OECD.Stat reports excess (deviation from the long-term average) deaths and COVID-19-related deaths on a weekly basis since the start of the pandemic. In this post, we compare data from Australia, Canada, Israel, the UK, and the USA and try to understand the data. There are two important measures: number and percentage change from average. These two figures allow for calculating the average number of deaths for a given week and for the whole studied period. Figure 1 presents the % change as a function of time for the overall value. One can see that in the UK the deviation is above 100% for some weeks at the beginning of the pandemic. The US has a peak of around 45% at the beginning and another three peaks of the same amplitude in 2021 and 2022.
Figure 1. Percentage change of excess
deaths: weakly estimates.
Figure 2. Same as in Figure 1 for COVID-19-related estimates
Figure 2 presents the same parameter for the COVID-19-related values. One can see a significant drop in amplitude in the UK and US. Therefore, it would be instructive to look into the numbers and compare the difference between the total excess deaths and COVID-19 related. Figure 3 displays these values for the UK and US. One can see that there is a difference between the % change values and the counted values. There is no big difference between numbers but the difference between percentage changes is larger than 2 times (Figures 1 and 2). What is the reason behind such a discrepancy? The most likely explanation is that the average values to calculate the % change are different and the average for the COVID-19 calculations has to be much larger.
When looking into the monthly CPI estimates
one often sees two values: seasonally adjusted and not seasonally adjusted. The
reason for the difference is simple: summer months are different from winter
months for many activities and there is no sense to compares agriculture-related values in July and February. One has to compare July for a given year
with the July data in the previous years. Then the trends and variations become
clearer. Instead of comparison with each year in the past separately the BLS introduces
the average values for the last 5 years. One can suppose that the same
procedure is applied to the weekly data reported by the OECD and they compare
excess deaths with the average values before the pandemic through the whole
pandemic period since the values from the pandemic years will disturb the
estimation of the pandemic influence. This assumption is likely accurate for
the total numbers in Figure 1 and one can calculate the average values from past estimates.
This is not the case for COVID-19-related excess deaths. To understand the OECD procedure and the results in
Figure 2, let us take the example of week 16 in 2020. The total excess number in
the USA is 23552.4 (the average value is not necessarily an integer) and this makes a 44.2% excess rate. The average value is roughly 100*23552.4/44.2= 53285. For
this average value, one can estimate the excess rate for COVID-19 deaths:
17221/53285=32.3%. The OECD gives 21% for this week. This is a tremendous discrepancy.
But we have enough data to explain it. Let us take the total number of deaths (not
the excess deaths) for week 16 as a sum of the average and excess values and
get 53285+23552=76837. For this current value, one obtains the excess death
rate of 22.4%, i.e. exactly as in the OECD table. For this value, the total
excess death rate 23552/76387=30.7%, not 44.4%. When using COVID-19-related cases
in the calculation of the influence of COVID-19 on the change in the overall
mortality rate one obtains a highly biased estimate. The COVID-19 effect can be
estimated only when using the averaged weekly values before the pandemic, but for
the same week in order to remove all known seasonal effects.
The OECD data cannot be considered reliable. They are rather constructed to hide high COVID-19 death rates in the OECD countries.
Figure 3. Numbers of excess deaths in the UK and USA: total and COVID-19 related.
No comments:
Post a Comment