sets will require students to use the statistical software package, STATA, which is
... In addition, you should plan on purchasing Naked Statistics: Stripping.
EIB E213: Econometrics The Fletcher School, Tufts University Fall 2013 Location: C205, M-W, 13:30-15:10 Draft: Final Syllabus will be Distributed on the First Day of Class
Instructor’s Information: Jenny C. Aker, Assistant Professor of Development Economics, The Fletcher School Email:
[email protected] or
[email protected] Web Page: http://sites.tufts.edu/jennyaker/ Office Hours: M-W 3:30-5 or by appointment Office Address: Cabot 603C Teaching Assistant: Jing Cai Staff Assistant: Sheri Callender For any questions about concepts, assignments or data, please sign up for office hours or see me right after class. If you cannot make office hours during the pre-assigned time slot, please e-mail me with the header “Office Hours Meeting”. I will not be able to respond to individual e-mails with questions about readings, class concepts, assignments or tests.
Course Description This course provides an introduction to basic econometric methods. These are the tools of data analysis that economists and other social scientists use to estimate the size of economic and social relationships and to test hypotheses about them, using real-world data. The goal of this course is to equip students with the facts, intuition and skills necessary to critically read econometric research produced by others and to conduct independent econometric research. Coursework includes quizzes, problem sets, midterm and final examinations and a group econometrics research project. The problem sets will require students to use the statistical software package, STATA, which is available in the Mugar computer lab. Reasonably priced versions of the software are also available for students (and is recommended so that you do not need to plan your study groups around the Mugar computer lab hours). Pre-requisites: Introductory statistics (at the level of EIB B205) is required. Basic multivariable calculus (at the level of EIB E210m) and introductory economics are strongly recommended but not required. All relevant statistical concepts will be reviewed as they arise, but the reviews will be brief. The basic calculus concepts employed in the course pertain to derivatives and what they tell us about the shape of relationships between variables in simple graphs. To assess whether your background is adequate, review the appendices on probability and statistics in the Wooldridge textbook, with a particular focus on probability density functions (pdfs), a concept that will arise repeatedly in class. In addition, you should plan on purchasing Naked Statistics: Stripping the Dread from the Data (Charles Wheelan), which provides an intuitive and useful overview of statistical and econometric concepts. I will offer a (brief) statistics review session online, as well as two additional review sessions throughout the semester. Requirements and Grading: There will be five online quizzes, five problem sets, a midterm, an inclass final exam and an econometrics research policy brief. Class sessions will be primarily lecture1
based and will rely heavily on class notes. Students are also expected to prepare for class by completing the required readings before each class and actively participating in class discussion. Lecture slide handouts will be posted on Trunk the day of class. Grades will be based upon the following breakdown: Problem sets: Quizzes: Midterm exam: Final exam: Policy brief:
15% 5% 35% 40% 10%
The problem sets can be completed in groups of no more than five (5) people and no fewer than three people. The midterm and final examinations will be closed book exams, but students will be allowed to prepare and use one 3X5 index card of formulas and notes for each exam. Timed quizzes will be completed online and will be individual. The policy brief will be in groups of two students, and will require a written document and an in-class presentation. Textbooks The required textbook for this course is: Wooldridge, Jeffrey, Introductory Econometrics: A Modern Approach, Fourth edition, South-Western College Publishing. The fifth edition was recently released in 2013, but is not required for this course. W4 refers to the fourth edition of Wooldridge in the readings. Reading an econometrics text is never easy, but it is essential that you make the effort to read this book. In the past, students have primarily relied upon the notes and not used the book (and many students will probably tell you not to purchase the book). This is your opportunity to learn the “written language” of econometrics. As with learning any language, reading goes very slowly at first, as you learn what the various symbols mean. If you make consistent effort, by the end of the semester you will find that your reading speed and comprehension have improved greatly. Developing such skills is of great value, because this course is only an introduction to econometrics. An additional required textbook is Charles Wheelan, Naked Statistics: Stripping the Dread from the Data, 2013, WW Norton and Company. Lecture Notes and Trunk. The lecture slides, problem sets, data, study questions and quizzes will be posted on the Trunk web pages for the course. Lecture slides will be posted the day of class. You should be enrolled automatically in the Trunk site shortly after you register for the course. A separate calendar with the specific date for each topic will also provided. Answer keys will not be provided for the study questions. Important or Unusual Dates Due to travel for fieldwork in Niger, there might be one class during the semester that will be cancelled (and rescheduled) in October or November, usually around the time of another school holiday (such as Columbus Day or Veterans’ Day). Students will be informed of this cancellation and re-scheduling approximately two weeks in advance.
