Covid 19 Edition
On Models and Muddles: Coping With the Coronavirus
Suddenly, epidemiologists are setting policy. Yet the emerging lessons are more general: the over-reliance on the models of experts, the mismanagement of public services, and the contrasting stresses faced by libertarian, statist, and communitarian approaches to governance.
The Coronavirus epidemic is a class ic application of the concept of radical uncertainty. Fortuitously, Radical Uncertainty, by John Kay and Mervyn King, has just been published: a brilliant account of how we became so credulous about economic models. We live in a world that has inevitably become too complex adequately to be captured in models. A world of both ‘known unknowns’ and ‘unknown unknowns,’ and hence of situations in which the most sensible response to the question ‘what should we do?’ is ‘I don’t know.’ At the onset of the current epidemic we could not put probabilities on which forms of social distancing would best limit its spread because we’d never done it before. We didn’t know how people would alter their behaviour in response to the appeal to ‘save the NHS’. We didn’t even know whether reducing the spread was desirable: perhaps fewer deaths now would be at the cost of more next winter. And these were just the ‘known unknowns.’ With a disruption like this, unknown unknowns are also lurking. We have no experience of the repercussions from shutdown in a modern economy and no meaningful way of assigning probabilities; nor of how people’s behaviour will evolve.
The perspective offered by radical uncertainty reorients our thoughts around two fundamental questions: ‘how to face unknown unknowns?’ and ‘how to face known unknowns?’ The answers to the former are to build resilience while encouraging rival teams of experts. Those to the latter are to learn from others, while investing in finding out the information that even others do not yet know. We needed to do all of these fast; we didn’t. This was not surprising: the deficiency was a symptom of a longstanding weakness.
Unknown unknowns: be prepared
A policymaker attuned to radical uncertainty expects to face unknown unknowns. By their nature, you do not know what they will be, but this is not a reason to do nothing. Resilience is about assigning responsibilities to the many entities best equipped to understand how to build the capacity to withstand shocks. After the mass terrorism atrocities in Paris, the French authorities realised that the city’s emergency health facilities could be overwhelmed. Build more facilities? Perhaps, but they also framed the question more broadly, which led them to ask the rail company, SNCF, to prepare to evacuate victims to provincial hospitals. SNCF put together a team that planned and practiced. That experience is now being used to transport critically ill patients to the spare capacity still available in Breton hospitals.
Similarly, after SARS, the authorities in Singapore realised that a health shock might arise at any time, at which point they would need the capacity for immediate large-scale response. So, they built it and waited: an entire hospital standing empty and ready, together with the testing capacity to track-and-quarantine.
The Swiss authorities are prepared not just for a health shock, but for disruptions to the supplies of essential services. Crucially, Swiss planning is designed to devolve responsibility for resilience throughout the economy, with the predominant role played by firms. There is no central stockpile: goods are kept in reserve in the warehouses of businesses distributed across the country. Some 250 senior managers at Swiss companies are trained to meet this role, reporting on their industries and co-ordinating to strengthening the supply chains. The whole system is estimated to cost a mere £10 per person per year.
With this Swiss strategy in mind, I noticed a press photo of Britain’s National Procurement Warehouse near Glasgow, which, it claims, supplies all British hospitals. Please let this not be true. Be thankful that we don’t have a Master of the Toilet Rolls: centrally planned national supply would have been a fiasco. The demand for toilet rolls hasn’t risen but shifted radically from hotels, offices and city-centre shops, to homes in suburbs; in response, supply has swiftly shifted accordingly through a myriad of devolved decisions close to changing information.
Companies are the right point at which to pin responsibility for resilience, but this requires active public policy. In the modern highly interdependent economy, CEOs do not factor in the costs to other firms of the risks they take. In aggregate, this places the economy close to a precipice where a common shock can push it over the edge. Hence, the consequences of the lockdown in Hubei: the province is a hub for pharmaceuticals and much other manufacturing.
