I realised a few weeks ago that it had been quite a long time since I last checked the official online coronavirus dashboard. There were periods when I would check the death stats every night, I was even motivated enough to create my own series of graphs on this site. I suspect a lot of people got into the habit of checking the dashboard every day; how many people are still doing that?

When the country was in confinement, the weekly briefings showed up innovative ways of measuring the lockdown by looking at things like average traffic levels. Could you measure interest in coronavirus through online behaviour?

I would guess that people tend to check the online dashboard when they’re concerned about the virus, and when they’re not concerned they don’t bother looking for the latest figures.
Of course, the question ‘how many people visited this website last month’ is definitely measureable. I was sure that Public Health England had the numbers, so I sent a freedom of information request.

They responded with this letter. I wrote their figures into a csv file and made a chart.

dashboard_viewer_chart

The FOI response came with two relevant observations:

  1. In November 2020 suspended analytics for a time while migrated the service.
  2. In March 2021 implemented the new GDS policy which is not to track analytics on users unless they allow it – you can see this in the drop off in users.

There’s a big jump from June to July 2020, and from then on a steady increase to a peak of 8.4 million visitors in January 2021. After March 2021 it’s less useful, although the refusal to track users is admirable.

The Google Trends data1 for ‘covid 19’ in the United Kingdom tells a different story.

google_trends_chart

According to Google searches, the highest interest was in April 2020 and it dropped off from then onwards.2

These could both be compared with the latest cases graph, using the numbers helpfully provided by the coronavirus dashboard.

case_numbers_graph

None of these graphs really line up, so to address the purpose of this exercise, there isn’t much to learn from patterns of online behaviour here.

  1. I’m aware this doesn’t look right as a bar chart, but not going to mess around with LibreOffice any further. 

  2. If we’re being picky about bias here, you can always point out that what search terms on Google are influenced by the search autocomplete algorithm.