Conjecture & surmise in imagined realities

By the time you start reading this column, the actual final results of the Lok Sabha elections may not yet be fully out. Exit polls are loud and clear (perhaps more loud than clear) in terms of which way people have chosen.  All exit poll outcomes (barring two) have returned a sweeping runaway victory for the NDA and a rout for the UPA. 

The exit poll results have led to extensive airing of firm views on the real deal with India’s politics. Even a seasoned poll pundit like Yogendra Yadav has said the Congress party must die for a change in the political situation. Despite being a politician himself, he is usually dispassionate enough to keep his aspirations and hope out of his assessment of the reality. However, are exit polls adequate for such firm conclusions to be drawn? 

Response to questions put to a sample of voters, forms the basis of drawing inferences and statistical conclusions as to the outcome of elections. While this exercise may feed the insatiable curiosity of the human mind, even with the best standards of integrity in the conduct of exit polls, astrology can remain a close competitor.

There is many a parallel in the real business world, and in particular, in business regulation, where similar extreme conclusions get drawn and policy choices are made, based on conjecture and surmise akin to exit polls. While exit polls results may at worst lead to some shooting of blood pressure in television studios, in the real world, expensive choices get made despite similar lacunae. Just two facets are worth noticing.

First, our ballot is a “secret ballot” for a reason. The law guarantees the secrecy of the vote cast so that the citizen casting her vote does so without any fear of reprisal. A citizen who is then asked to confess who she voted for, can only be expected to volunteer and waive the right to secrecy. Of course, a diehard loyalist would proudly declare her choice.  Indeed, many could answer truthfully, only because they are asked — computer hackers will tell you that the easiest way to get someone’s password is to simply ask for it. But those strong-minded or fearful about their secrecy being violated are bound to recoil. Often, speculative consumer choices are sought to be discerned by competition law regulators, say, when they have to consider if they should approve a merger of owners of two competing brands. They are prone to making the same mistakes as placing much in store by exit polls.

Second, the mere peppering of statistical tools that can give a semblance of science would not result in the outcome being more than conjecture or surmise. There could of course emerge realities in election outcomes that match the exit poll outcomes — but these would be more in the realm of being fortuitousness than in the realm of a scientific cause-effect relationship being established. Factors other than truthfulness too matter — for example, how one selects a sample, what size of sample is selected, how representative the sample is, and so on. Just consider how deeply controversial the selection of samples are in computation of television rating points, and of course how violent disagreements are over how to assess which news publication is read the most. The cause-effect confusion that boggles the mind, too comes up often in market abuse investigation — a typical vexed issue is determining whether a person who sold securities in large numbers was cutting losses or whether it was her sales that led to an unfair fall in prices.  

By the time you finish reading this column, the actual final results may be out. If the exit polls are proven right, remember it is a random match despite lacunae. If they are proven wrong, you know the randomness did not work out a match. Publication of exit poll results until completion of polling is banned for this very reason —their absence of reliability is inversely proportional their impact. 

What is vital is to remember is that real strategic decisions on whether to dump a product (winding up the Congress party), or hold someone guilty (say, of market manipulation) or discharge someone of suspicion (say, approval of a large merger), has to be founded on firmer reality and not solely on extrapolation from responses to sample-based surveys.  It is vital to own statistical tools and not let these tools own you. 
/> The author is an advocate and independent counsel. Tweets @SomasekharS