A Brief Overview of Emergence According to Systems Innovation, an emergent system can be defined as a process of change that involves non-linear, abrupt transitions as a system's overall structure and function is transformed into a new regime of behavior, exhibiting properties that could not have been predicted to arise prior to the transformation. Emergence has applications in art, science, philosophy and systems theory but to name a few. Markets, such as energy, flea, and stock markets, are examples of emergence on a grand scale. Regarding the stock market, it regulates the security prices of worldwide securities, yet it has no one individual or institution coordinating it, and thus exhibits emergence.
Because of its all-encompassing nature, there are a variety of definitions that exist in order to properly capture the concept. These various definitions are written to best suit the particular discipline that is being written for. According to Mark Kuperberg, the common characteristics of emergence can be summarized as follows:
There are at least two levels of organization;
There is a multitude of individual agents at the lower level of organization who operate by following simple rules; and
There is an aggregate outcome at the higher level that results from the interaction of these individual agents, but which is not easily derivable from the results that the individual agents follow. Many times, therefore, this aggregate behavior comes as a surprise to the observer because nothing in the rules at the lower level seems to predetermine the aggregate outcome.
As a concept, emergence might be a bit tricky to grasp. The following 2 minute video from Smithsonian Education helps to simplify the concept.
From the above video, we notice the following about ant colonies,and thus emergent processes: Ants are relatively simple components in the complex system of the ant colony. The individual behavior of each ant is essentially simplistic when juxtaposed to what the colony is doing. This is because the colony is capable of building nests, waging war with other colonies, and raising their own type of livestock among other reasons but to name a few! Ant colonies, therefore,show the benefit of harnessing the power of emergent systems.
Crowd Simulation in Unity
Crowds, which are self-organizing systems, also exhibit emergent behavior. In a crowd, individuals or agents are prone to exhibit behavior that is different from their individual, ordinary behavior. According to Reynolds, very simple behavior of the members of a crowd can result in pleasing and complex collective behavior. I demonstrate such behavior in a simulation in Unity, a cross-platform game engine. When observing the simulation, take note of the paths that form when the different colored agents interact with one another. That is where the crowd behavior displays itself.
The most interesting thing I find about emergence, is that it forms part of the components of agents-based-modelling, which is what we are going to discuss next.
A Brief Overview of Agent-Based Modelling
According to Wikipedia, an agent-based model is a class of computational models for simulating the actions and interactions of autonomous agents with a view of assessing their effects on the system as a whole. ABMs, are by their very nature, an extremely ideal framework to study emergent phenomena: in an agent-based-model, one models and simulates the behavior of a system's member agents and their resulting interactions, capturing emergent phenomena as the simulation is ran in a bottom-up manner.
One event that shook the economics profession in recent history would be the Great Recession. This recession demonstrated that the economy is complex, and thus not always in a state equilibrium. When applied to the field of economics, agent-based models exhibit both strengths and weaknesses. In the case of forecasting the price of a particular security, ABMs produce poor results. However, they are much more beneficial when it comes to explaining phenomena like bubbles, cycles, and the onset of bull and bear markets, among others.
Stock Market Simulation
After having discussed the theoretical aspects of agent-based models, it is time for an application. In order to achieve this, we will make use of Adaptive Modeler. This software application creates market simulation models in which thousands of virtual traders apply their
own trading strategies to real-world market data to trade, compete and adapt on a virtual market. The collective behavior exhibited in this software is an illustration of emergence, as was discussed earlier in this article.Adaptive Modeler comes with a variety of features so as to better understand and predict the prices of securities. For the sake of brevity, I will discuss some of the rudimental features of the software and leave some of more interesting features to interested readers.
After initializing the model, we get a population of 2000 trading agents, where each agent has its own trading rule, a starting capital of $100000, and no shares. This is what the model looks like when its ready. During evolution the model will process the historical quotes in the quote file as live streaming data. We will focus on three features offered by Adaptive Modeler.
1.General Overview
We will now evolve the model for 1000 days to see what we can gather from the population. In the video above we can observe several details about how the agents interact with one another. There's information on the populations's buy and sell orders, and a chart plotting the wealth distribution of the agents. While the stock returns for the S&P have a lognormal distribution, the wealth distribution of the agents has a normal distribution which is skewed to the right. The plot for the position distribution has a bimodal distribution, and this makes sense because agents can either be long or short the S&P.
2. Regarding the Trading Agents
We now evolve the market for another 1000 days. The most compelling data from an agent-based model is seen at the population level rather than in the individual agents. As was mentioned earlier, the purpose of an ABM is to study emergent behavior caused by the interaction of (relatively) simple agents.
In the above simulation, we observe how 3 agents compare in the market. Each agent's performance is determined by its own unique trading rule. Also, the agents are seen to periodically reset. This is because the particular agent has been replaced by a new agent in the breeding process. Agent breeding is an extensive topic and is not covered in this article. Overall agent one is the most successful agent, achieving a return of 20% at some point, while agent 13 is the least successful agent, achieving a total return of -15 at some point.
3. Regarding the Population
Finally, we observe how the agents wealth accumulates when plotted against their age, after evolving the model for another 1000 days. The most successful agents, which can be observed in the top right corner, have the greatest age. For individuals designing trading strategies, such an observation informs them to design with longevity in mind. For example, Such a design might cap the portfolio leverage at 25%.
Concluding Remarks
Regarding the similarities between ant colonies, economies and financial markets, one may summarize by saying that these systems are best understood as complex systems, especially when they are modeled through agent-based-models. As the use of Agent-Based-Models continue to increase in the world of finance, it is import to not disregard the current arsenal of tools that exists to study financial phenomena like recessions and economic phenomena like quantitative easing. What matters is a balance between these two resources of analysis, so as to best understand the particular issues at hand. Complement this article with this video on the use of agent-based-models to simulate emergency evacuation strategies.
References
Emergence, SystemsInnovation.io
The Two Faces of Emergence, swarthmore.edu
Smithsonian Learning Lab, learninlab.si.edu
Identifying Emergent Behaviour of Complex Systems, thenewstack.io
Emergent Crowd Behaviour, Research Paper
Unity For Everyone, youtube.com
Agent-Based Model, Wikipedia.com
Understanding the economy from bottom up, econpapers.repec.org
Altreva, Stock market forecasting software, altreva.com
Why the lognormal distribution is used to describe stock prices, financetrain.org
Binomial Distribution: What is it, statisticshwoto.org
Right-Skewed Distribution: What does it mean blog.prepscholar.org
Calculating Portfolio Leverage, tastytrade.com
Complex Systems, Wikipedia.com
Comments