Positivism and Sociology (RLE Social Theory) by Peter Halfpenny

Author:Peter Halfpenny [Halfpenny, Peter]
Language: eng
Format: epub
Tags: Social Science, Sociology, General
ISBN: 9781317651383
Google: 6P1TBAAAQBAJ
Publisher: Routledge
Published: 2014-08-21T05:57:47+00:00

CAUSE IN SOCIOLOGY

There is some tendency among sociologists to adopt a heroic Humean stance, for example, when they simply collect data and correlate variables, and perhaps test for statistical significance, but pay less explicit attention to substantive significance. In such cases, where no assessment is made of why some variables are correlated and others are not, there does seem to be an assumption that all (or all statistically significant) correlations are causal laws. Less heroic, but still Humean, are those who insist that only universal regularities have the correct form to be acceptable as laws. They seek to render correlation coefficients near unity by adding initial conditions and refining the measures of the variables under consideration (as described by Boudon, 1971, ch. 4). In effect, they restrict the range of their regularities and remove measurement errors in an attempt to achieve universality.

In contrast, much of the sociological literature concerned with avoiding spurious correlation (Lazarsfeld, 1955) and with inferring causality from correlation (Blalock, 1964) through ex post facto statistical analyses is predicated upon the assumption that social laws, if they are to feature in adequate explanations, have to express causal relations that are stronger than mere covariation, even universal covariation. Yet the conception of cause adopted often remains Humean. For example, in causal modelling or path analysis (which are, effectively, generalisations of Lazarsfeld’s techniques of elaborating statistical relationships to reveal spuriousness, and so on: H. A. Simon, 1954), the existence (or strength) of regularities is identified by using multivariate statistics to obtain measures of correlation between every possible pair of variables in the set under consideration. The values of the correlation measures are then used to choose between alternative causal models, that is, between alternative sets of assumptions (or theories) about which network of causes was responsible for the data. The overall aim is to delete from the network of all possible causal connections between the variables under consideration those that do not exist (or that do not exist above a chosen limit of strength). But then the problem facing all Humeans arises: statistical operations on observed data isolate regularities but do not distinguish between causal and non-causal conjunctions. Confronted by this problem, causal modellers typically turn to the sort of additional criteria invoked by Humeans to supplement regularity in their analyses of causality. One such criterion is causal priority interpreted as temporal order: some variables, it is argued, naturally occur prior to others and so can cause them but cannot be caused by them, this being the sociologists’ equivalent of Hume’s addition of sequence to constant conjunctions. For example, education measured in years of schooling comes naturally, it is said, before occupational status measured in terms of level of earnings and so, if the two are correlated, the former causes the latter but not vice versa. Alternatively, causal priority is sometimes interpreted as the theoretical order among the variables. It is argued that there is some theoretical reason, perhaps established in earlier research, for explicitly including or excluding some variables

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.