Of the world’s top 500 companies, 300 have facilities in Wuhan. Most of the Japanese firms operating in China reported that their supply chains were affected, and few had alternative arrangements in place to withstand prolonged interruption. The knock-on effects are unknowable because often companies do not know how their suppliers are supplied. The problem is not unique to Japan, and it is particularly acute in health provision: 70% of the blood thinners used by Italy come from China. But the state knows even less about interdependence that firms: it cannot be the mastermind. The Swiss strategy of stocking key inputs through a cadre of people whose purpose becomes national resilience is fully vindicated by the latest economics.
Is it because Switzerland and Singapore are rich that they have been able to indulge themselves in resilience? Rather, it is that they are rich because they have pragmatic and forward-looking public policy, one of the manifestations being that they are prepared for the unexpected. But most were not: the global economy is exposed to that precipice. We face an even larger danger than 2009, with even less capacity for international cooperation.
Unknown unknowns: build rival teams
British policy has been set by epidemiologists: their expertise is to build models which simulate the spread of a disease. But their forecasts can be no more reliable than the numbers fed into them. Their key use should be to identify which of the known unknowns matters most, and hence where we should put our energies in data collection. Instead, we got overly confident forecasts – we should build herd immunity, or according to a maverick Oxford model, we’d already built it. What the modellers should have said was that it was vital to establish two fundamental parameters – the incidence and the rate of contagion, — both of which required repeated testing of randomised samples, and without which mortality rates and hence sensible policy were not feasible.
If you plug in enough guesses for its parameters, a model will always come up with a clear prediction. The expert in charge of the model may be tempted to claim more than is justified. Facing many unknowns, the primary use of models is to highlight those among the known unknowns which critically affect policy choices. These then become the ones most urgent to resolve, whether by learning from others, or collecting data. The value of rival teams is that it flushes out these known unknowns, while discouraging exaggerated claims to knowledge.
Known unknowns: copy those with more experience
Since the epidemic started in China and rapidly spread around East Asia, there was soon a helpful amount of variation in policy response from which to learn. South Korea and Taiwan invested in mass screening. As high-income democracies, each provided experience from which other OECD countries could learn. Indeed, the demographic impact of the disease elsewhere in the OECD has turned out to be identical to that early experience. But there is a long history, both in Whitehall and our political parties, of reluctance to learn from others. Nor is it specific to health care: an unwarranted presumption of the superiority justifying British exceptionalism. We were trying to determine whether to go for herd immunity or suppression based on a British model rather than East Asian experience.
Known unknowns: find out
Whether for track-and-quarantine, release-from lockdown, or for the choices between herd immunity, shifting the peak, and suppression, we need mass testing. This takes us further into public mismanagement.
Who is currently infectious? We cannot rely upon self reporting or referral by doctors because many of the positive cases are asymptomatic. In Vò, an Italian town in which everyone was tested, only half the positives had symptoms. Who has acquired immunity? We need a massive capacity for testing: we don’t have it. We will soon need to reduce the soaring costs of lockdown by test-and-release, while continuing to sequester the over- 70s. Such tests would be self-administered, and so less accurate: as Prof Peto of the London School of Hygiene and Tropical Medicine wisely comments, ‘this would not matter’. If this is what it takes to reopen the economy, it is well worth it despite its inaccuracies.
There is no mystery as to why the United Kingdom is so far behind Germany. Whitehall and the main British political parties share a longstanding fixation with centralised control: it is an aspect of the same syndrome — the top knows best. The national testing centre at Colindale will develop the perfect test, and for this it naturally needs to guard its monopoly. The perfect test, from a scientific perspective, is completely accurate: no false positives, and no false negatives. But from a public policy perspective, currently the priorities are not accuracy but speed — time between a sample being taken and the result being ready, and scale – a test that can be done by the million.
