The Stock Market and Network Structure
Links: https://www.sciencedirect.com/science/article/abs/pii/S0927539810000368
Network, especially network structure, is highly applicable to a lot of common things in the real world. In A network perspective of the stock market, the authors investigate the links between each stock and attempt to construct a large network made up of stocks that influence each other in United States. I will first explain how the stock market can form a huge network like we talked about in class with different combinations of structures, then summarize how the journal attempts to form a better network with a higher degree of balance.
The stock market is extraordinarily large. Aside from the famous Dow Jones and S&P 500, NYSE alone boasts more than 3,500 stocks with a total market capitalization of US$30.1 trillion. In order to analyze the entire stock market, many analysts and machine learning experts would feed their AI with a huge amount of data from every stock. A network would then be created where each stock is an individual node. Stocks with a strong correlation or covariance would have a strong positive link while stocks with strong negative correlation or covariance would have a strong negative link. Stocks with close to 0 correlation would have no links. With this graph, however, analysts found that there is a lot of structural imbalance. One would naturally expect that, if stock A is highly correlated with stock B and B highly correlated with C, then A would be highly correlated with C. Yet, in this network, this is not true many times, resulting in a violation of strong triadic closure.
In order to prevent contradictions like the above, the journal recommended a different approach. Instead of using the data from every single stock, the authors recommend starting with stocks that have the biggest market capitalizations (sizes), then add in other stocks with very high correlation. This way, it is mathematically impossible for any violations of strong triadic closure or structural imbalance to exist. It will also simply a lot of relationships and reduce noise presented in the data. This will help a lot with portfolio managers deciding which stock to buy or sell.
Overall, the stock market presents an interesting case for the use of networks in analysis. The journal’s ideas provide a unique insight into how to filter out noise and imbalance within a large network.
Bibliography
Tse, C., Liu, J., & Lau, F. (2010). A network perspective of the stock market. Journal Of Empirical Finance, 17(4), 659-667. doi: 10.1016/j.jempfin.2010.04.008