The rise of Covid Quackery (Covackery perhaps?) has brought the use of ‘modelling’ to a greater prominence.
On the one hand, it has been entirely right and proper to pour heaps of derision on ‘models’ that predicted CoronaDoom. Or to laugh at the idiocy of the modelling that suggested that 20 million lives worldwide had been saved (in just one year) thanks to the Glory of Pfizer (PBUT - Profits Be Unto Them).
Then there’s the ‘climate’ modelling. Not the best record of ‘success’ there1.
The attitude has arisen, and it’s entirely understandable, that as soon as you hear the word ‘model’, you can pretty much dismiss the conclusions. Those of us who were somewhat shocked by the global response to a moderately serious viral infection, and the continued subsequent craziness, deception and tyranny, have lost faith in so many things.
Yet we can’t, and shouldn’t, dismiss modelling and the process of modelling out of hand, however bruised and battered we have been by their misuse during the Covid Era. Models and modelling have their place.
Those of us who want to restore some measure of balance, some sanity, need models to make some of our arguments. When I make the statement that the data doesn’t ‘look like’ what we’d expect from an effective vaccine, I am referring to some inner conception (some ‘model’) I have of what the observed data for an ‘effective’ vaccine would look like.
When we use excess death as a measure - because we can’t trust the official covid death statistics (for very good reason) - we’re using a model. We are using a model to try to get some estimate of how far outside of the ‘norm’ things are. The simplest ‘model’ for excess death is to establish some baseline using the average death toll for the previous 5 years and to look at deviations from that baseline. There are better (and more accurate) statistical tools we can use to refine this excess death ‘model’.
One of the big differences, other than sheer complexity, between these (simpler) models for things like excess death and vaccine efficacy, and the covid epidemiological models (and the climate models) is that these latter models are meant to be simulations of reality. The purpose of the former models for things like excess death is to gain some insight and not to serve as accurate predictors of ‘reality’.
Knowing what an effective vaccine ‘looks like’ - or ought to look like - is a useful thing.
Let’s look at the data for the ‘world’ - and accept that the data is probably not the best we could have done, given the wilful and deliberate deceptions that took place. We expect, don’t we, that a good vaccine improves things? Yet, the cumulative covid death data seems to be at odds with this.
I’ve added in simple gradient ‘guide’ arrows in yellow. In order to argue that the vaccine has been a ‘success’ we’d need to posit something like the following2
To argue either case requires modelling. In order to argue that the vaccines have been a rip-roaring success you have to have some model that describes ‘reality’ without the vaccines. In order to argue that the vaccines have been about as effective as a cardboard hammer, you have to have some model that describes the ‘reality’ of what a successful vaccine looks like.
One thing is absolutely certain - the vaccine did not actually improve things (in terms of cumulative death toll). The data shows this - there was no actual improvement. So we have 2 potential ‘explanations’ for the actual data
the vaccine was shite
the vaccine worked - things would have been so much worse otherwise
The first position is stronger. There was no improvement (the actual data). No further hypothesis is necessary - we just look at the data and say ‘nothing happened’. The vaccine had no discernible effect that is directly visible in the data (maybe even a slight detrimental effect). To conclude that the vaccine was, therefore, useless requires some additional supposition about what a ‘good’ vaccine looks like. But this position is stronger because we do have data about what a ‘good’ vaccine looks like.
The number of measles cases drops off a cliff after the introduction of the measles vaccine. And this is where we have between 60-70% vaccine uptake for a virus that is maybe 6X more infectious than covid. This, this, is what we expect from a successful vaccine.
The second position requires some mental gymnastics (which might, or might not, be reasonable) in order to explain why no improvement was apparent in the data. It should be fairly obvious from this chart for measles that the covid vaccine, whatever it actually did, operated nothing like as effectively as the measles vaccine.
But you only have to look at highly vaccinated (and highly masked) Japan to see that arguing for position 2 is going to be something of an uphill struggle.
I’m not saying it can’t be done but, in general, the more special pleading one has to indulge in, the weaker is one’s case.
Of course, using the cumulative ‘covid’ death data is not going to give us the most accurate picture - but it’s pretty stark nevertheless. The ‘benefit’ of vaccines cannot be directly seen in the data at a global level. And, like defining an ‘average’ temperature for the world is a bit of an an eye-raiser when talking about climate change, using the data for the entire world for covid is maybe not fine enough.
But when you do look in finer detail, the picture doesn’t look any better for the vaccines
Only something like 5% of countries show an improvement (in terms of mortality) pre- and post-vaccine. A real let-it-rip roaring success, if ever there was one.
The problems that exist when arguing the case for vaccine ‘success’ are quite substantial.
Why would we expect mortality to be significantly worse3 in the 2nd year when a significant fraction of the population have been exposed in the first year and have, therefore, some degree of exposure acquired immunity?
Why would we expect a virus to evolve to be more virulent in the normal course of things (without artificially imposed selection pressure like that of a leaky vaccine widely employed in the middle of a ‘pandemic’)? The general trend4, the evolutionary pressure, is in the direction of less lethality - more people surviving means more viral reproduction. In evolutionary terms a more successful strategy for the virus.
The existence of acquired immunity actually makes the case for vaccine ‘success’ much, much weaker. Because of acquired immunity we expect things to improve with time - 2nd and 3rd waves, if they occur, will be much less lethal at a population level. The fact that we do not see this in the data (little or no apparent change pre- and post-vaccine) is highly suggestive that the vaccine made things worse - it changed what should have been an improvement (even in the absence of vaccines) into no improvement at all.
How do we square this with the argument that is often advanced by the Prophets of Pfizer (PBUT)? The claim is that the vaccine has demonstrably prevented death - it lessens the severity of symptoms. Data is even presented to support this. But this position is clearly inconsistent with the overall global data as it stands. To bring it back into apparent consistency one must make all sorts of extra claims (such as the virus being very significantly more deadly in year 2)5.
Something is not adding up. Is it with the global cumulative death data, or is it with the data that purports to show vaccine effectiveness in terms of symptom reduction?
The benefit of simple models and using them to figure out, in broad-brush terms, what ought to have happened is that we can easily see inconsistencies6 like this one.
Models can be very powerful. But they have to (a) make sense and (b) have, at least, a passing resemblance to reality. The more complex a system you’re trying to model, the more difficult this is to achieve - and it’s particularly difficult when your model is trying to simulate reality to some extent, especially when it’s not 100% clear what basic fundamental variables and interactions are important7.
And, of course, whatever level of ‘modelling’ you do, it has to match ‘reality’. If your model doesn’t predict what actually happens, you’ve done something wrong. You hope that you’ve got it ‘mostly’ right and that only minor tweaks or additions are necessary. If your model predicts something that is out by an order of magnitude (Ferguson’s predictions for Sweden’s covid death toll, for example) then it’s pretty obvious that more than a minor ‘tweak’ is necessary, or that your initial assumptions are wildly off (or both).
Even if your model does match up with the observed data this, in itself, is no guarantee that your model is ‘correct’. It’s a necessary, but not sufficient, condition.
I want to defend models, in general, because, as a physicist, models are what I’ve used throughout my career. The various physical laws are, themselves, ‘models’ of reality - and they’re used to construct more complicated models that, we hope, describe (and explain) what is observed.
I might be wrong, but as a general rule if something has a significant effect it has, by definition, a significant effect8. Arguing that vaccines, or masks, have had a significant effect when no such significant effect can be directly seen in the data is interesting, to say the least.
The Emperor has outstandingly significant data - you just can’t see it.
Depends on how you define ‘success’. If you’re looking for a model that makes accurate medium to long term predictions, then forget it. If you’re looking for a model that scares the crap out people, then we might argue they’ve had more than a smidgeon of success.
Or to argue that the observed data is precisely what we would expect from a highly successful vaccine.
In order to claim high effectiveness the cumulative death figures would have to have been significantly worse in the 2nd year without the vaccine (the hypothetical blue line in the figure)
It’s not a fixed rule - but it is a trend. Furthermore, it’s a trend that has sound evolutionary reasoning behind it.
The other way to achieve this ‘consistency’ is to construct some complex model, claim it has anything to do with reality, and build in assumptions that give you the answer you want.
Inconsistencies are the bread and butter of any true scientist. We love them. We wake in the morning and kneel before the Science Gods saying “Please, Oh Holy Ones, let today be the day I find an inconsistency”.
And it’s not always clear exactly what the ‘rules’ governing interactions are. If you’re modelling something moving through the air, for example, you might want to consider air resistance. But what ‘model’ of air resistance should you adopt? Is the resistance proportional to the velocity, or to the velocity squared (or something else), for example?
No shit, Sherlock
In the 19th century and earlier, an "outbreak" meant that (for measles) about half of the deaths of a year were children (based on Gothenburg in Sweden); there are data here from 1749 re: measles and smallpox.
So what constitutes an outbreak needs to be taken into account, which modern models don't.
Image shows cases in Sweden 1911-2013. The red line is exponential moving average, green arrows are starting years for vaccinations.
[https://sv.wikipedia.org/wiki/M%C3%A4ssling#/media/Fil:M%C3%A4ssling_i_Sverige_1911-2013.jpg]
What is more interesting is that deaths due to measles drops sharplylong before any vaccinations start. Nutrition, living standards, clean water, and good waste processing took care of most deaths for most of the old killer diseases, and cam about very much due to the cholera-epidemics of the 19th century.
There's still an active cholera-pit in central Stockholm, a mass grave containing hundreds if not over a thousand dead. They were tossed in naked, and the pit was covered.
The entire site is a park, as it any exploitation of the otherwise quite valuable plot would necessitate digging up 200 year old corpses, semi-liquefied as one greatmass as they are to many for proper putrefaction and rot to set in, allof it riddled with still active, highly contagious strain of cholera: half the infected in that epidemic died.
It's a good thing they didn't have models back then, and had to use reason based in empircal experience instead.
Thank you Professor Rigger. This is the most informative and enlightened essay I have read since Thank God For The Time Lords. You are everything I am looking for in a teacher- learned, honest, rational, reasonable, sober, clear and concise, good humoured and literate. As a 75 year old student of life and literature it is very rare for me to read something which is coherent and free of grammatical errors. If there were a thousand more writers like you I wouldn’t be so concerned with the obvious destruction of the English language.
And thank you for knowing that DATA is a collective noun which should be followed by IS not Are. That is the icing on the cake. The cherry on top is that unlike everyone else in the world, except me, including everyone who should know better you speak of ACQUIRED IMMUNITY rather than the misleading term NATURAL IMMUNITY. (If it appears that I am shouting it’s because I am shouting for joy).
E&OE