How easy is it to understand this sentence you are currently reading? How easy it is to understand this sentence that has dependency arcs that are longer that make it more difficult to read? How about if my writing is magniloquent? Or what if I use normal words? Writing style matters for how easy it is to read text. This post asks if writing style can influence how long markets take to digest Bank of England monetary policy information. I find that Bank of England publications that summarise their content in the first sentence, and use less unexpected vocabulary, are associated with a faster time for swap markets to reach a new equilibrium price following the publication release.
The Monetary Policy Report (MPR), Minutes and other publications have material effects on asset prices (Hansen, McMahon and Tong (2019). But these moves in asset prices may take hours (or days) to materialise. The November 2021 MPR was 56 pages long. That publication was released simultaneously with the Minutes, which was 15 pages long. Subsequently, there was an hour long Q&A, the text of which was 14 pages long when transcribed. In other words, markets received a deluge of information. That information will only be fully reflected in asset prices when market participants have had time to read and digest the publications.
A discussion of what the Bank of England’s Monetary Policy Committee (MPC) chooses to say in these documents is well above this author’s pay grade. It is the result of a long process of deliberation by the MPC and staff. The content of that discussion, the outcome of the MPC’s decision, and the reasons behind it, are taken as fixed.
How the MPC chooses to communicate is a different issue (and indeed has been discussed on this blog before). This post asks if writing style can influence how long markets take to digest Bank of England monetary policy information. In other words, if the Bank of England writes more clearly, does that lead to a faster time for market prices to move to a new equilibrium?
Measuring post-publication market dynamics
How long market participants take to digest the Bank of England information is calculated by measuring the time it takes market prices of the 2 year sterling Overnight Index Swap (OIS) rate to stop moving following the publication(s).
Technically, I define a new equilibrium price to have been reached when the total change in price over an hour is statistically indistinguishable from zero.
I consider the release of the MPR (and any concurrent publications such as the Minutes) that occurred from 2009, when OIS data becomes available, to the end of 2019, when the available textual data ends.
Chart 1 shows how long it takes after each MPR (and Q&A) for the OIS market to settle.
Chart 1: Time for 2 year sterling OIS market to reach a new equilibrium following an Monetary Policy Report release
The joy of text
Clear writing has many aspects. The small sample means I can only test a few hypotheses as to what kind of writing is associated with quicker market reactions.
Luckily, previous work can act as a guide to what might be important. In an empirical exercise, Munday and Brookes (2021) find several aspects of central bank text to be significant for whether a communication is reported on in newspapers. These aspects of writing seem to matter because they make the text more readable.
The aspects of readability I consider in this post are:
- Average word prevalence. This measures how well known the words are in the publications. It is measured using the dataset of Brysbaert, Mandera, McCormick, and Keuleers (2019).
- Contextual expectancy score. This measures how expected (or unexpected) a word is given its context. This is measured using Spacey’s word vector engine.
- Dependency arc length. Dependency arcs map the relationships between words that readers must manage whilst reading a sentence. When reading a sentence, we process each word incrementally over time. Sentence structures that hamper this process make reading more difficult. Munday and Brookes (2021) find that long dependency arcs are related to reduced news coverage of Bank communication.
- Headlining score. This is a measure of to what extent the first sentence of a document summaries its contents. It is measured by the doc2vec similarity (Le and Mikolov (2014) of the first sentence and the rest of the document.
- Number of sentences. This measures how long the Bank of England’s text is.
I run a standard OLS regression of the length of time the market takes to reach a new equilibrium against the textual features outlined above.
The regression also includes controls for whether a monetary policy decision occurred on the day of the publication(s), and on the initial (30 minute) market surprise on the release of the Monetary Policy Report. Both of these are plausibly related to the time it takes for the market to reach a new equilibrium price, because they capture the Bank’s immediate monetary policy decision. I condition on them in order to focus on the communication component of the Bank’s information release.
Chart 2 shows the coefficients of the estimated regression. Each bar shows, all else equal, the association between one standard deviation increase in a feature, and the time the market takes (in seconds) to move to a new equilibrium. The black bars represent 95% confidence intervals. This regression is not without issues (discussed below), so these results should be interpreted as associations, rather than causal relationships.
There are two features that are significant at the 5% level and two at the 10% level.
Documents with higher contextual expectancy, first lines that summarise the entire document, words that are more prevalent, and are published on days without a monetary policy decision are associated with a shorter time for the market to reach a new equilibrium.
The length of dependency arcs, the initial market reaction, and, interestingly, the length of the document, do not display any association with the time taken for the market to digest the Bank’s information.
Chart 2: Effect on time to new equilibrium of writing style (regression coefficients with confidence intervals)
Without wanting to over-interpret these results, it does seem like particular aspects of writing style are important for market understanding.
The above analysis comes with several caveats, and so our results should be read in with them in mind.
Only correlations between some (handpicked) textual features and how long it takes for the market to settle have been presented. And, of course, correlation doesn’t imply causation. Indeed, there are plausible omitted variables: one could argue that if the Bank of England has a more complicated message to convey, it must write in a more complicated style.
Furthermore, the estimates of how long it takes the market to digest communication are simple, and influenced by news releases that occur after the publications (although these should only add noise to the estimates, not bias them).
Finally, the small sample does mean that the regression lacks power. Coefficients that just dip under a 5% or 10% significant level should not be over-interpreted.
These caveats notwithstanding this is initial evidence that writing style matters, adding to the existing body of work on this topic from the Bank of England (Haldane and McMahon (2018); Bholat et al (2018). Of course content matters, and the Bank of England’s message is of paramount concern when drafting communication. But, at the margin, when that message’s substance has been formed, the style it is presented in can help the market to understand it quicker.
Timothy Munday worked in the Bank’s External Engagement Division in Monetary Analysis when he authored this post.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.
If you want to get in touch, please email us at email@example.com or leave a comment below.