ABSTRACT
Measurement is the cornerstone of science and if social science is to justify its name it must aim for standards of measurement which bear comparison with those of natural science. In some cases there is little problem in meeting this requirement. For example, birth rates, life expectancy, and average hospital waiting times are well-defined concepts, which can be directly measured from readily available data. However, there is another class of concepts which are regarded as quantitative and yet which cannot be directly and unambiguously measured. Quality of the environment is such an example. We use the term in social debate in a way that implies it is something which one can have more or less of. If pressed to justify statements of this kind we would point to a whole collection of directly observable quantities such as levels of atmospheric pollutants, water quality, noise levels, contaminants in food, and so on. It thus appears that the term “quality of the environment” is a shorthand for something to which a constellation of other observable variables are assumed to contribute. It is, in short, a collective property of a set of variables. The problem is how to extract, in some sense, and then to combine into a single measure what each variable is contributing. The statistical problem is to provide a theoretical framework within which this can be done.
