index

index A quantitative social, economic, or political measure, often a weighted combination of a number of selected individual indicators for the domain of interest. One example is the Federal Bureau of Investigation's uniform crime index for the USA. Many countries have some kind of retail prices index which measures the rate of inflation in consumer goods. Indexes usually employ standardization to facilitate comparisons, for example showing the initial year of the time-series as base 100, or using the national figure as base 100 for presenting sub-national figures—as illustrated by standard mortality rates.

While these and many other simple indices are useful to sociologists interested in studying trends via longitudinal analysis, other more complex indices are of dubious utility, including (for example) the many available indices of deprivation and poverty. In the first place, the technical problems of combining multiple indicators of (say) deprivation into a meaningful single index are considerable. To mention only the most obvious, indicators may be closely correlated, although reflecting quite different processes; weighting, ranking, and collapsing of indicators is inevitably contentious, since these can all be done in many different ways (involving expert opinion, factor analysis, cluster analysis, and so on); and, finally, the causal analysis is undermined by the ecological problems (see ECOLOGICAL FALLACY) inherent in the resulting index (most clearly in the attempt to explain individual behaviour by reference to aggregated, usually district-level, data). In addition, many different indices have been created to cover the degree, extent, and intensity of deprivation, so that problems of interpreting the scores are further aggravated. The problems then involved in standardizing indicators across (say) countries simply underline the technical difficulties of creating meaningful composite indices.

Moreover, the availability of a growing range of sophisticated techniques for multivariate analysis renders these sort of composite indices increasingly obsolete. If relevant data are collected about (for example) unemployment, values, ethnicity, income, size and type of family, welfare dependency, and such like, from a sufficiently large sample, then the causal relationships between these characteristics can properly be treated as an empirical issue via the family of regression and other statistical techniques for causal modelling. The connections between structural location, behaviour, and beliefs can then be a matter for empirical investigation, and the researcher no longer needs to decide—a priori—how many of these attributes should be compounded before some putative condition of ‘deprivation’ is established.

Of course, politicians like to see league tables which rank local areas or regions in terms of a single index score purporting to reflect the degree of deprivation (or whatever), because such tables seem to provide a firm basis for public spending (and one which is easily understood). Sociologically dubious indices continue, therefore, to be a prominent feature of much policy research (see the report by UK Department of the Environment, 1991 Deprivation Index: A Review of Approaches and a Matrix of Results, 1995). However, whether they are good social science is quite another matter, and it is hard to see how such exercises help illuminate those causal relationships which constitute the primary interest of sociological research.