Author(s): Erik Andres-Escayola, Corinna Ghirelli, Luis Molina, Javier Perez, Elena Vidal
Date published: Jan 2023
SUERF Policy Brief, No 498
By Erik Andres-Escayola (European Central Bank), Corinna Ghirelli (Bank of Spain), Luis Molina (Bank of Spain), Javier Perez (Bank of Spain), and Elena Vidal (Bank of Spain)
JEL codes: D80, C43, E32, O11.
Keywords: Economic policy uncertainty, textual analysis, press coverage, Latin American economies, business cycles.
Download: SUERF Policy Brief, No 498 (0.52 MB)
This brief investigates the role that two key methodological choices play in the construction of textual indicators: the selection of local versus foreign newspapers and the breadth of the press coverage (i.e. the number of newspapers considered). The large literature in this field is almost silent about the robustness of research results to these two choices. The authors use as a case study the well-known economic policy uncertainty (EPU) index, taking as examples Latin America and Spain. They develop EPU measures based on press with different levels of proximity, i.e. local versus foreign, and corroborate that they deliver broadly similar narratives and impulse responses. Then, they show that constructing EPU indexes based on only one newspaper may yield biased responses. This suggests that it is important to maximize the breadth of press coverage when building text-based indicators, since this improves the credibility of results. In this regard, these findings are good news for researchers, given that they provide a justification for the combined use of a larger amount of data from local and foreign sources.
A broad and expanding literature in economics uses textual analysis to develop real-time measures of economic activity, policy uncertainty or social unrest. These measures are in turn confronted with economic and financial variables to show their relevance for economic and policy analysis and forecasting.
The literature typically focuses on one broad source of text: either the local press (for single-country studies) or the foreign-based press (for cross-country studies), with an interpretation of the results that abstracts from the type of data source. For instance, the well-known economic policy uncertainty (EPU) index pioneered by Baker et al. (2016) is based on local news. On the other hand, a very recent literature has constructed newspaper-based indicators using major international news sources (mostly leading Anglophone newspapers and networks).1
There is no clear consensus on how many sources should be used to construct text-based indexes or which type of press should be considered. Furthermore: the literature has developed many newspaper based economic indicators with no guidance on this matter, with the authors following their own feelings about it. Andres-Escayola et al. (2022) has tried to put some order on this issue by answering to the following questions: (i) Do sources with different levels of proximity convey different information?; (ii) Is it better to use a broad coverage or is it sufficient to rely on only one source? These questions are relevant since newspaper repositories are usually not free of charge and the production of newspaper-based economic indicators is growing fast. The answers to these questions provide some guidance for the construction of these types of indexes.
In their application, the authors focus on the well-known and broadly used EPU indicator by Baker et al. (2016) for a number of countries, and construct this indicator for each country on the basis of, alternatively, local and foreign newspapers.2
They then compare the narratives emerging from these indexes and examine their macroeconomic effects using a structural Bayesian vector autoregression (BVAR) framework.
One possible hypothesis is that news with different levels of proximity convey different messages to the readers and this might affect newspaper-based indicators. However, they show quite the opposite. Although EPU indexes based on local and foreign news show obvious differences, they reflect the same narrative overall and deliver extremely similar macroeconomic impulse responses. This finding should reassure researchers that they can also rely on foreign sources to construct EPU indexes. This option may foster the comparability of results across countries and lay the groundwork for cross-country studies of uncertainty.
To answer the second research question, the authors show that it is often the case that macroeconomic responses to EPU indexes based on only one newspaper lie outside the credible bands of responses to the EPU index based on all sources available. Figure 1 provides some examples for the case of Spain, Mexico and Brazil. This suggests that EPU indexes based on one single newspaper may provide biased results and that it is best to rely on multiple sources when constructing these indexes. Based on this evidence, the authors advocate maximizing the breadth of the press coverage when constructing newspaper-based indicators. That is, the larger the press coverage, the better.
