Covid19 spread in the USA — liberty vs science, state by state
Two ancient parables
I’d like to bring our attention to two ancient parables to initiate our discussion.
Rabbi Shimon bar Yochai taught a parable: Men were on a ship. One of them took a drill and started drilling underneath him. The others said to him: What are sitting and doing?! He replied: What do you care. Is this not underneath my area that I am drilling?! They said to him: But the water will rise and flood us all on this ship.
A second parable comes from Pirkei Avot: If I am not for me, who will be for me? And when I am for myself alone, what am I? And if not now, then when?
State of covid19 spread and knowledge - summary
This paper is a snapshot in time captured at the end of October, 2020. The pandemic, first observed roughly a year ago, has spread throughout every continent except for Antarctica. Focusing on the US, the US has 20% of both the world cases and deaths despite having only 4% of the world population. This paper explores some of the important dynamics at play in the US at this point. In particular, there are two diametrically opposed factions at work which have a strong influence on the spread of covid19 in the US today. One holds liberty as the highest principle, the other science-based necessary action.
Where do the two ideological factions stand in principle?
The one faction, represented by President Trump and so many others, but in general associated with the Right, refuses to mandate covid19 prevention practices. This faction holds that liberty is the highest value, and that masks are a violation of said liberty. To this faction, science is either not settled and/or not valid, doctors are not relevant and in Jared Kushner’s own words, it’s about a “getting the country back from the doctors”. They fear what they see as “tyranny”, judging from extensive anecdotal reading and social media interaction. If government can control masks, where will it end? They fear contact tracing, because the “big brother” government would be allowed access to people’s whereabouts
The other faction, represented by Dr. Fauci and the leadership of Gov. Cuomo, and in general associated with the Left, tries to follow the science as it learns, and tries to implement policy based on what the science is saying. They too are wary of individual freedoms but judge the greater good to be protecting the community from what they see as a clear and present threat.
These positions are clear. It is truly unfortunate for the US that this disparity exists. I would like to explore what this means in terms of data that has been measured.
Where do the two ideological factions stand in practice?
A mentioned above, the “liberty” faction refuses to wear masks or to practice social distancing. As discussed above, this group holds superspreader events, even using President Trump’s behavior as an example in nationwide stops at which covid19 surged, as reported by USA today, The Independent and Guardian from UK, and many others.
The “science” faction mostly tries to follow the latest medical/epidemiological recommendations.
Let’s see how these compare in practice.
Given the starkly contrasting ideologies along political lines, a first hypothesis is to assign “red” or “blue” to individual states and observe the presence or absence of correlations. One widely accepted index for doing so is the Cook partisan index, which measures, quantitively, the difference between red and blue. It is worth looking at two measures: the full duration of the pandemic, and the number of cases since the first wave looked reasonably under control. Jun 1 for this purpose is a convenient date in between the first and second waves.
Figs 1a and 1b are a visualization of the total covid19 cases from Jun 1 – Oct 24, 2020. Visually it is clear that red states dominate the most covid19 cases and blue states dominate the least covid cases.
Table 1 shows covid19 cases from Jun 1 – Oct 24, 2020, as a fraction of the population, ranked in order, with the specific partisanship. In this representation, red is right leaning, purple is even and blue is left leaning according to the Cook Partisan index.
Table 2 shows how many states are red, blue and even in the Cook partisan index. Note that 34 states appear to have blueshifted and 16 appear to have redshifted. This is discussed in somewhat more detail here.
Using these numbers, it is clear that the Top 15 are not blue (i.e. they are red/even), followed by Nevada at -1, the lowest possible amount of blue, followed by a further 7 red states. Nevada has tourists from all over, so it is not surprising that Nevada should be a representation from all over the US. Even neglecting Nevada, the statistical likelihood that the Top 15 are red/even – given there are 30/50 states which meet this criterion – is 1 chance in 2,127. If we neglect Nevada as tourist-agnostic, the likelihood of the Top 22 of the remaining 49 states being red is 1 chance in 102,697. It is statistically highly unlikely that this is a coincidence, and we have discussed the strong causation above. The conclusion is very clear. Covid19 negligent behavior, as is practiced in the red states, is strongly correlated with and causes causes many more covid19 cases. Clearly Vermont and Maine have adopted best practices.
Note the remaining two columns in Table 1 – the Nate Silver partisanship as of Oct 24, and the shift from the historical Cook index. Particularly visually striking is that the states with the most cases (#1-8 on the list in rank order) are in general getting even redder, with the exception of Florida, the largest population by far of the group. The next tier, from #9-17 in rank order, are getting bluer. So of the top 17 states, the top half is getting redder (more covid19 cases) and the bottom half is getting bluer (many covid19 cases, but not as many as the very top). Note that both Florida and Texas are experiencing a blueshift. And both Florida (8.9M c.f. 9.4 pre-registered) and Texas (9.7M c.f. 11.6M pre-registered) have seen early voting, for the first time, representing the overwhelming majority of the pre-registered electorate. Vermont, at the very bottom of Table 1, also has the largest blueshift of any state in the US.
