Listed vs. RDD Sampling

If you are a seasoned researcher, you likely know that even though the inferential statistics we use are based upon the assumption of probability samples, this assumption is routinely violated.�A probability sample is one in which it is possible to calculate the probability that any given person (element) will be selected. Whether conducting business-to-consumer or business-to-business marketing research, the key to being able to project the findings to the greater population is that, wherever possible, any one respondent is as likely to be chosen as any other.

While in some circumstances it may be necessary to set up quota groups and over sample a particular segment of particular importance (profitable customers), within that segment the same probability of being chosen should apply.�Many samples for business-to-business marketing research come from either customer or prospect lists that the sales and marketing department have on file.�Management of these lists can be very tricky, typically needing some level of sampling expertise.�

Where, you might ask, do sample management companies acquire telephone sample for general consumers?�The leading places are:

• Telephone directories

• Real estate information (deed & tax assessor)

• Voter registration

• Magazine subscription

• Surveys responses

These are great sample sources, but they exclude people who don't want to be listed and those that have recently moved, which can account for up to 50 percent of the consumer population. A potential bias is introduced when those listed may act, think and feel differently then those unlisted.

Random Digit Dial (RDD) sampling is an excellent way around the potential bias issues associated with listed sample.�An RDD sample company will generate random 4 digit numbers know as exchanges and will exclude 4 digit blocks that are know not to be assigned -- which is important as most blocks are unassigned.�As part of their business process, they periodically call through parts or all of the randomly produced active blocks, determining which numbers are "live" and delivering them to corporations and marketing research companies for projects requiring RDD sampling.

With over 108 million telephone households in the United States, approximately 30 percent with unlisted numbers, approximately 20 percent on the move annually and approximately 15 percent of the residential numbers in a typical directory disconnected, sample management can be challenging. Better sample management companies can help address representation issues with a variety of random digit list cleaning methodologies, which can go a long way toward delivering survey efficiency and results in fewer dialings, faster completion, and improved interviewer morale.