top of page

On Forecasts and Scenarios

In mid-August, about exactly 10 weeks ago, the wave of return travel was in full swing. Weekly testing numbers had almost tripled compared to early summer. The Robert Koch Institute (RKI) registered an average of about 1,200 new Corona cases daily—and a newspaper with four letters asked me: Where will we end up in autumn if this continues?

Back then, I calculated a number of up to 20,000 positive tests per day by the end of October. It was a huge headline that brought me angry letters. I was accused of dramatizing, I should be ashamed, spreading such panic, etc.

How easy it would now be to sit back and state: I had foreseen it back then.

Corona Forecast: 20,000 Infections Threaten in Autumn

The number of reported new Corona infections is growing slowly but steadily. Experts warn that 20,000 infections could occur in the autumn.

It's not that simple. Even statisticians cannot see into the future. We weigh data, we construct models that hold under certain assumptions. Under other assumptions, they can completely fail. If all of Germany suddenly voluntarily kept distance, wore masks, and regularly washed their hands, the situation would likely look different.

Now, 20,000 cases today do not mean the same thing as in the spring. The cases have shifted into younger age groups, treatment methods have improved, and a larger part of the dark figure is captured due to expanded testing. A 7-day incidence of 50 in October would probably correspond to a 7-day incidence of 10 to 20 in March and April.

Since 2012, the Bochum economist and RWI vice-president Thomas Bauer, Berlin psychologist Gerd Gigerenzer, and Dortmund statistician Walter Krämer have been scrutinizing statistics published monthly and their interpretation in the media. Since 2018, Katharina Schüller, the managing director, has also been involved…

It probably won't be long before someone accuses me of trivializing the situation due to this statement.

However, the problem is not (yet) the absolute number of cases, but the dynamics of the pandemic, which compared to the absolute number is much more robust against changes in the testing strategy or even against less than 100-percent accuracy of the tests. False positives simply cancel out when reproduction numbers are calculated.

Exponential growth means that we move from 20,000 to 25,000 cases just as quickly as from 1,000 to 1,250. However, 25,000 cases per day would still be unmanageable even if the pandemic stagnates at this level. Although a proportionately smaller number of severe cases are expected, current data suggest that, with a few weeks' delay, around 8,000 to 10,000 intensive care patients could be expected. Our health system could not cope with that.

Nevertheless, I view the upcoming "light lockdown" critically. In the spring, it was possible to reduce case numbers by about a factor of 10 within a month with strict restrictions. We closed schools and adhered to curfews. Such measures are not planned for November yet. Until we see the first effects, case numbers are likely to continue rising, possibly to about 25,000, before they start to decline. I consider a reduction by more than a factor of 5 very optimistic afterward. This means: we will hardly fall below 5,000 cases per day by the end of November.

These are relatively poor prospects for the price that is now to be paid—not because this price is so high, but because it will only bring a short gain in time and will stir up a lot of discontent. Especially among those who already do not "believe in Corona" and now feel vindicated in their theories because so far the case numbers have risen dramatically, but the number of deaths and intensive care patients still seems manageable.

So then it continues as before. A short phase of stability, then R oscillates again between 1.1 and 1.2, and by the end of February, the case numbers are back at today's level. Perhaps a flu wave will hit just then.

All this is not a prediction, but a possible scenario that I consider more likely than others. Whether it happens with a 30% probability or 70%, I cannot judge. I cannot see into the future.

But if I had to make a decision based on the data, I would advocate for a strict lockdown that rapidly reduces the number of cases to a few hundred per day. The cynical solution would be to just let the pandemic run its course, so that Corona deniers might finally understand the seriousness of the situation.

A compromise is sometimes the only enforceable and yet the worst solution.

I'm glad I'm a statistician and not a politician. I only have to calculate, not decide.


bottom of page