But what about deaths, you next ask?
Once again, primary data has been compiled here in an excellent time-evolving visualization done by Dan Goodspeed.
Figs 2a and 2b are a visualization of the total covid19 cases from Jun 1 – Oct 24, 2020. Visually it is clear that red states dominate the most covid19 cases and blue states dominate the least covid cases.
Table 3 shows covid19 deaths from Jul 1 – Oct 26 per 100,000 population, ranked in order, with the specific partisanship. Similar to Table 1, red is right leaning, purple is even and blue is left leaning according to the Cook Partisan index. Once again, even though the states aren’t in the same order as in Table 1, the correlation with red states with more mortality and blue states with less mortality is clear. The random probability that the Top 9 in mortality are all red is 1 in 256, and if we neglect Nevada due to tourist population, that the top 14 of 49 are red is 1 in 7130. These are statistically unlikely numbers. The difference and ratios between better and worse practices is striking — and this is human life at stake. Once again, Vermont and Maine have adopted best practices. Leaving aside the very extremes, there is a factor of 7 difference in covid deaths between the 10th and 90th percentile.
It is truly unfortunate that something as simple as preventing a lethal virus from transmitting is a political issue. This shouldn’t be. Preventing covid19 infection and death should be a humankind issue.
But isn’t covid19 data being misrepresented somehow? And isn’t the US doing all it can do?
In certain social media circles, there is doubt expressed as to the validity of case count or even death count – people point to accusations falsifying data or misclassifying covid19 cases or comorbidities.
First, the only systematic manipulation of data that I am personally aware of is the change in trajectory of case count with Trump took data reporting away from the CDC. I can go into detail about how, for example, Florida case counts changed trajectory within a week of this change in reporting, but that deaths did not track cases and I have posted this in detail on social media previously, but do not want to distract unduly from the main points here. My conclusion from data is that Florida data likely underestimated cases in July/Aug, but at some point the tsunami of cases may have caught up.
There are two primary indicators of real issues, independent of how cases and deaths are being logged and whatever errors may occur in said reporting: (1) excess deaths and (2) hospitalizations.
Fig 3a and 3b are extracted and reproduced from primary data and show (a) absolute deaths in 2020 in the US vs the 5 previous years and (b) the percentage growth from baseline vs time, both from Jan 5 – Sept 6, 2020, the last date for which it was easy to do the extraction. It is clear that the excess mortality, mainly in the northeastern US, was far above the baseline. Two waves of mortality are clear: the spring wave and the mid-summer wave. Once again mortalities are known to be going up in a third wave of mortality, not reflected in these graphs. For comparison, the excess deaths in South Korea is shown, where excellent covid19 practice is used, including contact tracing. There is a clear difference between the US and South Korea (and some other places) fully attributable to human practice. The US and the US Admin are not doing anywhere near all that is possible. There is no amount of testing now, Pres. Trump’s “go-to”, which can compensate for failed policy in the first place, which I pointed out seven months ago (in/near Fig 16). The cures for the pandemic are masks, social distancing and contact tracing.
It is clear that covid19 deaths are on the increase, as seen in Fig 3a and 3b. There are some days which have already seen 1000 deaths from covid, and with the trendline increasing, the US is clearly heading back there again.
There are also detailed reports, city by city, or state by state, showing excess mortality peaks. This New York Times reference is interactive and shows the excess death story clearly.
Here is another visualization tool, courtesy of New York Times
Let us be clear – the deaths from covid19 are clear and present, and significant compared with other causes in the US. Table 4 gives mortality rates per 100K from various causes. Leave the top row for the time being – we’ll return to that later. At the present time, using the most recent 2018 data available from the CDC, there are 868 deaths per 100K, all causes. If killers are ranked, in the worst state, Mississippi, the death rate is higher than the US aggregate for the #1 killer. For the US aggregate, covid19 is already the #3 killer with death rates not slowing. In the best state, Vermont, covid19 barely registers on this scale at this time.
With all these deaths, one would expect hospitalization issues in the US. Indeed, there are many reports of increased hospitalization, and hospitals at/near capacity and straining US limits in various locations – e.g. these references from Montana, El Paso, and these US reports from the NY Times on Oct 23 and Oct 25.
It is further claimed that the present Administration has a policy which prefers that this hospitalization data not be transparent, which supports the “liberty” narrative over the “science” narrative, which just leads to more cases, hospitalizations and deaths.
But isn’t the virus less deadly? Aren’t mortality rates decreasing?
It is true that at the spring peak, mortality rates were very high as science and medicine were coming to grips with covid19. That mortality rate has decreased from what was at one point well above 20% (over 50% in the graph (Fig 5), but there are statistical reasons why this was artificially high) to what is now approximately 4%. Let’s be clear. For those who contract covid19 (there is a case count), there is a 4% mortality rate, which is still quite significant, despite the medical progress to this point. And yes, this varies by age, demographic, etc. (not the point of this paper). Why?
There are many potential reasons, but one might be a Darwinian one relative to the virus. The virus needs a (human) host (let’s neglect zoonotic transmission). We can speculate that one way to encourage more transmission is for the virus itself to mutate to be somewhat less deadly than its predecessor mutations in order to encourage spread.
What does herd immunity mean?
There is one scenario in which the US develops herd immunity because no attempt is made to contain the virus in enough areas of the US that the virus propagates, through human transmission, throughout the US. Somewhere between 70-90% of the population, depending on the “R-factor”or R0 immunity number (herd immunity threshold is 1-1/R0 for this purpose) and other parameters of the virus, would be necessary. There are two principal methods of achieving this: (1) a vaccine is really available (it is not and cannot be ready immediately, despite how much Pres. Trump states, falsely, otherwise) or (2) there is no attempt to control the virus. Incredibly, the Trump administration plan, as articulated recently by Mark Meadows, is that “we are not going to control the pandemic”.
Let’s be clear, the herd immunity scenario in the absence of a vaccine would mean that a minimum of approximately 70% or so of the US population would contract covid19 and that 3-4% of these would die (slightly lower than 4% allowing for further improvements in treatment) – meaning something like 2.5% of the total US population – or roughly 9M people, would die. The total US death rate annually is 868 deaths per 100K of population, and 2.8M total annually, as a baseline. If this herd immunity took 1 year to develop, the death rate associated with covid19 herd immunity, based on the most recent understanding of mortality rate for closed cases, would be more than 3 times the baseline death rate in the US. If the total case count is underestimated at this time, then that would mean the mortality rate is lower, and if the case count is overestimated at this time, that would mean the mortality rate is higher. Whichever is the case, a death total of more than 3 times the baseline rate, even if that number is somewhat adjusted, is staggering. Returning to Table 4, the top row represents what this might look like if all the deaths took place in a 12 month period. This is what the “liberty” faction, as represented by Mark Meadows, Pres. Trump’s chief of staff, is advocating. Turning off the data collection hat for a moment, let’s be clear – herd immunity as advocated by the “liberty” faction is willful and reckless negligence – “live free and die”.
We need to take simple actions – masks, social distancing, handwashing – to contain the virus while waiting for more permanent measures, such as a vaccine. We have a choice between “live free and die” — i.e. “liberty” vs. “science”. The data is very clear.
Does social distancing and masks matter?
Yes! There are the clear scientific principles and there are so many clear incidents.
Let’s begin by pointing out that there are many transmission studies which show that masks trap covid19 viral particles. This much is settled science and clear. The open question is how that translates to reduced covid19 cases and deaths in the macroscopic community sense. I’ll give two examples where we can see an apparent difference between two different nearby populations.
Some clear evidence is in Kansas, where counties with masks saw an immediate reduction in cases within 14 days and those without masks initially saw no change in the Jul 12-Aug 3, 2020 time period, but then saw cases increase steadily throughout September.
Fig 6 is a direct adaptation of the primary raw data from counties representing 2/3 the Kansas population over a three week period. The left axis is the 7 day rolling average of daily cases per million population for counties with masks, the right axis the same for no masks. The no mask population had fewer cases to begin with, but noticed no change over the 3 week period. The masked population had higher cases, and noticed a statistically significant, clear, drop in cases at approximately the 14 day mark, shown by the dashed line. This is a clear case over a controlled population in geographical proximity. The evidence is quite clear. Masks reduce case count.
In Tennessee there has been a significant increase in hospitalizations associated with less mask use, as shown in Fig 7.
Other clear evidence is that there has been a statistically significant outbreak in counties following Trump’s known spreading rallies where masks were in general not worn, as discussed above. And indeed Trump held what even Dr. Fauci characterized as a superspreader event at the White House itself in support of Amy Coney Barrett — “The data speak for themselves — we had a superspreader event in the White House, and it was in a situation where people were crowded together and were not wearing masks.”
Why do masks work? What does 6ft mean?
Scientifically, the trove of information at this point is too deep to do justice to even a cursory skim, but very broadly, the following is widely agreed to at this time, an update and significant refinement of the understanding from the outset of the large case count in the NE US in the spring.
There are three principal modes of virus transmission: fomites (touching surfaces), droplets (large microscopic particles which can fall to the ground like a baseball) and aerosols (smaller microscopic particles which stay suspended in the air, more like a leaf blowing in the wind).
- Fomites are held to be not a significant source of virus transmission. That said, handwashing reduces the chance of transmission to near zero.
- Droplets are a source of transmission, and also are the historical and original source of the 6ft rule so widely adopted. But even with droplets, it’s about a 6ft radius from a particular point, depending on if there’s an airflow direction or breeze. If you’re downwind you need to be further away.
- But aerosol transmission is at least as important as droplets, and aerosols do not have a 6ft limitation. Early studies in restaurants and air flow direction demonstrated that covid19 could propagate. Where are aerosols significant? Anywhere that people are in a room for a long time together — e.g. schools, workplace, etc. According to recent understanding, aerosolized covid19 can remain in the air for up to 3 hours [https://www.health.harvard.edu/diseases-and-conditions/coronavirus-resource-center ]. People who are in the same environment (e.g. school/work) need to take additional precautions, starting with masks, so as to minimize the chance for covid19 transmission.
What does the mask do? The mask primarily prevents transmission of the covid19-containing particles which are otherwise free to form droplets or aerosols from the transmitter to those who breathe in the air. A secondary function is the last line of defense for the wearer to not breathe in such particle, but this is the secondary function.
Let’s be clear: the mask primarily protects others from the mask wearer; not the other way around. Those who say “but I don’t need a mask” are really demonstrating “I don’t care enough about you for me to wear a mask”.
But didn’t Dr Fauci say not to wear a mask?
No! Dr. Fauci has consistently been a proponent of mask use and of handwashing. The one exception, distorted by many people, is related to the horrific situation, earlier on in the pandemic, when even the doctors and front line workers could not get masks. At that time, it was necessary to make choices: clearly doctors and front line workers had priority on the scarce resource masks . Dr. Fauci is clear and consistent: wear a mask, practice social distancing.
Last words — heed the parables
We are living in challenging times. The covid19 pandemic is a clear, present threat to life throughout the world. In the USA, the knowledge exists to do something about covid19, but far too many insist it is an infringement of liberty to take the simple measures necessary, as indicated by science and demonstrated in abundance here through data. In short, many are drilling the boat under their own seats, endangering every one else. Each of us must ask “If I am only for myself, what am I?” We need to save the boat, together.
Thank you for reading this.
Additional sources used and data extraction procedures
In addition to the specific references already mentioned in the text, there are some excellent sources, most of which are updated dynamically. It is worth discussing what these sources measure.
There are many sources of covid19 information based on primary info, for example https://www.worldometers.info/coronavirus/country/us/ . In this work, also, the visualizations by Dan Goodspeed https://dangoodspeed.com/covid/total-cases-since-june are used, which measures the number of covid cases Jun 1 – Oct 24, and also color-correlated c.f the Cook partisan bias index, which work was based on primary data, in Fig 1 and Table 1. Similarly the covid19 death visualizations from Jul1 – Oct 26 from https://dangoodspeed.com/covid/total-deaths-since-july were used in this work in Fig 2 and Table 2. I strongly recommend Dan’s visualizations show the daily changes and evolution.
I’ve used csv data downloads when possible, but most of the data needed manual extraction (either 50 or 100 points per extraction, many extractions over many days). It is possible that a few, but hopefully not many, manual transcription errors were made.
The Cook partisan voting index tracks the voter percentage difference, state by state, based on the voting over the last two presidential election cycles. Numbers were extracted manually from https://en.wikipedia.org/wiki/Cook_Partisan_Voting_Index are used in this work.
Nate Silver’s team at https://projects.fivethirtyeight.com/2020-election-forecast/ provides state by state poll/popular vote forecast, state by state, as part of their “snake” visualization of the electoral college map. These numbers were extracted manually, state by state, Oct 24, for the Presidential race visualization, substrating Biden from Trump, and these numbers were used in this work as a measure of partisan bias on Oct 24, 2020. Note that Nate Silver’s percentage were multiplied by 100 to maintain consistency with Cook numbers above.
For both Cook and Nate Silver partisan bias, I’ve maintained the usual visualization – namely that negative numbers, to the left, represent the left, and positive numbers, to the right, represent the right.
Early voting has been well captured by Prof. Michael McDonald here https://electproject.github.io/Early-Vote-2020G/index.html and pre-registered voters were extracted from here https://worldpopulationreview.com/state-rankings/number-of-registered-voters-by-state
Some other covid19 partisan bias analysis has already been done here https://ltarof.medium.com/early-voting-correlations-state-by-state-early-votes-covid19-count-and-partisan-bias-b04e3cedc1ce together with hyperpartisan analysis, which may be the focus of another paper.
