Practical Econometrics
Practical Econometrics
In most economics courses it is never the intention to use or require advanced econometrics skills. However, a certain level of basic knowledge and basic standards is required when reading academic journals and making empirical assignments.
Econometrics is the discipline involved in empirical measurement and testing of economic relationships and the techniques to solve the associated problems. Estimation and measurement has a large theoretical basis. Different estimators and hypothesis tests depend on a clear understanding of their statistical (distributional) properties. Courses in econometrics provide these theoretical aspects in the form of mathematical/statistical derivations, proofs, and algebra.
Practical econometrics is not involved with derivations, proofs and matrix algebra. Practical econometrics relies on standard estimation techniques and tests, as they are implemented in commercial econometrics computer software. However, practical econometrics still requires the practitioner to have an adequate understanding of the issues involved in selecting the appropriate techniques and tests. In other words, computer software generates output but this has to be understood and evaluated. Garbage in - garbage out.
What we generally DO NOT need
For practical econometrics, it is generally not necessary to spend much time on the mathematical proofs and algebra in many of the econometrics textbooks. This task is taken care of by specialists (econometricians) and commercially available computer software (RATS, EViews, PCGive, etc).
Basic regression analysis using time series
An adequate understanding of the use of ordinary least squares estimation (OLS) is almost always sufficient, at least to start with. However, if you intend do some serious empirical research (for example, an empirical thesis) it is worthwhile to invest the additional effort and look into the uses of the more advanced techniques. For example, unit root tests, cointegration methods, instrumental variables (IV, TSLS), GMM, FIML, SUR, VAR/VECM, estimation with panel data, etc. Again, do not spend much time on proofs and algebra, but focus on the arguments for choosing between alternative estimation techniques, and find out how to do the work using your statistical computer package.
- Adequate understanding of the following concepts and how they affect the use of OLS in estimating a standard single equation
- stationary and nonstationary time series data, unit roots, cointegration
- serial correlation, autocorrelation
- heteroscedasticity, non-normality, outliers
- multicollinearity
- exogenous and endogenous variables
- Adequate knowledge of a number of approaches to deal with the aforementioned problems
- Adequate knowledge of testing economic hypotheses
A very accessible textbook, with particular emphasis on financial econometrics: C. Brooks, Introductory Econometrics for Finance. Cambridge Univ. Press, 2002. Help and information on some important topics can be found in the Working with EViews section. This section includes lecture slides and EViews examples.
Textbooks with particular emphasis on financial econometrics:
* T.C. Mills, The Econometric Modelling of Financial Time Series (2nd ed.). Cambridge Univ. Press, 1999. * J.Y. Campbell, A.W. Lo, A.C. MacKinlay, The Econometrics of Financial Markets. Princeton Univ. Press, 1997. * C. Gourieroux, J. Jasiak, Financial Econometrics. Princeton Univ. Press, 2001.
Other examples of textbooks are:
* R.S. Pindyck and D.L. Rubinfeld, Econometric Models and Economic Forecasts. McGraw-Hill, 1991. [Revised 4th ed. 1998]
* P. Kennedy, A Guide to Econometrics (5th ed.). Blackwell, 2003. * C. Dougherty, Introduction to Econometrics (2nd ed.). Oxford Univ. Press, 2002. * J.M. Wooldridge, Introductory Econometrics: A modern approach. South-Western College Publ., 2000. * R.C. Hill, W.E. Griffiths, G.G. Judge, Undergraduate Econometrics (2nd ed.). John Wiley & Sons, 2001. * F. Hayashi, Econometrics. Princeton Univ. Press, 2001.
The following internet references provide useful online help with collecting additional insight into regression analysis and econometrics. The presentation and level of difficulty varies. My search for better websites continues and suggestions are welcome.
Carefully select your desired topics. Focus on the practical implications and not on the algebra. Always try to make the link with the practical use of the computer software package EViews. Search for the appropriate commands used in EViews in any given situation.
An introductory statistics book and online tutorial for help in statistics courses.
* Electronic Statistics Textbook
* Using Econometrics: A practical guide
A.H. Studenmund's A Guide to Using EViews with Using Econometrics: A Practical Guide by R. R. Johnson. Welcome to the Companion Web Site featuring the web-based Guide to Using EViews with Using Econometrics: A Practical Guide. This guide presents a step-by-step explanation of how to replicate the econometric processes examined in the text with the data sets that are available via this web site.
* LSE Ec220 Introduction to econometrics
Undergraduate course by Christopher Dougherty, London School of Economics. Extensive set of teaching material with slides, datasets, study guide, handouts.
Undergraduate course in econometrics by Prof. Veitch, University of San Francisco. The emphasis is on intuitive understanding of techniques used by economists to analyze data.
* Statistics and introduction to econometrics
Bristol University.
Prof. Andrew J. Buck, Temple University. These lectures are for use in a 2 semester econometrics sequence at the graduate level.
* Econometric Resources on the Internet
John Kane, Oswego State University of New York. This site is designed to assist users of Econometrics: An Applied Approach (Boston: Houghton Mifflin, forthcoming 2001) and others who are interested in finding econometric resources on the internet.
More difficult material
* Time Series for Macroeconomics and Finance
Panel data methods
Panel data estimation has become increasingly important in empirical research in recent years, largely due to our need to increase the number of observations to reliably estimate certain economic relationships. Although in principle not radically different from normal regression analysis, panel estimation does require some adjustments in the estimation techniques.
Definitions:
* cross section: observing a group of more than one entity using one observation over some time period
* panel: observing a group of more than one entity over time (balanced or unbalanced panel, depending on the difference in number of observations over time)
* longitudinal: observing a group of the same entities over time
Pooled model (meaning all coefficients are assumed to be the same, with no group/unit or time effects represented by dummy variables)
SUR model (residuals from the group/unit equations in the panel are not independent but correlated and accounting for the correlation improvess the estimation)
Dynamic panel model (autocorrelation is present and modeled by including lags of the dependent variable)
I do not regularly work with panel data methods. The following references provide some useful further explanation and discussions of various problems.
* Yaffee (2003), A primer for panel data analysis,
http://www.nyu.edu/its/pubs/connect/fall03/yaffee_primer.html
* Petersen (2009), 'Estimating standard errors in finance panel data sets: Comparing approaches,' Review of Financial Studies, vol.22 (1) January 2009: 435-80.
* Programming Advice - Finance Panel Data Sets: based on STATA http://www.kellogg.northwestern.edu/faculty/petersen/htm/papers/se/se_programming.htm
* Arellano (2003) Panel Data Econometrics, Oxford University Press
* Frees (2004) Longitudinal and Panel Data, Cambridge University Press.
* Kling, University Tuebingen: Short introduction, panel data analysis
http://www.uni-tuebingen.de/uni/wwl/7.Session.ppt
* Greene, Stern - New York Univ.: B55.9912 Econometric analysis of panel data
http://pages.stern.nyu.edu/~wgreene/Econometrics/PanelDataNotes.htm
* ISER - EC968 Panel data methods
http://www.iser.essex.ac.uk/teaching/degree/ermij/ec968/index.php
* Edgerton, Lund University: Panel data course (2004)
http://www.nek.lu.se/nekded/Teaching/Panel/lecture_notes.doc