Within the next few weeks, the government will start to relax the coronavirus lockdown. Two big questions arise. The first, which has been widely discussed, is one that nobody, not even ministers, can yet answer: what will the new rules be? The second question, just as important, is receiving little attention—but can be answered: how can we tell if the new policy is working?
The need for an answer should be obvious. It will be vital to judge the impact of the new policy as quickly as possible. If it is working well, subsequent steps can be accelerated, and Britain can return to near-normal sooner. On the other hand, if relaxing the lockdown doesn’t work, then a decision to change course must be taken as early as possible, in order to avoid the danger of a second peak in the number of infections. How do we gather the information required to make this choice?
Let’s start with two big things we know about what has happened so far: first, countries with big, early testing programmes (notably Germany and South Korea) have suffered least; second, the British statistics about the spread of the virus and the numbers dying from it have been slow in coming and don’t provide the full picture.
People who die from the virus do so, on average, around four weeks after contracting it. If we rely on death statistics to measure the impact of easing the lockdown, then the timetable for measuring its impact is appallingly slow:
* After one-two weeks, the new policy will start to affect the number of infections.
* After five-six weeks, the number of deaths will start to respond to the new rules.
* After seven-eight weeks, given the time it takes for the number of deaths to feed through to the reported statistics, and for us to separate a real trend from daily fluctuations, we should know whether the new policy is working.
(It has become clear in the past few days that the rate for hospital deaths peaked around 8th April. This means that the infection rate probably peaked in early March. It was already beginning to decline before schools, restaurants, bars and most shops were forced to close. Would it not have been helpful to know this earlier?)
Applying that timetable to a new policy that takes effect, say, at the start of June, it will be late July before ministers can be sure whether their new rules are working—and hence whether to keep going, reimpose a full lockdown, or do something else.
Plainly, we shall need to know far sooner. This means detecting infection rates among the general public, rather than waiting for victims to die. And the best way to obtain infection rates is through a large-scale testing programme.
However, the current testing plans—for NHS staff, key workers and their families etc—will not provide this data. By definition, these are subsets of the population who are at risk precisely because of what they do. They obviously deserve priority when our testing capacity is limited. But to judge if a new policy is working, we must find out what is happening to the rest of us.
Last week the government announced that it will set up a panel of 25,000 people, rising eventually to 300,000, who will take blood tests at home at regular intervals. This an excellent idea. In time it will provide important information about the way the virus spreads and how many of us eventually contract it—even if we are lucky enough not to show any symptoms. However, it won’t tell us what we need to know in the weeks after each relaxation in the lockdown.
What, then, is to be done? Let me declare my bias. I spent 15 years as chairman, then president, of YouGov. We gathered information from thousands of people every day, not just on politics—the work for which YouGov is best known—but on a host of subjects. The way to find out fast whether any new pandemic policy is working is to do much the same.
We are now (weeks too late, but that’s another story) approaching the time when more than 100,000 people can be tested every day; and for these tests to be conducted at or near where people live. There will be enough capacity to test people who are not in any of the main risk groups without depriving those who need the test most.
Here’s what to do. Test a broadly representative sample of 10,000 members of the general public every day. As with normal polling, it would be a different sample each day; over the course of a week, data on 70,000 citizens would be gathered.
This is how the timing would work.
Once again, our starting point is that the number of infections would respond to the new policy within a fortnight. Tests can detect the infection around a week after exposure to the virus. With efficient data collection (YouGov, like any well-run research company—and, for that matter, the Office for National Statistics—knows how to process data almost in real time) we should be able to judge the effect of any new policy within three-four weeks of it taking effect. Ministers would save a month, compared with waiting for death statistics—a month that could be used to set new rules to save lives, to allow more businesses to reopen, or both.
Statistical purists will doubtless argue that it is impossible to get a perfect random sample of 10,000 Britons in a day. They are right. But for this plan to work, the daily sample does not need to be completely random. What matters is that the same kind of sample is used every day. The key thing is to obtain like-with-like comparisons of each day’s data. The level of infection measured each day does not have to be precisely accurate (although modern sampling methods, using sophisticated demographic information, can get pretty close to the true number). What we would be looking out for are significant changes in the reported rate of infection. As long as the sampling method stays the same, any change in the trend will tell us what we need to know.
Moreover, each week, by aggregating the daily figures to provide a sample of 70,000, we could explore what is happening in detail—by age, social class, region, whether people have returned to work and so on.
Ministers say they are guided by “the science.” I’m sure they say this in good faith, even though their use of the definite article implies a greater certainty than any true scientist would promise. What we do know is that good science relies on good statistics in order to get as close as possible to the truth. By commissioning the kind of research outlined here, ministers would be doing much to apply their mantra to the tough decisions they face in the weeks ahead.
This blog was first published by Prospect