Central Bank Communication and Financial and Macro Variables: The Case of European Inflation Targeters

ABSTRACT

This paper investigates the effect of central bank communication on financial (interest and exchange rates) and macro (inflation and production) variables in twelve European inflation targeters in 2012–2018. After deriving the tone and topics of communications from policy releases, a heterogeneous panel SVAR is estimated. The results show that the communication effect exists mostly for financial variables; a more pronounced response was found for domestic shocks. The study contributes to the literature by covering economies not at the center of the economic discussion, as well as utilzing a model which can address different types of shocks.

KEYWORDS:

• Central banks
• communication tone
• heterogeneous panel SVAR
• monetary policy effects; international and domestic effects

JEL CLASSIFICATION:

• E58
• C33
• F41

Acknowledgments

We thank Boris Fisera for excellent research assistance. We thank participants in the ’New trends in the economics of central banking’ conference in Paris (2021) for their valuable comments.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. Like Armenia, Kazakhstan, Moldova or Ukraine. As they are the most recent joiners to the IT, time series available and policy releases are not long enough to run econometric exercises.

2. The topics for the 10 topic model with three main keywords presented in parentheses are: inflation (inflation, core, target), prices (prices, oil, food), real sphere (growth, supply, demand), labor market (unemployment, employment, wages), interest rates (rate, cut, raise, percentage) decision (decision, board, vote), financial markets (financial, market, assets), abroad (global, ECB, Fed) exchange rate (EUR, USD, depreciation), and forecast/outlook (forecast, projection, probability).

3. $\mathit{IND}$ represents the industrial production index; $\mathit{CPI}$ represents the consumer price index; $\mathit{TONE}$ represents the measure of CB communication tone; $\mathit{IBOR}$ represents the interbank interest rate; and $\mathit{ER}$ denotes the bilateral exchange rate.

4. According to Ivanov and Kilian (2005), for T = 80 and a monthly frequency, the AIC and HQC outperform other information criteria in terms of overall accuracy.

5. This outcome is present even after trimming the bootstrap draws at 2 standard deviations. We are rather reluctant to trim the underlying data or manipulate our database by some other technique; instead, we address the potential issue stemming from the underlying heterogeneity at cross-country level by removing few countries potentially sensitive toward this issue from the sample (e.g., Turkey, Georgia, and Czechia).

6. The clustering at the individual-country level applied for the benchmark model was effective when seeking to distinguish not purely policy relevant paragraphs from the monetary policy-related text. However, after the clustering procedure was finished, we could not identify the same topics for all economies. Moreover, the larger number of missing observations at the individual country level prevented us from using this approach.

7. The keywords for the last topic allowed us to classify it as decision-making with keywords such as board, committee, meeting, minutes, decision, council.

8. We also generated impulse response functions for the fourth topic without attributing a special meaning to this $\mathit{TON}{E}_{\mathit{t},\mathit{i}}$ variable. This can be interpreted as an additional robustness check. Interestingly, we did not find any statistically significant impulse response for this particular $\mathit{TON}{E}_{\mathit{t},\mathit{i}}$ variable, a sort of counter-factual exercise that can indirectly provide further support for statistically significant findings attributed to the other three topics.

9. The results of this exercise are available upon request.

10. We used Macrobond data to estimate average Brent oil prices and ECB shadow rates as estimated by Wu and Xia (2020)

Additional information

Funding

This paper is the partial result of research project funded by the National Science Center no. 2018/31/B/HS4/00164.

Notes on contributors

Maria Siranova

Dr. Maria Siranova works as senior researcher at the Institute of Economic Research of the Slovak Academy of Sciences where she serves as the head of the Department of Macro-Financial Analysis. She also works as associate professor at the Institute of Economics (Commenius University in Bratislava). Her key areas of research include topics from empirical monetary economics, international finance with a special focus on capital flows and macroeconomic imbalances, and financial geography issues.

Magdalena Szyszko

Dr. Magdalena Szyszko is an Associate Professor at WSB Merito University in Poznan, Institute of Economics and Finance, specializing in monetary policy and its effectiveness. Her studies focus on central bank practices, including communication and its effect on macro variables and expectations of economic agents.

Aleksandra Rutkowska

Dr. Aleksandra Rutkowska is an Assistant Professor at the Poznan University of Economics, specializing in econometrics, data analysis, and machine learning models. With a Ph.D. in economics, she conducts research at the Department of Applied Mathematics, focusing on time series modeling including fuzzy models and natural language processing.