The ELDCARE study aims to investigate, at the ecological level, the relationships between socio-economic variables and cancer survival in patients aged 65 years and over. Survival data for patients diagnosed during the period 1985-1989 and followed up to 1994 were provided by 43 European Cancer Registries in 16 countries participating in the EUROCARE 2 project. Relative survival was computed by Hakulinen's methods. Data on socio-economic factors were collected by national statistics offices for the years around 1991. Pearson's correlation was used to study the relationships between cancer survival and socio-economic factors. We selected four groups of variables. The first group included macro-economic variables (such as Gross Domestic Product, GDP; Total Health Expenditure, THE); the second, the main characteristics of national health care systems; the third, demographic factors; and the fourth, variables on labour market organisation. The countries with the largest proportions of elderly populations, in Northern and Western Europe, spent more on health than the less affluent countries of Eastern Europe. GDP was strongly related to THE but a very high variability in Computed Tomography Scanners (CTS) among countries with similar THE was observed. Indeed, those countries with THE around US 1500 dollars per capita had survival rates for breast cancer ranging from 67 to 82%. Cancer survival in elderly patients in Europe was most strongly related to GDP and THE, especially for good prognosis cancers. Survival was strongly correlated with health care technologies, particularly CTS, but not with health employment. Survival was positively correlated with proportion of married elderly people (and negatively with widowed elderly), suggesting a role played by social support in influencing the prognosis of elderly patients. These results highlight how health outcomes in the elderly are a complex phenomenon, not determined only by GDP and THE, but affected by social organisation and life habits as well as economic development conditions.