Detailed course syllabus 1. Linear regression model. Ordinary least squares method. This introductory class covers basic definitions of econometrics, econometric model, and economic data. Classification of econometric models (single- and multiequation models, linear and nonlinear models, static and dynamic models) and types of economic data (that is, time series, cross section, and panel data) are discussed. Primary and secondary data sources are described. OLS (Ordinary Least Squares) estimation method is presented and then employed using gretl software package. The procedure of designing and estimating a single equation linear regression model is discussed, as well as interpretation of its estimates (for the cases of continuous, trend, and binary variables). Other estimation methods (that is, ML Maximum Likelihood, MM Method of Moments, and IV Instrumental Variables) are briefly mentioned. Several single equation linear regression models are estimated and interpreted by students in class using provided data sets. Class structure: lecture (2 hours) + computer lab (3 hours). Basic: J. Wooldridge, Introductory. A Modern Approach, South- Western, 2009 [chapters: 1 3 and 10] Supplementary: C. Dougherty, Introduction to, Oxford University Press, 2007 [chapters 1 2] Problems 4 and 5; additional questions and problems are available at the publisher s webpage (www.cengage.com). 1
2. Validation and testing of the linear regression model. This class is focused on evaluating single equation linear regression models estimated by OLS. General quality of an estimated model is evaluated on the basis of coefficients of determination (standard and adjusted) and information criteria (with focus on Akaike information criterion, AIC). Multicollinearity of explanatory variables is discussed and analyzed through variance inflation factors (VIF). Statistical tests of significance (t test for individual variables, F test for sets of explanatory variables) and specification tests (RESET, omitted variable, and Davidson-MacKinnon tests) are introduced and applied to models estimated on the basis of current economic data sets. Detection of outliers and remedial measures in presence of atypical observations are discussed. Several single equation linear regression models are evaluated by students in class. Computer lab exercises are designed to apply tests and methods presented during lecture to real-life economic data. Class structure: lecture (2 hours) + computer lab (3 hours). Basic: J. Wooldridge, Introductory. A Modern Approach, South- Western, 2009 [chapters: 4, 6, 7 and 9.1] Supplementary: M. P. Murray,. A Modern Introduction, Pearson Education, 2006 [chapters 7 8] Problems 3 and 5; additional questions and problems are available at the publisher s webpage (www.cengage.com). 2
3. Autocorrelation and heteroscedasticity of the error term. This class discusses tests of properties of the error term, and their consequences for OLS estimators. Serial correlation (autocorrelation), heteroscedasticity, and normality of the error term and described. Causes, consequences, and remedial measures are discussed in each case. Tests for serial correlation (Durbin-Watson, LM), heteroscedasticity (White) and normality (Jarque-Bera) are presented and applied to models based on current economic data. Finally, asymptotic properties of estimators and test statistics are described. All the tests described in the lecture are later carried out by students in computer labs; consequences and practical implications of the results are comprehensively discussed. On the basis of econometric software package (e.g. GRETL) printouts, the diagnostic algorithm for evaluation of a single equation linear regression models is presented, and typical estimation results are reviewed. Class structure: lecture (2 hours) + computer lab (3 hours). Basic: J. Wooldridge, Introductory. A Modern Approach, South- Western, 2009 [chapters: 5, 8 and 12] Supplementary: M. P. Murray,. A Modern Introduction, Pearson Education, 2006 [chapters 10 12] Problem 3; additional questions and problems are available at the publisher s webpage (www.cengage.com). Quiz to be announced (to take place next week). 3
