Our Experts' Overactive Imaginations: How Our Schools were Shuttered by a 10-Grade Science Project
The novel coronavirus pandemic has been the most disruptive phase of American life of this new millennium, generating half again as much unemployment as the great recession of 2008 and ten times as many American deaths as the September 11 attacks and the wars in both Iraq and Afghanistan combined. We, the American people, wise or foolish, have come to expect that our government will act in an informed, responsible manner, seeking to combine the insights of experts across a swath of domains: economic, public health, national security, and others. We do not expect perfection because the world is complex and finding solutions to these epoch-defining events is an imperfect and all-too-human endeavor. We expect our leaders to marshal our best and brightest to come to our defense and protect our interests. They have betrayed these expectations.
Our policy of social distancing, particularly the closure of schools to prevent the spread of COVID-19 by our youth, did not come from seasoned researchers, epidemiologists, or other experts. The policy of mass school closure in response to outbreaks of flu-like disease emerged from a high school sophomore’s science project.
Laura Glass, then 15 years old, entered her project in the Intel International Science and Engineering Fair in 2006. She was aided in its creation by her father, Robert Glass, a researcher at Sandia National Laboratories. Sandia Labs, a subsidiary of Honeywell, is a facility for the National Nuclear Security Administration. Robert Glass assisted his daughter in publishing the model (and subsequent work) in the Journal of Emerging Infectious Diseases in November, 2006.
The Glass-Glass model is a “social network” model. In essence, it is a computer simulation of a series of hypothetical towns of 10,000 people, each subjected to a flu outbreak. The makeup of each town is different in each simulation, but all towns have their social networks constructed using the same rules. Towns are made of a series of social networks, whose type and structure depend on what kind of group is represented. For instance, all towns contain households, child classes, teen classes, and adult work groups, among a variety of others. Households are assumed to be fully connected, while classes and work groups use ring networks; this means that disease can spread from one member of a household to any other, while the class and work groups are limited in their transmission to “neighbors.” The frequency of daily contact, and the number of possible contacts each individual has with others are random variables, determined by what kind of group they are in.
The Glass-Glass model’s findings are clear: teenagers (and to a lesser extent, young children) are the prime vectors for spreading influenza-like diseases. The model finds that school closures alone are only modestly effective at reducing outbreaks, reducing total cases by about 20%. However, school closures, coupled with a strict policy of keeping children at home, reduce cases by over 90% if compliance remains high. If compliance is only 50%, the model still finds a reduction in total cases of over nearly 70%. It would seem unambiguous: we must socially distance our children if we are to stop the spread of infectious diseases.
So what’s the problem?
The model assumes what it wants to prove. It’s not obvious by any stretch. The issue lies in how the groups that make up the towns are defined. The child and teen groups have significantly more “churn,” or interconnectivity, than do the adult groups. For instance, “child classes” (elementary education) are assumed to have one class per child, with an average of four links per member, with an average of six contacts-per-link each day. Teenagers are assumed to have six classes per day (exposing them daily to a far larger number of individuals than their younger counterparts), each with an average of four links per member, with one contact per link per day. Adults, in contrast, are assumed to operate in “work groups” with an average of six links per member, and a contract frequency of one per day per link. While such workgroups are likely representative of office workers and lab technicians, they certainly don’t seem to correspond to the realities of bartenders, taxi drivers, or others whose principle job is to interface with the public. Simply looking at the number of linkages in the groups, it seems obvious that teenagers would be the “super spreaders” of our simulated towns, assuming this network structure represents reality.
The model fails to account for a wide number of facets of public life that lend themselves to the transmission of diseases. The largest group in the model is 50 people. There are no concerts, no public transit, no public-facing service industries. The Glass-Glass model is a “toy” model, which perhaps serves as an illustrative model, but has no empirical basis whatsoever. The number of adults, children, and teenagers are proportional to US census results at the time, but the structure of the links between them seem entirely based upon arbitrary assumptions. The math is fancy, but the structural assumptions of this model are woefully insufficient to drive policy.
So how did this rickety stack of bad assumptions come to drive the United States’ response to the Coronavirus pandemic? Robert Glass was asked to prepare a brief for Secretary of Homeland Security Michael Chertoff concerning bioterrorism. This brief was intended for a cabinet-level exercise at the White House. In his own words: “I thought, ‘That’s exactly what Laura is working on. Her model was right there on the computer. I realized that was something important. I discussed it with her. She said, ‘Why don’t you close the schools?’ I was taking advice from my (high school) daughter.”
In 2008, the Glass-Glass model was cited by Neil Ferguson of Imperial College London in Nature, the most prestigious scientific peer-reviewed journal. Ferguson, an erstwhile physicist turned epidemiological prognosticator, is perhaps most famous for predicting 200 million deaths from the 2005 avian flu outbreak. Ferguson, dubbed “professor lockdown” during the current coronavirus pandemic, is responsible for the shift in the UK’s initial herd immunity policy to the current strategy of strict quarantine. He claims that this shift will save 510,000 lives in the United Kingdom alone, roughly 0.8% of its total population. Ferguson is also the source of the claim, since revised and derided, that the coronavirus would claim the lives of 2.2 million Americans. Incidentally, Ferguson was also identified as the vector that introduced the novel coronavirus to 10 Downing Street.
This brand of shiny, entirely simulated “faux-studies” cause us to import ruinous assumptions when we want to make real policy in the real world. The CDC reports that more children have died from influenza (100) than coronavirus (28) through 6/24/2020. Despite this, we have suspended the education of our children, lest they carry the disease to the elderly. Despite closures of schools, the pandemic has spread unabated, and a growing body of evidence suggests that social distancing had no statistically perceivable effect on the pandemic’s spread. We have been dragged along by those claiming to be experts, but whose expertise has no basis in empirical reality.
It is no longer enough for us uncritically to ask those who claim to be our “experts” what they know. They have and will continue to fail us. We have borne witness to security ratings agencies who claim that collateralized debt obligations are AAA rated, when they bear more in common with a Macau gambling den; quantitative “Masters of the Universe” and nobel prize winners who claim that their “12 sigma” strategies will fail once in a billion years when they are in fact merely piling on leverage; or epidemiologists who claim to be able to keep us safe when all they are doing is selling the illusion of safety. These “experts” have shown that their predictive powers are more like those of astrologists than scientists. It is now incumbent on us to ask how they know what they claim, lest we drown in a sea of snake oil and nice-sounding jargon.
Stay safe everyone,
Original Glass-Glass model study from 2006:
Coverage of the Glass-Glass model from 2006:
Neil Ferguson Citing Glass-Glass in Nature, 2008:
Later refinements and articles by Glass and Glass:
Robert Glass discussing daughter Laura’s role creating social distancing:
Niel Ferguson’s role in crafting UK COVID policy and prior bad predictions: