Once again a misleading media article regarding Covid & deaths in Sweden, this time by Reuters.
I’m not sure if media are too stupid to deal with basic numbers, or if they for whatever reason have a hidden agenda, therefore twisting every message and article according to that agenda. Regardless of which, I’ve now lost any and all trust I might previously,before 2020, have had for media, since time after time they fail – either because of stupidity or by intention – to present facts that are TRUE, TRUTHFUL and REPRESENTATIVE. Almost everything important presented by media nowadays seems to be part of some sort of Information Warfare,where the facts are spun intentionally to mislead – Technically True, but Semantically Misleading, e.g when death tolls year-to-year are compared using absolute numbers instead of population & age adjusted numbers. It seems to me that today’s media wants to control the narrative to suit their own purposes, instead of presenting as unbiased as possible facts.
Anyways, since the Reuters article above states that Sweden had 7.7% “Excess Deaths”, and as always fail to reveal how they’ve come up with that number, I thought I’d reverse engineer their “analysis” to see where that 7.7% number comes from. So, below, what follows are 5 different calculations on “Excess Deaths” for Sweden.
Why 5…? Well, as I’ve pointed out several times before, e.g. here, the notion of “Excess Deaths” is very ambiguous – before we can talk about “Excess”, we need to define what the “Normal” is, i.e. the Expectation. So, below, I’ll use 5 different Expectations, and as we will see, these lead to very different results in terms of “Excess Deaths”.
First, let’s look at the total number of deaths in Sweden for the past few years, including 2020:
year 2002 95009 2003 92961 2004 90532 2005 91710 2006 91177 2007 91729 2008 91449 2009 90080 2010 90487 2011 89938 2012 91938 2013 90402 2014 88976 2015 90907 2016 90982 2017 91972 2018 92185 2019 88766 2020 98124
By looking at these absolute deaths, we’ll notice that 2020 indeed seems to have a huge number of “extra” deaths, not least compared to 2019, and I’ve seen articles that claim “Sweden had 10.000 excess deaths 2020”. However, presenting facts in this fashion, while the statement is TRUE as well as TRUTHFUL, is simply not REPRESENTATIVE for the semantics of what we want to convey to the reader – was or wasn’t 2020 exceptional in terms of deaths. In other words, this is a typical case of presenting TECHNICALLY TRUE but SEMANTICALLY MISLEADING information.
Why ? Because the reader, seeing that deaths 2020 were almost 10000 more than 2019 is left with the impression that 2020 must have been really terrible, and this is simply not true. There are several things missing in this type of presentation of facts, using absolute numbers and with no clear definition of expectation, the two most significant failures are that the numbers, as presented, do not take neither population size nor population age structure into account. Both of these factors matter, tremendously, as the interested reader can read in other posts on this blog.
Basically, using absolute numbers to compare year-to-year deaths is nothing less than fraud. Whether it’s by accident or intentional, it really doesn’t matter. It’s simply fraud.
So, below I’ll use 5 different expectations for deaths as illustration that the method matters:
- Absolute number of deaths compared to average number of deaths 2015-2018
- Absolute number of deaths compared to average number of deaths 2015-2019
- Age group based mortality rate based on average age group mortality rates 2015-2018
- Age group based mortality rate based on average age group mortality rates 2015-2019
- Trend line (2015-2018) based age group mortality rates
This list is by no means exhaustive, there are other options for calculating what the Expectation (the “normal”) is, but hopefully these five methods illustrate the point that in order to allow the reader to determine whether 2020 (or any other year) was exceptional, you simply must present what you are comparing to, i.e. what’s your baseline and expectation.
Furthermore, the reason I use two different baselines above: 2015-2018 and 2015-2019, is that as we’ll see, if we consider 2020 being exceptional – an outlier – then we must also consider 2019 being at least as exceptional. Thus, IMO, the 2015-2018 baseline is the best, no-biased baseline available, since it is not distorted by the two outliers, 2019 and 2020.
So, here goes, here’s the Swedish Excess deaths, in absolute numbers as well as relative to Expectation:
So, depending on how you’ve defined the Expectation, Excess Deaths 2020 range from ~7000 (7.9% ) to ~ 1200 (1.2%) , while for 2019 the range is ~ -6700 to -2200 (NOTE the negative sign!)
So, which of these results gives the “best” answer….? Well, it’s definitely *not* the absolute numbers (as in the Reuters article above), regardless baseline, since it fails to take any other factor into account. So that leaves us with three remaining alternatives, the two age group mortality based ones, and the final trend line based one. Of these, I’d eliminate the age group mortality based on 2015-2019, since I consider 2019 being equally large if not larger outlier than 2020.
So, we are left with either the age group mortality based on 2015-2018 baseline, or the linear regression trend based expectation. If you are a believer in infinite growth (or, more correctly: infinite decline) you might want to chose the trend based result, which gives the following numbers for 2019 and 2020:
2019 : Excess deaths : – 4510
2020 : Excess deaths : 4356
If you do not subscribe to infinite growth, then you might prefer the age group mortality rate baseline of 2015-2018, which gives:
2019 : Excess deaths : – 6747
2020 : Excess deaths : 1176
The choice is yours.