4. Econometric prediction. Point and interval prediction. Mean error of prediction ex ante, errors ex post. Exponential smoothing. This class is centered on econometric methods of forecasting. It presents assumptions and properties of econometric forecasting, including test of stability of model parameters (Chow test). Necessary conditions for economic forecasting are enumerated. Point and interval forecasts are described, along with ex ante and ex post prediction errors (ME, MSE, MAE, MAPE, Theil s coefficient). Applications of standard and relative (percentage) errors and discussed. Selected statistical methods of forecasting (i.e., exponential smoothing by various methods) are also presented. Advantages of employing lagged explanatory variables and trend variables in forecasting models are reviewed. Several forecasting models are estimated and evaluated by students from the point of view of their predictive performance. Special attention is allotted to ex post forecast errors as the major empirical measure of forecast quality. Class structure: lecture (2 hours) + computer lab (3 hours). Basic: J. Wooldridge, Introductory. A Modern Approach, South- Western, 2009 [chapter 18.5] Supplementary: P. Kennedy, A Guide to, Blackwell Publishers, 1998 [chapter 18]; M. P. Murray,. A Modern Introduction, Pearson Education, 2006 [chapter 17]; additional reading assignments may be suggested by the instructor. Problems 2 and 4; additional questions and problems are available at the publisher s webpage (www.cengage.com). 4
5. Nonlinear models. Linearized models. Production and consumption functions. Marginal measures. Substitution. This class is focused on econometric nonlinear models, estimation methods for such models, interpretation of results, and testing. Differences between models linear and nonlinear in parameters / in variables, directly linear and intrinsically linear (linearized) models are discussed. Types of nonlinearities are presented, i.e. logarithms, squares, products of variables. Marginal effects and partial elsticities are introduced. Examples of nonlinear models are presented (Engel curve, S-curve, loglinear model, exponential model, logistic function). Estimation methods of nonlinear models are characterized. Example nonlinear models are estimated using econometric software (e.g. GRETL and its database) and tests of linearity and structural breaks are applied. The types of production functions as special cases of nonlinear models are presented (Cobb-Douglas, Constant Elasticity of Substitution, Translog function) and the related vocabulary is explained (returns to scale, isoquants, factor substitution, marginal rate and elasticity of substitution, capital-to-labor ratio, marginal productivity). Some exercises related to interpretation of results from estimated production models are done during the lab and more problems are left to prepare at home. Class structure: lecture (2 hours) + computer lab (3 hours). Basic: J. Wooldridge, Introductory. A Modern Approach, South- Western, 2009 [chapters 2.4 and 6.2 on different functional forms of variables] Supplementary: C. Dougherty, Introduction to, Oxford University Press, 2007 [chapters 4, 5], R.C. Hill, W.E. Griffiths, G.G. Judge, Undergraduate econometrics, John Wiley & Sons, 2001 [chapter 10], J.H. Stock, M.W. Watson, Introduction to, Pearson Education, 2007 [chapter 8]; additional reading assignments may be suggested by the instructor. Problems 2.7-2.9, 6.3-6.5, and C2.6, C6.10 on functional form of variables, additional questions and problems are available at the publisher s webpage (www.cengage.com). 5
6. Models of qualitative variables. Linear probability model, logit model, probit model, tobit model. This class introduces limited dependent variable models, namely the linear probability model, logit, probit, and tobit models. The interpretation of the qualitative dependent variables in such models is explained and differences between the specifications of these models are discussed. The methods to estimate the models (e.g. maximum likelihood, generalized least squares) are briefly mentioned. Different approaches to interpret estimation results, employing odds-ratios and marginal effects, and to predict binary variables are presented. Comparison and testing of the binary dependent variable models is done with the use of pseudo-rsquared and likelihood ratio statistics, classification tables, Cramer s rule, and count- R-squared statistic. Censored and truncated samples / variables are defined. The tobit model, explaining censored variables, is explained. Interpretation of marginal effects in tobit models is described. Example qualitative variable models are estimated using econometric software (e.g. GRETL and its database) during the lab. Different specifications are tested and compared. The results are interpreted and more problems are left to prepare at home. Class structure: lecture (2 hours) + computer lab (3 hours). Basic: J. Wooldridge, Introductory. A Modern Approach, South- Western, 2009 [chapter 7 on linear probability model and qualitative variables, chapter and chapter 17 on logit, probit, and tobit models, excluding section 17.3] Supplementary: Dougherty, Introduction to, Oxford University Press, 2007 [chapter 10], M. P. Murray,. A Modern Introduction, Pearson Education, 2006 [chapter 19], additional reading assignments may be suggested by the instructor. Problems 7.1, 7.2, C7.4 on qualitative variables, and problems C17.1, C17.2 on logit/probit models, C.17.3 on tobit models; additional questions and problems are available at the publisher s webpage (www.cengage.com). 6
