Statistics In The Social Sciences A Mixed Blessing?

This paper argues that the box of tools amassed by econometricians has much to offer, not only to economics but also to other social sciences. At the same time it argues for a reduced methodological role for statistical analysis. It scrutinises the scope for a proper 'estimation' of coefficients, and the more fundamental issue of the ability of statistical analysis to function as the 'tester' of theories par excellence. The analysis proceeds at two levels: (i) in terms of the conditions laid down by econometric theory itself, and by critiques of econometrics in the econometric literature, and (ii) in terms of the recent critique of logical positivism among philosophers of science. The paper concludes that the scope for genuine estimation of coefficients is often severely limited by the unavoidable omission of relevant variables which are not orthogonal to the included regressors. Such 'pernicious' omissions may result from either a lack of data or inherently non-quantifiable factors. This is particularly so where inter-disciplinary boundaries impose arbitrary limits on the variables considered.