Commentary: Socioeconomic deprivation and health and the ecological fallacy

BMJ 1994; 309 doi: (Published 03 December 1994) Cite this as: BMJ 1994;309:1478
  1. Ken MacRae
  1. Reynolds Building, Charing Cross and Westminster Medical School, London W6 8RP, reader in medical statistics.

    Socioeconomic deprivation and health and the ecological fallacy

    The three papers addressing the issue of socioeconomic deprivation and health published in this week's journal all conclude that deprivation is associated with increased mortality. Two of them are ecological studies in which the unit of observation is a geographically defined group.1 2 Such studies are susceptible to the ecological fallacy,3 which was first explained formally by Robinson.4 The advantages and disadvantages of ecological studies have been reviewed in depth by Morgenstern.5

    The problem is that the correlation between two variables when the group is used as the unit of analysis may be quite different from the correlation between those two variables when individual people are used as the unit of analysis. A simple example with two dichotomous variables demonstrates the fallacy (table). When the area is the unit of observation there is an association between exposure and disease, area 2 showing both greater exposure (300/1000) and greater incidence (300/1000) of the disease than area 1 (100/1000 and 100/1000). But when individuals within the areas are the units of observation there is no such association: in each area the same proportions of exposed and non-exposed individuals have the disease (10% in area 1 and 30% in area 2).

    Fortunately, substantial evidence using individuals as the unit of observation does exist to support the conclusion that ecological correlations between socio economic deprivation and health arise from associations among the relevant variables in individuals. One such piece of evidence comes from the paper by Sloggett and Joshi.6 They used data from the longitudinal study of the Office of Population Censuses and Surveys on a 1% sample of the population of England and Wales begun at the time of the 1971 census—that is, census records for the subjects in the longitudinal study were related to the subsequent deaths of these subjects. Sloggett and Joshi found that such socio-economic variables as access to a car, being an owner-occupier, unemployment, and lower social class explained nearly all the relation between the degree of deprivation in an area of residence in 1981 and subsequent premature death before 1990. Of course, these results add support to the Black report7 and to the research on this issue carried out since then (that published up to 1990 being reviewed by Davey Smith et al).8

    Relation between exposure and disease in two areas. Values are numbers of people

    View this table:

    So, those who would seek to criticise or ignore research on socioeconomic deprivation and health using ecological correlation studies in which appropriate indices of deprivation have been used cannot legitimately seek support from the ecological fallacy.


    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.
    6. 6.
    7. 7.
    8. 8.
    View Abstract