IS SUBJECTIVE WELL-BEING A PREDICTOR OF NONRESPONSE IN BROAD POPULATION SURVEYS

In connection with a medical screening operation for hypertension and diabetes among all adults (20+) in a medium-sized Norwegian county, all screenees were required to fill in a questionnaire containing questions about subjective well-being (overall life satisfaction and self-assessed health). Since screenees in practice constituted a ''captive audience'', questionnaire data are available for an unusually large proportion (88%) of the eligible population (N = 71 896). Information about socioeconomic background could be obtained for all screenees through record-linking. A second questionnaire was handed to all screenees as they left the screening site. The response rate (by mail) for this second questionnaire was quite ''normal'', appr. 80%. Working on the assumption that nonresponders and/or late responders for the second questionnaire are representative of persons not responding in ordinary broad population surveys, nonresponse biases in the means, variances and covariances of our measures (in the first questionnaire and the outside records) could be investigated. Bivariate results showed a slight overrepresentation of screenees with poor self-assessed health among those who returned the second questionnaire promptly by mail. However, multivariate analyses controlling for the age composition of person with less favorable self-assessed health reversed this pattern completely so that the only direct effect from subjective well-being remaining is a very slight tendency for persons in poor self-assessed health to fail to cooperate.