The Post: How better modelling could fix economic blind spots

Read the original article in the Post

Sometimes it takes a Queen to raise the tough questions. After the catastrophic 2008 global financial crisis (GFC), Elizabeth II asked British economists, with unusual bluntness: “Why did nobody see it coming?”

Among many other causes, including greed, hubris and free-market ideology, one reason for the blindness was simple: economists’ models flopped. Just before the crisis, America’s Federal Reserve modelled what would happen if US house prices fell by one-fifth, and concluded, “Not much.” In the event, as the effects of a 23% house-price collapse rippled around the globe, millions of people lost their jobs and the American economy alone took a $10 trillion hit.

These issues are aired in a book I read over the break, Making Sense of Chaos, by the pioneering complexity economist J Doyne Farmer, a colourful, bearded American whose research spans multiple domains and whose CV includes stints building computer systems to beat casinos at roulette and outperform the stock market.

Post-GFC, conventional economics has had many antagonists, especially on the left, but too often they have been outsiders decrying economics as nonsense, possibly even evil, certainly to be ignored. Farmer, as a sceptical insider, is far better placed to explain where it has gone wrong, and recommend improvements.

His overarching argument is that economists’ assumptions about human behaviour have generally been both too simplistic, modelling people as self-centred consumers with perfect foresight, and too individualistic, failing to account for the way that our actions intensify and exacerbate each other.

So far, so predictable, for anyone who has been reading the critiques of the last few decades. Where Farmer excels, however, is in explaining why economics fell into error – and it was mostly, he thinks, due to a lack of computer power.

The core assumptions – that people generally maximise their own happiness, predict the future perfectly, and reach “equilibrium” states in which the needs of sellers and buyers line up – were required to make economists’ mathematical equations solvable using pen and paper. Introducing more variables would have rendered the calculations impossible.

Fortunately, we can now replace these mathematical equations and hypothetical individuals, this attempt to predict the future using top-down abstractions, with a much more real-world-focused, ground-up approach based on big data and massive computational power.

Drawing on recent advances in both the latter categories, Farmer has spent much of this century helping pioneer a method called agent-based modelling, in which data about the decision-making of actual people are used to make models in which hundreds of thousands of realistic individuals (“agents”) interact with each other. Modern supercomputers then run repeated simulations of all this interaction, producing predictions of increasing accuracy.

When the Covid-19 pandemic hit Britain, for instance, Farmer’s team at Oxford University modelled its likely economic impact with startling precision, forecasting a fall in GDP of 21.5% against an actual contraction of 22.1%. One crucial advantage of their approach was that it captured what are known as the “emergent” qualities of an economic system: the dynamics that arise only when individuals interact with each other and which cannot be predicted by taking each person in isolation.

As Farmer acknowledges, ordinary people would probably have assumed that this is how economic modelling has always been done. It is this method that has given us weather forecasts of ever-increasing accuracy. And economics is slowly catching up.

Many of the world’s central banks, the equivalents of our Reserve Bank, have begun using agent-based models to better predict economic trends. Farmer’s modelling for the Bank of England helped show that debt-to-income limits – now used here and elsewhere to curb excessive mortgage lending – drastically reduce the likelihood of future housing bubbles.

Such modelling could also help fight climate change. Conventional economic models, Farmer argues, have typically overestimated the cost of reducing emissions and underestimated the benefits.

His team’s modelling suggests the green transition will actually save the world around $12 trillion compared to business as usual, partly because renewables are becoming so much cheaper than fossil fuels. And because scaling-up production generates opportunities for learning-by-doing and further innovation, a faster transition saves more money than a slower one. These results provide powerful support for a rapid economic shift.

Encouragingly, when Farmer gave a webinar to our Treasury last year, many of the participants were enthusiastic about his approach, even if they did point to local data gaps that could hinder its implementation. The Treasury also has in-house expertise in this area.

Better modelling won’t, of course, change the world by itself. Many of our economic problems are at heart political: we tolerate gaping income disparities because we think poor people are lazy, and disastrously high emissions because we’re careless about the natural world.

Nonetheless I found Farmer’s book to be a small ray of hope – a sign that science continues to advance, that policies will slowly improve, and that we can come to better understand our present and, with it, our future.