"Language Analysis Using AI Technologies like ChatGPT Increases Accuracy in Economic Research"

Analysis of Analyst Corporate Evaluation Reports by the Bank of Korea

[Asia Economy Reporter Seo So-jeong] A study has found that the accuracy of economic research can be improved through language analysis using artificial intelligence (AI) technology. In particular, recent natural language processing technologies such as ChatGPT are expected to bring significant innovation to the automation of economic analysis through text analysis techniques.

On the 16th, the Bank of Korea stated in its report "Industrial Monitoring Using AI Algorithms: Text Analysis of Securities Company Reports" (BOK Issue Note) that "Text is the most fundamental means of exchanging information, and since there is no limit to the scope of information conveyed, text analysis technology has very high utility in the economic field."

This study algorithmically collected 128,000 corporate analysis reports on 2,283 companies written by 1,079 analysts from 52 securities firms (2019?2022) and analyzed the qualitative information within the reports using natural language processing techniques. Through this, information such as industry-specific corporate conditions and factors affecting the business environment was derived.

As a result of estimating corporate conditions by industry based on the text analysis of securities company reports, the newly proposed text-based business condition index proved to be very useful in predicting macroeconomic indicators such as Gross Domestic Product (GDP) and Business Survey Index (BSI).

In particular, analyzing the overall industry text-based business condition index and the cyclical component of the Leading Economic Index revealed a causal relationship with the Leading Economic Index that does not appear in the KOSPI consensus forecast. Seobeom-seok, head of the Macroeconomic Model Team at the Bank of Korea, explained, "This suggests the possibility that the textual information provided by analysts reflects new information that numbers alone cannot convey."

Additionally, using text analysis algorithms, a table estimating factors of changes in the corporate business environment was created. The estimated change factor table clearly shows industry-specific issues, making it highly useful for understanding industry trends. Furthermore, text analysis can be applied to evaluate the impact of specific economic issues, industry similarity indices, and regional corporate business condition indices.

Seobeom-seok said, "If a vast amount of text information can be aggregated through algorithms, it will be possible to collect the thoughts of analysts, who are the primary producers of corporate information, in real time, greatly improving the work efficiency of economic research analysts who process this information secondarily." He added, "For more in-depth economic analysis using text in the future, it is necessary to analyze the information appearing in the text by linking it with economic theories and other background knowledge, and for this, the construction of large statistical models such as GPT will be required."

This content was produced with the assistance of AI translation services.