Unpacking Composition and Context

The difficulties of using ecological data have tended to make researchers cautious about interpreting the health effects associated with specific places – what we might term ‘place effects'. The ecological fallacy involves making inferences about relationships existing at the individual level on the basis of observations which are made at the aggregate level. A further consequence is that when we are dealing with health outcomes and potential explanatory factors at the aggregate level, much of the information which we are seeing is actually a proxy for missing individual-level data. Mitchell et al. (2000) make a useful distinction between those area variables which are summaries of the individual characteristics, such as percentage unemployment, and those which are genuinely features of the place and not reducible to the individual level, such as a changing industrial structure.

Numerous empirical studies have sought to prove or disprove the effects of place over and above those that can be ascribed to the aggregated characteristics of the people who live there. Macintyre et al. (2002) conclude that there is usually evidence of some place effect even in studies which have attempted to control for individual-level characteristics: generally speaking, both place and individual characteristics matter, and individual characteristics seem to matter most.

The distinction between these alternative explanations has become known as the ‘composition versus context' debate. Composition relates to the characteristics of the individuals who live there and context to the characteristics of the location itself. Macintyre et al. (2002) suggest that this distinction may still not be entirely clear but that ‘collective' aspects of local areas, through which individuals living in a particular locality share behaviour patterns and social norms are also an important element in understanding the place context.

print friendly pdf

“ For example, children in deprived areas may not play in the open air because their families do not have gardens or the resources to take them to play parks (a compositional resource based explanation); because too few public play parks are provided, and there are no good public transport links to those that do exist (a contextual resource based explanation); or because within the prevailing local culture play is not seen as something which is important to children, or it is not considered desirable or safe for children to play with strangers in public places (a collective explanation).” (Macintyre et al., 2002: 130).

Much work on attempting to measure contextual effects of place has been driven by data availability rather than any clear theoretical model of how contextual effects might operate. Mitchell et al. (2002) present an empirical GIS study aimed at investigating some of these effects using British mortality data while Macintyre et al. (2003) present an empirical study focused on housing tenure and the home and area environments within which people live, relating this to morbidity and mortality. These papers present helpful case studies of the need and complexity of unpacking composition and context issues in health GIS.