USA and Euro Area Uncertainty Indices
USA Economic Policy Uncertainty Indices
Buiding uncertainty indices using unsupervised machine learning in the fasion of Azqueta-Gavaldon (2017) has opened a new venue for extensive research. The purpose of this entry is to track the research which uses the uncertainty indices or the methodology to build new ones.
Research which uses the uncetainty indices (here):
- Xie, F., 2020. Wasserstein Index Generation Model: Automatic generation of time-series index with application to Economic Policy Uncertainty. Economics Letters, 186, p.108874.
- Husted, L., Rogers, J. and Sun, B., 2019. Monetary policy uncertainty. Journal of Monetary Economics.
- Saltzman, B. and Yung, J., 2018. A machine learning approach to identifying different types of uncertainty. Economics Letters, 171, pp.58-62.
Research which uses the methodology to build new uncertainty indicators:
- Echevarria Icaza, V. 2019. Desentrañando las causas de la incertidumbre de política económica en España: una aproximación usando Machine Learning . BBVA Research. Available here.
- Crocco, N., Dizioli, G. and Herrera, S., 2019. Construcción de un indicador de incertidumbre económica en base a las noticias de prensa. https://www.colibri.udelar.edu.uy/jspui/bitstream/20.500.12008/22124/1/tg-Grocco-Dizioli-Herrera.pdf
- Azqueta-Gavaldon A., Hirschbühl D., Onorante L., and Saiz L. 2019. Sources of economic policy uncertainty in the euro area: a machine learning approach, Economic Bulletin Boxes, European Central Bank, vol. 5.
- Azqueta-Gavaldon A. 2017: Financial Investment and economic policy uncertainty in the UK. IML '17 Proceedings of the 1st International Conference on Internet of Things and Machine Learning. https://dl.acm.org/citation.cfm?id=3158380
Euro Area Economic Policy Uncertainty Indices
This section contains the uncertainty indicators produced by Azqueta-Gavaldon A., Hirschbühl D., Onorante L., and Saiz L. (2020): Sources of economic policy uncertainty in the euro area: an unsupervised machine learning approach, Working Paper Series 2359, European Central Bank .
To see in detail how the indices were constructed, please see Section 2 (page 6) of the working paper. Nonetheless, a brief summary of the steps followed is offered here:
- Collect all news-articles that contain the words "economy" and "uncertainty" (in their respective language) for each corpus (collection of news articles for Germany, France, Italy and Spain).
- Run the word embeddings algorithm Word2vec to find semantically close words to "economy" and "uncertainty" and extend the size of the corpuses by including those new terms.
- Run the LDA algorithm to cluster each corpus into topics.
- Select those topics relevant to Economic Policy Uncetainty.
- Construct each topic as a time series:
- Sub-indices: sum the topic proportions of each topic per month and divide this topic proportion by the total numer of news-articles that contain the word "today" . This step produces 8 policy-related subindices: Monetary, Fiscal, Political, Geopolitical, Trade, European Regulation, Domestic Regulation, and Energy.
- Aggregate EPU index per country: the sum of the 8 sub-indices per month.
- Aggregate EPU index for the euro area: equalty weighted sum of the Indices and sub-indices across the four countries.
Notes: All indices (aggregates and sub-indices) are standardize to mean 100 and 1 standard deviation.
Notation: EPU_EA stands for the aggregate economic policy uncertainty index for the euro area while Monetary_IT would stand for the Monetary Policy uncertainty for Italy. Moreovoer, Monetary_EA stands for Monetary Policy Uncertainty accross the euro area countries here examined.