Every day from the 25th March until late June, the UK government has given briefings and shown publicly available data on the Covid-19 crisis. In this article, I will analyse how the graph on Global Death Comparison has been described and commented on during these briefings, what words were used and therefore what message the government was trying to convey. All briefings are available on BBC iPlayer and all slides are available on Gov.UK. The rest of the work has been my own.
This analysis was included in the paper “Cross-Country Comparisons of Covid-19: Policy, Politics and the Price of Life” (Balmford et al. 2020) published in The Journal of Environmental and Resource Economics (ERE), which can be found here.
The UK government broadcast from Number 10 offered an official update on the progression of the Covid-19 crisis in Britain. For a little over a month, between the 30th of March and the 9th May, these briefings included a Global Death Comparisons graph, which plots cumulative deaths attributed to Covid-19 across different countries, from the date of the 50th death in each country onward. During the briefing, a government spokesperson, often a scientific adviser, would describe and comment on the graph.
The focus and importance given to this section of the briefing changed over time, as did the words used to describe and interpret the graph. The following textual analysis will show that initially the graph was described with care, and shown as reliable evidence that the UK was in line with other European countries. As deaths in the UK increased more dramatically than elsewhere, the tone of the briefings changed, with the spokesperson frequently implying the data shown was unreliable and did not offer an accurate comparison. While such a change in tone may have been unintentional, it acted as a defence of the UK government handling of the Covid-19 crisis.
I started by transcribing the relevant sections, i.e. for as long as the graph was shown, in each daily briefing. All videos are available on the Government’s website and on BBC iPlayer and all slides are available on gov.uk. As a starting point for the transcription I took all verbal cues that indicated the speaker was moving on to the Global Death Comparison (GDC) graph, e.g.: “Next slide, please”, “As the next slide shows […]”. Since the GDC graph was usually the last slide in the slide pack, most transcripts end when the speaker either stops speaking or thanks the Minister leading the briefing that day. These snippets last on average between 30 seconds and 1 minute. All transcripts are available as Supplementary Material to the ERE paper here.
After producing 39 discreet .txt files, each containing the transcript for one daily briefing, I analysed them using the software AntConc, developed by linguist Laurence Anthony. This software allowed me to search for single words, clusters of words, and even word pairs which were used together frequently. For a complete tutorial on how to use AntConc, please visit the website linked above. The large majority of the words I searched for, I had already noted down as potentially interesting during the transcription phase, while others I identified along the way.
The transcription process was incredibly helpful as it allowed me to identify general trends that became apparent by reading all snippets of text in close proximity. During this process I noted down words or word pairs that I thought were particularly interesting, and grouped them in three main trends/semantic groups: initial positivity, comparisons are hard, a new better graph. I will explore these trends as I illustrate how the graph itself changed throughout its 39 days of air time.
I encourage you to read at least the first couple of pages of the Appendix and see if you can spot an initial trend. You’ll notice a lot of speakers don’t really focus on the graph, but rather either try to reassure the viewers, or reiterate the government’s initial message (“Stay at Home, Protect the NHS, Save Lives”). When they do focus on the graph, they spend quite a lot of time simply describing the graph, for example pointing out the line representing the UK and describing it in relation to other countries. Most importantly, they depict the UK as being part of a group of European countries all moving forward on the same path. One speaker even states that this comparison can be used to learn from other countries (Dr Jenny Harries, 05/04).
In order to check if my hunch was right, I searched for:
- Group: group, grouping, band, cluster, close, closely,
- Track: track, tracks, tracking, tracked, on track
- Tuck: tuck, tucks, tucking, tucked
- Sameness: same + situation, path, framing, etc.
The numbers are very small, granted, but a decreasing trend is identifiable nevertheless.
It is worth noting that, from the GDC graph’s first appearance on 30th March until 15th April, the UK line in the graph accounted only for deaths in hospitals. While there are some caveats written (in very fine print) on the graph itself – e.g.: that the data presented is lagged and only limited to one setting – these are never extensively voiced out by the spokesperson describing the graph. That is, not until around the 16th April, where the graph starts to show two lines for the UK, one for hospitals only and one for all settings.
This change in data reporting came as the novel coronavirus was ripping through care homes, potentially due to the government’s criticised mismanagement of Covid-19 cases in elderly patients (Discombe 2020; Grey and MacAskill 2020). Indeed this change came in reply to the demand for more inclusive and accurate reporting. Incidentally, around that time the speakers also started mentioning quite often that the data provided for all settings was severely lagged, implying it should not be taken at face value. This was due to the fact that the “all settings” data comes from the Office of National Statistics, which screens death certificates to check for mentions of Covid-19, rather than directly from hospitals (i.e. the NHS).
Shortly after, on 29th April, the graph changes again, now showing only one line for the UK: all settings . This happens around the time the UK registers the 2nd highest number of deaths in the world and it is now above all other European countries in the graph, below only the US. The graph now has a note pointing out that, “Different countries have different methods of counting Covid-19 deaths which means it is difficult to compare statistics across countries” and a short list of countries on the right, clarifying whether their data includes care homes deaths or not. Ironically, as is made clear in the full ERE paper, it is actually the UK which appears to under-report Covid deaths the most.
