- sample selection bias
- Non-random selection is both a source of bias in empirical research and a fundamental aspect of many social processes. When observations in social research are selected so that they are not independent of the outcome variables in a study, sample selection bias (sometimes labelled ‘selection effects’) leads to biased inferences about social processes. For example, information on wages is only available for people in employment, but working women are a non-random sample of all adult women: they have a larger investment in educational qualifications and higher potential earnings than do non-working women, so the decision to do paid work is not independent of earning potential. Sociologists have only recently begun to recognize the contaminating influence-as illustrated by, ‘An Introduction to Sample Selection Bias in Sociological Data’, American Sociological Review (1983) and, Making it Count (1985). Economists have been more alert to the impact of sample selection bias in distorting research results, for example from regression analyses, and have developed various techniques which are claimed to reduce the error. One of the most popular is Heckman's two-stage estimator, but this is not robust, and relies on strong assumptions about the way that selection occurs. Policy researchers have also developed models for assessing policy impacts and programme effects in experimental and non-experimental research which addresses the issue of selection effects (see, ‘Models for Sample Selection Bias’, Annual Review of Sociology, 1992).
Dictionary of sociology. 2013.