Dirty Data in Britain and the U.S.A.: The Reliability of "Invariant" Characteristics Reported in Surveys

In: Public opinion quarterly: journal of the American Association for Public Opinion Research, Band 39, Heft 4, S. 493-506

Abstract

With one exception, the reliabilities of ascribed social & background characteristics in the (Mich) Survey Research Center's American Panel Study of 1956-58-60 & the panel study of Political Change in Britain of 1963-64-66-70 were markedly higher than those for achieved characteristics. The variables considered (sex, race, age, occupation of father for both countries, state of birth, state where one grew up, size of place where one grew up, & education for the US) ranged in % reliable from 66-74% (occupation of father, education) to 98-99% (sex & race). Since relationships between variables are reduced by random error, unreliable measurement in survey data implies that the predictive power of achieved characteristics (eg, education) is reduced proportionately more than that of ascribed characteristics (eg, sex, race) & suggests a built-in tendency for policy-oriented research to find problems less malleable than actually may be the case. 5 Tables. AA.