Skip to content
  • Blog
  • 10th Nov 2020

A small number of people account for a large amount of coronavirus risk

Results from a survey of 3,702 English adults

The UK’s new national lockdown is expected to run until at least early December when the 3 tier risk system, which classifies different parts of the country as being Medium, High, or Very High risk for coronavirus, is then expected to return. 

We know public understanding of the three risk tiers is reasonably good. But understanding the guidance is one thing –  acting accordingly is another. 

Once lockdown ends, how compliant should we expect people to be with the guidance, in terms of limiting their number of risky social contacts? To estimate this, we ran a (pre-lockdown) survey about the social activities of 3,702 adults in England. 

How we measured risk

To create a measure of coronavirus transmission risk, we first asked our respondents how many different people they had met outside their household (or support bubble) in the previous 7 days (with the presumption that more meetings = more risk).

We then asked for details about these encounters, specifically:

  • Whether the meeting happened indoors or outdoors (indoors = more risk)
  • How long the meeting lasted (longer meetings = more risk)
  • Whether they kept their distance during the meeting (closer = more risk)
  • Whether they wore masks during the meeting (no masks = more risk)

To quantify the relative contribution of each factor, we applied a weight system similar to that used by MicroCovid, a COVID-19 risk calculator for daily activities. We estimated that transmission risk was:

  • 10 times higher for indoor vs outdoor meetings
  • 5 times higher for 30min vs 10min meetings
  • 5 times higher for meetings where people did not vs did stay at least 1m apart
  • 5 times higher for unmasked vs masked meetings

So, someone who recently met only 1 person outdoors for 10 minutes, with both parties wearing masks and staying more than 1m apart, would get a low risk score. Someone who sat indoors for 2 hours next to 5 other people, with no one wearing masks, would get a high risk score. A high risk score doesn’t necessarily mean a person is breaking the rules. For example, someone could follow the ‘rule of six’ by meeting 2 others once (lower risk), or by meeting different groups of 3-4 people several times a week (higher risk). 

One caveat: these are rough weightings, based on our understanding of the best available evidence on coronavirus transmission risk, but they are subjective and could reasonably be challenged. For this reason, we also conducted sensitivity analysis on the results you are about to see, and confirmed that they are robust (e.g. the results more-or-less hold even if you assume that masks are half as effective as we think they are, or that longer meetings are riskier than we think they are). However, we welcome suggestions on what alternative weights we should use.

What we found 

Of our sample of 3,702 adults, most said they had been keeping to themselves: in the past 7 days, 75% had either met no-one or only 1-2 people outside their household / support bubble (while also taking safety precautions like meeting outdoors and/or wearing masks). We classified these people as strong compliers’ – they accounted for only 1% of the total transmission risk among all respondents.

Conversely, we classified a small minority (8%) as ‘potential superspreaders’. These were people who met 3 or more people, typically without safety precautions (e.g. long meeting indoors, no masks, not keeping distance). This group accounted for 60% of the total transmission risk. Membership of this group was reasonably consistent across key demographics – it was not disproportionately men, or young people, or people from any particular ethnic group.


Our finding that a small number of people account for a large amount of coronavirus risk is consistent with evidence showing that coronavirus tends to spread in an ‘overdispersed’ way – most people who get the virus don’t pass it on, but a small proportion end up infecting many others. For example, a recent contact tracing study of 500,000 people in India found while 70% of people with the virus did not infect any of their contacts, 8% of infected people accounted for 60% of new infections (these numbers are, perhaps coincidentally, very similar to those in the above graph).

An implication of these findings is that coronavirus control measures may maximise their impact if they quickly reach people with the highest transmission risk. For example, Asia-Pacific countries such as Singapore, Japan, and South Korea have used backward contact tracing or ‘cluster busting’ strategies to help bring their own outbreaks under control. The rationale for backward tracing is that if ~70% of cases do not pass the virus on to others, then the efforts spent chasing the recent contacts of that group would be of less benefit than tracking down people with many recent high-risk contacts (e.g. hospital workers, church attendees, club-goers and migrant workers living in crowded dormitories,). 


Send us your suggestions