2
Course Outline I.
Introduction: What is econometrics? What is it good for? Wheelan, Introduction, Chapters 1, 2, Conclusion
II.
The simple two-variable linear regression model A. Statistics review 1: Probability, random variables, expected values, variances W4, Appendix B, p. 714-737 Wheelan, Chapters 4, 5, 5½, 7 B. Ordinary Least Squares (OLS) as a method for fitting the model to data W4, Chapter 2, p. 22-39; Appendix A, p.695-702 Wheelan, Chapter 11 (omitting the Appendix) C. The R-squared goodness of fit measure W4, Chapter 2, p.40-41 D. Statistics Review 2: Estimators and Desirable statistical properties for estimators W4, Appendix C, p. 747-759 E. OLS estimators as random variables F. Classical assumptions under which OLS estimators have the desirable properties W4, Chapter 2, p.46-59
III.
Multiple (k) variable regression models A. Introduction to multiple regression analysis W4, Chapter 3, p.68-73 B. OLS and goodness of fit in the K-Variable Model W4, Chapter 3, p.73-83, Chapter 6, p. 199-205 C. The classical assumptions revisited W4, Chapter 3, p. 84-104, W3, Chapter 3, p.89-109 D. Functional transformations of dependent and independent variables W4, Appendix A, p.702-711 ; Chapter 2, p. 43-46 ; Chapter 6, p. 189-199 3
E. Dichotomous (binary) independent variables W4, Chapter 7, p. 225-246; W3, Chapter 7, p.231-252 (W2, Chapter 7, p.218-240) F. Units of Measurement W4, Chapter 2, p. 41-43; Chapter 6, p. 184-189 G. Model specification IV.
Interval estimation and hypothesis testing A. Statistics review 3: Common families of statistical distributions W4, Appendix B, p. 737-744 Wheelan, Chapter 11 (Appendix) B. OLS under the normality assumption W4, Chapter 4, 117-120 Wheelan, Chapter 8 C. Confidence intervals and interval estimation W4, Chapter 4, p. 138-140; Appendix C, p. 762-769 Wheelan, Chapter 10 D. Testing hypotheses about a single parameter: the t test and statistical significance W4, Chapter 4, p. 120-138; Appendix C, p. 770-782 Wheelan, Chapter 9 E. The distinction between statistical significance and economic importance W4, Chapter 4, p. 135-138; Appendix C, p. 780-781 F. Testing hypotheses involving several parameters: the F test W4, Chapter 4, p. 140-154 G. Presentation of regression results W4, Chapter 4, p. 154-156
V. Models for binary dependent variables A. OLS when the dependent variable is dichotomous W4, Chapter 7, p. 246-251 4
B. Statistics review: maximum likelihood as an approach to creating estimators W4, Appendix C, p. 760-762 C. Probit and logit regression models for dummy dependent variables W4, Chapter 17, p. 574-580 D. Interpreting coefficients in probit and logit models W4, Chapter 17, p. 580-587 E. Interval estimation and hypothesis testing in probit and logit models VI. Omitted Variable Bias A. Nature of the problem and description of its consequences W4, Chapter 3, p. 89-94 Wheelan, Chapter 12 B. Omitted variable bias in program evaluation W4, Chapter 7, p. 251-254; W3, Chapter 7, p.258-260 (W2, Chapter 7, p.246-248?) C. Dealing with OVB: Considering whether the bias “works in your favor” D. Dealing with Omitted Variable Bias: Introducing proxy measures W4, Chapter 9, p. 306-313 E. Dealing with OVB: Using true or natural experiments W4, Chapter 13, p. 444-455 F. Dealing with OVB: Using fixed effects methods in panel and pseudo-panel data to eliminate bias W4, Chapter 13, p. 455-470 ; Chapter 14, p. 481-489 G. Dealing with OVB: Using instrumental variables techniques W4, Chapter 15, p. 506-525 VII. Other Problems with the Dependent and Independent Variables A. Including irrelevant variables 5
W4, Chapter 3, p. 89 B. Measurement error in the dependent and independent variable W4, Chapter 9, p. 315-322 C. Multicollinearity and other data weaknesses W4, Chapter 3, p. 94-99 VIII. The Problem of Heteroskedasticity W4, Chapter 8 IX. Summing Up
6
