Humans have a poor track record of predicting long term events. Examples abound.
The Peak Oil Theory from geophysicist Marion King Hubbert in 1956 predicted the hypothetical point at which global crude oil production would hit its maximum rate, with 1970 as that point for the United States.
Paul Ehrlich, the Stanford biologist predicted in his 1968 book, Population Bomb, that overpopulation would lead to hundreds of millions dying of starvation by the 1970s.
Michael Mann’s famous 1998 hockey stick graph showed an abrupt rise in global temperatures beginning in the 20th century, after a prior 900-year period of stability, to foretell an acute trajectory for future temperatures.
In March 2020, British epidemiologist Niel Ferguson predicted millions of global Covid deaths by the end of that year, including up to 300,000+ in Canada, after which followed a flurry of catastrophic mortality forecasting around the globe.
These examples have three things in common.
First - they have all significantly influenced government public policy, many with far reaching implications and damaging consequences.
Second - the predictions have all proven to be flawed, at minimum, or outright wrong – based on subsequent challenges by peer groups or the passage of time.
Third - they are all based on Models.
So what are these models we depend on so heavily?
Computer models are programmed digital representations of real-world situations, that can be gamed to represent different scenarios, after flowing data through them. Almost anything with variability, a long enough timeline and relevant data to be digested is a candidate.
Computer models are unseen powerhouses for the modern world. In government and business, they influence policy and drive decisions. Models are used to predict a near-limitless array of trends, behaviours and future outcomes – growth rates, stocking points, health events, weather levels, supply chain bottlenecks, climate patterns, risk potentials, voting tendencies, buying intentions, transportation flows, resource usage – and the list continues.
Modeling unlocks powerful capabilities to manipulate any number of variables and generate potentially millions of scenarios. They are indispensable tools for driving informed decisions and policies - when paired with earnest and transparent analysis.
But they do not produce definitive or immutable answers.
All models are wrong, some are useful. [James Box, British Statistician]
Model accuracy and utility depend on the underlying algorithm, the assumptions programmed in, scenario structures, quality of data flowing through, and the variables being tweaked. Model accuracy decreases as time horizon expands and assumption variability increases. And models are subject to human bias in their development, analysis and outcome reporting. Smart techs can make a model dance the Cha-Cha if they so choose.
Why does this matter?
We’ve all just lived through a period of draconian restrictions on our personal liberties, experienced impacts to our financial and mental well being, and suffered long term fiscal damage to our country…based on dire predictions about Covid.
Our Canadian government has already committed more than $40B to climate change and declared there is much more to come, laid their hand heavily on the scales regarding which “green” industries to promote and which industries are out of their favour, and are increasingly taxing and restricting businesses and citizens to accomplish their idealistic goals…based on dire predictions about global temperature rise and its causes.
As we continue to be beset with Sky is Falling warnings, bolstered by model outcomes that are delivered with great certainty - consider them a valuable input, but far from certain.
When you get anxious about the next cataclysmic prediction in the news, supported by colourful charts - think back to all the predictions we’ve gotten wrong in our well-meaning attempt to predict the future, and apply your own thinking framework to the issue. You won’t have all the data at your disposal, but you will have your pragmatism.
Stay tuned and stay pragmatic.