7. Time series econometrics. Stationarity and integration of time series. ARIMA models. The problems related to time series analysis using econometric models are discussed during this class. Stochastic processes, e.g. random walk and white noise, are defined. The problem of (non)stationarity of economic variables is introduced and basic tests of stationarity (different specifications of the augmented Dickey-Fuller test) are presented. The methods to check if the series is trend-stationary or difference-stationary and to measure the order of integration of time series are explained. Example time series are analyzed (in terms of their trends, stationarity, integration, and autocorrelation) using econometric software (e.g. GRETL and its database) during the lab. The autocorrelation and partial autocorrelation functions are presented. The methods of specifying autoregressive models, moving average models, and autoregressive integrated moving average (ARIMA) models are presented. Lag and difference operators and lag polynomials are briefly mentioned. Seasonal ARIMA models are defined. The optimal ARIMA models for some variables are selected using econometric software (e.g. GRETL and its database) during the lab. The empirical results are interpreted and more problems are left to prepare at home. Class structure: lecture (2 hours) + computer lab (3 hours). Basic: J. Wooldridge, Introductory. A Modern Approach, South- Western, 2009 [chapter 10.1, 10.5 on time series data, trends and seasonality; chapter 11.1-11.3 on stationarity and autocorrelation, AR and MA models, chapter 18.2 on testing for unit roots] Supplementary: M. P. Murray,. A Modern Introduction, Pearson Education, 2006 [chapters 17.2, 18], Stock, M.W. Watson, Introduction to, Pearson Education, 2007 [chapter 14]; additional reading assignments may be suggested by the instructor. Problems 10.1, C11.1, C18.2 on AR and MA models and testing unit roots, additional questions and problems are available at the publisher s webpage. 7
8. Advanced time series econometrics. Cointegration. Distributed lag models, error correction model Construction of dynamic econometric models is discussed during this class. Distributed lag models with a finite and infinite number of lags (Koyck model), and autoregressive distributed lag (ADL) models are presented. Short-run and long-run impact multipliers are defined and economic examples are provided. Specification methods for ADL models are presented, including the general-to-specific and specific-to-general approaches. The problem of spurious regression is introduced. The concepts of cointegration and cointegrating relationships are then presented as a way to analyze nonstationary variables in econometric models. The error correction mechanism in error correction models is then explained. Example ADL models are estimated and tested using econometric software (e.g. GRETL and its database), and empirical results are discussed during the lab. Impact mulitpliers are calculated. The best model specifications are selected and compared according to different criteria (e.g. information criteria, insignificant variables included in the model). Tests of stationarity are reviewed and cointegration links between variables are tested. The error correction models (ECM) are constructed. Some additional problems are left to prepare at home. Class structure: computer lab (3 hours). Basic: J. Wooldridge, Introductory. A Modern Approach, South- Western, 2009 [chapter 10.2 on finite distributed models and impact multipliers; chapter 18.1 on infinite distributed lag models, 18.3 on spurious regressions, 18.4 on cointegration and ECM] Supplementary: Stock, M.W. Watson, Introduction to, Pearson Education, 2007 [chapter 16.4]; M. P. Murray,. A Modern Introduction, Pearson Education, 2006 [chapter 18.7]; additional reading assignments may be suggested by the instructor. Problems 10.3, 10.6, 18.6 on distributed lag models, 18.2, 18.4, 18.5 on error correction models, C18.2 on cointegration; additional questions and problems are available at the publisher s webpage (www.cengage.com). 8
