RDD METHODOLOGY
 

- All 100-blocks (working blocks) that fall within a specified geographic area and with at least one known residential telephone number are identified and become the sampling frame of working blocks. Improving upon other methods which often require at least two known residential phone numbers, we include in the sampling frame exchanges and working blocks that have one or more known residential telephone numbers. This has the advantage of eliminating a potential source of bias that might be introduced by omitting 100-blocks with a single known residential telephone number. Moreover, this has virtually no practical impact on sample efficiency because less than 0.1% (1 out of 1,000) of all numbers are contained in 100-blocks with only a single known residential telephone number.

- All eligible exchanges in the geographic frame are identified and compared with the Master Exchange Database. Any exchange which is not identified as operational is dropped from the sampling frame. This eliminates exchanges which have been incorrectly included in the frame due to errors in phone number assignment.

- Corresponding time zone codes are appended to the eligible exchanges and working blocks in the geographic frame based on data contained in Master Exchange Database. Samples that cross more than one time zone may then be sorted by time zone to assist in call management and interview scheduling.

- Eligible working blocks are weighted according to the number of known residential telephones contained within each. Weights are cumulated across eligible working blocks.

- Working blocks are sampled in direct proportion to their weights. A working block with 80 known telephone numbers will have twice the probability of being selected as a working block with 40 known telephone numbers. This approach provides significant improvement in sample and dialing efficiency by increasing the probability of selection of those working blocks that are more densely packed with operating residential telephone numbers and decreasing the probability of selection of those that are more sparsely assigned with working numbers.

- Within each selected working block, one or more random numbers between 00 and 99 are generated and assigned to the working block. The quantity of random numbers created for each working block is in direct proportion to its weight in relation to the combined weights of all the eligible working blocks in the sample frame.

- The sample is pulled proportionately to either the population or household size of the states (for national samples) or counties (for single state samples) within the specified geographic frame.

- In order to reduce the probability of dialing business phone numbers in the sample, any number ending with ‘00’ is eliminated from the sampling frame.

- We do no other "enhancements," such as screening (predialing) all the numbers to detect ‘not in service’ recorded messages. This step would add significantly to our sample prices while creating many nuisance calls to residences where telephones often ring one time and are then automatically disconnected. Nor do we compare the sampled phone numbers with a list of business phone numbers and dedupe the matches. This can introduce bias by eliminating residences with home-based businesses sharing a common home/business telephone number. What’s more, business number purging has been shown to have negligible impact on dialing efficiency.

In sum, ASI supplies a highly representative RDD sample, including those enhancements that increase dialing efficiency (such as weighting working blocks) while avoiding those enhancements which are intrusive or may introduce sampling bias.




Welcome
Firm Profile
Online Panel Description
Online Consumer Panel
Online Business Panel
Consumer Phone
Business Phone
Request a Quote
Contact Us
e-mail me

|Welcome| |Firm Profile| |Online Panel Description| |Online Consumer Panel| |Online Business Panel| |Consumer Phone| |Business Phone| |Request a Quote| |Contact Us|