Don’t be deceived by Covid stats – we know a lot less than the numbers suggest

There is a concept at least as old as computing itself — Charles Babbage, the father of the field, expressed the sentiment, if not the words themselves — ‘garbage in, garbage out’. The idea is not a complicated one: no matter how advanced the calculation machine, no matter how good the statistical model, no matter how intricate the formulae, if the data on the way in isn’t reliable, the calculation that comes out will be suspect at best — and is liable to be outright wrong. We are regularly told that figures for economic growth were wrong and have to be dramatically revised. So what hope do we have of predicting coronavirus?

There is no reason to doubt either the skill or the dedication of the academics working to produce figures and forecasts to shape the response to the virus. But we do risk being misled by bad data. The daily influx of statistics, of infections, of deaths etc starts to build for most of us a false impression. It suggests to us that coronavirus is knowable, that we understand what is happening, that we have a plan. This impression begins to erode the second we examine the bedrock upon which it is built. Even the most reliable-looking of the figures we see tells us far less than it should.

It only takes a second of thinking about numbers of new cases to know they cannot be comparable across countries: how can they, if the death rate for the same disease is supposed to be ten times higher in Italy than in Germany? The infection numbers will only ever be the tip of an iceberg. In the UK, officially only those being considered for admission to hospital are eligible for coronavirus testing — with even NHS front-line workers unable to get tested before that stage, even if certain celebrities, politicians and others seem to get tests without much struggle.