella Esposito (2014) used the OECD's EPL data in their analysis of
labor market liberalization and employment outcomes.
For those of you who do not speak Italian,
Google Translate is your friend. It does a good job of translating
the article into English with one exception: It translates the phrase "lavoro a termine"
as "completed work." The proper translation is: "temporary employment."
Finally, for an American perspective on employment questions, you may also want to read
Schmitt's (2013) discussion of the relationship between the
minimum wage and employment outcomes.
(posted: 08 March 2017)
I have posted the
midterm exam questions
and the final exam questions.
As discussed in class, these are "take-home" assignments,
so together we will negotiate the submission date and other details.
OLS Theory and Example
(posted: 24 Feb 2017)
In class, we have been discussing
The Tuesday/Thursday class has discussed problems #1 and #4.
The Saturday class has discussed problem #1.
On Saturday (25 Feb), I will discuss problem #4 and use it
as an introduction to hypothesis testing.
After problem #4, I would like to discuss an example of regression
and hypothesis testing, so on Saturday (25 Feb) and on Tuesday (28 Feb)
I would like to discuss Mankiw, Romer and Weil's
tests of the
Solow Model. During that lesson, I will also show you how to
explore the MRW dataset in
Then, in the classes that follow, I would like to discuss
maximum likelihood and hypothesis testing. Those lessons will make
use of problems #2 and #3.
For background on these topics, please read Stock-Watson chaps. 1-5
and Kennedy chaps. 1-4..
Please also read chaps. 1, 2 and 3 in the Stock and Watson textbook.
And please install
and wxMaxima on your computers.
BLS Employment Situation Report
(posted: 03 Feb 2017)
This morning, the US Bureau of Labor Statistics released its
monthly "employment situation report." The New York Times has a
We will work with two labor market datasets this semester, so
the article provides context for some of the topics that we will
discuss. Please take a look at it and please also read my
"Analysis of the 'Biagi Law'" and
Schmitt's paper on the minimum wage.
Welcome to Econometrics
(posted: 30 Jan 2017)
Welcome to the course website. This site helps me organize the course.
I hope you find it helpful.
I have posted a copy of the
Please review it and please look at the notes and course materials
that I have posted below. I have also prepared a
problem set for you.
We will discuss the solutions to these problems in class.
Please acquire a copy of the
Stock and Watson textbook and the
Kennedy textbook as soon as possible.
To begin the course, I would like to start with a review of
statistics and probability, so please read
chaps. 1, 2 and 3 in the Stock and Watson textbook
before our next class session. To refresh your memory of statistics,
you may want to look at my
You must do econometrics to learn econometrics. A textbook is
helpful, but not sufficient. To learn econometrics, you must
actively explore a dataset.
The best dataset to give you is the dataset that I am most
passionate about. I want to know how employment protections affect
the employment opportunities available to workers.
This is a personal issue for me, but it's also an issue that affects
you, your family and your friends, so together let's explore the data
and find out how labor law shapes the career opportunities available
So as a starting point for discussion and as an introduction to
econometrics, please read my
"Analysis of the 'Biagi Law.'"
It tests an important null hypothesis and the dataset that the paper
explores is one of the datasets that we will explore together this
Then to place the discussion in an American context, please read
John Schmitt's (2013) paper on why the
minimum wage has no discernible effect on employment.
Having identified an avenue of inquiry, we must place it in a
theoretical context to, so please read the
wikipedia article on the minimum wage
and please study my "minimum wage
and the monopsonist" problem.
With theory and data in hand, we will then test the null hypothesis
that the minimum wage does not have a significant effect on
To conduct a statistical analysis (like the one just described),
you will need to install a few (open source)
statistical and mathematical software programs on your computer.
is a great statistical software package for learning econometrics.
It has an simple interface and ships with sample datasets.
It's a great learning tool and I highly recommend it.
As you get better at data analysis, you will want a
more flexible tool. I use the
which (as the name suggests) is a programming language.
For mathematics, I use
which provides a graphical interface to
a computer algebra system that specializes in symbolic operations
(as opposed to numerical computing).
Over the course of the semester, I will provide datasets for you
to work with in Gretl. I will avoid asking you to write your own
computer code, but to push you in the right direction,
I will provide R scripts and wxMaxima notebooks for you to review
(and tinker with).
take your time
Finally, please remember that we will cover these topics
over the course of the semester. So please begin reading,
but please do not rush through the reading. Take the time
to understand what you are reading. And enjoy it because
econometrics is a lot of fun.
I'm looking forward to working with you this semester.