menu
close

Author(s):

Olivier de Bandt | Banque de France
Jean-Charles Bricongne | Banque de France
Alan Cuzon |
Julien Denes | Banque de France
Annabelle de Gaye | Banque de France
Rémi Lefafta | ACT-ON Data
Marwan Menaa |
Pierre-Antoine Robert | Banque de France

Keywords:

inflation , Natural Language Processing , households and firms , expectations , Machine Learning

JEL Codes:

C53 , C55 , D84 , E31 , E58

This policy brief is based on Banque de France Working Paper No921. Working Papers reflect the opinions of the authors and do not necessarily express the views of the Banque de France or other institutions of affiliation.

The rise in inflation since 2021 could affect households’ and businesses’ perception of inflation. Our paper applies Natural Language Processing techniques (NLP) to the quasi-universe of newspaper articles for France, concentrating on the period 2004-2022, in order to measure inflation attention as well as perceptions by households and firms for that country. The indicator, constructed along the lines of a balance of opinions, is well correlated with actual HICP inflation. It also exhibits good forecasting properties for the European Commission survey on households’ inflation expectations, as well as overall HICP inflation. The method used is a supervised approach that performs better on our data than the Latent-Dirichlet-Allocation (LDA)-based approach of Angelico et al. (2022). The indicator can be used as an early real-time indicator of future inflation developments and expectations.

Central banks monitor inflation using a variety of indicators

The price stability objective of central banks, and in particular the ECB, means that they must be able to assess not only current and future inflation developments but also the perception of inflation of economic agents. As regards current inflation, the consumer price indices published by national statistical institutes such as INSEE are closely analysed, be it for headline inflation, its components or measures excluding the most volatile items. Projections are also conducted to assess the inflationary or deflationary risks to the economy.

Maintaining price stability also requires that the inflation expectations of economic agents, households or companies, remain anchored around the inflation target set by the central bank. To measure inflation expectations, several indicators are used: i) expectations from financial markets, measured using inflation-linked bonds or inflation derivatives, ii) expectations of forecasters based on surveys (e.g. the Consensus Forecast or the Survey of Professional Forecasters), iii) expectations of companies (e.g. in France, the new survey of business leaders conducted by the Banque de France) and iv) expectations of households from surveys for example the European Commission’s consumer survey or the European Central Bank’s Consumer Expectations Survey (CES). The analysis in this post focuses on the way inflation is perceived through certain forms of media and the short-term expectations that can be derived from this.

Text mining of alternative data to complement traditional sources

The media disseminate a wealth of data that may reflect the inflation perceptions and expectations of households and businesses. This includes traditional media such as television, radio and print or social networks. In the case of the press, processing this volume of data requires the use of data science techniques that analyse qualitative textual data and transform them into quantitative figures that can be used by economists in the form of indicators.

A recent work in this area, carried out at the Banque de France on French data, collects and analyses more than one million articles from the written press since 2003 (source Factiva) to construct indicators of perceived inflation (in the spirit of the work of Angelico et al. (2022) for Italy using Twitter).

The method is based on the selection of articles using keywords (related to the semantic field of “inflation” or “prices”). Filtering and classification algorithms are used to select only those that actually deal with inflation and not other subjects (e.g. literary ‘prizes’) and to derive the direction of inflation/prices that may be mentioned in the text (rising, falling or stable). Extensions have also been explored, to exclude articles that might reflect the views of central banks themselves, and focus more on households.

Chart 1 shows the simple observation of the number of articles, which reveals interesting dynamics. After filtering, we get a first indicator of “inflation attention”, measuring periods were the press dedicates more space to inflation developments. While the number of relevant articles is correlated with the inflation cycle, we can observe different spikes in 2015-2016 while inflation was decreasing. There were actually signals in press articles highlighting lower pressures on inflation.

Chart 1: Monthly number of articles on price developments after filtering and HICP evolutions: Intensity indicator

Sources: Factiva, Eurostat, authors’ calculations.

Textual indicators offer different information and sometimes leading inflation data

Other indicators, measuring the direction of price changes, are also calculated, based on the number of press articles mentioning a rise or fall in prices. These are “balances of opinion”, along the lines of business surveys, defined as (number of increases – number of decreases, or stability)/(total number of increases and decreases, or stability). These indicators can thus capture perceptions about current or future inflation. The Press indicator yields trends that are relatively close to those of observed HICP inflation and to the survey-based Household inflation expectations (see Charts 2.1 and 2.2).

Chart 2.1: Inflation Perception Press indicator (LHS) and HICP (%, y-o-y, RHS)

Sources: Factiva, Eurostat, European Commission, authors’ calculations.

Note: The Press and Household Survey indicators are constructed as opinion balances.

Chart 2.2: Inflation Perception Press indicator (LHS) and survey-based Household inflation expectations (1-year ahead, RHS)

Sources: Factiva, Eurostat, European Commission, authors’ calculations.

Note: The Press and Household Survey indicators are constructed as opinion balances.

