Assignment task: Applied Econometrics
Term Project Guidelines
This term paper/project is meant to be your first overture into the kind of work professional economists and data scientists do. You will be asked to develop a valid economic research question that can only be answered using econometric analysis of empirical data. To that end, you will have to, on your own, collect your data, select appropriate econometric techniques and methods, develop and test hypotheses, interpret and discuss your findings, and provide appropriate recommendations.
Remember, however, that econometrics is means to an end, not an end in itself. The "end" is to ask and answer an empirical question that is of interest to economists and lay people alike. Econometrics allows you to get to the answer empirically, but it will not ask the question or theorize about issues - this is the task of you, the econometrician. Your first goal should be to come up with a valid question to answer empirically. A typical empirical question looks like: "What is the relationship between X and Y, ceteris paribus (all else being equal)?" This relationship can be due to association (correlation) or causation. Extracting causation is a much more impressive scientific feat as compared to finding a correlation: it's more impactful to know that X causes Y rather than X correlates with Y. But it's also much harder to find evidence of causation.
Keep in mind that variables X and Y can't just be anything: for instance, you wouldn't, as economics or business student, ask about the impact of micro RNAs on protein output (medicine) or the impact of child delinquency on intra-familial relationships (sociology) or the impact of climate change on world's marine ecosystems (ecology). X and Y must be economic or business variables and they can't be just any economic variables. It doesn't make sense, for instance, to ask about the impact of primary school enrollment on tourism revenues, or total revenues generated by arcades on total U.S. crude oil imports. Economic theory must inform the selection of X, Y, and all other variables in the regression. This is what distinguishes econometrics from basic statistics or data mining.
I want this project to give you a glimpse into the world of professional economics, be it in academia or industry or think tanks. Wherever an economist works, he or she writes papers, analyses, briefs, etc., using econometrics. This is the "secret weapon" of our profession that makes economists extremely attractive to employers of all types and kinds.
Keep in mind that this project is supposed to represent the culmination of your learning of econometrics in this course. You will be asked to apply your econometrics skills to answer an economic research question. You should aim to be thorough and to impress me(and potential employers) with your econometrics skills. Showcase what you know. Remember that many employers ask for writing sample as part of the job application process; if executed well, you probably want to use this project for that purpose.
Project Structure:
Econometrics empirical projects have a standard template: (1) Introduction/background, (2) data/methods, (3) results/recommendations.
Introduction and Background of the Issue
This section explains to the reader what the research question is and what the paper is about. It provides a brief background of the issue at hand, describes what is currently known in the literature, what is currently not known in the literature, what your paper does to fill the gap, how your paper fills the gap. What are you trying to explore in the paper? Why is it important/relevant? What do you expect to find or not find? What are your (briefly summarized) findings? Please address these as you write.
This section should also develop testable hypotheses. Hypotheses are testable predictions. They come from theory, not random guesswork. For example, theoretically (from ECON 216) we expect that quantity of money affects prices: more money means higher prices in an economy. So, here's a testable hypothesis: faster money growth is associated with faster inflation rate. So, here you'd state that you expect a positive and statistically significant coefficient on money growth (independent variable) on inflation (dependent variable), and why.
Data and Empirical Methodology:
The longest section. It describes your data, where you found them, and why you chose those particular variables. It also discusses your econometric approach: the model, procedures, and/or estimation methods you will use, and justifies why you use them. Take note that we cannot simply throw a kitchen sink of various methods into the paper in hopes you'll find a "good" result, or just pick something with the highest R-squared. This way of doing things is unprofessional, amateurish, and frowned upon in professional communities, and will ultimately earn you a low grade. You have to be justified in selecting your models and your reasoning has to be grounded in theory. For example, you know from theory that the growth rate of money supply is a direct determinant of inflation. Do NOT, then, estimate a regression equation that explains inflation without money growth on the right-hand side. The control variables also must make sense: you wouldn't control for the number of pets per capita in the equation for inflation. Reading related literature also helps inform you what your empirical model should look like.
In this section, be sure to describe your dependent and independent variables (that is, your data), including sources. Talk about the expected signs on your coefficients (how you expect an independent variable would affect the dependent variable); this is typically expected only for your main independent variables. For others, just explain the rationale why you chose to include them in the model (why you need to control for them). Show descriptive statistics table. They should show mean, standard deviation, maximum and minimum for each variable, and nothing else (not necessary, really). Show graphs. Write the regression equation using MS Word equation editor. Talk about your data sources. Talk about your methods and why you use them. Discuss model limitations-nobody's model is perfect, either due to lack of data or something else.
Results and Recommendations:
You now have your results. Discuss them. What is your basic finding? Is your main finding statistically significant? Is it economically significant? Is it robust to including more controls? Why does your result matter? How do we interpret it? Why should we care? What is the implication for the society/world at large? Any econometric problems that you encountered? How might they affect the results? How did you resolve the problems? What are your recommendations (policy or otherwise) that you would craft in light of your findings? What should be done in order to optimize a particular outcome of interest (e.g., stable prices, increased GDP growth, lower crime rate, higher education rate, etc. - whatever your dependent variable is).
Finish the project off with a brief conclusion. Summarize what you found and remind the reader why it is important. You can restate your purpose and your research question, and simply recap what has already been said. This conclusion helps answer the question: "So what?" Discuss briefly the potential limitations of your study and discuss possible extensions. Do not present any new results here. Read your introduction and conclusion right after the other. This helps you see whether they are consistent with each other.
(Now)
a. Dependent variable: poverty rates by country
b. Panel dataset: multiple countries observed over multiple years
c. Source of data: World Bank World Development Indicators
d. Explanatory (independent) variables: read on Google Scholar related articles on the determinants of poverty across the world. Get inspired to build your model and on the selection of independent variables.
Go research the topic through library resources/search access (e.g., Google Scholar) to find relevant papers on the topic. Think about how you would approach the topic in question building on the papers you have found.
1. Provide a paper title, 100- to 150-word abstract of your paper, and dataset (Excel or Stata file) to me via e-mail. It's OK if the dataset isn't 100% complete, but it should be about 50-75% complete, and certainly at least your dependent and main independent variables.