Chapter 6 from the R.M. Dolin book, “Truth and Trust in Crisis.“
May 5, 2020: Three weeks is like forever in a crisis, especially after considering all that’s been lost, including, our inherent trust in bureaucrats, academicians, the media, and medical professionals. The cumulative COVID deaths according to CDC reporting has climbed to 68,934, and while this is tragic, it’s significantly less than the 500,000 Americans we were told would die from COVID by the end of April.
Three dramatic weeks ago, our trust in crisis managers Deborah Brix and Anthony Fauci was implicit; a trust so unwavering we formulate a hypothesis to disprove the COVID virus is human caused. We’re so confident all fingers for this catastrophe point at China, we develop a null hypothesis asserting if the virus behaves differently than a naturally occurring virus it’s because it’s human caused. To prove our null hypothesis, we develop an exponential model, because things in nature follow exponential patterns and discover the COVID virus is behaving like a natural virus, which doesn’t necessarily negate the possibility it’s human engineered.
Our analysis takes us beyond attempting to prove COVID’s origins to investigate why academic models are so hyperbolically wrong, and why there’s little interest in correcting them. Treating the world as a statistical population, we use the Italian COVID crisis as a subsample and build a statistical inference to predict that by COVID’s projected apex on April 14th, the U.S. will likely have experienced 30,500 deaths. This seems wildly implausible as experts assure us the cumulative death count by April 14th is going to be 250,000. When apex day arrives, however, the death count is 23,650, slightly below our prediction but inexcusably ten times below expert assessments.
What follows Apex Day is as shocking as it is unbelievable. By announcing COVID’s in remission, the CDC finds themselves at odds with government experts. Not to be upstaged, Deborah Brix announces CDC data cannot be trusted and counters by asserting COVID deaths are being under-counted even though her partner Anthony Fauci contends they’re being over-counted. To find out which side of trust CDC data lies, we’ll later use our exponential model to track COVID’s remission, to help make monthly projections, and compare our projections with the big three COVID models from the University of Washington (UW), University of Pennsylvania’s Wharton School of Business (Penn), and the Federal Emergency Management Agency (FEMA). We’ll even be so bold as to predict the day when COVID’s no longer an epidemic and look for validation in unlikely places.
After their embarrassing performance in April, you’d think responsible bureaucrats and academics would humbly ask for forgiveness and fade from the limelight, however, the intoxication of fame, coupled with American’s insatiable need for news, even when wrong, keeps COVID experts at the forefront of our crisis. Recall how recklessly unrestrained academia has behaved, first the UK model projects 2.2 million Americans will die from COVID by October. Not to be outdone, the UW model ups the ante by projecting 250,000 deaths in 14 days, which equates to a death rate two times higher than the UK model.
The same UW crew that rendered results 1,057% wrong in April, boldly roars back in May with a new projection, only this time, after having repeatedly embarrassed themselves and all of academia, they’re presenting an estimate so devoid of meaning it’s clear they no longer believe in their ability to make credible projections. UW projects[1] that by August 4th, between 95,000 and 243,000 Americans will have died from COVID. Notice they no longer provide a single number, which is an indication they’re learning to hedge their bets, only the range is so wide, the model is devoid of sufficient fidelity to be useful for anything other than parody; unfortunately, the government decides to use it for policy. Brix/Fauci[2] offer a somewhat narrower prediction that between 75,000 and 100,000 Americans will have died from COVID by August 4th, which is a significantly lower “worst case” estimate. Given how poorly previous Brix/Fauci and UW estimates have performed, one would expect credible journalists to challenge these dire estimates or at least question why they are so different, instead, we get compliant silence and a concerted media effort to censor any challenge to prescribed narratives
Adding upper and lower bounds is deftly strategic because the media never reports the lower number as it lacks hysterical impact; like they’re fond of saying, “if it doesn’t bleed, it doesn’t lead.” State and federal policy makers also focus on the upper bound to justify continued lockdowns and mandates. If August COVID numbers come in at the high range, our experts can tout their prowess at accurately modeling outcomes, if August numbers come in at the low range, they’ll take a victory lap for the success of their emergency measures, and if the numbers fall in between, it’s an optimal combination of both.
We are going to build five crisis models to not only demonstrate how easy they are to build, but also because each model provides elements of information that, when subject to scientific rigor, can be used to amass credible estimates of what’s likely to happen. Alone, anyone one of these models is insufficient, providing only a piece of the COVID prediction puzzle but their gestalt, i.e., their whole being greater then the sum of their parts, allows credible predictions to be made.
While each of the five models rely on the same data, they use different logic and analysis methods along with different assumptions. Each model yields different results based on the assumptions utilized. This does not make one model necessarily better than another as each provides insight. Knowing when one model is optimally appropriate comes down to experience and an understanding of the model’s limitations, which is where our scientific approach becomes paramount. This is a salient distinction from the Brix/Fauci and UW models. Brix/Fauci are medical doctors, not scientists, the UW model was developed in a university medical department, not a STEM department. The Penn model comes from a business school, and while likely skilled at statistics and accounting they are not trained in the scientific methods of data analysis. The FEMA model is developed by bureaucrats and likely lacks scientific rigor.
As with any scientific problem, we begin by stating our knowns, unknowns, and assumptions
- The first COVID death[3] in America occurs on February 29, 2020.
- As of May 5th, there have been 1,180,634 confirmed COVID cases with 68,934 reported deaths.
- The percentage of the population infected is 0.369%, and the percentage of those infected who die is 5.84%
- The percentage of Americans tested thus far for COVID is 1.62%
- The preCOVID number of Americans expected to die from something in 2020 is 2,863,859, or 7,825/day.[4]
- Herd immunity is achieved when 60% of the population becomes infected[5] and the virus dies for lack of new hosts.
- The official 2020 U.S. population is 328,200,000 (migrant population not counted).
We can use this data to generate our different COVID models, each providing elements of information that when combined, allow for an optimal prediction. I’m tempted to ask you to trust these models, but since we agreed to a zero-trust approach, you’re encouraged to run the numbers and make your own decision as to how best to apply each element of information into your gestalt.
The current iteration of expert models each project out to August, which makes me wonder if it’s because they believe that’s when herd immunity is reached. They’re not saying, but let’s test this hypothesis and see if we can generate an estimate of our own. Keep in mind, achieving herd immunity is one way this crisis ends.
Model 1: Assume CDC Data Is Correct
While we don’t know how many people have been infected, because we haven’t tested everyone, we know how many have died. Given that 1.62% of the population’s been tested and 1,180,634 cases are confirmed, if the entire population were tested there’d likely be 72,869,716 confirmed cases or 22.77% of the population. If ~73 million Americans are currently infected, it means the chance of dying once infected is ~0.09% and the projected number of deaths when herd immunity is reached becomes 186,284, which is above the upper estimates of both the UW and Brix/Fauci projections. The supporting model math is provided in Table 6.1. While this model provides an estimate on the number of COVID deaths likely to occur by the time herd immunity is achieved, it does not provide a timeline.
Table 6.1. Assuming the CDC Data is Correct.
Model 2: Assume Results of Stanford Study
A weakness in our first model’s logic is that the 5.2 million Americans currently tested for COVID were not randomly selected, they likely suspected they had COVID and then got tested, which skews the statistics. According to a Stanford University study, if a random sample of the population were to be tested, we’d find 50 to 85 times more infections than the CDC confirms are likely. If that becomes the basis for our death count estimate at herd immunity, then between 135,000 and 230,000 deaths would occur, which is close to the UW projections and bounds the Brix/Fauci projections as shown in Table 6.2.
Limitations of this model include that the upper and lower bounds are too wide to be meaningful, and it assumes a uniformly distributed population, i.e., that an average person in Manhattan experiences the same COVID journey as an average New Mexican
Table 6.2. Assuming the Results of the Stanford Study.
Model 3: Assume the CDC Overcounts COVID Deaths
The CDC has a history of sensationalizing data[6]. For example, in 2014, they predict 500,000 people will die of Ebola in West Africa when only 28,600 do. During COVID, the government incentives medical doctors and hospitals to attribute nonCOVID deaths to COVID,[7] which skews the data toward unrealistically higher death rates. Several states, including New York[8] and Pennsylvania[9], admit to inflating their COVID deaths, some by over 50%. If we project the number of deaths when herd immunity is reached based on the number of people likely to be infected, then half the number to account for confirmed rates of over-counting, we arrive at a projection of 93,142 deaths when herd immunity is reached. This estimate is below both the UW and Brix/Fauci projections as shown in Table 6.3.
A limitation of this model is that while some states intentionally overcount COVID deaths, not all states are and so we don’t really know what the data-falsification factor should be. Also, because the medical profession can’t be counted on to act honestly once incentivized, we know we cannot trust whatever they report and therefore know the data-falsification factor is not one.
Table 6.3. Assuming the CDC Over Counts COVID Deaths.
Model 4: Assume Actuarial Assessments
When one source of data gets corrupted, there usually exist other sources if we are creative. For example, highly accurate pre-COVID estimates exist for the number of Americans likely to die this year from heart disease, cancer, diabetes, etc., and the advent of COVID does not treat heart disease or cure cancer. Therefore, these pre-COVID predictions remain constant and COVID deaths are in addition to pre-COVID estimates. This obvious observation is going to be of paramount importance going forward as increasing evidence suggest CDC can’t be trusted.
Actuarial data provides what may be the most definitive alternate way to assess the number of COVID deaths when CDC data is suspect. We can use actuarial tables to determine the number of deaths expected to occur in the U.S. pre-COVID (from diseases, accidents, and other calamities), then subtract that from the number of deaths that have occurred from all calamities, including COVID. In theory, the delta represents the COVID caused deaths. What gets lost in this approach is information about the people who die from COVID who were going to die from something else and the people who die from something else that medical doctors attribute to COVID.
Prior to COVID, 2,863,859 Americans are expected to die from something in 2020, which means that by May 5th, 985,919 Americans are expected to die from nonCOVID caused calamities assuming a linear distribution. However, 984,462 Americans have died thus far, which means 1,457 fewer Americans have died so far this year than were expected pre-COVID. The limitation of this assessment is that there’s no distinction between people who die from COVID and those who die with COVID. Nonetheless, this is a valid measure of the impact COVID has on the country’s overall death count and is an astonishing outcome given the panic and hysteria being promoted.
Table 6.4. Actuarial Assessment.
This result indicates fewer Americans have died thus far in 2020 than were expected to die pre-COVID. A controversial conclusion that can be drawn from this result is that all the people who the CDC reports as having died from COVID thus far in 2020, were likely going to die this year from something else because no one who was going to die this year from something else is now alive because of COVID. While that’s of course not true at an individual level, in the aggregate, correlations do exist, and we’ll come back to this model often to see if the trend continues for the remainder of the year.
This model assumes that actuarial deaths are evenly distributed throughout the year. While that’s not likely true, it’s probably not far off. Even with this result’s caveats, it can be used to conclude that the overall impact of COVID relative to a normal non-COVID year is not as significant as we’re being told. This defies logic because after all the hysteria, it appears COVID is having minimal impact on American deaths thus far in 2020.
Model 5: Exponential Growth and Decay
For our last model, we’ll extend our simple exponential function, f(x)=ax, to both estimate the number of COVID deaths at herd immunity and when it’s reached. For this we use the knowledge gained during COVID’s assent up to the government proclaimed apex to project a likely descent down to herd immunity. After that we can integrate the curves to find the projected number of deaths. This involves a lot of calculus, but don’t worry, there won’t be a test at the end.
To determine COVID’s impact under uncertainty, we can compare the overall number of deaths in the U.S. this year from all causes against preCOVID expectations like we did in the actuarial model. Through May 5th, the overall number of U.S. deaths are 1,457 below what would have been expected had COIVD not happened. Projecting the number of deaths likely to accrue by August 4th, based on suspect CDC data, depends on three variables,
- When the COVID expansion apex was reached.
- The pace at which death rates decline while COVID’s in remission.
- Where the tapering tail of the exponential curve is truncated.
Figure 6.1 captures the rise and fall of COVID deaths per our exponential model. The blue segment represents COVID’s exponential growth up to its apex, which we previously generated. The red line represents COVID deaths from the apex to present, and the yellow line represents our model’s projection out to August 4th. It’s interesting that while I have castigated expert models for being outrageously wrong on their death count estimates, I applaud them for closely estimating when the U.S. COVID crisis peaks. As our analysis shows, the apex was very close to their April 14th estimate.
Figure 6.1. Exponential Model Predicting the End of COVID Pandemic.
