Study on Factor Analysis of Stock Price Fluctuations Based on Time Varying Coefficients Model

Stock investors seek to minimize risk and continually evaluate whether their target companies are suitable. To achieve this, it is crucial to understand the trends in corporate valuation. This study aims to gain valuable insights into the evaluation of portfolio companies through a factor analysis of stock price fluctuations. In this study, text data were used to analyze the factors behind stock price fluctuations. To model the data, it was necessary to quantify the text using methods such as bag-of-words. This made the specific information vague. Therefore, it was impossible to examine the specific factors that caused stock price fluctuations at any given time. For example, financial statements (e.g., investor relations materials) were provided as textual information to disclose information about how companies perform. As these contexts contain various content, each had a different influence on stock price fluctuations. However, if the data were used in model learning, differences in content might not be represented and might be considered to have the same effect as financial data. We focus on this issue of past studies and propose a new method to analyze the detailed information of the stock price for each time.. We first constructed a model using information other than textual data. Second, we focused on the model residuals to clarify the textual information to be focused on. The proposed model was based on a time-varying coefficient model that used numerical data such as economic trend indicators. In addition, we validated the proposed model by analyzing real-world data.