Literature Review

The study of economic policy uncertainty has become a focal point in contemporary research, given its far-reaching implications on financial markets, economic performance, corporate investment, labor market dynamics, and political polarization. The literature on measuring uncertainty has evolved significantly, encompassing various methods that can be broadly classified into three categories: financial measures, textual measures, and topic-based methods.

The literature on economic policy uncertainty measurement reflects a dynamic and evolving field. From financial measures to textual and topic-based methods, the journey illustrates a continuous quest for accuracy, efficiency, and comprehensiveness. The integration of machine learning techniques and the development of global indices represent significant milestones. However, the pursuit of real-time, universally applicable, and nuanced measures of uncertainty continues to inspire ongoing research and innovation.

Financial Measures

Financial measures of uncertainty often hinge on stock market volatility and discrepancies between expected and actual financial parameters. The VIX index, forecast errors in the stock market, and variations between expected and actual macroeconomic parameters collected through surveys are common tools in this category (Kaveh-yazdy and Zarifzadeh [2021]). These measures provide a quantitative lens to gauge uncertainty, linking it directly to observable financial market behavior.

Textual Measures

Textual measures represent a shift from quantitative financial data to qualitative analysis of documents such as monetary policy documents, social media posts, and news articles. The Economic Policy Uncertainty (EPU) Index, proposed by Baker et al. [2016], marked a significant advancement in this field. By measuring the frequency of articles containing specific uncertainty-related keywords, the EPU index quantified policy uncertainty across various domains.

Despite its innovation, the EPU index faced challenges, particularly in the labor-intensive manual process of article selection. This limitation led to the exploration of machine learning methods to automate the process. Azqueta-Gavaldón’s use of Latent Dirichlet Allocation (LDA) for topic modeling in uncertainty measurement offered a more efficient alternative to the EPU index (Azqueta-gavaldon [2017]). Subsequent research further embraced sophisticated techniques like support vector machines and logistic regression for precise document classification and tagging (Kaveh-yazdy and Zarifzadeh [2021]).

Topic-based Methods

The evolution of uncertainty measurement witnessed the emergence of topic-based methods, addressing the limitations of keyword-based approaches. Recent works such as Larsen [2021] leveraged LDA to categorize articles based on content and themes without pre-training or labeling. This method allowed the entire word mixture in an article to contribute to theme identification, differentiating between various types of uncertainty and offering a more nuanced understanding.

Future Directions and Challenges

While significant strides have been made in uncertainty measurement, challenges persist. The labor-intensive nature of the EPU index and its limited applicability to advanced economies have prompted further innovations. The World Uncertainty Index (WUI), covering 143 countries, and novel approaches using semantic clustering, word embeddings, and fuzzy k-means have been introduced to facilitate real-time measurement and updating of the index (Miranda-belmonte et al. [2023]).