Mathematically, the area under the curves represents the cumulative number of deaths, which can be represented as the integral of the exponent. In general, if the function is represented as f(x)=ax, then the integral with respect to x, is
From this equation, the area under the blue portion of the curve is 20,608, which correlates to the CDC reported number of COVID deaths on April 14th. The combined area under the blue and red curves is 73,431, which correlates to the CDC reported number of COVID deaths through May 5th. In other words, our simple model accurately simulates the CDC reported COVID crisis.
The yellow curve approximates the exponential rate of decline in COVID deaths from May 5th to August 4th. According to this approximation, by June 7th, the rate of COVID deaths declines to 85 deaths per day, with a cumulative total on August 4th of 85,656. Mathematically, the yellow curve extends to infinity, so must be truncated at some point. The CDC combines pneumonia, influenza, and COVID (PIC), deaths into a single number. Because a PIC number is used, zero deaths are not possible since both pneumonia and influenza are endemic. The question becomes, at what rate of PIC deaths can we truncate our model?
The minimum value would 85 deaths/per day because that’s how many Americans die each day from pneumonia and influenza in a typical year, so it’s the endemic death rate we accept as normal. Using this metric, our exponential model projects that the COVID epidemic ends on June 7th, 2020. If you think that’s an absurd assertion based on government, academic, and media hyperbole, hang on to your thinking cap, because it’s about to get bumpy.
This is a profound moment in our COVID assessment because we’re asserting that a month from now, COVID ceases to be an epidemic. Given government and media’s investment in hysteria, having the audacity to suggest the COVID crisis is about to end borders on preposterous. I share your sentiment, when our analysis first suggested this outcome, I ran, then re-ran the numbers multiple times looking for errors until eventually convincing myself the math is solid, the logic consistent, and our data backed up with evidence. We’re staking our flag on June 7th in stark contrast to what government, academia, the media, and medical professions are marketing. This then leads us to the matter of validation, specifically, how we decide when COVID has abated. As scientists we must comply with standards and certain thresholds must be met, such as independent confirmation of our outrageously preposterous projection.
The University of Washington projects 243,000 deaths will have occurred by August 4th. The White House COVID Task Force projects 100,000 deaths by this date. Our model, based on a first principles analysis, projects COVID ceases to be an epidemic on June 7th, and by August 4th, 85,656 PIC deaths will have occurred. This puts us below both the academic and government projections.
One likely reason the academic projection is significantly higher than the government prediction is because these two nonscientific estimates are driven by competing narratives. The academic projection is utilized primarily to justify media hysteria and government mandates, while the White House projection is driven by an incumbent president’s need to put the pandemic behind us in time for November elections. Our model, which is based on scientific evidence applied in a logically consistent framework, is devoid of narrative.
Based on available data and the gestalt of our five models the rate of COVID’s decay appears to be following its rate of expansion. In other words, the rate at which the country went from zero deaths on February 28, to the April 14th apex is nearly the same as the rate in which the virus is currently decaying, which is something we’d expect. If COVID’s remission precisely mirrors its expansion, the truncation date would be June 2nd, which makes our June 7th projection pretty darn close. It also means that states should be fully re-opened by early June, only don’t hold your breath, COVID’s quickly becoming a useful political pandemic with competing narratives and November’s not that far away.
[1] U.S. Death Toll Projections Double As Researchers Measure Impact Of Eased COVID-19 Lockdowns (forbes.com)
[2] Trump raises coronavirus fatality projection to as high as 100,000 – CBS News
[3] First death from coronavirus in the United States confirmed in Washington state – CNN
[4] United States Deaths Clock – IndexMundi
[5] Rethinking Herd Immunity and the Covid-19 Response End Game | Johns Hopkins Bloomberg School of Public Health (jhu.edu)
[6] Horowitz: Dr. Birx: ‘We’ve taken a very liberal approach to mortality’ – TheBlaze
[7] Minnesota doctor blasts ‘ridiculous’ CDC coronavirus death count guidelines | Fox News
[8] New York City coronavirus death toll jumps past 10,000 in revised count – Gadgets news – NewsLocker
[9] Pennsylvania Takes Hundreds of ‘Probable’ COVID Deaths Off Books After Coroners Come Forward (westernjournal.com)
