Study: Dems COVID19 Lockdown Measures Causing Most Deaths

Written by Joel Smalley MBA on June 27, 2020. Posted in Current News

New statistical study shows a pattern of adverse and counter-intuitive effects resulting from governmental measures (‘lockdown’) to limit the spread of COVID-19. In short, the more counter measures taken by a government the worse the outcome.

In ‘An empirical analysis of the relationship between excess mortality and lockdown severity in the United States‘ (June 2020) author, Joel Smalley identifies a statistical pattern whereby Democrat-run states have substantially worse mortality rates than Republican.

Summary

According to hypothetical epidemiological models, certain non-pharmaceutical interventions like population-wide social distancing combined with home isolation of cases and school and university closure ought to have a very substantial impact on COVID-19 mortality.

However, the results of analysis of empirical data on mortality and counter-measure severity of all 50 US states, actually shows a statistically significant INCREASE in mortality associated with HIGHER degrees of counter-measure severity.

The results also show that Democrat states have roughly double the average severity as Republican states and account for 75{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of all the excess mortality during the observable period of the SARS-CoV-2 epidemic.

It is evident that Democrat states have a much stronger tendency towards intervention and this has led to much poorer outcomes for citizens of those states. Given the significantly heterogeneous nature of SARS-CoV-2 in terms of risk demography, it would not be a surprise to see homogeneous interventions having no impact on outcomes. However, it is difficult to explain why outcomes should be materially worse.

The author tentatively proposes a hypothesis whereby an explanation may be assigned to the nature of collectivism versus individualism. In this instance, Republicans might be more likely to take appropriate responsibility for their own welfare, making decisions and taking actions according to their own perceived risks. Conversely, members of collectivist states may be more inclined to rely on the diktats of the state even though they may not be logical or reasonable. This conjecture would need much deeper investigation to be upheld.

Introduction

On 16^th March 2020, Professor Neil Ferguson’s Imperial College model forecast a worst-case scenario of 2.2 million American deaths from COVID-19 if no mitigation action was taken[i]. The report advised that the only option to avoid significant death was “population-wide social distancing combined with home isolation of cases and school and university closure”.

On 26^th March, Ferguson revised his UK death forecast down by 96{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} as a result of the implementation of such measures, which would put the US number in the region of 88,000. In fact, most countries and states went further than the recommended measures that only called for reductions of 75{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of contact outside the household, implementing draconian lockdown orders instead.

It is, therefore, a reasonable hypothesis that the degree to which the individual states of America applied the recommended mitigation should result in a range of mortality outcomes, significantly skewed in favour of those states that applied the measures most stringently.

Methods

In order to measure the outcome of the mitigation strategies employed by each state of America, it is necessary to have a measure of mortality that is free from bias and an objective measure of mitigation severity.

COVID-19 reported deaths are unreliable[ii]. However, excess deaths are an unbiased measure of deaths that occur as a result of a shock to the natural death process.

Modelling data from the Centers for Disease Control and Prevention (CDC)[iii], it can be seen that deaths in the USA follow a predictable, smooth process. The start and end of each year has a slightly higher rate than the middle summer months due to seasonal influenzas. The curves can easily be fitted with a 4-order polynomial. The polynomial for 2019, for example, has a an R² of 97{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117}, an extremely good fit.

Data analysts commonly compare an individual year with the average of the preceding five years in order to determine the excess (positive or negative) mortality for the year in question.

However, in the case of America, 2020 started almost exactly in line with 2019 so we have adopted the polynomial fit of 2019 as our baseline expectation of mortality against which we can derive the “excess” mortality for 2020. Using this method, it is very clear to see the COVID-19 “shock” that first occurred in week 12.

We apply the exact same methodology for each state’s mortality data (again from the CDC) to derive an excess number of deaths for each week, having observed similarly good degrees of fit of the expected curve. For the period in question, we are only interested in positive excess mortality during the period we know that SARS-CoV-2 was circulating through the population, i.e. from week 12.

Since it is quite apparent whether there is excess mortality and its duration, using this method, we tally the total excess for the observed period. This results in a range of excess death that spans a different number of weeks for each state (0 to 12). For Republican states, the average duration is 7.4 weeks and for Democrat states it is 7.9 weeks.

In order to standardise the data for comparative purposes, we express the resultant number relative to the average expected death per week for the same period. Thus, our target variable is the excess death expressed in number of weeks of expected death and ranges from 0 to 24.2.

To measure the severity of counter-measures, we count the number of weeks that lockdowns, i.e. “stay-at-home”, “shelter-in-place” or equivalent orders were effectively imposed by each state governor[iv] [v] [vi] [vii]. The number of weeks ranged from 0 to 14. For Republican states, the average duration is 4.0 weeks and for Democrat states it is 7.8 weeks.

