"Ecological" Inference

The Use of Aggregate Data to Study Individuals

Datos Bibliográficos

ID3765320
AutoresW Phillips Shively (Yale University, autor de correspondencia)
Año1969
Volumen63
Número4
Páginas1183-1196
Fecha de publicación1969-12-01
Peer ReviewedSí
Open AccessSí
TipoARTICLE
RevistaAmerican Political Science Review (JOURNAL)
Identificadores de la revistaISSN: 0003-0554 • E-ISSN: 1537-5943
EditorialCambridge University Press (CUP) (PUBLISHER)
DOI10.2307/1955079
OpenAlexW4249547423
IdiomaEN
Citas recibidas40
Referencias citadas9

Because they are inexpensive and easy to obtain, because they may be available under circumstances in which survey data are unavailable, and because they eliminate many of the measurement problems of survey research, data on geographic units such as counties or census tracts are often used by political scientists to measure individual behavior. This has involved us in the long-standing problem of inferring individual-level relationships from aggregate data, which was first raised by W. S. Robinson in the early nineteen fifties.In this paper, I shall first discuss the problem raised by Robinson. I shall then review three partial solutions to the problem-the Duncan-Davis method of setting limits, Blalock's version of ecological regression, and Goodman's version of ecological regression. Finally, I shall propose some ways in which Goodman's method may be used so as to reduce the problem of bias in its estimates, and make it a more reasonable tool for reserch.Our difficulty, as Robinson showed, is that we cannot necessarily infer the correlation between variables, taking people as the unit of analysis, on the basis of correlations between the same variables based on groups of people as units. For example, the "ecological" correlation between per cent black and per cent illiterate is +0.946, whereas the correlation between color and illiteracy among individuals is only+0.203

Aggregate (composite · Aggregate data · Biology · Data mining · Econometrics · Functional illiteracy · Geography · Inference · Mathematics education · Measure (data warehouse · Political science · Regression · Regression analysis · Statistics · Unit (ring theory · Artificial Intelligence · Census and Population Estimation · Computer Science · Ecology · Land Use and Ecosystem Services · Mathematics · Psychology