JOURNÉE D'ÉTUDES
(Écono)métrie: de l’arithmétique politique
à la révolution probabiliste
(Econo)metrics: from Political Arithmetic
to the Probability Revolution
Rome, 7 September 2012
École française de Rome
Piazza Navona, 62
00186 Roma
-----------------------------------
Abstracts
Quantifying the Social Sciences: An Historical and Comparative
Perspective
Alain
Desrosières (INSEE, Paris)
The
various social sciences have gradually been quantified since the
middle of the 19th Century. This quantification was seen as a symbol
of the attainment of a scientific status, comparable to that enjoyed
by natural sciences: “There is no science without measure",
said a slogan of the 19th Century. But this process followed
different paths for each of these disciplines. If the history of the
process of quantification is now well documented by several studies,
less frequent are the attempts to compare social sciences from this
point of view. The way each science did integrate the statistical and
probabilistic tools tells something not only about its own
epistemology and methodology, but also, in the sociology of science
perspective, about its actors, its networks, its norms, its criteria
of legitimacy, its controversies.
Without
actually replying to such a broad question, we will propose here a
small comparative panel including five disciplines: history,
sociology, political science, economics and psychology. Each one is
in itself a complex world, split into different trends, into
“schools” implying different paradigms; and controversies, when
they are not harsh conflicts, are usually present. However, what
distinguishes a disciplinary field is a relative agreement on what
people do not agree about, where the people concerned are quite
used to confrontation. On the other hand, confrontations between one
discipline and another are unusual, for reasons related to the
sociological boundaries of academic and scientific communities. Each
discipline is a disciplined world, largely confined to itself, with
its vocabulary, its paradigms, its institutions, its chairs, its
journals. That's why choosing the history of quantification as an
interpretative framework and as a symptom of something that would be
characteristic of these five worlds, may be a good idea, even if this
exercise is very simplistic.
Symptoms
and Measurement: Studying Economic Reality in France and Italy Before
WWI
Alberto Baffigi (Banca d’Italia)
In
the late eighties of the nineteenth century, scholars such as the
Austrian Franz Xaver von Neumann-Spallart (1837-1888) and French
Alfred de Foville (1842-1918) made important contributions that
favored the rapprochement between economics and statistics. The
economic depression in the seventies had strengthend scholar’s
awareness of the intrinsic instability of the economic system which
was born with the industrial revolution. This led economists to use
the information and statistical methods available to measure and
predict the behavior of economic activity. The scholars engaged in
the new discipline behaved, metaphorically, as the doctor who
performs diagnostics on the human body; they performed interpretation
of signs: economic semiology. This line of research witnesses the
need of the economists to bridge the gap between theory and economic
reality.
The history of economic semiology shows
that the development of this discipline was largely a French and
Italian business: the main founder of the discipline, De Foville was
French. The Italian Maffeo Pantaleoni, who was one of the main
promoters of this line of research, wrote an important theoretical
article about semiology, in 1892, in French, on Charle Gide’s Revue
d'économie politique. Pantaleoni was very influential and had an
impact on Rodolfo Benini, on Giorgio Mortara, Costantino Ottolenghi
and the Belgian statistician Armand Julin. In 1913, during the
fourteenth session of the International Institute of Statistics,
Julin presented a proposal to establish, within the Institute, a
special commission appointed to study the statistical methods related
to semiology. Among the sixteen subscribers of the proposal we find
Rodolfo Benini, Maffeo Pantaleoni and Lucien March. In a rapid survey
of the major contributions in the history of the discipline, also
listed Julin, in addition to his work, that of the two founders of
the discipline, Neumann-Spallart and de Foville, those of French
André Liesse (1854-1954) and Yves Guyot (1843-1928), and those of
Rodolfo Benini (1862-1956), Augusto Bosco (1859-1906) and Maffeo
Pantaleoni (1857-1924).
Although not strangers to the positivist
culture, proponents of economic semiology did not identify knowledge
with scientific induction. The facts bear, ultimately, no
informational content if they are not observed through the lenses of
a theory previously developed to detect and interpret them. As
pointed out by Rodolfo Benini, in its Principi di statistica
metodologica (1906), "regular patterns found in observed
cases cannot be extended to cases outside of our observational field,
without a bridge between the known and the unknown. This bridge is
the hypothesis". It is not useless here to remind that only 4
years earlier Henri Poincaré had expressed his conventionalist
epistemological view in “La science et l’hypothese”.
