My working life revolved around economics and stock markets. Over the years I developed a scepticism of the forecasts which were constantly dished out by analysts, governments, agencies and the media. That’s not to say forecasts do not have some use, but they always need to be taken with a healthy pinch of salt. I think most people intuitively know this, which is why it is endlessly frustrating to be constantly bombarded with economic forecasts by politicians and the media as if they have any real credibility.
Until now, I’ve put these inaccuracies down to the complex nature of economies and the sometimes irrational behaviour of mankind. The economically rational man is a useful tool for theories but breaks down when it comes to the real world. However, it appears there is a far deeper and fundamental problem that makes not just economic forecasting problematic but any predictions concerning phenomena with complex interactions. I discovered this intellectual treasure recently while reading a book on human consciousness (The Physics of Consciousness by Andrew Thomas).
Since Galileo and Descartes, science has largely followed a reductionist approach to problems. Reductionism means breaking down phenomena into their constituent parts to understand how the whole system works. This approach has been extremely effective in many scientific areas from particle physics at one end to cosmology at the other. This has been mimicked in the field of economics too (although with markedly less success!).
However, when we come to understanding the human brain and the thorny problem of human consciousness, the reductionist approach becomes inadequate. We have a good understanding of how neurons work, but a poor understanding of how they work together as a system to produce consciousness. This is because they react in a non-linear fashion. The best way to illustrate this is to consider two electrical components: resistors and transistors.
Resistors are linear. The output current of a circuit with a resistor in it will vary in a linear fashion with the input current. Not only that, a complex circuit of many resistors can always be simplified to a circuit containing a single resistor. Hence, phenomena which are linear in character can be subjected to a reductionist approach. Effectively, linear phenomena can always be simplified, hence they are much easier to analyse and, generally, predicted with a high degree of accuracy.
The best examples would be the movement of planets, stars and galaxies or subatomic particles are generally linear and predictable. Interestingly, in particle physics there is a very small level of randomness associated with quantum uncertainty, but at a macro level, it has no real impact in practical terms and can be ignored. I’ve also recently discovered that over very long time frames, solar systems and galaxies can also exhibit unpredictability and chaotic behaviour!
If we take a transistor, it is a non-linear component. For a current to flow through it, the voltage has to exceed a threshold amount. Any circuit with transistors cannot be simplified and hence a reductionist approach does not work as just examining each component in isolation cannot explain how the circuit works as a whole. It turns out that our neurons are similar to transistors, although they rely on electro-chemical reactions, not just electrical impulses.
To give an indication of how complex our brains are, we have about 100 billion neurons. However, each neuron has about 10,000 connections (synapses) to other neurons. Overall, our brains have about 1,000 trillion connecting nodes. The non-linearity of neurons, the vast number of them and the incredible number of connections means that by solely looking at neurons in isolation, in a reductionist approach, it is utterly inadequate to explain how the brain works and how consciousness develops.
Complex systems give rise to what is called emergent behaviour. Emergent behaviour is where a system is more than the sum of the parts and outcomes become more or wholly unpredictable. Enmeshed with emergent behaviour is chaos theory. Here we have to take an unexpected detour into meteorology and possibly one of the most important scientists that hardly anyone has heard of: Edward Lorenz, a meteorologist at MIT.
Arguably, Lorenz is the father of deterministic chaos theory in science. He was skeptical of the appropriateness of the linear statistical models in meteorology at that time, as most atmospheric phenomena involved in weather forecasting are non-linear. Lorenz constructed a relatively simple twelve factor computer model of weather systems. Because computers in the 1960’s were crude, he made some short cuts in calculations by rounding decimal places and starting in the middle of calculations. What he found was astonishing. Incredibly small differences in starting conditions by rounding decimal places had massive impacts on outcomes, immortalising “the butterfly effect” (a butterfly flapping its wings in Brazil causes a hurricane in Texas).
Lorenz’s discovery showed that even detailed atmospheric simulations cannot make precise long-term weather predictions. One of his most important conclusions was that physical systems can be completely deterministic and yet still be inherently unpredictable or chaotic, even in the absence of quantum effects, which was dubbed “deterministic chaos”. Lorenz’s research explains why it is so difficult to forecast weather and other complex phenomena accurately.
Essentially there is a measurement problem and a computational problem. The measurement problem is that measurements can never be precise enough to accurately predict outcomes from any initial state. In essence, you would have to know the initial state of every single particle with absolute precision. Taking it a stage further, in quantum physics, you can only be sure of either the position of a particle or its speed but not both at the same time (Heisenberg’s uncertainty principle). This means that absolute accuracy in measuring the state of every individual particle is impossible.
Not only is the measurement of the initial state of a system impossible but there is never enough computing power to simulate the outcome. A recent research paper (Ringel and Kovrizhi, 2017) demonstrated that there aren’t enough atoms in the universe to contain the data for a Monte Carlo simulation of two hundred electrons (not atoms!) when studying the quantum Hall effect. If you can’t simulate two hundred electrons, how can you hope to accurately model either movements of particles in the atmosphere or the interactions of the human brain?
This brings us back to the original issue of economic models and their efficacy and accuracy. If we can’t accurately model the human brain, the emergence of consciousness or account for the impact of chaos or complexity, why should we ascribe anything more than a passing interest in economic models? That’s not to say that models are completely useless, but they do have severe limitations, especially when it comes to non-linear and emergent phenomena like the behaviour of humans and especially in the realms of economics.
Next time you see an economic forecast in the media, just remember the phrase “garbage in, garbage out”. I’m afraid that just about sums up the “science” of economic forecasting. Forecasts are fundamentally flawed, because we can never fully understand the wonders of the human brain, however frustrating that might be. All the forecasts that you see on Brexit from the OBR , Treasury or whoever are essentially wild guesses and next to worthless.
There is also a good case to be made that the issue of non-linearity and deterministic chaos is a fatal blow to the climate change/climate emergency brouhaha. If atmospheric phenomena are non-linear and emergent, then, by definition, climate has to have the same characteristics (see below*). This means that predictions of what the earth’s climate will be like in ten, fifty or a hundred years time are intrinsically impossible. All the models are fundamentally flawed because they rely on reductive analysis and hence are virtually useless for non-linear phenomena. So next time you see predictions of a climate apocalypse, treat them with the same caution as economic forecasts.
* IPCC TAR Chap 14, Exec Summary: “…we should recognise that we are dealing with a coupled non-linear chaotic system, and therefore that the long-term prediction of future climate states is not possible.”