Lives, damned lives and statistics

Hongkonger, Going Postal
“.” by markus119 is licensed under CC BY 2.0

On January 11th, the BBC and Sky each made great play of ‘the worst excess deaths since the second world war’ announced for 2020. This clearly conveyed a message that the corona virus was indeed the worst thing to happen to the world in recent history.

I accept that this is a nasty flu bug and that precautions such as reducing contacts and distancing are sensible, as they would be for any other bug or virus.  You can make your own judgements on masks. And I don’t subscribe to conspiracy theories often –  experience suggests that cock-up is far more common than conspiracy.

But the statistics are open to question. By whom? Not by journalists, apparently. So let this minor academic pose an issue or two. (By the way, Radio four’s ‘More or Less’ programme is also recommended as a decent example of challenging headline statistics in a [usually] detached way).

This, then, is my take. Is it correct? That depends on whether you accept sources and assumptions. But I think it is worthwhile.

Here is the BBC: “The Covid pandemic has caused excess deaths to rise to their highest level in the UK since World War Two. There were close to 697,000 deaths in 2020 – nearly 85,000 more than would be expected based on the average in the previous five years. This represents an increase of 14% – making it the largest rise in excess deaths for more than 75 years.”

Let’s start with the last point. While it is true, it is misleading. 2019 was a low year for deaths following a notably high year in 2018. The ONS is, like nearly all national statistics collectors, distanced from government for reasons of confidence. ONS uses instead a five year moving average. This takes out precisely the sort of short term aberration used here by the BBC to heighten the story.

If we stay with the ONS, we find that they report (on the same day, note) an adjusted excess deaths figure of 75925, not 85000. They also warn that 2020 was a 53 week year statistically, not 52, and so to be careful with comparisons.

Meanwhile, the BBC also reported, on January 1st, that there had been 76530 covid deaths until the year end. This figure is close enough to 75920 to be regarded as the same. Yet it would suggest that EVERY SINGLE excess death during 2020 was down to covid, or at least those with a positive test during the previous 28 days. Surely not. Was there not a single death due to delayed cancer treatment or undiagnosed conditions, or due to heart attack where the patient delayed or did not call emergency services? Of course there was. ONS figures suggest more people died at home in 2020 –  a lot more- than would be expected. Their figure for these ‘non-covid excess deaths’ (up to 13 November, which seems to be the latest available), was 30785. This was for the period for around eight months from March 7th.  If we extrapolate this to the year end we get 38000 +/- .

Now, if (from ONS figures above) the total excess deaths for 2020 is 75920, and 38000 or so are non covid, then there have been just about as many non covid deaths in 2020 as covid deaths. Does this mean that lockdown has caused as many deaths as the virus? That’s a big leap, but let’s pose the question. As ever, things may not be as simple as they look, so beware.

“Ah, yes, but without lockdown there would have been many more covid deaths.” A fair response. What evidence is there for this? If we use (non lockdown) Sweden by way of analogy, ‘covid’ deaths for the year reported are 8273 of a total number of deaths of 95022 – 8% of the total. This compares to the UK’s 11% (75920/67000). So covid deaths in non lockdown Sweden are a lower percentage of total deaths than in the UK? This hardly suggests that lockdowns are effective. But is this ‘proof’ that lockdowns are ineffective? Well, perhaps – until you query why Sweden’s death rates are worse than their Nordic neighbours. There may be all sorts of reasons for this. The Swedes have a higher tendency to travel internationally than their Nordic neighbours. The Swedes have a higher immigrant community than their neighbours, with a consequent increase in arrivals of friends and relatives from elsewhere. So, plenty of opportunities for the virus to hitch a ride. And Sweden has a notably high proportion of residents in care homes. And, of course, this virus has a time to run yet.

If I put this analysis on TV news I would no doubt be introduced as an ‘expert’. I’m not. I’m positing on the basis of worthwhile but incomplete evidence. That doesn’t stop others though. Perhaps you saw “Coronavirus: Lockdowns could have saved up to three million lives across Europe, Imperial College study suggests”. Two clues as to why this should be treated with doubt: “could” and “up to”.  Imperial can’t be proved wrong, of course. But that doesn’t mean that they are right.

Hongkonger is exiled in the UK during the Wu Han disease (as it is often referred to in Hong Kong).

© Hongkonger 2021

The Goodnight Vienna Audio file