Learn Econometrics Fast

Preface AN INTRODUCTION TO APPLIED ECONOMETRICS

These guides are intended for students having no knowledge in econometrics and little knowledge in statistics and in probability.

For a long time, the tradition in France was to teach econometrics the hard way. In a first stage, students had to learn plenty of mathematical results on various classes of estimators and tests. During this time they had to believe that their arid investment will be profitable in the future and will allow them to deal with economic data and to answer economic problems. Later on they could turn to applications.

A problem with this method is that many students became discouraged during the first step of the process. Another problem was that many students who had reached the second step had a tendency to turn to very sophisticated and fragile methods when they faced simple practical problems. Sometimes, the results they reached were crazy when they mixed very complicated methods, with very elementary mistakes contradicting basic common sense. The most serious mistakes in econometrics, which can even be found in articles published by good journals, come from not spending time enough looking at the data, in a pragmatic way, without a priori and without the strong desire to apply complicated methods that are totally inappropriate for these data.

CONTENTS

Introduction

Chapter 1. Descriptive statistics

Graphs

Mean and other numerical summaries

The concept of robust summary

Chapter 2. Bivariate analysis

Correlation

An introduction to simple regression

Statistical aspects of regressions

Robustness of regression

Tutorial: the capital asset pricing model (CAPM) (Berndt, chapter 2)

Chapter 3. Multivariate analysis

Multiple regressions

Tutorial: Costs, learning curves and scales economies (Berndt, chapter 3)

Partial correlation

Fragility analysis

Regressions with dummy variables

•  Tutorial: Analysing the determinants of wages and measuring wage discrimination (Berndt, chapter 5)

•  A last example as a conclusion

Chapter 4. The econometric of time series

Regression with time lags: distributed lag models

Univariate time series analysis: the autoregressive model of order 1 (the AR(1) model)

Univariate time series analysis: the autoregressive model of order p (AR(p) model)

Regression with time series variables : the case with stationary variables

Regression with time series variables: the case with non-stationary variables: spurious regressions

Regression with time series variables: the case with non-stationary variables: Cointegration

Chapter 5. Exogeneity

An example

The concept of weak exogeneity

Comment

Identification and estimation

How to test the weak exogeneity of a variable: the Haussman test

Conclusion

Spécification, estimation et tests : l’approche de la Cowle Commission

Une première philosophie alternative : du général au spécifique (David Hendry)

Une deuxième philosophie alternative : l’analyse exploratoire des données (EDA)

Une troisième philosophie : l’analyse de fragilité ou de sensibilité (Leamer)

Conclusion

These guides are intended for students having no knowledge in econometrics and little knowledge in statistics and in probability.

For a long time, the tradition in France was to teach econometrics the hard way. In a first stage, students had to learn plenty of mathematical results on various classes of estimators and tests. During this time they had to believe that their arid investment will be profitable in the future and will allow them to deal with economic data and to answer economic problems. Later on they could turn to applications.

A problem with this method is that many students became discouraged during the first step of the process. Another problem was that many students who had reached the second step had a tendency to turn to very sophisticated and fragile methods when they faced simple practical problems. Sometimes, the results they reached were crazy when they mixed very complicated methods, with very elementary mistakes contradicting basic common sense. The most serious mistakes in econometrics, which can even be found in articles published by good journals, come from not spending time enough looking at the data, in a pragmatic way, without a priori and without the strong desire to apply complicated methods that are totally inappropriate for these data.

These notes follow a completely different principle. I will introduce econometrics through a series of simple applications. I will use little mathematics, and I will be little rigorous. I will appeal to the common sense and the intuition of the reader to introduce the basic concepts, methods and traps of econometrics. I will try to show that econometrics is simple, and thinking in an econometric way is the same as thinking in an economic way. Sometimes, the developments will be a bit tricky, and I hope as funny as the kind of riddles and puzzles you can find in newspapers and magazines. The book by Berndt (quoted among the references) is entertaining and pleasant to read (with much gossip on the profession, so you can discover that econometricians are also human beings). Finally, econometric methods give answers to economic questions, and these answers must be understandable and look convincing to people who are experts of these questions and not econometricians. So, introducing students to econometrics through applications is sensible.

There is a limit to the approach followed in these notes, and students are expected to feel it more and more when they progress in this course. Examples and intuition quickly meet their limits and to go further we must use logical and rigorous methods. So mathematics is unavoidable, and, after having read these notes students must learn a book of econometrics, which includes the mathematical foundations of this field. However, doing that in a second stage of learning, after having gone through these notes, will be a task much easier than starting directly with the mathematics of econometrics.

There are many user-friendly econometric software. I will advise you to use E-Views or Stata. Both are known and used in the whole world and it is much wiser to learn and use a software that is a world standard. Loosely speaking E-Views is well adapted to macroeconomics (time series-data) and Stata to microeconomics (individual data). However, the availability of such user-friendly software may encourage laziness and the absence of reflection. This is a pity because these software include powerful graphic capacities, and plenty of descriptive statistics, which are extremely precious to look at the data and to learn much of them[1].

Alexander Pope wrote: Little learning is a bad thing. Henri de Monfreid tells a nice story. When he was smuggling weapons on the Red Sea, he had a good friend, who was a policeman and the only French official on a small island near Djibouti. He told of his friend that he had the wisdom of men of the people who had not been spoiled yet by compulsory public education. Both comments are a bit arrogant and even reactionary, but they are basically true. Many applied econometricians use sophisticated methods, which were developed by experts in the field, and they apply them to their problems and data without further reflection. This imitation process has become very easy with the existing software and the availability of many programs on the websites of Eviews and Stata. In general these methods were correct for the problem they were designed for, and their developers did not make mistakes in their applications.  But the adaptation of these methods to problems they were not designed for can be awfully wrong. So, these applied econometricians make logical mistakes and draw silly conclusions. In my lifetime I read many strange papers of applied econometrics with results, which were meaningless, incomprehensible and unbelievable. These papers generally were in development economics and macroeconomics, but this can result from the fact that most of my readings are in these fields. These applied econometricians had a superficial knowledge of theoretical econometrics and tried to substitute recipes to logic. So, they had little learning[2]. However, they could have avoided their mistakes if they had not lost their common sense, the wisdom of the ignorant. These notes are quite insufficient to help you to solve the first problem. However, they will give you advice to help you not to progressively lose your common sense when you become more and more learned. These advices can be summed up in tow sentences. First, do not forget that you are an economist and that your econometric results must be explainable in plain French (or English) and without cheating, to a non-econometrician economist. Secondly, look carefully at the data and do not apply a method based on assumptions, which are contradicted by these data. In summary, econometrics must not make you lose your common sense.

Econometrics is a set of quantitative tools for analysing economic data. Economists need to use economic data for three reasons: 1) to decide between competing theories; 2) to predict the effect of policy changes; 3) to forecast what may happen in the future. Three examples: Have PC increased the productivity