Opinion Polls: Scientific Basis or Surrogate Campaigning?

What is the mathematical reasoning behind opinion polls in India? Which branch of statistics can give us insights into how vote shares can be converted into seats? For a complex multi-stage sampling process, how are poll agencies coming up with a precise “3% margin of error” estimate?

No one gives us answers to these questions—– because there are no legitimate answers. Psephologists, pollsters, data scientists etc have become the election-babas of India.

Is it surprising then that the opinion polls are manipulated?

According this tweet by Rajdeep Sardesai of CNN-IBN, the same agency did a national poll for different media outlets within a fortnight with different figures.

According to an article by Rukmini Shrinivasan  in Times of India the manipulation of opinion polls(perhaps with good intentions) was never a secret:

No agency wants to admit it but the statistical models that take them from voting preferences to seats include some element of subjectivity. In his early years, CVoter’s Deshmukh candidly admits that he applied his journalistic instincts to the numbers before arriving at a final seats tally. The agency now uses a proprietary algorithm, which he is still continuously fine-tuning , but there remains an element of something intangible. “I won’t call it a qualitative element, but insight gained from the quantitative data over time,” Deshmukh says. “The final figures can be decided by application of some political wisdom, but political wisdom should not dominate over findings collected from voters’ data,” Kumar says. (emphasis mine)

The only way to estimate voter preferences is by collecting a random and representative sample. If that is the case then why would anyone spoil the randomness of data by using their “political wisdom”? Adjusting the sample to demographics is one thing but to blatantly contaminate and manipulate the data under the garb of “political wisdom” or “journalistic instincts” works only in India.

It is not an easy exercise to arrive at seat estimates in a multi-party and representative democratic set up like India and some degree of manipulation is perhaps inevitable.  However, the News Express expose revelation that this manipulation can happen at the behest of some vested interests is very worrying.

Some analysts and journalists are quick to suggest that there should be more transparency in the methodology and the process should be monitored by an independent regulator (like Lokpal??).

No matter how transparent the methodology is, once an agency decides to manipulate the results of an opinion poll they can manipulate the survey data itself and I am not sure how any regulator can prevent this.  The only possible “regulator” for opinion polls is the credibility. If the seat projections of the opinion polls differ from the final results by a significant margin, these agencies should be out of business. However, if you look at the recent Delhi exit polls, C-Voter (Cong-20 and AAP-15), AC Neilsen (Cong-16 and AAP-15), ORG (Cong-20 and AAP-6) are nowhere near the actual results and none of these agencies even bothered to explain the discrepancy. The news channels continue to engage them and there is no incentive to be accurate.

The problem with the manipulation of opinion polls is that it directly interferes with the process of free and fair elections in two ways:

  1. There is something called “bandwagon effect” —- the probability of any individual adopting it[a particular choice] increases with the proportion who have already done so. Yogendra Yadav, based on the surveys of CSDS confirmed that Indian voters tend to back the winning horse. By manipulating the poll results (without any scientific basis) the agencies are creating a false bandwagon effect, thereby giving an edge to a political party.
  2. Election campaigning is regulated by EC with a model code of conduct which gives a level playing field to all the political parties. However, in many states, especially in South India, it is not uncommon for political parties to own news channels and these news channels are becoming centers of surrogate campaigning by repeatedly telecasting the favorable opinion polls.

Both of these are serious issues and while a complete ban on the opinion polls is not a solution, we should be open to the idea of a short blackout period just before the elections.

Also read : My previous post on why opinion polls should not be taken seriously.