🚨BREAKING🚨 UK Immigrants Dying at Substantially Higher Rates than Natives Since Spring 2020
New data released from ONS shows astonishing new information.
After fifteen months, I finally got the best mortality dataset available in England & Wales - daily deaths by date of occurrence since 1970, stratified by year of birth and sex1.
As advised by Dr Clare Craig, due to the possible confounding of immigration, they provided two datasets so we could derive a consistent dataset for analysis that was free from bias:
Any place of birth, which is available since 1970.
Stratified by place of birth (native or immigrant), which is available since 1988.
I always run a sanity check over the data before spending valuable time on analysis. This starts with plotting the aggregate data and checking for anomalies:
As we can see in Figure 1 above, there was apparently little trend in deaths between 1970 until the late 90s when the number of deaths trends lower, even though the number of immigrant deaths is rising. As we will eventually see once I start the modelling, this will be a function of births decades before. In other words, as the number of births decreases after WWI and WWII, inevitably the number of deaths must also eventually decrease.
It is important to note that my model of expected deaths will strictly be a function of number of births and mortality distribution of each birth cohort. This is the only reasonable way to accurately model expected mortality, rather than as a function of population at the time of death, which is not as intuitive, even though it is the most popular “correction” of absolute deaths used by the majority of statistics agencies.
You might already notice some anomalies in the immigrant mortality series. I wasn’t expecting these and was certainly not expecting to report insights on them but, as is my way, I report what I see, not what I’m looking for. More of that in a moment.
You can see that the full series is not stratified by natives and immigrants since that granularity of data was not available until 1988. In order to derive a consistent series of native mortality for the whole period (1970 to 2024), I had to model the ratio of immigrant deaths to native deaths to get the implied native deaths for the period 1970 to 1988.
Plotting the proportion of immigrant deaths to all deaths (Figure 2), the anomaly between 1993 and 1997 (red on the chart) is more apparent due to a better scale. I don’t know if this is a data error or can be explained with evidence so I have asked the ONS to check the data for that period as a matter of course2.
Excluding the data between 1993 and 1997, the rate of immigrant deaths is evidently consistent between 1988 and 2018. As such, we can reliably model 1988 to 1993 (a simple linear regression is reasonable) and use that to estimate immigrant deaths between 1970 and 1988.
When we subtract the modelled immigrant deaths from the 1970 to 1988 part of the full timeseries, we end up with a native mortality series between 1970 and 2024. We will eventually use this to produce our Gompertz distributions, confident that we have dealt with the confounding of immigration.
As we can see in the result (Figure 3), there are around half a dozen remarkable spikes in the data. I have labelled the four most significant, ignoring for the moment the importance of considering the mortality expectation at those times (i.e excess deaths rather than absolute deaths).
According to this raw mortality data, the worst burden of short-term mortality in England & Wales in the last 55 years, occurred right at the start of 1970. The “COVID” anomaly in April 2020 is pretty much in line with Dec 1989 and Jan 2000 (Millennium Bug!), so not totally unprecedented (notwithstanding the response).
Finally, we can return to the natives vs immigrants story. As per Figure 2, the proportion of immigrant deaths to all deaths in England & Wales is consistent between 2018 and 2020 (green shading). This means we can reasonably compare natives vs immigrants deaths between 2020 and 2024 without needing to detrend the immigrant deaths (assuming that 2018 to 2020 is the expected relationship going forward).
Normalising the two mortality series simply by using two axes with relative scales to fit the data between 2018 and 2020, allows us to observe differences in mortality outcomes since 2020.
As we can thus see in Figure 4, immigrants have fared substantially worse than natives since the emergence of “COVID”. In fact, during spring 2020, immigrant deaths were relatively 50% worse than natives. The natives baseline is around 1,600 deaths per day which rises to a peak of 3,200 during spring 2020 (double), whereas immigrant deaths rise from a baseline of around 200 to 600 (treble).
The relative demise of the immigrants is apparent again during the Jan-2021 spike and is persistently higher since summer of 2020, except curiously during winter mortality season 2022-23, where it matches the natives exactly.
The question that I cannot answer is why this anomaly might exist. If it was due to a virus, we would have to accept that the virus somehow discriminated against those who were not born in this country.
Obviously, that’s impossible so we would have to consider some other demographic explanation like ethnicity, deprivation and/or differences in metabolic and immunological health.
Or, we might have to consider the fact that the impact on mortality was not the direct result of a virus whatsoever? Moreover, we would have to consider why these differences only manifested material disparity since 2020 and not during other “viral epidemics”?
Regardless, the major implication of this is obviously that public health interventions are not instrumental in mortality outcomes but personal idiosyncrasies are. Makes you wonder why the public health agencies are not doing this analysis because it could save the “public” significant cost “next time”?
Next, we’re going to break the data down by sex and age (birth cohort) to see if there is any additional insight and build the excess death model as a function of the Gompertz mortality distributions.
Thanks again to everyone who contributed.
Update 13th June: ONS confirm the underlying data for this period was not correctly coded so this artefact is not genuine. Allowance should be made when working in this area of the curve.
Depending on ethnicity, don't know if it is a myth or trope but darker skin = less Vitamin D3, Indian Sub-continent people often diabetic and / or heart issues due to diet?
Lower Vitamin D levels?