So even before all the candidates were nominated, submitted nominations were reviewed/accepted/rejected the Clairvoyants Pollsters of CNN-IBN would like us to believe they can show us the light, never mind all previous attempts that lead us into the ditch of electoral outcomes.

Yesterday Offstumped questioned what was new about CNN-IBN’s methodology, well today we take a shot at dissecting that methodology which is described here.

So first off CNN-IBN throws a 3 letter word at us -PPS to explain why its selection of Assembly Constituencies at random is representative.

So what is PPS ?

This paper from a South African University has a good primer on Sampling and Sampling Design. The bottomline is when CNN-IBN picked the 75 assembly seats for the survey from a total of 224 it did so by giving smaller constituencies a higher weightage of being selected when compared to bigger constituencies so that when ultimately sampled there is no bias based on size on who gets selected to participate in the survey.

So you have 75 constituencies big and small, now what ?

CNN-IBN selected 4 polling booths from each using a different technique called “systematic random sampling”, never mind the oxymoron that is a technical term. The same paper above explains this technique as well. In summary its just a way of picking x number of people at fixed intervals starting with a randomly determined start point. There is a problem with this approach which the paper explains as most lists may have a bias in which they are organized.

So the 4 polling booths picked by CNN-IBN may not be as random as we imagine them to be.

Now to pick the actual participants in the survey the same technique was used to selected 30 voters from the updated voter lists after delimitation. Once again as explained above there maybe a bias in this selection as the technique suffers from any bias the list may have in terms of how entries are organized.

CNN-IBN further says that of the 9000 they picked based on the above only 5124 responded. Which means that close to 50% of the randomly picked participants did not participate in the survey.

So what we have here is a survey with only around 50% of a representative sample set with all its flaws based on the techniques described above.

Now we all know in India that there is usually a bias to who participates in an Opinion Survey and who doesnt. So despite CNN-IBN’s tall claims that the 5124 sample set matches the demographic in the state in terms of gender and religion and caste it by no means indicates their socio-economic status.

So there is no basis for CNN-IBN to have even come out with this survey in the first place for its sample set is defective and suffers a bias which is obvious from the non-participation in the survey.

Offstumped Bottomline: Unless CNN-IBN can prove that the sample sub-set that did not participate in the survey is similar to the sub-set that did it on the basis of socio-economic status, ideological bias and past voting history, there is no way it can claim on the basis of gender, religion and caste that its sample set is representative and that its prognostication for the outcome of polls in Karnataka is credible.  CNN-IBN must apologize to the people of Karnataka and withdraw its opinion poll.

Offstumped is written by Yossarin from the Indian National Interest

Be Sociable, Share!