Re-evaluating Working Exchange Design of RDD Sampling

Abstract

Working exchange design of sampling for RDD (random digit dialing) telephone survey uses a frame consisting of all possible telephone numbers in all the working exchanges and draws a telephone number sample in a single stage random sampling. This type of sample was believed to be practically useless, as the residential hit rate of such samples are very low.
In RDD sampling of recent years, however, it has become standard practice to apply computer assisted screening of non-working numbers to telephone number samples. Unfortunately, we can hardly find any papers which take into consideration the effect of screening of non-working numbers in evaluating sampling designs. If samples drawn with working exchange design are executed with computer assisted screening, their residential hit rate can reach a practical executable level. The design should be re-evaluated since it has become suitable for practical use.
The list-assisted design, which is recently used frequently because of its high household hit rate, has a problem of truncating telephone households in banks which contains only unlisted residential numbers. This paper tries to demonstrate that the working exchange design can attain a residential hit rate which is almost comparable with that of list-assisted design, as well as the full coverage of telephone household population.