What is the Effect of Trade on Economic Growth in Namibia?

Outline
• Ceteris paribus
• Randomized trial
• Cater, Greenberg, and Walker (2017)
• How to choose a research topic
Econometrics
• Economists use data to answer cause-and-effect questions.
“We approach these questions coolly, however, armed not with passion but with data.”
• Econometrics (or ’metrics) is disciplined data analysis paired with the machinery of statistical inference.
Ceteris Paribus
• Causal interpretations require other things equal (ceteris paribus) comparison.
• Example: Does indebtedness cause college students to drop out?
• Context: The proportion of American college students completing their degrees in a timely fashion has sharply dropped.
One-Minute Question
• Suppose that a researcher compares graduate rates between those with and without student loans.
• Would the result show the causal relationship between the amount of loans and graduate rates?
Answer to the Question
• No, because the comparison is not a ceteris paribus one.
• To investigate the causal relationship, one needs to compare ‘virtual twins’ who are identical in every way except that one student gets loans and the other paid tuition from savings.
Selection Bias
• Selection bias (or omitted variable bias) is the biggest obstacle of having the ceteris paribus condition.
• Example: Poorer students tend to have more student loans.
• Omitted variable: income level.
Strategies to Overcome Selection Bias
• Goal: To achieve ceteris paribus.
• Make the variables of interest unrelated to many other factors determining the outcome we mean to study.
• Strategy 1: Randomized experiment
• Change circumstances randomly to engineer ceteris paribus.
• Drawback: Expensive.
• More strategies:
• Regression, instrumental variables, regression discontinuity designs, differences-in-differencesCarter, Greenberg, and Walker (2017)
Randomized Trial Example
• Question: Does prohibiting computer devices in college classrooms boost student performances?
• Context
• Policy commitment to internet-enabled pedagogical reform in the 2016 National Education Technology Plan.
• Benefits and harms (multi-tasking and distraction in class) of connectivity in and out of classrooms
One-Minute Question
• How would you design a randomized experiment to claim the causal relationship between the computer use in classrooms and academic performance?
Methodology
Randomized trial at the US Military Academy at West Point
• Treatment group 1: No restrictions for laptops and tablets use
• Treatment group 2: Only permitted use tablets, which should remain flat on the desk
• Control group: No access to electronic devices
Ceteris Paribus in CGW (2017)
• Many sections of an identical course (principles of economics course)
• 450 students per semester, up to 18 students per section.
• Identical syllabus, course requirements.
• Same professor teaches 2-4 sections.
• Students cannot request professors.
• Students cannot switch sections.
• Students are unware of the computer policy.
• Strong performance means the assignment of military job and location of first choice. (Incentive to work hard.)
• Identical access to laptops and tablets.
• Each student is required to complete introductory computer science before taking principles of economics.
• Wifi connection available every building.
Results
• Exam scores among students in classrooms that permitted laptops and tablets were 0.18 standard deviations below the exam scores of students in classrooms that prohibited computers.
Tips on Writing Research Proposal
Items in the Research Proposal
• Present your research question.
• Provide the data sources you hope to use.
• Convince professor and students why you should care about the topic.
Finding Topic
• It should be something that personally interests you.
• There needs to be a specific causal question that can be asked.
• The data required to answer the question need to actually exist.