© Copyright - Daniel Courgeau - 2006 - 2012 - All rights reserved - Made by SCENARIO ORIGINAL
The problems linked to the notion of causality in social sciences are intricate.
The counterfactual approach elaborated in the early 1970’s by David Lewis may be defined as:
An event e causally depends on an event c if and only if, if c were not to occur e would not occur.
Its application to statistics and social sciences, largely used in the United States, raises many problems.
First, if we agree that there is no causality without manipulation, the use in demography of variables such as age, and sex, which do not verify the manipulability
criterions, cast doubt on it. For example, we have shown that the age effect on spatial mobility disappears quite entirely when we introduce different individual
characteristics, on which a political action may be possible.
In the second place, its use in statistics leads to a fatalist hypothesis, which may never be justified on the base of empirical data: it considers different potential
responses as predetermined attributes of an individual, waiting only to be uncovered by suitable experimentation, as said Dawid in 2000.
However we may avoid saying, as Herbert Smith in 2003, that we measure at the micro level, but we intervene at some higher level, as this may lead us to incorrect
political actions. In the example developed in the part Paradigms and axioms of this website, shows that a macro analysis may lead, if we want to act on the farmers’
mobility, to a policy prompting this mobility in the areas with a low proportion of farmers, while a multilevel analysis shows that such a measure would be ineffective.
The mechanist approach elaborated in the late 1990’s, but already used as mechanical mechanism in the XVIIth century by Descartes and Leibniz, permits to avoid
this use of unverifiable hypotheses. It may be defined as:
A mechanism for a phenomenon consists of entities and activities organised in such a way that they are responsible for the phenomenon.
Then two events are causally related if and only if there is a mechanism that connects them.
Let us say here to simplify, following Robert Franck in 2002, that a first step to put in evidence such a mechanism, is the systematic observation of a given social
phenomenon, as in demography since 350 years. The second step is to infer, from this observation, the formal structure that contributes to produce it: demography
has not yet attained this step. The third step will then study the social mechanism which generates the observed properties: some of these mechanisms had already
been studied in demography, but the failure to recognize these functions make such a study incomplete and above all not generally applicable. The last step is to
verify if the combination of functions which has been found permits to guide the causal analysis of the process and to increase its pertinence.
The points of view given by different paradigms in social sciences already permit a progress in this domain that it will be useful in a second step to axiomatize.
Our work in this domain is just beginning and will have to be pursued in the future.
References:
Lewis, D. 1973. Counterfactuals. Blackwells: Oxford.
Smith, H., 2003. Some thoughts on causation as it relates to demography and population studies. Population and Development Review, 29(3), pp. 459-469.
The counterfactual approach elaborated in the early 1970’s by David Lewis may be defined as:
An event e causally depends on an event c if and only if, if c were not to occur e would not occur.
Its application to statistics and social sciences, largely used in the United States, raises many problems.
First, if we agree that there is no causality without manipulation, the use in demography of variables such as age, and sex, which do not verify the manipulability
criterions, cast doubt on it. For example, we have shown that the age effect on spatial mobility disappears quite entirely when we introduce different individual
characteristics, on which a political action may be possible.
In the second place, its use in statistics leads to a fatalist hypothesis, which may never be justified on the base of empirical data: it considers different potential
responses as predetermined attributes of an individual, waiting only to be uncovered by suitable experimentation, as said Dawid in 2000.
However we may avoid saying, as Herbert Smith in 2003, that we measure at the micro level, but we intervene at some higher level, as this may lead us to incorrect
political actions. In the example developed in the part Paradigms and axioms of this website, shows that a macro analysis may lead, if we want to act on the farmers’
mobility, to a policy prompting this mobility in the areas with a low proportion of farmers, while a multilevel analysis shows that such a measure would be ineffective.
The mechanist approach elaborated in the late 1990’s, but already used as mechanical mechanism in the XVIIth century by Descartes and Leibniz, permits to avoid
this use of unverifiable hypotheses. It may be defined as:
A mechanism for a phenomenon consists of entities and activities organised in such a way that they are responsible for the phenomenon.
Then two events are causally related if and only if there is a mechanism that connects them.
Let us say here to simplify, following Robert Franck in 2002, that a first step to put in evidence such a mechanism, is the systematic observation of a given social
phenomenon, as in demography since 350 years. The second step is to infer, from this observation, the formal structure that contributes to produce it: demography
has not yet attained this step. The third step will then study the social mechanism which generates the observed properties: some of these mechanisms had already
been studied in demography, but the failure to recognize these functions make such a study incomplete and above all not generally applicable. The last step is to
verify if the combination of functions which has been found permits to guide the causal analysis of the process and to increase its pertinence.
The points of view given by different paradigms in social sciences already permit a progress in this domain that it will be useful in a second step to axiomatize.
Our work in this domain is just beginning and will have to be pursued in the future.
References:
Lewis, D. 1973. Counterfactuals. Blackwells: Oxford.
Smith, H., 2003. Some thoughts on causation as it relates to demography and population studies. Population and Development Review, 29(3), pp. 459-469.
© Copyright - Daniel Courgeau - 2006 - 2012 - Tous droits réservés - Réalisation SCENARIO ORIGINAL
© Copyright - Daniel Courgeau - 2006 - 2013 - Tous droits réservés - Réalisation SCENARIO ORIGINAL
© Copyright - Daniel Courgeau - 2006 - 2021 - Tous droits réservés - Réalisation SCENARIO ORIGINAL
