Election forecasting in a land of chaos!


By Chandre Dharmawardana


Can you predict the outcome of a toss of a coin, given enough data and a sufficiently large computer? Can you predict the outcome of the next election if we are given enough opinion polls and a sufficiently large computer? These are of course the sort of questions that might interest many people on the eve of a national election, in a country like Sri Lanka where political fever can end up in blue murder. We all know of astrologers, soothsayers, and election forecasters claiming to predict the outcomes even to near-100% accuracy – nothing less! They will justify it by saying, "look, I predicted the outcome of such and such a previous election with perfect accuracy, I am not like Mr. Pandit X who is a fake; I have a computer programme and I feed into it the election facts and even astrological data"!

A sure criterion of a "fake prediction" is that they rarely come with error bars or confidence limits. They may even predict a "46.3273% of the vote" to some party! The result is given to six significant figures, when no such measurement can be more accurate beyond two significant figures. In fact, we immediately understand that the forecaster has no understanding of what he is doing. It is a "pretence of knowledge", ever so common in this age of fake news that can be instantly loaded on to the social media, enabling them to acquire a global dimension. The prediction is quoted and re-quoted, and hence used even to influence the outcome, since people have a tendency to vote for the winning side. Political parties set up their own political predictions as part of the strategy, but disguised as if the predictions are coming from independent forecasters or astrologers. They can then say, "All predictions favour us".

Another feature of these fake election predictions is bias. When we say "bias", we mean cherry-picking of data to ensure that a certain outcome is obtained. It is like a detective who only selects the evidence that will incriminate a preselected accused person, ignoring the inconvenient evidence. But a scientist or statistician must follow all the evidence, even if they lead to conclusions contrary to one's strongly held beliefs. When Charles Darwin set off on the Beagle to collect data to prove that God had created the world of fauna and flora in all its splendour, he found evidence to the contrary, and took decades to come to deal with the implications. There is, however, the need to discard evidence that is contaminated or compromised, and make a judicious selection of the data. Galileo and Einstein knew what information should be prioritized, while recognizing the need to explain ALL the data.

Of course, prediction of elections is possible as soon as actual polls arrive. When even 10% of the vote is in, extrapolations can be made with increasing confidence. But to claim that one can make a 99% accurate prediction a few weeks ahead of a poll, in an uncontrolled (i.e. free) election, is an attack on our very rationality. Unfortunately, such claims abound in this age of "fake news" and should be resisted by every public-spirited person.

Unfortunately, this "pretence of having knowledge of the future" is common, not only among prophets and election forecasters, but even in many areas of social "sciences". These include political science and even economics. It was Bernard Shaw, who stated that if `you asked 10 economists to predict the direction the economy is headed, they will point in 12 directions'. The most telling indictment of this "pretence to knowledge" came from the Nobel winning economist, Friedrich von Hayek, in his Nobel acceptance speech, entitled a "Pretence of Knowledge". Even without naming names, he claimed that some of his highly distinguished colleagues were charlatans and deceits, or at best self-deluded fools! Basically, he stated that in dealing with complex systems, predictions of specific events, or even the catastrophic collapse of an economy, are impossible; even with the best computers and the most complete knowledge of the system. Leave aside predicting the outcome of elections or the downturns in economic systems; we are still unable to predict earthquakes and Tsunamis.

Poincare was one of the most famous mathematicians of the late 19th century, over-arching into the 20th century. It was he who wrote the letter of recommendation to Einstein, enabling the latter to get his first job in the Patent office in Switzerland. Poincare was also the first person to prove that even if we had all the data, and all the information about three billiards colliding, while the outcome is determined by the laws of mechanics, the outcome is in fact NOT predictable in most instances. The motion – deterministic but without the possibility of a useful prediction - is said to be "chaotic". Poincare's result was regarded as a bit of arcane mathematics, and was soon forgotten, especially in the excitement of the discoveries of the Quantum Theory and the Theory of Relativity. However, the ideas became poignantly relevant in the 1960s, when von Neumann, a renowned name of 20th century physics and the computer revolution returned to the work of Poincare. One could see chaos "happening" in computer simulations. Small planets were discovered whose trajectories are beyond prediction! Sinai, Mandelbrot and others opened a fascinating world of fractals and unpredictable "Sinai billiards". Murray Gell-Mann – Nobel Laureate of eight fold-symmetry fame, and others founded the study of complex systems, and averred that social systems are simply systems which are always at the edge of chaos. So Friedrich von Hayek was completely abreast of the latest thinking in his Voltaire-like indictment of his colleagues who claimed to formulate tools for predicting the future of economic systems.

Popular writers came forward, and coined the terminology of the appearance of "black swans" to characterize the unpredictable qualities of chaotic systems. In "regular" systems, swans are always white! In chaotic systems, there can be those unexpected black swans at any moment, in an uncontrolled way! The direction of history CAN be determined by Cleopatra’s nose.

Of course, Sri Lankans don't need a Poincare or a von Neumann to know about chaotic systems. They have it ready made, created by our politicians who find troubled waters to be most congenial for their moneymaking activities. No one even really knows the number of cabinet ministers – they are a non-countable set? Every election in Sri Lanka is a moment when the country is said to be "at the crossroads" - a sure sign of a chaotic system. It is at least a sign that the system is not controlled, since the outcome is foregone in "elections" held under dictatorships.

Of course, the public has heard of the misuse of statistics, and how "Statistics and damn lies" are equated. But statistics, if correctly used, is an essential tool of our world. Insurance companies, banks, businesses and public health administrators use highly trained statisticians, actuaries, and epidemiologists to evaluate and avert risk, determine the cost of compensation, or the amounts that have to be set aside for dealing with emergencies. Public-opinion polls are used for marketing and product design. Scientists use statistics in analysing their experiments. These are valid applications of statistical methods, and what I have written here is no indictment of such methods, in the sense that you may find in some nonsensical post-modernist writings.


animated gif
Processing Request
Please Wait...