Essays on Uncertainty and Communication in Monetary Policy

Abstract

If you tell a human being what will happen the next day, they will prepare for it. This simple idea has now become a cornerstone of economic policy. However, whether these preparations are crowned with success depends on many factors. First it is necessary that we understand what is going to happen. For this, good communication is inevitable. But what constitutes effective communication is much more challenging to answer. It depends on the sender and receiver of the message and the circumstances and forms it is conveyed. Second, the message that is sent out must be credible. Announcing an unrealistic target or not delivering on a commitment can undermine this credibility. This then effectively leads to a loss of communication capability in the future. In monetary policy, a central bank faces exactly this problem. Due to the financial crisis, many central banks have shifted their monetary policy to influencing the expectations of financial market participants and thus of society through forward guidance and purchase programs. Complex measures must be prepared appropriately for the target group and communicated via various channels. At the same time, it is of utmost importance that the announced measures are also implemented. However, because these measures have only been in existence for 15 years, we are still lacking a detailed understanding of the impact channels in some cases. We know that the announcement is at least as necessary as the actual implementation of the measures. As soon as an announcement is made about the future, society tries to prepare for it and thus anticipates the effect. However, the literature shows that there are still many unanswered questions, as sometimes the central bank’s intention differs from the markets’ reaction. This dissertation examines different questions based on four essays on the intersection of expectations, communication and how both can be measured empirically. I use established and new methodologies to empirically test the effects of monetary policy. More specifically, in the first essay, I use high-frequency data to measure the reaction of financial markets to European Central Bank (ECB) decisions and to quantify the macroeconomic effects. The second paper takes this idea forward and shows that these effects depend on the uncertainty at the time of the announcement. The third paper focuses on how central bank texts can be measured empirically. The final paper demonstrates the existence and explores the effects of a previously unknown monetary policy surprise in the euro area, triggered by a break in central bank communication style. My first paper examines the effect of ECB measures. Jens Klose and I use a VAR model to show the difference between conventional interest rate policy and communicative measures. We estimate monetary surprises for conventional, communication and asset purchase measures from the change in financial market variables during the announcement of ECB measures. Using these, we estimate a vector autoregressive model identified by external instruments. It shows that expansionary conventional and asset purchase measures lower the interest rate. This does not hold for communication measures: Contrary to expectations, expansionary measures do not increase inflation. By subdividing the surprises, we can show that the effect can be explained by the pattern of information shocks known from the literature. The comparison between the measures shows that the phenomenon does not occur with conventional measures. My second paper returns to the topic of information shocks. The literature speaks of an information shock when stock prices fall (rise) together with an expansionary (restrictive) shock. While it is evident that this pattern exists, the theoretical explanation is disputed in the literature. I incorporate uncertainty into the analysis of monetary policy surprises and can thus show that uncertainty can explain the observed pattern at the time of the decision. When uncertainty is high, information shocks occur more frequently. Furthermore, I integrate uncertainty into a VAR model and can thus show that identification by uncertainty yields the impulse responses known from the literature. These findings provide possible alternative explanations for why information shocks occur and illustrate that uncertainty is essential for the effectiveness of specific monetary policy measures. In the third paper, Johannes Zahner and I address how text data can be used in the quantitative analysis of central banks. Due to the focus of central banks on communication, the analysis of texts is becoming more and more important in research. At the same time, there are advances in computational linguistics that make it possible to use language for analysis in a more nuanced way than before. The work brings both strands together: We collect the largest text dataset on central bank communication and compute and evaluate a language model adapted to central banks. In this way, we enable researchers to capture language in detail. Furthermore, we show in three economic applications how the model can be used in a classical economic context despite its many dimensions. First, we demonstrate that the goals of central banks are reflected in their texts. Next, we investigate the effect of communication similar to Mario Draghi’s ’whatever it takes’ speech and show that credit spreads can be lowered in periods of high uncertainty. As a final application, we show that central bank speech is not free of social stereotypes. We find a gender bias, which is, however, declining due to a change in central bank communication. In my last paper, I focus on identifying monetary shocks in the European context. In 2016, the European Central Bank changed its communication structure and integrated the asset purchase programmes into its press release. I can show that due to this modification, two Quantitative Easing (QE) surprises occur from 2016: One during the press release and one during the press conference. Adapting the methodology established in the literature allows me to identify and compare the shocks. Interestingly, a difference emerges between the shocks. The short, preformulated message has a significantly more potent effect on stock prices than the shock during the press conference, where the communication is more spontaneous and detailed.