The economic semiology should be framed
in the contemporary epistemological debate that, aware of the process
of crisis of positivism, looked for a solution inside the science, in
the accurate definition of the problems of scientific research,
language and logic; these thinkers rejected irrational opposition to
positivism, keeping their positions far from idealism and from the
“post-modern” nietzschean position. The rise of economic
semiotics required a change of perspective, a way out of the doldrums
where the positivist metaphysics had ended up. Upstream of the drive
to study the symptoms of the economic movement, there was an
anti-realist and empiricist epistemological shift, there was logic
pragmatism and Empiriocriticism. Only this change of path could give
rise to semiology: the interpretation of signs requires a theory, a
logical theoretical framework within which to bring the symptoms
observed, which otherwise would be meaningless accidents.
The Early Years of the Bureau
of Agricultural Economics. Price and Crop Outlook Studies, 1922-1930
Eric Chancellier (University of Lorraine)
The U.S. Department of
Agriculture decided to create on July 1, 1922, the Bureau of
Agricultural Economics (BAE). The agricultural depression of the
early 1920’s, which brought change in relative prices of farm
products as well as a general lowering of the farmer’s purchasing
power, gave the signal for renewed emphasis on the study of the
dynamic forces which bring about price changes and for a study of the
possibilities of a reshaping of agriculture to adjust supply to
demand on the basis of a satisfactory price. Henry C. Taylor – head
of BAE – began an ambitious program of producing an annual outlook
for agricultural production and prices. He developed a program for
collecting agricultural statistics and for developing quantitative
methods in order to forecast prices commodity and crops (Taylor and
Taylor, 1954). This paper is organized as follows. The first section
will present the new method of the BAE about the analysis of prices.
We will use the works of Bean (1929a, 1929b) and Ezekiel (1923, 1924)
- BAE economists -. These authors develop a new analytical method -
multiple and partial correlation method - designed to better
understand the formation of agricultural prices. Our second section
will be devoted to the crop forecasting. For this, the BAE study
focuses on three directions: the estimation of acreage (Bean, 1930),
the estimation of crop yields (Smith, 1925a) and the impact of
weather on crops (Smith, 1925b). All these authors use extensively
the correlation tool to enable better predictions.
Bean L.H. 1929a. A
Simplified Method of Graphic Curvilinear
Correlation. Journal of the American Statistical
Association, 24(168) : 386-397.
_______. 1929b. The
farmers’ response to price. Journal of Farm Economics,
11(3): 368-385.
_______. 1930.
Application of a simplified method of correlation to problems in
acreage and yield variations. Journal
of the American Statistical Association, 25(172):
428-439.
Ezekiel M. 1923. On the
use of partial correlation in the analysis of farm management data.
Journal of Farm Economics, 5(4): 198-213.
_______. 1924. A
method of handling curvilinear correlation for any number of
variables. Journal of the American Statistical Association,
19(148): 431-453.
Smith B.B. 1925a.
Forecasting the acreage of cotton. Journal of the American
Statistical Association, 20(149): 31-47.
_______. 1925b.
Relation between weather conditions and yield of cotton in Louisiana.
Journal of Agricultural Research, 30: 1083-1086.
Taylor
H.C. et Taylor A.D. 1952. The story of agricultural economics
in the United States, 1840-1932. Men, services, ideas. Iowa State
College Press, Iowa.
Difficulties and
ambiguities of a probabilistic econometrics
Michel Armatte (Université
Paris Dauphine et EHESS/ Centre A. Koyré)
The
history of econometrics is now rich of many studies, but it has
suffered from an internalist methodological vision that singled it
out of the history of science and of the same socio-political context
where it developed, and from a hagiographic enthusiasm that made it a
too easy success story. As some recent researches suggest (eg Boumans
and Dupont-Kieffer, 2011), the now growing rejection toward a
paradigm that was depleted in the 1980s allows us to revisit this
history with new questions. We propose here to read the history of
econometrics in light of the questions concerning the possibility of
a probabilistic economics. The probabilistic economic modeling, seen through
its various dimensions – metaphysical (ontological vs. epistemic
chance), mathematical (the basis of probable calculation),
statistical (frequentist or subjectivist estimate), and pragmatic
(the social effects it produced) – has indeed been paid little
attention by standard historiography.