Simple correlation indicators point to a relatively high level of correlation over the sample April 2004 – August 2022 with HICP inflation and forecasts of professional forecasters (as given in the Consensus Forecast survey), respectively 77% for HICP and 67% for Consensus Forecast. Correlations seem a bit higher when benchmark inflation variables are taken with a lead, between 72% and 78%, signalling potential forecasting properties.

Textual indicators outperform traditional indicators to forecast households inflation expectations

Forecasting properties of textual indicators are analysed, by implementing different sets of regressions, with households inflation expectations as a dependent variable. Besides textual indicators, explanatory variables include inflation from the Consensus Forecast survey, inflation-linked swaps and lagged HICP (harmonized index of consumer prices). Oil growth rates are also controlled for. Whether all variables are included, or only the most significant variables are selected with automatic selection algorithms, textual indicators are always the most significant variable, outperforming other traditional variables, as done in Angelico et al. (2022).

The results obtained from text mining offer many opportunities

Using attention and directional indicators from the written press, inflation can be monitored on an infra-monthly basis, with rapidly available results (within a few days). The contribution of such indicators opens up interesting avenues of research on changes in the general inflation and price environment. For example, as shown by Korenok et al (2022), they could also help detect changes in the attention paid to inflation by economic agents.

About the authors

Olivier de Bandt

Olivier de Bandt is currently Director for Research at the Banque de France. He held various other management positions at the Banque de France (Director of economic forecasts, of international economics and cooperation) and at the ACPR (Director for research and risk analysis). He has published numerous articles in international academic journals in the fields of macroeconomics, financial economics, insurance economics, systemic risk and the effects of climate change. He taught as an associate professor at the University of Paris-Nanterre. He holds a PhD from the Department of Economics at the University of Chicago and graduated from Sciences Po Paris.

Jean-Charles Bricongne

Jean-Charles Bricongne is currently Deputy-Director in the Microeconomic and Structural Analysis Department in Banque de France. He graduated from the Ecole Centrale Paris, from Sciences Po Paris and holds a PhD from the University of Paris I Panthéon-Sorbonne. He previously worked as a seconded national expert on internal imbalances in the European Commission (DG ECFIN), especially on credit and housing. He was before counsellor of the General Director of Economics and Research, Head of a unit in charge of trade and competitiveness in Banque de France, and seconded expert on monetary and financial issues in the French statistical institute (INSEE). He has been a Senior Lecturer at Sciences Po Paris, an Adjunct Professor in Tours and Orléans Universities and is now an Adjunct Professor in Paris 1 Panthéon Sorbonne University. He is an associate researcher in LEO (Laboratoire d’Economie d’Orléans). He has written on various topics in applied trade analysis, globalization statistics, financial macroeconomics, housing, migrations, the links between financial and real economy, and datascience and economics.

Alan Cuzon

Alan Cuzon is a data engineer specialising in the design of data pipelines and the development of innovative tools to optimise analyses. He has also been involved in data science projects in the banking and healthcare sectors.

Julien Denes

Julien Denes was a Data Scientist at Banque de France, where his work focused on Natural Language Processing (NLP) and its application to creating new indicators at the service of economic research and public policy.

Annabelle de Gaye

Annabelle de Gaye is Senior Economist at the Climate Economics Unit of the Banque de France. Her research and policy work focuses on international projections, macroeconomic modelling and climate scenarios. Previously, she worked for the French Treasury at the French embassy in the US and also held Economist positions in investment banking and reinsurance companies. She holds a Master’s degree from Toulouse School of Economics and another one from the ENSAE.

Rémi Lefafta

Rémi Lefafta is a Data Analyst at ACT-ON Data, primarily focusing on HR and financial aspects. He holds a Master’s degree in Economics from the University of Tours. He has previously interned at the Banque de France as a Data Scientist, where he contributed to the application of NLP techniques.

Marwan Menaa

Marwan Menaa graduated from the University Paris Dauphine – PSL with a degree in Economic and Financial Engineering. Additionally, he holds an engineering degree from École Supérieure d’Ingénieurs Léonard-de-Vinci (ESILV). He has completed internships at reference institutions, notably at the European Central Bank, where he made significant contributions to the Directorate General of Monetary Policy Data Centre. Prior to that, he served as Economist / Data Scientist intern at Banque de France’s Directorate General of Statistics, Research, and International Affairs, further enhancing his skills in economics and quantitative analysis. Marwan’s expertise lies in economics, financial markets, and data science, with a particular focus on macroeconomic dynamics and inflation expectations. He employs advanced computational analysis techniques such as Natural Language Processing and Machine Learning to analyze complex economic phenomena and drive insightful research.

Pierre-Antoine Robert

Pierre-Antoine Robert is Auditor at the General Inspection of the Banque de France. He previously held an Economist position in the Business Cycles and Macroeconomic Forecasts Department of the Banque de France. He was lecturer in macroeconomic modelling at ENSAE and graduated from ENS, EHESS, ENSAE and Sciences Po Paris. His research and policy work focuses on macroeconomic modelling, forecasting and inflation expectations.

More on these topics

Tags: