Econometrics B (ØkB)
Econometrics B covers a number of statistical models relevant for empirical economic analysis as well as mathematical methods for analyzing them. The course gives a detailed account of classical principles for estimation and inference for both linear and nonlinear parametric models for cross section or panel data. Emphasis is placed on understanding the theoretical foundations for these methods, as well as the practical tools to implement them in an imperative programming language such as MATLAB or Python.
The course will be developed along the following four axes (example topics included):
1) Linear models for panel data
- Estimation with strictly and weakly exogenous regressors
- Random effects, fixed effects, first differences
2) Estimation methods and numerical tools for nonlinear parametric models
- Generalized Method of Moments (GMM)
- Numerical optimization
3) Discrete-choice models and models for demand
- Binary and multinomial response models
- Demand estimation
4) Non-/semiparametric and machine-learning methods
- Classical nonparametric methods
- Regularized linear regression
The course will provide the student with a statistical toolbox that can be used for the estimation of a wide range of reduced form and structural micro-econometric models.
BSc Programme in Computer Science and Economics
- The central assumptions in the linear panel data model with unobserved effects; including how these assumptions are tested and how potential violation affect identification and estimation of parameters.
- The specification, identification and estimation issues that arise in static and dynamic binary response model for panel data models with and without unobserved effects (e.g. scale and level normalization, state dependence, initial conditions, endogeneity, neglected heterogeneity).
- The specification, identification and estimation issues that arise in multinomial discrete choice models and their implied predictions (substitution patterns).
- The principle of M-estimation in terms of estimation and inference, as well as key examples of M-estimators such as maximum likelihood and non-linear least squares.
- The properties of simulation-based estimators (consistency, asymptotic normality, rates of convergence and smoothness) and how they can be applied in the context of discrete choice models.
- The most common numerical optimizers and solution algorithms and how they work in an estimation context.
- The differences in goals, methods and settings between the machine learning methods and the traditional econometrics and statistics methods.
- Identifying the characteristic properties of a given economic cross-sectional or panel data set.
- Assessing the identification strategies in existing research papers as well as in their own analyses.
- Assessing which estimator is best suited to address the problem.
- Derivinging estimators of the statistical model’s parameters using the principles of M-estimation and estimate and interpret the parameters. This may involve deriving the sample objective function (such as the likelihood function), the gradient and the information matrix and implement the estimation method numerically using optimization methods.
- Programming the estimator and estimate the parameters of the model.
- Reporting and interpreting the results (marginal effects, elasticities, counterfactual simulations).
- Constructing misspecification tests, analyzing to what extent an econometric model is congruent with the data, formulate economic questions as hypotheses on the parameters of the statistical model and test these hypotheses.
- Using an imperative programing language such as Python or MATLAB to implement the mathematical representation of the econometric models and estimation methods covered in the course. This involves computing parameter estimates that are typically not available in closed form as well as computation of partial effects, elasticities, standard errors, test statistics, and simulation of counterfactual predictions based on the estimated model.
- Presenting a statistical model and empirical results clearly and consisely. This includes using statistic and econometric terms in the correct way, giving statistically sound and economically relevant interpretations of statistical results, and presenting results in a way such that they can be reproduced by others.
- Suggest and construct an appropriate econometric model.
- Develop arguments supporting an identification strategy.
- Select a suitable estimation approach.
- Test specification(s) and economic hypotheses pertaining to the model.
- Read and critically evaluate research papers containing applied econometric cross-section and panel-data analyses.
The acquired skills in micro econometric theory and practice provide a strong background that enables students to do empirical analyses at a level suitable for the bachelor thesis, but also relevant for answering empirical economic questions that could be encountered in a government agency or the private sector.
The course is a combination of lectures and exercise classes.
There will be 3x2 hours of lectures and 3x2 hours of exercise
classes per week for 7 weeks.
The lectures cover the theory and the intuition behind the estimators, methods and econometric models. The exercise classes allow students to put the theory into practice through exercises, and also to obtain hands-on coding experience by implementing the estimators on real datasets using an imperative programming language such as MATLAB and Python.
Students are expected to prepare the exercises before coming to the exercise classes.
Prerequisites equivalent to the courses PoP, MatIntro, Introductory Statistics and Probability Theory (GSS), Econometrics A (ØkA), and Numerical Methods (NuMe).
Self Service at KUnet
- 7,5 ECTS
- Type of assessment
Written assignmentWritten assignment, 48 hoursThe exam consists of two written parts:
1. Resubmission of a revised mandatory assignment. Which assignment that is to be resubmitted is chosen randomly by the course organiser.
2. The second part of the exam is a 48-hour take-home exam.
- All aids allowed
- Marking scale
- 7-point grading scale
- Censorship form
- External censorship
Criteria for exam assessment
See Learning Outcome.
Single subject courses (day)
- Class Instruction
- Project work
- Course number
- 7,5 ECTS
- Programme level
- Block 3
- No limit
The number of seats may be reduced in the late registration period
- Study Board of Mathematics and Computer Science
- Department of Computer Science
- Faculty of Science
- Jesper Riis-Vestergaard Sørensen (4-6d7579764368667271316e7831676e)
Are you BA- or KA-student?
Courseinformation of students