We will start from two puzzling
questions:
- Historians should explain how a
paradigm supported by a small group of people (the Cowles
Commission), well trained in mathematics, probability and statistics,
was able to convince a global community of economists, who were
totally ignorant or allergic to any consideration of randomness and of
probability calculus, that the solution to economic issues went
through structural and probabilistic modeling. Only a bundle of
reasons related to the specific scientific regimes of WW2 and of the
Cold War can explain this.
- In 1859, the statistical physics of
Maxwell and Boltzmann, as well as the Darwinian theory of the
evolution of species, revolutionized these two disciplines by giving
chance - a chance which was not epistemic, but ontologic (or
objective, as Cournot said) - a constitutive place in their theory.
Following the studies of the Bielefeld group on the probabilistic
revolution, and some other historical works, it seems that the
economic thought did not experience a similar revolution either in
the nineteenth century, nor in the 1930s with the IES.
This paper will remind that the
econometricians' reference to Cournot concerned almost exclusively
his concept of a mathematical economics, and not the idea of
objective chance. On the other hand, as long as randomness and
probability are concerned, a reference to Quetelet and Yule would be
more fit to the econometrics of Tinbergen and Frisch, and of the IES,
and even of the Cowles Commission. The type of chance that presides
over the randomness affecting the economic relations could be
identified with an error or a noise disturbing a deterministic
relationship, rather than with an intrinsic variability of the
concerned actors and phenomena. The project of unifying the
mathematical and statistical approaches that was its creed was
largely hampered by the exclusion of the hazard, which was evident in
the works and correspondence of its founders I. Fisher, C. Gini, R.
Frisch and F. Divisia. We will reassess their positioning with regard
to randomness and probability, and, following Morgan, Le Gall, and
Mirowski, we will revisit the manifesto by Haavelmo (1944) and a text
by Marschak (1948) that are at the heart of the Cowles Commission's
creed in a probabilistic structural econometrics, in order to detect
the justifications, the interpretations, and the limitations
affecting the introduction of randomness in economics.
Frisch's Approach: Econometrics as the Science
of Measurement. Modelling as Intertwining between Theoretical Analysis
and Statistical Investigation
Ariane Dupont-Kieffer (Université Paris-Est, IFSTTAR, DEST)
This paper aims to investigate how Ragnar
Frisch has based econometrics on a specific articulation between
theoretical measurement and empirical measurement. His starting point
is to show that humankind has a constant need to believe in the
existence of regularities ruling the physical world as well as the
social one, and in the fact that the understanding of these
regularities allows human beings to influence their environment. The
knowledge of these regularities is rooted in Frisch’s perspective
on the synthesis of a reductionist reasoning and of a physicalist
approach. This conception of science begins with three requirements:
1) the use of mathematical tools for scientific investigation on the
equation between scientific laws and quantitative laws, 2) the
primacy of measurement procedures in the scientific work and 3) the
need to articulate theoretical measurement and the empirical one.
These three requirements explain the key role played by the model in
his process of scientific discovery. The model is then a “mediator”
according to the analysis developed by Morgan and Morrison (1999)
between economic theory - the corpus developed around the works of
Walras, Marshall, Pareto and mainly Irving Fisher - and the “reality”
as “reflected” in the statistical data, and modeling leads Frisch
to define a specific methodology of experiment.
Schumpeter, Frisch
and Lucas: The Oscillations of Economics while Dealing with Oscillations in the Economies
Francisco
Louçã, ISEG-UECE, Lisboa
The paper compares the
discussions between Joseph Schumpeter and Ragnar Frisch, in the
early thirties, and the discussions introduced by Robert Lucas and
the RBC school, more than fifty years later, on the nature (and
models) of oscillations in the economies.
Typically, economics
addressed these oscillations in the general equilibrium framework,
and the distinction between an impulse system and a propagation
apparatus was instrumental for such purpose. Yet, many
econometricians suspected and discussed this framework, even some of
the more unsuspected lawyers of general equilibrium.
These
discussions highlight early approaches of complexity in economics.