9. Input output balance. Major input output macroeconomic indices. Leontief model. This class focuses on input output balance tables (transaction tables), evaluating performance of economy on their basis, and forecasting. Structure of an input-output table for closed and open economy is discussed, and balance equations are derived. On the basis of input output tables and balance equations, measures of efficiency of production processes are discussed (including raw materials output ratio, imports output ratio, labour costs output ratio, profitability and labor productivity coefficients), and national accounts aggregates (gross national product, national income) are defined. Next, cost coefficients matrix, Leontief matrix, and Leontief model are introduced. First-type, second-type and mixed forecasts are calculated and applied to Polish macroeconomic data. Class structure: computer lab (3 hours). Basic: presentation slides and class notes Supplementary: O. Lange, Introduction to, PWN, 1978 [chapter 3] Problems 2 and 5; additional reading assignments may be suggested by the instructor. Quiz to be announced (to take place next week). 9
10. Macroeconomic models. Multi-equation models. Applied general equilibrium models. This class introduces simultaneous equations models and provides examples of their empirical applications. Issues of identification and estimation of structural simultaneous equations models are discussed, as well as simultaneity bias in OLS. Simultaneous equations estimation methods are classified in three major categories: single-equation methods, systems methods, and VARs (vector autoregressions). Single equation methods that is, OLS, ILS (Indirect Least Squares), IV (Instrumental Variables), 2SLS (Two-stage Least Squares), and LIML (Limited Information Maximum Likelihood) are then discussed in detail. Time series and panel data simultaneous equations models are described. Theoretical subjects are illustrated with applied general equilibrium models. Class structure: computer lab (3 hours). Basic: J. Wooldridge, Introductory. A Modern Approach, South- Western, 2009 [chapter 16] Supplementary: P. B. Dixon, B. R. Parmenter, A. A. Powell, P. J. Wilcoxen, Notes and Problems in Applied General Equilibrium Economics, North Holland, 1992 [chapter 2]; P. Kennedy, A Guide to, Blackwell Publishers, 1998 [chapter 10] Problems 2 and 4; additional reading assignments may be suggested by the instructor. 10
11. Decision problem. Optimization model. Linear optimization problem. Graphical method of solving LP model. Properties of LP models. This class introduces methods to solve some decision problems that can be formulated as linear programming (LP) problems. During the class the linear programming problem is defined and its components (constraints, objective function, decision variables, parameters) are explained. Several mathematical examples of problems with different feasible solution and optimal solution sets are presented. All types of graphical solutions for the two-variable problem are described (e.g. alternative optimum values, unbounded objective functions, infeasible problems). The graphical method to solve LP problems is explained and the important keywords are discussed (e.g. gradient vector, coordinates of the point, isoquants, maximum/minimum of the objective function, corner solution, binding constraints). Class structure: computer lab (3 hours). Basic: N. Balakrishnan, B. Render, R.M.Stair, Managerial Decision Modeling with Spreadsheets, Pearson Prentice Hall, 2007 [chapter 2 on linear programming models: definitions, solutions, graphical method] Supplementary: D. R. Anderson, D. J. Sweeney, T. A. Williams, K. Martin, An Introduction to Management Science. Quantitative Approaches to Decision Making, Thomson South-Western, 2008 [chapter 2]; additional reading assignments may be suggested by the instructor. Problems 2-13 to 2-19 on using the graphical method to solve LP problems, discussion questions 2-1 to 2-12 on definition and characteristics of LP problems and 4; additional questions and problems are available at the publisher s web page and Student CD-ROM included with the book. 11
