Econometrics I

Course Description

Econometrics I at IGIDR aims to teach linear regression as the first tool in econometric analysis. This covers estimation, inference and prediction of correlations between a given economic variable and another variable (univariate regression), or a set of other variables (multivariate regression). Estimation includes both the theoretical closed-form approach as well as the maximum likelihood approach. In the theoretical approach, the topics will cover the assumptions involved, and the limitations they impose in the analysis of economic models. These limitations include the problems of multicollinearity and heteroskedasticity. An elementary knowledge of probability, statistics and matrix algebra is helpful.

The course is based on the book Econometric Modeling: A likelihood approach by David F. Hendry and Bent Nielsen. The material for the course is also covered in Econometric Methods by Jack Johnston and John DiNardo.

Course Modalities

• Susan Thomas will keep office hours to meet students on Tuesdays (2-4pm) for answering questions. Students will be able to interact with the TA to a greater extent.
• Course grading will be based on a set of quizzes (25%), class project (10%) and a final examination (65%). Homework will be given during the classes. A quiz will be held at the end of each class. The quiz will be based on all the material taught and the material that is given in the homework. A solved quiz that is turned in as homework before the next class, will be given additional credit in the final course grade. There will be no makeup quizzes.
• The class project involves authoring a page on a subject from the course contents on the knol website. The list of topics and the selected authors.
  Key guidelines for this project:
  • Here's a sample: an entry on the Probit model. Use the broad structure of this entry: pose the question, show the analytics of the solution, do a Monte Carlo simulation where you construct examples illustrating the functionality that has been achieved (these examples would work as a guide to doing these things in R), drive T to large numbers and verify there is consistency, talk about the immediate extensions, offer a set of further readings.
  • Roughly speaking, the target size should be five to ten pages.
  • Assume that the reader knows some probability and statistics but not econometrics.
  • Be careful to give original sources for every fact that you have picked up.
  • Be very careful about rules about fair use. I have software which will detect plagiarism, and you will get zero credit for an entry where even a paragraph has been plagiarised.
  • The entry will go up on knol under your name. If it is low quality, it will hurt your reputation. Be sure to do it well. Ensure the writing quality and English is of top quality. Do read Strunk and White, and look at well written encyclopaedia entries in the library.
  • The entries that satisfy my quality standards will get named and linked to from my web page.
  • All the knol entries that we are building have links to each other. Do harness these linkages.
  • The date for the first draft of your write up to be submitted is 29 September. Feedback on the draft will be discussed in the week of 8 to 12 October.
• This is intended to be a computationally intensive course. The software used in the course is called R. This will be introduced during course. All estimation during the course will be done in R, and the programs as well as the data will be made available.

Course content

Introducting the econometrics problem using the Bernoulli model

• Sample and population distributions.
• Distribution functions.
• Eg., the Bernoulli model.
• Inference in the Bernoulli model
• Asymptotic theory.

The regression model

• Continuous distributions
• The likelihood function, L.
• Estimation by maximising L.
• Inference

Univariate regression

• The econometric model
• Interpretation
• Correlation vs. Causation
• Matrix algebra of univariate regression
• Comparison of the maximum likelihood approach vs. the theoretical form.

Finite sample inference and simulation approaches

• Monte Carlo simulation
• Testing in cross-sectional regression
• Testing in time series regression

Multivariate regression

• The econometric model using three variables
• Matrix algebra of multivariate regression
• Interpretation
• Partial correlations
• Multiple correlations
• Properties of estimators
• Inference

Misspecification analysis

• Test for normality
• Test for identical distributions
• Test for functional form
• Test for exogeniety

Strong exogeniety

• The bivariate normal distribution and model
• Inference with exogenous variables

Issues in empirical modelling

• Aspects of empirical modelling
• Interpreting estimation results
• Problems that arise with time series data
• Misspecification in time series: test for independence
• Identification of structural models
• Under-identified/Over-identified structural models
• Structural breaks

Automatic model selection

• Model formulation and misspecification testing
• Removing irrelavant variables and keeping relevant variables
• Selection bias

Forecasting

• Forecasting from a regression
• Forecasting in changing environments

Course materials