A method for improving the results of primary investigation by including the reasons for non-response in sample design and weighting (German)
For years now, one has been able to note ever decreasing rates of exhaustiveness in random surveys. The decrease in exhaustiveness is mainly a phenomenon of the respondents, caused by continually increasing inclination towards non-response. Up to now, the methods known for approaching this problem usually involve considerable increases in expenditures and a prolongation of dead-line, and can therefore be used but to a limited degree. Our proposal comprises a means of eliminating the decrease of representativeness due to non-responses without such an increase in expenditures. The method is therefore based on the inclusion of behavioural factors in factorial sample adjustment. In this way it can overcome the difficulty generally found in the usual demographic weightings: the latter must abstract the different behaviour of identical demographic groups, whereas our method uses different behaviours as criteria for weighting. The improvements achievable using this method are demonstrated by results of surveys and model calculations.
- This could also be of interest