Inference of the Us and Chinese Stock Markets Using Statistical and Computational Methods

Abstract

In the contemporary era, people have strong incentives to explore the underlying principles of stock markets and China and the US are the 2 largest economies across the world. So, it is the stock markets in these two countries that we need to explore and study in this paper. In order to test whether the trends of the US and Chinese stock market are predictable and identify the difference between these two markets, we employed various models to study the S&P 500 and CSI 300 indexes' trends. Specifically, in this paper, we included the Markov chain, hidden Markov model (HMM), logistical regression with lasso, autoregressive integrated moving average (ARIMA) and support vector machine (SVM) to achieve our target.Therefore, we obtained several interesting key findings in our paper. We found that the Chinese stock market is more likely to be affected by technical indicators instead of historical information, as logistical regression with lasso selected most of the technical indicators and the estimated order of the Markov chain is zero when modelling the CSI 300 trends, which is different from the US stock market. Also, the AUC value of SVM outperformed other models used in the US stock market, at 0.731, while ARIMA model resulted in high AUC values in both markets, at 0.606 and 0.622 for the US and Chinese stock markets respectively. So, we confirmed that the Chinese stock market is less efficient than the US stock market. What's more, to predict the future trends in the US stock market, SVM could be the best choice, while ARIMA model works effectively for both markets.