sophychen79/Weak-supervision-in-Economic-Policy-Uncertanity

The project applies weak supervision approaches to econmic policy uncertanity

Weak Supervision in Analysis of News: Application to Economic Policy Uncertainty

Abstract

The need for timely data for economic decisions has prompted most economists and policy makers to search for supplementary sources of data. In that context, text data is being explored to enrich traditional economic data sources due to its abundance and ease to collect. Our work focuses on studying the capability of textual data, in particular news pieces, for detecting and measuring economic policy uncertainty. Understanding economic policy uncertainty is of great importance to policy makers, economists and investors since it influences their expectations about the future economic fundamentals with impact on their policy, investment and saving decisions. This research tackles the data bottleneck challenge that has hindered the adoption of machine learning in measuring economic policy uncertainty from text data. We test various approaches to classifying news pieces in regards to presenting economic uncertainty content. We propose a solution involving a weak supervision approach, which expresses domain knowledge and heuristics through labeling functions. These labeling functions are used to generate probabilistic labels that can be used for training an end model without need for human annotated data.After we generated a weak supervision based economic policy uncertainty index that we used to conduct extensive econometric analysis along with the Irish macroeconomic indicators to validate whether our generated index foreshadows weaker macroeconomic performance.

Authors

Paul Trust, Ahmed Zahran and Rosane Minghim

Contacts

Paul Trust:120222601@umail.ucc.ie
Ahmed Zahran:a.zahran@cs.ucc.ie
Rosane Minghim:r.minghim@cs.ucc.ie