Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Wage Sentiment Indices Derived from Survey Comments via Large Language Models

Taihei Sone

TL;DR

This paper develops a Wage Sentiment Index (WSI) for Japan by applying Large Language Models to the Economy Watchers Survey (EWS) and validating it as a leading indicator of wages from the Monthly Labour Survey (MLS). It extends the prior Price Sentiment Index (PSI) framework and introduces a scalable data architecture to incorporate additional data sources. The study demonstrates that WSI models based on LLMs outperform baselines and discriminative pretrained models in predicting wage dynamics, with Granger-causality tests confirming predictive lead relationships, particularly for standard WSIs derived from LLMs. The findings suggest that LLM-driven sentiment indices can enhance the timeliness and effectiveness of economic policy design, while the proposed architecture supports future expansion to newspapers and social media and potential cost savings through alternative models.

Abstract

The emergence of generative Artificial Intelligence (AI) has created new opportunities for economic text analysis. This study proposes a Wage Sentiment Index (WSI) constructed with Large Language Models (LLMs) to forecast wage dynamics in Japan. The analysis is based on the Economy Watchers Survey (EWS), a monthly survey conducted by the Cabinet Office of Japan that captures real-time economic assessments from workers in industries highly sensitive to business conditions. The WSI extends the framework of the Price Sentiment Index (PSI) used in prior studies, adapting it specifically to wage related sentiment. To ensure scalability and adaptability, a data architecture is also developed that enables integration of additional sources such as newspapers and social media. Experimental results demonstrate that WSI models based on LLMs significantly outperform both baseline approaches and pretrained models. These findings highlight the potential of LLM-driven sentiment indices to enhance the timeliness and effectiveness of economic policy design by governments and central banks.

Wage Sentiment Indices Derived from Survey Comments via Large Language Models

TL;DR

This paper develops a Wage Sentiment Index (WSI) for Japan by applying Large Language Models to the Economy Watchers Survey (EWS) and validating it as a leading indicator of wages from the Monthly Labour Survey (MLS). It extends the prior Price Sentiment Index (PSI) framework and introduces a scalable data architecture to incorporate additional data sources. The study demonstrates that WSI models based on LLMs outperform baselines and discriminative pretrained models in predicting wage dynamics, with Granger-causality tests confirming predictive lead relationships, particularly for standard WSIs derived from LLMs. The findings suggest that LLM-driven sentiment indices can enhance the timeliness and effectiveness of economic policy design, while the proposed architecture supports future expansion to newspapers and social media and potential cost savings through alternative models.

Abstract

The emergence of generative Artificial Intelligence (AI) has created new opportunities for economic text analysis. This study proposes a Wage Sentiment Index (WSI) constructed with Large Language Models (LLMs) to forecast wage dynamics in Japan. The analysis is based on the Economy Watchers Survey (EWS), a monthly survey conducted by the Cabinet Office of Japan that captures real-time economic assessments from workers in industries highly sensitive to business conditions. The WSI extends the framework of the Price Sentiment Index (PSI) used in prior studies, adapting it specifically to wage related sentiment. To ensure scalability and adaptability, a data architecture is also developed that enables integration of additional sources such as newspapers and social media. Experimental results demonstrate that WSI models based on LLMs significantly outperform both baseline approaches and pretrained models. These findings highlight the potential of LLM-driven sentiment indices to enhance the timeliness and effectiveness of economic policy design by governments and central banks.

Paper Structure

This paper contains 16 sections, 2 equations, 28 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Literature
  3. Method
  4. How to Calculate the WSIs
  5. How to Classify Comments When Calculating the WSIs
  6. Baseline Model
  7. Discriminative Pretrained Language Models
  8. LLMs
  9. Data
  10. The EWS
  11. The MLS
  12. Data Architecture
  13. Results & Discussions
  14. WSI Time Series
  15. Granger Causality Test
  16. ...and 1 more sections

Figures (28)

  • Figure 1: Distribution of Economic Judgment
  • Figure 2: Number of Records by Region
  • Figure 3: Japan Map with Regions
  • Figure 4: Monthly Count Trend by Economic Judgment
  • Figure 5: Monthly Percentage Trend by Economic Judgment
  • ...and 23 more figures