What is happening with Covid infection rates?

As the ONS publishes its latest COVID – 19 insights Owl tries to make sense of what, superficially, appear to be conflicting data from: the ONS; daily published case rates from the government dashboard and the Zoe symptom app.

From the ONS:

Coronavirus (COVID-19) cases continued to increase in England and have increased in Wales and Northern Ireland in the week ending 24 July 2021, but decreased in Scotland.

The estimated percentage of the community population (those not in hospitals, care homes or other institutional settings) that had COVID-19 was:

  • 1.57% (1 in 65 people) in England, up from 1.36% (1 in 75 people) last week
  • 0.62% (1 in 160 people) in Wales, up from 0.47% (1 in 210 people) last week
  • 1.48% (1 in 65 people) in Northern Ireland, up from 0.59% (1 in 170 people) last week
  • 0.94% (1 in 110 people) in Scotland, down from 1.24% (1 in 80 people) last week

From daily case rates:

Since the end of April, daily case rates (the data given most publicity via the government dashboard) have been volatile, swinging up and then down, best illustrated graphically:

From the Zoe symptom tracker app:

This is now showing, as with the ONS data, much higher prevalence than the current reported case rates but in the past few days has indicated a turning point, again best illustrated graphically:

How might we interpret all this?

The first point to make is that there is no perfect measure of the extent of infection in the population. All  these are good data sources but each has its strengths and limitations.

For example, the ONS data come from a randomised sample i.e. includes those who are showing no symptoms but it is a lagging indicator in that it takes time to collect the samples and collate the information. So it is a week or so behind. 

Case rate reporting has a shorter time lag but is a record of those coming forward for testing i.e it misses asymptomatic and mild cases. The government list of symptoms is also regarded by many experts as being too restrictive. With all the publicity given to the “pingdemic” it is possible that some may be less likely to present for testing. It is worth pointing out that both the ONS and Zoe data indicate much higher rates (about twice for the Zoe data). 

The Zoe data has the advantage of daily reporting from a very large sample, about a million contributors, but it uses symptoms as proxy measures and the sample, although large, is self-selecting.

Pandemic data doesn’t usually move in such dramatic fashion as shown by the case rate data without an underlying intervention or event such as a lockdown. As a community moves towards herd immunity, one would expect infection rates gradually to plateau, then gradually to fall.

Big swings most likely reflect some sort of behavioural change. The problem with interpretation is that we have had a number of these, some positive, some negative: school closure; a heat wave; people gathering to watch Euro 2021 etc. There are other changes stemming from “Freedom Day” which may only just be beginning to affect the data: opening up of enclosed spaces to large gatherings such as nightclubs, pubs and cinemas.


Owl’s interpretation, for what it’s worth, is that, despite the publicity given to case rates, there is a high level of Covid in circulation. The implication of this is that incautious behaviour could trigger a rapid surge in cases, a superspreading event. However, there now seems to be evidence that, all else being equal, we may be on the threshold of some sort of herd immunity and could see a sustained gradual decline. The ONS shows numbers declining in Scotland.