In a wonderful book “Economyths”, author, David Orrell, writes on ‘complexity’ and the dangers of models for predictions of economic phenomena.

“So if the traditional reductionist approach doesn't work, what is the alternative? Emergent phenomena have been widely studied by complexity scientists, through the use of techniques such as cellular automata or agent-based models. Cellular automata are computer programs that typically divide the screen into a grid of cells. The evolution of the system is governed by simple rules that describe how one cell affects its neighbours. While the laws are simple at the local level, the emergent behaviour at the global level can be extremely complex, and can't be modelled directly using equations. Cellular automata have been used to study a wide range of phenomena, including turbulent fluid flow, avalanches, the spread of forest fires, and urban development.

Agent-based models consist of multiple software 'agents' that could represent, say, investors in the stockmarket. The agents are allowed to influence each other's behaviour, just as in reality investors communicate with those around them. They make decisions based not on uniform laws, but on fuzzy heuristics or rules of thumb. Agents can also learn and adapt their behaviour, in the same way that investors become more conservative after being burned by a market fall. It is therefore impossible to assign them a fixed and independent demand curve of the sort required by the 'law of supply and demand'.

The collective effect of the agents is again to produce emergent behaviour that is often quite surprising, and that can lead to useful insights about how the system works. Agent-based models have been used to reproduce the boom/bust behaviour of markets, and have found many other applications in areas from transport to cancer therapy. Programmes in complexity are starting to appear at business schools and institutions like the London School of Economics. The first way to revive economics, then, is to encourage this trend, and in the process rid the field of its quasi-Newtonian pseudo-laws.

One drawback of this type of research is that it has none of the icy glamour and prestige of great Newtonian mathematical laws. It is unlikely that anyone will ever win a Nobel Prize for an agent-based model. Nor does complexity theory offer a single unified approach. Models are seen more as patches, each which captures an aspect of the complex reality.

Also, while the complex systems approach is useful for simulating many aspects of the economy, it is unlikely that it will prove to be much better than orthodox theory at predicting the course of something like the housing markets. The reason is that the exact behaviour of a system depends on all the exact details, and the only way to predict a system would be to reproduce it on the computer. That's the point of emergent properties: they can't be predicted by a simple equation. Instead, complexity scientists search for pockets of predictability - aspects of the system that are amenable to prediction.

Complexity research has many implications for economics (most of the conclusions of this book are based on a complexity viewpoint), but its most devastating consequence is that it throws a spanner in the entire mechanistic approach for modelling complex systems like the economy. Newton's blueprint for numerical prediction, again, was to reduce a system to its fundamental components, discover the physical laws that rule them, express as mathematical equations, and solve. But reductionist method doesn't work for emergent properties. There are no fixed laws - only general fuzzy principles that can be roughly captured by rules of thumb but rarely conform to neat mathematical equations. The message of the Pythagoreans- that all can be reduced to number - turns out not to be true.”