Abstract

Understanding the foreign investment dynamics is essential to compile Balance of Payments Statistics and its analysis. This paper proposes a methodology to help understand the dynamics of foreign investment through multimodal data from both unstructured and structured data. The unstructured data from online news is analyzed in advance by applying text mining to provide a leading indicator to proxy the likely future trends of foreign investment. On the other hand, the reporting data (such as International Trade Reporting System, and External Debt Reporting System) and other administrative/business data, as structured data, are also utilized to construct the current foreign investment realization. We demonstrate the analysis in aggregate value or more detailed focus such as portfolio investment, direct investment, etc. Our analysis shows that the Balance of Payments Statistics quantitatively confirms leading and prompt indicators. In addition, we also performed further investigation that can capture the evidence to support the foreign investor behavior analysis by extracting information from the news and the realization from structured data. We suggest that our methods can be used to enrich the Balance of Payments Statistics, in relation with the foreign investment.