Models may not be Robust

In our effort to understand the world around us, we observe, we form hypotheses and we test whether or not our explanations match reality. From there, we use models to test our understanding of natural systems. Since the early 20th century, we have tried to understand financial systems using this scientific form of reasoning. Practitioners and academics have used the tools available to them to explain the movements of financial markets and develop a model that reflects their understanding.

Practitioners (traders and investors) gain experience in the markets, which reflect reality. They are mostly able to perceive what has happened, even if they are not always able to assume why. They use their reasoning and common sense build on their experience. This type of knowledge is reflected in aphorisms and sayings that are passed back and forth between practitioners. Sayings like “sell in May and go away”, “don’t fight the tape”, or “buy the rumour, sell the news” reflect an accumulation of experiences that have shown these aphorisms to be helpful. The main drawback to this type of accumulated, popular wisdom is that people don’t always process their experiences rationally. Think, for example, of a gambler who “almost won”. Almost winning is exactly the same as losing, in reality, but it is experienced very differently. This type of experience could skew the perceptions that support popular wisdom.

Academics and models try to explain and reproduce reality. Because reality is complex, they simplify and they take shortcuts in an effort to reduce the complexity and develop a model that is easy to use. These models are then used to try to predict the future. As we have seen, our models are sometimes helpful, but are never failproof. (Take as an example the saying that “economists have predicted nine of the last five recessions.”) Contrast this with a physical, not social, science. Models in physics, chemistry or biology are able to exactly explain and predict outcomes with known variables. With social sciences, however, the greatest variable is human action. Economists tried to reduce this by inventing “homo economicus” or rational man, someone whose choices are logical and preferences are consistent. As behavioural economics has shown, that man doesn’t exist.

Because movements of financial markets are based on human behaviour, random fluctuations do not fit within the same realm of randomness as Brownian motion or other outcomes that can be described by the bell curve. The bell curve is a tool that academics use to reduce the potential outcomes of a random variable that can’t be known. If the distribution is normal, the validity of the outcome will not be affected and, although the answer is not known exactly, it will always fit within known parameters. If the random fluctuations, however, do not fit the bell curve, it cannot be used as a shortcut to describe the parameters of potential outcomes. This has, in fact, been shown to be the case in many ways. Just a couple are the market crash of 2008, which had a standard deviation of 12, meaning it should only happen less than once in the lifetime of the Earth; the fact that measured volatility is not consistent over time; and the fact that correlations change with circumstances.

Close enough may be okay for some things, but it’s not good enough for my money. I’d rather use no model at all, than a model that is known to be flawed. But this seems to not be the case for the financial industry. Instead of rejecting a flawed model, ever more complex measures and ratios have been developed, always based on the same assumptions, ie. normal distribution and rational decision making. When forecasts are wrong, excuses are made such as: the direction was right, but the magnitude was wrong; the movement was right, but the timing was wrong; or the prediction would have been correct, but the markets were manipulated by the government. Those are all ways of “almost winning”, which as we recall is exactly the same as not winning, especially when money is at stake.

Instead of trying to predict the future based on models that do not accurately represent reality, it is more robust to heed popular wisdom that is based on experience and keep in mind that the outcome could be vastly different. I don’t mind admitting that financial markets in specific and human behaviour in general are systems that we can’t model because we don’t understand them sufficiently. Facing this reality allows a practitioner to make decisions that are not based on erroneous assumptions.

Advertisements

One thought on “Models may not be Robust

  1. Pingback: The Antithesis of Robustness: LTCM | Financial Thinking

Comments are closed.