VR Logo

Perils of Poll Prediction

Soothsayers Inc has been severely struck by the electorate

To loosely paraphrase Hollywood actor Danny DeVito in a movie that he acted in, ‘No one can foretell the future. That’s why it’s called the future.” That’s a lesson that’s always been there for all manner of soothsayers and their followers to learn. You couldn’t tell from watching the news channels on TV, but the electorate has made an even bigger joke of psephologists and pollsters than it has of Lalu Prasad Yadav, Ram Vilas Paswan and the Left.

I feel this rather strongly as an investment analyst and advisor. Election pollsters may crawl out of the woodwork only once in a while, but their investment equivalents are in business all the time, even though they have been wrong almost all the time. Over the last two years, they’ve been wrong first on the way down, and then on the way up.

The basic problem is not that this or that psephologist or analyst doesn’t know how to do his job, but that this job can’t be done, at least in any useful way. Almost by definition, this job doesn’t exist. In a manner of speaking, the Indian elections are what would be called a non-linear system in science or maths. In non-linear systems, small changes in inputs produce large changes in the output. Moreover, these changes are neither proportional to, nor even necessarily in the same direction as the changes in input.

This may sound like a contradiction, but my hunch is that in a technical sense, various pollsters did a reasonably competent job. A first-past-the-post parliamentary election system with a huge number of fragmented battlefields is a severely non-linear system. At any reasonable sample size, perfectly reasonable sampling errors would magnify to such huge deviations in the final results that the entire exercise will be meaningless. There’s no hope of making a correct prediction, except once in a while, and except by chance. As the results of the past many elections prove, this is an exercise that can’t succeed.

Or maybe not. Whether this exercise was a success or not actually depends on what its goal was. I’m just naively assuming that the goal was to make correct predictions. Perhaps the goal was simply to entertain people and generate viewership. And I suppose it was a success at that. In fact, the polls’ failure makes them more interesting, and even gives newspaper columnists a whole new subject to write about.

However, there should be no such ambiguity about the purpose of predictions about the markets. Those should have a precisely-defined goal—to help people make more money, or at the very least, to help them lose less money. Unfortunately, the stock markets are an even more non-linear system. Not only does it have a wide variety of poorly-knowable inputs, it is a perpetual system. That means its own outputs are continuously becoming further inputs.

Interestingly, while both systems seem immune to a systematic analysis, in my experience they both yield to the instinctive understanding that people have. In these elections, practically every experienced journalist who had traveled across the country would privately say that the Congress was likely to sweep the polls. Once upon a time, this would have been the view that the media would have carried. Unfortunately, we’ve developed a mindset where we would rather trust half-baked systems with a veneer of science and ignore the instincts of the experienced.