12. Main types of optimization problems: product mix, diet, transportation problem, investment portfolio etc. Application of SOLVER from Excel spreadsheet. During this class the most popular real-life linear programming (LP) problems are reviewed. The examples of a diet problem, investment problem, scheduling problem, blending problem, transportation problem and assignment problem are presented and solved. Different characteristics related to specific problems are highlighted, e.g. integer (or natural, or binary) values of decision variables in some problems, balanced and unbalanced versions of transportation problems. The focus is on defining the real-life problems using mathematical formulas. Some problems are to be prepared by students at home. One possible software solution designed to solve LP problems is the application SOLVER in the Excel spreadsheet program. This application is presented during the class and some example problems are solved using the software and some are additionally solved graphically to compare the results. Class structure: computer lab (3 hours). Basic: N. Balakrishnan, B. Render, R.M.Stair, Managerial Decision Modeling with Spreadsheets, Pearson Prentice Hall, 2007 [chapter 3 on different real-life decision problems, chapters 2 and 3 on examples using Excel] Supplementary: D. R. Anderson, D. J. Sweeney, T. A. Williams, K. Martin, An Introduction to Management Science. Quantitative Approaches to Decision Making, Thomson South-Western, 2008 [chapter 4]; additional reading assignments may be suggested by the instructor. One of the problems 3-1 to on 3-6 optimal product mix, problem 3-8 on the production plan, 3-10 on the lest expensive sampling plan, 3-14 or 3-15 on portfolio selection, 3-19 on optimal scheduling, 3-47 on transportation problem ; additional questions and problems are available at the publisher s web page and Student CD-ROM included with the book. 12
13. Post-optimization analysis. Some specific post-optimization problems are presented during this class. These problems are graphically illustrated using simple example problems. Sensitivity of the optimal solution to changes in coefficients of the objective function, to changes in a constant term in the linear constraint, and to changes in the number of linear constraints are discussed. Example problems are solved during the class. Regions (intervals) of stability of optimal solutions and stability regions for the base structure of the optimal solutions are found in specific example problems. Dual problems are shortly mentioned and the shadow price is defined and found using example problems. Some additional problems are left to prepare at home. Class structure: computer lab (3 hours). Basic: Basic: N. Balakrishnan, B. Render, R.M.Stair, Managerial Decision Modeling with Spreadsheets, Pearson Prentice Hall, 2007 [chapter 4 on sensitivity analysis of decision problems] Supplementary: D. R. Anderson, D. J. Sweeney, T. A. Williams, K. Martin, An Introduction to Management Science. Quantitative Approaches to Decision Making, Thomson South-Western, 2008 [chapter 3]; additional reading assignments may be suggested by the instructor. Problems 4-10, 4-16, 4-18 (or any other problems between 4-10 and 4-37) on sensitivity analysis of optimal solution in LP problems, 4-19 and 4-20 on dual problems and shadow prices, discussion questions 4-1 to 4-9 on the theory of sensitivity analysis in LP problems; additional questions and problems are available at the publisher s web page and Student CD-ROM included with the book. 13
14. Model of multi-activity project. Critical path method. This class discusses management techniques for multi-activity projects. General introduction to challenges of managing large investment or production projects is provided. Then, network diagram of a multi-activity project is explained, and its applications described with special attention paid to event schedule design, critical time evaluation, and slacks and floats analysis. Basic techniques of time-cost analysis of a project are introduced. Links between multi-activity project modeling and linear programming methods (see topics 11 13) are explained. Finally, advantages and disadvantages of the critical path method in project management are discussed. Class structure: computer lab (3 hours). Basic: N. Balakrishnan, B. Render, R.M.Stair, Managerial Decision Modeling with Spreadsheets, Pearson Prentice Hall, 2007 [chapter 7] Supplementary: D. R. Anderson, D. J. Sweeney, T. A. Williams, K. Martin, An Introduction to Management Science. Quantitative Approaches to Decision Making, Thomson South-Western, 2008 [chapter 10]; H. Kerzner, Project Management. A Systems Approach to Planning, Scheduling, and Controlling, Van Nostrand Reinhold, 1995 [chapter 12] Problems 4, 6, and 8; additional questions and problems are available at the publisher s web page and Student CD-ROM included with the book. 14