This article caught my eye too.

I've tried to level some criticisms and questions at it, which you can find here:

https://mega.nz/file/wARRWIxJ#gMyN1z1SZsfXCRgod16YHe4Ye2jWbQ49N7RVYyUSphE

I'm not claiming he's totally wrong, but there are things that I think need looked at.

Ultimately, if we're to claim we're right with any conviction, we have to engage with, and refute articles like these. This is in addition to the principle arguments that we have on our side.

"You would expect a national lockdown to stop an infectious virus spreading. Afterwards, you would expect the data to look the way it looks. The idea may be dull and disappointing, but it has widespread scientific support. Sadly, when lockdown ended, it also explains why the virus returned."

I'm not really seeing the evidence in the data that national lockdowns (as they have been implemented) have a significant effect. I think they have some effect, sure. After all, if a country implements a brutal lockdown (maybe even welding people into their homes) you would expect this to have a very significant effect on transmission and infection/mortality rates.

What we've had is somewhere in between doing very little (as we would during any typical influenza season) and a full-on brutal lockdown. The question here is the relationship between some parameter we might call "lockdown severity" and things like infection rate. Is it a linear relationship? If a lockdown is at 50% on the severity scale between nothing and totalitarian, is there a 50% effect on infection rate? It almost certainly is not a linear relationship like this.

It might be, for example, that you need to get to 80% of a brutal lockdown in severity before you start seeing any significant effect. It might be that you only need to be at 20% severity before seeing a significant effect. So what does the data seem to be telling us?

The argument that Leo Benedictus uses to suggest that the UK lockdown worked is appealing, but naïve. The argument is that deaths peaked at 8th April and this was 16 days after lockdown - and that this, therefore, had been caused by lockdown (note, this is associational - it might be true but we cannot definitively make the claim it is true). He is attributing the reversal of the mortality trend (from increasing to decreasing) to lockdown. His argument is that, from studies, covid deaths typically occur 2-3 weeks after infection and therefore this peak is broadly when we would expect it to occur.

Sounds plausible, right? Well there's much more to understanding the dynamics of a curve than looking at local maxima. The real question is when did the rate of increase of the curve change? That's a fundamental change. It's a bit technical mathematically but we can draw an analogy with acceleration.

If you're in a car and you put your pedal to the metal your speed increases. If you apply constant positive acceleration the rate at which your speed increases is a positive constant. Now ease your foot off the gas a bit and you can keep your speed constant. Take your foot off the pedal entirely and the car starts slowing down (friction from the road). You're still going forward but your speed is getting less and less - and now you're experiencing a negative acceleration (we might want to use the term deceleration here).

So the question really is about when did we change from an increasing rate of mortality to a decreasing rate of mortality? You can easily estimate this from the data we have - and it's around 31st March/1st April. At this date the "acceleration" of the mortality curve goes negative .

In plain terms, just a week after lockdown, we see the dynamics of the mortality curve are already fundamentally changing. If one argues for a 2-3 week lag, as Benedictus does, then this significant change to the dynamics of the mortality curve could not have been due to lockdown. Something was fundamentally changing before lockdown could possibly have had an effect.

The important thing to note is that the overall shape of the mortality curve is broadly consistent across several European countries. In particular, we see exactly the same kind of shape with a peak in mortality even for Sweden. If we're going to argue that lockdowns caused the reversal (the peak) in the UK, then what caused the reversal in Sweden?

Now it's possible that lockdown measures shift the peak a bit (make it occur sooner), or lower the height of the peak - I can't say, I haven't done a full look and analysis of the data. Perhaps when I have the time I should. But, crucially, the assertion that lockdowns "cause" the peak is, in my interpretation of the data, a false assertion.

Fundamental to many of the "mainstream" arguments is a type of reasoning that looks like "it would have been worse if we didn't do X". Maybe - but that's an unprovable assertion. If we're going to get any kind of scientific evidence for the efficacy of lockdowns this kind of argumentation just doesn't cut the mustard. We have to try to use the data we have and not some counter-factual speculation.

The more important question for me is not whether lockdowns actually work (I don't believe a UK-style lockdown achieved very much at all except buggering up things for people not suffering, or at significant risk, from covid), but really whether they are a proportionate response? There seems to have been an implicit assumption that lockdowns are a sensible and correct response to a virus that is not all that different in overall threat than a bad influenza. Even if lockdowns worked beautifully - this important question would remain.

It's quite easy to get any data to support the decision you want by picking a carefully chosen arbitrary cut off date to stop looking afterwards which is exactly what they've done.

Stop looking after lockdowns ends therefore not examining longer term effects that take a few weeks to manifest (the R spike etc).

Its 12 year old level or analysis.

You can use his method to prove Germany "won" world war 2 by stopping looking after 1940 or someone's sports team won a match if you stop looking after the time they took the lead.