Stock market Photo: TUCHONG

A profound understanding of the essence and laws of the financial system is a major issue of concern for China. A complex system approach should be adopted to study the financial-economic system, as this system depends on a multiplicity of factors with complex interconnections, and interacts with the social environment extensively and profoundly.

Computational experimental finance

With the rapid advancements in computing power and hardware facilities, computational experimental methods have emerged as a viable approach for modeling complex financial systems that traditional analytical models often fail to describe.

Computational experimental finance aims to model financial market participants “from the bottom up” to explore the complex laws of evolution and micro-level formation mechanisms of financial systems. This field has witnessed the emergence of several key models.

Individual learning models use intelligent algorithms to simulate investors’ learning behavior and construct artificial stock markets through multi-stage learning, with the goal of reproducing the complexities of real financial systems, and understanding the micro mechanisms behind the operating principles of the stock market. 

Population evolution models are based on the integration of ecological concepts into computational finance. Researchers first divide system participants into different populations by preferences, prior beliefs, and wealth status, then simulate evolutionary processes according to trading rules to reproduce and understand specific market phenomena.

Individual learning models and population evolution models are relatively simple because they do not consider more complex interactions between individuals. By contrast, complex network models can help better understand the laws governing the evolution of financial systems.

Among them, scale-free networks can be used to model networks composed of highly heterogeneous nodes. This modeling method has been applied to analyzing the contagion of credit default risk and the vulnerability of financial systems, as it can describe the significant uneven distribution emerging in complex systems as a whole.

At present, big data technology allows researchers to observe system phenomena with finer granularity, and upgraded computer technology facilitates the modeling of financial systems from a micro-level perspective. In the future, computational experimental finance is likely to make progress in areas such as the impact of heterogeneous participants on financial markets, formation mechanisms of irrational factors, and the modeling of market feedback mechanisms.

Social finance

In recent years, progress has also been made in social finance, one of the frontier areas of research on complex financial-economic systems. Social finance views financial markets as systems of cultural evolution, exploring their laws of evolution in a comprehensive manner that factors in the influence of dynamic socio-cultural characteristics, such as information, strategies, beliefs, and market patterns. Multiple types of social finance research are closely related to complex financial-economic systems.

The first type is called the mass model, as it is the popular formulation of the operating principles of the world. The mass model is ubiquitous and easy to understand, and often appeals to perception and emotion, such as various “myths” about stock price trends.

The second type of research examines the ways individuals are influenced by the social transmission bias of economic signals. An important takeaway from this type of research is that wealth-related information asymmetry discourages consumption, which is consistent with the increase in savings due to wealth inequality observed in empirical research.

The third type of research focuses on the impact of self-reinforcing transmission bias on individual decision-making. Investors in financial markets tend to brag about their own investment success, leading those who receive the signal to adopt underperforming, high-risk strategies.

The fourth type of research emphasizes that the limited carrying capacity of existing means of information transmission makes it difficult for the social communication network to correct subtle differences in individual beliefs.

Classical finance already acknowledges that learning from others is at the core of financial markets. Social finance explicitly centers on how social interaction shapes the thoughts and behavior of system participants, as well as the evolutionary processes of these characteristics. Therefore, it can help explain important phenomena in real financial systems, such as market price bubbles and anomalies in the predictability of stock market returns.

New opportunities

Research on complex financial-economic systems is faced with a number of challenges, particularly in three aspects. The first concerns data acquisition and processing. The second involves model establishment and validation. The third pertains to prediction and decision-making. Machine learning outperforms traditional economic models in addressing these challenges, as it provides more flexible model configuration to deal with complex interaction between variables and ambiguous functional forms.

Previous financial-economic research mostly adopted economic or statistical models trained on small datasets, and focused only on relevant information within their target economic mechanism, resulting in a vast amount of untapped data. However, small models lack the capability to capture the fundamental laws governing ultra-high dimensional, dynamic, and complex financial-economic systems. In this regard, large language models (LLMs) have the potential to offer news tools and methodologies.

Sentiment analysis of public opinion data is a promising application of LLMs in financial-economic research. Researchers can integrate web scraping tools with LLMs to automatically collect content posted on social media, and train the models to better understand valuable information conveyed by the language, thereby identifying the affective leaning of each piece of information.

Research on complex financial-economic systems can expect a bright future, given the increasingly complex internal and external environments of the financial system, advances in computing and data storage technologies, and the growing practical need for clarification of complex financial-economic phenomena.

Yang Xuewei (professor) and Zhu Peng are from the School of Management and Engineering at Nanjing University.

Edited by WANG YOURAN