RDP 2021-05: Central Bank Communication: One Size Does Not Fit All 1. Introduction

Central banking used to be a rather secretive business. As Janet Yellen (2012) noted in a speech ‘In 1977, when I started my first job at the Federal Reserve Board … it was an article of faith in central banking that secrecy about monetary policy decisions was the best policy’. But times have changed. There has been a gradual evolution towards greater transparency in central banking practice and an increasing emphasis on the quality of communication (Eijffinger and Geraats 2006; Filardo and Guinigundo 2008; Dincer and Eichengreen 2009).

This evolution has been driven by two primary forces. The first is a literature that has highlighted how expectations are central to the efficacy of monetary policy. If central banks communicate effectively, markets are able to anticipate the policy implications, and should respond to information contained in new economic data when it occurs rather than disruptively when central banks make policy announcements (Hawkesby 2019). Reflecting this, countries that are most effective in this practice tend to experience less interest rate volatility and smaller reactions to monetary policy changes (Blinder et al 2008).

The second force is the transition towards independent central banks and the associated need for transparency in support of democratic accountability (Blinder et al 2001). As independent public institutions, central banks are ultimately accountable to the public. To support this accountability they need to reveal enough about their analysis, actions and internal deliberations so that interested observers can understand each monetary policy decision as part of a logical chain of decisions leading to some objective and, thereby, assess their performance (Woodford 2005; Bernanke 2010; Preston 2020).

Central banks do, however, confront a trade-off. Monetary policy is not simple. A comprehensive explanation may not be simple or easily understood, but a simple explanation may not be accurate. In practice, central banks have tended to provide more explanation over recent years. For example, the length of the RBA's Statement on Monetary Policy (SMP) has increased from just over 10,000 words, in its first iteration in 1997, to over 30,000 words in the latest issues. A review of the Federal Reserve Bank's communication also reveals a general trend towards longer and more complex documents, to the point where the reader requires a university-level education to understand the content properly (Davis and Wynne 2016; Haldane 2017). A related challenge is that it can sometimes be difficult to decide who a central bank's audience is. Is it, for example, market economists who spend their time interpreting central bank actions for their financial market clients? Or is it the general public who are deciding whether to take out a mortgage or to invest in some new equipment for their business? Or companies and unions deciding how they will approach wage negotiations?

Despite the increased emphasis on communication and the many questions in the area, there has been relatively little study of the communication quality of central bank documents and even fewer answers on what makes for effective communication in this area.^[1] To fill this gap we analyse central bank communications using surveys and a novel application of machine learning techniques. We focus on how different audiences, in particular economists and non-economists, perceive the readability of and the degree of reasoning in various economic communications. Ours is the first work that attempts to measure the degree of reasoning in central bank communication and, consequently, also the first that considers the relationship between readability, reasoning and audience. We discuss the reasons for this particular delineation, and related work, in more detail in the next section.

We have three main results. First, we find that simple readability indices, such as the Flesch–Kincaid (FK) grade level, are barely correlated with individuals' survey ratings of text readability. This suggests that a focus on simple readability metrics, as is common, may fail to increase a broad measure of readability. Furthermore, we find that the readability of a text is generally uncorrelated with the degree of reasoning in the text. Thus, a focus on readability alone runs the risk of undermining the achievement of transparency to the extent that it de-emphasises the importance of the content of a document.

Second, the way people comprehend a document depends on their knowledge of economics. We find that there is no correlation between the way the economists and non-economists in our sample perceive the readability and reasoning of the same piece of text. Our machine learning results reflect this observation: the textual elements associated with more readable paragraphs vary between economists and non-economists. Put another way, one needs to emphasise different techniques when writing for economists and non-economists; one size does not fit all.

Finally, there seems to be a trade-off between readability and reasoning – at least within a given paragraph. We find, for example, that the introductions of documents tend to be more readable but contain less reasoning while conclusions tend to be less readable but contain more reasoning. This highlights that different parts of a document have different objectives and it would be difficult to achieve these multiple objectives with a single style of writing within a single paragraph. Importantly, the application of any single metric to an entire document will fail to capture the need for different emphases at different points. Consequently, we emphasise that text quality metrics – including our own – should be used to inform rather than prescribe.

While this is the first study that we are aware of that systematically assesses these 3 aspects of central bank communication, our results are individually unsurprising. In other respects, however, they are new. In particular, we have not seen any previous work that considers the various aspects of effective central bank communication simultaneously and, in particular, that acknowledges the trade-offs that exist between the various individual objectives. No one document or style of writing is best for every paragraph, audience or communication objective.

Footnote

There is, of course, a huge literature on effective communication in general. [1]