Around the same time (end of April) most speakers started including a caveat in their description of the graph, noting how comparisons are difficult due to the fact that different countries measure coronavirus deaths differently. Moreover, they pointed out the data presented are just crude numbers and not rates (e.g deaths per million people), and therefore more difficult to interpret. To demonstrate this trend, I searched for:
- Comparison,comparisons, compare, comparative in the same sentence as difficult, difficulty
- Interpret in the same sentence as difficult, difficulty
- Different measures, different systems, different methods (and similar expressions)
- Number, numbers in the same sentence as crude, absolute
- Number, numbers in the same sentence as rate, rates
- Caveat, caveats
These are obviously small numbers, but the graph above clearly shows that during most of April, when the UK was shown below other European countries on the graph, the speakers did not think necessary to mention that comparing deaths across countries is difficult. The idea that comparisons are difficult spikes in late April and is then repeated almost daily up until the very last day.
The solution to this difficult comparison dilemma presents itself shortly after. Speakers start mentioning a new graph that is often referred to as “excess death” or “excess mortality”. Importantly such data is unaffected by country-specific testing regimes. This graph would provide a better comparison and therefore a clearer understanding of the global situation. According to the spokespeople, however, this will not be available for some time (some mention a few months to a year). To demonstrate this I searched for:
- Excess death, excess deaths, excess mortality, all cause mortality, all mortality
- Size, sizes, population, populations in the same sentence as adjust, adjusted
- Comparison, comparisons in the same sentence as true, accurate, real
- Clearer understanding
Again, this concept seems to surface in late April and then spike in early May, and is then mentioned in 5 of the last 7 briefings.
Combining both data on difficult comparisons and the need for a better metric together, the result is striking:
Lastly, on 10th May, the UK Prime Minister gives a speech introducing the government’s plan to ease lockdown. On that day he does not show GDC data and afterwards the graph is no longer shown.
This analysis revealed two macro trends. Initially, the government’s spokesperson presenting the GDC graph placed a lot of importance on the graph itself, focusing on describing the UK’s position compared to other countries and highlighting the common path all countries were following in cumulative deaths. This could have been aimed at reassuring the public that the high numbers of infections and deaths in the UK were in line with the course of the pandemic elsewhere in the world, and therefore inevitable. During this initial period the speakers did not shy away from using this graph to make a comparison, describing the UK’s position as on the same path as all other countries. This comparison was presented as reliable and positive.
Towards the end of April, however, we can see a shift in how this graph is presented. This could be attributed to two main factors: first, the dramatic increase in deaths in the UK, which reached second highest in the world (behind only to the US); second, the inclusion of “all settings” data in the graph, which brought the public’s attention to the care home crisis. As a result of this, speakers start what could be described as a discrediting campaign, pointing out almost daily how such comparisons are unreliable, discouraging interpretation of these data, and suggesting the existence of another, accurate and true comparison that would however (conveniently enough) not be available any time soon.
To be clear, I am not suggesting that simply contrasting officially recorded deaths is the best comparison. A more effective way would indeed be comparing deaths adjusted by population (which they do show, but only twice), or comparing excess deaths (as in the Financial Times, The Economist and the ERE paper). However, I still find it very interesting how this is only pointed out by the speakers once the graph starts showing the dramatic increase in UK deaths, while before that it was assumed such comparisons were not only reliable, but also something the UK could learn from.
This textual analysis allowed me to quantitatively evidence a change in attitude on the government’s part towards the Global Death Comparison’s graph.
Not unlike Snowhite’s Evil Queen, the UK government is happy to show and endorse the Global Death Comparison graph for as long as it makes the country appear in line with the rest of Europe. As soon as the graph starts showing a different picture – one where the UK is on a separate, steeper trajectory than the rest of Europe – the UK government’s spokespersons swiftly start discrediting the graph and redirecting the public’s attention to another, indefinitely unavailable graph. This discrediting strategy culminates in the removal of the GDC graph from daily briefings after the announcement of lockdown easing measures on 10th May.
This change in attitude seems to fall within a broader strategy from the UK government to defend its management of the pandemic – which has been criticised throughout – to the point of cleansing and censuring the information made available to the public.
I hope studies similar to this one will be able to provide the data necessary to shed light on how the UK government’s managed the pandemic and how lessons can be learned for the future. A cross-country analysis of how different countries spoke about global comparisons, and its correlation to the pandemic’s course in those countries, may well prove fruitful.
Anthony, L. (2019). AntConc (Version 3.5.8) [Computer Software]. Tokyo, Japan: Waseda University. Available from https://www.laurenceanthony.net/software
Balmford, B., Annan, J.D., Hargreaves, J.C. et al. (2020) Cross-Country Comparisons of Covid-19: Policy, Politics and the Price of Life. Journal of Environmental and Resource Economics
Discombe, M. (2020) Government has misled public over UK deaths being lower than France (Online). Available from https://www.hsj.co.uk/coronavirus/government-has-misled-public-over-uk-deaths-being-lower-than-france/7027404.article. Accessed 14 June 2020.
Grey, S., MacAskill, A. (2020) Special Report: In shielding its hospitals from COVID-19, Britain left many of the weakest exposed (Online). Available from https://www.reuters.com/article/us-health-coronavirus-britain-elderly-sp/special-report-in-shielding-its-hospitals-from-covid-19-britain-left-many-of-the-weakest-exposed-idUSKBN22H2CR. Accessed 14 June 2020.