Econometrics

History of econometrics

The company of econometrics and Cowles Commission

If the economic scene often turned to mathematics to apprehend its object of study, one can consider that econometrics starts to develop in its modern form in the Années 1920. The Norwegian economist Ragnar Frisch will play a big role in the birth of this discipline and its Institution nalisation. Besides one allots often the paternity of the term to him. In 1930, accompanied by Irving Fisher, it founds the Société of econometrics ( Econometric society ) whose essential object is “ to support the studies in quantitative matter which tend to bring the theoretical point of view closer to the point of view of Empirique in the exploration of the economic problems ”, then in 1933, it creates the review Econometrica which will be the principal vehicle of the econometric thought.

Work of the Cowles Commission for research in economics (group of research created in 1932 with university of Colorado, which settles at the university of Chicago then at the university of Yale, whose work concern the relation between the economic theories, mathematics and statistics) will allow the development of techniques of estimates for systems of simultaneous equations. Trygve Haavelmo, prize winner of the P " Nobel Prize of économie" in 1989, belonged to this group of research.

Establishment of the discipline and modern period

After the Second world war, econometrics develops very quickly. This growth will be carried out on several Niveau X.

It first of all acquires a strong fame thanks to attribution many " Nobel Prize of économie", with economist S considered for the mathematical formalism their work.
It gains ground in the field of the economic scene which becomes increasingly dependant on the Mathématiques.
It acquires a Statut increasingly important Institutionnel. In France for example, its Enseignement is exempted in many Université S.
the econometric techniques improve, which returns them more Opérationnelle S, and what in fact of the invaluable tools of assistances to the Décision S economic S.

One can distinguish two great periods in modern econometric research. Until the end of the Years 1970, work is directed towards the specification and the Solvabilité of models Macroéconomique S with simultaneous equations. Then, following the introduction of rational anticipations and the Critical of Lucas, research turns to the Microéconomie and analyzes it time serieses.

During the first period, econometric research relates primarily to the conditions of estimate of the models Macroéconomique S of simultaneous equations comprising a random element. Chronologically, the projections will be carried out as follows:

1939 : Tinbergen builds a model of the Business cycles comprising equations of behavior. Each equation is estimated by means of the method of least squares.
1944 : Haavelmo poses the general terms of solvency.
1945/1950: Klein presents the first models whose solution is obtained by the method of the maximum of probability.
1949 : Koopmans determines the conditions of Solvabilité in the case of a linear model.
1954 : Theil introduces the method of the square least doubles.
1955 : first estimated model designed by Klein/Goldberger.
In the Years 1960 and Years 1970, the projection of the Technologies the information involves the appearance of models Macroéconomique S designed at ends of Prévision. For example, the model of Brookings includes/understands 400 equations. After 1970 were used standard models like that of Wharton.
the Années 1970 also devote the questioning of the traditional models Macroéconométrique S. In particular because, following their inefficiency to explain and envisage the consecutive stagflation with the Oil crises, they will be shown not to have sufficient foundations Microéconomique S. Lucas watch for example as of 1972 the bond between anticipations of the Economic agents and the variation of the coefficients structural of the models Macroéconométrique S. Its conclusion is whereas any measurement of Economic policy leads to a change in the behavior of the Agent S, and why consequently, these Agent S is capable to counter the government policies by anticipating them. What reduces the monetary Intérêt considerably budget policy and .

Methods of econometrics

The methods of econometrics are very varied. In mathematical economy, one finds very diverse mathematical tools like the Algèbre, the analyzes, the Game theory and the Probabilité. From the point of view of the Statistical applied to the economy, one will be able to retain the descriptive Statistique, the Data analysis and the Statistique mathematics. The econometric techniques in a strict sense, result generally from the Statistique mathematics. One will also retain the importance of the Matrix algebra.

In the econometric techniques in a strict sense, one initially finds the models of Regression linear traditional which rest on many Hypothèse S, then more complex models, which make jump to one these same assumptions.