Figure 1: Robustness results regarding the breadth of press coverage: IRFs of GDP to EPU shocks
Note: Each panel depicts the median impulse responses (IRF) of GDP to a rise of one standard deviation in the EPU index in Spain, Brazil and Mexico, respectively. In each panel, the dashed grey line depicts the median IRF of GDP to shocks in the EPU constructed based on all available press (pooling together local and foreign sources) and the grey area depicts its corresponding 84%–16% credible set. For Spain, we consider a set of Anglophone newspapers as foreign sources. For Mexico and Brazil, foreign sources are represented by both Anglophone and Spanish newspapers. All other lines represent the IRFs of GDP to shocks in the EPU index constructed considering only one local newspaper at a time. Filled symbols indicate statistical significance within the 84%–16% credible set, while empty symbols represent not-significant estimates. The horizontal axis measures quarters since the shock.
Andres-Escayola E., C. Ghirelli, L. Molina, J. J. Pérez and E. Vidal. (2022). “Using newspapers for textual indicators: which and how many?”, Bank of Spain Working Paper N. 2235.
Baker, S. R., N. Bloom, and S. J. Davis (2016). “Measuring Economic Policy Uncertainty”, The Quarterly Journal of Economics, 131(4), pp. 1593-1636.
Barrett, P., M. Appendino, K. Nguyen, and J. de León Miranda (2020). “Measuring Social Unrest Using Media Reports”, IMF Working Papers 2020/129, International Monetary Fund.
Caldara, D., and M. Iacoviello (2022). “Measuring geopolitical risk”, American Economic Review, 112(4), pp. 1194-1225.
Mueller, H., and C. Rauh (2018). “Reading Between the Lines: Prediction of Political Violence Using Newspaper Text”, American Political Science Review, 112(2), pp. 358-375.
Ahir, H., N. Bloom, and D. Furceri (2019). “The World Uncertainty Index”, Working Papers 19-027, Standford Institute for Economic Policy Research.
About the authors
Erik Andres Escayola is a research analyst in the External Developments division at the European Central Bank. Previously he worked as a Research Assistant in the International Economics and Euro Area Department at the Bank of Spain focusing on analysing Latin American economies. He holds a MSc. in Economics from the University of Mannheim and BSc. in Economics from the University of Barcelona.
is a research economist in the Economic Developments Department of the Bank of Spain (Macroeconomic Modelling Unit). Before joining the Bank of Spain in 2017, she worked as a research fellow at the Joint Research Centre of the European Commission and as a Post-Doctoral researcher at the University of Louvain la Neuve (Belgium). She holds a Ph.D. from the University of Ghent (Belgium). See https://sites.google.com/site/corinnaghirelli/home
Luis Molina is a senior economist in the International Economics and Euro Area Department of Bank of Spain, and he is currently specialised in the analysis of Latin American economies and of emerging markets. He holds a MSc in Public Sector Economics from Institute for Fiscal Studies (Ministry of Finance of Spain). His research areas include macroeconomic analysis, emerging markets, economic development and growth, and the effects of social and political strains on economic performance.
Javier J. Pérez
is Director of the International Economics and Euro Area Department of the Bank of Spain, and member of the International Relations Committee of the Eurosystem. Before joining the Bank of Spain in 2008, he worked at other institutions, including the European Central Bank, the University Pablo de Olavide of Seville (Spain), the Research foundation centrA (within the Spanish regional public administration), and the University Complutense of Madrid. He holds a Ph.D. in Economics (with distinction) from this latter University. Javier has also been member of the Monetary Policy Committee of the Eurosystem, and of its substructure dealing with public finances, the Working Group on Public Finance. He contributes regularly to policy and academic conferences and publications (see: http://bit.ly/JavierJPerez
Elena Vidal is an economist working as secondee in Latin American desks at the OECD. Previously she worked as an analyst in the International Economics and Euro Area Department at the Bank of Spain focusing on analysing Latin American economies. She holds a MSc. in International Economics from the UNED, MSc. in Business from the Université Paris Dauphine and BSc. in Business and Economics from the University Autonoma of Madrid.
Examples include the political violence index of Mueller and Rauh (2018), the geopolitical risk index (Caldara and Iacoviello, 2022), theWorld Uncertainty Index (Ahir et al., 2019) and the Reported Social Unrest Index (Barrett et al., 2020).
When considering foreign sources, they select the major international newspapers produced in the US, the UK and Canada, since the Anglophone press is the most standard international source used in the literature.