Abstract:

We have devised and developed a new method for creating random digit dialling samples for telephone surveys in our agency. The method itself, which we have named ‘Number Propagation’, is very simple. We take the phone numbers of respondents from our face to face omnibus survey and add and subtract one (and two and three) from each number. This generates a new set of numbers which we use for telephone surveys. The idea of generating random numbers by simply adding one to each of a set of known numbers is not new, but we believe that our method is set apart by the very high quality of the original face to face omnibus sample. It is also distinguished by the amount of information we know about the original numbers (we refer to them as seed numbers). We have explored the relationships between the characteristics of the seed and generated numbers in terms of the households we find at the end of the telephone. These relationships have allowed us to produce deliberately skewed samples for some surveys. The need for random dialling is brought about by the increasing proportions of telephone subscribers who are not listed in directories. This has been an issue in the United States for a little longer than elsewhere, so we begin our paper by looking at the development of random dialling systems there. Recent academic work in the UK has uncovered difficulties in exporting this experience and we believe that our system provides a real alternative. At the time of writing our technique has been in use for over a year providing samples for our telephone omnibus survey. We are able therefore to present an analysis of calls to nearly 150 numbers, which, together with a brief examination of fieldwork productivity, makes us confident that ‘Number Propagation’ really does work.