Models of linear regression

The models of linear Regression seek to determine a linear relation between one or more explanatory variables (one used in the beginning the term of exogenous variables, but the exogeneity is a protean concept, one distinguishes the weak exogeneity today --who allows the regression -- strong exogeneity -- who allows the forecast -- and finally the superone -- useful in analysis of economic policy, to see Engle, Hendry & Richard, 1983: " Exogeneity" , Econometrica, 51, pp. 277-304) and one or more variables given (or endogenous) starting from a whole of N qualitative data or quantitative. It is considered in general that the differences between the observations and the relations between the variables can be explained by various sources of errors:

the aggregation of the behaviors of the Economic agents which do not take to all them rigorously identical Décision in manner,
the existence of errors of measurement of the variables,
the existence of the explanatory variables which are not include in the relation,
errors which come from what the true relation is not linear.

In these last three cases one says that the relation was badly specified. These sources of errors of which considered as random. They are called random elements. Very often it is supposed that of these random elements around the theoretical value, a normal Loi follows, which makes it possible to make estimates of the parameters of the model of adjustment, and to carry out tests. The objective is then to find relations linear between the variables which minimize the random error. Various methods exist. One can for example use the Method of least squares.

The technique of regression in the traditional linear models is based on some fundamental assumptions:

the relation between the endogenous variable and the explanatory variable is linear.
the constancy of this relation
independence between the random elements and the explanatory variables
the endogenous variables observed have a random element.
the expectation of the random elements is null.
the variance of the random element is constant. It is the assumption of homoscedasticity
the random elements are statistically independent between them (at various dates, for various individuals…).
the random elements are distributed according to a normal law. This property results from the central limit theorem.
the explanatory variables are not correlated between them.

Historically, it was considered that the explanatory variables were free from random elements. It is not any more the case since the years 1980, the properties of the variables endogenous and explanatory must be similar: they are random processes " similaires" , i.e is stationary, is which can be easily transformed of stationary processes of the same manner (one speaks about integrated processes).

When one makes jump one or more of these assumptions, one enters particular models econometic, of which nonlinear models.

Nonlinear model of regression

Partial Kernel Least Public gardens
Support Vector Machines
Maximum Least Public gardens
a posteriori
closer neighbors (fuzzy)
Neural Networks Regression

Nonparametric model

Tests of causality

Dynamic models and Time serieses

AR (Autoregressive Model): The observations delayed of the variable to explain $Y \,$ are included in the model.

MY ( Moving Average in English, Moving average in French): The residues of the previous periods are in the model.

ARMED: This class of model generalizes the processes AR and MY.

VAR (autoregressive process vectorial): Process AR but generalized with the multivariate case.

Principal results and applications of econometrics

The econometric tests brought an interesting lighting to economic theories from which certain economic policies were drawn. For example, the monetarists will call into question the relevance of the curve of Philips while being based on statistical data of long run. In the same way, the econometric tests strongly will weaken the bond, crucial at the keynésiens, between national consumption and national revenue.

Concerning the applications, econometrics will find outlets in the Finance and the budgetary Economic policies and financial. The model developed by Black and Scholes for example, makes it possible to calculate the value of the options. Models Macroéconométrique S were also set up as of the Années 1950 to envisage the impact of the economic policies. The question arises consequently of knowing if the use of the macroeconometric models has an impact on its object of study. Problem that the monetarists and the theorists of rational anticipations as Lucas will not fail to underline.

Problems epistemological S of econometrics

It is at the beginning with an aim of making coincide the economic theories with the data, that the economists set up a whole whole of econometric techniques. Their objective was to determine which economic models could be regarded as most probable. Thus, while being based on a selection process of the models of the type poppérien (the least probable models in front of being gradually eliminated), they hoped to be able to approach the economic Réalité as well as possible without having to intervene on it . However, this improvement was far from being systematic. This, in spite of an increasing perfection of the econometric techniques. The mathematical high-tech in econometrics should not thus mask the fact that the economy cannot claim to become: on the one hand an applied science as well as the Physique, the Chimie, the Biologie or the experimental Psychologie, and on the other hand a science of as precise Observation and “neutral” as the Cosmology or the Demography.

Without claiming with exhaustiveness, one can see several reasons there:

Of the reasons of an ethical nature

At the Macroeconomic level , an experiment of great width could plunge a country in the recession and thus engage the life of million people. At the Microeconomic level , any intervention would compete with the Agent S already implied on the ground. Generally, any form of Expérimentation which implies human people poses obvious problems ethical S.

