Econometrics for Policy Analysis

Author

Jared Greathouse

Published

2024-11-04

1 Syllabus: PMAP 4041, Fall 2024

Note

Instructor: Jared Greathouse. Office Hours/Meetings: By Request. Location: Classroom South, Room 200. 2pm-3:15pm. See the source code for this page here.

Every day, governments pass laws/public policy to affect some outcome of interest. Policy usually touches thousands if not millions of people. From traffic-circles to pop/sugar sweetened beverge taxes, vaccine mandates and universal pre-k programs, cannabis legalization to minimum wages, public policy impacts us all from birth to death.

Policy is never self justifying. It demands evaluation. If California bans tobacco smoking in public, or if New York City implements gun control, presumably we would agree these likely impact outcomes like tobacco use or homicide rates, ideally decreasing both of them.

If California’s anti-tobacco policy didn’t affect smoking rates at all (or worse, if more people began to smoke) or if gun control has 0 impact on homicide rates (or increased them, paradoxically), then surely these could not be justified in the very first place. Before we continue, understand fundamentally these outcomes being affected are the point. The only reason that we, as a society, do policy is precisely because we think policy affects (or should affect) people somehow. If political science studies “who gets what where”, one summation of policy studies might be “what works?” But what policies should we care about? How can we know if they work? This is the starting point for empirical policy analysis. This class discusses the theory and process for how statistical analysis of data may be used to answer policy questions.

1.1 Course Philosophy and Structure

I believe the best way to demonstrate knowledge of policy analysis is through writing. As such, there will be no quizzes or in-class exams. Why? It is unrealistic. In real life, rarely do we have an hour and 30 minutes or a ten minute quiz window on icollege to write a full summation of our ideas or think through a question. Typically, we have much more time and resources to help us. In fact, proper use of resources, instead of memory, is what makes a good analyst. Good analysts do not need to remember everything, but they do need to be good at finding answers. The class is broken up into two sections: in the first section, we go over basic probability, correlation, and regression, as these are the tools you’ll work with to actually be able to write your paper. The remainder of the class covers other topics in research for policy analysis.

1.1.1 Paper Breakdown

In this spirit, I give to you one assignment. Specifically, you will write a paper where you will apply the statistical concepts we cover to answer questions about a real, existing policy. You will turn in each sub-section of the paper to me well before the first draft is due. I will give you feedback on each section, which I expect you to improve upon with subsequent iterations. So then, when it is time to turn in the full first draft, you will be able to incorporate my previous suggestions. I expect you to send me your section/paper (or at least ask to meet with me about it) as often as you need. To facilitate this, I have three potential datasets you may use.

  1. The first dataset deals with the impact of Texas’s abortion restrictions on the number of births, derived from here. The goal here is to see how Texas’ monthly births would look if SB8 were never passed.
  2. The second dataset deals with the impact of a tax on pop/soda on the total employment in San Francisco. The goal with this paper is to see how SF’s employment trends would look in the event that the SSB tax were never passed. The data are derived from this paper.
  3. The third dataset deals with the impact of terrorism on the GDP per Capita of the Basque Country, Spain from 1975-1997. The goal of this paper is to see how Basque GDP would look if the terrorism never happened.

Should there be another policy you wish to study, or another research question you have in mind, you may do this instead, subject to public data on that question existing and my permission. Here is an example paper of how yours should look, more or less. The text file versions of these three datasets are now online (as of August 29, 2024)

The paper consists of 5 sections: The intro, data, methods, results, and discussion sections. Each section is worth 4% each, meaning the sections themselves are worth 20%. The full first draft, then, is 20%. Here is what each section must look like:

  • Intro: Describes the general problem of interest. Should answer “Who did the policy/intervention, what policy/intervention was done, when did the policy happen, where did the policy happen, and how is the policy exepected to affect whatever outcome we care about?” I.e., if you are doing option 3, you will need to do some reading on why we would expect terrorism to affect a local economy. If you are doing option 2, you will need to read on why a tax on pop might affect local levels of employment. Also, say why having answers to how this policy affects things matters– why does the (wo)man on the street care about how abortion bans affect birth rates? Whatever the context, the goal of your paper is to see how some outcome would look if a policy of interest did not happen.

  • Data: This section describes the dataset you’re using. Say how many units of analysis there are (i.e., how many unique cities, countries, states, counties are in your dataset). Say the time period the data was collected from, if applicable. Say when the treatment began and for which state/entity. Say how each variable was measured (i.e., if the main variable is education, is it a 0 1 variable for college educated or not, or is it 1 … 16 to represent the years of education).

  • Methods: Describe your statistical model. For most people, this presumably will be difference-in-differences.

  • Results: Summarize your results. Did the policy work or not? If so, by how much? was the effect size small or large? How uncertain are we about the estimates?

  • Discussion/Conclusion: talk about what your results mean and why they matter. See this paper for an example of a more comprehensive conclusion.

1.1.2 Class Grade Breakdown

  • 40% of your grade comes from the best first version of the paper. The 5 sections themselves are 4% as I say above, and the full best first version is 20%.
  • 55% for the final paper (a revision of the above, not broken into sections), and
  • 5% for attendence.
  • I will drop the first submission if your final paper is an improvement.

1.2 Additional Details

  1. If I feel the concept is important, it’ll be in the lecture notes or we will discuss it. I will also sometimes assign external readings to be done before class.

  2. There is no required textbook (aside from this one!) for this course. Various free textbooks exist that go over the statistical material that we do such as Introductory Econometrics with R, Introductory Statistics, Intro to Modern Statistics, Regression and Other Stories, Intro to Econometrics, Intro to Political Science Research Methods, and many others. The Policy Department at Georgia State also recommends Introduction to Research Methods or Research Methods for the Social Sciences. The corresponding lecture will focus on the content that each respective chapter covers. Note that these books cover different aspects of the course in different levels of depth (Gelman’s book Regression and Other Stories is obviously mainly about regression, one of the last math topics we cover, whereas the others are more rudimentary).

  3. The same is true for software– I don’t care which of these you use, but the only ones I know well are Stata, Python, and (to a lesser degree) R. For this classroom, there’s already Stata on the computer, and I’ll provide code blocks frequently to illustrate concepts and ideas. For Stata users, Statalist is a great resource for Stata. R also is backed by a vast statistician community.

1.3 Helpful Notes from Me

  1. Sun Tzu said every battle is won before it is fought. To reverse the perspective, as Ben Franklin said, if you fail to prepare, prepare to fail. The fact that the paper is the primary assignment you have, in effect, means that I expect quality analyses written in a professional manner. I do not expect perfection, or anything at a level beyond what we cover, but preparation for the paper is your best friend in this course. Choosing your paper topic as early as possible, asking me for feedback on current draft iterations will help me, help you.

  2. As corollary to the preceding points, please do contact me if you have questions. Policy data analysis is what I do in my research every day. I love what I do, and I love discussing this topic with others. If you have any questions about the ideas we cover in class or have any difficulties, you may always meet with me or contact me otherwise. However, I can only help you if you reach out to me.

  3. With this said, do not simply communicate with me. Feel free to communicate with your classmates as well. This is something I only really learned the value of as a PHD student, so I figured I would advise the same to you.

1.4 Class Schedule

Below is the schedule. All readings for Econometrics for Policy Analysis (EPA) should be done before class. I will specify if anything else must be read before class.

1.4.1 Week 1

  • 08-26-2024 (Monday)

Introductions

  • 08-28-2024 (Wednesday)

Paper Discussion

1.4.2 Week 2

  • 09-02-2024 (Monday)

University holiday. No class.

  • 09-04-2024 (Wednesday)

Required: EPA, C2 and EPA C3.

Completely Optional: IS C2, IS C3 (skim), IDS C2, IDS C3, especially “Discrete Probability” and “Random Variables”.

Briefly covers the use of data for policy analysis. A refresher on averages. Also covers t-tests, standard errors, and confidence intervals.

  • 09-06-2024 (Friday) The introduction section for your paper is due.

1.4.3 Week 3

  • 09-09-2024 (Monday) - No Class, Jared’s Out of Town. Read the chapter on basic Asymptotic Theory (mainly the Law of Large Numbers and the Central Limit Theorem), EPA C4.

  • 09-11-2024 (Wednesday)

Stata Day.

1.4.4 Week 4

  • 09-16-2024 (Monday)

Correlation, Coeffcients, and Association (EPA, C6)

Here we cover basic correlation in 2 Dimensions, mainly using scatterplots.

  • 09-18-2024 (Wednesday)

Required: EPA C6, OLS Explained, estimation (here through (not including this section) here)

Optional: (ROS, C7), IS, C10.

1.4.5 Week 5

  • 09-23-2024 (Monday)

Inference for OLS. Gauss-Markov Assumptions of OLS.

  • 09-25-2024 (Wednesday) No class.

1.4.6 Week 6

  • 09-30-2024 (Monday)

Intro to Treatment Effects

  • 10-02-2024 (Wednesday)

Difference-in-Differences - Read the DID chapter as well as this link through (and including) section 9.2.1

1.4.7 Week 7

  • 10-07-2024 (Monday)

Review of DID and go over methods section.

  • 10-09-2024 (Wednesday) The data and methods sections for your paper is due.

Read: Presenting Results

1.4.8 Week 8

  • 10-14-2024 (Monday)

Midpoint (Last Day to Withdraw). Discuss Results section.

1.4.9 Week 9

  • 10-21-2024 (Monday) Ethics in Research: Replication (discuss replication paper, sections 1, 3, and 4)

  • 10-23-2024 (Wednesday) Ethics in Research: P-Values

Read this paper.

1.4.10 Week 10

  • 10-28-2024 (Monday) Data Measurement First submission due today. Must be in Word document or PDF (the file must have a .docx or .pdf extension, I will not read it otherwise). Watch this video.