In order to test for confounding explanatory variables, we also measured excess death, as calculated above, relative to the proportion of deaths of nursing home residents since it represented a significant number of COVID-19 deaths[viii].

Results

The total number of excess deaths using this method by summing the excess deaths of each state amounts to 119,302 which is broadly in line with the reported COVID-19 deaths of 127,000[ix]. 25{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} of these deaths occurred in Republican states and 75{154653b9ea5f83bbbf00f55de12e21cba2da5b4b158a426ee0e27ae0c1b44117} in Democrat states.

Plotting excess deaths against lockdown duration reveals a significantly positive correlation which is contrary to the hypothesis.

In fact of the 12 states that have experienced no excess death at all during the period in question (Alaska (R), Arkansas (R), Hawaii, Idaho (R), Kentucky, Maine, Montana, North Carolina, North Dakota (R), Oklahoma (R), South Dakota (R), and West Virginia (R)), 5 of them (Arkansas (R), Kentucky (D), North Dakota (R), Oklahoma (R), and South Dakota (R)) had no ostensible lockdown. In fact, only three “no-lockdown” states had any excess at all (Nebraska (R), Utah (R) and Wyoming (R)) and were all at the low end of the range.

There were three outliers. New York City (24.2 standard excess vs 10 lockdown weeks) and New Jersey (11.3 standard excess vs 12 lockdown weeks) are both omitted from the chart due to scale. Hawaii, with its unique geographical properties benefited with no excess mortality for the period but actually has endured the longest duration of lockdown at 14 weeks which is very difficult to explain.

There is no materially discernible error in any of the state data.

Testing for confounding in the relationship between excess mortality and proportion of deaths in nursing home residents did not reveal anything of statistical significance.

Conclusion

The hypothesis that “population-wide social distancing combined with home isolation of cases and school and university closure” should lead to significantly better outcomes in terms of mortality from COVID-19 cannot be supported given the empirical data that is available.

On the contrary, the empirical data very strongly suggests that mortality outcomes are improved with fewer interventions.

The argument that, relying only on the hypothetical model itself, it is still possible to claim that the counter-measures were responsible for what might have been an even larger number of deaths is difficult to accept where the empirical data shows such a strong contrarian correlation.

Moreover, it is even more difficult to reconcile the fact that all of the states that effectively deployed no significant measures at all have resulted in virtually no excess deaths. There is low feasibility that this could be due to confounding or particular attributes of those states since they all share their borders with states with otherwise very different outcomes.

It is evident that Democrat states have a much stronger tendency towards intervention and this has led to much poorer outcomes for citizens of those states. Given the significantly heterogeneous nature of SARS-CoV-2 in terms of risk demography[x], it would not be a surprise to see homogeneous interventions having no impact on outcomes. However, it is difficult to explain why outcomes should be materially worse.

A possible reason could be in the nature of collectivism versus individualism, where Republicans might be more likely to take appropriate responsibility for their own welfare, making decisions and taking actions according to their own perceived risks, whereas members of the collectivist states may be more inclined to rely on the diktats of the state even though they may not be logical or reasonable. This conjecture would need much deeper investigation to be upheld.

[i] https://evidencenotfear.com/evidence/

[i] https://www.npr.org/2020/05/01/847415273/south-coronavirus-related-restrictions-by-state

[ii] https://edition.cnn.com/2020/03/23/us/coronavirus-which-states-stay-at-home-order-trnd/index.html

[iii] https://www.aljazeera.com/news/2020/03/emergencies-closures-states-handling-coronavirus-200317213356419.html

[iv] https://www.nytimes.com/interactive/2020/us/states-reopen-map-coronavirus.html

[v] https://www.forbes.com/sites/theapothecary/2020/05/26/nursing-homes-assisted-living-facilities-0-6-of-the-u-s-population-43-of-u-s-covid-19-deaths/#5ec19e7d74cd

[vi] https://www.worldometers.info/coronavirus/country/us/

[vii] https://spiral.imperial.ac.uk:8443/bitstream/10044/1/77482/14/2020-03-16-COVID19-Report-9.pdf

[viii] https://www.scientificamerican.com/article/how-covid-19-deaths-are-counted1/

[ix] https://data.cdc.gov/

[x] https://evidencenotfear.com/evidence/

About the author: Joel Smalley holds an MBA from the University of Toronto and  works as a Blockchain architect and early stage, polymath data-driven technologist, specializing in fintech, healthtech and IoT. He is currently CEO of Supermoney Ltd and CIO and CTO of Toucan Labs. 

PRINCIPIA SCIENTIFIC INTERNATIONAL, legally registered in the UK as a company incorporated for charitable purposes. Head Office: 27 Old Gloucester Street, London WC1N 3AX. 

Please DONATE TODAY To Help Our Non-Profit Mission To Defend The Scientific Method.

Trackback from your site.