Of the reasons which relate to the specificity of the Economic agents

Certain theories and economic predictions cannot be directly tested by the Expérimentation, because it is not possible to retort all the economic phenomena under conditions of laboratory, in particular when the object of study is too vast (size of a nation or more). In addition, as soon as they take note of the economic laws, the Agent S can modify their Comportement S (the typical example is the Prophétie car-director). Moreover, the social behaviors complex and are always subjected to a strong share of Aléa S, Irrationalité, and Free-referee. Who more is, the economic Réalité is often very changing. It is then difficult to compare the economic variables over long periods, because the economic conditions can vary whole with the whole. Lastly, behind the economic choices very often a choice subjacent Politique hides which occurs inside a specific social environment; what does not fail to raise the problem to know if the economic laws are laws Naturelle S, or if they are the result of the voluntary intervention of the Economic agents.

Of the reasons epistemological S

The variables selected to encircle the economic object can be more or less directed according to certain purposes specific to the observer (this one will show only one aspect partial and partial of economic reality). Another point, various tests can lead to the validation of antagonistic assumptions, because the variability of the econometric techniques, the historical configurations, of the available data, and theoretical uncertainties, always do not make it possible to slice between various models. And finally, the rejection of a model necessarily does not imply that this model is unrealistic. It indicates only, that under certain conditions, which will not necessarily reproduce, a model has strong a Probabilité of being unrealistic or false. Last point, on the methodological level, the recourse to mathematics in economy is subject to deposit for many reasons (variability of the behavior, complexity of the object of study, erratic fluctuations of the prices, not very robust correlations between the variables, etc)

In short, econometrics is well far from being a “hard science”. However, if one disregards these many difficulty, it should be admitted that it showed a certain effectiveness to measure the impact of the macro-economic policies. The part “validation of the theories” being the subject as for it of interminable controversies.

In addition, it will be noticed that the success of econometrics inevitably does not achieve the unanimity in the economists. Certain thinkers of the Austrian school like Ludwig von Mises, dispute the interest of the formalization of the economic behavior. Moreover, as John Kenneth Galbraith had noted it, the professional economy is organized hierarchically: economy hétérodoxe at the base, néo-liberal economy at the top and the most mathematical forms of the néo-liberal economy to the point. One suspects well that this classification Hiérarchique creates dissensions within the Communauté of the economists, whose certain members refuse to adopt a mathematical formalism considered to be sometimes excessive and superfluous.

It as should be noted as recently, the experimental economy, news connects economy consisting in carrying out experiments of laboratory to test the micro-economic models, came to compete with econometrics on the ground of the validation of the theories. The results which emanate from this new discipline are often in obvious contradiction with the assumptions which underlie the models ecometric founded on models of general stability, which tends to reduce the range Heuristique of microphone-econometrics.

“Nobel Prize” of economy

The Prix of the Bank of Sweden in economic scenes in memory of Alfred Nobel (“Nobel Prize” of economy) was decreed on several occasions for econometric work:

Jan Tinbergen and Ragnar Frisch received the first “Nobel Prize” of economy decreed in 1969 (macroeconometry);
Lawrence Klein received the price in 1980;
Daniel McFadden and James Heckman shares the price in 2000 for their work in microeconometry;
Robert Engle and Clive Granger received the price in 2003 for their economic analyzes of the time serieses. Engle is a pionner method ARCH and Granger developed the method of cointegration.

See too

Internal bonds

History of the economic thinking
Function Cobb-Douglas
Method of least squares
linear Regression
nonlinear Regression
Translation of terms of English-French econometrics

External bonds

Company of econometrics
the gate of econometrics
research in econometrics in France
John J. Siegfried, has First Lesson in Econometrics , Journal off Political Economy , 78:6 (November-December 1970), pp. 1378-1379
Cours of Econometrics of Master Recherche to the Institut of political studies of Paris. Resources on econometrics.
Association of Econometrics Applied, Lira in line
Course of Econometrics
Site of Exercises, Course and Annals of Mathematics for Economists, University of Lille

Be-X-old: Эканамэтрыка Simple: Econometrics

Random links: