Predictions, predilections, and Indian general elections

Predictions are awfully difficult, even with the reassuring crunch of numbers! And yet, they are our fatal predilection, whether the challenge is for our pleasure or for that of others. Be it the stock market, astrology, or gambling, our insatiable curiosity for the mysterious and unknown is probably what propels such abandonment.

Take the ongoing Lok Sabha polls, going on for perhaps too long now, and the concomitant barrage of prophecies from TV channels, from YouTube “influenczars” and from downright bored individuals, all of whom shoot off their incendiary numbers into the tinder of social media. These soothsayers speak from relative ignorance and mostly from a fascination for a particular party, be they pidis or bhakts.They speak with the confidence that they can actually fool all the people all the time, with Abraham Lincoln being the ultimate fool, poor man. What is most horrendous is that all of them have declared random numbers like 180, 200, 250, 272, 303, 320, 362, 370, and the ace of spades, 400, long before all the votes have come in. This is why I liken election predictions in India to gambling. It is matka guessing in a satta bazar. Parrot talk, or tarot talk, it leads to no concrete outcome. It does not affect the choices made by the voters, choices one suspects were cast in stone, long before the elections were even announced. And yet, these delirious djinns persist, as their minds are made up too. The happy side to this sad fantasy is that it is time-bound, and we can all reclaim our normal lives on June 4.

Lighter comments aside, there is a more serious side to prediction, based on the assumption that future events are influenced to some or a greater extent by what has happened before. History is our greatest teacher. The art and science of good prediction lie in identifying which past (or present) event will likely influence an event to come. Let us take the Indian monsoon, an annual event of major economic importance to our nation, and one that has just begun this year. It would help the government enormously if it knew a few months in advance if the monsoon were to be normal, below average, or will punch above its weight.

At present, the Indian monsoon is predicted using a combination of atmospheric models, sea surface temperature analysis, historical data, and satellite observations. Key factors include the El Niño-Southern Oscillation, the Indian Ocean Dipole, and wind patterns. Advanced computational models simulate these variables to forecast rainfall and monsoon onset accurately. However, there are totally different methods of monsoon prediction that would not be considered ‘scientific’ by western definition, but are yet chillingly accurate. Traditionally, some Indian communities predict the monsoon by observing the nesting behaviour of certain birds, like the Indian weaver bird (Baya Weaver). It is believed that if these birds build their nests on higher branches, a heavier monsoon will ensue. Conversely, lower nests indicate lighter rainfall. This method relies on the birds’ instinctive response to environmental cues, and seems to be largely reliable. Nature lovers might like to correlate this occurrence with the Japanese kō or 72 micro-seasons, for further poetic justification.

Prediction is a key aspect of science, and the smaller the number of known events, phenomena, and parameters used to predict a large number of future events accurately, the better the value of the prediction. At the risk of introducing an equation in this column, one can say the value of a prediction, V = (N/n) – 1, where one is able to correctly predict N outcomes with n parameters. According to this equation, the value of a prediction is zero if one needs n parameters to predict N outcomes.

Let’s apply this equation to the present Lok Sabha election. We could say that N = 543, in other words, we want to correctly predict the results from 543 constituencies. How many parameters would we need, or what is the value of n, for a good prediction? What our present-day pollsters are doing, in trying to improve the accuracy of their guesses, is to increase the number of parameters to absurd levels. They ask whether all the Muslim women in Kishanganj will vote like their menfolk, whether the hot summer days in Ramagundam will harm voting, or whether all the Gounders in Kovai will stay loyal to the AIADMK. The problem with such micro-parameters is that they cannot be quantified properly. Informal surveys on the ground may not be the answer because all respondents may not answer truthfully. We see ludicrous statements from politicians—some say that low turnouts hurt the BJP, while others say that low turnouts hurt the Congress. Others say that a low or high turnout may help or hurt a party, depending on the state. Still others say that whatever happens, Congress is winning. The problem with any or all of this is that none of it can be either proven or disproven—this makes it non-scientific. The reader will appreciate that as n increases with all these unprovable parameters, the value of the prediction quickly goes to zero. It is easier to say that there is only one parameter in this election, and that is the prime minister himself. Applying the equation, the value of my prediction is (543/1) – 1 = 542, which in practical terms means a sizable majority for the BJP. I think I’m doing pretty well for myself with this prediction. June 4 will tell.

What I would have liked to see, and what seems missing, is a more accurate assessment of ‘swing’ constituencies in bellwether states. No expert needs to predict that the BJP will win in Gandhinagar or that it will lose in South Kolkata. On the other hand, if the BJP were to win in Thiruvananthapuram, Bengaluru Rural, Hyderabad, or Central Chennai, this would signify that winds of change are blowing across the country. The very high voter turnout in Baramulla is another pointer. An ability to extrapolate from this to an accurate prediction about the Kashmir Valley in particular and the country in general, is what is missing. Once again, what I would like to see on June 4 is how soon any particular TV channel is able to call the election, say by 9, 10, or 11 am? The sooner it is able to do this, the better its reports will be and the less fluff and bluff.

Election predictions are notoriously hard due to the changing nature of the parameters, something we do not usually see in science. The Truman-Dewey US presidential election of 1945, was made memorable by the photograph of the victor, Harry Truman, brandishing a copy of the Chicago Tribune that screamed the wrong headline, “Dewey defeats Truman”. So close was the vote. At the end of World War II, the UK voted against Winston Churchill, their doughty PM, who had saved them from Nazi tyranny. He was hustled out of office when the Labour Party won by a substantial majority. Can we ever imagine a similar situation in India? And yet, some of our own elections have indeed yielded surprising results, and one recollects 1971, 1977, and 2019, when the decisiveness of the mandate was quite something else.

So, let’s take all these prognostications in stride and await the authentic verdict.

The author is an Emeritus Professor in the Indian Institute of Science, Bengaluru and is the author of ‘Bharat: India 2.0’ published in 2021. He has an H-index of 104. Views expressed in the above piece are personal and solely those of the author. They do not necessarily reflect Firstpost’s views.