Germany doesn’t have a national testing centre — it is entirely decentralised. Given a 65-year litany of failures, not just in healthcare but across the board, the persistence of Whitehall’s faith in the capability of centralised control persists: if anything, it has been getting worse as Brown and Osborne centralised decisions in the Treasury. Shifting responsibilities down the system not only enables rapid scale up, it has a further huge advantage: the power of decision is closer to the coalface of practitioner experience. We learn not just from accumulating and analysing codifiable knowledge — the domain of the expert. We learn by doing, or by trying to do, things that we can’t do and that force us to experiment. A decentralised system learns from a myriad of failed experiments running in parallel, and so it learns fast: teams copy other teams that have hit on something that works well enough to get the job done.
But the key known unknowns are a containment strategy and a vaccine. Both answers are global public goods and so liable to be undersupplied. We simply do not know whether the typical OECD country is best doing some variant of Swedish or South Korean policy. Sweden is insisting that people take personal responsibility for their safety, building herd immunity at a pace that protects the health system, and encouraging self-isolation among the elderly. South Korea and New Zealand kept the numbers down sufficiently to do track, quarantine and contain. To establish which of these strategies is best we need variation and modularity. The European Commission should be encouraging member countries to adopt different approaches. The political herd immunity to which governments are prone is that it is safer to fail with a policy that others are following than to fail with a distinctive policy. The Commission should explain this and congratulate the Swedish government on trying an approach which may turn out to be better for Europe than lockdown, and which looks well-suited to emulation in Africa. We don’t know and it may be years before we do.
But we can only learn from variation if there is no contamination between countries trying different approaches, and for this we need to close borders to population movement. Singapore became re-infected by new arrivals; New Zealand has been more cautious. Again, this is something that the Commission should be advocating. Yet to date, it has been advocating precisely the opposite: common strategy, and open borders. These familiar, ideology-driven political priorities have overridden the forward-looking pragmatism needed for learning. We don’t know how to make a vaccine, but both strategies depend on the vaccine cavalry arriving as soon as possible. To find a vaccine we need experiments in parallel. One of the more useful contributions of economics has been to show why markets, which work so well for toilet paper, do not work well for the discovery of vaccines. We need to override this market, not rely on it. Public money needs to finance many rival research teams, and as some start to look promising, each of them should be further financed to prepare for mass production. Normally, this stage doesn’t happen until after there is a clear winner with a patent. But each government has an incentive to leave this to others: we need the fragile entities we have for multinational coordination to focus on this specific task. While the finance is best organised from each government to its nationally-based pharmaceutical firms and universities, the Commission and the G7 should both be trying set mutual commitments to rapid scale-up of finance.
Beyond Muddles and Models
Britain, like most Western countries, will muddle through this epidemic: much of the course has now been set. Lockdown is unlikely to run through to the Autumn because there is growing awareness that locking down parts of an interdependent economy can have devastating knock-on effects which themselves cause mounting health problems. The latest concern here is that the attempts not to overburden the NHS may have resulted in such a sharp fall in treatment for other conditions, with many beds left empty, that this has caused more deaths than the virus itself. Such is the nature of unknown unknowns.
All of us, in Europe and beyond, are probably entering a phase of lockdown-ease-lockdown-ease, differentiated by age groups, until testing has scaled up to permit test-and-release. Time will be bought until a vaccine is discovered. Fortunately, Britain is good at muddling through: we have countless such national narratives on which to draw. Neither now nor later, should people indulge in blame: nobody knew what to do, and mistakes were inevitable. But we, like others, do need to learn, and to learn from others, because the unexpected will happen again. A common refrain in economics is that ‘it takes a model to beat a model’, and that attitude encapsulates the fundamental error not only of British policy, but of quite a few of us around the world. Faced with events like this, all models will be wrong. They have their uses, but the lesson is to think beyond the models, not just within them.
Paul Collier is Professor of Economics and Public Policy at the Blavatnik School of Government, Oxford University. This is an adapted version of an article which was the cover feature in the Times Literary Supplement, published on 24 April.