  • 10-30-2024 (Wednesday) Data Measurement continued Read this chapter

1.4.11 Week 11

1.4.12 Week 12

  • 11-11-2024 (Monday) Review of Previous Material

  • 11-13-2024 (Wednesday) Review of Previous Material

1.4.13 Week 13

  • 11-18-2024 (Monday) Paper Week

  • 11-20-2024 (Wednesday) Paper Week

1.4.14 Week 14

  • 12-02-2024 (Monday) Special Topics- Synthetic Control Methods, Day 1

  • 12-04-2024 (Wednesday) Special Topics- Synthetic Control Methods, Day 2

1.4.15 Week 15

  • 12-09-2024 (Monday) Final Paper Submission Due
  • 12-11-2024 (Wednesday)

1.4.16 Week 16

  • 12-16-2024 (Monday)

1.5 Course Policy

1.5.1 Assignments

  • Assignments are due to icollege the day it’s listed. Late work is a 0 until it is turned in.

1.5.2 Attendance

  • Up to 2 unexcused absences are allowed.
  • To obtain approval for an excused absence from class (including religious holidays, family emergencies, or illness), a written notice needs to be sent to the instructor via email in advance.

1.5.3 Changes to Schedule

  • Syllabus and course schedule are subject to change.

1.5.4 Type of Course

  • This is an in-person course at Classroom South, Room 200.

1.5.5 Technology

  • This course will primarily use Stata, however I don’t care if you already know another software (e.g, R or Python, which are free) and wish to use that instead. In particular, Stata may be bought for 48 dollars for a 6 month license (but please talk to me first if your wish is to go to graduate school). Stata may also be accessed via GSU’s VLAB.
  • GSU computers are available in libraries and computer labs across campus (including in the AYSPS Building) for easy access.
  • Please reach out to me immediately if you have any challenges accessing or utilizing the technology for this course.
  • I will deduct as many points from an assignment as I feel you’ve used Chat GPT or other AI services to write the text of your paper. Of course, Chat GPT may certainly assist in the writing of statistical code, but it is YOUR responsibility to ensure the results are accurate. In the first place, you should rely on the notes, since code is provided in the notes for all major things you must do (in Stata, anyways).

1.5.6 Support Statements

1.5.6.1 Inclusivity Statement

We understand that students in our program come from a variety of backgrounds and perspectives. AYSPS is committed to providing a learning environment that respects diversity. To build this community, we ask all members to: - Share their unique experiences, values, and beliefs - Be open to the views of others - Honor the uniqueness of their colleagues - Appreciate the opportunity that we have to learn from each other in this community - Value each other’s opinions and communicate in a respectful manner - Keep confidential discussions that the community has of a personal (or professional) nature

1.5.6.2 Students with Disabilities

Students who wish to request accommodation for a disability may do so by registering with the GSU Access & Accommodations Center (AACE). Students may only be accommodated upon issuance of a signed Accommodation Plan by AAACE. Students are responsible for providing a copy of that plan to instructors of all classes in which accommodation is sought. To register for accommodations, please follow this link: GSU Access & Accommodations Center.

For more information, contact AACE located at Student Center East, Suite 205, 55 Gilmer Street, Atlanta, GA 30303.
Phone: 404-413-1560
Email: access@gsu.edu

1.5.6.3 Remote Academic Coaching

The Office of Disability Services also offers free remote academic coaching. To learn more about these services, go to GSU Disability Services or watch a Coaching Video.

1.5.6.4 Veterans & Serving Military

Georgia State honors its military and veteran men and women returning to pursue their education. Students who are veterans serving in the military and their dependents are encouraged to avail themselves of a full range of college services and activities through the Military Outreach Center (MOC).

For assistance or guidance while attending GSU on campus or online, contact the Atlanta Campus Military Student Advocate, Randy Barrone, at 404-413-2331. Please be sure and let me know ASAP if or when there is any possibility of you being activated and deployed. Thank you for your service!

For more information, contact the GSU Military Outreach Center
Phone: 404-413-2331
Email: rbarrone@gsu.edu
Website: GSU Military Outreach
Address: Dahlberg Hall, 30 Courtland Street, Suite 217, Atlanta, GA 30303

1.5.6.5 Online Course Evaluations – Student Surveys

Your constructive assessment of this course plays an indispensable role in shaping improvements of all courses within this program and your educational experiences at Georgia State. Please take the time to fill out the online course evaluations. We appreciate honest and open feedback.

1.5.6.6 GSU Policy Prohibiting Students from Posting Instructor-Generated Materials on External Sites

The selling, sharing, publishing, presenting, or distributing of instructor-prepared course lecture notes, videos, audio recordings, or any other instructor-produced materials from any course for any commercial purpose is strictly prohibited unless explicit written permission is granted in advance by the course instructor. This includes posting any materials on Chegg, Course Hero, OneClass, Stuvia, StuDocu, and other similar sites. Unauthorized sale or commercial distribution of such material is a violation of the instructor’s intellectual property and the privacy rights of students attending the class and is prohibited. The GSU Faculty Senate approved this policy on August 21, 2020.