Saturday 21 March 2020

The Problem with Mathematical Modelling in Covid-19 and Economics

"Mathematical modelling" is everywhere at the moment, and it should make us as nervous as medical staff going to work without protective gear. It's a bad tool for policy making in a time of crisis. Models are abstract and never specific, but it is the specifics which determine matters of life and death. Whilst an agent-based model might give some idea of the exponential rise in cases, or the overloading of health systems, there will always be missing variables and causal mechanisms which are misunderstood. Some of those missing variables will account for why nearly 800 people died in Italy yesterday (following the nearly 700 the day before) two weeks after their lockdown, or why half of the 1500 people in France in intensive care are aged under 60. At the same time, the statistics from different countries don't seem to be comparable: China under-reported the extent of the death rate from the virus and Russia is determined to say that it is a "foreign problem", underplaying its significance, while building hospitals on the side.

Epidemiologists trying to predict the consequences of COVID-19 with models will cite the caveat that it is "only a model", but the fact is that it's only "only a model" if it isn't used to inform policy. As soon as policy takes heed of a model, the model becomes part of the situation. Ideas like "herd immunity" arise from models, and have informed policy. The outcry that this policy was effectively eugenics has been sufficient to cause some back-tracking.

Mathematical modelling is affecting policy decisions beyond simplistic models of transmission. At the root of government policy - particularly in the Anglo-Saxon world - is economic thinking which is also underpinned by mathematical modelling. Where do the epidemiological models and the economic models meet? That seems to have been the question puzzling the British and US governments in the last week or so. It is, of course, the wrong question. Herd immunity didn't simply arise from a particular slant on an epidemiological model; it was a compromise between the epidemiological model and the economic model: let the virus sweep through the country, let everyone continue their lives as normal, let people "eat virus", let the old die, let's save on the care and pensions, protect the market and the banks and all will be well. To any country with a "pension reform" headache, I can't believe this thought hasn't crossed the minds of their leaders, and the less scrupulous they are, the less they seem to do about the virus.

Frankly, to the calculating mind of Dominic Cummings, Boris Johnson, Donald Trump and a few others, this seems like a good plan. Until you think that it might be your parents in a makeshift hospital with no ventilators. The problem from Cummings and Johnson, and for any modelling geeks out there, is that Coronavirus is not abstract. It's not like a hedge fund whose victims are nameless in far-away countries. It threatens people we love. And much as the logic of capitalism dictates that we are all individuals engaged in a kind of Darwinian struggle for material success, love creates bonds which do not obey the individualist logic of the modeller. There is no variable which can represent its effects.

This is not to say that models are useless. But it is to say that the most sensible attitude to them is to ask how they might be wrong, rather than to ask what they predict. They are useful in the sense that they promote critical discussion among concerned individuals thinking about how to act effectively. So the critical thing is the organisation of human beings around the model. Get a load of politicians who want quick-and-easy answers, and the model could be lethal.

Unfortunately, in economic modelling, they have never been used like this. Much to the dismay of great minds in economics, including Marshall, Keynes and Hayek, bad mathematics took over the discipline of economics promising policy guided by numbers - models presented ways of removing the uncertainty of policy-making. But in removing the uncertainty, they threw away most of the information in the system.

What it left us with was a rationalistic, linear and shallow picture of human life. Looking at the world rather like Harry Lime looks down on those "little dots" from the Ferris Wheel in the Third Man, each of us was reduced to a kind of "variable set", each with our motivations and histories, and each of which could stop moving or disappear at any point without any effect on the others.

Coronavirus tells us what's missing: it cuts to the heart of our mistakes in modelling. The virus reminds us what we always knew but preferred to ignore: "we are all connected". Model-based capitalism told us the opposite. The crisis of this pandemic is only just beginning. Our very understanding of money, the economy, ownership, debt and wealth depend on the modeller's deceit that we are individuals acting with one another according to rules codified in law, executed through a market. Coronavirus takes us back not to social law, but to natural law: the bonds of love across generations matter more than any amount of money. It blows apart what Veblen saw as the atavistic behaviour of the leisure classes once and for all.

Love isn't some new variable which can be factored into the model. What happens in meaningful social interaction is the coordination of expectations, and love plays a powerful role in forming expectations. Money is, by contrast, merely a codification of expectations: artifice. But when "heart speaks to heart" (as Newman put it) - as it surely is now - there is no need to artificially codify expectations. We know the truth of the world. Maybe it is the "implicate order" of nature that we perceive. But we know. And nothing else matters.

There is a modelling question to be asked here. But it is not about extra variables. It is more about what the heart speaking to heart really is. What is it that enables us to tune-in to one another? Indeed, what is it that drives us to modelling in the first place?

The way forwards from Coronavirus will be a meta level of understanding.

No comments: