Indexing Economic Fluctuation Narratives from Keiki Watchers Survey

TL;DR

The paper addresses how causal narratives extracted from the Keiki Watchers Survey can augment traditional diffusion indices to forecast economic fluctuations. It adapts a climate-focused narrative framework to a Japanese economic survey, extracting cause–effect relations, building cross-topic causal chains, and aggregating them into 156 time-series via a time-decayed weighting scheme. The resulting Economic Fluctuation Narrative Indices exhibit strong associations with the Diffusion Index, with notable performance for cumulative lagging indicators (e.g., a top chain achieving $r=0.81$). These findings demonstrate the value of narrative causality for economic analysis and point to future work in Granger-causality testing and cross-language narrative index development.

Abstract

In this paper, we design indices of economic fluctuation narratives derived from economic surveys. Companies, governments, and investors rely on key metrics like GDP and industrial production indices to predict economic trends. However, they have yet to effectively leverage the wealth of information contained in economic text, such as causal relationships, in their economic forecasting. Therefore, we design indices of economic fluctuation from economic surveys by using our previously proposed narrative framework. From the evaluation results, it is observed that the proposed indices had a stronger correlation with cumulative lagging diffusion index than other types of diffusion indices.

Figures (8)

• Figure 1: Overview: The methodology begins with extracting causal information from the Keiki Watchers Survey, followed by constructing causal chains to form economic fluctuation narratives. These narratives are indexed to calculate the Economic Fluctuation Narrative Index, which is then analyzed for correlations with the Diffusion Index (DI). The results are visualized using heatmaps to highlight the relationships between the narratives and the DI.
• Figure 2: Chain Example: This figure illustrates an example of constructing a causal chain. When two explanation texts from different topics contain causal relationships and the cosine similarity between the result expression of the earlier sentence and the cause expression of the later sentence exceeds 0.5, they are considered part of the same chain. In this example, the earlier topic, "Job Offer Movement," and the later topic, "Sales Volume Movement," are linked, forming a narrative.
• Figure 3: Heatmap of Pearson Correlations between Economic Narratives and Cumulative Lagging Diffusion Index: The heatmap illustrates Pearson correlations between indices derived from economic narratives and the Cumulative Lagging Diffusion Index (DI). Rows represent front causality (cause expressions), and columns represent rear causality (result expressions). For example, the index calculated from the narrative linking "Number of Visitors" as the front causality and "Sales Volume Movement" as the rear causality shows a Pearson correlation of 0.81 with the Cumulative Lagging DI. This highlights the strong relationship between specific economic causal chains and the DI.
• Figure 4: Heatmap of Pearson Correlations between Economic Narratives and Lagging Diffusion Index
• Figure 5: Heatmap of Pearson Correlations between Economic Narratives and Cumulative Coincident Diffusion Index