DigiPoll’s RDD Methodology

 

The core principle of Random Digit Dialling Sampling (RDD) method is targeting Exchange Information Numbers (EIN). Each EIN is attached to a certain area (example from the Yellow Pages (below). The last four numbers are randomized:

Accurate information with regards to use exact geographical spread of EIN is proprietary to Telecom. DigiPoll has successfully tested the ranges and is now able to identify accurately the areas where such interchanges operate.

This system allows very accurate representation of the area surveyed. The reason for this is the application of the telephone exchange area system by which the calls are scattered through the entire area and the responses reflect the distribution and densities in each area, havoiding the need for complex weighting systems based on wards to achieve a geographically representative sample.

There are more exchange areas and connections in densely populated suburbs. Similarly, the newly developed areas are adding more connections to existing exchanges and many numbers are turned back as “not in service” numbers .

The same is the case with sparsely populated areas. Overall this method helps to ensure that the sample reflects the most current distribution of the population and therefore the socio-economic composition which is geographically dependent.

 

 RDD versus Clustered Randomisation

The origin of clustered randomisation was the need to create samples for populations living in large geographical areas. Nation or state wide surveys were typical to that. The cluster randomisation method involves a selection of localities which represent a metro / non-metro population share as measured in a general census. According to this, some neighbourhoods in rural communities are targeted for canvassing. However the major drawback for cluster randomisation has been deficiency of the sample in terms of the representation of a wide range of socio-economic grouping. Although cluster randomisation is justified for face-to-face surveys, it is definitely impractical for telephone surveys where geographic scope is not a limitation for canvassing.

When access to telephones has become common, the Random Digit Dialing (RDD) was offered as a solution. RDD means the creation of Random numbers on the basis of the first 3 digits of a telephone number which are part of a Telephone Exchange Areas which exist on a common telephone cable in a given exchange area. Theoretically, any household with a telephone has an equal chance of being selected for a sample. However in the real world, competing telephone companies have been reluctant to provide current and comprehensive lists of the Telephone Exchange Areas. As a result, survey companies have been using a method, which can be defined as clustered RDD. This clustered RDD follows the old cluster randomisation method by which RDD was applied to only selected known exchange areas. Other survey researchers elected to create randomisation from electronic white pages but the electronic white pages being used, are outdated.

To the best of its knowledge, DigiPoll has been offering a unique sampling solution. DigiPoll applies an algorithm, which first identifies potential Telephone Exchange Areas and secondly, determines the probability of usage within that exchange area. The end result of this sampling method is, when repeated many times over, the actual population of the country will be reflected. Our Random Digit Dialing (RDD) sampling methodology helps reflect a good socio-economic profile of the community.

The reason for this is the application of the telephone exchange area system by which the calls are scattered through the entire area and the responses reflect the distribution and densities in each area. There are more exchange areas and connections in densely populated suburbs. Similarly the newly developed areas are adding more connections to existing exchanges and many numbers are turned back as not in service numbers. Same is the case with sparsely populated areas. Overall this method helps ensure that the sample reflects the most current distribution of the population and therefore the socio-economic composition, which is geographically dependent.

To achieve a high response rate, DigiPoll will create a database of numbers that we believe will be adequate to get required responses. These numbers will be churned through over and over until the sample size is achieved. The higher call-back attempts ensure better representation of the population.

Time and again this methodology has been proved to be accurate and successful primarily through various pre-election polls and the findings of numerous national surveys conducted on a variety of socio-economic issues.

The difference between clustered RDD and scattered RDD can be summarised with consistency of omission of localities as compared to the random omission of localities in scattered RDD. The random inclusion or exclusion of localities in a sample follows the principle of pure random sampling, whereas cluster sampling has permanent bias. Even during the 2011 general election DigiPoll was the only major pollster to predict the rise of NZ First and that they will make it to the parliament. This success has been repeated since the first poll done in 1996. Herald-DigiPoll, Marae-DigiPoll and Te Karere-DigiPoll are now well-established brands, which people trust and respect in the field of public opinion polling.